Functional inactivation of endogenous MDM2 and CHIP by HSP90 causes aberrant stabilization of mutant p53 in human cancer cells

Abstract

The tight control of wild-type (wt) p53 by mainly MDM2 in normal cells is permanently lost in tumors harboring mutant p53 (mutp53), which exhibit dramatic constitutive p53 hyperstabilization that far exceeds that of wtp53 tumors. Importantly, mutp53 hyperstabilization is critical for mutp53′s oncogenic gain-of function in vivo. Current insight into the mechanism of this dysregulation is fragmentary and largely derived from ectopically constructed cell systems. Importantly, mutp53 knockin mice established that normal mutp53 tissues have sufficient enzymatic reserves in MDM2 and other E3 ligases to maintain full control of mutp53.

We find that in human cancer cells endogenous mutant p53, despite its ability to interact with MDM2, suffers from a profound lack of ubiquitination as the root of its degradation defect. In contrast to wtp53, the many mutp53 proteins which are conformationally aberrant are engaged in complexes with the HSP90 chaperone machinery to prevent its aggregation. In contrast to wtp53 cancer cells, we show that in mutp53 cancer cells this HSP90 interaction blocks the endogenous MDM2 and CHIP E3 ligase activity. Interference with HSP90 either by RNAi against HSF1, the transcriptional regulator of the HSP90 pathway, or by direct knockdown of Hsp90 protein or by pharmacological inhibition of Hsp90 activity with 17AAG destroys the complex, liberates mutp53 and reactivates endogenous MDM2 and CHIP to degrade mutp53. Of note, 17AAG induces a stronger viability loss in mutp53 than in wtp53 cancer cells. Our data supports the rationale that suppression of mutp53 levels in vivo in established cancers might achieve clinically significant effects.

Introduction

Missense mutations in the p53 gene occur in over 50% of human cancers. In normal unstressed cells the level of wild-type (wt) p53 protein is very low due to rapid turnover by its main physiologic E3 ligase MDM2, which is only interrupted when needed in response to stress. This tight control is permanently lost in tumors harboring mutant p53 (mutp53), which exhibit a dramatic constitutive p53 stabilization compared to wtp53 tumors. Mutp53 hyperstability was always assumed to be entirely due to the loss of p53-mediated transactivation of MDM2, itself a wtp53 target gene that forms an autoregulatory loop (1). Surprisingly, however, recently generated knockin (KI) mice expressing mutant p53 R172H in all tissues clearly establish that mutp53 is inherently unstable in normal cells. Despite mutp53’s impairment to transcriptionally induce MDM2, only tumors but not normal tissues of these mice display constitutive stabilization of mutp53 (2–4). Expression of murine and human MDM2 is controlled by two different promoters: the constitutive p53-independent P1 promoter and the p53-responsive P2 promoter (5). Thus, in these KI mice the constitutive p53-independent transcription of MDM2 from the P1 promoter alone is apparently sufficient to degrade mutp53 in normal tissues. This finding eliminates the notion that mutp53’s transcriptional inability to induce sufficient levels of MDM2 is the sole or even primary cause for mutp53 hyperstability. Rather, upon malignant conversion some undefined additional alteration(s) must occur that stabilize mutp53.

Compared to p53 null mice, mutp53 KI mice show an oncogenic gain-of function (GOF) phenotype (2, 3). In agreement, depletion of mutp53 by siRNA in human tumor cells leads to suppressed tumor growth in culture and in xenografts, and to enhanced chemosensitivity (6, 7). Importantly, mutp53 hyperstabilization is critical for manifestation of mutp53’s GOF in vivo. In support, constitutive MDM2 deficiency in p53 R172H/R172H mice (in short p53H/H mice) causes earlier tumor onset, increased tumor incidence and metastasis, and shortened survival compared to MDM2-proficient p53H/H mice, implying that GOF depends on mutp53 levels (4). Thus, tumor-specific stabilization of mutp53 is a critical determinant of mutp53’s GOF.

Little conclusive insight currently exists about the precise mechanisms responsible for dysregulating mutant p53 protein levels in cancer cells. In fact, this important question constitutes a major unexplored area in the p53 field, with the exciting prospect that advances have high translational potential that might be exploited for a mutp53-directed cancer strategy. The existing studies provide only fragmentary insights mostly derived from ectopically overexpressed mutp53, analyzed in constructed non-physiologic cell systems of wtp53 or null p53 background. In contrast to wtp53, many mutp53 protein species are conformationally aberrant. To prevent aggregation, this dictates their engagement in stable complexes with heat shock proteins Hsp90 and Hsp70, members of the HSP90 chaperone machinery that is upregulated and activated in cancer (8, 9, 10). By analyzing the endogenous status, we show here that in human cancer cells harboring mutp53, this HSP90 interaction, while on the one hand stabilizing the mutp53 conformation, on the other hand blocks its degradation by inhibiting constitutive MDM2 and CHIP E3 ligase activity. Interference with the HSP90 pathway, using either RNAi against the upstream regulator or against Hsp90, or with a pharmacological Hsp90 inhibitor, destroys the complex, liberates mutp53 and reactivates endogenous MDM2 and CHIP for mutp53 degradation

