Modulation of Gene Expression by Tumor-Derived p53 Mutants

More than 50% of the p53 mutations found in human cancers are missense mutations, and a majority of these tumors retain and overexpress the full-length mutant protein. The percentage of missense mutations found in the p53 gene is significantly larger than that found in other tumor suppressor genes (1, 2, 3, 4). The large percentage of missense mutations combined with the elevated level of expression of the tumor-derived p53 mutants suggests that mutant proteins perform a vital oncogenic role and are therefore selectively overexpressed. This follows the gain of function hypothesis, which predicts that mutations in the p53 gene not only destroy the tumor suppressor function of the wild-type (WT) protein but that the mutant protein may also gain oncogenic functions. This hypothesis also predicts that tumors with mutant p53 proteins may be more aggressive or that patients with such tumors have a poorer prognosis than patients with tumors lacking the p53 protein. This is true for cancer patients with tumors containing p53 mutations (5, 6).

WT p53 is widely accepted to act as a cell cycle inhibitor or an apoptotic agent when accumulated at a relatively higher than normal level under stress situations (1, 3, 4, 7). Tumor-derived p53 mutants are mostly defective in the growth suppressor function. This stems from a loss of transcriptional activation and repression of genes normally regulated by WT p53 (4, 7, 8). On the other hand, we and others have shown that p53 mutants derived from cancer can transcriptionally activate promoters of genes involved in cell growth regulation (3, 4, 9, 10, 11).

Broadly, three types of p53 mutations can be identified (12, 13): (a) loss of function, where the tumor suppressor activities of p53 are abolished; (b) dominant negative, where hetero-oligomeric complex formation between WT and mutant p53 results in the inactivation of WT p53 present in the cells; and (c) gain of function, where mutant p53 procures a dominant oncogenic role that does not depend on complex formation with WT p53.

The gain of function hypothesis for mutant p53 can be tested in cells devoid of endogenous p53 (10, 14, 15). For example, using murine cell systems [10(3) p53-null cells] it was shown previously that constitutive expression of mutant p53 increased oncogenicity of the cells, as manifested by increased tumorigenicity when injected subcutaneously into nude mice (10, 14). Because there was no endogenous WT p53 function for the mutant to eliminate, it was previously taken to mean that mutant p53 must have been acting directly to stimulate unregulated cell growth. The NH₂-terminally located transactivation domain and the COOH-terminally located sequences overlapping the oligomerization domain and the nonspecific nucleic acid binding domain of p53 mutants are required for this gain of function (11, 14, 16). Thus, the data generated thus far indicate a direct relationship between transactivation properties of mutant p53 and gain of function.

With the discovery of p53 family members p73 and p63 (17, 18), the gain of function activities of p53 mutants can also be explained by interaction of mutant p53 with these proteins. Some such evidence already exists (17, 18). However, as reported recently, we could discover very small amounts of p73-mutant p53 hetero-oligomers in H1299 cells, suggesting that factors other than interaction with p73 may also be involved (19). We have not measured p63 levels in these cell lines yet.

To test the hypothesis that mutations in p53 can activate transcription of growth-promoting oncogenic genes in human cells, we generated stable cell lines expressing mutant p53-R175H, p53-R273H, or p53-D281G using p53-null H1299 lung cancer cells. Compared with vector-transfected cells, the presence of mutant p53 induced an increased growth rate in H1299 cells. To define the molecular pathway by which mutant p53 achieves these effects, microarray hybridization analyses were conducted using Affymetrix (Santa Clara, CA) U95Av2 arrays. Using statistical significance and cluster analyses, a striking pattern emerged revealing that all three p53 mutants induced expression of a common set of genes, a significant number of which are involved in cell growth, survival, adhesion, and angiogenesis, and expression of which is inhibited or unaffected by WT p53. Real-time quantitative polymerase chain reaction (QPCR) analysis of representative genes verified the microarray data.

