HSP70 Binding Sites in the Tumor Suppressor Protein p53*

Mutations within conserved regions of the tumor suppressor protein, p53, result in oncogenic forms of the protein with altered tertiary structures. In most cases, the mutant p53 proteins are selectively recognized and bound by members of the HSP70 family of molecular chaperones, but the binding site(s) in p53 for these chaperones have not been clearly defined. We have screened a library of overlapping biotinylated peptides, spanning the entire human p53 sequence, for binding to the HSP70 proteins, Hsc70 and DnaK. We show that most of the high affinity binding sites for these proteins map to secondary structure elements, particularly β-strands, in the hydrophobic core of the central DNA binding domain, where the majority of oncogenic p53 mutations are found. Although peptides corresponding to the C-terminal region of p53 also contain potential binding sites, p53 proteins with C-terminal deletions are capable of binding to Hsc70, indicating that this region is not required for complex formation. We propose that mutations in the p53 protein alter the tertiary structure of the central DNA binding domain, thus exposing high affinity HSP70 binding sites that are cryptic in the wild-type molecule.

The wild-type p53 protein is a tumor suppressor that arrests the cell cycle in response to DNA damage through sequencespecific DNA binding and transcriptional activation of specific genes, including those encoding proteins involved in the cell cycle, in cell growth arrest after DNA damage, and in apoptosis (reviewed in Ref. 1). Inactivation of p53 through deletions or point mutations (2)(3)(4) or interactions with other cellular or viral proteins (5-9) is a key step in many human cancers. A variety of mutations within conserved regions of the protein disrupt the sequence-specific DNA binding function of p53, either by altering residues critical for binding or by disrupting the overall tertiary structure of the molecule (10). A number of mutant p53 proteins with altered tertiary structures form stable complexes within cells with Hsc70, the major constitutively expressed member of the HSP70 1 family (11)(12)(13). In some cases, mutant p53 proteins have also been shown to associate with the heat shock-inducible family member Hsp70 (14 -16).
HSP70 molecular chaperones perform numerous functions in cells including stabilization of newly synthesized or unfolded polypeptides, facilitation of translocation of nascent chains across membranes, mediation of the assembly or disassembly of multimeric protein complexes, and targeting of proteins for lysosomal degradation (reviewed in Refs. [17][18][19][20]. These chaperone functions of HSP70 proteins are regulated and possibly targeted to different cellular locations by specific co-chaperones that are members of the DnaJ-like protein family (reviewed in Ref. 21). Hsp40, a mammalian homologue of DnaJ, is associated with the Hsp70-p53 complex in a carcinoma cell line (22). Hsc70 may also cooperate with another molecular chaperone, Hsp90. In a mammary tumor cell line, both Hsc70 and Hsp90 associate with mutant p53 in the cytoplasm, but the complex dissociates upon translocation of p53 into the nucleus (23). Recent studies in vitro and in vivo demonstrate a role for Hsp90 in the achievement and/or stabilization of the mutant conformation of a p53 molecule containing a point mutation (24). Hsc70 and Hsp90 have previously been shown to function together in regulating the conformation and activity of steroid hormone receptors (reviewed in Refs. 25 and 26). In this case, the chaperone-receptor complex dissociates upon hormone binding, exposing nuclear localization and DNA binding sites on the receptor molecule. The biological significance of the formation of HSP70-p53 complexes is unknown, although Hainaut and Milner (27) suggested that Hsc70 may play a role in the regulation of p53 conformation.
Although the sites of interaction have not been clearly defined, the interaction of mutant p53 with HSP70 proteins seems to involve conserved regions within both proteins, since a mutant murine p53 has been shown to associate with DnaK, the major HSP70 family member in Escherichia coli (28), and Xenopus laevis p53 can bind to mammalian HSP70 proteins (29). Various approaches have been used to identify the HSP70 binding site(s) within the p53 molecule. Deletion analysis by Sturzbecher et al. (30) showed that loss of Hsc70 binding occurred upon removal from p53 of N-terminal residues 13-66. Consistent with these results, Lam and Calderwood (31) identified a potential binding site within a synthetic peptide (P17G) corresponding to p53 residues 13-30, which include a highly conserved domain of p53 (Fig. 6A). However, further experiments by Sturzbecher et al. (30) showed that a double mutant lacking residues 13-66 as well as residues 200 -219 bound to Hsc70, suggesting that although amino acids 13-30 represent a potential site for Hsc70 binding, there must be additional binding sites in other regions of the molecule. We have previously shown that the peptide P17G binds to both Hsc70 and DnaK and that the major determinants for this binding are in the central region of this peptide (32). Hainaut and Milner (27) suggested that the C-terminal 28 amino acids of p53 are required for association with Hsc70, because Hsc70 fails to complex with p53 mutants with alterations in these residues. However, at least one of these mutants was reported to have a wild type conformation (23), so that lack of binding to Hsc70 may reflect inaccessibility of binding sites elsewhere in the molecule rather than loss of a specific site at the C terminus. Deletion of this region of p53 or binding to the bacterial HSP70, DnaK, activates the molecule for sequence-specific DNA binding, suggesting that the activation by DnaK occurs through binding to the C terminus of p53 (12,34). Direct peptide binding studies defined a likely binding epitope for DnaK near the C terminus and showed that Hsc70 also recognizes this region but with lower affinity than DnaK (32,35). We showed in addition that a peptide corresponding to a sequence in the central region of the molecule bound to both Hsc70 and DnaK (32). In the present study, we confirm the existence of potential binding sites for HSP70 proteins in the N-and C-terminal regions of p53 but show that the great majority of high affinity sites lie in the central, DNA binding domain of the molecule.

