The Pan-Bcl-2 Inhibitor (−)-Gossypol Triggers Autophagic Cell Death in Malignant Glioma

Defects in apoptotic signaling pathways and overexpression of antiapoptotic genes play fundamental roles in the development of many types of cancer, including gliomas (1-4). Caspases are the major executioners of classic apoptosis (type I cell death), but it is now clearly evident that both caspase-dependent and caspase-independent forms of programmed cell death exist (5, 6). Autophagy is a form of cellular self-digestion where cellular constituents are engulfed in double membrane–containing vesicles called autophagosomes (7). Their vesicular content is subsequently digested by lysosomal proteases after fusion of autophagosomes with lysosomes. Autophagy is a complex, multistep process that is genetically regulated by the Atg genes (7). Members of the phosphatidylinositol 3-kinase family are also major regulators of autophagy. Whereas class III phosphatidylinositol 3-kinases are required for autophagosome generation (7), class I phosphatidylinositol 3-kinases inhibit autophagy through activation of the kinase mammalian target of rapamycin (mTOR; ref. 8), thereby also enhancing proliferation and survival.

The role of autophagy is highly contextual, as it can exert both cytoprotective and death-promoting effects, the latter of which implicating overactivation of autophagy as an alternative cell death pathway (9, 10). Indeed, induction of autophagic cell death (type II cell death) by autophagy stimulators has been recently discussed as a concept to exploit caspase-independent programmed cell death pathways for the development of novel anticancer therapies, especially those directed against malignant gliomas (11, 12).

By antagonizing activation of the pore-forming proteins Bax and Bak, antiapoptotic Bcl-2 family members inhibit the mitochondrial pathway of apoptosis (13-15). Recent findings show that antiapoptotic Bcl-2 family members also can form a complex with Atg6/Beclin1 (16-19), and formation/dissociation of this complex may play an important role in modulating autophagy in tumor cells. Bcl-2 and Bcl-xL sequester Beclin1 by binding to its BH3 domain and prevent it from forming a multiprotein complex essential for vesicle nucleation during the early steps of the autophagic process, thereby inhibiting autophagy (16, 18, 19). Consequently, small-molecule inhibitors of prosurvival Bcl-2 proteins binding to their respective hydrophobic BH3 grooves—also termed BH3 mimetics—are capable of activating both apoptosis and autophagy (15, 18).

There are several synthetic and natural BH3 mimetics, most of them targeting Bcl-2 and/or Bcl-xL (15, 20). Gossypol is a natural polyphenolic compound and BH3 mimetic derived from cottonseeds that possesses proapoptotic effects in various in vivo and in vitro models (21-24). Gossypol acts as a pan-Bcl-2 inhibitor and can inactivate Bcl-2, Bcl-xL, Mcl-1, and Bcl-w (15, 20). There are two enantiomers of gossypol, (+)-gossypol and (−)-gossypol, the latter being more potent as an inhibitor of tumor growth (15).

(−)-Gossypol (AT-101) has shown single-agent activity in various types of cancer (15, 20), and its potential impact for glioma therapy is currently being investigated in phase I/phase II clinical trials. In cancer cells with an intact apoptotic machinery, (−)-gossypol has been reported to induce apoptotic cell death (21, 25).

Here, we investigated the sensitivity of apoptosis-resistant malignant glioma cells to single-agent treatment with (−)-gossypol and to (−)-gossypol in combination with temozolomide (TMZ). Our data show that (−)-gossypol induces mitochondrial dysfunction in the absence of effector caspase activation to trigger an autophagic type of cell death, which is further potentiated by TMZ in MGMT (O⁶-methylguanine-DNA methyltransferase)-negative cells.

The mTOR inhibitor rapamycin, racemic gossypol, bafilomycin A1 (BAF), and TMZ were obtained from Sigma-Aldrich. (−)-Gossypol (98% purity) was acquired from ENZO Life Science. The caspase substrate acetyl-DEVD-7-amido-4-methylcoumarin (Ac-DEVD-AMC) was purchased from Bachem. Tetramethylrhodaminmethylester (TMRM) was from MobiTec. BH3 mimetics BH3I-2′ and HA14-1 were purchased from Calbiochem (Merck Biosciences) and ABT-737 was from Symansis. C₂-Ceramide was from Merck Chemicals Ltd. Lysotracker Red DND-99 was obtained from Invitrogen. All other chemicals were used in analytic grade purity from Sigma-Aldrich.