Stable cell lines were generated after transfection of H1299 with mutant p53 expression plasmids (or expression vector alone), which contains a neomycin resistance gene as described previously (14, 16). Stable cell lines were made for the following: p53-R175H, p53-R273H, p53-D281G, and p53-D281G [L22Q/W23S (11, 20)] using neomycin (G-418) at 400 μg/mL. A control cell line was also generated (HC-5). After 3 to 4 weeks of selection, individual colonies were isolated and checked for expression of p53 by Western blot analysis using the monoclonal antibody PAb1801. Selected mutant p53-expressing clones were expanded into cell lines and used for additional experiments.

Recombinant adenoviruses expressing WT p53 and β-galactosidase were generated at the Massey Cancer Center in the laboratory of Dr. Kristoffer Valerie as described elsewhere (21).

In a 10-cm dish, 3 × 10⁶ cells were infected with the recombinant adenoviruses (21) expressing either WT p53 or β-galactosidase at a multiplicity of infection of 10. At 20 to 24 hours, RNA was extracted for microarray analysis as described below. Total RNA was isolated using Trizol reagent (Invitrogen, Carlsbad, CA) following the manufacturer’s protocol and checked by 1.2% agarose-Tris-borate-EDTA gel electrophoresis.

Expression profiles of mutant p53-expressing cells were compared with H1299 cells stably transfected with vector alone. Cells (approximately 3 × 10⁶ cells at the time of preparation) were plated 36 to 40 hours before RNA preparation. To compare with the mutant p53 expression profiles, H1299 cells were also infected with recombinant adenovirus expressing WT p53 and compared with cells infected with adenovirus expressing β-galactosidase as a control. One culture (approximately 3 × 10⁶ cells) per microarray sample was used. All microarray hybridization analyses were performed with biological duplicates of each mutant p53-expressing cell line using Affymetrix U95Av2 chips by either the GeneChip Core lab at University of California San Diego or Virginia Commonwealth University’s institutional Nucleic Acid Research Facility. The U95Av2 arrays represent 12,625 human gene sequences and expressed sequence tags characterized previously, in which each gene is represented by a probe set consisting of 16 to 20 probes. The general procedures for microarray hybridization and analysis have been described elsewhere (22).

Data analysis was done using Affymetrix Microarray Analysis Suite version 4. All arrays were normalized by a correction to a set value of median total hybridization intensity. Changes in gene expression were detected in terms of statistical significance of change in expression for a given gene between two compared microarrays [S-score (23, 24)]. S-score analysis takes into account signals detected by 16 to 20 multiple probe pairs for individual genes, as well as intensity-dependent and -independent noise. The general procedures have been described elsewhere (22). Briefly, the normalized files generated by the Microarray Analysis Suite version 4 software were analyzed by the S-score program (24). Two hybridizations from each cell line were performed, and changes in gene expression were detected in terms of S-score (23, 24). S-scores are derived to have a mean value of 0 (representing no change) with a SD of 1. Genes with an average intensity difference value of <50 in at least one sample were filtered out, resulting in 1,724 genes. The S-scores generated from our mutant p53-expressing cell lines were then analyzed for significance across replicate experiments by using a permutation method performed with the Significance Analysis of Microarray (SAM) program from Stanford University [Stanford, CA (25)]. The settings for this analysis were as follows: one-class response, unlogged data, 300 permutations, K-nearest neighbor imputer of 10, and a random number seed of 123456789. Once the program reported the list of ranked genes, the “delta value” was adjusted to a stringent false discovery rate (FDR) of 0.3%. Cluster analysis was done using the Cluster and TreeView programs⁴ to provide a graphical display of the expression patterns (26, 27). Genes reported by SAM were analyzed by hierarchical clustering with average linkage grouping. For our analysis, the arrays themselves were not clustered. Functional grouping of the identified genes was done by manual editing of GeneOntology categories obtained through the DAVID annotation tool (28).⁵

