Interference with the HSF1/HSP70/BAG3 Pathway Primes Glioma Cells to Matrix Detachment and BH3 Mimetic–Induced Apoptosis

Glioblastoma is the highest grade (grade IV) glioma and the most aggressive and frequent primary brain tumor in humans (1, 2). Diffuse infiltrative growth into the surrounding brain tissue which prevents complete tumor resection is a major complication in the treatment of this cancer (3). Patients with glioblastoma are treated with combined radiochemotherapy, which prolongs survival for several months only (4, 5), as the intrinsic apoptosis resistance of residual tumor cells impedes their elimination (6–9), thereby significantly contributing to the rapid and constant manifestation of tumor recurrence.

The Bcl-2 homology domain 3 (BH3) mimetic class of drugs competitively disrupts the interaction between proapoptotic and antiapoptotic proteins of the Bcl-2 family (10, 11), thereby either facilitating or directly inducing apoptosis. The (−)enantiomer of gossypol (AT-101) is a natural polyphenolic compound derived from cottonseeds and acts as a pan-Bcl-2 antagonist inactivating Bcl-2, Bcl-xL, Mcl-1, and Bcl-w (11, 12). In cancer cells with an intact apoptotic machinery, AT-101 induces the mitochondrial/intrinsic pathway of apoptosis (11–13). Furthermore, AT-101 activates the autophagic pathway by releasing BECN-1 from its inhibitory interaction with Bcl-2 (14, 15). We have previously demonstrated that AT-101 triggers a nonapoptotic, autophagy-dependent cell death in malignant glioma cells that was significantly reduced by gene silencing of ATG5 and BECN-1 (7, 16). However, it is currently unclear why AT-101 does not induce apoptosis in glioma cells, and the resistance mechanisms underlying this phenomenon remain to be uncovered.

Autophagy is an evolutionarily conserved process in which cellular constituents are delivered to the autophagosomal–lysosomal pathway for bulk degradation (17). This type of autophagy characterized by formation of autophagosomes and their subsequent fusion with lysosomes is termed macroautophagy. Other forms of autophagy are microautophagy and chaperone-mediated autophagy (CMA; ref. 18). BAG3 (Bcl-2–associated athanogene 3, also called Bis)—a HSP70 co-chaperone—mediates the degradation of damaged client proteins by a selective form of macroautophagy, thus maintaining cellular function by recruitment of the macroautophagy pathway (19). There is evidence that BAG3 is overexpressed in several types of cancers including glioma (20, 21). In addition, BAG3 was shown to support cell survival and underlying the resistance to chemotherapy in several human cancers (22, 23), in part, by supporting the antiapoptotic activity of prosurvival Bcl-2 family members through preventing their degradation via the proteasomal pathway (21, 24).

Here, we investigated the role of the HSF1/HSP70/BAG3 pathway in resistance of glioma cells to apoptosis induced with the BH3 mimetics AT-101 and ABT-737. Our data show that silencing of BAG3 efficiently increased AT-101–induced cell death and reactivated apoptosis in U251 and U343 glioma cells. We also report that the apoptosis-sensitizing effects of BAG3 silencing can be mimicked by interrupting the HSF1/HSP70/BAG3 pathway with KRIBB11 (a HSF1 inhibitor) and YM-1 (a BAG3/HSP70 complex inhibitor; refs. 25, 26).

ABT-737 was obtained from Santa Cruz. AT-101 [>98% purity, the (−)enantiomer of gossypol (for chemical structure, refer to ref. 27)] was acquired from Tocris (Bristol). KRIBB11 and MG-132 were purchased from Calbiochem. Staurosporine was obtained from Alexis Biochemicals. Wortmannin was received from Alexis Pharmaceuticals. Z-Val-Ala-dl-Asp-fluoromethylketone (z-VAD) was purchased from Bachem. Laminin-rich extracellular matrix (lrECM; BME Growth Factor Reduced PathClear) was ordered from Biozol. YM-1 (2-((Z)-((E)-3-Ethyl-5-(3-methylbenzo[d]thiazol-2(3H)-ylidene)-4-oxothiazolidin-2-ylidene)methyl)-1-methylpyridin-1-ium chloride), an analog of MKT-077 (for structure, refer to ref. 2), temozolomide, and all other chemicals were acquired from Sigma-Aldrich.

