Hsp90 Is Essential for Restoring Cellular Functions of Temperature-sensitive p53 Mutant Protein but Not for Stabilization and Activation of Wild-type p53

Several signaling pathways that monitor the dynamic state of the cell converge on the tumor suppressor p53. The ability of p53 to process these signals and exert a dynamic downstream response in the form of cell cycle arrest and/or apoptosis is crucial for preventing tumor development. This p53 function is abrogated by p53 gene mutations leading to alteration of protein conformation. Hsp90 has been implicated in regulating both wild-type and mutant p53 conformations, and Hsp90 antagonists are effective for the therapy of some human tumors. Using cell lines that contain human tumor-derived temperature-sensitive p53 mutants we show that Hsp90 is required for both stabilization and reactivation of mutated p53 at the permissive temperature. A temperature decrease to 32 °C causes conversion to a protein conformation that is capable of inducing expression of MDM2, leading to reduction of reactivated p53 levels by negative feedback. Mutant reactivation is enhanced by simultaneous treatment with agents that stabilize the reactivated protein and is blocked by geldanamycin, a specific inhibitor of Hsp90 activity, indicating that Hsp90 antagonist therapy and therapies that act to reactivate mutant p53 will be incompatible. In contrast, Hsp90 is not required for maintaining wild-type p53 or for stabilizing wild-type p53 after treatment with chemotherapeutic agents, indicating that Hsp90 therapy might synergize with conventional therapies in patients with wild-type p53. Our data demonstrate the importance of the precise characterization of the interaction between p53 mutants and stress proteins, which may shed valuable information for fighting cancer via the p53 tumor suppressor pathway.

The p53 tumor suppressor gene encodes a 393-amino acid transcription factor that is activated in response to cellular stress such as oncogenic activation (1), DNA damage (2), and hypoxia (3) leading to induction of downstream effects including cell cycle arrest, DNA repair, and apoptosis, thus preventing the proliferation of damaged cells. Wild-type p53 protein has four well characterized functional domains that are responsible for regulation of function: (i) the acidic, highly protease-sensitive N-terminal transactivation domain, which is responsible for interaction with the transcriptional machinery (4 -6); (ii) the central DNA binding core domain (7,8); (iii) a tetramerization domain (9,10); and (iv) a basic, C-terminal negative regulatory domain, the modification of which modulates sequence-specific DNA binding (11)(12)(13). The core region (residues 90 -290) consists of the sequence-specific DNA binding domain and contains four of five regions that are evolutionarily conserved among vertebrates (14). The p53 gene is the most frequently mutated gene in human cancer, occurring in about 55% of human tumors (15)(16)(17). The majority of mutations are single amino acid substitutions in the DNA binding core domain of p53 (18), producing a full-length protein that is incapable of binding DNA and is therefore nonfunctional as a transcriptional activator/repressor. The importance of the core region for p53 function as a tumor suppressor is underlined by the fact that 95% of more than 10,000 described mutations of the TP53 gene are clustered in the DNA binding domain. The mutations in the core region found in human tumors can be classified into two groups: (i) the contact mutants with changes in residues that are directly involved in contacting DNA, and (ii) the conformation mutants with changes in residues that are important for maintaining the structural stability of the core domain. Mutants can also be classified into three broad phenotypes based on their thermodynamic stability and DNA binding properties: (i) DNA contact mutations that have little effect on folding or stability of p53 protein (e.g. R273H); (ii) mutations that cause a local distortion, mainly in proximity to the DNA binding site (e.g. R249S); and (iii) mutations that cause global unfolding (e.g. mutation in the ␤-sandwich, e.g. V157F) (19). Many of these point mutations lead to loss of the wild-type protein conformation and can be detected by the loss of reactivity to the monoclonal antibody PAb1620 (20) or by the gain of reactivity to the monoclonal antibody PAb240 that recognizes mutant conformations of p53 protein (21).
The large frequency of mutants in cancer and their broad phenotypes open a promising strategy in cancer therapy that leads to rescue of p53 mutants and restoration of the tumor suppression activity. In particular, pharmacological reactivation of mutant forms of p53 protein which restore transcriptional activity toward proapoptotic genes has the potential to eliminate tumor cells through induction of cell death. Investigating the molecular mechanisms involved in the stabilization of wild-type and mutant p53 is therefore crucial for developing strategies to restore wild-type protein conformation as a basis for the development of new drugs targeting p53 in human cancer.
