Defective p53 Post-translational Modification Required for Wild Type p53 Inactivation in Malignant Epithelial Cells with mdm2 Gene Amplification*

Mdm2 gene amplification occurs in benign and chemotherapy-responsive malignant tumors with wtp53 genes as well as in breast and epithelial cancers. Mdm2 amplification in benign tumors suggests that it is not sufficient for p53 inactivation in cancer, implying that other defects in the p53 pathway are required for malignancy. We investigated mechanisms of wtp53 protein inactivation in malignant conversion of epithelial cells by comparing clonally related initiated cells with their derivative cancerous cells that have mdm2 amplification. Deficiencies in p53 accumulation and activities in response to DNA damage were not due simply to Mdm2 destabilization of p53 protein, but to continued association of DNA-bound p53 with Mdm2 protein and lack of binding and acetylation by p300 protein. The aberrant interactions were not because of mdm2 amplification alone, because DNA-bound p53 protein from initiated cells failed to bind ectopically expressed Mdm2 or endogenous overexpressed Mdm2 from cancerous cells. Phosphorylations of endogenous p53 at Ser18, -23, or -37 were insufficient to dissociate Mdm2, because each was induced by UV in cancerous cells. Interestingly, phospho-mimic p53-T21E did dissociate the Mdm2 protein from DNA-bound p53 and recovered p300 binding and p21 induction in the cancerous cells. Thus wtp53 in malignant cells with mdm2 amplification can be inactivated by continued association of DNA-bound p53 protein with Mdm2 and failure of p300 binding and acetylation, coupled with a defect in p53 phosphorylation at Thr21. These findings suggest therapeutic strategies that address both p53/Mdm2 interaction and associated p53 protein defects in human tumors that have amplified mdm2 genes.

Mdm2 gene amplification occurs in benign and chemotherapy-responsive malignant tumors with wtp53 genes as well as in breast and epithelial cancers. Mdm2 amplification in benign tumors suggests that it is not sufficient for p53 inactivation in cancer, implying that other defects in the p53 pathway are required for malignancy.

We investigated mechanisms of wtp53 protein inactivation in malignant conversion of epithelial cells by comparing clonally related initiated cells with their derivative cancerous cells that have mdm2 amplification.
Deficiencies in p53 accumulation and activities in response to DNA damage were not due simply to Mdm2 destabilization of p53 protein, but to continued association of DNA-bound p53 with Mdm2 protein and lack of binding and acetylation by p300 protein. The aberrant interactions were not because of mdm2 amplification alone, because DNA-bound p53 protein from initiated cells failed to bind ectopically expressed Mdm2 or endogenous overexpressed Mdm2 from cancerous cells. Phosphorylations of endogenous p53 at Ser 18 , -23, or -37 were insufficient to dissociate Mdm2, because each was induced by UV in cancerous cells. Interestingly, phospho-mimic p53-T21E did dissociate the Mdm2 protein from DNA-bound p53 and recovered p300 binding and p21 induction in the cancerous cells. Thus wtp53 in malignant cells with mdm2 amplification can be inactivated by continued association of DNA-bound p53 protein with Mdm2 and failure of p300 binding and acetylation, coupled with a defect in p53 phosphorylation at Thr 21 . These findings suggest therapeutic strategies that address both p53/Mdm2 interaction and associated p53 protein defects in human tumors that have amplified mdm2 genes.
p53 protein activity, interaction with other cellular factors, and transcription of genes involved in growth suppression can be abolished by p53 gene mutation and/or defects in p53 signaling pathways. Mutational inactivation of p53 occurs in ϳ50% of all human cancers (1), primarily during the late stages of tumor development, demonstrated by the lack of mutations in most colorectal adenomas and high frequency of p53 mutation in colorectal carcinomas (2). Furthermore, studies using the two-stage murine skin tumorigenesis model identified a wild type p53 (wtp53) genotype in almost all papillomas and in early well differentiated squamous cell carcinomas produced by chemical carcinogen exposure followed by promotion (3,4). Mutations and loss of p53 heterozygosity arose later in ϳ50% of moderately or poorly differentiated carcinomas, raising the question of how p53 protein is inactivated in the sporadic epithelial cancers that have wtp53 genes.
One mechanism of wtp53 protein inactivation that occurs in the absence of p53 gene mutations is mdm2 gene amplification. Overexpression of Mdm2 protein can result in decreases in p53 protein accumulation and activation by targeting p53 protein for ubiquitin-mediated degradation (5), as well as by interfering with the ability of the wtp53 protein to transactivate downstream target genes (6). Amplification and overexpression of the mdm2 gene and protein is seen in ϳ7% of all cancers. Whereas the most common tumors with mdm2 amplification are benign soft tissue sarcomas (20 -30%) and osteosarcomas (15-20%), which are commonly chemo-sensitive (7), amplification in malignant epithelial-derived cancers, including nonsmall cell lung cancer, esophageal squamous cell carcinomas, and breast carcinomas occur at lower frequencies (ϳ5-15%), generally in the presence of wtp53 (Refs. 8 and 9 and references therein). The high incidence of mdm2 amplification in benign lesions, and the more aggressive, malignant phenotype displayed by p53-null or mutant tumors with complete p53 protein inactivation, suggests that mdm2 amplification alone is not sufficient for complete inactivation of p53 protein. This implies that epithelial derived tumors with mdm2 amplification that have undergone malignant transformation must harbor an additional defect(s) in wtp53 functional pathways.
This observation is consistent with studies using transgenic mice expressing the mdm2 gene from a K14 promoter that targeted the basal layer of the epidermis. These mice displayed increased papilloma formation induced by chemical carcinogenesis (DMBA followed by TPA) and were predisposed to the appearance of premalignant lesions. However, only a small fraction (ϳ5%) progressed to squamous cell carcinomas (10). These results imply that tumors overexpressing the Mdm2 protein require further functional inactivation of the wtp53 protein to undergo malignant conversion and underscore the importance of studies focused on the mechanism of p53 loss of function in epithelial cells with amplified mdm2.
