Cisplatin abrogates the geldanamycin-induced heat shock response

Heat shock protein (Hsp) 90 is a molecular chaperone that contributes to cellular homeostasis by participating in several processes, including folding nascent proteins, stabilizing unfolded proteins to prevent aggregation, and facilitating intracellular trafficking. ATP hydrolysis is critical for Hsp90 chaperone functions, driving clamp-like reactions to facilitate protein folding (1). Benzoquinone ansamycins such as geldanamycin (GA) and its derivative 17-allylaminogeldanamycin (17-AAG) bind to the NH₂-terminal ATP-binding domain of Hsp90 and lock it into an ADP-bound conformation (2–6). At least two measurable cellular responses ensue. First, proteins that rely on Hsp90 are degraded by the proteasome after treatment with GA. These proteins, referred to as clients, are normally stabilized and folded into a functional state by the Hsp90 chaperone complex. However, GA binding disrupts the interaction of these proteins with the Hsp90 chaperone complex, preventing normal folding and leading to ubiquitylation and subsequent degradation (7–9). Second, GA treatment induces up-regulation of many proteins through the action of the transcription factor heat shock factor-1 (HSF-1; refs. 10–13). These newly synthesized proteins then work to restore cellular homeostasis after disruption of Hsp90 function.

17-AAG is currently in phase 2 clinical trials as a single agent, and phase 1 trials in combination with other cancer therapeutics. Despite promising preclinical studies showing significant antitumor activity in vitro and in vivo (14–16), 17-AAG has produced little clinical effect thus far as a single agent (8, 17). This lack of activity has prompted studies to determine possible resistance mechanisms to 17-AAG. Although tumor cell sensitivity to 17-AAG might be mediated, at least in part, by the presence of overexpressed client proteins such as Her-2 (18, 19), recent studies have indicated that the up-regulation of stress-responsive proteins, particularly Hsp70 and Hsp27, after Hsp90 inhibition, might also be responsible for the poor activity observed in 17-AAG clinical trials (10, 12, 20). Interest in these stress response proteins as a factor in resistance to Hsp90-directed therapy has been supported by data from several studies. Not only has Hsp70 been shown to inhibit changes in conformation and localization of Bax, thereby preventing apoptosis, but down-regulation of Hsp70 also sensitizes tumor cells to 17-AAG (21, 22). Moreover, Hsp27 up-regulation has been shown to contribute to 17-AAG resistance through a glutathione-mediated mechanism (23). Additionally, KNK437, a benzylidene lactam compound that suppresses the cellular heat shock response, has been shown to sensitize cells to Hsp90-directed agents (21).

Taken together, these studies indicate that circumventing HSF-1–mediated up-regulation of stress-responsive proteins provides an opportunity to increase the efficacy of Hsp90-directed therapy. After treatment with GA, HSF-1 is released from a Hsp90-containing heterocomplex that normally serves to repress HSF-1 transcriptional activity (24–26). HSF-1 then collects in the nucleus within large and brightly staining nuclear stress granules to eventually drive transcription of stress-responsive genes (27, 28). After trimerization and posttranslational modifications such as phosphorylation, HSF-1 binds to highly conserved promoter sequences called heat shock elements (HSE; refs. 29, 30) and stimulates transcription of stress-inducible proteins, including Hsp70 and Hsp27, up to 1,000-fold compared with unstressed conditions (31). Each of these activation steps for HSF-1 offers a prospect for pharmacologic intervention that could enhance efficacy of Hsp90-directed agents by limiting Hsp70 and Hsp27 up-regulation.

