Celastrol targets proteostasis and acts synergistically with a heat-shock protein 90 inhibitor to kill human glioblastoma cells

Abstract

Glioblastoma multiforme is a devastating disease of the central nervous system and, at present, no effective therapeutic interventions have been identified. Celastrol, a natural occurring triterpene, exhibits potent anti-tumor activity against gliomas in xenograft mouse models. In this study, we describe the cell death mechanism employed by celastrol and identify secondary targets for effective combination therapy against glioblastoma cell survival. In contrast to the previously proposed reactive oxygen species (ROS)-dependent mechanism, cell death in human glioblastoma cells is shown here to be mediated by alternate signal transduction pathways involving, but not fully dependent on, poly(ADP-ribose) polymerase-1 and caspase-3. Our studies indicate that celastrol promotes proteotoxic stress, supported by two feedback mechanisms: (i) impairment of protein quality control as revealed by accumulation of polyubiquitinated aggregates and the canonical autophagy substrate, p62, and (ii) the induction of heat-shock proteins, HSP72 and HSP90. The Michael adduct of celastrol and N-acetylcysteine, 6-N-acetylcysteinyldihydrocelastrol, had no effect on p62, nor on HSP72 expression, confirming a thiol-dependent mechanism. Restriction of protein folding stress with cycloheximide was protective, while combination with autophagy inhibitors did not sensitize cells to celastrol-mediated cytotoxicity. Collectively, these findings imply that celastrol targets proteostasis by disrupting sulfyhydryl homeostasis, independently of ROS, in human glioblastoma cells. This study further emphasizes that targeting proteotoxic stress responses by inhibiting HSP90 with 17-N-Allylamino-17-demethoxygeldanamycin sensitizes human glioblastoma to celastrol treatment, thereby serving as a novel synergism to overcome drug resistance.

Figure 1

Celastrol mediates a ROS and JNK-independent mechanism of cell death in human glioblastoma cells. (a) In a number of different cell types, it has been proposed that celastrol generates ROS and subsequently activates stress-dependent pathways (e.g., JNK) and inhibits pro-survival signaling (e.g., decreased AKT/PKB expression). (b) U251N cells were treated with celastrol at the indicated concentrations (0.3–10 μM) and labeled with Hoechst (nuclear stain). Scale bar in top left panel (200 μm) is representative for all panels shown. (c) U251N cells were treated with celastrol at the indicated concentrations (0.3–10 μM) in serum-containing and serum-deprived media for 24 h. Cell viability was assessed with the MTT assay. LD50: 4.07±0.29 μM (Serum−) versus 3.16±0.09 μM (Serum+); P=2.67 × 10⁻¹² (see Statistical analysis in Materials and Methods section). (d), U251N cells were treated with serum-deprived media with or without 50 μM L-BSO, an inhibitor of gamma-glutamylcysteine synthetase, for 3 h, after which celastrol was added at various concentrations (0.3–10 μM) for another 24 h. Cell viability was assessed as in panel (c). L-BSO treatment did not have a significant effect (LD50: 3.65±0.24 μM without L-BSO versus 3.25±0.27 μM with L-BSO; P=1.06). (e) Intracellular GSH content was assessed as described in the Materials and Methods section, following 24 h treatment with serum-containing or serum-deprived media in the presence and absence of 50 μM L-BSO and 3 μM celastrol. L-BSO led to a significant reduction in intracellular GSH relative to untreated cells. No significant change in GSH content was noted for celastrol alone. (f) U251N cells were treated with serum-deprived media with or without SP600125, an inhibitor of JNK, at the indicated concentrations for 3 h, after which celastrol was added at various concentrations (0.1–10 μM) for another 24 h. Cell viability was assessed as in panels (c) and (d). No significant effect was seen with SP600125 (LD50: 2.5±0.2 μM without SP versus 2.2±0.58 μM with 10 μM SP and 1.89±0.71 μM with 20 μM SP; P=0.0834). Average values and S.D. are reported for triplicate measurements (N=3). Results are representative of at least three independent experiments. ***P<0.001