Material and Methods

Human cancer cells

Cancer cell lines MCF7 (breast), RKO and HCT 116 (colon), U2OS (osteosarcoma), as well as immortalized MCF10A (breast) and MRC5 (diploid fibroblasts) contain functional wtp53. Conversely, breast cancer MDA 231(R280K), MDA468 (R273H), T47D (L194F) and SKBR3 (R175H), prostate cancer DU145 (P223L, V274F), pancreatic cancer PANC1, bladder cancer 5637 (R280T) and ovarian cancer EB2 cell lines all harbor mutp53. Stable mutant SW480 (p53 R273H/P309S) inducibly express shp53 under the control of a tetracycline-regulated promoter upon adding Tetracyclin in culture (1.0 mg/mL) or feeding it to nude mice (2 grams per liter in drinking water) (7). Stable MDA231-Luc (control) and MDA231-shp53 cells were a gift from Dr. S. Deb. Cells were cultured in 10% FCS/DMEM. Where indicated, cells were treated with 25 μM ALLN (Calbiochem) for 3h. CHX (Sigma) was added to the medium (final 50 μg/ml). UbAL (BioMol Interntl), a specific inhibitor of deubiquitinases, was included in all buffers. Treatment with 5 μM 17AAG (17-allylamino-17-demethoxygeldanamycin, LC Laboratories) was for 24h, 5 μM Camptothecin for 3h and Nutlin (20 μM, Sigma) for 24h. Cell viability was determined by CellTiter-Blue assay (Promega) in a 96-well format (10,000 cells/well, seeded 24h prior). Proliferation was measured by cell counts. Nude mice were injected subcutaneously with MDA231- or SW480 shp53 cells or vector controls (10⁶ cells per injection site, 6 sites per mouse). Tumors were harvested after 12 or 20 days.

Plasmids

CMV-MDM2 plasmid carrying a Neomycin resistance gene (11) was transfected with lipofectamine (Invitrogen). Stably transfected clones were selected in 700 μg/ml G418 (Gibco). The p53 R280K plasmid was subcloned into a retroviral REBNA puro vector. Phoenix A cells were transfected with mut p53R280K REBNA or empty vector. After 48h, supernatants containing the retroviral particles were collected and used to infect T47D cells overnight, followed by puromycin (1 μg/ml) selection 48h later.

RNA Interference

Pools of 4 different siRNA duplexes specific for human MDM2 (Ambion) and CHIP (Dharmacon) were transfected with Lipofectamine 2000. Cells were harvested 48h later and analyzed. For Hsp90 silencing, 5637 and MDA231 cells were transfected with 10 pmol of SilencerR Select siRNAs (Ambion) and analyzed 3 days later.

Immunoblots, immunoprecipitations

For immunoblots, equal total protein of crude cell lysates (2.5–5 μg) was loaded. When loading was normalized for equal amounts of non-ubiquitinated p53, a first quantitation immunoblot was run prior to the second definitive immunoblot. For nuclear/cytoplasmic fractionations the Pierce kit was used. Antibodies were FL393 and DO1 for p53, SMP14 for MDM2 (Santa Cruz), HAUSP (Calbiochem), p14Arf ( Abcam), p53 Ser15 (Cell Signaling), MDMX (Bethyl Lab), Hsp70, Hsp90 and histone deacetylase HDAC (all Affinity Bioreagents), PCNA, tubulin, actin, rabbit IgG (all Sigma). For detecting endogenous complexes, crude lysates were immunoprecipitated with 1 μg antibody for 2h. Beads were washed 3 times with SNNTE plus 2 × RIPA (50 mM Tris, 150 mM NaCl, 1% Triton X-100, 0.1% SDS, 1% NaDeoxycholate, pH 7.4) prior to immunoblotting. Immunofluorescence was performed as described (12).

Results and Discussion

Tumor-derived endogenous mutp53 shows complete lack of ubiquitination, causing its profound degradation defect

Although stabilization of mutant p53 was noted previously, the ubiquitination status of endogenous mutant p53 remains controversial. While early reports noted higher stability of mutp53 (13), recent studies using ectopic expression suggest that mutp53 is more ubiquitinated than wtp53 in cancer cells (4,14, 15). Moreover, until now, studies on the regulation of mutp53 stability were mostly limited to genetic analysis of KI mice (2–4) or ectopic overexpression of mutp53 in tumor cells (1, 13, 14). Thus, to characterize the degradation of endogenous mutp53, we probed a panel of randomly chosen human cancer cell lines expressing either wtp53 or mutp53. Mutp53 tumor cell lines exhibit a dramatic constitutive p53 stabilization ranging from 10–20 fold above wtp53 cancer lines (Suppl. Fig.1A). To compare their p53 ubiquitination status side-by-side, immunoblots from total cell lysates were normalized for comparable amounts of non-ubiquitinated p53. While wtp53 ubiquitination was readily detected, mutp53 ubiquitination remained undetectable in all lines, even after prolonged exposure (Fig. 1A). Moreover, proteasome inhibition by ALLN treatment led to marked accumulation of ubiquitinated p53 only in wild-type cancer cells, while mutp53 remained non-ubiquitinated (Figs. 1B and 1C right). Putative mutations in the RING domain of MDM2 were excluded in all six mutp53 lines, eliminating mutational inactivation of MDM2 as possible explanation. Ubiquitinated wtp53 mainly localizes to the cytoplasm, while the nucleus preferentially harbors non-ubiquitinated p53 (12, 16) (Fig. 1C left). Mutp53 accumulates mainly in the nucleus. To further exclude that the dramatic accumulation of non-ubiquitinated mutp53 in the nucleus might mask a putative ubiquitinated pool in the cytoplasm, we performed fractionations. Again, ubiquitinated wtp53 was mainly located in the cytoplasm where it was further stabilized by ALLN (Fig. 1C left) (12). In contrast, cytoplasm and nucleus were completely devoid of ubiquitinated mutp53, and remained unresponsive to ALLN (Fig. 1C right). Taken together, this indicates that the severely impaired degradation of mutp53 is due to grossly defective ubiquitination by MDM2 and possibly other related E3 ligases, thereby causing its aberrant stabilization.

