ΔNp73, A Dominant-Negative Inhibitor of Wild-type p53 and TAp73, Is Up-regulated in Human Tumors

The p53 family member p73 has significant homology to the p53 tumor suppressor. Human full-length p73 (TAp73) shares 63% amino acid identity with the DNA-binding region of TP53 including conservation of all DNA contact residues, as well as 38 and 29% identity with the tetramerization domain and transactivation domain, respectively (1). TAp73 also shows functional homology to p53. Ectopically overexpressed TAp73α and TAp73β (two COOH-terminal splice variants) largely mimic p53 activities, including the induction of apoptosis, cell cycle arrest, and the transactivation of an overlapping set of target genes (2, 3). Moreover, endogenous TAp73 is able to integrate death stimuli from three different pathways: cellular and viral oncogenes, some forms of DNA damage, and T cell receptor hyperactivation. We and others have shown that deregulation of the oncogenes E2F1, cMyc, and E1A induces apoptosis and growth suppression in tumor cells in a p53-independent manner by transcriptionally inducing and activating endogenous TAp73 proteins (4–7). Moreover, during E2F1-mediated apoptosis in primary mouse embryo fibroblasts (MEFs),* a supra-additive cooperation exists between wild-type p53 and TAp73 (5). Although wild-type MEFs show 77% apoptosis after forced E2F1 expression, p53^−/− MEFs (containing TAp73) and p73^−/− MEFs (containing p53) alone show reduced killing ability of only 12 and 15%, respectively. The excessively weakened killing ability of p73^−/− MEFs, despite the presence of wild-type p53, is consistent with an important synergistic signal emanating from TAp73 that cooperates with p53 to induce oncogene-triggered death. Because oncogene deregulation of E2F1 is one of the most common genetic alterations in human tumors, this finding provides a possible physiologic cause for TAp73 overexpression in tumors. Furthermore, TAp73 is activated to mediate apoptosis by a restricted spectrum of DNA damage. Endogenous TAp73 is activated in response to cisplatin, and γ-IR in a pathway that depends on the nonreceptor tyrosine kinase c-abl (8–10). Moreover, doxorubicin stabilizes TAp73 protein by acetylation (11). Conversely, cells deficient in c-abl do not up-regulate their p73 and are resistant to killing by cisplatin. On the other hand, TAp73 is not activated by UV, actinomycin D, and mitomycin C, all of which activate p53 (1, 8). Lastly, peripheral T cells undergo apoptosis after hyperstimulation of their T cell receptors. This cell fate is mediated via the E2F1–TAp73 pathway (6). Consistent with this notion, E2F1 null mice exhibit a marked disruption of lymphatic homeostasis with increased T cells and splenomegaly (12, 13). Thus, TAp73 might function independently but synergistically with p53 in a tumor surveillance pathway in vivo.

However, despite this strong functional homology, data from human tumors and p73-deficient mice argue against a classical Knudson-type tumor suppressor role for the TP73 gene. TP73-deficient mice lack a spontaneous tumor phenotype (14) and inactivating mutations in human tumors are extremely rare (>900 tumors analyzed to date; for review see reference 15). Moreover, although all normal human tissues studied express very low levels of p73, multiple primary tumor types and tumor cell lines overexpress wild-type p73, including cancers of the breast, lung, esophagus, stomach, colon, bladder, ovary, liver, bile ducts, ependymal lining, myelogenous leukemia, and neuroblastoma (15). It is important to point out that to date, most studies identifying p73 overexpression in primary human tumors have examined total levels of p73, with only a few exceptions that specifically measured TAp73 (16, 17). Importantly, in the mouse, an N terminally truncated ΔNp73 protein has recently been found, generated from an alternative promoter in intron 3 and lacking a transactivation domain (18). ΔNp73 is the predominant form in the developing mouse brain and is the only form of p73 in the neonatal brain and sympathetic ganglia (14, 18). Mouse ΔNp73 plays an essential antiapoptotic role during developmental p53-driven neuronal death in vivo by acting as a dominant-negative inhibitor of p53 (18). Functional studies have shown that ΔNp73 is required to counteract p53-mediated neuronal death during normal “sculpting” of the developing mouse neuronal system (18). In primary neuronal cultures, withdrawal of the obligate survival factor nerve growth factor leads to p53 induction with p53-dependent cell death but a concomitant decrease of ΔNp73. Importantly, sympathetic neurons are rescued from cell death after nerve growth factor withdrawal or recombinant adenoviral p53 infection when ΔNp73 levels are maintained by viral delivery (18). Given the existence of this powerful transdominant p53 inhibitor in the mouse, the possibility arose that this isoform might in part be responsible for the overexpression seen in human tumors and, in fact, be the crucial component. Therefore, we sought the human counterpart of ΔNp73, determined whether human tumors express it and determined its potential role in cancer.