In this study, we show that combining cisplatin (CDDP) with GA or 17-AAG results in synergistic tumor cell killing. To define a mechanism for this synergy, we have investigated the contribution of HSF-1–mediated heat shock response up-regulation. Our data indicate that CDDP blocks GA-induced HSF-1–mediated transcription, resulting in decreased stress-responsive protein levels after treatment. The decreased transcription observed when GA is combined with CDDP is due to CDDP-mediated abrogation of HSF-1 chromatin binding, thereby preventing up-regulation of stress-responsive transcripts for genes such as Hsp70 and Hsp27. We have also identified melphalan (MEL) as another agent that enhances 17-AAG efficacy by blocking the up-regulation of stress-responsive proteins. Taken together, our data show that chemotherapeutics such as CDDP and MEL may be useful for preventing 17-AAG resistance by blocking the up-regulation of stress-responsive proteins such as Hsp70 and Hsp27.

Reagents were obtained from the following sources: GA and 17-AAG from Dr. V.L. Narayanan (Drug Synthesis and Chemistry Branch, National Cancer Institute, Bethesda, MD); CDDP, MEL, and camptothecin (CPT) from Sigma; and ECL enhanced chemiluminescent reagents from Amersham Pharmacia Biotechnology.

H9010 mouse monoclonal antibody recognizing Hsp90 was previously described (32). The remaining antibodies were purchased from the following suppliers: peroxidase-coupled affinity-purified goat anti-mouse and goat anti-rabbit secondary antibodies from Kirkegaard & Perry; mouse monoclonal anti-Hsp70, rabbit polyclonal HSF-1, and mouse monoclonal anti-Hsp27 from Stressgen; and mouse monoclonal anti-actin from Sigma.

A549 and HeLa cells were cultured in RPMI 1640 with 10% fetal bovine serum, 100 μg/mL streptomycin, and 100 units/mL penicillin. Hsp70^−/− and Hsp70^+/+ murine fibroblasts were cultured in DMEM-high glucose with 1 mmol/L sodium pyruvate, 10% fetal bovine serum, 100 μg/mL streptomycin, and 100 units/mL penicillin.

HeLa cells were transfected with small interfering RNA (siRNA) as previously described (23). Briefly, cells were plated in six-well plates at a density of 5 × 10⁵ per well and allowed to adhere for 20 to 24 h. Control siRNA #1 (400 nmol) or Hsp70-specific siRNA (23) was complexed with 10 μL of Lipofectamine 2000 (Invitrogen) in 0.5 mL of Opti-MEM (Invitrogen) for 10 min. Cells were incubated for 4 h with complexed lipid-siRNA, after which 1 mL of Opti-MEM containing 35% fetal bovine serum was added. The next day, cultures were washed once with serum-free medium, and fresh medium was added. Cells were trypsinized and replated for clonogenic assays or immunoblotting the next day, as described below.

A549 cells, or HeLa cells transfected with control or Hsp70-specific siRNA, were trypsinized and plated in 60-mm tissue culture plates to a density of 1,500 per plate. After cells were allowed to adhere for 22 to 24 h, drugs were added as indicated to final concentrations from 100-fold concentrated stocks. After a 24-h incubation, plates were washed twice with serum-free medium and then incubated in fresh medium until colonies were visible. The plates were washed once with PBS and stained with Coomassie brilliant blue. Visible colonies were counted, with typical plating efficiencies of 17% to 26%.

3-(4,5-Dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium salt (MTS) assays were carried out using the CellTiter 96 Non-Radioactive Cell Proliferation Assay (Promega) according to the manufacturer's instructions. Briefly, cells were plated in 96-well plates at a density of 500, 500, and 1,000 for A549, Hsp70^−/−, and Hsp70^+/+ fibroblasts, respectively. After cells were allowed to adhere for 24 h, drugs were added as indicated to final concentrations from 100-fold concentrated stocks. After a 24-h incubation, plates were washed twice with serum-containing medium, supplemented with fresh medium, and incubated for 5 additional days. Dye and stop solutions were added as directed by the supplier. Cell survival was estimated by absorbance, which was read at 570 nm.