U-251 MG (formerly known as U-373 MG; ECACC 09063001) hereafter called U251 and Gos-3 (DSMZ No.: ACC 408) hereafter called U343 were a kind gift of Dr. Werner Paulus, Institute of Neuropathology, University of Münster (Münster, Germany). Both cell lines were STR-profiled in November 2015 by ATCC. U87 cells were obtained from ATCC and STR-profiled in May 2016. Glioma cells were grown and subcultured in DMEM GlutaMAX, with 10% heat-inactivated FCS, 1% glutamine, and 1% penicillin/streptomycin (all: Gibco/Invitrogen) added to the media. Cultivation occurred in a humidified incubator at 37°C and 5% CO₂.

Lentiviral shRNA specific for BAG3 (SHCLNV-NM_004281; TRCN0000007294; Sigma Aldrich) was used for transduction of U251, U343 and U87 cells. Transduction was performed as previously described (29).

The cells were seeded on 13-mm coverslips and cultured for 24 hours. After respective treatment, cells were fixed with paraformaldehyde (4% PFA, 4% sucrose) and then permeabilization was performed with 0.1% Triton X-100 and cells were stained with DAPI (AppliChem). For immunostaining of cytochrome c, a mouse monoclonal anti–cytochrome c antibody (clone 6h2.b4, BD Biosciences) and a Cy3-labeled secondary anti-mouse antibody (Jackson ImmunoResearch Laboratories) were used. Subsequently samples were mounted on a microscope slide. For analysis of cytochrome c release, samples were analyzed using a Nikon C1i confocal microscope. Digital images were obtained using EZ-C1 Nikon software and processed with ImageJ. For assessment of cellular morphology, cells were stained with DAPI and Texas Red-X Phalloidin (ThermoFisher Scientific), and samples were analyzed using a Leica SP8 laser-scanning microscope (Leica).

Cells were lysed for protein analysis with SDS lysis buffer containing protease and phosphatase inhibitors. An applicable amount of protein (60–80 μg) was applied on a 10%, 12%, or 15% SDS-PAGE, respectively. Subsequently, proteins were separated by electrophoresis at 135 V and then blotted to nitrocellulose membranes (Protean BA 83; 2 lm; Schleicher & Schuell) at 15 V for 35 minutes in Towbin buffer [25 mmol/L Tris, 192 mmol/L glycine, and 20% methanol (v/v)]. Membranes were blocked for 1 hour at room temperature in blocking buffer (5% BSA, 50 mmol/L Tris-HCl, pH 7.5, 150 mmol/L NaCl, and 0.05% Tween-20) and then blots were probed overnight at 4°C with a rabbit polyclonal anti-BAG3 antibody diluted 1:2,500 (Biozol/Abnova), an anti-BAG1 antibody used according to Gamerdinger and colleagues (19), a rabbit polyclonal anti-Bax antibody, a rabbit monoclonal anti–Bcl-2 antibody, a rabbit monoclonal anti–Bcl-xL antibody, a mouse monoclonal anti-EZH2 antibody, a rabbit polyclonal anti-HSF1 antibody, a rabbit polyclonal anti-HSP70 antibody, a rabbit monoclonal anti–Mcl-1 antibody, a rabbit polyclonal anti-LC3 antibody (Sigma), a rabbit monoclonal anti-p65 antibody, a rabbit polyclonal anti-survivin antibody (R&D Systems), a mouse monoclonal anti-actin antibody (Sigma) diluted 1:500, a mouse monoclonal anti-GAPDH antibody (Calbiochem) diluted 1:10,000, and a mouse monoclonal anti-tubulin antibody (Sigma) diluted 1:10,000. Unless stated otherwise, antibodies were obtained from Cell Signaling Technology and diluted 1:1,000. After incubation with secondary antibodies, Cy2-/Cy3-conjugated goat anti-mouse and goat anti-rabbit, respectively (LI-COR), were diluted 1:10,000 for 1 hour at room temperature and signal was detected using an Odyssey Imaging System (LI-COR Biosystems).