Figure 1. Tumor-derived mutant p53 proteins show a complete lack of ubiquitination, causing their profound degradation defect.

A. Ubiquitination of endogenous p53 in a panel of wild-type and mutant human cancer cell lines. p53 loading was normalized for comparable amounts of non-ubiquitinated p53. DO1 immunoblot.

B. Proteasome inhibition (by ALLN) results in accumulation of ubiquitinated p53 only in wtp53 cells, while mutp53 proteins remain completely non-ubiquitinated. Normalized loading and blotting as in A.

C. Left, Ubiquitinated wtp53 is mainly located in the cytoplasm. In contrast, both cytoplasm and nucleus are completely devoid of ubiquitinated mutp53. Right, Ubiquitinated wtp53 resides mainly in the cytoplasm and is stabilized by ALLN-mediated proteasome inhibition. In contrast, mutp53 is non-ubiquitinated in both compartments and remains so even after ALLN treatment. Immunoblots normalized for p53 loading. HDAC and Hsp90 as nuclear and cytoplasmic markers.

Neither HAUSP or p14Arf expression nor p53 modification can explain mutant p53 hyperstability

To gain more insight into the mechanism of mutp53 stabilization in human cancer, we analyzed MDM2, HAUSP and p14Arf, key molecules in the regulation of p53 stability. MDM2 is a classic target gene of wtp53. As expected, due to the lack of transcriptional activity of mutp53, levels of MDM2 mRNA and protein were downregulated in mutp53 cancer cell lines compared to wtp53 cancer lines, but only by about 2–3 fold (Fig. 2A and Suppl. Fig. 1B). Importantly and parallel to normal tissues of mutp53 knockin mice which express wt-like low levels of mutp53 (4), although downregulated, all mutant human cancer lines express MDM2 protein constitutively. This renders the complete lack of mutp53 ubiquitination disproportionally severe and leaves it unexplained.

Figure 2. Neither HAUSP or p14Arf expression nor p53 modification can explain mutant p53 hyperstability.

A. All mutp53 cell lines have lower (about 2-fold) but readily detectable levels of MDM2.

B. Top, HAUSP expression is similar in wild-type and mutp53 lines. Bottom, p14^ARF is undetectable in all wtp53 cell lines and upregulated in some but not all mutp53 cell lines.

C. Constitutive DNA damage does not contribute to mutp53 stabilization. Ser15 phosphorylation is undetectable in the majority of mutp53 cell lines.

D. Mutp53 cancer cells express lower amounts of acetylated p53 than wtp53 cells. Loading normalized for comparable amounts of non-ubiquitinated p53 and blotted with p53 Acetyl 373/382 antibody. A–D Immunoblots, actin, tubulin and PCNA as loading controls.

Since hyperstability of mutp53 might not only be caused by reduced ubiquitination but also by increased deubiquitination, we analyzed HAUSP, the major p53 deubiquitinase. However, levels of HAUSP and its interaction with p53 were similar in wt and mutp53 cells and thus did not contribute to mutp53 hyperstability (Fig. 2B and Suppl. Fig. 1C).

p14Arf protein levels, normally undetectable, are elevated in cells with a perturbed p53 signaling axis (17, 18), raising the possibility that the MDM2 antagonist Arf could contribute to hyperstability of mutp53 (4). We therefore examined Arf levels in mutant and wtp53 cells. As expected, Arf was undetectable in all wtp53 cells (Fig. 2B bottom). Although Arf was upregulated in mutant EB2, MDA435 and DU145 cells, it remained undetectable in T47D, MDA231 and SK-BR3 cells. This lack of correlation between Arf expression and mutp53 status indicates that Arf upregulation is not a generic mechanism for hyperstabilization of mutp53, although it might contribute in individual cancers when deregulated.

Alternatively, it was suggested that constitutively activated DNA damage signaling in mutp53-harboring tumors could account for stabilization of mutp53 via chronic Ser15 phosphorylation (19). We therefore assessed Ser15 status in unstressed cell panels. However, the majority of mutp53 lines and all wtp53 lines lacked constitutive Ser15 phosphorylation. Although two of the six mutant lines did show variably increased phosphorylation (Fig. 2C), genotoxic stress (Camptothecin) induced Ser15 phosphorylation similarly in both wtp53 and mutp53 cells (Suppl. Fig. 2A). Thus, this modification does not contribute to generic mutp53 stabilization. Acetylation of the C-terminal lysines of wtp53 via competitive inhibition of their ubiquitination was proposed as an important mechanism for wtp53 stabilization in response to DNA damage. Moreover, acetylation of wtp53 interferes with its interaction with MDM2 (20). To test whether putative hyperacetylation of mutp53 may cause its ubiquitination defect, we analyzed the acetylation status of normalized amounts of mutant and wtp53. However, mutp53 cells generally express lower amounts of acetylated p53 than wtp53 cells (Fig. 2D), excluding hyperacetylation as contributor to the specific hyperstability of mutp53.

