Inhibition of MDM2 by hsp90 contributes to mutant p53 stabilization.

Stabilization and overexpression are hallmarks of mutant p53 found in nearly 50% of human tumors. Mutations in the conformation-sensitive core domain of p53 often lead to association with molecular chaperones such as hsp70 and hsp90. Inhibition of hsp90 function accelerates mutant p53 degradation. We recently found that expression of p53 core domain mutants inhibits MDM2 degradation, suggesting that mutant p53 can modulate MDM2 functions. In this report, we show that mutant p53 mediates formation of MDM2-p53-hsp90 complexes. Release of MDM2 from the p53-hsp90 complex after DNA damage restores MDM2 but not p53 turnover, whereas dissociation of hsp90 by geldanamycin increases the degradation of both MDM2 and mutant p53. Mutant p53 degradation after hsp90 inhibition requires MDM2 expression. The interaction between MDM2 and hsp90 is disrupted by the 2A10 antibody, which recognizes a site on MDM2 important for binding to alternative reading frame (ARF). Expression of mutant p53 prevents MDM2 from binding ARF and accumulating in the nucleolus in an hsp90-dependent fashion. These results suggest that hsp90 recruited by mutant p53 conceals the ARF-binding site on MDM2 and inhibits its ubiquitin-protein isopeptide ligase function, resulting in the stabilization of both mutant p53 and MDM2.

The p53 tumor suppressor is mutated in about 55% of human tumors. The majority of mutations are single amino acid substitutions in the DNA-binding (core) domain of p53 (1). In normal cells, p53 is present at very low levels due to rapid degradation mediated by MDM2. MDM2 binds to p53 and promotes its ubiquitination by acting as a ubiquitin E3 1 ligase (2)(3)(4). Expression of MDM2 is activated by p53 at the transcription level (5,6). Therefore, MDM2 functions as a negative feedback regulator to maintain p53 at low levels. In response to stress or DNA damage, p53 is stabilized through multiple mechanisms, such as phosphorylation of p53, expression of the MDM2 inhibitor ARF, and inhibition of MDM2 expression (7). ARF binding inhibits the ubiquitin E3 ligase function of MDM2 and sequesters MDM2 into the nucleolus (8 -10).
In general, tumor cells with mutant p53 accumulate p53 to high levels. The inability of mutant p53 to induce sufficient MDM2 expression is an important mechanism that contributes to the stabilization of p53 (11). However, other studies suggest that binding of heat shock protein hsp90, which is a common feature of p53 mutants, may also play a role in the stabilization of mutant p53 (12)(13)(14). Inhibition of p53-hsp90 binding using benzoquinone ansamycin antibiotics (geldanamycin) that bind specifically to the ATP-binding domain of hsp90 can lead to enhanced ubiquitination and degradation of mutant p53 (14,15). Therefore, mutant p53 may be resistant to degradation in part due to binding of hsp90.
Understanding the mechanism of mutant p53 stabilization may have practical significance in addition to explaining a tumor-specific phenomenon. Although the major consequence of p53 mutation is loss of tumor suppressor function, accumulation of high level mutant p53 during tumor development may also have positive effects on cell proliferation. Mutant p53 can inhibit the function of wild type p53, possibly by oligomerization (16). Ectopic expression of mutation p53 in p53-null cell lines can increase the tumorigenic potential and drug resistance (17)(18)(19). Mutant p53 can also activate the c-myc promoter and overcome the mitotic spindle checkpoint in normal human fibroblasts (20,21). These observations suggest that mutant p53 has gain-of-function properties that enhance cell transformation.
hsp90 is an abundant protein important for protecting cells from stress such as high temperature. Additionally, hsp90 regulates many important signaling pathways in the absence of heat shock. It is found in complexes with v-Src, c-Erb2, RAF-1, Wee1, Cdk4, Bcr-Abl, and the glucocorticoid receptor (22). hsp90 inhibitor geldanamycin and radicicol have potent anti-tumor activity (22). These hsp90-binding compounds can promote rapid c-Erb2 and RAF-1 degradation through ubiquitin-dependent proteasomes, resulting in the inhibition of mitogen-activated protein kinase pathway (23)(24)(25). Therefore, hsp90 binding can inhibit the turnover of many important cellular proteins by the ubiquitination pathway through an unknown mechanism.