For this study, human grade 3 to grade 4 malignant astrocytoma cell lines U87, U343, LNT-229, and MZ-54 (26) were grown in high-glucose (4.5 mg/L) DMEM with 10% heat-inactivated FCS, 100 units/mL penicillin, and 100 mg/mL streptomycin and maintained in a humidified incubator at 37°C and 5% CO₂. Primary cortical astrocyte cultures were prepared from 6-week-old Wistar rats as described previously (27) and were cultured in BME basal medium with 10% heat-inactivated FCS, 100 units/mL penicillin, 100 mg/mL streptomycin, and 5% glucose.

The expression plasmid green fluorescent protein-light chain 3 (GFP-LC3; ref. 28) was stably transfected using Metafectene reagent according to the manufacturer's instructions (Biontex). For selection of stably transfected cell lines (U343 GFP-LC3 and MZ-54 GFP-LC3), gentamicin was used. Expression plasmids mRFP-GFP-LC3 (29), TOPO-Mcl-1 (30), and mRFP-Mito (BD Biosciences) were transiently transfected using Metafectene reagent according to the manufacturer's instructions. Twenty-four hours after transfection, cells were subjected to the respective treatments as indicated.

Cells were fixed with 4% paraformaldehyde and permeabilized with 0.1% Triton X-100 in PBS. Cells were stained with the monomeric cyanine nucleic acid stain TO-PRO-3 at a final concentration of 1 μmol/L (Invitrogen/Molecular Probes) to allow ultrasensitive detection of double-stranded nucleic acids. For immunostaining of cytochrome c, a monoclonal anti–cytochrome c antibody (clone 6h2.b4, BD Biosciences) and a Cy3-labeled secondary antimouse antibody (Jackson ImmunoResearch Laboratories) were used. For detection of the active monomeric form of Bax, the monoclonal anti-Bax antibody (clone 6A7) and a Cy2-labeled secondary antimouse antibody (Jackson ImmunoResearch Laboratories) were used. For confocal microscopy, cells were mounted on 13-mm coverslips, cultured for 24 hours, transfected, treated, and finally analyzed using a Nikon C1i confocal microscope. The fluorescence of GFP, RFP, TO-PRO-3, Cy2, and Cy3 was recorded with the suitable filter sets (GFP fluorescence and Cy2: excitation 488 nm, emission 509 nm; TO-PRO-3: excitation 642 nm, emission 662 nm; RFP and Cy3: excitation 554 nm, emission 568 nm). Digital images were obtained using EZ-C1 Nikon software. For quantification of cytochrome c release and Bax activation, 100 cells from three different cultures were scored for each treatment (300 total).

Lentiviral vector stocks specific for Beclin1 (SHVRS-NM_003766), Atg5 (SHVRS-NM_004849), and Mcl-1 (SHVRS-NM_021960; Sigma Aldrich) were used for transduction of glioma cells. The target sets included five sequences for different small hairpins. The pLKO.1-puro control transduction particles (SHC001V) did not contain a hairpin insert and were used as a negative control. For the transduction, cells were plated in 96 well-plates and transduced the following day at a multiplicity of infection of 10. New medium was added to a final volume of 100 μL containing hexadimethrine bromide (Sigma-Aldrich) at a final concentration of 8 μg/mL. Cells were incubated for 24 hours before changing the medium. After overnight incubation, cells were washed, trypsinated, and transferred to six-well plates, after which puromycin (Calbiochem) was added at a final concentration of 5 μg/μL.

SDS-PAGE and Western blotting were done as described previously (31). The resulting blots were probed with a mouse monoclonal anti–Bcl-2 antibody diluted 1:50 (Santa Cruz Biotechnology), a rabbit polyclonal anti–Bcl-xL antibody diluted 1:500 (BD Biosciences), a rabbit polyclonal anti–Mcl-1 antibody diluted 1:200 (Santa Cruz Biotechnology), a rabbit polyclonal anti–Bcl-w diluted 1:200 (Stressgen Bioreagents), a rabbit polyclonal anti-Beclin1 antibody (Cell Signaling) diluted 1:1,000, a rabbit polyclonal anti-Atg5 antibody (Cell Signaling) diluted 1:1,000, a mouse polyclonal anti-mTOR antibody diluted 1:1,000 (Cell Signaling), a mouse monoclonal anti-LC3 antibody (Sigma-Aldrich) diluted 1:1,000, a mouse monoclonal anti-p62 (lck ligand; BD Biosciences) diluted 1:1,000, a mouse monoclonal anti–α-tubulin antibody (clone DM 1A, Sigma-Aldrich) diluted 1:1,000, and a mouse monoclonal anti–glyceraldehyde-3-phosphate dehydrogenase (GAPDH) antibody (Abcam) diluted 1:1,000.