Figure 2

Caspase-3 is dispensable for celastrol-mediated cell death in human glioblastoma cells. (a) U251N cells were incubated with either serum-containing (Serum+) or serum-deprived (Serum−) media or treated with 1 μM STS for 12 h or 3 μM celastrol for 24 h in serum-deprived media. Total cell lysates were collected and immunoblotted for PARP-1 cleavage, characteristic of caspase-3 activation. Actin is used as a control for protein loading. (b) U251N cells were treated with serum-deprived media with or without the caspase-3 selective inhibitor, Ac-DEVD-CMK, at the indicated concentrations for 3 h, after which celastrol was added at various concentrations (0.1–10 μM) for another 24 h. Cell viability was assessed by MTT assay. Caspase-3 inhibition led to a small but significant effect in shifting the dose–response to celastrol relative to treatment with celastrol alone (LD50: 1.12±0.19 μM with DEVD versus 1.45±0.28 μM with 20 μM DEVD and 1.35±0.14 μM with 50 μM DEVD; P=0.000113). Average values and S.Ds. are reported for triplicate measurements (N=3). Results are representative of at least three independent experiments

Figure 3

Celastrol-mediated glioblastoma cell death is dependent on thiol reactivity and is abrogated by co-treatment or preincubation of the drug with free thiols. (a) U251N cells were pretreated with either 2 mM NAC (NAC PreTreat) or serum-deprived media (Serum− and NAC Co-Treat) for 24 h. Media was replenished with celastrol at various concentrations (0.3–10 μM) with (NAC Co-Treat) or without (Serum−, NAC PreTreat) 2 mM NAC for a subsequent 24 h. Cell viability was assessed with the MTT assay. Significant differences in cell viability are noted when NAC is co-treated with celastrol relative to cells treated with celastrol alone (LD50: 12.6±161 μM; P=1.6E-23) but not when NAC is removed following pretreatment and before the addition of celastrol (LD50: 3.38±0.11 μM without NAC versus 3.55±0.19 μM with NAC pretreatment; P=0.0856). (b) U251N cells were pretreated with either 2 mM NAC or serum-deprived media for 24 h. Media was replenished with paraquat at various concentrations (0–1 mM) for a subsequent 24 h. Cell viability was assessed with the MTT assay. Significant rescue of cell viability was noted with NAC pretreatment compared with no pretreatment for the two highest doses of paraquat. (c) U251N cells were treated with 1 mM Trolox, 50 μM DTT or serum-deprived media in combination with celastrol at various concentrations (0.1–10 μM) for 24 h. Cell viability was assessed as in panels (b) and (c). No significant effect was seen with trolox (P=0.0102), whereas the presence of DTT significantly delayed the response to celastrol (LD50: 5.65±0.08 μM without DTT versus 12.6±14.1 μM with DTT; P=5.13 × 10⁻⁸) (see Statistical analysis section in Materials and Methods). Average values and S.Ds. are reported for triplicate measurements (N=3). Results are representative of at least three independent experiments. **P<0.01. (d) Chemical structure for celastrol–NAC adduct (6-N-Acetylcysteinyldihydrocelastrol) formed upon Michael addition of NAC to C6 electrophilic center of the celastrol quinone methide moeity

Figure 4

Celastrol blocks protein degradation and promotes accumulation of poly-ubiquinated substrates in human glioblastoma cells. (a) U251N cells were treated with serum-containing and serum-deprived media in the presence and absence of 500 nM 17-AAG (Hsp90 inhibitor), 10 μM MG-132 (proteasome inhibitor) and various concentrations of celastrol for 24 h. Total cell lysates were collected and immunoblotted for ubiquitin. (b and c) Cells were treated with serum-containing and serum-deprived media in the presence and absence of 10 μM MG-132 and 3 μM celastrol for 3–24 h. Total cell lysates were collected and immunoblotted for LC3B (a) and p62 (b). Actin is used as a control for protein loading. (d) In situ immunolabeling of p62 in U251N cells reveals accumulation of aggresomes, as indicated by white arrowheads. Cel, Celastrol 3 μM for 8 h; Rap, Rapamycin 200 nM for 24 h. Scale bar is representative for all four panels. (e) U251N cells were treated with increasing concentrations of celastrol for 24 h and analyzed for lysosomal content as described in the Materials and Methods section. Lysosome number was normalized to untreated cells (100%). Significant differences relative to untreated cells are denoted. (f) U251N cells labeled with LysoTracker Red DND-99 are depicted with cytoplasmic membrane outlines highlighted to clarify cell delineations. Scale bar in the first panel (50 μM) is representative for all the four panels shown. (g) U251N cells were treated with serum-deprived media with or without 50 μM CQ (inhibitor of lysosomal acidification) or 5 mM 3MA (inhibitor of the class III P13K, Vps34) for 1 h after which celastrol was added at various concentrations (0.5–10 μM) for another 24 h. Cell viability was assessed using the MTT assay. No significant differences were observed with CQ or 3MA relative to celastrol alone (LD50: 3.66±0.31 μM with serum deprivation versus 3.50±0.30 μM with CQ and 3.69±0.24 μM with 3MA; P=2.728). (h) U251N cells were treated with serum-deprived media with or without 1 μg/ml CHX, an inhibitor of protein translation, for 1 h, after which celastrol was added at various concentrations (0.3–10 μM) for another 24 h. Cell viability was assessed using the MTT assay. A significant difference in the response to celastrol was noted with CHX-treated cells relative to treatment with celastrol alone (LD50: 5.01±0.24 μM without CHX versus 14.8±88.3 μM with CHX; P=4.38 × 10⁻⁴) (see Statistical analysis section in Materials and Methods). Average values and S.Ds. are reported for triplicate measurements (N=3). Results are representative of at least three independent experiments. **P<0.01, ***P<0.001