In cells expressing wild type p53, MDM2 is degraded rapidly with a half-life of about 0.5 h (26). We recently found that MDM2 in tumor cells with mutant p53 is stabilized by interaction with p53. In this study, we show that formation of MDM2-p53-hsp90 ternary complex is important for the stabilization of MDM2 by mutant p53. MDM2 degradation can be accelerated by inhibition of hsp90 and DNA damage, which disrupt the interaction between MDM2 and hsp90 by different mechanisms. These results reveal a novel mechanism of MDM2 and mutant p53 stabilization in pathological conditions and implicate a similar scheme by which hsp90 may cause stabilization of other cellular proteins.
Adenovirus Infections-Recombinant adenovirus-expressing mutant p53 was amplified using 293 cells. The titer of the crude lysate was determined by serial dilution and detection of cytopathic effects on 293 cells in 96-well plates. The ability of the p53 175H mutant to inhibit MDM2 degradation through complex formation was characterized in our previous experiments (27). Cells were infected with diluted crude viral lysate at 100 plaque-forming units/cell. MDM2 protein levels were determined 24 h after addition of the viruses. To determine the half-life of MDM2, 75 g/ml cycloheximide was added to the cultures, and samples were collected at different time points for Western blot.

MDM2 Is Stabilized in Tumor Cells
Expressing Mutant p53-MDM2 expression is activated by wild type p53. It is often assumed that cells with mutant p53 should express very low levels of MDM2 due to loss of p53 function. We directly compared the levels of MDM2 in a panel of tumor cell lines expressing mutant p53, wild type p53, or null for p53. The result shows that several cell lines expressing mutant p53 have significantly higher levels of MDM2 than the p53-null H1299 and Saos2 cells (for example, DLD1 has Ͼ4-fold higher MDM2 than H1299 by quantitative analysis) (Fig. 1A), reaching levels similar to or even higher than two cell lines with wild type p53.
Stable transfection of H1299 cells with p53 175H mutant also resulted in a 4-fold increase of MDM2 level (Fig. 1A). Higher MDM2 levels were also observed in H1299 cells stably transfected with hot spot mutant p53 248W and p53 281G (data not shown). Because p53 mutants do not stimulate MDM2 transcription, the result suggests that MDM2 expression levels remain elevated in cells with mutant p53 due to enhanced translation or stabilization.
The rates of MDM2 degradation in six cell lines were compared after treatment with the protein synthesis inhibitor cycloheximide. The results show that cells with mutant p53 contain significantly stabilized MDM2 compared with MCF7 and HT1080, which have wild type p53 (Fig. 1B). Although the MDM2 in MDA468 cells (p53 273H ) is significantly more stable than MCF7 and HT1080, it is less stable than the MDM2 in other mutant p53 cell lines (Fig. 1B). The p53 273H mutant is also less stable than other p53 mutants in this assay. As discussed below, this difference correlates with the ability to bind hsp90. Wild type p53 in H1080 cells has a biphasic mode of degradation with an initial rapid decrease followed by stabilization of the remaining p53 (Fig. 1B). This may be due to rapid depletion of MDM2 in this cell line following addition of cycloheximide. These results confirm and extend our recent observation that expression of p53 core domain mutants induces stabilization of MDM2 (27), which contributes to moderate accumulation of MDM2 in cells with mutant p53.
MDM2 Forms a Complex with hsp90 by Binding to Mutant p53-Conformational changes of mutant p53 often result in association with chaperones such as hsp70 and hsp90 (12)(13)(14)29). Because MDM2 binds to the N terminus of p53 and hsp90

FIG. 1. Tumor cells with mutant p53 express stabilized MDM2.