QPCR was conducted using the LightCycler (Roche, Nutley, NJ). Details of this method are described elsewhere (16). Briefly, cDNA was synthesized using the Thermoscript reverse transcription-polymerase chain reaction system (Invitrogen). Primers were designed using Oligo 5 (Molecular Biology Insights) and synthesized by Sigma Genosys (St. Louis, MO). Reactions were performed in triplicate using SYBR Green dye. Primers used were as follows: (a) asparagine synthetase (ASNS), 5′-AGAGATTCTCTGGCGACCAAAAGA-3′ and 5′-CTGGGTAATGGCGTTCAAAGACTT-3′; (b) angiopoietin 1 (ANGPT1), 5′-GATGTCAATGGGGGAGGTT-3′ and 5′-CTCTGACTGGTAATGGCAAAAATA-3′; (c) cyclin B2, 5′-CTGCCACGCTTTTTCTGATG-3′ and 5′-GACTTGTACCAGCCAATCCA-3′; (d) c-myc, 5′-GCCGCCGCCAAGCTCGTCTCAGAG-3′ and 5′-GCTGCTGGTGGTGGGCGGTGTCTC-3′; (e) asparaginyl-tRNA synthetase (NARS), 5′-GCCGGATGAGTTGTGTC-3′ and 5′-ACCCCAATTAGTTCCCAGAA-3′; (f) RNA polymerase II, polypeptide E (RNA polII E), 5′-CCCACCGACCAGATCTTTG-3′ and 5′-GACGACGTGCTCAGGGACTC-3′; (g) cell division cycle 25A (CDC25A), 5′-AAGGCCCATGAGACTCTT-3′ and 5′-AAACTTGCCATTCAAAACAG-3′; (h) ras homologue gene family member G (rhoGAP), 5′-AGTACATCCCCACCGTGTTC-3′ and 5′-GGACTGGCAATGGAGAAAC-3′; (i) apoptosis inhibitor 5 (API5), 5′-CCGACCTAGAACAGACCTT-3′ and 5′-GCCAACAATTTCAATACCTC-3′; and (j) associated molecule with the SH3 domain of STAM (AMSH), 5′-TAGATGTGTTCCCAACCTTA-3′ and 5′-GTTGGCACTGGCTAACTG-3′.

Cells were seeded at a density of 1 × 10⁵ cells per 60-mm plate. Five plates of each cell line were plated initially. Whenever needed, one plate of each was trypsinized, and cells were suspended in 1 mL of serum-containing media; cells were counted each day using a hemocytometer, and the total number of cells was calculated and plotted. Media in the remaining plates were changed every day. SDs were calculated from three independent experiments run simultaneously.

Cloning of the presumptive gene promoters was done using genomic data sequence available in the National Center for Biotechnology Information (NCBI) database. Briefly, sequence-specific primers were designed to amplify [by polymerase chain reaction (PCR)] a genomic DNA fragment of up to 1 kb upstream sequences containing the transcriptional start site for each of the promoters. Genomic DNA from MCF-7 cells was extracted using DNAzol reagent (Invitrogen) and used in the PCR reaction as template. Primers were designed using the Oligo 5 (Molecular Biology Insight, Cascade, CO) program. The PCR fragment was then purified and cloned into the pGL3 luciferase reporter plasmid (Promega, Madison, WI) using the indicated restriction enzymes. Sequences of the fragment were confirmed by DNA sequencing (Amplicon Express, Pullman, WA); clones bearing no bp mutations were selected for further use. The following primers sets were used: (a) angiopoietin 1 (ANGPT1), 5′-CGGGGTACCCAGGAGGTTTTTATGTGGAA-3′ and 5′-CCCAAGCTTAATGGCAGCGAGGAA-3′ (KpnI and HindIII); (b) integrin α₆ (ITGA6), 5′-GGAAGATCTAGCCTTCATGCCACCTACAC-3′ and 5′-CCCAAGCTTGCCACCTTCGCCTCCTC-3′ (BglII and HindIII); (c) cyclin B2, 5′-CGGGGTACCTGGGCTGATTATTAGACGAA-3′ and 5′-GGAAGATCTACGGGGAAGGCAAGA-3′ (KpnI and BglII); (d) DNA polymerase δ subunit 2 (POLD2), 5′-CGGGGTACCAGGCCCAGGGAAGTAGCAGA-3′ and 5′-GGAAGATCTGCCCACCGACCCAGGAG-3′ (KpnI and BglII); and (e) transcription factor E2F-5, 5′-CGGGGTACCATGAAAACCAACCCTAAAACTCCA-3′ and 5′-CCCAAGCTTGATCCAGAACGCCGTCCTT-3′ (KpnI and HindIII). Luciferase construct for the asparagine synthetase and EBAG-9 were kind gifts of Michael Kilberg (University of Florida, Gainesville, FL) and Satoshi Inoue (Saitama Medical School, Saitama, Japan), respectively (29, 30).