Selective impairment of MDM2 E3 ligase activity in mutp53 but not wtp53 cancer cells

Consistent with the lack of ubiquitination, the half-life of mutp53 is dramatically increased compared to wtp53 (Fig. 3A). Moreover, other bona fide substrates of MDM2, i.e. MDMX and MDM2 itself, are also more stable in mutant compared to wtp53 cancer cells (Fig. 3A) and are insensitive to proteasome inhibition in mutant but not in wt p53 cancer cells (Fig. 3B). Thus, major physiologic substrates of MDM2 exhibit degradation deficiencies in mutp53 cells. During stress, DNA-damage induces auto-ubiquitination and self-degradation of MDM2 as part of the stabilization mechanism of wtp53 (21). However, while camptothecin destabilized MDM2 in wtp53 cells, this was not the case in mutp53 cancer cells, again supporting their selectively impaired MDM2 activity (Fig. 3C). Of note, mutp53 is fully competent for binding to MDM2. We did not observe dramatic differences in the physical interaction between endogenous mutp53 and MDM2, as reported earlier (13, 14). The amounts of co-precipitated p53 simply reflected the respective steady state levels (Suppl. Fig. 2B). In sum, this strongly suggests that functional impairment of endogenous MDM2 is a major factor responsible for the aberrant stabilization of mutp53 in cancer cells.

Figure 3. Selective impairment of MDM2 E3 ligase activity in mutp53 cancer cells.

A. Bona fide physiologic substrates of MDM2 (p53, MDMX and MDM2 itself) exhibit degradation deficiencies in mutp53 (MDA468) compared to wtp53 (HCT116) cells. Cycloheximide (CHX) chase for the indicated times. Actin, loading control. * denotes a nonspecific band.

B. Non-responsiveness of MDM2 and MDMX to proteasome inhibition and DNA damage in mutp53 cancer cells indicates selective impairment of the MDM2 E3 ligase activity in mutp53 cancer cells. Left, MDM2 activity is selectively impaired in mutp53 cancer cells. While wtp53 cells stabilized MDM2 more than 3-fold upon ALLN, no MDM2 stabilization occurred in mutp53 cells. Right, Corresponding densitometry quantitation of relative MDM2 levels is shown. Bottom, In contrast to wtp53 cells, MDMX levels failed to stabilize in mutp53 cells in response to ALLN. Cells treated with 25 μM ALLN for 12 h followed by immunoblotting.

C. DNA damage destabilizes MDM2 due to autoubiquitination in wtp53 but not in mutp53 cancer cells. Camptothecin (5 μM) for 2h. Immunoblot.

D. Normal mutp53 harboring p53H/H MEFs (which express somewhat higher constitutive levels in culture than wtp53 MEFs), properly stabilize mutp53 in response to genotoxic stress and proteasome inhibition. Immunoblots of cells treated with 5 μM Camptothecin or 25 μM ALLN for 6h.

E. Endogenous mutant p53 is readily degraded by overexpressed levels of ectopic MDM2. Top, immunoblots after transient transfection of MDM2 with and without 25 μM ALLN for 6h. Bottom, immunofluorescence of mutp53 MDA468 cells transfected with MDM2 (red). The transfected cell does not exhibit p53 staining (green). A–E Immunoblots, actin and tubulin as loading controls.

On the other hand, normal H/H mouse embryo fibroblasts harboring the R172H mutation of p53 (H/H MEFs) properly stabilize mutp53 in response to genotoxic stress and proteasome inhibition, similar to wtp53-harboring MEFs (Fig. 3D). Likewise, mutp53 also properly stabilizes upon irradiation in normal spleen and thymus of H/H mice (4). Together, this confirms that in normal cells MDM2 retains its ability to control mutp53 stability, while this regulation is lost once cells become transformed.

Of note, supraphysiologic levels of ectopic MDM2 readily degraded endogenous mutp53 in cancer cells (Fig. 3E, lanes 7 and 8). Conversely, proteasome inhibition by ALLN blocked ectopic MDM2-mediated p53 degradation similarly in mutant and wild-type cells (Fig. 3E, compare lanes 1, 2 with 5, 6), confirming earlier reports (13, 14). This also reaffirms that the defect that causes mutp53 hyperstability lies with blocked endogenous MDM2 activity and not with its substrate.

