Hsp90 chaperones wild-type p53 tumor suppressor protein.

Immortalized human fibroblasts were used to investigate the putative interactions of the Hsp90 molecular chaperone with the wild-type p53 tumor suppressor protein. We show that geldanamycin or radicicol, specific inhibitors of Hsp90, diminish specific wild-type p53 binding to the p21 promoter sequence. Consequently, these inhibitors decrease p21 mRNA levels, which lead to a reduction in cellular p21/Waf1 protein, known to induce cell cycle arrest. In control experiments, we show that neither geldanamycin nor radicicol affect p53 mRNA levels. A minor decrease in p53 protein level following the treatment of human fibroblasts with the inhibitors suggests the potential involvement of Hsp90 in the stabilization of wild-type p53. To support our in vivo findings, we used a reconstituted system with highly purified recombinant proteins to examine the effects of Hsp90 on wild-type p53 binding to the p21 promoter sequence. The human recombinant Hsp90 alpha-isoform as well as bovine brain Hsp90 were purified to homogeneity. Both of these molecular chaperones displayed ATPase activity and the ability to refold heat-inactivated luciferase in a geldanamycin- and radicicol-sensitive manner, suggesting that post-translational modifications are not involved in the modulation of Hsp90alpha activity. We show that the incubation of recombinant p53 at 37 degrees C decreases the level of its wild-type conformation and strongly inhibits the in vitro binding of p53 to the p21 promoter sequence. Interestingly, Hsp90 in an ATP-dependent manner can positively modulate p53 DNA binding after incubation at physiological temperature of 37 degrees C. Other recombinant human chaperones from Hsp70 and Hsp40 families were not able to efficiently substitute Hsp90 in this reaction. Consistent with our in vivo results, geldanamycin can suppress Hsp90 ability to regulate in vitro p53 DNA binding to the promoter sequence. In summary, the results presented in this article state that chaperone activity of Hsp90 is important for the transcriptional activity of genotypically wild-type p53.

The p53 tumor suppressor protein is a transcription factor, which regulates cellular response to stress, abnormal cell proliferation, and DNA damage (1,2). More than 50% of human cancers possess the mutated p53 gene (3), and the inactivation of p53 function leads to cell transformation (4,5). Most oncogenic p53 mutations are located in its DNA binding domain, inhibiting ability of this protein to initiate transcription (3). However, p53 may also be rendered inactive by other cellular events, such as the sequestration of wt-p53 1 in the cytoplasm, because of increased nuclear export (6) or the association of p53 to cytoplasmic proteins (7,8). Recently it was shown that p53 is transported in HEK cells to the nucleus along microtubular tracks by cytoplasmic dynein and the Hsp90 molecular chaperone (9), which is similar to the mechanism described for rapid ligand-induced movement of the glucocorticoid receptor to the nucleus (10,11). Processes that inhibit p53 nuclear transport and proteolysis also lead to the inactivation of p53 protein (12,13).
In the response to various stresses, such as ionizing radiation, UV and hypoxia, p53 is activated, stabilized, and imported into the nucleus, where it promotes transcription of several genes whose products induce cell cycle arrest, DNA repair, or apoptosis (14). In a non-stress situation, the level of p53 in the cells is mainly regulated at the post-translational level by MDM2 (12,(15)(16)(17).
Hsp90 is an abundant molecular chaperone important for protecting cells from stress, such as high temperature. Additionally, Hsp90 regulates many signaling pathways. Hsp90 is found in a complex with several oncoproteins, including v-Src, c-Erb2, Raf-1, Akt, Bcr-Abl, and tumor suppressor p53 (18 -22). The Hsp90 inhibitor 17-allylaminogeldanamycin (17-AAG) is currently in phase II of clinical trials as a potential anti-tumor drug. Hsp90 inhibitors usually induce ubiquitination and degradation of Hsp90 client proteins (23). In tumor cells, Hsp90 exists in a functionally distinct conformational form that is much more efficiently recognized by 17-AAG. This form of Hsp90, which possesses elevated ATPase activity, is found in a multichaperone complex with cochaperones: p23, Hop, and probably others (24).
It has been known for years that genotypically mutant p53 co-immunoprecipitates with members of the Hsp70 and Hsp90 families (19). Such interactions lead to the formation of a p53 multichaperone complex that is responsible for the stabilization and sequestration of p53 in the cytoplasm (25)(26)(27). Binding of molecular chaperones to mutant p53 inhibits the ability of MDM2 to promote p53 ubiquitination and degradation, resulting in the stabilization of both p53 and MDM2 (28,29). It also has been shown that Hsp90 directly associates with the MDM2 protein (30). Hsp90 inhibitors can partially disrupt these interactions, which results in the degradation of mutant p53 (31,32). With the use of highly purified proteins, we have identified intermediate reactions that lead to the assembly of molecular chaperone complex with p53 protein possessing wild-type or mutant sequence. The presence of Hsp90 in a complex with wt-p53 inhibits binding of Hsp40 and Hsc70 to p53. However, the conformational mutant of p53, which possesses low affinity toward Hsp90, can form a stable multichaperone complex in which Hsp90 is bound to mutant p53 indirectly (mut p53-Hsp40-Hsc70-Hop-Hsp90). Several independent methods, such as surface plasmon resonance, immunoprecipitation, ELISA, and cross-linking were used to demonstrate that Hsp90 directly, in the absence of any other co-chaperones, can associate with genotypically wt-p53 but not with mutant p53 protein (33). The accompanying article by Muller et al. (34) supports our findings. Moreover, it has been shown by NMR that Hsp90 associates with a truncated version of wt-p53. It was suggested that the p53 core domain bound to Hsp90 is predominantly unfolded and lacking helical or sheet secondary structure (35).