EXPERIMENTAL PROCEDURES
Proteins and Cell Extracts-Bovine brain Hsc70 and biotinylated Hsc70 were obtained from StressGen Biotechnologies Corporation, and DnaK was obtained from Epicentre Technologies. Cytoplasmic extracts were prepared from KHOS-240S human osteosarcoma cells that had been heat-shocked for 2 h at 42°C. The cells were washed twice with phosphate-buffered saline, scraped from the culture dish, and then lysed by 20 strokes of a homogenizer in 500 l/100-mm dish of hypotonic buffer (20 mM HEPES (pH 7.2), 10 mM KCl, 5 mM MgCl 2 , 0.2 mM phenylmethylsulfonyl fluoride, and 10 g/ml leupeptin (36)).
Peptides-The synthesis, N-terminal biotinylation, and sources of the peptides listed in Table I were as described previously (32,37). The peptide library representing the entire human p53 sequence (38) was obtained from Chiron Mimotopes P/L (Victoria, Australia). The library consists of 77 peptides (numbered 3-79) corresponding to overlapping 15-amino acid segments of the human p53 sequence, linked to biotin at the N terminus via an additional peptide spacer of Ser-Gly-Ser-Gly. Each peptide shares a 5-amino acid overlap in the primary p53 sequence with the previous peptide.
Assays of Peptide Binding to HSP70 Proteins-Direct binding of N-terminally biotinylated peptides or competition of binding of biotinylated peptide by excess unmodified peptides was measured as follows. Bovine Hsc70 or E. coli DnaK (1 g/assay) was incubated for 15 min at 37°C with biotinylated p53 peptides at a final concentration of 5 or 50 M in the absence (direct binding) or presence (competition) of a 10-fold excess of unmodified peptide in a final volume of 10 l containing 40 mM Hepes (pH 7.0), 75 mM KCl, and 4.5 mM Mg(OAc) 2 . In some cases, cytoplasmic extracts containing Hsc70 and Hsp70, prepared as described above from heat-shocked human osteosarcoma cells, were incubated with biotinylated peptides at final concentrations of 50 M. The HSP70-peptide complexes were then separated from free peptides by nondenaturing polyacrylamide gel electrophoresis (PAGE; 6% acrylamide) and transferred to nitrocellulose. Complexes of HSP70 proteins and biotinylated peptides were visualized by incubation with peroxidaseconjugated streptavidin and the ECL detection system (Amersham Corp.). The amount of each biotinylated peptide bound was quantitated by densitometric scanning of the bands corresponding to Hsc70 or DnaK, respectively, and expressed relative to the amount of binding to Hsc70 of peptide 28, which was included as an internal control in each separate set of assays.
Competition by Peptides of Co-immunoprecipitation of Mutant p53 and HSP70 -Human KHOS-240S osteosarcoma cells (which express a mutant form of p53 (39)) were heat-shocked by incubation at 42°C for 2 h and then labeled at 37°C for 2 h with 300 Ci of [ 35 S]methionine/ 100-mm culture dish. The cells were washed with phosphate-buffered saline and lysed on ice for 20 min in 1 ml of 50 mM Tris-HCl, pH 8.0, containing 1% Nonidet P-40, 0.5 units/ml aprotinin, 100 M leupeptin, and 500 M phenylmethylsulfonyl fluoride. After centrifugation for 10 min at 14,000 ϫ g to remove cell debris, aliquots (80 l) of 35 S-labeled lysate were precleared by incubation with an irrelevant antibody and protein A-Sepharose for 2 h at 4°C. Precleared extracts were incubated overnight at 4°C, either in the absence of added peptide (control) or in the presence of peptide P17G (660 M), V10 (590 M), or S16D (660 M).
Assays of Association of Wild-type p53 and Deletion Mutants with Biotinylated Hsc70 -Wild-type human p53 and N-and C-terminal deletion mutants were translated in vitro from cDNAs cloned into pBluescript SK Ϫ (40). The reactions were performed in a final volume of 50 l in the presence of [ 35 S]methionine (Amersham, 40 Ci/translation) using the Promega TNT kit according to the manufacturer's instructions. Aliquots (5 l) of the in vitro translated 35 S-labeled p53 species were diluted to a final volume of 20 l in buffer A (150 mM NaCl, 50 mM Tris, pH 8, 1% Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride) and incubated for 20 min at 37°C and 30 min at 4°C with 1 g of biotinylated Hsc70 (StressGen) in the absence or presence of 3 mM o-phenanthroline. Streptavidin-agarose (Sigma; 55 l of a 20% suspension in buffer A plus 0.05% bovine serum albumin) was then added, and the mixtures were incubated with shaking at 4°C for 45 min to facilitate the binding of the biotinylated Hsc70 molecules and any associated p53 species. The streptavidin-agarose beads were sedimented and washed three times with buffer A, and associated 35 S-labeled p53 species were analyzed by SDS-PAGE (8% acrylamide) and fluorography.