Cells were plated on 100-mm dishes, allowed to adhere for 22 to 24 h, and then treated as described. Adherent cells were lifted from plates by scraping, combined with nonadherent cells, pelleted at 250 × g for 5 min at 4°C, rinsed once with ice-cold PBS, and then lysed in lysis buffer containing 10 mmol/L HEPES (pH 7.4), 20 mmol/L sodium molybdate, 150 mmol/L KCl, 10 mmol/L MgCl₂, 0.1% NP40, 1 mmol/L Na₃VO₄, and protease inhibitors (Complete, Mini, EDTA-free, tablets; Roche). After a 10-min incubation on ice, the detergent-insoluble fractions were pelleted at 18,000 × g for 2 min at 4°C. Total protein concentration of supernatants was estimated by the bicinchoninic acid method (33). Aliquots containing 50 μg of protein were separated by one-dimensional SDS-PAGE, transferred to nitrocellulose, probed with antibodies, and visualized by enhanced chemiluminescence as previously described (34).

Cells were treated as indicated and then harvested by scraping. Cells were washed once with ice-cold PBS, resuspended in Buffer A [10 mmol/L HEPES (pH 7.9), 10 mmol/L KCl, 0.1 mmol/L EDTA, 0.1 mmol/L EGTA, 100 μmol/L DTT, 500 μmol/L AEBSF], and incubated for 15 min on ice. NP40 was then added to a final concentration of 0.2% (v/v). After each sample was vortexed, and sedimented at 14,000 rpm for 1 min, the supernatant was collected for cytoplasmic fraction. Nuclei were resuspended in ice-cold Buffer B [20 mmol/L HEPES (pH 7.9), 25% glycerol, 0.4 mol/L NaCl, 1 mmol/L EDTA, 1 mmol/L EGTA, 100 μmol/L DTT, 500 μmol/L AEBSF], vortexed for 20 min at 4°C, and sedimented at 14,000 rpm for 5 min. The supernatant was collected for nuclear fraction. The total protein concentrations of cytoplasmic and nuclear fractions were estimated by the bicinchoninic acid method (33). Aliquots containing 50 μg of protein were subjected to electrophoresis using 10% Ready Gel precast gels and diluted Tris-glycine buffer from Bio-Rad. The separated proteins were transferred to nitrocellulose, probed with antibodies, and visualized by enhanced chemiluminescence as previously described (34).

Cells were treated for 24 h, washed, and incubated in drug-free medium as indicated. Cells were harvested by scraping. Adherent and nonadherent cells were combined and pelleted at 250 × g for 5 min at 4°C, rinsed once with ice-cold PBS, and fixed in 3:1 methanol/acetic acid overnight at room temperature. Fixed cells were applied to coverslips and then stained with 1 μg/mL Hoechst 33258 in 50 mmol/L Tris-HCl (pH 7.4 at 21°C) containing 50% (v/v) glycerol. Apoptosis was determined by examining slides by fluorescence microscopy and recording the number of cells that showed chromatin condensation or nuclear fragmentation. At least 500 cells were counted per slide.

Cells were stained with the rabbit polyclonal anti-HSF-1 antibody according to He et al. (35) with the following changes: A549 cells were plated on glass coverslips and allowed to adhere for 24 h. Cells were treated for 24 h with drugs as indicated, washed once with microtubule-stabilizing buffer [3 mmol/L EGTA, 50 mmol/L PIPES, 1 mmol/L MgSO₄, 25 mmol/L KCl], and fixed in the wells with 4% paraformaldehyde in PBS (v/v) for 30 min at room temperature. Cells were blocked for 1 h at 37°C in blocking buffer containing the following: 5% normal goat serum, 1% glycerol, 0.1% bovine serum albumin, 0.1% fish skin gelatin, and 0.04% sodium azide. Primary antibody was diluted in blocking buffer and then incubated with cells for 1 h at 37°C. After three washes in 1× PBS, cells were incubated with Alexa Fluor 568 goat anti-rabbit IgG (Invitrogen) for 1 h at 37°C, washed with PBS, and then fixed for an additional 5 min in 4% paraformaldehyde at room temperature. Images were captured using a 100× oil objective on a Zeiss LSM 510 confocal laser scanning microscope.