To identify defects in wtp53 function that contribute to the malignant progression of epithelial tumors with mdm2 gene amplification, two cell lines of common lineage, an aggressive squamous cell carcinoma (03R) (11,12) that displays mdm2 gene amplification in the presence of wtp53 genes and its initiated precursor (03C), were analyzed for defects in wtp53 protein activation and function. The p53 functional defect in 03R cells was not because of Mdm2 destabilization of p53 alone, because equalizing levels of p53 protein between 03C and 03R cells by ectopic expression failed to induce p21 expression. Endogenous Mdm2, p300, and p53 proteins isolated from cells exposed to DNA damage were analyzed in vitro for binding to a biotin-labeled DNA probe sequence, revealing a DNA-bound p53-Mdm2 complex in 03R cancer cells that failed to associate with p300, whereas DNA-bound p53 from 03C precancer cells bound p300 but not Mdm2. The overexpression of Mdm2 alone was not responsible for the continued DNA-bound p53-Mdm2 complex, because Mdm2 overexpression in 03C precancer cells did not result in Mdm2 association with DNA-bound p53. The continued association between DNA-bound p53 and Mdm2 was not the result of failed p53 phosphorylation at Ser 18 , -23, or -37 (Ser 15 , -20, and -33 in humans, respectively), which all were detected in 03R cancer cells after UV treatment. However, the exogenous expression of p53 protein mimicking phosphorylation at Thr 21 (18 in humans) prevented association of Mdm2 with DNA-bound p53, formed a DNA-bound p53-p300 complex, and induced p21 induction in response to DNA damage. This implies that defective phosphorylation of p53 at Thr 21 and the aberrant interaction between DNA-bound p53 and Mdm2 that inhibits p300 binding can combine to inactivate wtp53 protein at the malignant conversion stage of epithelial carcinogenesis.

MATERIALS AND METHODS
Cell Culture and Treatment-Non-transformed murine epidermal cell strain 291 exhibits phenotypically normal differentiation and morphology in vitro and in vivo (13). Initiated 291.03C (03C) and tumorigenic squamous cell carcinoma 291.03R (03R) were derived from exposure of 291 cells to 7,12-dimethylbenz[␣]anthracene and cells were maintained under optimal growth conditions, as previously described (11,13,14). Sequencing of the p53 and mdm2 genes in 03C and 03R cells was performed with three sets of overlapping primers spanning each gene using reverse transcriptase-PCR based methods, as previously described (12). The p53-null NK-1 epidermal cells, obtained from A. Balmain (UCSF, San Francisco, CA), were maintained under the same growth conditions used for 291 cells. Cells in log phase growth (at ϳ60% confluency) were treated with 4 gray of IR using a 137 Cs source irradiator or 135 J/m 2 of ultraviolet B (UV) light, using two Westinghouse FS20T12 sun lamps emitting primarily in the 290 -350 nm range.
Plasmids and cDNA Constructs-The green fluorescent protein (GFP) 1 reporter plasmids were constructed by inserting either two p21derived p53 responsive elements, pGFP-p53x2 (GCTCAAGCTTCGAA-TTCTAGAGAACATGTCCCAACATGTTGGGCGTCGGCTGTCGGGG-AACATGTCCCAACATGTTGCGGGCATTGATCCGAGGTCCACTTC-GCTATATATTCCCCGAGCTCCTATCTACACGG) (15), or mutated p53-binding sequences, pGFP-p53mut (containing MG 15 (16)), into the pEGFP-1 vector (BD Biosciences). The pCHDM1A plasmid, expressing N-terminal HA-tagged human Mdm2, was kindly provided by Hua Lu (OHSU, Portland, OR). Where indicated, either pCMVp53 or pIREp53 was used for the expression of murine wild type p53. The Ser codons of murine p53 at amino acid positions 23 and 37, and the Thr codon at 21, were changed to Glu (phospho-mimic) by PCR site-directed mutagenesis and expressed using the pIRE2-GFP vector (BD Biosciences). The entire PCR-derived p53 coding sequence in each vector was sequenced at the OHSU Molecular Microbiology and Immunology Research Core Facility (OHSU MMI-RCF).
Transient Transfection and GFP Reporter Assay-Murine epidermal cells were transiently transfected using LipofectAMINE (Invitrogen), according to the manufacturer's instructions. Briefly, cells in 100-mm dishes at ϳ50 -60% confluency were transfected with 5 g of the indicated plasmid (5 l of LipofectAMINE/1 g of DNA) and subjected to DNA damage ϳ16 -18 h post-transfection and analyzed after DNA damage at the time points indicated. GFP expression from pEGFP-C1 (constitutive GFP expression from CMV promoter) was used to determine that transfection efficiencies were similar in NK-1, 03C, and 03R cells (25 to 35% of cells). The GFP expression from pEGFP-C1 and the GFP reporter plasmids were analyzed by fluorescent microscopy using an Olympus IX70 microscope (Olympus America Inc.) and Magnafire Mega-Pixel digital camera and imaging and control software (Media Cybernetics). After image analysis, whole cell lysates were prepared from cells transfected with GFP reporter constructs and separated by 9% SDS-PAGE followed by immunoblotting for p53, GFP, and hsp70 expression.
Flow Cytometry-The 03C and 03R cells were harvested 24 h after treatment and fixed using a Cytofix/Cytoperm kit, according to the manufacturer's instructions (BD Pharmingen). Fixed cells were pelleted and stained at 22°C for 20 min with 5 g/ml Hoechst (33342 Molecular Probes) in phosphate-buffered saline. Treatments were performed in triplicate and each series profile generated was compiled from ϳ20,000 events after gating on single cells. The total cellular DNA content was detected using a FACScan Vantage flow cytometer (BD Biosciences) at the OHSU Flow Cytometry Core Facility. Data were analyzed with WinList 2.0 software (Verity Software House).
Colony Formation Assay-Aliquots of viable cells harvested for cell cycle analysis were collected and plated at 300, 1000, and 3000 cells/ 60-mm dish in triplicate. Twelve days after plating, the cells were fixed with 100% methanol, stained with 10% Giemsa, and colonies were counted.