The level of p53 protein is mainly regulated at the posttranslational level by the MDM2 protein, whereas expression of MDM2 is activated by p53 at the transcription level, forming a negative feedback loop to maintain p53 protein at low levels under normal conditions (22,23). MDM2 binds to p53 and promotes its ubiquitination by acting as ubiquitin E3 ligase (24,25). It is well established that most stress signals that stabilize and activate p53 interfere with (i) the ability of MDM2 to promote p53 degradation, (ii) induction of the MDM2 inhibitor p14ARF, and (iii) inhibition of MDM2 expression. Phosphorylation of p53 on serine 15 by the ataxia-telangiectasia mutated kinase and serine 20 by human Chk2 kinase after DNA damage play critical roles in p53 stabilization by interfering with MDM2 binding. Stabilization of p53 also occurs in tumor cells with mutated p53, where missense mutations of p53 in the DNA binding core domain cause conformational change and stable association of this protein with molecular chaperones such as Hsp70 and Hsp90 (26,27). It was shown that Hsp90 contributes to the accumulation of mutant p53 and that its binding to p53 inhibits the ability of MDM2 to promote p53 ubiquitination and degradation, resulting in the stabilization of both mutant p53 and MDM2 (28 -30). Hsp90 appears to conceal the p14ARF binding site in the central domain of MDM2 in Hsp90⅐p53⅐MDM2 complexes and prevents MDM2 from ubiquitinating p53 and itself (31). In addition, Kamal et al. (32) reported that Hsp90 derived from tumor cells has a 100-fold higher activity than does Hsp90 from normal cells. Inhibition of Hsp90 has been shown recently to be of potential therapeutical significance in various types of human cancer (33,34). The relationship between p53 mutants and Hsp90 activity is supported further by the recent demonstration that Hsp90␤ gene repression is mediated by wildtype p53 (35), suggesting that the development of p53 mutation and increased Hsp90 activity can be coincident processes. Hence, modulation of Hsp90-assisted p53 folding has the potential for regulating mutant p53 activity in tumor cells. In this study we analyze the role of Hsp90 in mutant p53 conformational changes and stability of the wild-type p53 protein. We show that Hsp90 is required for both stabilization of mutated p53 and reactivation of human tumor-derived temperature-sensitive p53 mutants at the permissive temperature. Furthermore, we show that Hsp90 is necessary for transactivation activity of reactivated mutant p53 protein and its ability to induce downstream genes after a variety of genotoxic stresses, including cancer chemotherapeutic agents. These data highlight the importance of Hsp90 in regulating p53 function and the potential for manipulating the Hsp90-p53 interaction as an adjunct to cancer therapy.

MATERIALS AND METHODS
Cell Culture and Treatment-BT474 and MCF-7 are human breast cancer cell lines, and H1299 cells were derived from lung cancer. MCF-7 cells have a wild-type p53 gene, BT474 cells carry a temperaturesensitive mutation in the p53 gene (E285K) that adopts a wild-type conformation at 32°C but has a mutant conformation at 37°C. H1299 cells are p53-null. The BT474-RGC and H1299-RGC cell lines were established by stable transfection with a p53-responsive reporter construct (pRGC⌬fosLacZ) that allows quantitative measurement of p53 transcriptional activity by ␤-galactosidase assay. Cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 300 mg/liter L-glutamine. Cells were grown at two different temperatures (37 and 32°C) to 80% confluence before experimental treatments. Roscovitine, 6-benzylamino-2(R)-[[1-(hydroxymethyl)propyl]amino]-9-isopropylpurine, was added from a 50 mM stock solution in dimethyl sulfoxide to a final concentration of 20 M. Cycloheximide was added from a 20 mg/ml stock solution to a final concentration of 10 g/ml and geldanamycin from a 2 mM stock in dimethyl sulfoxide to a 2 or 4 M final concentration. Control cells received an equivalent volume of dimethyl sulfoxide. For UV-C irradiation, BT474 and BT474-RGC cells were washed with phosphate-buffered saline (PBS) 1 and then irradiated in the absence of medium using a model 1800 Stratalinker (Stratagene, La Jolla, CA). Cytostatic drugs were add to the cell culture medium for a 12-h incubation at the following final concentrations: 1 g/ml doxorubicin, 2 g/ml oxaliplatin, and 20 g/ml camptothecine. MCF-7 cells at 80% confluence were also stressed at 41°C for period of 12 h.
Transfection of BT474 and H1299 -The BT474-RGC and H1299-RGC cell lines were derived from BT474 and H1299 parental cells by stable transfection with the plasmid pRGC⌬fosLacZ coding for ␤-galactosidase controlled by the p53-responsive RGC promoter element (kindly provided by Prof. D. P. Lane). Integration of pRGC⌬fosLacZ was provided by cotransfection with pCMV-neo vector providing resistance to aminoglycoside antibiotics. Plasmids pRGC⌬fosLacZ and pCMV-neo were mixed in the ratio of 5:1 and transfected into BT474 and H1299 cells using Lipofectamine 200 reagent (Invitrogen). Stable transfectants were selected at 2 mg/ml G418 sulfate (Invitrogen). Transient transfection of H1299 cells with the plasmids pcDNA3-wtp53, pcDNA3-E285K p53, and pCMV-A138V p53 were performed using Lipofectamine 2000 reagent.