RNA was harvested from cells using the RNeasy kit from Qiagen. RNA (250 ng) was used for each condition, and then one-step reverse transcription-PCR (RT-PCR) was carried out using SuperScript One-Step RT-PCR with Platinum Taq from Invitrogen according to the manufacturer's instructions. Primers were as follows: Hsp90, 5′-GCCTCTGGTGATGAGATGGT-3′ (forward) and 5′-CATGGAGATGTCACCAATCG-3′ (reverse); Hsp70, 5′-CGACCTGAACAAGAGCATCA-3′ (forward) and 5′-AAGATCTGCGTCTGCTTGGT-3′ (reverse); Hsp27, 5′-GGACGAGCATGGCTACATCT-3′ (forward) and 5′-GACTGGGATGGTGATCTCGT-3′ (reverse); and p21, 5′-GACACCACTGGAGGGTGACT-3′ (forward) and 5′-CAGGTCCACATGGTCTTCCT-3′ (reverse). Conditions for RT-PCR were as follows: 1 cycle of 50°C for 30 min, then 25 cycles of 94°C for 15 s, 60°C for 30 s, and 72°C for 60 s, followed by 1 cycle of 72°C for 5 min.

Chromatin immunoprecipitation (ChIP) was carried out using the ChIP Assay kit from Millipore according to the manufacturer's instructions. Briefly, 15 to 20 million cells were cultured for each condition. After drug treatment, cells were washed and placed in fresh medium. DNA cross-links were formed by adding formaldehyde to medium to a final concentration of 1% (v/v) and incubating for 10 min at 37°C. Cells were then washed, harvested by scraping, lysed on ice for 10 min using the kit SDS lysis buffer, and sonicated with five pulses at maximum power to shear DNA, with 1 min cooling between each 8-s pulse. After lysates were cooled on ice for 3 min and sedimented at 14,000 rpm for 10 min, the supernatant was diluted to 1 mL with ChIP dilution buffer from the kit. Lysates were precleared at 4°C for 1 h using protein A-Sepharose beads (Sigma). Lysates were then rotated with HSF-1 polyclonal antibody (Stressgen) overnight at 4°C, supplemented with the beads from the kit, and rotated overnight again at 4°C. After beads were washed as directed, cross-links were reversed as indicated in the kit directions. DNA was precipitated using phenol-chloroform extraction followed by ethanol precipitation. PCR to amplify the HSE DNA bound to HSF-1 was carried out using PCR supermix (Invitrogen) with the following primers that result in an ∼183-bp band: forward, 5′-GAAGACTCTGGAGAGTTCTG-3′; reverse, 5′-CCCTGGGCTTTTATAAGTCG-3′.⁶

Statistical analysis consisted of the two-tailed paired t test. A P value of <0.05 indicated statistical significance. Synergy for clonogenic assays was determined by the median effect method. In brief, cells were treated with serial dilutions of each drug individually and with both drugs simultaneously or sequentially at a fixed ratio of doses. The fractional survival (f) was calculated by dividing the number of colonies in drug-treated plates by the number of colonies in control plates. Log [(1/f) − 1] was plotted against log [drug dose]. From the resulting graphs, the x intercept (log IC₅₀) and slope m were calculated for each drug and for the combination by the method of least squares and then used to calculate the doses of the individual drugs and the combination required to produce varying levels of cytotoxicity, and then the combination index (CI) was then calculated using the CalcuSyn program (Biosoft; ref. 36).

To examine the potential effect of combining GA and CDDP in vitro, we performed clonogenic assays using A549 cells. We compared the cytotoxicity of the GA + CDDP combination to the effect of the two agents alone using the median effect method (36), which determines whether the cytotoxicity for the combination is greater than (CI < 1), equal to (CI = 1), or less than (CI > 1) the additive effect of the individual agents. For these experiments, we used the ratio of 1:20: for GA:CDDP as determined by the ratio of the IC₅₀ for each agent in A549 cells.