To further test the idea that mutp53-harboring cancer cells suffer from a selective inhibition of their MDM2 activity, we forcibly equilibrated MDM2 levels in mutant and wtp53 tumor cells. First, MDM2 levels in wtp53 breast cancer cells (MCF7) were downregulated by siRNA to match those of mutp53 breast cancer cells (MDA231 and MDA468) (Fig. 4A). If MDM2 levels were the sole determinant as was previously assumed, one would now expect wtp53 to hyperstabilize to levels matching those of mutp53. Surprisingly, however, wtp53 stabilized by less than 2-fold in MCF7 cells, far below the ~20 fold constitutive stabilization of mutp53 levels in MDA231 and MDA468 cells (Fig. 4A). Similar results were obtained for other wtp53 cells (RKO and HCT116, Suppl. Fig. 1D). Conversely, we corrected the lower MDM2 levels in mutp53-harboring MDA231 back to those of wtp53-harboring MCF7 cells by generating stable MDM2 clones that express about 2-fold higher MDM2 levels (Fig. 4B). (MDA231 cells do not have elevated p14Arf as shown in Fig. 2B, thereby excluding a possible negative effect on MDM2 activity). However, in all successfully established MDA231 clones mutp53 levels remained unaffected and ubiquitination non-detectable, even after challenge with ALLN (Fig. 4B). This is despite the fact that ectopic MDM2 undergoes effective complex formation with endogenous mutp53 (Fig. 4C). Likewise, ectopic MDM2 and endogenous MDMX again display (self)-degradation defects in all mutp53 MDM2 clones, judged by the poor (for MDM2) or absent (for MDMX) stabilization after ALLN treatment, in contrast to wtp53 MCF7 cells (Fig. 4B). Thus, in contrast to markedly supraphysiologic MDM2 expression (Fig. 3E), physiologic levels of overexpressed MDM2 in mutp53 cancer cells again are functionally inhibited. This suggests that a saturatable cellular mechanism leads to MDM2 inactivation in mutp53 cancer cells.

Figure 4. MDM2 activity is functionally impaired in mutant p53 cancer cells.

A, B. Decreased levels of MDM2 are not the main determinant of mutp53 stability in cancer cells. A. MDM2 levels in wtp53 breast cancer cells (MCF7) were downregulated by siRNA to match those of mutp53 cells (MDA231, MDA468). However, knockdown of MDM2 in MCF7 cells elevates wtp53 levels only by less than 2-fold, and does not reach the 20-fold stabilization present in mutp53 cells.

B. Conversely, the lower MDM2 levels in mutp53 MDA231 cells were corrected back to those of wtp53 MCF7 cells. Stable MDM2 clones express 2- fold higher MDM2 levels. However, in all established MDA231 clones, mutp53 levels remained unaffected and ubiquitination non-detectable, even after challenge with ALLN (3 independent clones of 7 shown). Cells treated with 25 μM ALLN for 12h.MDM2 and MDMX display the same (self)-degradation defect as seen in Fig. 3, in contrast to wtp53 MCF7 cells. Bottom, Densitometry quantitation of MDM2 levels of above blot in the absence or presence of ALLN by densitometry of long and short exposures, respectively. Tubulin, loading control.

C. Ectopic MDM2 undergoes effective complex formation with endogenous mutp53. MDM2 immunoprecipitation from MDA231-MDM2 clone 21 or wtp53 MCF7 cells, followed by immunoblots. PCNA, loading control.

Tumor-specific stabilization of mutp53 is caused in part by the HSP90 molecular chaperone machinery

Guarding the proteome against misfolding, aggregation and illicit interactions induced by proteotoxic stress such as reactive oxygen species, hypoxia and acidosis, the heat shock family of molecular chaperones guide proper conformational folding of nascent polypeptide ‘clients’ into mature proteins, assist in the productive assembly of multimeric protein complexes and regulate the cellular levels of their clients by promoting degradation. Normal chaperone function is subverted during oncogenesis to allow initiation and maintenance of malignant transformation and enable cancer cell survival since cancer cells are in a constant state of proteotoxic stress, both from an adverse microenvironment (hypoxia, acidosis) as well as from within (conformationally aberrant oncoproteins, high levels of ROS, spontaneous DNA damage, aneuploidy). Thus, their proteins and in particular their oncoproteins require massive chaperone support to prevent aggregation and promote survival (23). Hence, in addition to their oncogene addiction, cancer cells also show addiction to heat shock proteins. Among chaperones, heat shock protein 90 (Hsp90) is unique because many of its clients are conformationally labile signal transducers with crucial roles in growth control, cell survival and development. Most importantly, HSP90 plays a key role in the conformational stabilization and maturation of mutant oncogenic signaling proteins. These encompass steroid hormone receptors, receptor tyrosine kinases (i.e. HER-2), signaling kinases (Bcr-Abl, Akt, Raf-1), and mutant p53 (8, 23). Hsp90 is the core protein of the multi-component chaperone machinery HSP90 (that includes Hsp70 and others), a powerful anti-apoptotic system that is highly upregulated and activated in cancer. Hsp90 is a dynamic ATPase. ATP binding to the N-terminal domain of Hsp90 and its subsequent hydrolysis drives a conformational cycle that is essential for the HSP90 chaperone activity. Co-chaperone Hsp40 stimulates the associated Hsp70 ATPase activity; chaperone Hsp70 helps fold nascent polypeptides and adaptor Hop mediates interaction of Hsp90 with Hsp70 (7). Importantly, upregulation of HSPs and in particular Hsp90 is an almost ubiquitous feature of human cancers (23). Moreover, structural and affinity differences exist, as revealed by the fact that Hsp90 purified from tumor cells has a 100-fold stronger binding affinity to small molecule ligands of its ATP-binding pocket than does Hsp90 protein purified from normal cells. Of note, tumor Hsp90 is entirely engaged in mulichaperone complexes due to an increased load of mutant clients, whereas normal cell Hsp90 is largely uncomplexed and free (23, 24).