Wild-type p53 is a structurally unstable protein, which undergoes conformational changes at elevated temperatures (36,37). We have proposed that during heat shock, cytoplasmic p53 possessing the wild-type sequence could temporarily adopt a mutant conformation, subsequently initiating the formation of a multichaperone complex that could partially stabilize wt-p53 (19). Results from a recently published article by Wang and Chen (38) support our hypothesis. They found that heat shock inhibited p53 ubiquitination and initiated the accumulation of p53 at the post-translational level. Two factors influence these events during heat stress: 1) ATM-dependent phosphorylation of p53 and 2) formation of the chaperone complex with genotypically wt-p53, which adopts a conformation characteristic to that of a mutant protein (38). The evidence for Hsp90 binding to mutant p53 is conclusive, whereas the exact nature of cellular interactions between Hsp90 and genotypically wt-p53 possessing either wild-type or mutant conformation still remains to be elucidated.
In this study, we demonstrate that the chaperone activity of Hsp90 is required for wt-p53-dependent transcriptional activity. Specific Hsp90 inhibitors, geldanamycin and radicicol, inhibit p53 activity as the transcription factor by dissociation of p53 from its target DNA promoter sequence sites. Results from the reconstituted in vitro system clearly show that Hsp90 positively regulates p53 DNA binding to a specific promoter sequence after incubation at physiological temperature of 37°C. Moreover, this Hsp90 activity is ATP-dependent and inhibited by geldanamycin.

MATERIALS AND METHODS
Cell Culture Experiments-K15, a line of human fibroblasts immortalized by stable expression of hTERT, was a kind gift from Prof. H.
FIG. 1. Hsp90 inhibitors decrease p53 activity as the transcription factor. K15 cells were cultured in the presence or absence of 2 M camptothecin (CPT), 3 M geldanamycin (GA), or 3 M radicicol (R) for the indicated time periods. A, time course of total p53 (p53), p53 phosphorylated at Ser-15 (Ser15-P-p53) and p21/Waf1 protein (p21) induction was analyzed by Western blotting. Tubulin was detected as the control of total protein level. B, cellular p53 mRNA levels 3.5 h after the addition of camptothecin were quantified by real-time RT-PCR. The results were normalized to the glyceraldehyde-3-phosphate dehydrogenase mRNA. C, p53 binding to the p21 promoter sequence 4 h after camptothecin treatment was detected by ChIP and real-time PCR. D, cellular p21/Waf1 mRNA levels 3.5 h after the addition of camptothecin was quantified by real-time RT-PCR. All results are representative of at least three independent experiments.
Kampinga. The cells were cultured at 37°C in a humidified atmosphere containing 5% CO 2 , in Ham's F10 medium with 15% fetal bovine serum. For inhibition of topoisomerase I and activation of p53, 2 M camptothecin (Sigma) were used. For inhibition of Hsp90, 3 M geldanamycin (Sigma) or 3 M radicicol (Sigma) were used.
Chromatin Immunoprecipitation (ChIP)-ChIP assay was done as described in Ref. 39 with minor modifications. Briefly, K15 cells were cultured in 100-mm plates. After the experimental treatment, the cells were cross-linked with 1% formaldehyde and cross-linking was stopped by addition of glycine. The cells were lysed in radioimmune precipitation assay buffer, and DNA was disrupted into pieces of 500 -600 bp by sonication. The lysate was cleared by centrifugation and protein A-Sepharose beads were added to the lysate together with 1 l of anti-p53 antibody (DO-1, Santa Cruz Biotechnology). After overnight incubation at 4°C, the beads were washed, cross-linking was reversed, and DNA was purified on silica gel columns (A&A Biotechnology). Real-time PCR was performed to detect the p21 promoter fragment using the following primers: p21-ChIP-U, GTGGCTCTGATTGGCTTTCTG, p21-ChIP-L, CTGAAAACAGGCAGCCCAAG, annealed at 55°C with 1 mM MgCl 2 .