Specific Binding to Hsc70 and DnaK of Biotinylated p53
Peptides-Initially seven biotinylated peptides with sequences (Table I) corresponding to various segments of wild-type and mutant p53 proteins were tested for their ability to bind specifically to purified DnaK and Hsc70 or to HSP70 family members present in an extract of heat-shocked osteosarcoma cells. The nonbiotinylated versions of these peptides were previously demonstrated by other assays to bind to DnaK, Hsc70, and the endoplasmic reticulum-located HSP70 family member, BiP (32). Peptide P17G and subpeptides P6S, T6L, and K6G correspond to sequences near the N terminus of wild-type murine p53, while peptides S16D and S10S correspond to sequences in the C-terminal region of wild-type human p53. S2V10 corresponds to a point-mutated sequence from the highly conserved region II in the central core domain of human p53 (41).
The N-terminally biotinylated peptides were incubated with the HSP70 proteins, and HSP70-biotin peptide complexes were separated from free biotinylated peptides by nondenaturing PAGE, transferred to nitrocellulose, and visualized with streptavidin peroxidase and an enhanced chemiluminescence detection system (see "Experimental Procedures"). Three of the peptides (P6S, K6G, and S10S) showed no interaction with any of the HSP70 proteins when tested at a concentration of 50 M (Fig. 1). The remaining four peptides all displayed significant binding to Hsc70 (which migrates as a triplet representing different oligomeric forms (Fig. 1A)), to DnaK (which migrates as a closely spaced doublet (Fig. 1B)), and to Hsc70 and/or Hsp70 in a cytoplasmic extract (Fig. 1C). The specificity of binding of these biotinylated peptides to HSP70 family members was highlighted by their differential binding to the Hsc70 and Hsp70 proteins present in a complex mixture of cytoplasmic proteins (Fig. 1C). Peptides P17G and T6L bound only to Hsc70 in the cytoplasmic extract, whereas S2V10 and S16D bound to both Hsc70 and Hsp70, indicating target specificity differences even between these closely related proteins. The peptide-associated species of higher mobility than Hsc70 may represent BiP, particularly since this band was undetectable by biotin S16D binding (we have previously reported that BiP does not bind peptide S16D (32)). The results obtained in these experiments are summarized in Table I, where they are compared with previous data obtained using an assay based on competition by the nonbiotinylated versions of these peptides of the binding of HSP70 proteins to reduced and carboxymethyl-ated lactalbumin, a model unfolded polypeptide (32). The two sets of data are qualitatively in good accordance.
Specificity of the binding of biotinylated peptides was also tested by competition with excess unlabeled peptides (Fig. 2). Hsc70 was incubated with 50 M biotinylated T6L, P17G, S2V10, or S16D, respectively, in the absence (Ϫ) or presence of a 10-fold excess (500 M) of the corresponding unlabeled peptide or of peptide V10, which binds to both Hsc70 and DnaK (see Table I and Ref. 32). This sequence has previously been shown to be presented by major histocompatibility complex class I molecules on the surfaces of tumor cells expressing the mutant p53 protein (42). Binding of each of the biotinylated peptides was specific and saturable, since it could effectively be competed by an excess of the same unlabeled peptide, or by peptide V10. Similar results were obtained for DnaK (not shown). Peptides that did not bind to the HSP70 proteins in the direct binding assay (e.g. P6S and K6G) were also unable to compete with the binding of a biotinylated peptide (data not shown). Thus, competition of biotinylated peptide binding exhibits similar specificity to direct binding.
Lam and Calderwood (31) previously showed that excess peptide P17G could effectively compete the co-immunoprecipitation of Hsc70 and mutant p53. The data shown in Fig. 3 demonstrate that peptides P17G, V10, and S16D can prevent co-immunoprecipitation of both Hsc70 and Hsp70 with fulllength p53 from extracts of heat-shocked osteosarcoma cells. Consistent with their relative efficacies in competing the binding of biotinylated peptides (Fig. 2), peptide P6S did not affect the association of the HSP70 proteins with full-length p53, while peptide S2V10 did interfere with this association, but with lower affinity than peptide V10 (results not shown).