Our data indicated that simultaneous exposure to GA or 17-AAG and CDDP resulted in synergistic antiproliferative effects, especially at doses near the IC₅₀ for 17-AAG + CDDP and IC₉₀ for GA + CDDP (CI = 0.549 ± 0.197 and 0.296 ± 0.134, respectively; Fig. 1A) but not at low concentrations. Subsequently, we examined A549 cells treated continuously with DMSO (control), 100 nmol/L GA, 30 μmol/L CDDP, or 100 nmol/L and 30 μmol/L of CDDP for apoptotic nuclear changes. After Hoechst 33258 staining, more nuclear fragmentation was evident in cells treated with the combination treatment compared with either GA or CDDP alone at 72, 96, and 120 h (Fig. 1B). From these data, we conclude that the synergy observed when GA is combined with CDDP is consistent with enhanced cell death due to combination treatment.

To examine the mechanistic basis for this synergistic interaction, we determined the effect of CDDP on the stress response, which is an important determinant of sensitivity to GA and 17-AAG. When lysates from cells collected for Hoechst staining were examined for stress-inducible proteins, the increased expression of Hsp70 and Hsp27 observed after GA treatment (Fig. 2A, lanes 2, 6, and 10) was much less prominent after CDDP (lanes 3, 7, and 11) or the combination treatment (lanes 4, 8, and 12). These data suggested that treatment with CDDP prevented up-regulation of stress-inducible proteins that occurs after GA treatment. To better define the dose dependence of the CDDP-mediated block of the heat shock response during GA treatment, lysates from cells treated for 24 h with DMSO (Fig. 2B, lanes 1 and 7), GA alone (lanes 2 and 8), increasing doses of CDDP (lanes 9–12), or GA and increasing CDDP (lanes 3–6) were blotted for stress-inducible proteins. As shown in Fig. 2B, CDDP blocked the induction of Hsp70 and Hsp27 by GA during simultaneous treatment even at doses as low as 3 μmol/L. Interestingly, after treatment with CDDP, we observed a mobility shift for HSF-1 that is often associated with phosphorylation (Fig. 2B). It is notable that the most extreme mobility shift for HSF-1 occurred at the highest dose of CDDP even without GA present, suggesting that phosphorylation might be responsible for the decreased HSF-1–mediated transcription after treatment with the combination (37).

To determine whether the decreased Hsp70 and Hsp27 protein levels observed after combination treatment reflected a decrease in the mRNAs, we did RT-PCR using total mRNA collected from cells treated with DMSO, GA, CDDP, or GA combined with increasing doses of CDDP. As shown in Fig. 2C, Hsp90β transcription remained steady with GA, CDDP, or combination treatment. Treatment with CDDP abrogated the transcription of Hsp70 and Hsp27 at 3, 10, and 30 μmol/L (lanes 6, 7, and 8, respectively). Conversely, treatment with CDDP increased transcription of p21, a cell cycle regulator that is transcriptionally regulated by p53 after DNA-damaging chemotherapy, including CDDP (38). These data show that CDDP causes decreased transcription of stress-inducible HSE-containing genes that are up-regulated during GA treatment, but not other genes such as Hsp90β and p21, suggesting that CDDP-induced down-regulation of HSE-containing genes is unlikely to be due to decreased global cellular transcription.