A, increased MDM2 levels in cell lines with mutant p53. MDM2 and p53 were detected in the cells by Western blot using monoclonal antibody 3G9 and DO-1. Identical amounts of total protein were loaded for each cell line, and the same filter was used for the detection of both MDM2 and p53. H1299-175H was established by stable transfection of H1299 with the p53 175H mutant. B, stabilized MDM2 in mutant p53 cell lines. The rates of MDM2 and p53 degradation were determined by Western blot after treatment with cycloheximide (CHX) for the indicated times. Identical amounts of total protein were loaded for each time point.
hsp90 Contributes to Mutant p53 Stabilization most likely binds to the conformation-sensitive core domain of p53, it is possible that MDM2 is also sequestered into complexes containing mutant p53 and hsp90. Consistent with this hypothesis, MDM2 immunoprecipitation and Western blot analysis showed that both hsp90 and hsp70 were coprecipitated with MDM2 in cells expressing mutant p53 but not in cells with wild type p53 or null for p53 ( Fig. 2A). Since MDM2 does not interact with hsp90 and hsp70 in the absence of mutant p53, it suggests that the interaction is indirect and is bridged by mutant p53. hsp90 coprecipitated with MDM2 was detectable using an hsp90 ␣-specific antibody (2D9). Due to lack of a ␤-specific antibody, it is not clear whether hsp90␤ also interacts with MDM2.
The coprecipitation of MDM2 with hsp90 in MDA468 cells (p53 273H ) was weaker than that of other cell lines ( Fig. 2A). To determine whether this is due to weaker binding between the p53 273H mutant and hsp90, cell lysate was immunoprecipitated using anti-p53 pAb421 antibody cross-linked to protein A beads (to prevent the Ig heavy chain from interfering with p53 detection). The relative amounts of p53 and the coprecipitated hsp90 were compared by Western blot of the precipitates. The p53 273H mutant coprecipitated less hsp90 compared with other p53 mutants. These results show that the ability of mutant p53 to bind hsp90 correlates with the stability of MDM2 and p53.
hsp90-p53 binding can be disrupted by geldanamycin and radicicol (30). Both compounds bind to the ATP-binding site of hsp90 and cause its dissociation from client protein (31). If MDM2 binding to hsp90 is mediated by mutant p53, the interaction should also be sensitive to inhibition by geldanamycin and radicicol. As shown in Fig. 2C, when cells were pretreated with radicicol, the coprecipitation of hsp90 with MDM2 was significantly inhibited. A similar effect was also observed using geldanamycin (data not shown).
hsp90 Selectively Binds to Mutant p53 Exposing the pAb240 Epitope-Point mutations of p53 in the DNA-binding domain cause disruption of conformation, exposure of an internal epitope recognized by antibody pAb240, and loss of the pAb1620 epitope (32). To confirm the conformational change of mutant p53 in our cell lines and to determine hsp90-binding specificity of different forms of p53, we compared the efficiency of pAb240 and pAb1620 to coprecipitate p53, MDM2, and hsp90. Cell lysate was precipitated with mutant-specific pAb240, wild type-specific pAb1620, and pan-specific pAb1801.
Coprecipitation of hsp90 and MDM2 with p53 was determined by Western blot of the precipitate. The result showed that mutant p53 reacted strongly with pAb240 and weakly with pAb1620, whereas wild type p53 had an opposite reactivity profile (Fig. 3). MDM2 was coprecipitated with both pAb240 and pAb1620-reactive mutant p53, suggesting that the binding is independent of p53 core domain conformation. Interestingly, hsp90 only coprecipitated with pAb240-reactive mutant p53 but not with the small amount of pAb1620-reactive mutant p53. hsp90 also did not bind to wild type p53, even though a small amount of wild type p53 was reactive to pAb240 in this experiment. These results suggest that partially denatured wild type p53 (possibly occurred during cell lysis) is not a good target for hsp90. The point mutations cause additional structural changes that attract hsp90 binding. Inhibition of hsp90 Stimulates MDM2 and Mutant p53 Degradation-hsp90 binding is important for maintaining the stability of many cellular proteins. Disruption of hsp90-mutant p53 binding also leads to destabilization of p53 (13,14). MDM2 functions as a ubiquitin ligase E3 to promote its own degradation and the degradation of p53 (4,33). These observations led us to test the possibility that MDM2 stabilization is due to interaction with hsp90. As expected, treatment of DLD1 cells with geldanamycin and radicicol increased the rates of MDM2 and mutant p53 degradation (Fig. 4A). Treatment of cells expressing wild type p53 did not result in further destabilization of p53 or MDM2 (data not shown). MDM2 degradation induced by radicicol was blocked by MG132 (Fig. 4B), suggesting that it was mediated by proteasomes.