Supernatant was loaded onto a Q-Sepharose column equilibrated with buffer A and bound proteins eluted with linear gradient of 0 -0.5 M KCl in buffer A. Fractions containing Hsp90 were salted out with 30% (NH 4 ) 2 SO 4 following centrifugation at 70,000 ϫ g for 20 min. The supernatant was loaded onto a butyl-Sepharose column, which had been equilibrated with buffer A containing 30% (NH 4 ) 2 SO 4 and bound proteins were eluted with linear gradient from 30 to 0% (NH 4 ) 2 SO 4 . Hsp90 ␣-MBP was applied onto amylose resin (NEB) and eluted with 10 mM maltose. MBP tag was cleaved with factor Xa protease (NEB) according to the manufacturer's suggestions. Hsp90␣ protein of more than 95% purity was concentrated on a Resource Q FPLC column and dialyzed against buffer B: 25 mM Hepes, pH 7.5, 10% glycerol, 150 mM KCl, 1 mM DTT. Bovine brain Hsp90 was purified as described (40). Human recombinant Hsc70 (HSPA8), and Hdj1 were overexpressed and purified as described (33). Human recombinant Hsp70 (HSPA1A) was purified exactly as Hsc70, after overexpression in BL21 RIL E. coli strain from pET11b-Hsp70 construct, a kind gift from Prof. R. Morimoto. pMALc2x-hdj2 and pET30a-hdj3, constructs encoding human Hdj2 and Hdj3 were kind gift of Prof. K. Terada. Both proteins were purified from E. coli as previously described (41,42). All these chaperones tested positive for activity by the luciferase refolding assay (see below). p53 human recombinant protein was purified essentially as described (43).
ATPase Assay-ATPase activity was measured as previously described (44). 10 M Hsp90 was incubated in 20 l of buffer: 40 mM Hepes pH 7.5, 150 mM KCl, 5 mM MgCl 2 , 10 mM ATP, 0.5 Ci of [␥-32 P]ATP/ 100 l reaction buffer. Geldanamycin at a concentration of 500 M was added where indicated, and the reaction was carried out at 37°C. At time points 0 -120 min, 1-l samples were spotted on PEI-cellulose plates. Plates were resolved in 1 M LiCl: 1 M HCOOH, 1:1, dried and spots corresponding to non-hydrolyzed ATP and free phosphate were cut out, and radioactivity was measured in a liquid scintillation counter (Packard Bioscience). All results were corrected to the spontaneous ATP hydrolysis.

FIG. 2. ATPase and luciferase refolding activities of Hsp90␣
and Hsp90 purified from bovine brain (bov) are inhibited by geldanamycin. A, Hsp90 proteins at concentration 10 M were incubated in reaction buffer with 10 mM ATP and 0.1 Ci of [␥-32 P]ATP at 37°C, and 0.5 mM geldanamycin was added when indicated. At indicated time points 1/20 of the reaction mixture was spotted on PEI cellulose plates. After resolving and drying the plates were cut, and spots corresponding to free phosphate and non-hydrolyzed ATP were counted for radioactivity. B, Hsp90 preparations were incubated with 2 mM ATP or 100 M geldanamycin. Luciferase (12,66 mg/ml) was 400 times diluted with this mixture and incubated for 5 min at 50°C. After cooling down, the denatured luciferase was 6 times diluted with the renaturation buffer containing Hsc70, Hdj 1, ATP, and the ATP regeneration system using concentrations described under "Materials and Methods." The refolding reaction was carried out at 25°C. At indicated time points 1/6 of the reaction was tested for luciferase activity using the luciferase substrate. Controls (not shown) contained Hsp90s alone, with no other chaperones, and had refolding activity similar to "no Hsp90" reactions.
p53 DNA Binding Assay-The DNA binding activity of p53 was quantified by EMSA (gel-shift) assay. 50 ng of human recombinant p53 was diluted in the final volume of 5 l of EMSA buffer: 50 mM Tris pH 7.5, 5% glycerol, 50 mM KCl, 5 mM MgCl 2 , and 2 mM DTT. Samples were supplemented optionally with up to 5 g of Hsp90 (human recombinant ␣ or from bovine brain), other chaperones (see Fig. 5C) or BSA and 0 -20 mM ATP. Such 5-l samples were then incubated at 4 or 37°C for 1 h in a thermocycler. The activation step followed that included addition of 15 l of mixture containing: 1ϫ EMSA buffer, 0.2 Mdpm of 32 P-labeled p21 sequence (below), 1 g of nonspecific 44-bp dsDNA (sequence below, usage based on Ref. 46), and 100 ng of the antibody pAb421 (Ab-1; Oncogene). 20-l samples with the specific p21 DNA were afterward incubated for 5-10 min at room temperature, loaded onto a 4% native polyacrylamide Tris borate gel and electrophoresed at 15 mA for 2 h at 4°C. Gels were dried and exposed overnight to the Biomax MS-1 Kodak film (Sigma).