Flow cytometry was performed as previously described (29).

Caspase-3–like protease activity was determined as previously described (29).

Freshly prepared p-HEMA solution (1 mg/mL in 95 % ethanol) was pipetted into the appropriate wells and dried at 37°C for 2 days with lids in place. After washing with PBS, cells were seeded and remained in suspension throughout the experiment.

For measurement of three-dimensional (3D) clonogenic survival, cells were embedded in 0.5 μg/μL lrECM supplemented with DMEM, 10% FCS as previously described (30, 31). After 24 hours, cells were either treated with 5, 10, or 15 μmol/L AT-101 and 0.2 % DMSO as solvent control for 2 hours or with 1, 10, or 100 μmol/L temozolomide and 0.1% DMSO control for 24 hours. To determine long-term survival, cells were washed 5 times and incubated for 8 days, followed by microscopic counting of 3D grown colonies >50 cells. Representative photographs of typical colony formation were obtained using an Axio Observer microscope (Carl Zeiss). Calculation of plating efficiencies were as follows: numbers of colonies formed divided by numbers of cells plated. Relative plating efficiencies were normalized to U251 or U343 control plating efficiencies, respectively. Surviving fractions after inhibitor or chemotherapy drug treatment were calculated as follows: numbers of colonies formed/[numbers of cells plated (treated) × plating efficiency (DMSO control)].

For immunoprecipitation (IP) against GFP, cells were transfected with GFP-Bax (32) and GFP-HSP70 (33) as described elsewhere (34). Subsequently, cells were lysed with lysis buffer (20 mmol/L HEPES, pH 7.4, 150 mmol/L NaCl, 0.1% Triton-X) for 20 minutes on ice. After 5 repeated freeze–thaw cycles, GFP-trap beads (Chromotek) were added and incubated for 1 hour at 4°C under gentle agitation. Beads were washed 3 times with ice-cold lysis buffer and 3 times with ice-cold PBS. Finally, beads were eluted in SDS-loading buffer containing 5% 2-mercaptoethanol (Sigma) on a shaker for 30 minutes at 95°C. For analysis, SDS-PAGE and Western blotting were used.

IP against active Bax was performed as previously described (34).

Cells were seeded and cultured for 24 hours; after treatment, cells were lysed and cytosolic as well as nuclear extract fractions were prepared using the NE-PER Kit (ThermoFisher Scientific) according to the manufacturer's instructions.

Cells were seeded and cultured for 24 hours after 2-hour preincubation with mitomycin to avoid proliferative effects. After respective treatment, cell adhesion was determined using CytoSelect 48-Well Cell Adhesion Assay (ECM Array, Fluorometric Format) from Cell Biolabs according to the manufacturer's instructions.

Orthotopic transplantation of glioma cells was performed as described previously (35, 36). Briefly, 5 × 10⁵ U343 cells expressing either shctrl (n = 14) or shBAG3 (n = 15), respectively, were transplanted into the right striatum of 8- to 10-week-old athymic nude mice (Foxn1^nu/nu). The mice were constantly monitored for clinical symptoms. Tumor-bearing mice were sacrificed when their clinical symptoms appeared and considered as a death event for the Kaplan–Meier analysis. Forty days after tumor transplantation, 3 mice of each group were sacrificed and the brains were collected, fixed, and embedded in paraffin for histologic analyses by routine hematoxylin and eosin (H&E) staining. All animal experiments were performed in accordance with the German animal protection law approved by the Regional Administrative council (Regierungspräsidium Darmstadt).

Data are given as means ± SEM. For statistical comparison, one-way ANOVA followed by Tukey test was used employing IBM SPSS software (IBM). To compare 3D plating efficiencies of ctrl versus BAG3KD cells, the unpaired 2-tailed t test was applied (GraphPad Prism, GraphPad Software, Inc.). Kaplan–Meier analyses were performed using GraphPad Prism. P < 0.05 was considered to be statistically significant.