Importantly, many mutp53 proteins are damaged in their conformation-sensitive core domain and form abundant stable complexes with Hsp90 in tumor cells (8, 10). In contrast, wild-type p53 is unable to form stable Hsp90 complexes and does so only transiently and with a few components. For example, the A1-5 fibroblasts expressing the temperature-sensitive p53 A135V mutant showed that the HSP90 components Hsp90, Hsp70, co-chaperone p23 and cyclophilin 40 only co-immunoprecipitate with mutant p53 (at 37°C) but not with wild-type p53 (at 30°C) (8, 9, 10).

This stable mutp53-specific interaction with HSP90 chaperones in cancer cells has been speculated to be linked to mutp53’s aberrant stabilization. In a preliminary immunoprecipitation study, Peng et al presented circumstantial evidence, although no direct proof, that MDM2 might be inactivated by being trapped within a trimeric complex of mutp53-MDM2-Hsp90, and proposed that Hsp90 binding conceals the Arf-binding site on MDM2, thereby somehow inhibiting its ligase function (25). Interpretation of this study, however, was made difficult by the fact that MDM2 was found stabilized in tumor cells with mutant p53 for reasons that are unclear. While we and others find MDM2 downregulated in mutp53 tumor cells, our findings nevertheless fully endorse that mutp53 hyperstability in cancer cells is strongly dependent on heat shock support because it inhibits mutp53 ligases, as described below. HSF1, the master transcriptional regulator of the inducible heat shock response, controls all stress-inducible chaperones including HSP90 (26). HSF1 is frequently upregulated in human tumors and the HSF1-mediated stress response plays a causal, broadly supportive role in mammalian oncogenesis (23, 26). We find that shRNA-mediated knockdown of HSF1 in mutp53 cancer cells, which in turn downregulates Hsp90 and Hsp70 protein, induces rapid destabilization of mutp53 (Fig. 5A and Suppl. Fig. 3) and reduces its half-life (Suppl. Fig 3A). Moreover, this relationship follows a direct dose-response. The stronger the HSF1 knockdown and therefore the Hsp90 and Hsp70 knockdown (obtained by repeat rounds of retroviral HSF1-shRNA infection), the stronger the mutp53 destabilization (Fig. 5A).

Figure 5. Stabilization of mutp53 in cancer cells is caused by the HSP90 chaperone machinerythat inhibits the MDM2 and CHIP E3 ligase activity.

A. HSF1 knockdown, which leads to downregulation of Hsp90 and Hsp70 proteins, induces destabilization of mutp53. The eukaryotic genome contains separate genes encoding constitutively expressed and inducible Hsp90. Only inducible Hsp90 transcription is controlled by HSF1 (23). Note that the Hsp90 antibody recognizes both constitutive and inducible Hsp90, explaining the only partial obliteration of Hsp90 after HSF1 knockdown.

B. Hsp90 interference by17AAG causes release of mutp53 from the Hsp90 complex. Hsp90 (or IgG control) immunoprecipitation (1 mg protein each) from MDA231 cells before and after 17AAG treatment (5 μM for 24 h). HAUSP, loading control.

C. 17AAG induces ubiquitination of mutp53 in a dose-depended manner. Cells were treated with the indicated concentrations of 17AAG for 24h.

D. Left, 17AAG induces degradation of mutant p53 (and MdmX) in a dose dependent manner. 17AAG induces degradation of mutant but not wild-type p53 in cancer cells. Right Cells treated as indicated for 24h; 2 μg of mutp53 and 20 μg of wtp53 cell extracts probed. 17AAG induces degradation of mutp53 but does not decrease Hsp90 levels. Under normal conditions, HSF1 is also a client of HSP90 but held in an inactive HSF1-HSP90 complex. 17AAG inhibits HSP90 chaperone activity and promotes the release of HSF1 for transcriptional activation of heat shock proteins, including Hsp90. This explains the slight increase in Hsp90. HAUSP, loading control.

E. Downregulation of Hsp90 protein by siRNA destabilizes mutp53 in MDA231 and 5637 cancer cells. Scr si, scrambled siRNA. Three days post transfection, cells were harvested and analyzed by immunoblots. Actin, loading control.

F. 17AAG decreases the half-life of mutp53, Mdm2 and MdmX. Cycloheximide chase (CHX) of MDA231 and MDA468. Cells treated with 5 μM 17AAG for 16h, then 17AAG plus 50 μg/ml CHX for the indicated times.

G. Nutlin partially prevents 17AAG-induced destabilization of mutp53, indicating MDM2 reactivation upon 17AAG. PANC1 cells harboring mutp53 were treated with 5 μM 17AAG with or without 10 μM Nutlin for 16h. Immunoblot. Actin, loading control.