For measuring effector caspase activity, treated cells were lysed in 200 μL of lysis buffer [10 mmol/L HEPES (pH 7.4), 42 mmol/L KCl, 5 mmol/L MgCl₂, 1 mmol/L phenylmethylsulfonyl fluoride, 0.1 mmol/L EDTA, 0.1 mmol/L EGTA, 1 mmol/L DTT, 1 μg/mL pepstatin A, 1 μg/mL leupeptin, 5 μg/mL aprotinin, 0.5% 3-(3-cholamidopropyldimethylammonio)-1-propane sulfonate (CHAPS)]. Fifty microliters of this lysate were added to 150 μL of reaction buffer [25 mmol/L HEPES, 1 mmol/L EDTA, 0.1% CHAPS, 10% Sucrose, 3 mmol/L DTT (pH 7.5)]. The fluorigenic substrate Ac-DEVD-AMC was added at a final concentration of 10 μmol/L. Accumulation of AMC fluorescence was monitored over 2 hours using an HTS fluorescent plate reader (excitation 380 nm, emission 465 nm). Protein content was determined using the Pierce Coomassie Plus Protein Assay reagent (KMF). Caspase activity was expressed as a change in fluorescence units per microgram of protein per hour.

Cell death was detected by flow cytometry after Annexin V-Fluos/propidium iodide (PI) double staining (Roche Applied Science; Sigma-Aldrich). All cells that were positive for Annexin and/or PI [i.e., cells from all quadrants except the bottom left one (Q3)] were considered dead. For analysis of mitochondrial membrane potentials (ΔΨ_m), cells were stained with 30 nmol/L TMRM (Invitrogen) for 15 minutes followed by flow cytometric analysis. To quantitatively detect changes in activation of the autophagosomal/lysosomal pathway, acidic vacuoles were stained with 25 nmol/L Lysotracker Red DND-99 (Invitrogen) for 30 minutes and washed twice with PBS, and the net amount of acidic vesicles was determined by flow cytometric analyses. In all cases, a minimum of 10⁴ events per sample were acquired. Flow cytometric analyses were done on a FACSCanto II (BD Biosciences) followed by analysis using FACSDiva software (BD Biosciences).

Cells were plated in 96-well-plates at 2,000 per well and cultivated for 24 hours before onset of the treatment. After treatment, 20 μL of the MTT stock solution (5 mg/mL) were added to 100 μL of medium in each well, followed by incubation at 37°C with a 5% carbon dioxide atmosphere for 3 hours. Following incubation, the medium containing the MTT reagent was removed, cells were solubilized by adding n-propyl alcohol/1 mol/L HCl (24:1), and lysates were gently mixed for 30 minutes. The absorbances at 560 nm were measured with a HTS fluorescent plate reader.

Data are given as means ± SEM. For statistical comparison, one-way ANOVA followed by Tukey's test was used using SPSS software (SPSS GmbH Software). P values <0.05 were considered to be statistically significant.

To analyze whether inhibition of antiapoptotic Bcl-2 family members is sufficient to induce glioma cell death, four different small-molecule BH3 mimetics were used in this study (15, 20). To block the activity of Bcl-2 and Bcl-xL, we applied the two synthetic, cell-permeable inhibitors BH3I-2′ and HA14-1. The BH3 mimetic HA14-1 has been described as a specific inhibitor of Bcl-2, whereas BH3I-2′ was previously shown to bind and inhibit Bcl-2 and Bcl-xL (32, 33). To investigate the dose-dependent cytotoxic effects of the BH3 mimetics in nontransformed cells, we performed initial experiments with primary astrocytes (Fig. 1A). In analogy to previously published data, BH3I-2′and HA14-1 at concentrations of 20 and 30 μmol/L, respectively, were not capable of inducing cell death in astrocytes, as well as in malignant glioma cell lines U343 and MZ-54 after 48 hours (Fig. 1A and B; refs. 26, 34). We also used the BH3 mimetic ABT-737, which is capable of inhibiting Bcl-2, Bcl-xL, and Bcl-w, but not Mcl-1 (35, 36). ABT-737 had moderate cytotoxic effects at concentrations of 10 μmol/L in astrocytes, but induced significant cell death at a higher concentration of 20 μmol/L (Fig. 1A). At 10 μmol/L, ABT-737 (35) also induced cell death in U343 and MZ-54 cells (Fig. 1B). In addition to the three synthetic BH3 mimetics, we used the natural BH3 mimetic gossypol, a pan-Bcl-2 family inhibitor. For our initial analyses, we compared the cytotoxic effects of racemic (±)-gossypol with those of the biologically active enantiomer (−)-gossypol in nontransformed astrocytes (Fig. 1A). To this end, concentrations of 10, 15, 20, and 30 μmol/L of (±)-gossypol and (−)-gossypol were added for 48 hours to cultures of primary astrocytes. Concentrations of up to 20 μmol/L were insufficient to induce cell death in astrocytes, whereas 30 μmol/L of (±)-gossypol and (−)-gossypol elicited significant cell death in the cultures (Fig. 1A). Interestingly, (−)-gossypol was capable of efficiently triggering cell death in U343 and MZ-54 cells at concentrations of 15 μmol/L (Fig. 1B).