Because the combination treatment caused a decrease in stress-responsive proteins such as Hsp70 and Hsp27, two proteins previously implicated in 17-AAG resistance (21, 23), we assessed whether blocking the induction of Hsp70 would contribute to greater 17-AAG sensitivity. When HeLa cells transfected with Hsp70 siRNA were treated with 17-AAG for 24 h and examined using colony-forming assays, a 3-fold decrease in IC₅₀ was observed relative to control-transfected cells (IC₅₀ = 23.4 ± 11.7 nmol/L and 85.0 ± 32.0 nmol/L, respectively; Fig. 3A). Likewise, MTS assays showed that the IC₅₀ of mouse embryonic fibroblasts lacking Hsp70 (Hsp70^−/−) is at least 10-fold lower than in isogenic cells containing Hsp70 (Hsp70^+/+; IC₅₀ = 4.3 ± 2.9 nmol/L and 63.7 ± 14.9 nmol/L, respectively; Fig. 3B; ref. 19). Interestingly, MTS assays also revealed that the IC₅₀s of Hsp70^−/− cells treated with CDDP alone or GA + CDDP are not significantly different (IC₅₀ = 365 ± 130 nmol/L and 180 ± 144 nmol/L, respectively; P < 0.37; Fig. 3C), consistent with CDDP effect being mediated in part by abrogation of the heat shock response. Together, these data indicate that Hsp70 contributes to GA resistance, and imply that blocking Hsp70 induction through inhibition of HSF-1 action might increase GA sensitivity.

Previous studies have implicated HSF-1–regulated transcription in the heat shock response (10, 12, 30, 31). Many steps are required for HSF-1 activation, including phosphorylation, trimerization, and localization to the nucleus, where HSF-1 binds to DNA containing HSE promoters. Our data, combined with work by Bagatell et al. (39), suggest that CDDP blocks HSF-1–induced up-regulation of Hsps after GA treatment. However, the precise mechanism for this decrease in stress-inducible proteins has not been addressed to date. We first assessed whether HSF-1 trimerized after treatment. Nondenaturing gel electrophoresis followed by immunoblotting was used to examine the oligomeric state of HSF-1 in cells treated with DMSO, 100 nmol/L GA, 30 μmol/L CDDP, or GA and CDDP simultaneously. As indicated in Fig. 4A, HSF-1 was present in both monomeric and trimeric forms in the cytoplasmic portion of cells, with less monomer occurring in cells treated with either GA or GA + CDDP (Fig. 4A, left, lanes 2 and 4, respectively). However, CDDP alone did not cause a loss of the monomer (Fig. 4A, left, lane 3). Strikingly, an accumulation of the trimerized form of HSF-1 was found within the nuclear fractions in cells treated with GA or GA + CDDP (Fig. 4A, middle, lanes 2 and 4, respectively), which indicated activation and nuclear localization. This accumulation of trimerized HSF-1 in the nucleus explained the decrease of the monomeric form in the cytoplasmic fraction. To control for loading, the fractions were combined to show total protein for each treatment (Fig. 4A, right). These data show that HSF-1 was trimerized, and thereby potentially activated, in the presence of GA + CDDP despite the observed decrease in stress-responsive gene transcription versus GA alone.

Next, we examined whether HSF-1 forms nuclear stress granules after treatment with GA, CDDP, or GA + CDDP (27, 28). Immunofluorescent staining for HSF-1 revealed that no stress granules could be observed in cells treated with DMSO or CDDP (Fig. 4B, left). Conversely, treatment with GA or GA + CDDP resulted in stress granule formation, shown as punctate staining in the cell nucleus (Fig. 4B, right, white arrows). Taken together, our data indicate that CDDP is able to block GA-induced transcription of stress response genes, such as Hsp70 and Hsp27, but does not seem to block HSF-1 trimerization or localization to nuclear stress granules.

Based on the preceding results, we hypothesized that CDDP might affect HSF-1 binding to DNA. To assess whether HSF-1 can bind HSE sequences in cells treated with CDDP, we did ChIP. Immunoprecipitated HSF-1 from untreated cells or cells treated with DMSO or CDDP had low binding to the HSE (Fig. 4C, lanes 4 and 6, respectively). As expected, treatment with GA increased binding compared with DMSO alone (lane 5). Strikingly, HSF-1 chromatin binding is decreased in cells treated with GA + CDDP versus GA alone (Fig. 4C, lanes 7 and 5, respectively). These data are consistent with CDDP inhibiting HSF-1 binding to HSE-containing DNA even in the presence of GA.