H. siRNA-mediated knockdown of MDM2 and CHIP rescue 17AAG-induced destabilization of mutp53.

ATP binding to the N-terminal domain of Hsp90 with subsequent ATP hydrolysis drives a conformational cycle that is essential for HSP90’s client binding and chaperone activity. Specific Hsp90 inhibitors such as the ansamycin antibiotics geldanamycin and 17AAG competitively bind to the N-terminal ATP-binding pocket and stop the chaperone cycle, leading to client protein degradation. 17AAG is a potent and highly specific Hsp90 inhibitor currently in phase I-III clinical trials for refractory multiple myeloma and several solid cancers including breast cancer (27). We find that in all mutp53 human cancer cells tested, 17AAG specifically induces endogenous mutp53 to be released from HSP90 (e.g.Fig. 5B), followed by its efficient ubiquitination (Fig. 5C) and degradation (Fig. 5D) (10, 22). To further support the specificity of the 17AAG data, we examined the effect of Hsp90 levels on mutp53 stability. Indeed, downregulation of Hsp90 protein by siRNA destabilizes mutp53 in MDA231 and 5637 cancer cells (Fig. 5E). Of note, 17AAG does not downregulate Hsp90 levels in mutp53 cells (Fig. 5D). Accordingly, 17AAG markedly shortens the half-life of mutp53 (Fig. 5F). Interestingly, this is not the case for cancer cells with wild-type p53. In fact, 17AAG has opposite effects, downregulating mutp53 while upregulating wtp53 (Fig. 5D). This 17AAG effect on wtp53 is consistent with previous data (28). In primary tumor cells derived from mouse models of medulloblastoma that either retain or are null for wtp53 function, inhibition of Hsp90 by a derivative molecule, 17DMAG, induced wtp53-dependent apoptosis (29).

Inhibition of MDM2 and CHIP by HSP90 is largely responsible for stabilization of mutant p53

In normal cells HSP90 chaperones regulate the protein levels of their clients in part by directly recruiting ubiquitin ligases and presenting them for proteasome-mediated degradation. The chaperone-dependent E3 ligase CHIP (carboxy-terminus of Hsp70-interacting protein) binds to Hsp70 and is a resident part of the HSP90 complex, normally promoting degradation of clients such as glucocorticoid and androgen receptors, c-ErbB2 (30) and phosphorylated tau (31). Importantly, CHIP’s degradative function can become defective in tumors.

As shown above, interference with the HSP90 chaperone function by 17AAG triggers mutp53 degradation by freeing mutp53 from complexation with Hsp90, which apparently enables the reactivation of E3 ligases (Figs. 5B-F). Which endogenous ligase(s) are responsible? Our evidence implicates both MDM2 and CHIP reactivation in 17AAG-mediated degradation. First, 17AAG-reactivated MDM2 leads to self-degradation of MDM2 and its physiological substrate MDMX (Fig. 5F). Moreover, 17AAG-mediated mutp53 destabilization is partially reversed (rescued) by Nutlin (Fig. 5G), or by siRNA-mediated MDM2 or CHIP knockdown (Fig. 5H), and almost completely rescued by synergistic interference with both ligases (combined siRNAs and Nutlin, Fig. 5H, lane 5). In further support, HSF1 knockdown-mediated mutp53 degradation is again partially reversed by Nutlin (Suppl. Fig. 3B) and simultaneous knockdown of either MDM2 or CHIP ligases (Suppl. Fig. 3C). Together, our findings indicate that both MDM2 and CHIP are the major endogenous E3 ligases for mutp53, although CHIP appears to be the more effective one. Both are presumably active in normal cells of p53H/H KI mice harboring mutp53. In cancer cells, mutp53 is trapped in stable interactions with upregulated and activated HSP90 that effectively inhibits MDM2 and CHIP activity, leading to its aberrant stabilization. Ectopic expression studies previously implicated CHIP as an alternative E3 ligase for ubiquitination and degradation of mutp53 (14, 32, 33). Likewise, pharmacological interference with Hsp90 by17AAG reactivates both ligases to degrade mutp53. This is similar to the functional redundancy of chaperone-associated E3 ligases that promote degradation of glucocorticoid and androgen receptors, as revealed after deletion of CHIP (34).

17AAG reduces cell viability more profoundly in mutp53 compared to wtp53 cancer cells

To further test the notion that mutp53 levels are the major determinant of its oncogenic gain-of-function (GOF) and to test the dependence of established tumors on maintaining these high levels, we evaluated the consequences of downregulating mutp53 by i) shp53 and ii) pharmacological destabilization via 17AAG. In strong support that GOF indeed depends on highly stabilized mutp53, we and others consistently find that downregulation of mutp53 by shRNA strongly inhibits the malignant phenotype of human cancer cells in vitro and in vivo (Figs. 6A-D). For example, stable and tetracycline-inducible knockdown of endogenous mutp53 in breast (MDA 231) and colon (SW480) cancer cells by shp53 RNA interference dramatically inhibits cell proliferation (Fig. 6A) and invasion (Fig. 6B) in culture, and strongly inhibits tumor growth in nude mouse xenografts in vivo (Figs. 6C, D). In agreement, mutp53 knockdown in SKBr3, HT29, SW480, MiaPaCa-2 and MDA231 cells also caused strong inhibition in proliferation, clonogenicity and soft-agar assays in vitro (6, 7). It also induced strong chemosensitization towards conventional genotoxic drugs (6, 7) and inhibited metastatic spread in mouse xenografts (35). Collectively, these data imply that tumors are addicted to their high levels of mutp53 and support the rationale that suppression of mutp53 levels in vivo might achieve clinically significant effects, particularly when combined with other anti-cancer therapies.

Figure 6. 17AAG reduces cell viability more profoundly in mutp53 compared to wtp53 cancer cells.