Of the four investigated BH3 mimetics, only ABT-737 induced detectable effector caspase activity in U343 cells (37), indicating that gossypol induced a caspase-independent type of cell death in glioma cells (Fig. 1C) expressing Bcl-2, Bcl-xL, Mcl-1, and Bcl-w (Fig. 1D). Staurosporine was used as a positive control for induction of apoptotic cell death. To confirm that (−)-gossypol and ABT-737 targeted antiapoptotic Bcl-2 family members, we analyzed the activation of Bax, with the 6A7 antibody detecting the active monomeric form of Bax (Fig. 1E and F). To this end, we transiently transfected U343 cells with a plasmid encoding mRFP-Mito. Twenty-four hours later, cells were treated with 15 μmol/L (−)-gossypol and 10 μmol/L ABT-737 for 48 hours, fixed, and subjected to confocal analysis. Bax clustering and colocalization with mRFP-Mito were detected in ∼25% of (−)-gossypol–treated cells and ∼12% of ABT-737–treated U343 cells.

We then investigated in more detail the role of Mcl-1 in cell death resistance. To this end, we established stable Mcl-1–knockdown (KD) cell lines (Fig. 2A) and transiently overexpressed Mcl-1 in U343 cells (Fig. 2A). U343 control cells, Mcl-1–overexpessing cells, and Mcl-1 KD cells were treated with (−)-gossypol and ABT-737 for 48 hours, after which whole-cell lysates were analyzed for expression levels of Mcl-1. Western blot analysis indicated a pronounced decrease of Mcl-1 protein levels after (−)-gossypol treatment (Fig. 2B). Fluorescence-activated cell sorting (FACS) analyses with AnnexinV/PI revealed a significant increase of cell death from ∼12% to ∼30% and from ∼38% to ∼70% in Mcl-1 KD cells after treatment with ABT-737 and (−)-gossypol, respectively (Fig. 2C). These findings underscore the important role of Mcl-1 in cell death resistance of glioma cells.

To evaluate whether gossypol-induced cell death was associated with autophagy, we used U343 and MZ-54 cells stably transfected with GFP-LC3 and treated them with 15 μmol/L (−)-gossypol for up to 48 hours (Fig. 3A and B). Confocal microscopy analysis indicated that (−)-gossypol triggered vesicular localization of GFP-LC3. Quantitative analysis revealed that approximately 80% to 90% of U343 and MZ-54 cells displayed vesicular formation of autophagosomes after treatment with (−)-gossypol, whereas this percentage was much lower after treatment with the BH3 mimetics BH3I-2′ and HA14-1 (Fig. 3B). We also stained acidic vacuoles with Lysotracker Red and subsequently measured the extent of acidic vesicles by FACS analysis (Fig. 3C and D). Induction of autophagy is associated with a net increase in the percentage of cells displaying high Lysotracker fluorescence, as established for detection of rapamycin-induced autophagy elsewhere (38). The percentage of cells highly labeled with Lysotracker was significantly increased by (−)-gossypol. Again, in comparison with the three synthetic BH3 mimetics (HA14-1, BH3I-2′, and ABT-737), activation of the autophagosomal/lysosomal pathway was clearly much more potent after treatment with (−)-gossypol.