Luciferase assays were done as described previously (16). Saos-2 (p53-null human osteosarcoma) cells were plated at equal density 24 hours before transfection. Culture media were changed on the day of transfection. Cells were transfected with 100 ng of the appropriate promoter-luciferase plasmids and 250 ng of an expression plasmid (pCMVBam) for WT p53, p53-R175H, p53-R273H, p53-D281G, or vector alone (16). Cells were harvested 48 hours after transfection using Reporter Lysis Buffer (Promega). Luciferase activity was detected using a Luminometer from Turner Designs. Cell extracts were normalized based on total protein concentration; the vector reading was arbitrarily set to 100.

H1299 cells (3 × 10⁵) expressing p53-R175H (H-R175H-72) or vector alone (HC-5) were transfected by LipofectAMINE 2000 with small interfering RNA (siRNA) directed against p53 or a nonspecific sequence (33 nmol/L, final concentration) 24 hours after plating. Cells were transfected a second time 24 hours later. Cells were plated at equal density (3,000 cells per well) 48 hours after the first transfection (day 2) in 96-well plates. Cell viability was measured using a (4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxy-phenyl)-2-(4-sulfophenyl)-2H-tetrazolium assay kit (CellTiter 96 Aqueous Non-Radioactive Cell Proliferation Assay Kit, Promega) for the next 5 days following the manufacturer’s protocol. Mutant p53 protein levels were detected using antibody PAb1801 from a parallel experiment; β-actin (AC-15; Sigma, St. Louis, MO) was used a loading control. Cell viability was expressed as a percentage of HC-5 control cells transfected with the same siRNA. siRNA sequences used were as follows: (a) control siRNA, 5′-CAUGUCAUGUGUCACAUCUCTT-3′ and 5′-GAGAUGUGACACAUGACAUGTT-3′; and (b) p53 siRNA, 5′-GCAUGAACCGGAGGCCCAUTT-3′ and 5′-AUGGGCCUCCGGUUCAUGCTT-3′ (31).

The p53MH program was used as described by Hoh et al. (32). The settings were moving average and bootstrap replicates of zero, no gap weights, smoothing factor of 0.1, and no sequence filtering. The scoring method used was a likelihood probability. Sites with a score of <80% were filtered out. Sequences of the hMDM2 and WAF1 promoters were used as controls for the analysis (data not shown). The control sequences were large enough to include known p53 binding sites.

Earlier data indicate that expression of mutant p53 in cells normally devoid of p53 may give a growth advantage (3, 10, 14). We tested that idea using mutant p53-expressing stably transfected H1299 cell lines (Fig. 1,A). Cells were plated at equal densities; one plate of each cell line was seeded for each day of the assay. The data in Fig. 1 show that mutant p53-expressing H1299 cells proliferate at a higher rate compared with the control cell line (HC-5). This increased growth rate was also observed with different clones of H1299 expressing the same p53 mutant and H1299 cells expressing different mutant p53 forms (R175H or R273H), suggesting that this characteristic is not due to a clonal variation (Fig. 1 B; data not shown).

Expression of the transactivation-deficient mutant p53-D281G (L22Q/W23S) did not increase the growth rate of cells as well as those expressing p53-D281G. This mutant p53 is defective in gain of function as measured by the lack of increased tumorigenicity when constitutively expressed in 10(3) cells (11). It has also been shown that this particular p53 mutant is defective in transactivation of the multiple drug resistance gene 1 (MDR1) and c-myc promoters, which are up-regulated by p53-D281G (11, 14). Because p53-D281G (L22Q/W23S) is defective in transactivation, the lack of growth enhancement implies that transactivation is needed for this property of mutant p53.