Primary tumors and normal tissues were collected at University Hospital State University of New York, Stony Brook in compliance with and approved by the Institutional Review Board. Freshly harvested tumors (minimum 60% tumor cells) and normal tissues were immediately snap frozen in liquid nitrogen and stored at −80°C until RNA extraction. The human breast cancer cell line MDA 231, the p53 null lines H1299 and SaOs2, and the wild-type p53 lines U2OS, RKO, and HeLa cells were maintained in DMEM/10% fetal calf serum.

Total RNA was extracted from tissues as previously described (17). 1 μg total RNA, 10 pmoles each of a radiolabeled isoform-specific upstream primer, and a common TP73 exon 4 reverse primer were used in a 10-μl reaction (Titan Kit; Roche) for 25 cycles to ensure linearity of the assay (unpublished data) and subjected to phosphoimage analysis. Primer sequences were 5′-TGC TGT ACG TCG GTG ACC-3′ (sense ΔNp73), 5′-CGA CGG CTG CAG AGC GAG-3′ (sense TAp73), and 5′-TCG AAG GTG GAG CTG GGT TG-3′ (antisense for both). ΔNp73 and TAp73 amplicon sizes were 175 and 365 bp, respectively. For Ex2Del p73, primers were 5′-GCG CCA GGC CAG CCG GGA CGG AC-3′ (sense) and 5′-CGC GGC TGC TCA TCT GGT CCA TGG TGC-3′ (antisense), which yielded a 320-bp band. Band intensities were normalized to their respective glyceraldehyde3-phosphate dehydrogenase (GAPDH) values. In matched samples, the up-regulation in tumors was calculated by dividing normalized ΔNp73^Tumor by ΔNp73^Normal, TAp73^Tumor by TAp73^Normal, or Ex2Del p73^Tumor by Ex2Del p73^Normal. To calculate the preferential up-regulation of ΔNp73 in tumors, the formula ΔNp73^Tumor/ΔNp73^Normal divided by TAp73^Tumor/TAp73^Normal was used. The analogous formula was used for Ex2Del p73. Some samples were independently repeated and yielded highly reproducible results. Genomic sequencing of p53 exons 5–9 was performed with the Amplimer Panel (CLONTECH Laboratories, Inc.). Immunocytochemical staining for p53 was done on fixed tissue sections from the same tumor mass.

A pcDNA3-based expression plasmid for ΔNp73α was generated with an NH₂-terminal Flag tag. pcDNA3-Fp53 expressing human Flag-tagged wild-type p53, pcDNA3-p73α and pcDNA3-p73β expressing hemagglutinin-tagged human TAp73α and TAp73β (provided by G. Melino, University of Rome, Rome, Italy), murine stem cell virus expressing a tetramerization domain mutant of ΔNp73α called ΔNp73L322P (corresponding to L371P in TAp73α), and the p53/p73-specific reporter PG13-Luc were previously described (4). PG13 is a reporter plasmid containing 13 tandem repeats of the p53 consensus DNA binding site. H1299 and SaOs2 cells were transfected by Fugene (Roche). Luciferase activity was normalized for renilla luciferase activity (Promega).

Hela and H1299 cells were transfected with expression plasmids for wild-type p53 (0.5 μg) or TAp73β (0.5 μg) with either 1.5 μg ΔNp73α or empty vector (see Fig. 3 C). The indicated molar ratios were used and green fluorescent protein (GFP) was cotransfected in all cases (see Fig. 3 B). Total cell lysates were prepared 24 h later and subjected to immunoblot analysis. Gel loading was normalized for equal vimentin or GFP levels (20 μg per lane). Antibodies to p73 were monoclonal ER15 (recognizes amino acids 380–495 of human p73α isoforms and detects TA and ΔN forms; Oncogene Research Products), polyclonal anti-p73 (raised against the COOH-terminal; Chemicon), and the polyclonal anti-ΔNp73 (raised in rabbit against the exon 3′ peptide LYVGDPARHLATA and immunopurified, and does not cross react with p53 or any TAp73 isoform). Antibodies against p53 (DO-1 and PAb 421), HDM2 (IF2), p21Waf1, and 14-3-3σ were from Oncogene Research Products and Flag antibody (M2) was from Sigma-Aldrich. For the normalization of protein loading, blots were reprobed with α-vimentin (BioGenex) or GFP (CLONTECH Laboratories, Inc., see Fig. 3 B). For coimmunoprecipitations (see Fig. 5, A–C), SaOs2 and U2OS cells were transfected with 2.4 μg of the indicated plasmids in single transfections, or 0.6 μg p53 plasmid plus 1.8 μg ΔNp73α in cotransfections. 24 h later, 600 μg lysates were subjected to immunoprecipitation with 1 μg of the indicated antibodies and analyzed by Western blot as previously described (19).