DNA Affinity Immunoblotting-DNA binding was analyzed by DNA affinity immunoblotting (DAI), a sensitive in vitro technique for measurement of endogenous DNA-binding proteins (19). Briefly, the p53 consensus binding sequence, 5Ј-TTCGAGAGGCATGTCTAGGCATGT-CT-3Ј (hup53), or mutant binding sequence, 5Ј-TTGAGGTCAGGCAG-TGCACTGCAC-3Ј (mutp53), and complement were synthesized in the OHSU MMI-RCF, biotin-labeled on one strand with a single biotin residue using 5Ј biotin-TEG (Operon, Qiagen), annealed, and used as 26-bp DAI probes. Where indicated, 134-bp DAI probes (p53x2 and p53C) were used to avoid potential steric hindrance between the streptavidin-biotin and the p53-p300 complex. These probes were generated by PCR using plasmids pGFP-p53x2 (specific probe), detailed above, and pGFP-p53C (control probe) (GCTCAAGCTTCGAATTCTTGAGGT-CAGGCAGTGCACTGCACGGCGTCGGCTGTTGAGGTCAGCAGTGC-ACTGCACTTGAGGTCAGGCAGTGCACTGCACTTCGCTATATATTC-CCCGAGCTCCTATCTACACGG, italics indicates sequence changes from p53x2) as templates with biotinylated reverse primer ([Bio-TEG]CCGTGTAGATAGGAGCTCGGG) and forward primer (GCTCAA-GCTTCGAATTC). Following PCR amplification the 134-bp probes were visualized by electrophoresis and gel-purified using a gel purification kit according to the manufacturer's instructions (Qiagen). A 200-g aliquot (unless otherwise indicated) of nuclear extract or whole cell lysate was diluted in 1ϫ DNA binding buffer (DBB) (20 mM Tris (pH 7.2), 1 mM EDTA, and 0.06% Triton X-100 supplemented with 50 mM NaF, 1 mM Na 3 VO 4 , 5 mM dithiothreitol, and 10 g of salmon sperm DNA) or 1ϫ DBB-NaCl (DBB prepared with 50 mM NaCl and 3% glycerol), respectively, with a final ammonium sulfate/NaCl level equal to 100 mM and glycerol level equal to 7%. Reactions reached steadystate during incubation for 30 min at 4°C with 60 nM biotin-labeled DNA. DAI studies using extracts from cells exogenously expressing Mdm2 were performed with 100 g of salmon sperm DNA. The DNAprotein complexes were precipitated for 2 h at 4°C with 0.1 mg of streptavidin magnasphere paramagnetic particles (Promega), washed with 1ϫ DBB-AS (DBB prepared with 50 mM ammonium sulfate and 3% glycerol) or 1ϫ DBB-NaCl for nuclear or whole cell lysate precipitates, respectively, and analyzed by immunoblotting, as described above. The DNA-bound p53 was quantitated and values were averaged from three independent experiments, utilizing OptiQuant software. To determine -fold increase in p53 bound sequence specifically to DNA, lysates of IR-or UV-treated cells were compared with their respective untreated cell lysates. In addition, DNA-bound p53 was compared with the total p53 protein in an equivalent amount of cell lysate as follows: ((DNA-bound p53/total p53) ϫ 100 ϭ relative p53 DNA binding index (RDB)).
Immunoprecipitation-Immunoprecipitation of p53 from whole cell extracts, to analyze association with Mdm2, was carried out as reported previously (20) using PAb421 monoclonal antibody (1 g) incubated with lysate overnight at 4°C. The NaCl was diluted to a final concentration of 125 mM by the addition of 1ϫ DBB. The immune complexes were precipitated with protein A-Sepharose (Amersham Biosciences) and detected by immunoblotting for p53 and Mdm2 as described above.
Immunoprecipitation of p53 for the detection of phosphorylated and acetylated forms was performed as described above using 1.5 mg of lysate incubated with a mixture of PAb122 and PAb6.2 (21) agaroseconjugated antibodies.

Activation of Wtp53 Is Compromised in 03R Cancer Cells-
We previously reported increased p53 mRNA and decreased immunoprecipitable p53 protein in 03R tumorigenic cells, compared with initiated cell precursor 03C and normal progenitor cell 291, all of which had the wtp53 genotype (12). The p53 gene in initiated 03C and tumorigenic 03R cells was confirmed as wild type in the cells used for these experiments by sequencing as described under "Materials and Methods" (data not shown). To identify the functional consequences of the p53 expression abnormalities, the ability of 03C precancer and 03R cancer cells to undergo G 1 cell cycle arrest and suppress colony formation in response to IR or UV treatment was analyzed. The 03C cells induced both G 1 growth arrest and suppression of colony formation, whereas 03R cells failed to demonstrate either activity in response to IR or UV treatment (Fig. 1, A-C). Also in response to IR or UV treatment, non-transformed 291 (results were equivalent to 03C and not shown) and initiated 03C cells accumulated p53 protein, as identified by immunoblotting ( Fig.  2A, lanes 4 -6 and 13-15, respectively), in contrast to 03R cells that failed to accumulate p53 in response to IR ( Fig. 2A, lanes 7-9) and accumulated ϳ2.5-fold less p53 than 03C cells in response to UV ( Fig. 2A, lanes 16 -18). The NK-1 (p53-null epidermal keratinocytes) cells displayed no detectable p53 ( Fig.  2A, lanes 1-3 and 10 -12, respectively), as expected (17).

FIG. 1. The 03R cancer cells fail to induce a G 1 growth arrest or suppress colony formation in response to IR and UV treatments.
A, untreated, IR-(4 gray), or UV-treated (135 J/m 2 ) 03C (initiated) and 03R (cancer) cells from triplicate cultures were collected at 24 h, fixed, stained with Hoechst solution, and analyzed by a FACScan flow cytometer for DNA content. The average % of cells in G 1 , S, and G 2 are shown ϮS.D. within triplicates. The sub-G 0 DNA content in each cell type remained unchanged following DNA damage (see Supplemental Materials Fig. S1). B, an aliquot of the untreated, IR-treated, and UV-treated 03C and 03R cells collected for flow cytometry in A were plated at 300 viable cells/60-mm dish in triplicate and stained 12 days after plating with Giemsa. C, Colonies that formed (Ն1 mm) in B were counted and expressed as the percentage of colonies formed relative to the untreated control.
Next the transcriptional activity of p53 in 03C and 03R cells was analyzed by immunoblotting for p21 protein accumulation in response to IR, and by activation of a p53-directed GFP reporter plasmid (15) in response to UV, because of the p53independent accumulation of p21 protein in NK-1 cells in response to UV ( Fig. 2A, lanes 10 -12), as has been previously described (22). Increases in p53 protein in 03C cells correlated with an increase in p21 protein after IR treatment ( Fig. 2A,  lanes 4 -6) and high levels of GFP expression from the GFP reporter 5 h after UV treatment (Fig. 2B, panel D, and Fig. 2C, lane 6). In contrast, the 03R cells displayed negligible differences in p21 accumulation in response to IR ( Fig. 2A, lanes 7-9) and no induction of GFP from the p53-specific reporter in response to UV (Fig. 2B, panel F, and Fig. 2C, lane 8). The 03C and 03R cells demonstrated equal transfection frequencies (ϳ30%, Fig. 2B, panels A and B), and no GFP expression from the negative control pGFP-Mut (Fig. 2C, lanes 1-4). These results indicate that while p53 protein can accumulate (although at reduced levels) in 03R cells in response to UV treatment, this p53 protein is transcriptionally inactive.