When cell lysate was prepared in the presence of the ubiquitin isopeptidase inhibitor iodoacetamide, hsp90 inhibition caused the appearance of high molecular weight forms of both mutant p53 and MDM2 (Fig. 4B). This is consistent with a previous report (15) of mutant p53 polyubiquitination stimulated by hsp90 inhibitors. The high molecular weight MDM2 was detected using two different MDM2 monoclonal antibodies 3G9 and 4B11, suggesting that they also represent polyubiquitinated MDM2. To confirm this result, DLD1 cells were transiently transfected with an HA-ubiquitin expression plasmid FIG. 2. MDM2 forms a complex with hsp90 in a mutant p53dependent fashion. A, cell lysates were immunoprecipitated with MDM2 antibody 2A9, and the coprecipitated hsp90 and hsp70 were detected by Western blot. The same filter was stripped and probed for MDM2. B, the p53 273H mutant has lower hsp90 binding affinity. Cell lysates were immunoprecipitated using p53 antibody pAb421 crosslinked to protein A beads. Mutant p53 and coprecipitated hsp90 were detected by Western blot using the same filter. C, dissociation of hsp90 from MDM2 by radicicol. Cells were treated with 20 M radicicol for 1 h and immunoprecipitated using MDM2 antibody 2A9 and p53 antibody pAb421. The coprecipitated hsp90 was detected by Western blot .   FIG. 3. hsp90 binds to mutant p53 that exposes the pAb240 epitope. Cell lysates were immunoprecipitated using the indicated p53 antibodies, fractionated on SDS-PAGE, and analyzed for coprecipitation of hsp90 and MDM2 with p53 by Western blot. pAb419, SV40 T antigen-specific control antibody; pAb1801, pan-specific; pAb1620, wt p53-specific; pAb240, mutant p53-specific.
hsp90 Contributes to Mutant p53 Stabilization and treated with geldanamycin. MDM2 was immunoprecipitated by anti-MDM2 antibody and then analyzed by anti-HA Western blot. The level of high molecular weight HA-ubiquitin precipitated by the MDM2 antibody was significantly increased after geldanamycin treatment (Fig. 4B). Therefore, similar to their ability to induce mutant p53 ubiquitination, hsp90 inhibitors also induce ubiquitination of MDM2. These results suggest that dissociation of hsp90 promotes p53 and MDM2 ubiquitination, which in turn leads to their degradation by proteasomes.
Although the ability of hsp90 inhibitors to promote mutant p53 degradation is well established, the factor that mediates mutant p53 ubiquitination and degradation has not been defined. The simplest interpretation of the results described above is that mutant p53 degradation induced by inhibition of hsp90 is mediated by re-activated MDM2. To test this possibility, mouse embryo fibroblasts derived from p53-null (35.8) or p53/MDM2 double-null mice (174.1) were infected with the p53 175H adenovirus. After inhibition of hsp90 by radicicol, mutant p53 became unstable in p53-null cells but remained stable in the p53/MDM2 double-null cells (Fig. 4C). This result suggests that when MDM2 binds to the p53-hsp90 complex, its ability to promote p53 degradation is inhibited by hsp90. Dissociation of hsp90 from mutant p53 reactivates MDM2, resulting in the degradation of mutant p53. (27) show that inhibition of MDM2 degradation by mutant p53 is dependent on complex formation. If MDM2 accumulation is due to mutant p53-mediated contact with hsp90, then inhibition of MDM2-mutant p53 binding should relieve MDM2 from the inhibitory effect of hsp90 and accelerate MDM2 degradation. To test this hypothesis, DLD1 cells expressing mutant p53 were treated with topoisomerase I inhibitor CPT, which can induce phosphorylation of p53 and inhibit MDM2 binding (34). Cycloheximide was added 4 h after treatment with 0.5 M CPT to determine the rate of MDM2 degradation. The result shows that CPT treatment accelerated the turnover of MDM2 (Fig. 5A) but had no effect on the stability of mutant p53 (Fig. 5A). Degradation of MDM2 induced by CPT was blocked by MG132 (Fig. 5A), indicating that it is mediated by proteasomes. A similar result was also observed after ␥-irradiation-induced DNA damage (data not shown).