Previous studies have shown that the clinically relevant drug TMZ is capable of killing glioma cells (39), and we next asked whether (−)-gossypol was capable of enhancing cell death triggered by TMZ. The sensitivity of gliomas to TMZ is tightly correlated to the expression of the DNA repair enzyme MGMT (40). Therefore, we compared MGMT-positive U87 cells (41, 42) with MGMT-negative U343 cells (43) and analyzed the extent of cell death after treatment with 5, 10, and 15 μmol/L of (±)-gossypol and (−)-gossypol either with or without 100 μmol/L TMZ for 96 hours. The combined treatment of TMZ and (−)-gossypol exerted additive effects on the viability of U343 cells (Fig. 4A, top) and, to a lesser extent, on U87 cells (Fig. 4A, bottom). These combined effects were confirmed by FACS analyses for PI uptake and Annexin binding (Fig. 4C). Of note, the obtained FACS profiles showed no major Annexin V–positive, PI-negative cell fraction after treatment with TMZ, (−)-gossypol, and the combination of the two drugs, suggesting the absence of apoptotic cell death (Fig. 4B). Because Annexin-positive, PI-negative cells have been observed in other models of autophagic cell death (44), we also used a caspase activity assay, with staurosporine serving as a positive control for caspase activation. Similar to single-agent treatment with higher concentrations of gossypol (Fig. 1C), the combinatorial treatment with 5 μmol/L (−)-gossypol and 100 μmol/L TMZ did not induce measurable effector caspase induction, confirming the absence of apoptotic cell death (Fig. 4B).

To further characterize the mode of cell death induced by gossypol and TMZ, we analyzed U343 GFP-LC3 cells for formation of LC3-positive vacuoles, nuclear alterations, and subcellular distribution of cytochrome c by confocal fluorescence microscopy (Fig. 5A and B). Representative z-stack confocal pictures of all labelings (TO-PRO-3, GFP-LC3, and cytochrome c) indicated that (−)-gossypol, TMZ, and the combined treatment increased the formation of autophagosomes (Fig. 5A, arrowheads). Most of the cells revealed loss of the filamentous staining pattern of cytochrome c, indicating fragmentation of mitochondria (Fig. 5A, arrows), whereas a fraction of cells had released their cytochrome c (Fig. 5B). Quantitative analyses indicated that ∼5% of U343 cells and ∼15% of U343 cells had released their cytochrome c after 96 hours of treatment with (−)-gossypol and the combined treatment with (−)-gossypol and TMZ in our confocal analyses, respectively (Fig. 5C), a time point at which ∼40% and ∼75% of cells were already dead as measured by FACS analysis (Fig. 4B, top). Collectively, these data suggest that release of cytochrome c may be a late event in (−)-gossypol–induced cell death (37).

mTOR is an established main regulator of autophagy (11, 45). Because mTOR serves to inhibit the autophagic pathway, mTOR inhibitors such as rapamycin can activate autophagy (11). To characterize the effects of rapamycin on proliferation of glioma cells, we performed MTT assays and FACS analysis of PI uptake and AnnexinV binding. U87 cells were treated with rapamycin at concentrations of 0.1, 1, 10, and 1,000 nmol/L and cultured for 24, 48, and 72 hours (Fig. 6A). Although our MTT assays indicated that rapamycin elicited a time-dependent, antiproliferative effect, it was rather moderate, even in the highest applied concentration of 1,000 nmol/L (Fig. 6A). In further flow cytometry experiments, we analyzed whether rapamycin-mediated mTOR inhibition might act in a synergistic manner with (−)-gossypol to induce autophagy and cell death. U87 cells were seeded and exposed to rapamycin (0.1, 1, 10, and 1,000 nmol/L) with or without 5 μmol/L (−)-gossypol for 96 hours. Although rapamycin did significantly enhance cell death induced by (−)-gossypol, the 0.potentiating effects were again moderate (Fig. 6B).