For the immunoprecipitation of ΔNp73 from tumors and normal tissues (see Fig. 2 D), frozen tissue was mechanically pulverized under liquid N2, resuspended in 1.5 ml TENN buffer (50 mM Tris, 5 mM EDTA, 150 mM NaCl, 0.5% NP-40, pH 8.0) containing a protease inhibitor cocktail (Roche), and homogenized using an electric homogenizer followed by sonication for 2 min. The lysates were centrifuged twice at 14,000 rpm for 15 min. 2 mg total protein in 500 μl TENN was precleared four times using 70 μl each time of a 1:1 mixture of protein A and G slurry (GIBCO BRL). The last supernatant was incubated with 1 μg ER15 antibody for 30 min at 4°C, followed by the addition of 40 μl slurry of Prot G beads. The mixture was then rotated overnight at 4°C. For controls, lysates were immunoprecipitated with monoclonal GC15 specific for p73β (Oncogene Research Products) or Flag antibody (M2). Pelleted beads were washed four times in RIPA buffer (50 mM Tris, 150 mM NaCl, 1% Triton X-100, 0.1% SDS, 1% NaDeoxycholate, pH 7.4), boiled in sample buffer for 3 min, and loaded onto an SDS-PAGE gel. Membranes were immunoblotted with polyclonal anti-ΔNp73 (1:20) and bands were verified by reblotting with ER15 (1:200). For coimmunoprecipitation of a ΔNp73–p53 complex from tissues (see Fig. 5 D), 2 mg total protein in 500 μl TENN was precleared as described above and incubated with a 1:1 mixture (40 μl slurry) of agarose beads covalently coupled to the monoclonal p53 antibodies 1801 or DO-1 (Santa Cruz Biotechnology, Inc.). For controls, 40 μl Prot G agarose slurry was used. Membranes were immunoblotted with polyclonal anti-ΔNp73 and bands were verified by reblotting with ER15.

Hela and SaOs2 cells were seeded into 8-well chamber slides and cotransfected with 300 ng of the indicated pcDNA3-based expression plasmids per well using Fugene (see Fig. 4 A). Control wells (vector alone) received 600 ng. After 16 or 24 h, cells were stained with Annexin V or Tdt-mediated dUTP-X nick-end labeling (TUNEL), respectively, according to the manufacturer's instructions (Roche). Expression was determined by immunofluorescence in duplicate wells. Transfection efficiency was reproducibly ∼30% of cells, similar among all constructs and evenly distributed throughout the wells. Annexin V or TUNEL positive cells (494 fields at 40×) and plasmid-expressing cells (15 random fields, >500 cells) were counted and the percentage of apoptosis in transfected cells was determined after correction for background with vector alone.

RKO cells were seeded into 8-well slides and transfected with either vector, the irrelevant control plasmid cRel, or with ΔNp73α using Lipofectamine Plus (GIBCO BRL; see Fig. 4 D). LcRel is a transcriptionally and apoptotically inactive Flag-tagged, truncated version of the transcription factor cRel, fused to a mitochondrial targeting sequence (19) and cloned into pcDNA3. 5 h after transfection, cells were either stressed with 5 μM camptothecin or left untreated overnight before fixation with paraformaldehyde. Wells were processed for TUNEL staining followed by immunofluorescence staining with polyclonal anti-ΔNp73 (1:20) and donkey anti–rabbit IgG (H+L)-rhodamine conjugate (Jackson ImmunoResearch Laboratories). Apoptosis was quantitated as the percentage of ΔNp73α-expressing cells and compared with vector-transfected cells.