Further study of p53 transcriptional activity in the clonal epidermal cell model revealed that increases of p53 protein in 03C cells correlated with an increase in Mdm2 protein after IR and UV treatments ( Fig. 2A, lanes 4 -6 and 13-15, respectively), whereas no increase in Mdm2 protein occurred in NK-1 cells after treatment ( Fig. 2A, lanes 1-3 and 10 -12). Induction of Mdm2 was undetectable in 03R lysates ( Fig. 2A, lanes 7-9  and 16 -18), consistent with the defect in p53 induction in these FIG. 2. Accumulation of p53 protein from 03R cells does not lead to transcriptional activation. A, the NK-1 (p53-null), 03C, and 03R cells were harvested before (Ϫ) and at the indicated times after IR or UV treatment. Proteins (40 g) were separated by SDS-PAGE and detected by direct immunoblotting using antibodies to Mdm2, p53, p21, and hsp70. Fold induction for each protein was determined by densitometry from three independent experiments using OptiQuant software relative to the untreated lysate of each cell type (value of 1) and hsp70 internal loading control. B, the 03C and 03R cells were transiently transfected with either a GFP expressing vector pEGFP-C1 (C1, panels A and B) or a p53-responsive GFP reporter plasmid, pGFP-p53x2 (px2, panels C-F). After transfection with the p53 reporter, cells were analyzed by fluorescence microscopy for expression of GFP before (Ϫ) and 5 h after (ϩ) UV treatment. Micrographs shown are of identical fields (including phase contrast) before and after treatment that provide a representative example of the population of transfected cells. C, cellular proteins were isolated from 03C and 03R cells transfected with pGFP-p53x2 (px2) or pGFP-p53Mut (Mut), separated by 10% SDS-PAGE, and immunoblotted for expression of p53 and GFP protein in response to UV treatment (n/a, not applicable).
cells. While not induced by IR or UV, Mdm2 protein in 03R cells was constitutively overexpressed at ϳ10-fold higher levels than in 03C ( Fig. 2A, compare Mdm2 in lane 7 to lane 4) and was identified as wild type in both 03C and 03R cells by sequencing as described under "Materials and Methods" (data not shown). Southern analysis revealed that the mdm2 gene was amplified in 03R cells ϳ14-fold compared with 291 and 03C cells (Supplemental Materials Fig. S2), associating amplification of the mdm2 gene with malignant conversion in the 03R lineage.
The overexpression of Mdm2 protein may result in p53 protein functional defects, such as increased destabilization and/or cytoplasmic localization of p53 protein, resulting in transcriptional inactivation. However, mdm2 amplification does not always inactivate wtp53 (23), suggesting that Mdm2 overexpression itself is not responsible for the failed p53 transactivation. Equivalent basal levels of total p53 protein in 03C and 03R cells suggest that the p53 defect in 03R cells is not merely the result of increased Mdm2-directed degradation of p53. Using another approach to address this issue we tested the ability of exogenously overexpressed p53 to recover p21 induction in 03R cells alone or in response to IR. After equalizing the levels of p53 protein between NK-1, 03C, and 03R cells by transient transfection of a murine p53 expression plasmid (Fig. 3, lanes 3-4, 9 -10, and 11-12, respectively), p21 protein was induced in response to IR in transfected 03C cells expressing exogenous p53 (Fig. 3, lanes 9 and 10) as in untransfected 03C cells ( lanes  5 and 6). The p53-null NK-1 cells overexpressing p53 also displayed a recovery of p21 induction in response to the expression of p53 (Fig. 3, compare lanes 1 and 2 to lanes 3 and 4) indicating that the exogenous p53 protein is capable of transactivation. However, the 03R cells, even in the presence of p53 protein nearly equivalent to that induced by DNA damage in 03C cells, did not recover the ability to induce the p21 protein in response to IR (Fig. 3, lanes 11 and 12). In addition, the inactivation of p53 was not a result of cytoplasmic sequestration as indirect immunofluorescence studies showed endogenous p53 and Mdm2 proteins co-localized in the nucleus in both 03C and 03R cells (Supplemental Materials Fig. S3). These results indicate that Mdm2-mediated destabilization of p53 protein alone is not responsible for the defect in p53 function, implying other functional defects in the malignant conversion of the 03R cancer cells.
Aberrant Association of p53 and Mdm2 Proteins during Sequence-specific DNA Binding in 03R Cancer Cells-Another mechanism by which Mdm2 could be inhibiting p53 function is by interfering with the recruitment of p53 co-factors to the DNA-bound p53 protein necessary for efficient p53-directed transactivation. Previously, Jin et al. (24) demonstrated Mdm2 association with the p21 promoter in a p53-dependent manner in lung carcinoma H1299 cells, using chromatin immunoprecipitation assays. However, Zauberman et al. (25) have shown that transcriptionally active p53 in the wild type conformation, expressed in normal rat embryonic fibroblasts, binds DNA in the absence of Mdm2, as determined using McKay assays (26). The analysis of DNA binding and simultaneous association with Mdm2 protein by DAI revealed that, in precancerous 03C cells, the amount of p53 protein bound to the sequence-specific DNA probe (hup53) after IR or UV treatment increased (Fig.  4A, lanes 1-2 and 5-6, respectively) and this increase correlated with the induction seen in total p53 protein shown in Fig.  2A. Mdm2 protein was minimally detectable in the p53-DNA complexes identified by DAI using IR-and UV-treated 03C cell lysates (Fig. 4A, upper panel, lanes 1-2 and 5-6, respectively). Based on these results, Mdm2 proteins of non-transformed (291, data not shown) and initiated precancer cells did not associate with transcriptionally active p53 proteins in complex with DNA.
In contrast, p53 protein from 03R cancer cells treated with IR displayed no increase in sequence-specific DNA binding, indicating that a failure to accumulate p53 protein correlates with a lack of induction of p53 sequence-specific DNA binding (Fig. 4A, lanes 3 and 4). Yet p53 induction in 03R cells in response to UV correlated with increases in DNA-bound p53 protein (Fig. 4A, lanes 7 and 8, compared with Fig. 2A, lanes 16  and 17). This indicates that the p53 protein in 03R cells can accumulate in response to UV treatment and bind sequence specifically to DNA. However, the DNA-bound p53 proteins from IR-and UV-treated 03R cell extracts were in complexes with Mdm2 as shown by DAI (Fig. 4A, upper panel, lanes 3-4  and lanes 7-8, respectively). The presence of Mdm2 in the DNA complexes depended upon p53 protein, as indicated by the correlated increases in p53 and Mdm2 proteins in complex with DNA after UV treatment of 03R cells (Fig. 4A, lane 8), and the absence of p53 and Mdm2 proteins bound to DNA with the negative control mutp53 probe (Fig. 4A, lanes 9 -11). Co-immunoprecipitations of p53 with PAb421 verified that p53 and Mdm2 can associate normally in 03C cells in the absence of sequence-specific DNA (Fig. 4, B and C, lanes 1 and 2) and also can co-immunoprecipitate in the 03R cells. But in parallel DAI experiments, Mdm2 associated with DNA-bound p53 only in 03R cells (Fig. 4, B and C, lanes 7-8) and not in 03C (lanes 5-6). Thus, Mdm2 from 03C cells associated with p53 protein as required for Mdm2 regulation of p53 stability. However, the p53 proteins in 03C cells had either become modified after IR and UV treatment such that association with Mdm2 was prevented or had dissociated from Mdm2 protein upon sequencespecific binding to the DNA probe. The interaction between DNA-bound p53 and Mdm2 proteins in 03R cell lysates correlates with the transcriptional inactivation of wtp53 observed during malignant transformation in the murine epidermal model of tumorigenesis. These results suggest a mechanism for inhibition of p53 transcriptional activity in cells by Mdm2 and p53 protein association in transcriptionally inactive trimeric complexes with p53-specific DNA.