mTOR can form two major signaling complexes: mTOR/Raptor (TORC1) and mTOR/Rictor (TORC2; ref. 46). Because rapamycin can only inhibit the TORC1, but not the TORC2, complex (46), we additionally applied LNT-229 glioma cells with a stable knockdown of mTOR (Fig. 6C; ref. 47). In comparison with vector-transfected control cells, there was an increase in the basal levels of acidic vesicles as determined by FACS analysis of Lysotracker fluorescence in LNT-229 mTOR-KD cells (Fig. 6D), which displayed efficiently suppressed protein levels of mTOR as analyzed by Western blot (Fig. 5C). Likewise, induction of autophagy by (−)-gossypol and (−)-gossypol plus TMZ was also increased in LNT-229 mTOR-KD cells (Fig. 6D). In congruence to enhanced induction of autophagy, cell death triggered by (−)-gossypol, TMZ, and a combinatorial treatment with (−)-gossypol and TMZ was also potentiated in LNT-229 mTOR-KD cells versus control cells (Fig. 6E). To investigate whether this potentiation was associated with enhanced mitochondrial dysfunction, we measured mitochondrial membrane potentials (ΔΨ_m) after uptake of TMRM by flow cytometry. Treatment with staurosporine (6 hours, 3 μmol/L) served as a positive control. Similar to the effects on autophagy and cell death, the extent of mitochondrial dysfunction was visibly higher in mTOR-KD cells versus control cells (Fig. 6F). In contrast to autophagic cell death, knockdown of mTOR paradoxically reduced the amount of cell death induced by vincristine, a drug known to promote apoptotic cell death in glioma cells (Fig. 6G; ref. 48), as has been previously reported (47). Thus, the effect of mTOR knockdown on autophagy observed after treatment with (−)-gossypol is specific and does not reflect generally reduced resistance toward cell death in these cells. Rather, these data support a dual function of mTOR that may positively or negatively modulate cell death in a stimulus-dependent manner (49).

Our data obtained thus far had suggested that activation of autophagy and cell death coincide and are tightly correlated in gossypol-induced cell death of malignant glioma cells. However, these observations did not prove that autophagy played a death-promoting role in this type of cell death. To analyze a potential contributory role of autophagy in caspase-independent cell death induced by gossypol, we performed a lentiviral knockdown of both Beclin1 and Atg5 (7, 9) in three different glioma cell lines (U87, MZ-54, and U343; Fig. 7A). Immunoblotting using antibodies against Atg5 and Beclin1 indicated efficient silencing of both genes. Control cells, Beclin1-KD cells, and Atg5-KD cells were treated with 5 μmol/L (−)-gossypol, 100 μmol/L TMZ, and a combination of the two drugs for 96 hours. C₂-ceramide served as a positive control for induction of autophagy. As to be expected, drug-induced autophagy was strongly reduced in Beclin1-KD and Atg5-KD cells in comparison with control cells (Fig. 7B). Cell death analysis with double staining of Annexin V/PI and flow cytometry indicated that knockdown of Beclin1 and Atg5 dramatically decreased the amount of cell death in all three investigated cell lines (U87, MZ-54, and U343; Fig. 7B), indicating that autophagy indeed played an essential death-promoting role in this type of cell death.

To exclude the possibility that the observed increase of cells with LC3-positive vacuoles under (−)-gossypol treatment was caused by a reduction of autophagic flux due to impaired fusion of autophagosomes with lysosomes, rather than by activation of autophagy, we performed a number of additional experiments. To study the autophagic flux, we monitored the conversion of the cytosolic form of LC3 (LC3-I) to the membrane-bound lipidated form (LC3-II) and the degradation of p62, which is selectively degraded by autophagy, by Western blots both in the presence and in the absence of the lysosomotropic agent and autophagic flux inhibitor BAF (Fig. 8A). U343 control cells and U343 Beclin1-KD cells were treated with 5 μmol/L (−)-gossypol with or without 100 μmol/L TMZ for 72 hours and then further treated with BAF for 4 hours. Whole-cell lysates were subjected to immunoblotting with specific antibodies against LC3 and p62. The U343 control cell line indicated elevated conversion of LC3-I to LC3-II and degradation of p62 after treatment with (−)-gossypol and combined treatment with (−)-gossypol and TMZ, again confirming activation of autophagy (Fig. 8A). The levels of LC3-II and p62 were increased in the presence of BAF, indicating an intact autophagic flux under treatment with (−)-gossypol and (−)-gossypol plus TMZ. In contrast, there was neither degradation of p62 nor an increase of LC3-II in U343 Beclin1-KD cells in the absence of BAF (Fig. 8A). To further validate these findings, we used an LC3 tandem fluorescence construct allowing to discriminate between autophagosomal and autolysosomal LC3 (29). U343 wild-type cells were treated with 5 μmol/L (−)-gossypol and 100 μmol/L TMZ (72 hours) and further treated with 10 nmol/L BAF for 4 hours to monitor the autophagic flux at the single-cell level. Untreated U343 cells displayed bright, diffuse GFP and mRFP fluorescence signals (Fig. 8B). U343 cells treated with both drugs displayed enhanced formation of autophagosomes (colocalized fluorescence of GFP and mRFP) and a parallel increase of autolysosomes (only mRFP fluorescence). BAF significantly reduced the number of mRFP-positive autolysosomes under treatment with (−)-gossypol and TMZ, indicating active, ongoing autophagy in the absence of BAF. Finally, we analyzed Lysotracker Red uptake at the single-cell level to show that the lysosomal function was not disrupted by (−)-gossypol and TMZ. In our confocal analyses, U343 cells treated with (−)-gossypol and TMZ revealed a selective lysosomal signal of Lysotracker Red, indicating lysosomal integrity. This lysosomal signal was largely lost when cells were treated with the lysosomal inhibitor BAF (Fig. 8C). These data suggest that the autophagosomal/lysosomal pathway is fully active under treatment with gossypol and TMZ in glioma cells.