As a starting point, BAG3 protein levels were determined in a panel of different glioma cell lines by Western blot analysis (Fig. 1A, left). BAG3 was highly expressed in most cell lines, with U251 cells showing the highest BAG3 protein levels, whereas U343 cells showed a medium/high BAG3 expression. To further investigate the oncogenic function of BAG3, cell lines with stable lentiviral knockdowns (KD) of BAG3 were established in U251 and U343 cells (Fig. 1A, right). Overexpression of BAG3 in U251 and U343 cells was also confirmed in comparison to SH-SY5Y cells that express low endogenous levels of BAG3 but strongly induce BAG3 expression after treatment with the proteasome inhibitor MG132 (ref. 37; Fig. 1B). Indeed, MG132-induced BAG3 levels in SH-SY5Y cells did not exceed basal BAG3 expression levels in the 2 glioma cell lines. Protein analysis by Western blotting also revealed a slight, further increase of BAG3 expression following treatment with the BH3 mimetic AT-101 (48 hours) in U251 and U343 cells (Fig. 1B).

In comparison to BAG3-proficient control cells, BAG3-depleted U251 cells showed a significant increase of early apoptosis and overall cell death after 24- and 48-hour treatment with AT-101 (Fig. 1C). In combined treatment with the pan-caspase inhibitor z-VAD, total cell death was strongly decreased, pointing to a caspase-dependent form of cell death induced by AT-101 in BAG3KD cells. Quantification of cell death revealed higher amount of cells in the bottom right panel representing early apoptotic cells in U251 BAG3KD cells (Fig. 1D). A caspase-3–like activity assay was performed under the same conditions (Fig. 1E). After 48 hours of AT-101 treatment, DEVD cleavage was significantly higher in U251 BAG3KD cells than in control cells and combined treatment with z-VAD completely blocked caspase-3–like activity.

Flow cytometry after Annexin-V/propidium iodide (PI) staining showed similar effects in U343 cells: after 48 hours of AT-101 treatment, a significant increase of early apoptosis and cell death could be observed in BAG3KD cells compared with control cells, which was significantly reduced after combined treatment with z-VAD (Fig. 1F). These results obtained with both KD cell lines were confirmed with polyclonal BAG3KD cultures in U251 and U343 cells (data not shown). To confirm these results in another independent cell model, transient lentiviral BAG3 depletion was performed in U87 cells (Supplementary Fig. S1A). Here, cell death was significantly enhanced in BAG3-depleted cells after AT-101 treatment (Supplementary Fig. S1B). We also observed a minor elevation of apoptosis under control conditions in U87 BAG3KD cells, underlining the cytoprotective role of BAG3.

To analyze induction of the intrinsic pathway of apoptosis, we monitored cytochrome c release from the mitochondria after treatment with AT-101 in the presence of z-VAD to prevent caspase activation and to limit cell shrinkage after cytochrome c release. After AT-101 treatment, cytochrome c release (indicated by cytoplasmic staining) was more pronounced in U251 BAG3KD cells as compared with control cells (Supplementary Fig. S2). In line with these observations, mitochondrial membrane potentials were significantly reduced in U251 (Fig. 2A) and U343 (Fig. 2B) BAG3KD cells compared to control cells after 24 and 48 hours of AT-101 treatment, respectively. These data indicate an increased permeabilization of the mitochondrial membrane in BAG3KD cells treated with AT-101. Furthermore, we also performed an IP against active Bax that was considerably increased in U251 BAG3KD cells compared with control cells (Fig. 2C), again demonstrating robust induction of the mitochondrial apoptosis pathway in BAG3KD cells.

Western blot analysis was performed in U251 and U343 cells to investigate the possible effect of AT-101 on the protein levels of Bcl-2 proteins and on induction of autophagy. Bcl-2 and Bcl-xL protein levels were strongly reduced after AT-101 treatment (Fig 2D). Combined treatment of AT-101 with the proteasome inhibitor epoxomicin did prevent the degradation of Bcl-2 and Bcl-xL. In contrast, addition of the autophagy inhibitor wortmannin did not show any effects on the levels of Bcl-2 and Bcl-xL. Therefore, the observed AT-101–triggered degradation of Bcl-2 and Bcl-xL seems to be mediated via the proteasomal pathway. In U251 BAG3KD cells, an almost complete depletion of Bcl-2 and Bcl-xL after 48-hour AT-101 treatment was detected, whereas Bcl-2 and Bcl-xL protein levels were less significantly reduced in control cells (Fig 2E). U343 BAG3KD cells showed a similar decrease under these conditions. Analysis of the LC3 switch revealed no discernible difference between the control cells and the BAG3KD cells (both in U251 and in U343), suggesting that bulk (macro)autophagy was not influenced by BAG3 depletion. Also, no difference of BAG1 expression levels were detected in BAG3KD cells compared with control cells. Therefore, a hypothetic “back”-shift from BAG3 to BAG1 expression in the BAG3KD cells did not occur.