Antibodies-The antibodies used in this study are listed below.
• The anti-␤-galactosidase monoclonal antibody BG-2 was used as negative control. • The final antibody concentrations used for Western blotting were 1 g/ml, and immunoprecipitation used 1 g of antibody.
Analysis of p53-dependent Transcriptional Activity-⌻ measure p53dependent transcriptional activity, ␤-galactosidase activity was determined in the human breast cancer cell line BT474-RGC or H1299-RGC stably transfected with the p53-responsive reporter construct (pRGC⌬fosLacZ). Cells were washed with PBS, lysed in 0.25 M Tris-HCl, pH 7.5, by two freeze-thaw cycles and two 2-s ultrasound pulses at maximum intensity. The lysates were cleared by centrifugation at 20,000 g for 30 min at 4°C. Protein concentrations in supernatants were measured using the Bio-Rad protein assay (Bio-Rad Laboratories), and ␤-galactosidase activity was assessed using a colorimetric assay with ONPG as substrate and absorbance at 420 nm (41). All experiments were performed in duplicate on at least three separate occasions.
SDS-PAGE and Immunoblotting-Treated and control cells were collected by centrifugation (200 g, 10 min, 4°C), washed three times with ice-cold PBS, and lysed by 150 mM NaCl, 50 mM Tris-HCl, pH 8.0, 5 mM EDTA, 1% Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride buffer supplemented with protease inhibitor mixture (Sigma). 20 g of total protein lysate was mixed with 2ϫ Laemmli sample buffer. Proteins were separated by SDS-PAGE on 10% or 12.5% gels and transferred onto nitrocellulose membranes in a Bio-Rad Trans-Blot SD semidry transfer cell for 1 h applying 200 mA in transfer buffer (240 mM Tris, 190 mM glycine, and 20% methanol). Prestained molecular mass markers (Bio-Rad) were run in parallel. The blotted membranes were blocked in 5% milk and 0.1% Tween 20 in PBS for 2 h at room temperature and probed overnight with specific monoclonal antibodies or rabbit polyclonal serum. To confirm equal protein loading, immunodetection was performed with the anti-actin AC-40 monoclonal antibody. After washing three times in PBS plus 0.1% Tween 20, peroxidaseconjugated rabbit anti-mouse immunoglobulin antiserum or swine antirabbit immunoglobulin antiserum (DAKO) diluted 1:1,000 in 5% milk and 0.1% Tween 20 in PBS was used as the secondary antibody. To visualize peroxidase activity, ECL reagents from Amersham Biosciences were used according to the manufacturer's instructions.
Immunoprecipitation-Cells were lysed in 150 mM NaCl, 50 mM Tris-HCl, pH 8.0, 5 mM EDTA, 1% Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride buffer supplemented with protease inhibitor mixture (Sigma). Immunoprecipitation of conformationally different forms of p53 protein was performed using mouse monoclonal antibodies DO-1 (recognizing all p53 forms in native conditions), PAb240 (recognizing mutant conformation of p53 in native conditions), and PAb1620 (recognizing wild-type p53 form in native conditions), as described previously (21). For both preabsorbtion of the lysates and isolation of the antibody-p53 complex, protein G-Sepharose beads (Amersham Biosciences) were used. For preparation of the lysate for immunoprecipitation of Hsp90 in complex with MDM2 and/or p53 we used lysis buffer (10 mM Tris-HCl, pH 7.4, 1 mM MgCl 2 , 0.2% Tween 20 v/v, 10 mM sodium molybdate, 1.0 mM phenylmethylsulfonyl fluoride, and a protease inhibitor mixture (Sigma).
cDNA Macroarray Analysis of Gene Expression-Briefly, total RNA was isolated with the Qiagen RNeasy system. Following the instructions provided with the Human p53 Signaling Pathway GEArray Kit (Superarray, Inc., Bethesda, MD), the isolated RNA was then used to generate biotin-labeled cDNA probes by reverse transcription. After hybridizing the biotinylated cDNA samples to the gene-specific oligomers prespotted on the nylon membrane, the chemiluminescence signal was generated with the streptavidin-alkaline phosphatase system and detected using Amersham Biosciences x-ray film.
Densitometry-Protein and bound biotinylated cDNA levels were evaluated by densitometric analysis of photographic material. TotalLab software (Nonlinear Dynamics) was used to compute relative level of protein or biotinylated cDNA.