Identification of Additional HSP70 Binding Sites in p53-To identify all potential binding sites in p53 for binding to Hsc70 and DnaK, we utilized a peptide library representing the complete wild-type human p53 protein sequence. This peptide library has been used previously to identify the Mdm2 binding site in the N-terminal region of p53 (38) and to define the epitopes recognized by monoclonal anti-p53 antibodies (43). The library consists of 77 different peptides containing 15residue segments of the p53 amino acid sequence (see Fig. 6B).
a Data are from densitometric quantitation of the bands in Fig. 1 and were normalized relative to the binding of peptide S2V10, which was designated as moderate affinity (ϩ). Peptides were classified as negligible affinity (Ϫ) if no binding was observed or low affinity (Ϯ) for 0.5-fold or less relative to S2V10 binding. High affinity binding was classified into two categories, ϩϩ for 3-5-fold higher binding than S2V10 and ϩϩϩ for 6-fold or greater increased binding relative to S2V10. b Data are from Fourie et al. (32). Shown is the range of concentrations of peptide at which 50% inhibition of the binding of reduced and carboxymethylated lactalbumin (RCMLA) to the HSP70 protein was achieved. The ranges of apparent K d values corresponding to the ratings given in this column are as follows: ϩϩ, Ͻ200 M; ϩ, 200 -500 M; Ϯ, 500 M to 1 mM; Ϫ, Ͼ Ͼ1 mM. Where actual values are given, the apparent K d has been estimated from a binding curve over a range of peptide concentrations.
FIG. 1. Binding of biotinylated peptides from p53 to Hsc70 and DnaK. Bovine Hsc70 (A), DnaK (B), or a cytoplasmic extract from heat-shocked osteosarcoma cells containing Hsc70 and Hsp70 (C) was incubated for 15 min at 37°C with 50 M biotinylated p53 peptides, P17G, P6S, T6L, K6G, S2V10, S10S, and S16D (see Table I for sequences). Free biotinylated peptides were then separated from HSP70-peptide complexes by native PAGE, followed by transfer to nitrocellulose and detection with streptavidin-peroxidase and ECL. The additional lane in C is an immunoblot with anti-HSP70 monoclonal antibody MA3-006 (Affinity Bioreagents) to indicate the positions on the gel of the Hsp70 and Hsc70 species. Note that binding of peptide P17G to any of the HSP70 proteins results in their increased mobility on native PAGE. Similar observations were made for this peptide by Lam and Calderwood (31). The peptides consecutively overlap by 5 amino acids and are attached to a biotin molecule via a 4-amino acid spacer (Ser-Gly-Ser-Gly-) at the N terminus.
Assays measuring the binding of the biotinylated library peptides to Hsc70 and DnaK were performed as described above for the experiment shown in Fig. 1, at peptide concentrations of both 5 and 50 M for binding to Hsc70, and at 5 M for binding to DnaK. The SDS-PAGE analysis of the assays performed with 5 M peptides are shown in Fig. 4, and the results of densitometric quantification of the data are presented in Fig. 5. The analysis using 5 M peptides showed that no high affinity sites for Hsc70 binding are present within residues 1-110 of human p53 (which are included within peptides 3-24). As discussed earlier, HSP70 binding site(s) had previously been identified within peptide P17G, which corresponds to residues 13-31 of murine p53 (Refs. 31 and 32; see Table I and Figs. 1 and 2). The weak binding observed with biotinylated peptides that include the corresponding residues of human p53 (peptides 5-8; see Figs. 4 and 5) suggests that the amino acid differences in the human sequence (perhaps particularly the substitution of Asp for Gly at residue 21; see Table I) result in a lower affinity interaction with Hsc70. The majority of the high affinity sites for binding to Hsc70 are contained within peptides 25-58, which include residues 111-290 of p53 and correspond to the central, highly conserved, sequence specific DNA-binding domain of the molecule (Fig.  6A). High affinity binding was also observed for peptides 77-79, corresponding to the C-terminal 23 amino acids of p53, which contain peptide S16D (see Table I). Binding of many of the peptides corresponding to the central and C-terminal regions of p53 apparently displayed saturation at 5 M concentration, since no further increase in binding was observed at 10-fold higher concentrations, while increased binding to Hsc70 was observed at 50 M for some lower affinity peptides such as those corresponding to the N-terminal region corresponding to peptide P17G (data not shown). The affinities of the central and C-terminal p53 peptides are thus in the highest range previously observed for peptide binding to Hsc70, i.e. less than 10 M (32).
The patterns for binding of the p53 library peptides to Hsc70 and DnaK were qualitatively similar, although some quantitative differences were observed, consistent with our previous observations of overlapping and individual specificities for HSP70 proteins (32). The differences were most apparent among the weaker binders. For example, peptides 5, 6, and 7 showed significantly more binding to DnaK than Hsc70, as did peptides 19, 21-24, and 61-64, while peptides 67 and 68 appeared to have higher affinity for Hsc70 than DnaK (Figs. 4  and 5). Among the stronger binders, peptides 29, 30, 31, 55, and 56 displayed significantly greater binding to DnaK than Hsc70, while peptides 38, 44, 45, and 46 appeared to have a lower affinity for DnaK than for Hsc70. In general, DnaK displayed a broader pattern of recognition, binding a greater number of sites that lie outside the central core of the p53 molecule.