Because BAG3 was previously shown to regulate cell adhesion (38), we investigated the effects of BAG3 depletion on cell adhesion properties next. Confocal imaging after F-actin staining with phalloidin revealed strongly altered cell morphology in U251 BAG3KD cells compared with control cells (Fig. 3A, see also Supplementary Fig. S2). In comparison to BAG3KD cells, BAG3-proficient cells displayed an increased cell size and more flattened and tightly attached cellular morphology. In contrast, BAG3KD cells appeared more rounded and loosely attached with many membrane protrusions. After 48 hours of AT-101 treatment, remnant parts of the actin skeleton were still detectable in shrunken but still viable control cells, whereas most BAG3KD cells were completely shrunken and rounded in the absence of an intact actin cytoskeletal structure. To quantitatively assess matrix binding, we performed a cell adhesion assay with 5 different extracellular matrix (ECM) proteins (fibronectin, collagen I, collagen IV, laminin I, and fibrinogen; Fig. 3B). This assay clearly demonstrated a significantly reduced adhesion of U251 BAG3KD cells to all investigated ECM proteins under control conditions which was further enhanced by AT-101 treatment for 24 hours, a time point at which the amount of cell death is still relatively moderate (Fig. 1C). To further determine the role of matrix detachment in AT-101–induced cell death, cell adhesion was impeded by coating cell culture dishes with p-HEMA. Under these conditions, U251 ctrl cells showed a robust increase of phosphorylated FAK, whereas an almost complete block of FAK phosphorylation was observed in BAG3KD cells cultivated in suspension (Fig. 3C). In suspension cultures, cell death was significantly increased in comparison to monolayer cultures, and BAG3-depleted cells showed a further increase of cell death after 5 μmol/L AT-101 treatment for 24 hours (Fig. 3D). To evaluate the long-term effects of BAG3 depletion in response to treatment with AT-101 or temozolomide on glioblastoma cell survival, we applied a 3D clonogenic assay more closely approaching physiologic conditions (39–41). BAG3 depletion significantly reduced the 3D plating efficiency of U251 cells relative to control cells, whereas clonogenic survival of U343 cells was not affected (Fig. 3E; Supplementary Fig. S3A). Treatment with increasing concentrations of the BH3 mimetic AT-101 significantly reduced clonogenic survival of U251- and U343 BAG3-depleted cells in comparison to control cells (Fig. 3F; Supplementary Fig. S3B). Similarly, BAG3 depletion significantly sensitized U251 cells to temozolomide treatment, whereas treatment of U343 BAG3KD cells with 10 and 100 μmol/L temozolomide resulted in decreased colony formation (Fig. 3G; Supplementary Fig. S3C).