DNA Damage Induces MDM2 but Not Mutant p53 Degradation-Previous results
To determine how much MDM2 can be eliminated by DNA damage, DLD1 cells were treated with CPT and two other DNA-damaging agents for 16 h. This resulted in a significant decrease in MDM2 level, although p53 level also decreased moderately (Fig. 5B). Therefore, in contrast to the induction of MDM2 by DNA damage in cells with wild type p53, DNAdamaging treatments actually decrease the level of MDM2 by promoting MDM2 degradation. As expected from the ability of DNA damage to inhibit MDM2-wild type p53 binding, CPT treatment also significantly inhibited MDM2-mutant p53 and MDM2-hsp90 coprecipitation but had no effect on p53-hsp90 coprecipitation (Fig. 5C). These results are consistent with the interpretation that MDM2 sequestration into mutant p53-hsp90 complex leads to inhibition of MDM2 turnover.
MDM2 Is Sequestered into Complexes with Mutant p53 in Cells-If MDM2 stabilization is dependent on binding to mutant p53 and hsp90, one would expect that the majority of stable MDM2 should be bound to mutant p53. To test this prediction, DLD1 (mutant p53) and HT1080 (wt p53) cell lysate were subjected to repeated rounds of immunoprecipitation using p53 antibodies, followed by immunoprecipitation with MDM2 antibody to detect the remaining free MDM2. The result shows that in DLD1 cells, most of the MDM2 was depleted by the p53 antibody (Fig. 6), suggesting that most of it was bound to mutant p53. A similar result was also obtained using T47D cells expressing mutant p53 (data not shown). In contrast, the p53 antibody only captured a small fraction of MDM2 in HT1080 (wt p53) cells, and the majority of MDM2 can only be recovered using an MDM2 antibody (Fig. 6). Therefore, in  hsp90 May Interact with the ARF-binding Site of MDM2-Although MDM2 does not bind hsp90 in the absence of mutant p53 (Fig. 2), the two proteins may physically interact after being brought to close proximity by mutant p53. This interaction may inactivate MDM2 by blocking an MDM2 functional domain. Therefore, a series of MDM2 monoclonal antibodies were tested for the ability to coprecipitate hsp90.
We reasoned that if a region of MDM2 is in close contact with hsp90, antibodies that recognize this region should displace hsp90 and thus would not coprecipitate hsp90. Among 14 monoclonal antibodies recognizing different regions of MDM2 (35) (Fig. 7 and data not shown), 1G2 (epitope located between 339 and 383) and 2A10 (2 epitopes: 258 -260 and 393-395 (36)) did not coprecipitate hsp90, although their ability to precipitate MDM2 was similar to other antibodies. Both antibodies were also able to coprecipitate mutant p53 and hsp70 (Fig. 7), suggesting that their effect on hsp90 is specific. The results suggest that the central region of MDM2 may directly interact with hsp90. Alternatively, the conformation of the region may be altered by interaction with hsp90, so that only the MDM2 that is not in complex with hsp90 can be precipitated by these antibodies.
Interestingly, the 2A10 antibody epitope in the central acidic domain of MDM2 (residue 258 -260) is in close proximity to the ARF-binding site (residue 212-244) (8). 2A10 inhibits the binding between MDM2 and an ARF N-terminal peptide in vitro (8). This suggests that the interaction between hsp90 and MDM2 may also prevent MDM2-ARF binding. This was further tested in the next experiment.