To determine whether expression of mutant p53 remains directly related to the enhancement of growth rate in H1299 cells, we used p53-specific siRNA. H1299 cells expressing p53-R175H were transfected with a double-stranded ribonucleotide sequence specific for p53 (or a nonspecific sequence as a control; see Materials and Methods) 24 and 48 hours before plating (31). Cell viability and level of p53 were measured every day for up to 5 days after plating. Data presented in Fig. 2 show that as the mutant p53 protein level decreases after siRNA transfection (Fig. 2,A, compare Lanes p53 with Lanes C for days 2–5), the number of viable cells remains similar to that of the control cells (Fig. 2B, bottom panel, days 3–5). This continues up to day 5, after which the transient effect of the siRNA diminishes. As the mutant p53 protein level begins to increase (Fig. 2,A, compare Lanes p53 on days 6 and 7 with Lanes p53 on days 2–5), cell viability also rises (Fig. 2,B, bottom panel, days 6–7), indicating that continuous mutant p53 expression is necessary for the observed phenotype. Thus, there is a direct relationship between mutant p53 level and growth enhancement in H1299 cells. Control siRNA itself does not affect mutant p53-mediated growth enhancement compared with vector control cells (Fig. 2 B, top panel).

In a parallel experiment, we tested whether protein levels of c-myc, a target of mutant p53 (Table 1), decreased after transfection with siRNA directed against p53 (Fig. 2 C). Western blot analysis shows that both p53 and c-myc levels decrease after transfection. Because the growth rate of mutant p53-expressing cells decreases after siRNA treatment, this suggests a possible involvement of c-myc in mutant p53-mediated growth enhancement in H1299 cells.

Because data described above and previously (3, 11, 14) suggest that mutant p53-mediated transactivation is necessary for the growth-promoting functions of mutant p53, we determined gene expression profiles modulated by p53 mutants to identify a possible pathway used by the mutant p53 proteins in inducing oncogenesis. Microarray hybridization analyses of three H1299-derived cell clones expressing p53 mutants (p53-R175H, p53-R273H, and p53-D281G) were performed. The gene expression profiles of vector-transfected H1299 cell lines were used as controls. Affymetrix U95Av2 arrays (HG-U95Av2) were used. Hybridization signals were normalized and filtered as described in Materials and Methods, and changes in gene expression were detected in terms of S-score for a given gene between two compared microarrays (23, 24). The magnitude of the S-score represents the significance of change; thus, an absolute value S-score of >2 represents a P value of <0.045, uncorrected for multiple comparisons. Scatter plot analyses (data not shown) of the S-scores show a very high degree of consistency of changes in gene expression between biological duplicates. The scattergrams also indicate a relatively large number of changes in gene expression in each of the mutant p53-expressing cell lines versus the control cell lines.

To identify genes reproducibly affected by the expression of any mutant p53 in these cell lines, we analyzed the S-score data by a permutation method, SAM (25). A SAM analysis of the S-scores has the advantage that small and subtle consistent changes reproduced across the array samples can be detected. A one-class analysis and a stringent FDR of 0.3% were used to focus on the most reproducible changes evoked by the p53 mutant across all hybridizations. We identified 108 (88 up-regulated and 20 down-regulated) genes with altered expression in the mutant p53-expressing cell lines. This type of analysis excludes those genes differentially affected by the mutant proteins themselves; therefore, a cluster analysis of this list only reveals two major clusters (Fig. 3). In parallel, microarray data obtained from H1299 cells infected with adenovirus expressing WT p53 (or β-galactosidase control) were analyzed in a similar manner (data not shown) and compared with the transcriptional profiles from the mutant p53-expressing cell lines to determine the transcriptional response of the identified genes in the presence of WT p53. Table 1 shows a representative example of the genes identified as up-regulated in H1299 grouped in functional categories. Authenticity of the microarray analyses was verified by QPCR for a representative group of gene targets (Fig. 4 A). We have recently reported the up-regulation of ASNS by mutant p53 (19), and it is used here as a control for the QPCR analysis.