p300 Fails to Associate with DNA-bound p53-Mdm2 Complexes-p53 activation and stability is regulated not only by FIG. 3. Exogenous expression of p53 protein in 03R cells does not recover p21 induction. The NK-1, 03C, and 03R cells were transiently transfected with the concentration of pCMVp53 expression plasmid indicated, resulting in equivalent levels of p53 expression in each cell line. After transfection, cells were harvested before (Ϫ) and 3 h after treatment with IR (ϩ). Forty g of each transfected lysate in parallel with untransfected lysate was separated by 12% SDS-PAGE and subsequently immunoblotted for p53, p21, and hsp70. Fold induction for proteins from 03C and 03R cells was determined by comparing with 03C untreated/untransfected, whereas proteins from NK-1 cells were compared with untreated/untransfected NK-1 cells, each adjusted for hsp70 loading control (n/a, not applicable).
Mdm2, but also by p300. While it has previously been demonstrated that a competition between Mdm2 and p300 exists for binding to p53 (27), the p300 protein can bind p53 and Mdm2 proteins to form a ternary complex (28) and can acetylate p53 protein, increasing p53 DNA binding and transcriptional activity (18, 29 -31). The ability of Mdm2 and p300 to alter p53 sequence-specific DNA binding has been analyzed by electrophoretic mobility shift assay (32); however, the ability of endogenous Mdm2 and p300 to form a ternary complex with DNA-bound p53 has not been addressed.
The 03C and 03R cells expressed p300 at similar steady-state levels (Fig. 5A). DAI analysis of 03C cell lysates identified p300 in complex with p53 bound to the p53x2 sequence-specific probe (Fig. 5B, lanes 1-3) that failed to precipitate a significant level of Mdm2 protein (Figs. 4A, lanes 1-2 and 5-6, and 5B, lanes  1-3). The association of p300 with DNA-bound p53 was shown to be specific for p53 and not for the DNA by the lack of p300 binding to the negative control probe p53C, which failed to bind p53 protein (Fig. 5C). Whereas a ternary complex has been demonstrated to regulate p53 stability (33), it appears that endogenously active p53 protein bound to a sequence-specific DNA probe associated primarily with p300, potentially freeing the p53 transactivation domain from Mdm2 interference (34). In contrast, the DNA-bound p53 from 03R cells that efficiently formed DNA-p53-Mdm2 complexes (Figs. 4A, lanes 3-4 and  7-8, and 5B, lanes 4 -6) did not associate with p300 (Fig. 5B,  lanes 4 -6). The association of p300 and DNA-bound p53 from 03C cells correlated with induction of p53 acetylation at Lys- FIG. 4. Aberrant p53 and Mdm2 protein interaction observed during sequence-specific DNA binding. A, the 03C and 03R whole cell lysates treated with IR or UV were subjected to DAI by incubation with wild type (hup53) or control (mutp53) biotinylated oligonucleotide probes, precipitation of DNA protein complexes with streptavidin beads, separation by 9% SDS-PAGE, and detection by immunoblotting with antibodies specific to p53 or Mdm2 proteins. Mutp53 control probe similarly failed to bind 03C proteins (data not shown). Fold activation of p53 sequence-specific DNA binding is shown relative to p53 from each paired untreated lysate. The faint lower band present in all lanes in the upper panel is a nonspecific band that interacts with Mdm2 antibody 2A10. B and C, untreated (Ϫ) or 3-h IR-(B) or 5-h UV-(C) treated cells were either immunoprecipitated with PAb421 (lanes 1-4) or subjected to DAI with hup53 probe (lanes [5][6][7][8] and analyzed for the association of p53 (PAb242-horseradish peroxidase-conjugated antibody (43)), and Mdm2 by immunoblotting. To nearly equalize the amount of p53 protein bound in each type of experiment, 200 g of total protein was used for immunoprecipitation (IP), whereas 300 g was used for DAI. Molecular mass markers (kDa) are shown to the right of the figure. 379 (target of p300 (35)) in response to IR and UV treatment (Fig. 6, lanes 1-6), whereas 03R cells failed to undergo efficient p300-directed acetylation in response to DNA damage (Fig. 6,  lanes 7-12). These results indicate that DNA-bound p53 that is transcriptionally inactive can readily associate with Mdm2 but not p300 and implies that Mdm2 association with DNA-bound p53 can interfere with the association of DNA-bound p53 and p300.
Defective p53 Implicated in Aberrant p53/Mdm2 Association in 03R Cancer Cells-Although we showed that the p53 and mdm2 genes are both wild type, the abnormal association between p53 and Mdm2 proteins in cancer cells could arise because of post-translational defects in either or both proteins. To determine whether the Mdm2 protein highly expressed in 03R cells was responsible for the defective DNA-bound p53/Mdm2 interaction, Mdm2 proteins from 03R cells (Fig. 7A, lane 4) were combined with activated, DNA-bound p53 proteins from IR-treated 03C cells (Fig. 7A, lane 6) and analyzed for their ability to associate. When the cell lysates from IR-treated 03C and 03R cells were combined (Fig. 7A, lane 9), the level of p53 protein bound sequence-specifically to DNA was increased to the levels observed with IR-treated 03C lysate alone (Fig. 7A, compare lane 6 with lane 9). However, there was no increase in the level of Mdm2 bound to the p53-DNA complex, indicating that the activated p53 protein from 03C cells was not bound by the Mdm2 protein from 03R cells. This result indicates that the function of Mdm2 is intact in the 03R cancer cells, whereas activation of wtp53 is defective.
The continued association observed between DNA-bound p53 and Mdm2 in the 03R cancer cells may also occur because of Mdm2 overexpression and/or defects in p53 post-translational modifications that have been described to disrupt the association between p53 and Mdm2 proteins (for review see Ref. 36).