HSF1 is a stress-induced transcription factor acting upstream of HSP70 and its co-chaperone BAG3. Hypothetically, Mcl-1 is stabilized by HSP70/BAG3 (24) and thereby degraded after inhibition of HSF1. To further analyze the contribution of the HSF1/HSP70/BAG3 pathway, the HSF1 inhibitor KRIBB11 was used in comparison to the experiments performed with the BAG3 depletion models. HSP70, BAG3, and Mcl-1 protein levels all were reduced by KRIBB11 in a time- and dose-dependent manner in U251 control cells (Fig. 4A). In combined KRIBB11 treatment with AT-101 and ABT-737, early apoptosis and total cell death were significantly induced in U251 cells (Fig. 4B) and U343 cells (Supplementary Fig. S4A), respectively. In cultures treated with the combination of KRIBB11 and AT-101 for 48 hours, the percentage of dead cells was on a similar level as in BAG3KD cells treated with AT-101 alone. Similar sensitizing effects of KRIBB11 were also observed in combination with ABT-737 (Fig. 4B; Supplementary Fig. S4A). A caspase-3–like activity assay was performed under the same conditions (Fig. 4C). After 24 and 48 hours of combined KRIBB11 and ABT-737 treatment, DEVD cleavage was significantly higher than single ABT-737 treatment in U251 control cells. U251 BAG3KD cells (with low Mcl-1 levels) already showed a higher level of caspase-3–like activity after single treatment with ABT-737, which was not further induced by additional treatment with KRIBB11. Mitochondrial membrane potentials were significantly reduced in U251 cells after combined treatment with KRIBB11 and AT-101 for 48 hours. Here, the mitochondrial dysfunction reached a similar level as in BAG3KD cells treated with AT-101 alone (Fig. 4D). In combined treatment of KRIBB11 with ABT-737 for 24 and 48 hours, a significant reduction of mitochondrial function was detected in U251 cells as well as in U343 cells (Fig. 4D; Supplementary Fig. S4B).

For direct targeting of the HSP70/BAG3 complex, the novel compound YM-1 was used. YM-1 was recently shown to selectively disrupt the HSP70–BAG3 interaction (26). Here, we confirm the disruption of the HSP70/BAG3 complex by YM-1 after 2 and 24 hours in glioma cells (Fig. 5A) and the consequent degradation of Mcl-1 after YM-1 treatment (Fig. 5B). Combined treatment of YM-1 with AT-101 and ABT-737 for 24 and 48 hours significantly increased cell death in U251 cells, as detected by Annexin-V staining followed by flow cytometry (Fig. 5C). Because of the autofluorescence properties of YM-1, PI staining unfortunately could not be performed. Similar results were obtained in U343 cells (Fig. 5D): cell death was significantly increased after treatment with AT-101 and ABT-737 in combination with YM-1, compared with single-agent treatment with AT-101 and ABT-737, respectively. The pan-caspase inhibitor z-VAD evoked a strong reduction of cell death after combined treatment with YM-1 and BH3 mimetics (Fig. 5C and D). To further confirm apoptosis induction, a caspase-3–like activity assay was performed (Fig. 5E). As previously observed for KRIBB11, a significant increase of DEVD cleavage was observed after combined treatment of YM-1 with AT-101 and ABT-737, demonstrating reactivation of BH3 mimetic–induced apoptosis by YM-1.

Our data obtained so far suggested that the stability of antiapoptotic Bcl-2 family proteins (Figs. 2E and 4A), prevention of Bax activation (Fig. 2D), and reduced matrix attachment (Fig. 3) are key mediators of BAG3-dependent glioma cell survival. In addition, we also investigated other possible mechanisms underlying the sensitization of glioma cells after interfering with the HSF1/HSP70/BAG3 pathway. Despite the hypothesis that BAG3/HSP70 may serve to retain Bax in the cytosol (21), we could not detect direct complex formation between HSP70/BAG3 and Bax in our IPs (Supplementary Fig. S5A). Survivin and the NF-κB pathway regulator IKKγ are other established BAG3 client proteins previously implicated in regulation of apoptosis (20, 26, 42). However, we observed no difference regarding the basal levels and stress-induced degradation of survivin in U251 BAG3KD cells compared with control cells (Supplementary Fig. S5B). Similarly, there were no detectable differences in the translocation of p65 from the cytosol to the nucleus, used as a marker for NF-κB activation (Supplementary Fig. S5C). Therefore, apoptosis induced by HSF1/HSP70/BAG3 pathway interference occurs independent of survivin and the NF-κB pathway in glioma cells.

To verify our hypothesis that BAG3 is a promising target for treating glioma, we orthotopically transplanted U343 control cells (ctrl) or BAG3-depleted U343 cells (Fig. 6A) into the right striatum of athymic nude mice. In mice transplanted with U343 ctrl cells, the first death occurred 35 days after transplantation (Fig. 6B; ref. 36). In contrast, BAG3 depletion delayed the occurrence of the first death event to 47 days after transplantation. Importantly, only 6 of 14 mice (42.85%) transplanted with U343 ctrl cells survived the observation period of 8 weeks, whereas 14 of 15 mice (93.33 %) survived in the group with U343 BAG3KD cells. Histologic analyses of 3 representative mice taken 40 days after transplantation revealed that transplanted U343 BAG3KD cells had formed only small tumors at this time point. In contrast, tumors in the U343 ctrl group were much larger and displayed a more invasive phenotype than those of the BAG3KD group (Fig. 6C and D).