Potential Hsc70 Binding Sites in p53 Are Frequently Located in ␤-Strands within the Hydrophobic Core- Fig. 6B shows the sequences of the peptides that make up the p53 library, aligned under the complete p53 sequence and designated by their relative shading as moderate affinity or high affinity for binding to Hsc70. Shown in boxes on the p53 sequence are the putative binding sites recognized by Hsc70, which have been identified by comparing the relative binding affinities of the overlapping peptides. In each case, we looked for a stretch of 7 residues (the apparent length of the HSP70 recognition motif (44)) that was present in one or more overlapping peptides that bound to Hsc70 but absent in neighboring peptides that bound Hsc70 with lower or negligible affinities (occasionally taking into account the fact that for an individual peptide, the N-terminal spacer (SGSG-) might contribute to the binding motif). Also shown above the p53 sequence are the secondary structure elements (␤-sheets and ␣-helices) and the residues (Cys 176 , His 179 , Cys 238 , and Cys 242 ) that coordinate the bound Zn 2ϩ atom, all of which have been defined by analysis of the threedimensional structures of the isolated core domain (10) or the oligomerization domain (45)(46)(47).
As noted above, there are no high or moderate affinity sites for binding of Hsc70 present within the N-terminal 110 residues of human p53. The three-dimensional structure of this domain has not been defined, but proteolysis experiments indicate that it has a relatively unstructured conformation (48). The great majority of the potential Hsc70 binding sites lie within the core domain, which corresponds to residues 102-292 (see rows 2 and 3 of Fig. 6B). Most interestingly, 10 of the 15 high affinity Hsc70-binding peptides in this domain (which contain 8 of the 11 proposed binding sites) involve residues that form ␤-strands in the folded core domain, while 2 of the remaining 3 proposed sites include either residues that coordinate the Zn 2ϩ atom or residues that fold into the ␣-helix, which contacts the major groove of the DNA helix during sequencespecific DNA binding by p53 (10). The single, moderate affinity Hsc70 binding site identified within the oligomerization domain (which corresponds to residues 319 -360 in row 4 of Fig.  6B) also aligns with the single ␤-strand in this domain. The remaining high affinity site, at residues 379 -385 near the C terminus of p53, lies in a region whose three-dimensional structure has not been defined, although it is more solvent-exposed and protease-sensitive than the core and oligomerization domains (48).
Analysis of Binding of p53 Deletion Mutants to Hsc70 -Although we have located a relatively large number of potential Hsc70 binding sites within the human p53 sequence, we do not know which of these sites might be important during the in- teraction of p53 and Hsc70. To establish whether the potential Hsc70 binding sites in the N and C termini of the p53 molecule are required for binding of the protein to Hsc70, four deletion mutants of p53 that lack either 75 or 150 residues from the N or C terminus of the polypeptide (40) were tested for binding to biotinylated Hsc70 in the absence or presence of o-phenanthro-line, which can disrupt the wild-type conformation of p53 by chelating the Zn 2ϩ ion that stabilizes the core domain (33,49).
Wild-type p53 and the N-and C-terminal deletion mutants were translated in vitro in the presence of [ 35 S]methionine and then incubated (in the presence or absence of o-phenanthroline) with biotinylated Hsc70. Streptavidin-agarose was then used to precipitate the biotinylated Hsc70 and associated 35 Slabeled p53 molecules. The relative amounts of the p53 species that were coprecipitated were determined by fluorography. Fig.  7A shows an example of the fluorogram obtained for the ⌬C75 and ⌬C150 deletion mutants. In each case, the amount of the C-terminal deletion mutant associated with Hsc70 in the presence of o-phenanthroline (lanes 3 and 6) is higher than that observed in the absence of either Hsc70 (lanes 2 and 5) or o-phenanthroline (lanes 1 and 4). The results for all of the p53 species were quantitated by densitometry and are shown in Fig. 7B. In each case, the central column (shaded) indicates the amount of p53 polypeptide co-precipitated with biotin Hsc70 in the presence of o-phenanthroline and is compared with the amount precipitated in the absence of Hsc70 (left column, standardized to 1 arbitrary unit), or o-phenanthroline (right column). No significant difference was observed in the amounts of full-length p53, ⌬N75, or ⌬N150 precipitated with streptavidin-agarose in the presence or absence of biotinylated Hsc70 or o-phenanthroline. These results indicate that, under the conditions of this experiment, none of the potential Hsc70 binding sites present within full-length p53 or within the Nterminally truncated proteins are available for interaction with Hsc70. However, an increased amount of labeled ⌬C75 and ⌬C150 proteins was precipitated with streptavidin-agarose in the presence of biotinylated Hsc70 either in the absence (a 50 -80% increase) or presence of o-phenanthroline (a 200 -400% increase). This demonstrates that the C-terminally truncated p53 molecules contain sites that can interact with biotinylated Hsc70 and that the availability of these sites is increased following chelation of the Zn 2ϩ atom by o-phenanthroline. Thus, although the C-terminal 150 amino acids of p53 contain potential high affinity binding sites for Hsc70 (within peptides 55-58 and 77-79; see Fig. 4), these sites do not appear to be required for stable interaction with Hsc70. We conclude that the high affinity Hsc70 binding sites present in the central DNA binding domain of the molecule corresponding to peptides 25-39 (amino acids 116 -195) and 44 -47 (amino acids 206 -235) are sufficient to promote a stable interaction between p53 and Hsc70. These sites are presumably cryptic in the wild-type molecule, even following removal of the Zn 2ϩ atoms by chelation, but can be exposed by conformational changes induced by deletion of C-terminal sequences or by point mutations (11)(12)(13)50). An alternative explanation for the increased association of the C-terminal deletion mutants with the chaperone is that Hsc70 preferentially associates with lower oligomeric forms of p53. Hsc70 has been shown to selectively complex with p53 dimers and possibly monomers, but not with higher molecular weight forms of mutant p53 (27). ⌬C75 and ⌬C150 lack the tetramerization domain and have been shown to be incapable of oligomerization (40).