A high resistance against classic caspase-dependent apoptosis is a major hallmark of malignant gliomas (2, 3). Activation of the multiprotein proapoptotic Bcl-2 family members Bax and Bak is a critical step in the activation of the mitochondrial (intrinsic) pathway of apoptosis and precedes mitochondrial release of cytochrome c and Smac/Diablo and ensuing apoptosome formation (13). The intrinsic pathway is antagonized by antiapoptotic Bcl-2 family members, which are known to be highly overexpressed in malignant gliomas (34, 50). In addition to their role in regulation of apoptosis, prosurvival Bcl-2 family members are also capable of antagonizing caspase-independent types of cell death and autophagy (16-19, 34, 50). We have previously shown that the resistance of malignant glioma cells to caspase-independent cell death is tightly correlated to the expression levels of Bcl-2 and Bcl-xL (34).

To analyze whether inhibition of prosurvival Bcl-2 proteins was per se sufficient to induce glioma cell death, and whether it was capable of sensitizing glioma cells to chemotherapeutic drugs, we used four different BH3 mimetics with different binding profiles to Bcl-2 family members. The BH3 mimetics BH3I-2′ and HA14-1 are specific for Bcl-2 and Bcl-xL (15, 20), whereas ABT-737 can inactivate Bcl-2, Bcl-xL, and Bcl-w, but not Mcl-1 (15, 20). We also used the natural pan-Bcl-2 inhibitor (−)-gossypol capable of inhibiting all four antiapoptotic Bcl-2 family members (15, 20). Of note, (−)-gossypol (AT-101), which is clinically developed by the company Ascenta, is orally applicable and well tolerated (51). Phase I/phase II clinical trials with (−)-gossypol (AT-101) are currently ongoing in malignant glioma, and a follow-up will reveal the possible effect of (−)-gossypol (AT-101) on long-term survival.

Interestingly, glioma cells displayed only limited sensitivity to single-agent treatments with the BH3 mimetics BH3I-2′ and HA14-1, whereas the pan-Bcl-2 inhibitor (−)-gossypol was capable of potently inducing cell death. These data suggest that inhibition of Bcl-2 and Bcl-xL did not suffice to trigger glioma cell death in the absence of metabolic stress or additional death stimuli. ABT-737, which also inhibits Bcl-w, also elicited detectable cell death in U343 and MZ-54 cells, albeit to a lower extent than (−)-gossypol. Therefore, inhibition of Mcl-1 may be crucial for overcoming the intrinsic cell death resistance of malignant glioma. In line with this hypothesis and previously published findings (35), silencing of Mcl-1 was required to efficiently trigger ABT-737–dependent cell death.

Due to inhibition of mitochondrial outer membrane permeabilization and mitochondrial dysfunction, prosurvival Bcl-2 family members prevent increased reactive oxygen species production, thereby acting in an antioxidative manner. In addition, they serve to modulate the cellular redox homeostasis through elevated expression of antioxidant enzymes and by increasing the levels of cellular glutathione (52). Therefore, inhibition of prosurvival members of the Bcl-2 family by BH3 mimetics, especially by pan-Bcl-2 inhibitors, might also hamper the antioxidative capacity of cells. Given the fact that oxidative stress is also known to stimulate autophagy in several experimental paradigms (53, 54), it is presumably involved in the prominent induction of gossypol-dependent autophagy observed in our experiments. Indeed, (−)-gossypol was previously shown to be a potent inductor of oxidative stress in colorectal cancer cells (22). Interestingly, we observed significantly reduced protein levels of Mcl-1 after treatment with (−)-gossypol, indicating stress-triggered degradation of Mcl-1, probably through the coordinated action of stress-induced c-Jun NH₂-terminal kinases and glycogen synthase kinase 3 (55).