The HSP70 co-chaperone BAG3 mediates a selective form of autophagy for the degradation of damaged client proteins known to accumulate under conditions of cellular stress, thereby consolidating cells by recruitment of the macroautophagy pathway (19, 20). Recently, BAG3 was identified as a critical mediator of inducible resistance in cancer cells that survived after concomitant blockage of the constitutive protein quality control pathways by mitigating proteotoxicity via selective autophagy (43). By modulating the stability of a variety of substrates, BAG3 is also involved in a number of other cellular processes including apoptosis, cell adhesion, cell motility, and proliferation with wide ranging implications for development, aging, and tumor progression (20). The key function of BAG3 in the regulation of apoptosis and autophagy furthermore suggest that BAG3 may play an important role in the crosstalk between these 2 pathways. The BAG3 gene is a transcriptional target of the stress-induced transcription factor HSF1 which was found to be upregulated in glioma (44). In addition to glioblastoma, overexpression of BAG3 has been observed in different types of tumors including pancreatic carcinoma, leukemia, small cell lung cancer, and thyroid carcinoma, compared with very low basal levels of BAG3 in non-malignant cells (20, 21, 45–47). The role of BAG3 in promoting cell survival may therefore enhance therapy resistance in glioblastoma and several other human neoplasms. One key mode of BAG3 action in this context is the sustained overexpression of prosurvival proteins, such as members of the Bcl-2–like subgroup of the Bcl-2 family, survivin, and IKKγ (20, 24, 26). These client proteins are stabilized via preventing their degradation in the proteasomal pathway, thereby supporting their antiapoptotic activity.

BH3 mimetics can induce apoptotic or nonapoptotic cell death, and in some tumors, this apparent dichotomy appears to be correlated with the expression levels of prosurvival Bcl-2 family members (14, 15). The role of autophagy in modulation of BH3 mimetic–induced cell death is currently controversial (12). Previous studies have shown a cell type- and context-dependent autophagy, which play either a death-promoting or cytoprotective role. Despite the fact that the BH3 mimetic and pan-Bcl-2 inhibitor AT-101 is competent to induce apoptosis in different cell models, we found that AT-101 induces a nonapoptotic, autophagy-dependent cell death in glioma cells (7, 16). The mechanisms underlying this block of apoptosis activation in glioma cells hitherto remained unidentified, but we hypothesized that in addition to the antiapoptotic Bcl-2 proteins, there may be other factors preventing Bax activation following AT-101 treatment. Indeed, in the study by Festa and colleagues, the authors proposed a model in which Bax is sequestered in the cytosol in complex with BAG3 and HSP70, thereby preventing activation of the mitochondrial pathway of apoptosis (21). However, we could not detect any interaction of HSP70/BAG3 with Bax in our co-IP experiments performed in U251 glioma cells.

To further address the role of BAG3 in cell death resistance and the potential shift to apoptotic cell death, we performed stable lentiviral BAG3 depletion in 2 glioma cell lines: effector caspase induction and apoptosis was efficiently restored in U251 glioma cells with high BAG3 expression, with less pronounced effects in U343 cells expressing lower endogenous levels of BAG3. BAG3 depletion led to a robust downregulation of Mcl-1, Bcl-2, and Bcl-xL but had no discernible effects on the levels of survivin and activation of the NF-κB pathway. These sensitizing effects of the BAG3 knockdown were associated with enhanced activation of Bax, cytochrome c release, and increased mitochondrial dysfunction, confirming activation of the intrinsic pathway of apoptosis in BAG3KD cells.