To determine whether the aberrant interaction between DNAbound p53 and Mdm2 was the result of mdm2 amplification and overexpression, a tagged human Mdm2 (HA-Mdm2) was exogenously overexpressed in the 03C precancer cells at levels nearly equivalent to endogenous Mdm2 expression in the 03R cancer cells (Fig. 7B, lanes 3-4, compared with lanes 5-6; HA tag results in size difference), and the resulting cell extracts were analyzed by co-immunoprecipitation and DAI. The immunoprecipitation of p53 protein with PAb421 resulted in the co-precipitation of endogenous Mdm2 from GFP-transfected 03C cells (Fig. 7C, lane 1) and levels of exogenous HA-Mdm2 protein from 03C cells were nearly equivalent to endogenous Mdm2 precipitated from 03R cells (Fig. 7C, lane 2 compared  with lane 3). No p53 or Mdm2 protein was precipitated using the GFP immunoprecipitation control antibody (data not shown). DAI analysis of the identical cell extracts detected high levels of p53 bound sequence-specifically to DNA (Fig. 7C, lanes  4 -6) and not to the mutp53 negative control (Fig. 7C, lanes  7-9). In contrast to the association of exogenous HA-Mdm2 and p53 by co-immunoprecipitation, the HA-Mdm2 did not associate with p53 bound sequence-specifically to DNA from 03C cells (Fig. 7C, lane 5), whereas, as before, Mdm2 did complex with DNA-bound p53 from 03R cells (Fig. 7C, lane 6). Furthermore, 03C cells even in the presence of excess Mdm2 protein still induced the accumulation of p21 protein in response to IR (Supplemental Materials Fig. S4). These results imply that mdm2 gene amplification alone is not responsible for the defective p53 transcriptional activation of aberrant DNA-bound p53-Mdm2 complex observed in 03R cells, indicating that other defects in p53 signaling and regulation must exist. Therefore, we investigated phosphorylation sites of p53 within the putative Mdm2 binding site.
The Phosphorylation of p53 at Ser 18 6. Loss of p300 association with DNA-bound p53 in 03R cells resulted in loss of p300-directed acetylation. Approximately 1.5 mg of the indicated lysate was first immunoprecipitated with a mixture of p53-specific antibodies (PAb122-and PAb6.2-agarose-conjugated antibodies); precipitating proteins were then analyzed by 9% SDS-PAGE and immunoblotted with specific antibodies for total immunoprecipitated p53 protein (total p53, immunoblotted with 242-horseradish peroxidase-conjugated antibody) or p53-Lys 379 acetylation. Immunoblots for acetyl-p53 forms were repeated at least twice on independently prepared lysates. The immunoblots shown for each cell type are derived from the same lysate preparation.
FIG. 5. DNA-bound p53-Mdm2 complexes fail to associate with p300 protein in 03R cells. A, 40 g of the indicated nuclear extract was separated by either 7 (p300) or 9% (Mdm2 and p53) SDS-PAGE and immunoblotted with specific antibodies for p300, Mdm2, and p53. B, the 03C and 03R untreated (Ϫ), 3-h IR-, and 5-h UV-treated nuclear extracts were subjected to DAI with the p53x2 probe and proteins were separated by 7% SDS-PAGE and analyzed by immunoblotting as in A. C, the indicated 03C and 03R nuclear extracts were analyzed by DAI as performed in B for the ability of p53 to associate with a nonspecific DNA template (p53C) and to precipitate Mdm2 and/or p300. damage activating p53 for sequence-specific DNA binding and inhibiting the association of p53 and Mdm2 (reviewed in Ref. 36). The p53 N-terminal amino acids required for the p53/ Mdm2 interaction are highly conserved between residues Pro 13 to Pro 27 (human) with only one amino acid (Asp 21 to Gly 21 ) lacking identity, implying that signaling pathways for regulation of the p53/Mdm2 interaction are similar between species. Previous studies of p53 phosphorylation, using p53 peptides or recombinant proteins (37)(38)(39)(40)(41)(42)(43), indicate a potential role for defects in Ser 18/15 , Ser 23/20 , and Thr 21/18 phosphorylation in the continued association of Mdm2 with DNA-bound p53 and the malignant progression of 03R cancer cells in association with mdm2 gene amplification. To identify changes in p53 posttranslational modifications that may occur differentially in the 03C precancer to the 03R malignant cell in response to IR and UV, the phosphorylation of endogenous p53 protein from these cells were compared using phospho-specific antibodies. The phosphorylation of p53 at Ser 18 was identified by direct immunoblotting of cell lysate (Fig. 8A), whereas identification of Ser 23 and Ser 37 phosphorylation required enrichment of the p53 protein by immunoprecipitation followed by immunoblotting (Fig. 8B).
The phosphorylation of p53 at Ser 18 was up-regulated with similar kinetics in both 03C and 03R cells after either IR or UV treatment (Fig. 8A). Therefore, 03R cells are not deficient in pathways required for phosphorylation of p53 at Ser 18 after IR or UV, and phosphorylation of Ser 18 alone is not responsible for stabilization of wtp53 in non-transformed 291 cells (data not shown) or initiated 03C cells. We next analyzed the modification of p53 at Ser 23 and Ser 37 , where Ser 23/20 has been proposed to interfere with the p53/Mdm2 interaction. Phosphorylation of p53 at Ser 23 and Ser 37 in 03C cells was induced in response to IR or UV treatment (Fig. 8B, lanes 1-6 and 13-18, respectively). Whereas 03R cells displayed a loss of Ser 23 induction and a delay and reduction in the level of Ser 37 phosphorylation in response to IR (Fig. 8B, lanes 7-12), both Ser 23 and Ser 37 phosphorylations were observed in response to UV treatment (Fig. 8B, lanes 19 -24). The DNA-bound p53-Mdm2 complex occurs even following UV treatment (Fig. 4A, lane 8) indicating that Ser 23 and Ser 37 modifications alone do not result in the dissociation of Mdm2 and DNA-bound p53 but do correlate with increases in p53 accumulation and DNA binding activity not seen in 03R cells in response to IR, which lacks these phosphorylations.
Defective Phosphorylation of p53 at Thr 21 Implicated in the Continued Association of Mdm2 and DNA-bound p53 in 03R Cancer Cells-The induction of p53 phosphorylation at Ser 23 and Ser 37 observed in 03C and 03R cells in response to UV treatment prompted us to use site-directed mutagenesis to investigate the importance of Thr 21 in the aberrant association between Mdm2 and DNA-bound p53. The exogenous expression of WTp53 or phospho-mimic proteins (Thr or Ser substituted with Glu) p53-T21E or p53S23E/S37E in 03R cells resulted in ϳ6-fold higher levels of exogenous p53 protein expression compared with basal levels of p53 in GFP-transfected 03R cells (transfection efficiency and negative control; Fig. 9A, lanes 3-8 compared with lane 1). However, only the 03R cells ectopically expressing p53-T21E recovered p21 protein induction in response to IR (Fig. 9A, lanes 7 and 8), indicating that p53 protein transcriptional activity can be recovered in the presence of mdm2 amplification.