We determined whether the triple mutant p53-D281G (L22Q/W23S) has retained the ability to up-regulate expression mutant p53 target genes identified by our expression analysis. We have used QPCR analysis and compared expression levels of a number of target genes among vector-transfected, p53-D281G–expressing, and p53-D281G (L22Q/W23S)–expressing H1299 cell clones. As shown in Fig. 4,B, in all of the cases examined, the presence of the two additional mutations within the transactivation domain caused a significant loss of up-regulation. Combined data presented in Figs. 1, 2, and 4 indicate that transactivation by tumor-derived p53 mutants is directly related to the ability of the mutant protein to increase the growth rate of H1299 cells in which they are expressed.

To determine whether the promoters of some of the genes whose expression was found to be up-regulated are transactivated by mutant p53 in transient transcriptional assays, seven promoters or presumptive promoters of up-regulated genes were assayed. The presumptive promoters for angiopoietin 1 (ANGPT1), integrin α₆ (ITGA6), cyclin B2, DNA polymerase δ subunit 2 (POLD2), and E2F-5 were cloned from genomic DNA as described in Materials and Methods using information obtained from the NCBI database. Functions of the promoters of estrogen-responsive A9 (EBAG9; ref. 29) and asparagine synthetase (ASNS; ref. 30) and the presumptive promoters for ANGPT1, ITGA6, cyclin B2, POLD2, and E2F-5 were studied using a luciferase reporter plasmid and p53-null human osteosarcoma Saos-2 cells (16) in the presence and absence of WT or mutant p53 expression plasmids. Data in Fig. 5 show varying degrees of transactivation by the p53 mutants of the promoters studied. Although there is some variation among p53 mutants, they all have the capacity to transactivate the promoters, suggesting a uniformity of function; WT p53 repressed all of the promoters tested. The transcriptional data from these promoters further verify the earlier results of the microarray hybridization analysis.

We have observed that H1299 cells expressing p53-D281G (L22Q/W23S) do not have a significant growth advantage over vector-transfected cells, but expression of p53-D281G in H1299 cells generates a cell line with an increased cellular growth rate (Fig. 1 B). This confirms the notion that mutant p53-mediated transactivation is related to growth advantage (gain of function) observed in cells in which mutant p53 is expressed. This hypothesis predicts that transcriptional changes as a result of mutant p53 expression lead to the observed gain of function.

To understand the mechanism behind the phenotypes observed in cells expressing mutant p53, we have performed microarray analysis to study the transcriptional changes that occur in cells when mutant p53 is expressed in them. Consistent with this idea, we have identified a group of genes regulated by p53 mutants in the absence of WT p53 using DNA microarray hybridization. QPCR data (Fig. 4) suggest that our microarray hybridization analysis is reliable. Some of the identified genes are involved in cell growth, survival, adhesion, and angiogenesis (Table 1). Our finding that all three p53 mutants transactivate a common set of genes (Fig. 3) in the absence of WT p53 is unique and provides highly significant information toward understanding mutant p53-induced growth advantage, suggesting the existence of a common pathway used by a group of p53 mutants to aid in oncogenesis. In vivo transcriptional changes induced by p53 mutants have also been reported previously by us and others (16). Of the genes identified in our microarray, MCL1 EBAG9, ANGPT1, ITGA6, NFκB2, and E2F-5 expression has been reported as up-regulated in various cancer types (33, 34, 35, 36, 37, 38). Because information is not available in the literature, it remains to be seen whether these genes are up-regulated in the natural context of the p53 mutation, i.e., in human cancers carrying a p53 mutation.

Although the exact molecular mechanism behind the gain of function properties of p53 mutants has yet to be completely clarified, transactivation of growth-promoting genes remains a strong candidate. Tumor-derived p53 mutants p53-R175H, p53-R273H, and p53-D281G confer increased growth advantage and decreased drug sensitivity in the absence of WT p53 (Fig. 1,B; data not shown; ref. 15). Because the mutant p53-D281G (L22Q/W23S) is defective in transactivation (11) and fails to induce growth enhancement in cells in which it is expressed (Fig. 1 B), the overall conclusion from our work is a direct relationship between the described phenotypes and the transactivation ability of mutant p53.