To further scrutinize these synergistic, proapoptotic effects of BH3 mimetics in combination with ablation of BAG3 function, we employed the selective HSF1 inhibitor KRIBB11 (25) and the specific HSP70/BAG3 small-molecule inhibitor YM-1 (26) that prevents formation and function of the HSP70-BAG3 module. Inhibition of HSF1 by KRIBB11 significantly decreased the protein levels of HSP70, BAG3, and Mcl-1, indicating that constitutive overexpression of BAG3 is driven by HSF1 in glioma cells. KRIBB11 was also able to mimic the sensitizing effect of BAG3 depletion after combined treatment with AT-101 and ABT-737. In analogy to these results, YM-1 induced the degradation of Mcl-1 and displayed synergistic proapoptotic effects with AT-101.

Anoikis is a form of apoptosis that is induced by detachment of adherent cells from the extracellular matrix (48). It was previously demonstrated that BAG3 silencing reduces cell motility and ECM adhesion of breast and prostate carcinoma cells (38). The ability of BAG3 to regulate cell adhesion was proposed to rely on multiple interactors of BAG3 through different structural domains in this context (20). Glioma cells have also been shown to be susceptible to anoikis in several paradigms, for example, inhibition of integrins with cilengitide (49). In our experiments, KD of BAG3 in U251 cells led to a pronounced reduction of ECM attachment as demonstrated on 5 different matrices (fibronectin, collagen I, collagen IV, laminin I, and fibrinogen). This reduction in ECM attachment was accompanied with a reduced cell size and less flattened and tightly attached cellular morphology. We hypothesize that this loss of cell–matrix interactions could prime glioma cells to anoikis, which is in line with the finding that matrix detachment was further enhanced after treatment with AT-101. Moreover, sensitivity to AT-101 was increased in suspension cultures, and BAG3-depleted cells U251 showed an almost complete block of FAK phosphorylation—a major regulator involved in antagonizing anoikis in cancer cells after matrix detachment (50). In line with these findings, BAG3 depletion limited the 3D plating efficiency, strongly decreased in vivo tumor growth in an orthotopic mouse model, and sensitized glioma cells to AT-101 and TMZ treatment in 3D clonogenic survival assays. Interestingly, anoikis can be inhibited by high expression levels of Bcl-2 (48), suggesting that enhanced expression of Bcl-2 may be directly involved in promoting BAG3-dependent anoikis resistance. In other pioneering work, it was also demonstrated that Bcl-2 enhances glioma invasion via modulation of MMP expression (51), providing an additional possible explanation for the reduced aggressiveness of BAG3-depleted tumors in vivo.

On the basis of the findings of our study, we propose that by stabilizing antiapoptotic Bcl-2 family members and promoting anoikis resistance, the HSF1/HSP70/BAG3 pathway may play a pivotal role for the cell death resistance of glioma. Pharmacologic intervention with BAG3 and HSP70 function is an interesting approach to reactivate apoptotic cell death in glioma and the HSF1/HSP70/BAG3 pathway is a potential target for future therapies.

J.P. Steinbach has received speakers bureau honoraria from Medac, Roche, and Boehringer; is a consultant/advisory board member for Roche, Mundipharma; and has provided travel grants for Roche and Medac. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M.C. Burger, F. Gessler, D. Kögel

Development of methodology: P. Antonietti, M.C. Burger, D. Kögel

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): P. Antonietti, B. Linder, S. Hehlgans, I.C. Mildenberger, J.P. Steinbach

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): P. Antonietti, B. Linder, S. Hehlgans, F. Gessler, M. Mittelbronn, D. Kögel

Writing, review, and/or revision of the manuscript: P. Antonietti, B. Linder, S. Hehlgans, I.C. Mildenberger, M.C. Burger, S. Fulda, J.P. Steinbach, F. Gessler, F. Rödel, M. Mittelbronn, D. Kögel

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.C. Burger, F. Gessler, M. Mittelbronn, D. Kögel

Study supervision: F. Gessler, D. Kögel

We thank Gabriele Köpf for excellent technical assistance and David G. McEwan for providing assistance with the Leica SP8 laser-scanning microscope.

This study was supported by the Deutsche Forschungsgemeinschaft (SFB 1177 on selective autophagy) to D. Kögel and S. Fulda, a grant of the Medical Faculty, Goethe University Frankfurt (Förderung von Nachwuchsforschern) to F. Gessler and by a grant of the BMBF to S. Fulda.

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.