RESULTS
Temperature-sensitive Phenotype of p53 in BT474 Cells-Point mutations in residues that are important for maintaining the structural stability of the core domain of p53 disrupt its native conformation, leading to exposure of the epitope recognized by the mutant-specific antibody PAb240 and simultaneous loss of the epitope recognized by the wild-typespecific antibody PAb1620 (21). The cell line BT474 bears a temperature-sensitive mutant p53 (E285K) (42) that accumulates in the nucleus at 37°C, and its high level is unchanged after induction of cellular stress at this temperature. At 32°C, this p53 mutant is reactivated and can perform wildtype p53 functions. To verify the use of these cells as a model for the reactivation of wild-type p53 function, BT474 cells were cultured at 37 or 32°C with or without treatments that stabilize and activate wild-type p53. Proteins and mRNA were extracted for analysis of p53 levels and the expression of downstream target proteins and genes by Western blotting and cDNA macroarray analysis. Fig. 1 presents the results of Western blotting for p53 and the two downstream proteins, MDM2 and p21 WAF1 in BT474 cells compared with MCF-7 cells that contain a wild-type p53 gene. Cells were untreated or exposed to UV light or the Cdk inhibitor, roscovitine, both strong activators of wild-type p53 (Fig. 1A). In MCF-7 cells, basal levels of p53 are very low but are increased by UV or roscovitine, and this increase is associated with induction of p21 WAF1 and MDM2 proteins. In BT474 cells maintained at 37°C, the basal level of p53 is much higher, p53 levels are not increased after treatments, and there is no induction of p21 WAF1 or MDM2 proteins. After 20 h at 32°C, the level of p53 is decreased, and treatment with UV light or roscovitine at 32°C causes stabilization of p53 protein which is associated with induction of p21 WAF1 and MDM2 proteins. We also used the BT474-RGC cell line (see "Materials and Methods"), which is stably transfected with a reporter construct expressing ␤-galactosidase under the control of a p53-responsive promoter element to test the usefulness of this approach as a simple method to assess and quantitate the transcriptional activity of the temperature-sensitive p53 mutant (E285K) directly. BT474-RGC cells are identical to the parental BT474 cell line in terms of their growth characteristics and in their expression of all studied proteins at both temperatures. Thus, at 37°C (mutant p53 conformation), ␤-galactosidase activity was very low, whereas at 32°C (wild-type p53 conformation), ␤-galactosidase activity was elevated and was further increased more than 2-fold by treatment with UV light or 20 M roscovitine, confirming the use of ␤-galactosidase as a convenient marker of direct transcriptional activation by p53 in these cells (Fig. 1B). To investigate the generality of transcriptional activation of downstream genes, we used a targeted macroarray. Analysis of BT474 cells demonstrated a lack of induction of all p53 target genes by roscovitine at the nonpermissive temperature. Most p53-responsive genes are induced after culturing for 20 h at 32°C compared with cells grown at 37°C, compatible with p53 activation at the permissive temperature. Interestingly, although both p21 WAF1 and MDM2 were further induced by treatment with roscovitine at the permissive temperature, other genes that are proapoptotic were not further increased in these cells, including Apaf-1, Bax, PIDD, and NOXA (Fig. 2).
Post-translational Stabilization of MDM2 in the Presence of Mutant p53-The expression of MDM2 protein is activated by wild-type p53 but not by mutant p53 proteins that lack transactivational activity. Paradoxically, BT474 cells at 37°C have high steady-state MDM2 protein levels even though p53 is not active at this temperature, a phenomenon described previously by Peng et al. (30) in some cell lines carrying mutant p53 because of enhanced translation or stabilization of MDM2 protein. To address the mechanisms by which MDM2 protein A, cells were lysed after 20 h of initial treatment, and proteins were separated using electrophoresis and detected by Western blotting using antibodies DO-1 for detection of p53, 2A9 for detection of MDM2, 118 for detection of p21 WAF1 , and AC-40 for actin. B, p53-dependent transcriptional activity in BT474-RGC was measured using ␤-galactosidase activity. Cells were lysed after 20 h of initial treatment, and ␤-galactosidase activity was assessed using a colorimetric assay with ONPG as substrate and absorbance at 420 nm. All experiments were performed in duplicate on at least three separate occasions. levels are elevated in the presence of mutant p53 in BT474 cells, we measured mRNA levels on a commercial oligonucleotide array. In direct contrast to protein levels, MDM2 mRNA is low in cells maintained at 37°C and is increased 6-fold at 32°C. Treatment with roscovitine fails to induce MDM2 mRNA at 37°C but induces a 2-fold further increase in cells with wild-type p53 conformation at 32°C (Fig. 3A). These data demonstrate that MDM2 protein levels are regulated by post-translational mechanisms in BT474 cells and suggest that MDM2 protein is degraded rapidly in the presence of wild-type p53 but is stabilized in the presence of mutant p53. To investigate protein degradation as the mechanism for maintaining low levels of MDM2 in the presence of wild-type p53, we performed a more detailed analysis of MDM2 protein levels by Western blotting. The results confirmed the relative lack of full-length MDM2 protein in untreated cells at 32°C compared with cells maintained at 37°C. The analysis also revealed an increased proportion of shorter MDM2 protein forms in untreated BT474 cells maintained at 32°C which were virtually absent from cells at 37°C. Treatment with roscovitine reduced the amount of these shorter products and simultaneously increased the amount of full-length protein (Fig. 3B). The data are compatible with degradation of MDM2 in untreated cells with wildtype p53, and MDM2 stabilization as the main factor for elevated protein levels in cells expressing mutant p53.