To nearly equalize the amount of p53 protein bound in each sample and in each type of experiment, 150 g of total protein was used for IP in lanes 1 and 3 and 175 g was used in lane 2, whereas 200 g was used  for DAI in lanes 4, 6, 7, and 9 and 240 g in lanes 5 and 8. Molecular mass markers (kDa) are shown to the right of the figure.   FIG. 7. Defective p53 protein implicated in the aberrant DNAbound p53-Mdm2 complex. A, the 03C and 03R untreated (Ϫ) and 3-h IR-treated (ϩ) lysates were subjected to DAI using hup53 in the lanes indicated as performed in Fig. 4A. In lane 9, IR-treated lysate from 03C and 03R cells (03Cϩ03R) was combined (400 g of total protein) and the mixture of lysate was analyzed by DAI as performed in Fig. 4A. Forty g of the indicated lysate was separated by electrophoresis in the first 4 lanes indicating the total p53 and Mdm2 protein levels in 03C and 03R cells. Proteins were identified by immunoblotting as in Fig. 4A and -fold increase in protein accumulation and binding was determined relative to 03C-untreated lysate for both direct lysate and DAI. B and C, the 03C and 03R cells were transiently transfected with human HA-tagged Mdm2 (03C cells) or GFP (03C and 03R cells) and UV-treated 18 h post-transfection. Five h after treatment both untreated (Ϫ) and treated (5) cells were harvested and lysed. B, 40 g of the indicated lysate was separated by 8% SDS-PAGE followed by immunoblotting for p53, Mdm2, and hsp70. C, the indicated 03C and 03R cell lysates were either immunoprecipitated (IP) using PAb421 (lanes 1-3) or subjected to DAI using hup53 (lanes 4 -6) or mutp53 probes (lanes 7-9) and analyzed for the association of p53 and Mdm2 by immunoblotting for p53 (PAb242-horseradish peroxidase-conjugated antibody) and Mdm2 as described above in B. Immunoprecipitation of p53 followed by immunoblotting for Mdm2 with a mixture of Mdm2 antibodies resulted in the enrichment of background bands that migrated more slowly than Mdm2 (upper two bands in C, lane 1).
The increase in p21 accumulation may result from the inhibition of Mdm2 binding to DNA-bound p53. By DAI, exogenously expressed WTp53 and phospho-mimic p53-S23E/S37E and p53-T21E bound sequence-specifically to the p53x2 DNA probe (Fig. 9B, lanes 3-8). The increase of DNA-bound p53 protein in WTp53-transfected 03R cells led to a proportional increase in the amount of Mdm2 precipitated with no additional p300 association observed (Fig. 9B, lanes 3 and 4). Interestingly, the p53-S23E/S37E phospho-mimic protein also failed to interact with p300 and led to a proportional increase in the amount of Mdm2 associated with DNA-bound p53 (Fig. 9B, lanes 5 and 6) consistent with the interpretation of the data from UV-treated 03R cells that phosphorylation of p53 at Ser 23 or Ser 37 does not prevent the association of Mdm2 and DNAbound p53. Individual Glu substitutions at Ser 23 or Ser 37 produced similar results to with the p53-S23E/S37E double phospho-mimic protein (data not shown).
In contrast, the DNA-bound p53-T21E mutant protein expressed in untreated and IR-treated 03R cells did not associate with additional Mdm2 protein (Fig. 9B, lanes 7 and 8), maintaining only the low level of Mdm2 associated with the endogenous p53 bound to the DNA probe observed in GFP-transfected 03R cells (Fig. 9B, lanes 1 and 2). Significantly, the failure of DNA-bound p53-T21E protein to associate with Mdm2 restored the capability of p300 to associate with DNAbound p53 (Fig. 9B, lanes 7 and 8) in the cancerous cells. Thus, Mdm2 association with DNA-bound p53 can interfere with the association of p300 with DNA-bound p53.
Taken together, these results imply that defects in phosphorylation of p53 at Thr 21 allow for the continued association of DNA-bound p53 and Mdm2 proteins. The association of Mdm2 with DNA-bound p53 then inhibits p300 association with DNAbound p53 and p300-directed acetylation, contributing to the malignant progression of cancer cells bearing mdm2 gene amplification.

DISCUSSION
Epithelial tumors that display mdm2 gene amplification, generally in the presence of wtp53 genes, are primarily malignant and similar to tumors with p53-null or mutant genotypes that lack functional p53 protein (8). However, amplification of the mdm2 gene is predominantly found in benign tumors implying that additional defects in the signaling and activation of wtp53 are required to drive the malignant progression of epithelial tumors harboring mdm2 amplification. Additional defects in wtp53 activation were analyzed in a clonal model of multistage carcinogenesis containing non-transformed, initiated, and malignant (with mdm2 amplification) derivative epidermal cells of a common lineage (12) giving the opportunity to determine alterations in post-translational modification and protein interaction in context of development of a sporadic cancer.
The interaction between Mdm2 and the N terminus of the p53 protein has been shown to destabilize p53 protein and inhibit exogenous p53 protein transcriptional activity by concealing the transactivation domain (18,20). The current results indicate that Mdm2 protein is unable to conceal the p53 transactivation domain of DNA-bound p53 in non-cancerous cells, demonstrated in 03C cells by p53 protein accumulation, induction of p53 downstream genes and growth inhibition, and lack of Mdm2 association with p53-DNA complexes, consistent with the results from Zauberman et al. (25). In contrast, p53 protein from tumorigenic 03R cells bound sequence-specifically to probe DNA in complexes with Mdm2, providing a novel mechanism for the functional inactivation of p53 protein in cancer cells with mdm2 amplification.
The 03R cancer cells that overexpressed Mdm2 protein ϳ10fold exhibited ϳ50% less induction of p53 protein in response to DNA damage than the 03C cells. The lower levels of p53 accumulation in response to DNA damage may have resulted from Mdm2-mediated destabilization of induced p53. However, the A, 40 g of the indicated 03C and 03R whole cell lysates collected after either IR or UV treatment were separated by 9% SDS-PAGE and immunoblotted for total p53 protein and phospho-Ser 18 p53. B, the Ser 23 or Ser 37 phosphorylation of p53 protein was analyzed as described in the legend to Fig. 6 for p53 acetylation. Immunoprecipitations were followed by immunoblotting with specific antibodies for phospho-p53-Ser 23 , phospho-p53-Ser 37 , or total p53 (total p53; PAb242-horseradish peroxidse-conjugated antibody). The p53 forms shown for each cell type and treatment condition are derived from the same lysate preparation. Experiments were performed at least twice in independent cell preparations, with identical results. reduced levels of p53 protein were not responsible for the inactivation of p53 protein in the 03R cells, because exogenous expression of p53 at levels nearly equivalent to those seen in 03C cells did not recover p21 induction in 03R cancer cells. Furthermore, overexpression of Mdm2 protein alone did not result in the formation of DNA-bound p53-Mdm2 complexes in 03C precancer cells. Combined, these results indicate that Mdm2 overexpression and destabilization of p53 alone is not sufficient for the functional inactivation of wtp53. In support of this observation, it has been previously shown that tumor cells lines harboring mdm2 gene amplification, including HCT-116 (colon carcinoma), CaCl 7336 (melanoma), and SJSA (osteosarcoma) that overexpress Mdm2 by 20 -100-fold, still accumulate transcriptionally active p53 in response to DNA damage (23). Taken together, these findings indicate that mdm2 amplification alone is not responsible for the complete functional inactivation of wtp53 protein and that other defects in p53 pathways must exist in malignant cells.