Mutant p53 Blocks MDM2-ARF Binding and Nucleolar Localization-MDM2 interaction with ARF results in targeting to the nucleolus (10). If hsp90 blocks the ARF-binding domain on MDM2, it may prevent ARF binding and affect the nucleolar localization of MDM2. Therefore, the effect of mutant p53 on MDM2 localization was examined. Endogenous MDM2 in p53null H1299 cells was found to localize to the nucleoli (data not shown), possibly due to high level ARF expression in this cell line (10,37). Our previous experiment showed that stable transfection of p53 175H mutant caused a diffused nuclear accumulation of endogenous MDM2 (27), suggesting that mutant p53 expression alters MDM2 localization. The nucleolar staining pattern was maintained when MDM2 expression was in- Cell lysates were subjected to three consecutive rounds of immunoprecipitation using MDM2 antibody 3G9 or p53 antibody pAb421. The pre-cleared lysate was then precipitated by pAb421 or 3G9, respectively. The MDM2 captured in each step was detected by Western blot.
FIG. 7. Two MDM2 antibodies interfere with hsp90 binding. A, epitope locations of MDM2 monoclonal antibodies tested for coprecipitation of hsp90. B, 1G2 and 2A10 do not coprecipitate hsp90. DLD1 lysate was immunoprecipitated using the indicated antibodies, and the same filter was blotted for MDM2, hsp90, and hsp70. A duplicate IP sample was analyzed for the coprecipitation of p53 using protein Ahorseradish peroxidase to reduce the overlapping IgG background. pAb419 is a control antibody against SV40 T antigen. C, a graphic interpretation of the result.
hsp90 Contributes to Mutant p53 Stabilization creased by stable transfection of MDM2 plasmid alone (Fig. 8,  top panel). When MDM2 was stably cotransfected with the p53 281G mutant, a diffused nuclear staining was observed in most of the MDM2-positive cells (Fig. 8, middle panel). Treatment of H1299 cells cotransfected with p53 281G and MDM2 by geldanamycin for 4 h restored MDM2 localization in the nucleolus (Fig. 8, bottom panel) but had no effect on p53 localization (which remained diffusely nucleoplasmic, data not shown). In additional experiments, we found that endogenous MDM2 in C33A cells (p53 273C ) can also be induced to redistribute to the nucleolus by geldanamycin or DNA-damaging treatments ( Fig.  9A and data not shown). Therefore, mutant p53 can block nucleolar accumulation of MDM2, which is dependent on the recruitment of hsp90.
If hsp90 contacts the ARF-binding sites on MDM2, it may prevent MDM2-ARF interaction. To test this possibility, MDM2 was precipitated from C33A cells using the 2A9 antibody, and ARF coprecipitation was determined by Western blot. The result showed that ARF does not coprecipitate with MDM2 in C33A lysate (Fig. 9B). However, when cells were pretreated for 4 h with geldanamycin, efficient MDM2-ARF coprecipitation was detected. This correlated with inhibition of hsp90-MDM2 binding (Fig. 9B). Therefore, hsp90-MDM2 interaction results in the exclusion of ARF, suggesting that hsp90 conceals the ARF-binding site on MDM2. DISCUSSION Accumulation of stable mutant p53 is a prominent feature of many tumors. The major factor responsible for degradation of wild type p53 is MDM2. Reduced MDM2 transcription due to p53 mutation may contribute to mutant p53 stabilization (11). However, a direct comparison of MDM2 levels in multiple tu-mor cell lines shows that cells with mutant p53 are not totally devoid of MDM2 expression. In fact, the steady-state levels of MDM2 are often higher in cells with mutant p53 than in certain p53-null cells and are comparable to some cells with wild type p53. MDM2 expression can be induced by p53-independent mechanisms such as p53 homologs p63 and p73, acti- hsp90 Contributes to Mutant p53 Stabilization vated ras, and growth factors (39 -41). Therefore, mutant p53 may be intrinsically resistant to MDM2-mediated degradation in order to accumulate in the presence of basal level MDM2. Alternatively, MDM2 function may be compromised after binding to mutant p53.