Control immunoprecipitations (not shown) indicated that all of the in vitro translated p53 species could be recognized by antibody PAb240, considered to be specific for the "mu-tant" conformation of p53 (51). In vitro translated wild-type p53 can display either a wild-type or a "mutant" conformation, depending on the batch of reticulocyte lysate, although the factor(s) responsible for this effect have not been identified (52). Because we found that only the C-terminally truncated p53 species interacted with Hsc70, there is no direct correlation between the PAb240 ϩ conformation and the exposure of Hsc70 binding sites. A possible explanation is that p53 molecules translated in vitro are in a conformational equilibrium between the "native" and "mutant" states. For the wild-type molecule and the N-terminal truncation mutants, this equilibrium greatly favors the native state, and Hsc70 does not interact. However, the PAb240 antibody (whose K d would be in the nanomolar range, in contrast to that of Hsc70, which is in the micromolar range (32)) could bind to the very small proportion of p53 molecules that are in the mutant conformation and shift the equilibrium by mass action, so it eventually complexes a significant proportion of the p53 molecules. For the p53 proteins with C-terminal truncations, it appears that the conformational equilibrium is shifted further toward the "mutant" state that exposes Hsc70 binding sites, especially when the structure is further loosened by chelation of the Zn 2ϩ atom by o-phenanthroline. DISCUSSION The p53 protein contains three major functional domains: an N-terminal domain, which is responsible for transcriptional transactivation by p53 (53); the central core domain required for specific DNA binding (10); and the C-terminal domain, which contains nuclear localization sequences (54), the oligomerization domain (45), and regulatory phosphorylation sites (12,(55)(56)(57). Many different cellular and viral proteins interact with p53 (reviewed in Ref. 58), and binding sites for a number of these proteins, including Mdm2 (38, 59), human papilloma virus E6 (60), and Tms1 (61), have been identified within the p53 amino acid sequence. However, the binding site(s) for the molecular chaperone Hsc70 that become exposed when p53 adopts a "mutant" conformation (11-13, 50) have not been clearly defined, although sites within the N-and C-terminal domains of the molecule have been proposed (27, 30 -32). In this study, we present the first analysis of the whole p53 molecule for potential binding sites for Hsc70 and its bacterial homologue, DnaK.
We have confirmed the presence of potential binding sites for these HSP70 proteins in the N-and C-terminal regions of p53, but, in addition, we identify for the first time a large number of high affinity sites in the central DNA binding domain of the molecule. Interaction of Hsc70 with sites in the central domain of p53 is consistent with the ability of this chaperone to discriminate between wild-type and mutant conformations of p53 (50) or between folded and heat-denatured forms of the wildtype protein (62). Structural analysis of p53 has revealed that the N-and C-terminal domains are relatively unstructured (48), and accordingly, antibodies to these regions are not conformation-specific. By contrast, the central region takes up a defined conformation critical for the DNA binding function of p53 (10), and antibodies to this domain are conformation-specific (43,50,51,63,64). Studies using such conformationsensitive antibodies indicate that, even in the wild-type p53 molecule, the core domain has a metastable structure that can be perturbed by treating the protein with chelating or oxidizing agents (33,49) or by raising the temperature (65). These conformational changes are accompanied by a decrease in the specific DNA binding activity of the protein. Point mutations in p53, particularly within the core domain, also frequently disrupt the conformation of the molecule and alter its capacity to bind DNA and transactivate gene expression (48). However, adoption of a "mutant" conformation by p53 is not necessarily an all-or-none or irreversible phenomenon. Antibodies specific for this conformation frequently immunoprecipitate only a proportion of the population of mutant polypeptides, indicating that some of the molecules remain in a "wild-type" conformation in which the antigenic epitopes are buried (43). Furthermore, many p53 mutants can refold into a wild-type form when subjected to appropriate conditions (66).