For p53 activated by CKII, the activation mix was made, containing per every 4 l of volume: 50 ng of p53, 7 units of CKII (Calbiochem), 0.3 units of creatine kinase, 150 mM of phosphocreatine (ATP regeneration system; Roche Applied Science), 1-5 mM of ATP and 1ϫ EMSA buffer. CKII activation was done for 30 min at 25°C, and then each 4 l of CKII activation reaction were supplemented separately with Hsp90/BSA in 1ϫ EMSA buffer up to 5 l. Afterward, a 1-h incubation step either at 4 or 37°C was performed, followed by addition of 15 l of DNA mixture without the antibody. Electrophoresis was performed as described above.
For testing the nonspecific p53 DNA binding activity, 0.2 Mdpm of nonspecific, radiolabeled 44-bp dsDNA (below) was used per sample instead of labeled p21 sequence in the activation step. The remaining part of the experiment was performed as described before for specific DNA but no additional unlabeled DNA, antibody, or CKII was used in this case.
Best results with geldanamycin in the EMSA assay were obtained when prior to the addition of p53, Hsp90 was preincubated with 83-500 M GA in the presence of 1ϫ EMSA buffer for 30 min at room temperature. After the addition of p53 and ATP in 1ϫ EMSA buffer, the final GA concentrations were 25-150 M. Since the stock solution of GA contained 100% Me 2 SO, in GA titration experiments all samples were supplemented with the same amount of Me 2 SO as added with GA.
Sequences were annealed to form double-stranded DNA in a thermocycler using the following program: 5 min 94°C, 5 min. 50°C, 4°C.
Presence of the dsDNA was tested with a 16% polyacrylamide Tris borate gel electrophoresis. Sequences used in the EMSA assay were labeled with the T4 polynucleotide kinase (PNK; Fermentas) as described in the producer's manual.
ELISA-Investigation of the p53 conformation was carried out using a two-site ELISA. First the wells were coated with wt-p53 conformation specific pAb1620 monoclonal antibody or DO-1 (both of mouse origin, Oncogene Science) at 50 ng per well in carbonate buffer pH 9.2 at 4°C for 16 h. The wells were blocked for 1 h at room temperature with 100 l of blocking wash buffer (25 mM Hepes-KOH pH 7.6, 5 mM DTT, 150 mM KCl, and 2 mg/ml BSA). This was followed by titration of increasing amounts of human recombinant p53, either kept at 4°C or incubated at 37°C for 1 h. The p53 dilutions were done in ELISA reaction buffer (25 mM Hepes-KOH pH 7.6, 5 mM MgCl 2 0.05% Triton X-100, 5 mM DTT, 150 mM KCl, 2 mg/ml BSA). Detection of p53 protein was carried out using the FL-393 antibody (rabbit origin, Santa Cruz Biotechnology) for 1 h diluted in blocking wash buffer at room temperature. This was followed by addition of anti-rabbit IgG-horseradish peroxidase secondary antibodies (Santa Cruz Biotechnology). Analysis of bound antibodies was performed by colorimetric detection with the TMB peroxidase EIA substrate kit (Bio-Rad), followed by absorbance measurements with a microplate reader (Bio-Rad) at 450 nm.

Inhibition of Hsp90 in Human Fibroblasts Decreases the
Wild-type p53 Transcriptional Activity-To investigate the possible interaction of the Hsp90 molecular chaperone with wt-p53 in vivo, we used K15 human fibroblasts. K15 cells express genotypically wt-p53, which can be immunoprecipitated with wt-p53-specific pAb1620 antibody, but not with the mutant p53-specific pAb240 antibody (data not shown). This suggests FIG. 3. Wild-type human recombinant p53 loses its wild-type conformation upon incubation at 37°C. A, p53 was incubated for indicated periods of time at 4 or 37°C and subsequently immunoprecipitated with the wild type-specific antibody pAb 1620. Detection after SDS-PAGE and blotting of the immunoprecipitates were performed with antibody DO-1, detecting total amount of the immunoprecipitated p53. Lower amounts of the wild-type conformation p53 was immunoprecipitated when p53 was kept at 37°C for a longer time. B, ELISA wells were coated with anti-p53 antibodies pAb 1620 (50 ng, wild type-specific) or DO-1 (50 ng, pan-specific), and increasing amounts of p53 were added to the wells. Upon incubation at 37°C for 1 h the amount of mutant conformation p53 rises (decrease in pAb 1620 antibody reactivity), which corresponds to the lower curve on the graph (open circles) in comparison with p53 kept at 4°C (filled circles). This effect was not because of decreased antibody accessibility to p53 caused by aggregation, since DO-1 antibody binding did not significantly change after 37°C incubation (open and filled squares). All presented values are means of four repeats. The OD is normalized appropriately to the maximum values for both antibodies.

p53-Hsp90 Functional Interactions
that in K15 cells, p53 exists in a functional, wild-type conformation. The functionality of p53 in the K15 cells was further demonstrated by its ability to induce p21/Waf1 expression and cell cycle arrest upon camptothecin treatment or ␥ irradiation ( Fig. 1 and data not shown).