Hsp90 Contributes to Stabilization of Soluble Temperaturesensitive Mutant p53 Protein-The p53 protein is conformationally flexible as well as labile, and the Hsp90 chaperone is known to bind both mutant and wild-type p53 in vivo to stabilize these conformations. Binding of molecular chaperones to mutant p53 and MDM2 inhibits the ability of MDM2 protein to promote p53 ubiquitination and degradation, potentially resulting also in stabilization of MDM2. We analyzed complex formation among mutant p53, Hsp90, and MDM2 using immunoprecipitation followed by Western blotting. Our data show the participation of Hsp90 in complex formation and stabilization of both mutant p53 and MDM2 at 37°C (Fig. 4). The p53-specific monoclonal antibody DO-1 identified MDM2 as well as Hsp90 in complex with p53, and the monoclonal antibody 2A9 specific for MDM2 protein immunoprecipitated both p53 and Hsp90. The antibody AC88 specifically recognizes Hsp90 that is not bound to substrates (43). As expected, this antibody did not pull down any p53 protein. The goat polyclonal antibody N-17 recognizes the N terminus of Hsp90 and immunoprecipitation of cell lysates with this antibody coprecipitated both p53 and MDM2. Taken together, these data clearly show that mutant p53, MDM2, and Hsp90 coexist in a trimeric complex in BT474 cells at 37°C. To investigate the functional consequences of Hsp90 for stabilization of p53 and MDM2 we used geldanamycin, a specific inhibitor of Hsp90, and analyzed the effect of this compound on complex formation and protein stability. Initially, we demonstrated that geldanamycin was effective in these cells. Fig. 4 shows that a 12-h incubation of BT474 cells with 4 M geldanamycin efficiently disrupted the complex of mutant p53 with Hsp90 after immunoprecipitation with the p53-specific antibody DO-1. The amount of p53 was unchanged in these cells, but the amount of p53 that coimmunoprecipitated with MDM2 was reduced, whereas the level of free Hsp90 detected with AC88 was increased.
Having determined the effectiveness of geldanamycin in disrupting the Hsp90⅐p53⅐MDM2 complex, we next performed a time course analysis to study the role of Hsp90 in regulating the stability of p53 and MDM2. MDM2 protein was decreased to nondetectable levels 4 h after treatment with 4 M geladanamycin. The level of mutant p53 protein was also lowest between 2 and 4 h, but there was a subsequent stabilization of mutant p53 with steady-state levels restored by 16 h (Fig. 5, A and B).
In the presence of cycloheximide as an inhibitor of protein synthesis, geldanamycin caused a steady decline in mutant p53 to about 10% of the initial level. These data indicate that Hsp90 directly stabilizes both preexisting and newly synthesized MDM2 protein, whereas Hsp90 is required to stabilize preexisting mutant p53 but is not required to stabilize newly synthesized mutant p53. These data suggest that in the presence of geldanamycin, MDM2 protein is able to drive degradation of mutant p53 protein that is not newly synthesized and has been allowed to interact with Hsp90 (Fig. 5, A  and B) (44,45). Therefore, Hsp90 acts to protect mutant conformations of p53 from MDM2-mediated degradation. The rate of degradation of preexisting mutant p53 diminishes because MDM2 is itself unstable in the absence of Hsp90. In addition, newly synthesized p53 can accumulate to restore steady-state levels in the absence of Hsp90. If Hsp90 indeed has a role in maintaining mutant p53 stability, as is clearly indicated in Fig.  5, such newly synthesized protein should not be correctly folded and may exist as protein aggregates as was published by other groups previously (45).