Regulation of the p53/Mdm2 interaction has been previously shown to be altered by p53 post-translational modifications (for review see Ref. 36). Previously, the individual phosphorylation of a human p53 peptide at Thr 18 , but not at Ser 15 , -20, or -37 has been shown to inhibit the association of p53 and Mdm2 proteins (37)(38)(39)(40). In addition, the expression of a phospho-mimic p53 D18T,D20S or phospho-defective p53 A18T or p53 A20S either prevented or increased Mdm2 association with p53, respectively (41)(42)(43)(44). In the current study, the complementary analyses of p53 phospho-mimic protein and endogenous phosphorylated p53 forms and their complexes with DNA in vitro, Mdm2 and p300 allowed for a novel view of p53 loss of function in malignant conversion. We show that phosphorylation of p53 protein in 03R cells at Ser 18 , Ser 23 , or Ser 37 , identified by phospho-p53 antibodies did not prevent the association of DNA-bound p53 and Mdm2 protein, because all of these phosphorylations occurred in re-sponse to UV treatment. The inability of Ser 23/20 to disrupt the DNA-bound p53-Mdm2 complex in the current study is further supported by the report that introduction of the murine phosphodefective p53 missense mutation p53 A23S into murine ES cells did not alter p53 accumulation, response to DNA damage, or ability to induce apoptosis (45).
Significantly, the p53-T21E phospho-mimic protein bound sequence-specifically to DNA and prevented the association of DNA-bound p53 and Mdm2. Furthermore, the p53-T21E phospho-mimic protein increased association of DNA-bound p53 and p300, and demonstrated p21 accumulation in response to IR. These results extend previous findings by demonstrating that Thr 21 phosphorylation can release the p53 transactivation domain from Mdm2 inhibition when bound to DNA and that Thr 21 phosphorylation can overcome a mechanism of p53 inactivation identified in epithelial tumors cells with mdm2 amplification. They also imply that defects in the phosphorylation of p53 at Thr 21 exist in 03R cancer cells and contribute to the continued association of DNA-bound p53 and Mdm2, leading to the functional inactivation of wtp53 protein in tumor cells with mdm2 amplification.
Besides disrupting the p53/Mdm2 interaction, phosphorylation of the p53 N terminus has been described to increase p53 interactions with CBP/p300 and TAF II 31 (6,46,47). This may occur by alteration of the p53 transactivation domain into an open conformation that is more conductive to CBP/p300 association (48), implying that the p53-T21E protein exogenously expressed in 03R cancer cells not only disrupted the association of DNA-bound p53 and Mdm2 but also attracted p300.
The ability of p53 protein to associate with either Mdm2 or p300 has been demonstrated by means of mammalian yeast two-hybrid assays, in which a competition was observed between p300 and Mdm2 for binding to a GAL4-p53 fusion protein bound to GAL4 DNA binding sites (27). Using DAI we were FIG. 9. Substitution of Glu for Thr at amino acid position 21 in p53 prevents the association of DNA-bound p53 and Mdm2 but results in an interaction with p300 protein. A, the 03R cancer cells were transiently transfected with GFP, WTp53, p53-S23E/S37E (double phospho-mimic), or p53-T21E expression vector and then analyzed by immunoblotting for p53, Mdm2, p300, p21, and hsp70 expression following IR treatment. Immunoblotting of p53-phospho-mimic proteins was performed with PAb122 only, as the epitope for PAb242 was lost following amino acid substitution. Fold increase in p53 protein expression and p21 induction was determined relative to untreated GFP-transfected 03R cells or each individual untreated sample, respectively, and hsp70 internal loading control. B, the indicated cell lysates were analyzed by DAI as described in the legend to Fig. 5B using only PAb122 to detect p53 protein by immunoblotting. The levels of p53, Mdm2, and p300 protein in B were quantified relative to untreated, GFP-transfected 03R cells. able to assess the association of DNA-bound p53 with Mdm2 and/or p300 using endogenous proteins characterized from cells at the initiated and malignant stages of tumorigenesis. In the current study, endogenous p53 bound sequence-specifically to DNA probes either complexed with p300, associated with p53 transcriptional activity in 03C precancer cells, or with Mdm2, associated with lack of p53 transcriptional activity and acetylation in 03R cancer cells. This implies that Mdm2 association with DNA-bound p53 can interfere with the association of p300, thereby reducing p53 transcriptional activity. Evidence for this interference is demonstrated by the ability of the p53-T21E phospho-mimic to prevent Mdm2 association and increase p300 association with DNA-bound p53, where continued Mdm2 association results in defective p53 function. However, a trimeric complex of p53, Mdm2, and p300 without DNA association has also been identified using exogenous protein expression and this complex inhibited the p300-directed acetylation of the p53 protein (32,49,50). In addition, the formation of a trimeric complex between p53, Mdm2, and p300 may be necessary for the efficient degradation of the p53 protein, as p300 has recently been shown to act as an E4 ligase for p53, leading to p53 polyubquitylation (28,33). Transcriptionally active DNA-bound p53 protein may fail to associate with Mdm2, allowing for increased p300 acetylase activity and decreased competition between acetylation and ubiquitylation of C-terminal p53 lysine residues (51).
Other questions remaining include the identification of the endogenous levels of Thr 21 phosphorylation in 03R cancer cells by phospho-specific antibody, when it becomes available, and identification of the kinase(s) responsible for the in vivo phosphorylation of Thr 21/18 . Candidate kinases for the phosphorylation of p53 at Thr 21/18 include casein kinase 1 and human vaccinia-related kinase, which have both been shown to phosphorylate p53 protein in vitro at Thr 21/18 (40,(52)(53)(54). Because of the ability of p53-T21E phospho-mimic protein to disrupt the DNA-bound p53/Mdm2 association and rescue p21 accumulation in 03R cancer cells, it will also be of interest to determine whether small molecule inhibitors of the p53/Mdm2 interaction can rescue p53 function in malignant tumor cells with mdm2 amplification. As tumor cells with mdm2 amplification commonly have wtp53 genes, this population of tumors could potentially reactivate wtp53 function in response to small molecule inhibitors in a similar fashion to the expression of p53-T21E in 03R cancer cells.
Taken together, mdm2 amplification alone was not sufficient for p53 inactivation but cooperated with defective Thr 21 phosphorylation and continued Mdm2 association with DNA-bound p53, resulting in the complete functional inactivation of wtp53 protein and malignancy. Dissection of wtp53 inactivating mechanisms in the presence of amplified mdm2 could lead to new molecular approaches to disrupt the Mdm2/p53 association. Furthermore, the Mdm2-p53 complex with DNA may potentially occur without mdm2 amplification, suggesting a novel mechanism of p53 inactivation and therapeutic target in tumors that maintain wtp53 genes.