A–D. Knockdown of mutant p53 by shRNA inhibits proliferation and invasion of human cancer cells in vitro and in vivo. shp53-mediated stable knockdown of mutp53 in MDA231 compared to parentals, assayed for proliferation in vitro (A) and in tumor xenografts in nude mice (C, D). C. Average tumors weight (n=12 each) at day 20. Bars represent mean +/−standard error. All tumors show significant reduction of mutp53 levels (examples shown). B. Knockdown of mutp53 in MDA231 cells by Tet-inducible shp53 inhibits their invasion. Matrigel Boyden chambers and immunoblot (bottom). D. Knockdown of mutp53 in SW480 colon cancer cells by Tet-inducible shp53 inhibits their growth in vivo. Tumor xenografts were harvested at day 12. Bottom, p53 immunostaining (DO1) in parental and knockdown xenografts after oral tetracycline fed to nude mice. The shp53 tumor shown corresponds to the single larger tumor in graph above.

E. Knockdown of mutp53 by 17AAG (2 μM) and/or Tetracycline-inducible shp53 RNAi decreases cell viability of mutp53 SW480 cells proportional to the extent of mutp53 destabilization (bottom). CTB assay. Cells treated with Tetracyclin for 24h or left untreated, then concurrently with empty vehicle and/or 2 μM 17AAG for 24h.

F. 17AAG inhibits cell viability more profoundly in mutp53 than in wtp53 cancer cells. CTB assay. Cells treated with 2 or 5 μM 17AAG for 24h or left untreated.

G. 17AAG loses killing efficacy when its ability to degrade mutp53 is overwhelmed by excess ectopic mutant p53. The indicated T47D cells were treated with 0.625 μM 17AAG or mock-treated for 48h. Top, Cell viability assay (CTB). Bottom, Corresponding immunoblot of empty vector or mutp53 R280K overexpressing T47D cells. HAUSP, loading control.

H. Proposed model of regulation of mutp53 stability by the HSP90 multi-chaperone machinery in cancer cells. See text for explanation.

Importantly, we find that destabilization of mutp53 via Hsp90 interference by 17AAG markedly inhibits the viability of SW480 colon cancer cells (Fig. 6E, compare column 1 with 3). This is again dose-dependent, since 17AAG at 2 mM cooperates with further reduction of mutp53 levels by tetracycline-inducible shp53 in reducing cell viability (Fig. 6E, compare column 2 with 4). Importantly, in a side-by-side comparison of mutant and wild-type p53 harboring cancer lines, 17AAG reduces cell viability more profoundly in mutp53 cancer cells. Moreover, 17AAG at the same effective concentration is non-toxic towards normal cells such as MRC5 (Fig. 6F). Thus, this data suggests that 17AAG might have more potent anticancer effects in mutp53 tumors compared to wtp53 tumors. In support of a causal link between 17AAG targeting mutp53 and 17AAG cytotoxicity, 17AAG largely loses its killing efficacy ( Fig. 6G top) when its ability to degrade mutp53 is overwhelmed by excess amounts of ectopically expressed mutant p53 (‘overstuffed’) (Fig. 6G bottom). At the concentration used, the excessively high level of mutant p53 has exhausted 17AAG’s ability to degrade it and concomitantly squelches 17AAG’s ability to affect cell viability. As expected, 17AAG retains some remnant efficacy, suggesting a partial p53-independent component of 17AAG action.

In sum, given that the tumor-specific aberrant accumulation of mutp53 is the basis for its GOF in malignancy and chemoresistance (6, 7, 35), understanding its underlying mechanism is critical for therapy of mutp53-harboring cancers. Based on our results we propose the following model as a likely scenario (Fig. 6H). Normal tissues in p53H/H knockin mice that harbor missense mutant p53 are able to efficiently control their mutp53 levels, despite the fact that their MDM2 levels are diminished since MDM2 is only supported by constitutive P1 promoter-driven transcription (4). Of note, mutp53 tumor cells are facing the same MDM2 situation, i.e. lower MDM2 levels that are only P1 promoter-driven due to impaired p53 transcriptional activity. Therefore, tumor-specific stabilization of mutant p53 proteins - which contributes to driving the tumor phenotype – largely or exclusively depends on a second alteration that these cells undergo upon their transformation. This alteration is the addiction of malignant cells to support from the activated heat shock machinery for their survival. In contrast to wild-type p53, the aberrant conformation of many mutant p53 proteins makes them dependent on heat shock support so that they stably engage in complexes with the highly activated HSP90 chaperone to prevent their aggregation. Intimately linked to this conformational stabilization, however, is the fact that this interaction also acts as a large protective ‘cage’ against degradation, thereby enabling mutp53’s GOF. The E3 ligases MDM2 and CHIP, which in principle are capable of degrading mutp53, are also trapped in this complex in an inactive state. Since mutp53 is fully competent to bind to MDM2, HSP90 likely binds to pre-existing mutp53-MDM2 complexes. Alternatively, chaperone-bound mutp53 could recruit MDM2. Depleting HSP90 components or binding of 17AAG to HSP90 destroys the complex, releases mutp53 and enables MDM2/CHIP-mediated degradation. However, although unlikely, formally it cannot be completely excluded that despite the same MDM2 situation as in normal tissues, the lower MDM2 levels in tumor cells might also play a minor role in mutant p53 hyperstability. In aggregate, these data provide encouraging evidence for the possibility of mutp53-directed anticancer therapy that targets an essential co-factor of its stabilization rather than mutp53 itself. We present a rationale for further pharmacological improvement in small molecule inhibitors of HSP90 chaperones. Such drugs, generally well-tolerated and some already in clinical trials, might represent an attractive mutant p53-targeting strategy for those 50% of cancer patients, particularly when combined with other anti-cancer agents.