Several studies showed that the ability of MDM2 to promote p53 turnover is tightly linked to its own instability. Deletion of the central domains or the C-terminal RING finger of MDM2 not only destroys its ability to degrade p53 but also causes stabilization of MDM2 itself (3). Point mutations in the RING finger that inactivate the E3 ligase function also block the self-ubiquitination of MDM2 (33). Previous reports (12)(13)(14) showed that inhibition of hsp90 binding leads to destabilization of mutant p53. We recently found that expression of mutant p53 causes stabilization of MDM2 by formation of p53-MDM2 complex (27). These observations suggest a connection between mutant p53 stabilization and MDM2 inactivation. The results described in this study confirm the role of hsp90-binding in mutant p53 stabilization and suggest that inhibition of MDM2 by hsp90 is important for stabilization of mutant p53.
Our results are consistent with a model in which mutant p53 binds to both hsp90 and MDM2, and hsp90 then inhibits the ubiquitin ligase activity of MDM2, blocking the ubiquitination of both MDM2 and mutant p53 (Fig. 10). Dissociation of hsp90 by geldanamycin activates MDM2, leading to ubiquitination and degradation of both MDM2 and mutant p53. Inhibition of mutant p53-MDM2 binding by DNA damage releases MDM2 from the hsp90-p53 complex, allowing MDM2 to promote degradation of itself but not p53. Our results do not rule out the possibility that geldanamycin first induces mutant p53 degradation by another factor, which removes its protective effect on MDM2. However, the inability of hsp90 inhibitors to accelerate mutant p53 degradation in MDM2-null cells favors the interpretation that hsp90 directly blocks the ability of MDM2 to degrade mutant p53. This mechanism complements the current model based on loss of MDM2 feedback and explains the ability of mutant p53 to tolerate various levels of basal MDM2 expression.
It should be noted that the ability of hsp90 to block mutant p53 degradation is not absolute and can be overcome by high levels of MDM2. Cotransfection experiments showed that mutant p53 was degraded by MDM2 when the MDM2 expression plasmid was in significant excess to p53 (42). High level MDM2 expression obtained by transfection of MDM2 cDNA (2) or induction by the p53 homolog p73 2 can also reduce endogenous mutant p53 level in tumor cell lines. In Li-Fraumeni patients carrying wild type and mutant p53 alleles, mutant p53 does not accumulate, presumably due to sufficient amounts of MDM2 induced by the remaining wild type p53 allele. However, cotransfection of equal amounts of mutant p53 and MDM2 plasmid resulted in stabilization of MDM2 and lack of mutant p53 degradation (27). Therefore, the protective effect of hsp90 may only be obvious when mutant p53/MDM2 ratio exceeds a certain threshold.
Although the mechanism by which hsp90 inhibits MDM2 is still unclear, the requirement for complex formation suggests that hsp90 may directly contact MDM2 to block its ability to promote p53 and self-ubiquitination. The inability of the 1G2 and 2A10 antibodies to coprecipitate hsp90 provides indirect evidence of a close contact between hsp90 and a central domain of MDM2. Interestingly, 2A10 binds to a site on MDM2 that is important for interaction with ARF, which is an inhibitor of MDM2 ubiquitin ligase function (8). Microinjection of 2A10 into cells activates wild type p53, presumably due to inactivation of MDM2 (8). Overexpression of ARF can cause MDM2 and p53 stabilization. Therefore, hsp90 may interact with the ARFbinding site on MDM2 and inactivate MDM2 by a mechanism similar to ARF. This is consistent with the ability of mutant p53 to block MDM2 accumulation in the nucleolus and coprecipitation with ARF, which is dependent on recruitment of hsp90.
hsp90 binding prevents degradation of many signaling proteins in the absence of stress (22). Its role in inhibiting MDM2mediated degradation of mutant p53 suggests that hsp90 may stabilize client proteins by inhibiting their E3 ligases. The RING finger protein CHIP has been shown to associate with hsp70 and plays a role in targeting the cystic fibrosis transmembrane-conductance regulator and glucocorticoid receptor for ubiquitination and degradation, possibly by acting as a ubiquitin ligase (38,43). Unlike MDM2, which is highly specific in targeting p53, CHIP may act as a more general E3 ligase that target a variety of misfolded proteins recognized by hsp70. It will be interesting to determine whether hsp90 regulates the function of CHIP similar to MDM2.