Treatment of K15 cells with 2 M camptothecin (CPT), a drug which induces DNA damage by inhibition of topoisomerase I, results in a p53 up-regulation as shown by Western blot analysis (Fig. 1A). When the camptothecin treatment was performed in the presence of specific Hsp90 inhibitors, geldanamycin or radicicol (3 M), p53 was induced to somewhat lower levels, and the time course of the induction was slower (Fig. 1A and results not shown). Importantly, the relative amount of p53 phosphorylated at serine 15 was not affected by the Hsp90 inhibitors (Fig. 1A). This suggests that Hsp90 might act on p53 directly, rather than through ATR, DNA-PK, or ATM kinases, which are known to phosphorylate p53 at this position (47,48). To exclude the possibility that the lower p53 levels were caused by a decrease of its transcription, we quantified p53 mRNA by real-time RT-PCR. As shown in Fig. 1B, treatment of cells with geldanamycin or radicicol did not significantly change the level of p53 transcripts, suggesting that those drugs down-regulate p53 at the protein level. By fluorescence microscopy and cellular fractionation we showed that geldanamycin or radicicol do not affect the nuclear localization of p53 (results not shown). These results indicate that Hsp90 might play a role in the stabilization of the wt-p53 protein.
To test whether Hsp90 indeed can influence the transcriptional activity of p53, we analyzed the binding of p53 to the p21 promoter by ChIP. As shown in Fig. 1C, the camptothecininduced binding of p53 to the p21 promoter was almost completely disrupted by geldanamycin. The ChIP experiment for p21 promoter was also successfully performed using anti-Hsp90 antibodies, suggesting a functional interaction between wt-p53 and Hsp90. 2 Further, we have shown by real-time RT-PCR that geldanamycin or radicicol strongly inhibit the expression of p21 mRNA following the camptothecin treatment (Fig.  1D). This decrease in the p21 level could also be observed by Western blot (Fig. 1A). Taken together, described results suggest that while to some extent Hsp90 stabilizes genotypically wt-p53 protein level, a stronger effect is visible on the transcriptional activity of the wt-p53. These phenomena evidently depend on ATP, since both drugs effectively compete with ATP for binding to Hsp90.
ATPase and Chaperone Activities of Hsp90 Can Be Inhibited in Vitro-To answer the question, how does Hsp90 affect the transcriptional activity of p53, we used a reconstituted in vitro system, with highly purified recombinant proteins, to monitor p53 DNA binding to the p21 promoter sequence. For these tests, we used the human recombinant Hsp90 ␣ isoform, and for control experiments, we also purified Hsp90 from bovine 2 A. Helwak, unpublished results.  9 -11). The experiments were performed as described under "Materials and Methods." All EMSA bands in this and the following figures (unless indicated otherwise) correspond to the size of p53 tetramer bound to the DNA, supershifted by the activating antibody Ab421.
brain. Purified recombinant Hsp90␣ as well as bovine Hsp90 possess the ATPase activity, which is inhibited by geldanamycin ( Fig. 2A) and radicicol (results not shown). To test the molecular chaperone activity of purified Hsp90 proteins, we used a modified version of a previously described luciferase refolding assay (45). When Hsp90␣ or bovine Hsp90 was present during heat denaturation of luciferase (5 min, 50°C), the efficient refolding of luciferase was observed at reduced temperatures in the presence of Hsc70 and Hsp40 (Fig. 2B). This refolding reaction was dependent on the presence of Hsp90 during the denaturation process and was severely inhibited when geldanamycin was preincubated with Hsp90␣ or bovine prior to heat denaturation of luciferase (Fig. 2B).
Hsp90 Proteins Promote Specific DNA Binding Activity of Wild-type p53 at 37°C-The binding of the genotypically wildtype recombinant p53 protein to a p21 promoter sequence was monitored using the EMSA assay. As described before (49), we used antibody Ab421 or CKII to activate wt-p53 binding to the promoter DNA, and DO-1 antibody to perform p53 supershifts as specificity controls (not shown). p53 possesses two DNA binding sites, one sequence-specific, located in the central, core domain of p53 and another non sequence-specific, located at the C-terminal domain of p53 protein. To eliminate the effect of this second nonspecific p53 binding site, all p53 DNA binding assays were performed in the presence of a short DNA competitor. In many previous in vitro p53 DNA binding studies long (Ͼ100 bp) DNA competitors have been used, such as poly(dI-dC) (50), pBluescript vector (49), or salmon sperm DNA (51).
We also initially performed DNA binding experiments with poly(dI-dC) and pBluescript. However, as shown by Anderson et al. (46), long competitor DNA molecules inhibited the entry of p53-DNA complexes into the EMSA gel, which caused a decreased quality and reproducibility of our results. This problem was solved by the introduction of short, 44-bp competitor DNA fragments to the DNA binding reaction.