A picture emerges of a wild-type p53 molecule with a core domain poised to respond to cellular messages (1, 67, 68) that are delivered when proteins such as Mdm2 or Tms1 bind to sites in the N-or C-terminal domains of the molecule or when residues near the C terminus are phosphorylated. These messages are relayed allosterically to the core domain, which then undergoes conformational changes that alter its DNA binding activity. Interaction of the C-terminal regulatory domain and the core domain was proposed based on the NMR structure of the oligomerization domain (45,46). Furthermore, proteolytic removal of the C terminus activates sequence-specific DNA binding (12), suggesting that the C-terminal domain interacts with the core domain in the native molecule and that disruption of this interaction increases the DNA binding activity of the core. Binding of DnaK to the C terminus of wild-type p53 also activated DNA binding (12), but Hsc70 did not have the same effect. 2 This is consistent with our previous results showing that the affinity of Hsc70 for the C-terminal sequences is lower than that of DnaK (32,35).
In its active state, the core domain consists of two ␤-sheets, which serve as a scaffold for the structures that interact with the DNA: 1) two large loops that are stabilized by a tetrahedrally coordinated Zn 2ϩ atom and 2) a loop-sheet-helix motif (10). Presumably, in the inactive state, which can be promoted either by binding of cellular proteins or by the introduction of mutations, the structure of the core domain is loosened and altered so that specific DNA binding contacts can no longer be 2 T. R. Hupp and D. P. Lane, unpublished results.  35 S-labeled p53 was incubated for 20 min at 37°C in the absence or presence of 3 mM o-phenanthroline (OP) and 1 g of biotinylated Hsc70. Streptavidin-agarose was then added to precipitate biotinylated Hsc70 and any associated p53 species. The precipitates were analyzed by SDS-PAGE and fluorography. The results obtained with the C-terminal deletion mutants ⌬C75 and ⌬C150 are shown in panel A. The additional species of higher apparent molecular weight observed for each of the p53 mutants may have arisen by premature initiation of translation at alternative methionines and were observed in previous experiments with the same DNA constructs (Sang et al. (40)). The densitometric quantitation of the amounts of wild-type and truncated p53 species co-precipitated with Hsc70 is shown in panel B. The data in B were normalized by expressing the amount of p53 coprecipitated relative to that observed in the presence of o-phenanthroline and the absence of Hsc70. made, and previously buried epitopes become available to conformation-specific antibodies. For many mutants, the structural loosening is sufficiently extensive to expose potential binding sites for Hsc70, which, as we have shown in this study, are concentrated in the core domain. These sites correspond to sequences that in the active, DNA-binding conformation are folded into secondary structural elements, particularly ␤-strands, that are wholly or partly buried in the interior of the protein. Thus, the Hsc70 binding sites are usually cryptic in the wild-type p53 molecule but can be exposed by changes in tertiary and/or quaternary structures induced by point mutations or deletions such as those found in human cancer, by heat denaturation, or by other unknown changes to the wild-type molecule that occur, for example, in normal cells during stimulation of growth (67), in certain reticulocyte lysate preparations (52), or following phosphorylation (12,69).
The proposed epitopes for Hsc70 binding were identified by comparing the relative affinities of the p53 library peptides, searching for a stretch of 7 residues (the apparent length of the HSP70 recognition motif (44)) that is present in one or more overlapping peptides that bind to Hsc70 but is absent in neighboring peptides that bind Hsc70 with lower or negligible affinities. For the most part, the heptameric sequences shown in Fig. 6 fit either the consensus motif of bulky hydrophobic/ aromatic residues present in two or more alternating positions proposed for binding of the ER HSP70 protein BiP (37) or the consensus motif of a cluster of 3-5 hydrophobic residues (often anchored by a central leucine) flanked by basic residues proposed for binding to DnaK (70). No such consensus motif has been developed for Hsc70, although we have shown that a central, large hydrophobic residue is crucial for peptide binding to Hsc70 as well as to BiP and DnaK (32). In addition, Takenaka et al. (71) have described a small set of Hsc70-binding peptides that have in common the presence of hydrophobic residues flanked on one side or the other by basic amino acids. We note that the proposed Hsc70 binding sites in p53 have hydrophobic residues either concentrated in the center of the heptamer or dispersed throughout the epitope, and the majority contain a basic residue in the first or second position of the heptamer. We have previously shown that BiP, Hsc70, and DnaK display both common and divergent peptide binding specificities (32), so that we would expect a degree of overlap in their binding sites. Indeed, we have shown that Hsc70 and DnaK bind to many of the same p53 library peptides, although there is not complete overlap, with DnaK displaying a broader pattern of recognition, binding a greater number of sites that lie outside the central core of the p53 molecule.