Reactivation of Temperature-sensitive Mutant p53 Is Associated with Conformation Changes of Core Domain and Is
Dependent on Hsp90 -To study the role of Hsp90 in maintaining wild-type p53 conformation we measured the effect of geldanamycin on the levels of mutant and wild-type p53 and transactivation of the p53 target genes p21 WAF1 and MDM2 (Fig. 6). Mutant and wild-type conformations of p53 were assessed by immunoprecipitation with PAb240 and PAb1620. Shifting the temperature of BT474 cells to 32°C induces a rapid and transient increase in the level of wild-type p53 conformation (PAb1620-positive). The overall levels of p53 are reduced, as are the levels of mutant p53 conformation (Fig. 6A). The increase in wild-type conformation is followed by an increase in p21 WAF1 and MDM2 levels, indicative of transcriptional activation. Treatment with roscovitine at 32°C induces a more prolonged induction of wild-type p53 protein and a more profound induction of target genes (Fig. 6B). In the presence of geldanamycin (Fig. 6C), switching BT474 cells to 32°C does not lead to accumulation of wild-type p53 conformation, and the mutant form is slightly decreased at 2-6 h before increasing to FIG. 2. Analysis of gene expression. A targeted macroarray was used to investigate the generality of transcriptional activation of downstream genes. BT474 cells grown with or without 20 M roscovitine at either 37 or 32°C were used for total RNA isolation with the Qiagen RNeasy system. Isolated RNA was then used to generate biotin-labeled cDNA probes, which were hybridized to the gene-specific oligomers prespotted on the nylon membrane (Human p53 Signaling Pathway GEArray Kit). The level of cDNA probes was evaluated by densitometric analysis of photographic material. TotalLab software was used to compute relative levels of expression.
steady-state levels, as seen previously for cells maintained at 37°C. Geldanamycin treatment also abolished the induction of p21 WAF1 and MDM2, and in fact MDM2 levels are decreased rapidly, similar to the effect of geldanamycin on cells maintained at 37°C. In addition, roscovitine was unable to induce the wild-type p53 conformation and transcriptional activity in the presence of geldanamycin (not shown). These data clearly show that the reactivation of the E285K temperature-sensitive p53 mutation in BT474 cells requires Hsp90 activity.
To investigate whether the activation of other p53 mutants is also dependent on Hsp90, whether other cell types show a similar dependence, and whether other p53-inducing agents also require Hsp90 for stabilization and activation of p53 we used the ␤-galactosidase reporter system in BT474 and H1299 cells. Fig. 7A shows the time course of reporter enzyme activity and dependence on geldanamycin in BT474-RGC cells shifted FIG. 3. Characterization of MDM2 expression. Identification and characterization of expression of MDM2 by macroarray or Western blotting are shown. A, BT474 cells grown with or without 20 M roscovitine (Rosc) at either 37 or 32°C were used for total RNA isolation with the Qiagen RNeasy system. Isolated RNA was then used to generate biotin-labeled cDNA probes that were hybridized to the gene-specific oligomers (left panel). ␤-Actin is shown as an internal control. The level of cDNA probes was evaluated by densitometric analysis of photographic material to compute relative levels of mRNA (right panel). B, BT474 cells were lysed, and proteins were separated using electrophoresis and detected by Western blotting using antibody 2A9 for detection of MDM2 protein (left panel). Densitometric analysis of the full-length and degraded MDM2 proteins is shown in the right panel. to the permissive temperature of 32°C. ␤-Galactosidase activity induced within 12 h at 32°C is increased further by roscovitine, peaking at about 24 h in both cases. Induction requires Hsp90 activity because levels are unchanged in the presence of geldanamycin (Fig. 7A). To show that these effects are seen with other p53 mutants, we used H1299-RGC cells that are p53-null and contain the RGC-LacZ reporter construct stably integrated into the genome (Fig. 7B). Transient transfection of these cells with the temperaturesensitive mutant A138V or E285K leads to minimal basal ␤-galactosidase activity at the nonpermissive temperature and a more than 10-fold induction of activity after switching to 32°C for 24 h. This induction requires Hsp90 in both cases. Interestingly, transfection of these cells with a wild-type p53 plasmid leads to induced enzyme activity that is insensitive to geldanamycin. To investigate whether the induction of reactivated mutant p53 at the permissive temperature is always dependent on Hsp90, we treated BT474-RGC cells with a variety of DNA-damaging and chemotherapeutic drugs. In all cases, enzyme activity was induced by more than 2-fold compared with untreated cells at the permissive temperature. Geldanamycin abolished this induction and reduced the levels to those of control geldanamycin-treated cells (Fig. 7C). Identical results were seen in cells grown at 30°C, which further activates p53 in control cells (Fig. 7D).