It has been previously shown that p53 protein can exist in a constant state of equilibrium between wild-type and mutant conformation (52,53). It is possible that elevated temperatures could shift this equilibrium toward a mutant conformation, hence the amount of p53 possessing the wild-type conformation should be decreased. Indeed, purified human recombinant wt-p53 protein loses its wild-type conformation upon incubation at 37°C and higher temperatures (Fig. 3 and results not shown). The immunoprecipitation of the wt-p53 by pAb 1620, specifically recognizing the wt-p53 conformation, is significantly reduced following incubation of wt-p53 at 37°C (Fig. 3A). The same effect was also observed using the modified ELISA test. In this case less p53 was detected by the conformation-specific pAb 1620 whereas comparable amount of the protein was detected by DO-1 at both temperatures (Fig. 3B). Consistent with these experiments, 1-h incubation at 37°C completely abolished the DNA binding activity of the genotypically wt-p53, as tested by the gel-shift assay (Fig. 4A, lane 2). The presence of increasing amounts of Hsp90 during this incubation step at 37°C significantly enhances the binding of p53 to the p21 promoter sequence (Fig. 4A). This regulation of p53 DNA bind- FIG. 5. Bovine brain Hsp90 has the same effect in vitro on p53 as Hsp90␣. GTP does not replace ATP in this reaction, and other human chaperones cannot substitute Hsp90. A, in lanes 5 and 6 bovine brain-purified Hsp90 was used instead of Hsp90␣. 2.5 g of Hsp90s and 0 -5 mM ATP were used. Remaining reaction conditions were similar to Fig. 4. B, lanes 1-3 contained 5 mM GTP that does not substitute for 5 mM ATP (lane 4) in the Hsp90␣-dependent recovery of p53. Nevertheless, it is apparent that GTP has some protective activity by itself, so p53 is not inactivated as much as in the presence of ATP or no nucleotide (see A). C, reactions carried out as in A; 1 h at indicated temperature, 5 mM ATP in all lanes. Bovine brain Hsp90 (lane 3) was used at 2.5 g (5.6 M). Human recombinant Hsp70 (HSPA1A) and Hsc70 (HSPA8) were used at 6 M in lanes 4 and 5 as substitutes for Hsp90. Combinations of chaperones from Hsp70 and Hsp40 family (Hdj1, 2, and 3) known to efficiently refold denatured luciferase were used in lanes 6 -11, at molar ratio Hsp/c70:Hsp40 2:1 (6:3 M). Lane 12 represents unspecific Hdj1 binding to the p21 promoter DNA in the absence of p53.

p53-Hsp90 Functional Interactions
ing by Hsp90 was found to be ATP-dependent, with at least 3 mM ATP required for the distinct effect (Fig. 4B). When we used poly(dI-dC), as a nonspecific competitor, the ATP dependence of this reaction was less pronounced (results not shown).
The results of the ATPase and luciferase refolding assays with Hsp90␣ and bovine Hsp90 indicate that both preparations possess ATP-dependent chaperone activity. Similarly, both Hsp90␣ and bovine Hsp90 enhanced p53 DNA binding to the promoter sequence (Fig. 5A). We also determined whether GTP could substitute for ATP in that reaction in order to rule out the involvement of a second, recently proposed nucleotide binding site on Hsp90 (54). GTP was unable to substitute for ATP during Hsp90-dependent binding of p53 to the promoter sequence (Fig. 5B). In addition, geldanamycin, which blocks the N-terminal nucleotide-binding site of Hsp90, inhibits the Hsp90dependent binding of p53 to the p21 promoter DNA (Fig. 6). Other recombinant human chaperones from Hsp70 and Hsp40 families were not able to efficiently substitute Hsp90 in this reaction (Fig. 5C), neither alone (results for Hdj proteins not shown) nor in different Hsp/c70-Hdj combinations, known to refold denatured luciferase (41,42) and suspected to alter the activity of such chaperone machines (55).
Similar to antibody-activated p53, the DNA binding of CKIIactivated p53 was also positively regulated by Hsp90 (Fig. 7). In the control experiment (Fig. 7A, lane 2), we observed that CKII-treated p53 protein appeared to be more resistant to incubation at 37°C as compared with p53 activated by the pAb421 antibody, which is in agreement with earlier sugges-tions (43). Surprisingly, the nonspecific binding of p53 to dsDNA, for which the unstructured C-terminal part of the protein is responsible (56), is also inhibited after incubation at 37°C (Fig. 7B). However, unlike sequence-specific DNA binding, Hsp90 in presence of 5 mM ATP did not positively regulate the nonspecific DNA binding activity of the wt-p53 (Fig. 7B). These results suggest that the observed Hsp90-mediated enhancement of p53 DNA binding at 37°C is restricted to the p53 core domain that is responsible for binding to specific, promoter-derived DNA molecules. These results fit the domain-specificity of Hsp90-p53 interaction, described in the accompanying article by Muller et al. (34).