Figure 5

Celastrol promotes proteotoxic stress in a panel of human glioblastoma cell lines and GSCs. (a) U343 and (b) U87 cells were treated with serum-containing and serum-deprived media in the presence and absence of 50 μM CQ (inhibitor of lysosomal acidification), 500 nM 17-AAG (Hsp90 inhibitor), and 3 μM celastrol for 24 h. Total cell lysates were collected and immunoblotted for proteasomal substrates (ubiquitin) and autophagy substrates (p62). Actin is used as a control for protein loading. (c) U343 and (d) U87 cells were treated with serum-deprived media with or without 1 μg/ml CHX, an inhibitor of protein translation, for 1 h, after which celastrol was added at various concentrations (0.3–10 μM) for another 24 h. Cell viability was assessed using the MTT assay. A significant difference in the response to celastrol was noted with CHX-treated U343 cells relative to treatment with celastrol alone (LD50: 5.30±0.28 μM without CHX versus 7.32±0.1 μM with CHX; P=8.36 × 10⁻⁶) (see Statistical analysis section in Materials and Methods). Average values and S.Ds. are reported for triplicate measurements (N=3). (e) GSCs were treated with serum-deprived media with or without 1 μg/ml CHX 1 h, after which celastrol (3 μM) was added for another 24 h. Cell viability was assessed using the MTT assay and reported as the percentage of control (untreated cells for celastrol−CHX and CHX-treated cells for celastrol+CHX). A significant difference in the response to celastrol was noted with CHX-treated GSCs as noted in the annotations. At right, the response to 1 μg/ml CHX alone is observed to reduce cell viability and proliferation by∼30–40%. Average values and S.Ds. are reported for triplicate measurements (N=3)

Figure 6

Celastrol upregulates the expression of heat-shock response genes, which delay cell death in human glioblastoma cells. (a) U251N cells were treated with serum-containing or serum-deprived media in the presence and absence of 1 μg/ml CHX, an inhibitor of protein translation, for 1 h, after which cells were treated with celastrol for 3–24 h. CQ (50 μM) for 24 h controls for protein accumulation following inhibition of autophagy. Total cell lysates were collected and immunoblotted for Hsp72. Actin is used as a control for protein loading. (b) U251N cells were treated with serum-containing and serum-deprived media in the presence and absence of 500 nM 17-AAG (Hsp90 inhibitor) or 10 μM MG-132 (proteasome inhibitor) for 24 h and 3 μM celastrol for 3–24 h. Total cell lysates were collected and immunoblotted for Hsp90. Actin is used as a control for protein loading. (c) U251N cells were treated with serum-deprived media with or without 500 nM 17-AAG for 1 h, after which celastrol was added at various concentrations (0.3–10 μM) for another 24 h. Cell viability was assessed using the MTT assay. 17AAG significantly shifted the dose–response to celastrol treatment relative to treatment with celastrol alone (LD50: 1.03±0.12 μM without 17-AAG versus 0.69±0.11 μM with 17-AAG; P=1.23 × 10⁻⁶). Average values and S.Ds. are reported for triplicate measurements (N=3). Results are representative of at least three independent experiments

Figure 7

Celastrol causes proteotoxicity via a thiol-sensitive mechanism in human glioblastoma. Poly-ubiquitinated protein aggregates along with the autophagy substrate and stress sensor, p62, accumulate as a result of defective protein degradation quality-control mechanisms, namely autophagy and proteasomal degradation. The disruption in proteostasis is further substantiated by the activation of cytoprotective heat-shock responses (HSP72/90)