The degree to which the potential Hsc70 binding sites map to known secondary structural elements in the p53 molecule is remarkable. Eight of the eleven proposed binding sites in the DNA binding domain involve residues that form ␤-strands in the folded structure (10). These ␤-strands pack together to form the hydrophobic core of the domain, consistent with the binding sites only being available to Hsc70 when the structure of the core domain is disrupted. Many of the hydrophobic and aromatic residues present in the potential binding epitopes (e.g. Val 143 , Leu 145 , Val 157 , Val 218 , Ile 232 , Tyr 234 , Tyr 236 , Ile 255 , and Phe 270 ) correspond to amino acids that Cho et al. (10) designate as making van der Waals interactions that stabilize the core. Furthermore, a subset of these hydrophobic residues (e.g. Val 143 , Val 157 , Tyr 234 , Tyr 236 , and Phe 270 ) have been targets for substitution in mutants that apparently affect p53 function by destabilizing the structure of the core domain (10). Two of the remaining three potential Hsc70 binding sites in the core domain include residues involved in coordination of the Zn 2ϩ atom (including Arg 175 , which is a hot spot for mutation) or residues that make up the ␣-helix involved in sequence-specific DNA binding by p53 (10). The single, moderate affinity Hsc70 binding site present within the oligomerization domain also aligns with the single ␤-strand identified as a crucial element for tetramerization of p53 (45). Again the potential binding site contains hydrophobic and aromatic residues (e.g. Phe 328 , Leu 330 , Ile 332 , Phe 338 , and Phe 341 ) that Clore et al. (45) designate as making the van der Waals interactions that stabilize the interaction between this ␤-strand and the ␣-helix, which is the only other secondary structural element in the oligomerization domain. Leu 330 is substituted in a tumor-derived mutant p53 that is one of the minority that contains an amino acid alteration that maps to the oligomerization domain (45). The remaining high affinity site, at residues 379 -385 near the C terminus of p53, lies in a region whose three-dimensional structure has not been defined. However, Clore et al. (45) suggest that the C-terminal residues of p53 normally interact with the core domain and inhibit specific DNA binding, either by preventing the core from adopting the optimal conformation for sequence-specific DNA recognition or by masking the DNA recognition site. This interaction can be disrupted and DNA binding can be potentiated by removal of the C-terminal residues by proteolysis, by mutagenesis, by modification of the C terminus by phosphorylation, by binding to an antibody that recognizes an epitope in this region, or by binding to DnaK but not to Hsc70 (12). Since our p53 peptide binding data and the data reported by Hansen et al. (35) indicate that binding site(s) for both DnaK and Hsc70 are located near the C terminus, it seems likely that the two chaperones bind to different (although perhaps overlapping) sites in this region and that the DnaK binding site is exposed in the wild-type conformation of p53, while that for Hsc70 is at least partially buried (possibly by the interaction between the C terminus and the core domain).
Any of the potential Hsc70 binding sites within the p53 molecule could be responsible for the observed interactions between the two full-length proteins. Our analysis of the binding of Hsc70 to deletion mutants of p53 suggests that sites within the C-terminal 150 residues are not required for the interaction, but does not rule out the possibility that these sites are available and recognized in other p53 mutants, such as those with amino acid substitutions within the core domain. Indeed it is possible that in different point mutants distinct portions of the core domain are opened up, so that a different population of potential Hsc70 binding sites becomes exposed. What presumably is important is that potential Hsc70 binding sites are not exposed on the wild-type p53 molecule, since complex formation with the chaperone would probably interfere with the normal functioning of p53. Our results suggest that this has been achieved during evolution by excluding high affinity Hsc70 binding motifs from those sequences in p53 (the N-terminal domain and large segments of the C-terminal domain) that do not take up a defined tertiary structure in the wild-type molecule and by localizing the potential binding motifs within secondary structural elements that are cryptic in the properly folded molecule. p53 would not normally encounter bacterial chaperones, and thus it would not matter that DnaK recognizes some sites that lie outside the central core of the p53 molecule within the less structured N-and C-terminal domains.
What then is the role of Hsc70 binding in the normal physiology of p53? It is likely that, in common with the majority of cellular polypeptides (72), p53 interacts with Hsc70 and its co-chaperones, including Hsp40 (22), soon after the nascent chain begins to emerge from the ribosome (reviewed in Refs. 17 and 20). Hsc70 would recognize and bind to some or all of the potential sites that we have identified while they are exposed on the prefolded molecule, preventing its intra-and intermolecular aggregation until all segments of the chain necessary for folding are available. Cycles of binding to Hsc70 and controlled release (perhaps to the temporary care of other chaperones, such as Hsp90 or cytosolic chaperonins) would facilitate efficient folding of p53 (20). Interaction of p53 with Hsc70 would cease once all potential binding sites become hidden in the interior of the partially or completely folded protein. It is possible that the wild-type protein never interacts with Hsc70 (or its heat shock-induced cognate, Hsp70) again during its normal life cycle, unless it becomes unfolded during cellular stress. Alternatively, Hsc70 might interact transiently with the mature p53 to facilitate (but not initiate) the conformational changes that modulate the cellular activities of the protein (35). We therefore propose that prolonged binding of p53 to the chaperone, such as that seen with tumor-derived mutants, is not a marker of a normal physiological event but rather the reflection of the derangement of the structure of the protein's core.