The Mode of p53-Hsp90 Interaction Suggests Dynamic Mechanism for Positive Regulation of wt-p53 by Hsp90 -The ability of Hsp90 to restore p53 binding to the promoter sequence at 37°C cannot be explained by a simple, passive protection of the wild-type conformation of p53 at elevated temperatures, analogous to that described for RNA polymerase protection by DnaK (57,58). In the mentioned case, the presence of ATP diminishes to a great extent the DnaK-dependent protection of RNA polymerase at elevated temperatures. However, as shown in Fig. 8, the same concentration of ATP that is required for Hsp90-mediated binding of p53 to the promoter sequence ( Fig.  4B) also shifts the equilibrium of p53-Hsp90 complex formation toward its dissociation. The weaker effect of AMP-PNP, the ATP analog, indicates that at least nucleotide hydrolysis and probably its exchange is important for this reaction (Fig. 8). Consequently, 5 mM AMP-PNP could not substitute for ATP in the Hsp90-dependent enhanced DNA-binding of p53 in the EMSA assay (results not shown). Hsp90 also did not cause p53 supershifts in the EMSA assay when Hsp90 was present in reactions (Figs. 4 -7), which may suggest a transient nature of the protein-protein interaction. Taken together, these results indicate that the dynamic repeated binding and dissociation of the Hsp90 to p53 is responsible for the positive regulation of p53 DNA binding activity at physiological temperatures of 37°C.

DISCUSSION
The significance of cellular interactions of p53 with heat shock proteins in normal and tumor cells remains unclear. It has been demonstrated that mutant p53 associates to and is stabilized by a multichaperone complex (31,33). p53 protein, encoded by the non-mutated gene, has also been found associated to Hsp90 and presumably other chaperones at elevated temperatures, probably due to the conformational transition of p53 from wild-type to a form characteristic to the mutant protein (38). However, it was not possible to co-immunoprecipitate wild-type conformation p53 with molecular chaperones from cell lysates (26,29,59). A recently published study suggests that Hsp90 inhibitors influence the degradation of p53 temperature sensitive mutant at permissive temperatures in which p53 presumably adopts a wild-type conformation (60). In contrast, articles by other groups show no significant effect of Hsp90 inhibitors on genotypically wt-p53 level at physiological temperature (26,38,59), although these inhibitors may cause conformational change of wt-p53 (30).
In this article we show the effect of geldanamycin and radicicol on the activity of genotypically wt-p53 in human fibroblasts. Whereas the presence of geldanamycin or radicicol had minor effects on the cellular wt-p53 level and its phosphorylation at Ser-15, these Hsp90 inhibitors dramatically influenced p53 activity as the transcription factor, measured by chromatin immunoprecipitation as well as quantitative analysis of p21 mRNA and protein levels. The observed effect of Hsp90 inhibitors on the p53 transcriptional activity is not caused by the inhibition of the nuclear transport of p53 (results not shown). These in vivo data suggest that Hsp90 may play a role in the FIG. 6. The in vitro p53 rescuing activity of Hsp90␣ and bovine brain Hsp90 can be inhibited by increasing amounts of GA. A, 1.25 g of human Hsp90␣ was preincubated for 30 min with the indicated amount of GA (see "Materials and Methods") and used for p53 rescue from 37°C inactivation in the presence of 5 mM ATP (lanes 3-8).
In all other lanes reactions were carried out identically with 5 mM ATP but excluding Hsp90. Lane 9 is a control confirming lack of the negative effect of GA on p53 binding to the DNA in the absence of Hsp90. B, reactions as in A but bovine brain Hsp90 is used instead of Hsp90␣. In the case of this isoform less GA was required to fully inhibit Hsp90. For explanation of differences in overall intensity of bands between A and B see "Materials and Methods." regulation of p53-promoted transcription. To examine the possible effects of Hsp90 on p53 DNA binding activity, in vitro DNA binding studies were performed with purified human Hsp90␣ recombinant protein and Hsp90 purified from a bovine brain (a mixture of bovine ␣ and ␤ Hsp90 isoforms). Both Hsp90 protein preparations displayed similar ATPase and luciferase refolding activities that were inhibited by geldanamycin.
The gel shift assay was used to monitor in vitro p53 binding to the p21 promoter DNA sequence. A wt-p53 conformation, recognized by pAb 1620, is essential for this activity. Incubation of the p53 protein encoded by the non-mutated sequence, at 37°C decreases the amount of p53 protein found in an immunocomplex with pAb 1620 as well as the p53 DNA binding. In fact, incubation of wt-p53 for 1 h at 37°C completely abolished p53 binding to the p21 promoter-derived sequence. Interestingly, the presence of increasing amounts of Hsp90 during the incubation of p53 at 37°C can positively regulate p53 DNA binding to the promoter sequence, whereas other human recombinant chaperone proteins from Hsp70 and Hsp40 families were not able to substitute for Hsp90 activity. This activity is ATP-dependent and can be inhibited by geldanamycin. These effects correlate with our in vivo results, where geldanamycin inhibited p53 binding to the chromatin as well as transcription from the p21 promoter. In order to examine the possibility that transient Hsp90 interactions are required for positive regulation of p53 DNA binding to the promoter sequence at 37°C, we monitored the direct binding of Hsp90 to p53 in the presence or absence of ATP. Similar to Hsp70-substrate complex formation (61,62), the presence of ATP shifted the binding/dissociation equilibrium toward dissociation. These results suggest that the influence of Hsp90 on p53 DNA binding cannot be explained by the passive protection of wt-p53 conformation, caused by static association with Hsp90. We showed that the presence of ATP, which induces dissociation of Hsp90 from p53, also promotes the ability of p53 to bind to the DNA promoter sequence at 37°C.