These data demonstrated that Hsp90 activity is required for producing conformationally active p53 during the reactivation of mutant p53 molecules. To study the role of Hsp90 in regulating stability and conformation of wild-type p53, we measured p53 conformational changes in MCF-7 cells in the presence or absence of roscovitine and geldanamycin with or without a heat shock (Fig. 8). These experiments revealed that at the physiological temperature of 37°C, roscovitine treatment causes high induction of both total p53 and of conformationally wild-type p53, seen by immunoprecipitation with DO-1 and PAb1620, respectively. The increase in wildtype p53 protein after roscovitine is accompanied by induction of p21 WAF1 protein. There is also a slight increase in conformationally inactive protein, detected with PAb240, although this form remains a minor component of total p53 (Fig. 8B). Geldanamycin treatment did not influence the total level of p53 at 37°C, nor did it inhibit the ability of roscovitine to induce wild-type p53 protein and p21 WAF1 . In contrast, at 41°C, the induction of wild-type p53 conformation and induction of p21 WAF1 are abolished by geldanamycin, and there is an accumulation of mutant conformation p53 detected with PAb240. Control MCF-7 cells, cells treated with 20 M roscovitine and/or 2 M geldanamycin, were grown either at 37 or 41°C and lysed for immunoprecipitation (IP). A, cell lysates were immunoprecipitated using pan-specific anti-p53 antibody DO-1, wild-type specific anti-p53 PAb1620, and mutant specific anti-p53 PAb240 antibodies. Immunoprecipitated p53 isoforms were detected by immunoblotting using anti-p53 polyclonal antibody CM-1. Detection of p21 WAF1 protein was performed in cell lysates by Western blotting (WB) with monoclonal antibody 118. B, density ratios between wild-type p53 conformation immunoprecipitated with PAb1620 and mutant p53 conformation immunoprecipitated with PAb240 were plotted.

DISCUSSION
Mutations in the p53 gene are the most common abnormality in human tumors. The vast majority of mutations arise within the central core domain of p53 and affect protein conformation, DNA binding, and transcriptional activity. Therefore, a major goal of p53 research is to identify factors that influence the conformation and stability of both mutant and wild-type p53 molecules. In particular, reactivation of mutant p53 molecules to restore wild-type function represents a potentially useful therapeutic strategy. A variety of studies have suggested roles for Hsp23, Hsp40, Hsp70, and Hsp90 in regulating p53 conformation. Geldanamycin and its analogs 17-AAG and 17-DMAG are specific inhibitors of Hsp90 which can inhibit cancer growth and promote apoptosis (32). These agents could be used to treat cancers bearing mutant p53 by impairing solubility, nuclear transport, and abrogating mutant gain of function. Here, we show roles for Hsp90 in regulating the stability of temperaturesensitive p53 mutants and in the conformational change of p53 during reactivation at the permissive temperature. MDM2 stability is also critically dependent on Hsp90, although the stability of endogenous wild-type p53 is not affected by Hsp90.
We initially demonstrated the temperature-sensitive character of p53 (E285K) in BT474 cells, where cultivation at 32°C causes mutant p53 to adopt a wild-type conformation and activates transcriptional activity of mutant p53. Reactivation is incomplete in this system, and there is a persistence of denatured p53 mutant protein, which has been demonstrated with other temperature-sensitive p53 mutants (46). Immunoprecipitation demonstrated the presence of complexes between p53 (E285K), MDM2, and Hsp90. The generation of complexes between p53 and chaperones including Hsp90, Hsp40, and Hsp23 has been demonstrated with other p53 mutants, indicating that they occur commonly in human cancers (29,44). Unlike mutant p53, wild-type p53 is a labile protein caused by rapid proteolysis by the ubiquitin-proteasome pathway, with MDM2 acting as the E3-ubiquitin ligase (47), and only low concentrations of p53 usually reside in the cell. MDM2 also mediates its own ubiquitination and degradation. BT474 cells contain high levels of MDM2 at the nonpermissive temperature, and in the absence of Hsp90 activity MDM2 is degraded rapidly, pointing to a critical role for Hsp90 in stabilizing MDM2. The data suggest that MDM2 is not active as an E3-ubiquitin ligase when complexed with Hsp90, which thereby inhibits both selfubiquitination and p53-ubiquitination. In keeping with this hypothesis, mutant p53 protein is initially unstable in the absence of Hsp90 activity because of high levels of free MDM2. Because MDM2 is itself unstable under these conditions, p53 levels subsequently increase as the levels of MDM2 decrease (Fig. 5). Blocking protein synthesis in the presence of geldanamycin indicated that Hsp90 specifically protects preexisting mutant p53 from degradation and is not required to allow stabilization of newly synthesized mutant p53. In addition, mutant p53 is stable even at the permissive temperature of 32°C, when p53 reactivation has occurred (Fig. 6A), suggesting that mutant p53 forms a stable complex with Hsp90 which is difficult to dissociate even in the presence of wild-type p53 protein. Taken together, the data indicate dual roles for Hsp90 in protecting mutant forms of p53 from MDM2-mediated degradation, allowing the accumulation of mutant p53 to the high levels seen in tumor cell lines and primary tumors that carry a mutation in the p53 gene (48). A schematic representation of these findings is depicted in Fig. 9.