There is also experimental evidence pointing to alternative mechanisms involved in gossypol-induced cell death. The fact that gossypol can kill MEFs in a Bax/Bak–independent manner may suggest that its cytotoxic activity is not solely related to its capacity to bind to prosurvival Bcl-2 proteins (37, 56). However, it was also proposed that similar to the short Nur77-derived peptide NuBCP-9 (57), (−)-gossypol can change the conformation of Bcl-2, thereby transforming it into a proapoptotic molecule not requiring the presence of Bax and Bak (56). In addition to the potential effects of BH3 mimetics on oxidative stress (as discussed above), these observations may shed further light on the death-promoting effects of (−)-gossypol in Bax/Bak–deficient cells.

The DNA repair enzyme MGMT is inactivated by epigenetic silencing in a significant portion of malignant gliomas, and the MGMT status is an established predictive marker for the clinical response to the alkylating drug TMZ and for clinical outcomes (40, 58-60). In line with these observations, (−)-gossypol augmented cell death induced by TMZ in MGMT-negative U343 cells, whereas the combined effects of the two drugs were much lower in U87 cells, known to express MGMT after TMZ treatment (42). Interestingly, it was proposed that the cytotoxic effects of TMZ, at least in part, are exerted through induction of autophagy and type II cell death (39).

Although stimulation of autophagy may primarily constitute an endogenous protective stress response under conditions of metabolic stress, enforced overactivation of autophagy can also lead to autophagic type II cell death in many experimental paradigms (10). In our experiments, cell death triggered by (−)-gossypol and (−)-gossypol in combination with TMZ in glioma cells displayed characteristics of type II cell death, as it was associated with nuclear condensation, cell shrinkage, translocation of GFP-LC3 to autophagosomes, and ensuing autophagic flux of LC3 to autolysosomes. Cell death was also associated with cytochrome c release in a subfraction of cells, but occurred in the absence of apparent lysosomal damage and detectable effector caspase activation. These data suggest that neither lysosomal permeabilization nor activation of the intrinsic pathway of apoptosis was required for gossypol-induced cell death in glioma cells.

A major endogenous regulator of autophagy is the kinase mTOR, and the disassembly of mTOR complexes TORC1 and TORC2 has been implicated in activation of autophagy (61). In our experiments, inhibition of mTORC1 by rapamycin elicited only modest effects on cell proliferation and on the sensitivity to combined treatment with (−)-gossypol and TMZ in U87 cells. In contrast, inactivation of mTORC1 and mTORC2 by stable RNA interference against mTOR sensitized LNT-229 cells against induction of autophagy and cell death triggered by (−)-gossypol and (−)-gossypol in combination with TMZ. These data suggest that mTORC2 has an essential contributory role in the inhibition of autophagy and autophagy-dependent cell death in glioma cells (62). In a reciprocal fashion to our results obtained with the mTOR-KD cells, knockdown of Beclin1 and Atg5 in U87, U343, and MZ-54 cells potently diminished the extent of cell death induced by (−)-gossypol and the combined treatment with TMZ, indicating that autophagy indeed significantly contributed to this type of cell death.

Collectively, our data suggest that pan-Bcl-2 inhibitors induce caspase-independent, autophagic cell death in apoptosis-resistant malignant glioma cells. Based on the differential effects of (−)-gossypol in comparison with the more specific inhibitors and on our Mcl-1 knockdown data, neutralization of Mcl-1 is likely essential for efficient killing of glioma cells.

No potential conflicts of interest were disclosed.

We thank Hildegard Schweers for excellent technical assistance, Dr. Noboru Mizushima (The Tokyo Metropolitan Institute of Medical Science, Tokyo, Japan) and Dr. Tamotsu Yoshimori (National Institute of Genetics, Shizuoka, Japan) for providing the GFP-LC3 and mRFP-GFP-LC3 plasmids, and Dr. Ulrich Maurer (Institute of Molecular Medicine and Cell Research, Albert Ludwigs University Freiburg, Freiburg, Germany) for providing the Mcl-1 expression plasmid.

Grant Support: Deutsche Krebshilfe grant 108795 and Wilhelm Sander Stiftung grant 2005.067.1 (D. Kögel).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.