In further experiments, we examined the effect of combining GA with the DNA cross-linking agent MEL or the topoisomerase I poison CPT. Both of these drugs can block transcription and cause cancer cell death (40, 41). MTS assays showed that GA increased the cell death observed with both MEL and CPT (Fig. 5A and B). For MEL, the IC₅₀ decreased from 103 ± 29.1 nmol/L in the absence of GA to 10.5 ± 12.2 nmol/L in the presence of GA. For CPT, the IC₅₀ decreased from 303.4 ± 90.2 nmol/L in the absence of GA to 98.7 ± 53.8 nmol/L in the presence of GA. Western blotting revealed that Hsp70 up-regulation is blocked in a dose-dependent manner when MEL is added to GA (Fig. 5C, lanes 5 and 6). Conversely, CPT did not block Hsp70 up-regulation as dramatically as MEL even at high doses (Fig. 5C, lanes 7 and 8). These data suggest that abrogation of stress response induction by MEL may cause added cell death when it is combined with GA. CPT, on the other hand, seems to increase cell killing when combined with GA through mechanisms other than abrogation of HSF-1–mediated transcription. This conclusion agrees with studies by Flatten et al. (42), which showed that 17-AAG and CPT derivative SN-38 were synergistic when combined, and concluded that this synergy resulted from down-regulation of Chk1.

Previous studies have yielded somewhat conflicting results when 17-AAG and CDDP were combined in various cell types. Vasilevskaya et al. initially reported additive effects when cisplatin was combined with 17-AAG, although the combination was even found to be antagonistic in some cell lines (8, 17). Variations between cell lines were reported to result from differential caspase activation, possibly attributable to differences in p53-mediated stimulation of apoptosis after treatment. In contrast, Bagatell et al. (39) reported synergistic effects in neuroblastoma and osteosarcoma cell lines when 17-AAG was combined with CDDP. Likewise, our data show that the combination of 17-AAG and CDDP is synergistic in vitro using non–small cell lung cancer cell line A549. Quantitation of apoptotic cells (Fig. 1B) indicates increased cell killing by the combination.

Previous studies have implicated Hsp70 (21, 22) and Hsp27 (23) in resistance to 17-AAG. Our experiments showed that CDDP, when combined with 17-AAG, could block HSF-1–induced Hsp70 up-regulation (Fig. 2), in agreement with the results of Bagatell et al. (39). Furthermore, we show that CDDP blocks the up-regulation of Hsp27, another HSE-regulated gene, while still causing up-regulation of p21, a gene that is increased in response to p53 activation after CDDP treatment (38). The survival data (Fig. 3) showing that Hsp70 knockdown and lack of effect on CDDP sensitivity in the Hsp70^−/− cells further support that CDDP is affecting the heat shock response. The relatively selective inhibition of GA-induced Hsp70 and Hsp27 induction provides a plausible explanation for the observed synergy between CDDP and GA or 17-AAG.

There are several possible mechanisms by which CDDP could block HSF-1 activity. First, CDDP could cross-link HSF-1 to the Hsp90 chaperone complex, which would sequester HSF-1 as an inactive monomer, thereby repressing its transcriptional activity. A previous study showed that treatment with GA + CDDP resulted in high-molecular weight cross-linked Hsp90 (39). It was unclear from that study whether HSF-1 is contained in these complexes. Our results indicate by two separate criteria that HSF-1 is activated in response to GA + CDDP treatment (Fig. 4). These observations suggest that HSF-1 has been released from Hsp90 as a result of GA treatment and that CDDP is interfering downstream of this activation. It is important to note that the synergistic interaction between GA and CDDP is observed only at higher concentrations, which may reflect a concentration-dependent effect of CDDP on the heat shock response. However, the effects of CDDP on HSF-1–mediated up-regulation of stress-responsive genes were studied using concentrations that caused synergistic cell killing.