SaOs2 cells in 60-mm plates were transfected by Fugene with 1.5 μg each of the indicated expression plasmids. Vector-only plates received 3 μg plasmid. 24 h later, cells were transferred to 100-mm dishes and placed under G418 selection (500 μg/ml) for 21 d, fixed, stained with crystal violet (20), and photographed. All foci were counted.

H1299 and U2OS cells were transfected with plasmids encoding wild-type p53, ΔNp73α, or a combination of the two. 24 h later, nuclear extracts were prepared by Dounce homogenization as previously described (21). As probes, we used double stranded oligonucleotides of the p53 consensus DNA binding sequences: 5′-GGGCATGTCCGGGCATGTCC-3′ (called p53CON) or 5′-CCTGCCTGGACTTGCCTGGCCTGCCTGGACTTGCCTGG-3′. EMSA was performed with 1 ng 32 P–labeled probe and 10 μg nuclear extract in binding buffer (40 mM NaCl, 10 mM morpholino propane sulfonic acid, pH 7.0, 0.1% NP-40, 1 mM EDTA, 2.5% glycerol, 0.5 μg poly[dI-dC]) for 30 min at room temperature. Samples were loaded onto a native 6% polyacrylamide gel and electrophoresed at 4°C and 200 V for 4 h. Where indicated, 0.5 μg p53-specific monoclonal antibody PAb 421 was included to produce a supershift. Specific competition experiments included unlabeled p53CON in 50-fold molar excess of radiolabeled probe. Nonspecific competition consisted of 50-fold molar excess of the scrambled p53CON 5′-GGGAATTTCCGGGAATTTCC-3′ or the mutant sequence CCTTAATGGACTTTAATGGCCTTAATGGACTTTAATGG (mutated nucleotides are underlined).

Wild-type p53-harboring RKO cells were seeded into 8-well slides and transfected with 360 ng wild-type p53 expression plasmid with 200 nM of either ΔNp73 antisense oligonucleotide directed against exon 3′ with extension into the adjacent 5′ untranslated region (UTR) (antisense, 5′-A*C*C*G*ACGTACAGC*A*T*G*G-3′; stars indicate the phosphorothioate-modified bases), or ΔNp73 sense oligonucleotide (sense, 5′-C*C*A*T*GCTGTACGT*C*G*G*T-3′; both HPLC-purified, Operon Inc.) using Lipofectamine Plus or Oligofectamine (GIBCO BRL) according to the manufacturer's instructions. Vector alone was also used. 20 h after transfection, cells were subjected to immunoblotting or TUNEL staining. Apoptosis was quantitated using a Nikon E800 microscope equipped with a Bio-Rad confocal laser scanning system. For each sample, the total fluorescence of five random fields per duplicate well was acquired with a 20× lens using identical acquisition and photomultiplier settings. Data were processed to calculate the integral optical density using the Laser Pix software package (Bio-Rad Laboratories). In other experiments, RKO cells were transfected with 200 nM each of the antisense and sense oligonucleotides listed above. After 8 h, cells were DNA damaged by adding 1 μM camptothecin for an additional 16 h before TUNEL staining. Apoptosis was determined as described above.

Mouse ΔNp73 differs from TAp73 by a novel exon 3′, which replaces the first 3 exons and is spliced in frame to exon 4 of the TP73 gene (18). To clone the human homologue of mouse ΔNp73, we performed a GenBank search using mouse exon 3′. By sequence alignment of a human genomic P1 artificial chromosome clone containing TP73 (sequence data are available from GenBank/EMBL/DDBJ under accession no. AL 136528), we identified a region with 77% identity to the 5′ UTR of mouse ΔNp73 mRNA (sequence data are available from GenBank/EMBL/DDBJ under accession no. Y 19235). This allowed us to predict the human exon 3′ and design isoform-specific primers for human ΔNp73. Full-length ΔNp73α cDNA, spanning exons 3′–14 including 103 bases of 3′ UTR, was cloned by RT-PCR from total RNA of human placenta and MDA 231 breast cancer cells, and sequence was confirmed (Fig. 1 A). To confirm that ΔNp73 derives from a separate promoter upstream of exon 3′, we performed 5′ rapid amplification of cDNA ends and sequenced the 5′ UTR upstream of exon 3′ (Fig. 1 C). Human exon 3′ encodes 13 unique amino acids with almost complete identity to mouse exon 3′ (12 out of 13 residues are identical; Fig. 1 A). The human ΔN promoter contains the predicted TATA box 25 nucleotides upstream of the transcriptional start site, which is located 7.6 kb downstream of exon 3.