Reactivation of temperature-sensitive p53 is dependent on Hsp90-mediated folding of newly synthesized p53 because no native protein conformation is seen, and there is no induction of p53 target genes at the permissive temperature in the presence of geldanamycin (Fig. 6C). In addition, stabilization of the native conformation by therapeutic agents is also critically FIG. 9. A schematic representation of dual roles for Hsp90 in protecting mutant forms of p53 from MDM2-mediated degradation. A, temperature-sensitive mutant p53, MDM2, and Hsp90 complex formation at 37°C. B, reactivation of nascent p53 at 32°C is dependent on Hsp90. C, inhibition of Hsp90 with geldanamycin blocks reactivation of mutant p53. This inhibition leads to disintegration of a complex of p53, Hsp90, and MDM2 proteins and is responsible for blockade of conversion of mutant conformation. dependent on Hsp90 activity, implying that therapies that aim to reactivate mutant p53 will be incompatible with Hsp90 antagonist therapy. Even in the presence of Hsp90, reactivation of mutant p53 as a treatment for human cancers will be limited by the transience of p53 activity, so that reactivated mutant p53 should also be stabilized for an effective outcome. However, not all downstream genes are induced further after reactivation and stabilization. In particular, stabilization of reactivated p53 is highly efficient in inducing p21 WAF1 and MDM2 mRNAs and in transactivating the RGC-LacZ promoter but does not further elevate proapoptotic gene expression. This apparent selectivity of induction has been described for other temperature-sensitive p53 mutants (49) and may be the result of persistence of denatured p53 after reactivation exerting sequence-specific dominant negative effects (50). It is also possible that genetic background or cell type-specific effects are responsible (51-53), although identical results were obtained using two different mutants in two different cell lines derived from different anatomical sites. Further studies are required to determine the prevalence and mechanism for promoter selectivity because this would severely limit the effectiveness of reactivation treatment.
We also studied the role of Hsp90 in regulating the activity of wild-type p53 both with and without stabilization. Recent reports using purified proteins have demonstrated that Hsp90 is capable of folding wild-type p53 (54,55). In contrast, because geldanamycin does not reduce the amount of wild-type p53 in MCF-7 cells, our results demonstrate that Hsp90 is not required for wild-type p53 folding at physiological temperatures. Similar findings were obtained with HCT116 colon cancer cells that contain wild-type p53 (56) and human fibroblasts in vivo (55). Nevertheless, Hsp90 clearly has an important role in p53 folding during heat shock at 41°C. Most importantly, we showed that Hsp90 is not required to stabilize and activate wild-type p53 after the administration of chemotherapeutic drugs at physiological temperature.
In conclusion, the novel cell lines BT474-RGC and H1299-RGC provide a simple and rapid approach for screening the effects of agents that are designed to target p53 in human cancer. We have used these systems to demonstrate the critical role played by Hsp90 in regulating the reactivation of temperature-sensitive p53 mutants. In contrast, Hsp90 does not play a major role in maintaining basal levels of latent wild-type p53 or in the stabilization and activation of wild-type protein after treatment with DNA-damaging agents or therapeutic drugs in vivo. Although inhibition of Hsp90 induces the degradation of mutant p53 by the dissociation of p53 and MDM2 from Hsp90, lack of Hsp90 activity also abolishes the reactivation of temperature-sensitive p53 mutants. Thus, the observed efficacy of antagonist Hsp90 treatment in cancer may be due partly to increased degradation of mutant p53, removing its potential to act in a dominant manner in tumor cells. However, treatment of tumors with Hsp90 antagonist therapeutics abrogates simultaneous attempts to stabilize reactivated p53 mutants and restore wild-type tumor suppressor function. Therefore, inhibition of Hsp90 and reactivation of p53 as targets for cancer therapies may be mutually exclusive, at least for tumors containing the class of mutants studied herein. On the other hand, because Hsp90 is not required to maintain or activate wild-type p53 protein, Hsp90 antagonists should have no effect on the beneficial actions of p53-stabilizing agents in tumors that contain wild-type p53. Moreover, because Hsp90 can disrupt a variety of other oncogenic alterations outside the p53 pathway, such as ErbB2, Raf-1, Akt, and Bcr-Abl (57-61), our data indicate that conventional chemo-or radiotherapies will synergize with Hsp90 inhibition therapeutics in tumors with wild-type p53. Indeed, our findings provide a molecular mechanism to account for the antagonism seen between geldanamycin/ cisplatin combination treatment in colon cancer cells containing mutant p53 compared with the additive effect of these two agents in cells with wild-type p53 (56). Therapeutic compounds that enable both chaperone-mediated refolding of p53 and release of native p53 from the chaperone machinery should prove highly effective.