The concentrations of geldanamycin and radicicol, which are sufficient to inhibit Hsp90-dependent in vivo transcriptional activity of p53 are lower than the concentration of Hsp90 inhibitors used in our in vitro assays, suggesting that we are only reconstituting the minimum Hsp90 chaperone systems in vitro. We are in the process of testing the hypothesis that the presence of Hsp90 co-chaperones could influence the inhibitors' affinity to Hsp90 and Hsp90-dependent p53 binding to the FIG. 7. Only the specific p53 DNA binding, activated by both Ab421 and CKII, is rescued by Hsp90 in vitro. A, specific binding of p53 to the labeled p21 promoter DNA was activated by the antibody Ab421 (lanes 4 -6; supershifted by the antibody) and CKII-dependent phosphorylation (lanes 1-3; not supershifted) and both could be inactivated by a 1-h incubation at 37°C (lanes 2 and 5). Although difference between binding strength at 4°C and 37°C is smaller in the case of CKII than in Ab421 activation (phosphorylation may protect against heat inactivation, see text for reference), Hsp90␣ rescues p53-specific binding in both cases (lanes 3 and 6). 5 mM ATP was present in all reactions, and in the case of CKII activation an ATP-regeneration system was used to maintain the ATP concentration. B, a 32 P-labeled nonspecific DNA sequence, and no nonspecific competitor, was used instead of the specific p21 promoter-derived DNA (see "Materials and Methods"). p53 binding to the nonspecific DNA was also inactivated by 37°C (2), but it cannot be restored by Hsp90␣ (3). All reactions contained 5 mM ATP. Protein amounts and ratios for A and B are as in Fig. 5A.   FIG. 8. ATP dissociates the Hsp90-p53 complex. Hsp90 (500 ng) was coated onto an ELISA plate in 50 ml of buffer (25 mM Hepes-KOH pH 7.5 and 100 mM KCl). The incubation proceeded for 2 h at room temperature. After the washing and blocking procedure (25 mM Hepes-KOH pH 7.5, 100 mM KCl, and 2 mg/ml BSA), the indicated amounts of wt-p53 in the reaction buffer (25 mM Hepes-KOH pH 7.5, 150 mM KCl, 10 mM MgCl 2 , 5% glycerol, 0.05% Triton X-100, 1 mg/ml of BSA) were added in the presence of 5 mM ATP or its analog AMP-PNP for 1 h at 25°C. The detection of p53 bound to Hsp90 was followed as described (33). All values are means of four repeats.

p53-Hsp90 Functional Interactions
promoter sequence. However, our results already indicate that the influence of Hsp90 on wt-p53 activity should be taken into consideration while using Hsp90 inhibitors in the therapeutic treatment of cancer, especially if cancer cells possess wt-p53.
There are at least two possibilities of explaining the mechanism for Hsp90 positive regulation of p53 DNA binding to the promoter sequence at 37°C. First, Hsp90 inhibits p53 aggregation or catalyzes the disaggregation of p53 protein at elevated temperatures. Such a mechanism was previously discovered for chaperones belonging to prokaryotic (57,58) and eukaryotic (63) Hsp70 families; and now is shown for Hsp90 in presence of p53 by Muller et al. (34). Second, the Hsp90 association with wt-p53 could induce the partial unfolding of p53. Following the dissociation of this Hsp90-p53 complex in the presence of ATP, p53 could spontaneously refold back into a wild-type conformation with a high affinity for the p21 promoter sequence. A similar mechanism of molecular chaperone action was proposed in the case of Hsp100 involved in protein folding and proteolysis (64). The unfoldase activity of Hsp100 molecular chaperone was eventually demonstrated by a subsequent study (65). Recent data from the Ted Hupp laboratory (30) suggest that Hsp90 in the presence of MDM2 could indeed partially unfold p53. We propose that partial unfolding of p53 could be catalyzed by Hsp90, and the subsequent spontaneous refolding of p53 back into a wild-type-like conformation may prevent p53 aggregation thus increasing p53 DNA binding to the promoter sequence (see Fig. 9). More importantly, these chaperone-mediated actions would decrease the probability for the formation of kinetically trapped, mutant-like intermediates and that would allow a shift in the conformational equilibrium toward the active, wt-p53 conformation. These events would ultimately promote the p53 transcriptional activity and allow for the ubiquitination and degradation of p53 protein. In addition, the retention of p53 in a wild-type conformation by transient Hsp90 interaction would also inhibit the formation of a multichaperone-p53 complex, which prevents p53 degradation and import to the nucleus (Fig. 9).