Another possibility for the observed gain of function properties is the involvement of p73 and p63, two p53 family members (17, 18, 39). p73 and p63 share significant sequence similarity with p53 and are able to transactivate promoters with a WT p53 consensus sequence in their promoter region, such as p21 and MDM2. As with WT p53, overexpression of p73 or p63 leads to apoptosis and cell cycle arrest (40). Although there is sequence similarity between the oligomerization domains of p53, p73, and p63, WT p53 does not hetero-oligomerize with p73 and p63 (41). More recent reports suggest that p53 mutants can interact with p73 and p63 and that these interactions occur through association via the DNA-binding domain of the mutant protein, thus inactivating p73 (and p63) function (42, 43). There is evidence to suggest that this interaction is responsible for the decreased sensitivity to chemotherapeutic drugs that is observed in cells overexpressing mutant p53 (44). It is possible, however, that the phenotype observed in mutant p53-expressing H1299 cells may require a combination of transcriptional activation and protein-protein interaction with p73, p63, and perhaps other proteins. Our studies suggest that although p73 is present in H1299 cells, interaction of p73 with mutant p53 is minimal (19).

It is known that mutant p53-mediated transactivation does not require WT p53 DNA-binding sites, showing that the mechanism of transactivation by mutant p53 is distinct from that of transactivation by the WT protein (3, 14, 45). We envision at least three molecular explanations for the up-regulation of genes in the presence of mutant p53. (a) It may be that p53 mutants differ in their ability to recognize and bind responsive elements, although there is no strong evidence thus far for any mutant p53 response element in the literature. In this case, mutant p53 proteins may possess a DNA binding ability altogether distinct from the WT protein. (b) A second alternative exists whereby mutant p53 may become locked onto transcription factors such as Sp1, ATF/CREB, or nuclear factor-κB (46, 47, 48) to activate promoters. As such, differences in levels of transcription can be accounted for by the identity of the transcription factors involved or by the relative affinity of the mutant p53 for the transcription factor. (c) It is possible that mutant p53-mediated induction of mRNA, as judged by microarray or QPCR analysis, is the result of stabilization of RNA by a posttranscriptional mechanism, although no direct evidence exists for this mechanism either. The results of our promoter analysis suggest that RNA stabilization is not a mechanism of mutant p53-mediated gene up-regulation, at least for the target genes studied here.

Using an algorithm used by Hoh et al. (32) to find presumptive p53-binding sites on the upstream sequences (approximately 2,000 bp upstream from the 5′ end of the mRNA) of genes listed in Table 1, we observe that only in a few cases are there reasonable sites that may bind WT p53 (data not shown). These sites may play a role in WT p53-induced repression of transcription from these promoters, a prediction that remains to be tested in the future. In other cases, WT p53 must be repressing transcription by using protein-protein interactions (49, 50). Whether mutant p53 utilizes any of the WT p53 binding sites is an open question, although this is perhaps unlikely.

The observations reported here help formulate the role of mutant p53 in oncogenesis, where mutant p53 may play an active role in the progression of cancer. The ascribed function also attempts to explain why tumors expressing mutant p53 have a poorer prognosis. In a normal cell, both the alleles of p53 are WT, and p53 has full potency as a “protector” against stress situations such as DNA damage principally acting as a transcriptional activator. Because of a mutation in one of the alleles affecting the coding sequence of p53, a faulty mutant protein with one amino acid substitution is produced. Mutant p53 becomes stabilized; this weakens the transcriptional function (and the protector capability) of the WT protein, leading to the eventual loss of the remaining WT allele. The mutant protein begins working as a potential oncogenic agent, transcriptionally activating growth-promoting, angiogenic, and invasive genes, leading to aggressive oncogenic phenotypes. The identification of targets common to multiple p53 mutants is indicative of common or perhaps overlapping pathways to mutant p53-mediated tumorigenesis. Definition of this pathway and the molecular mechanisms dictating it will provide strong candidates for therapy development.

We thank Arnold Levine, Bert Vogelstein, Satoshi Inoue, Michael Kilberg, and Sohei Kitazawa for providing us with cells and plasmids.