In further experiments, we showed that CDDP interferes with HSF-1 binding to DNA. This could occur through adduct formation on HSF-1 or the HSE-containing DNA. First, CDDP binding could prevent HSF-1 trimerization or interfere with HSF-1 DNA-binding activity. This would thereby block the induction of HSE-containing genes. Second, CDDP has been shown to preferentially bind to guanine and adenine nucleotides (43). Because HSEs are known to contain GA-rich consensus sequences composed of inverted repeats of the pentamer 5′-nGAAn-3′ (29–31), CDDP may be stochastically more likely to bind to the HSE promoter, resulting in semiselective inhibition of stress-responsive protein up-regulation. By causing DNA adducts in this region, CDDP could act as a direct steric hindrance for HSF-1 binding to the promoter element, thereby causing decreased transcription of stress-responsive genes.

Another possibility is that CDDP is affecting upstream signaling events that modulate HSF-1 activity. HSF-1 phosphorylation is a key regulator of transcriptional activity. To date, phosphorylation at two sites, Thr¹⁴² and Ser²³⁰, has been shown to increase transcriptional activity of HSF-1 up to 3-fold (37). Our data suggest that CDDP treatment is inducing phosphorylation of HSF-1 (Fig. 2B), which could indicate activation. However, it is possible that CDDP is stimulating the phosphorylation of one or more of the other known sites on HSF-1, including Ser¹²¹, Ser³⁰³, Ser³⁰⁷, Ser³²⁶, and Ser³⁶³, which decrease HSF-1 activity (37). Interestingly, a study by Vasilevskaya et al. (44) showed that inhibition of c-Jun NH₂-terminal kinase kinase increased survival after treatment with GA + CDDP and that constitutively active c-Jun NH₂-terminal kinase signaling pathways were sufficient to increase cytotoxicity of the GA + CDDP combination. Because c-Jun NH₂-terminal kinase was previously shown to phosphorylate HSF-1 on Ser³⁶³, which inhibits transcriptional activity (37), it is possible that CDDP may be causing c-Jun NH₂-terminal kinase–induced HSF-1 phosphorylation, resulting in decreased transcriptional activation. This decrease in transcription could in turn prevent up-regulation of stress-responsive proteins such as Hsp70, which has been correlated with resistance to both 17-AAG and CDDP (10, 12, 20).

Our studies have raised another possibility for pharmacologic manipulation to enhance 17-AAG activity. Hsp70 and Hsp27 have been previously implicated in resistance to Hsp90-directed agents (21–23). Targeting these proteins directly could represent a new strategy for increasing 17-AAG efficacy. Alternatively, targeting the up-regulation of these proteins through inhibition of stress response–induced transcription could provide even greater sensitivity to Hsp90-directed therapy. Although agents that block the stress response, such as quercetin and KNK437, exist (12), neither is a specific inhibitor of HSF-1. Therefore, identifying new compounds that specifically inhibit HSF-1 activity while maintaining low toxicity may be the key to promoting 17-AAG as an effective therapy for cancer treatment. Until then, it may be possible to enhance the cytotoxicity of 17-AAG in the clinic with standard chemotherapeutics such as CDDP or MEL that can block, at least partially, the HSF-1–mediated heat shock response that occurs after treatment with Hsp90-targeted agents.

C. Erlichman was coinvestigator for a 17-AAG clinical trial funded by Conforma Therapeutics. The other authors have no potential conflicts of interest to disclose.

We thank Dr. Tej K. Pandita for providing Hsp70^−/− and Hsp70^+/+ cell lines, Drs. Scott Kaufmann and Larry Karnitz for scientific discussion and critical reading of the manuscript, and Dr. Martin Fernandez-Zapico for assistance with the ChIP assay.