Unique cDNA primers were designed for amplification of ΔNp73 from tissues by semiquantitative RT-PCR (Fig. 1 B). Then, we determined if expression levels of ΔNp73 were tumor specifically up-regulated using a spectrum of human tumor pairs that were matched with the patients' normal tissues of origin (Table I and Fig. 2 A). They included 35 cancers (cancers of the ovary, endometrium, cervix, vulva, vagina, breast, kidney, and colon) and 2 large benign ovarian tumors (serous cystadenoma and serous cystadenofibroma). Indeed, in 27 of the 37 pairs (73%), ΔNp73 was specifically up-regulated between 2- and 150-fold in the patients' tumors compared with their matched normal tissues (Table I, second column). In a subset of our matched pairs, we also determined the expression of the previously described Ex2Del p73 (23). Ex2Del p73 is another isoform of p73 that lacks most of the transactivation domain, but in contrast to ΔNp73 it is generated from the same promoter as TAp73 by splicing out exon 2. Ex2Del p73 was shown to be up-regulated in some ovarian and vulval cancers and breast cancer cell lines and is a transdominant inhibitor of p53 (22–24). As already seen with ΔNp73, Ex2Del p73 was also specifically up-regulated 2.2–20-fold in 11 of 22 analyzed tumors (50%) compared with their matched normal tissues either alone or concomitant with ΔNp73 (Table I, third column, and Fig. 2 A). Taken together, 30 of 37 tumor pairs (81%) exhibited tumor-specific up-regulation of ΔNp73 and/or Ex2Del p73. On the other hand, because we previously showed that breast cancers can overexpress TAp73 (17), we next used isoform-specific RT-PCR to simultaneously measure ΔNp73 and TAp73. TAp73 was up-regulated in 18 of 37 tumors (49%; Table I, fourth column) compared with its respective normal tissues of origin. However, although tumors with deregulated ΔNp73 and/or Ex2Del p73 also tended to show up-regulation of TAp73, the magnitude of the ΔNp73 and/or Ex2Del p73 up-regulation was much higher than that of TAp73 in the majority of cases. Among the 31 tumor pairs with up-regulation of any one or all of these three p73 transcripts, 22 tumors (71%) exhibited preferential up-regulation of ΔNp73 (19 tumors; Table I, left side of fifth column) or Ex2Del p73 (3 tumors; Table I, right side of fifth column). These included 13 ovarian cancers and 2 large benign ovarian tumors, 5 endometrial cancers, 2 breast cancers, 2 cervical cancers, and 2 vulvar cancers. Moreover, 14 of these 22 tumors (67%) exhibited exclusive up-regulation of ΔNp73 and/or Ex2Del p73. (“preferential” up-regulation means that the up-regulation of ΔNp73 and/or Ex2Del p73 was more marked than the up-regulation of TAp73). In one case, the tumor showed preferential ΔNp73 up-regulation of >2,700-fold compared with TAp73 expression (tumor no. 4). Only nine tumors (29%) exhibited an inverse ratio with a disproportional rise of TAp73 compared with their rise in ΔNp73. Thus, taken together, in our series of 31 matched tumors with some form of tumor-specific deregulation of the TP73 gene, 22 tumors (71%) exhibit either exclusive or preferential up-regulation of dominant-negative p73 isoforms.

In our paired tumor series, we had an overall p53 mutational rate of 38%, as determined either by direct sequencing of the PCR-amplified DNA binding domain or by immunocytochemically shown nuclear overexpression (Table I). This prevalence is in agreement with the reported rates of p53 mutations in these tumor types (∼40%; reference 25). We reasoned that if ΔNp73 and Ex2Del p73 were indeed p53- and TAp73-directed inhibitors and would act as oncogenes, one would expect to see their expression preferentially in wild-type p53 tumors. Of the 22 tumors with preferential up-regulation of ΔNp73 and/or Ex2Del p73, 21 tumors were available for p53 mutational analysis. Of these 21 tumors, 15 tumors harbored wild-type p53 (71%). In contrast, among the nine cases with preferential up-regulation of TAp73, six tumors harbored mutant p53 (66%). We then analyzed this set of 30 tumors for a correlation between overexpression of dominant-negative forms of p73 and concomitant wild-type p53 status. When all tumors were included, a statistical trend but no significance was found by Student's t test. When the unusual tumor number 4 is excluded, which is characterized by a singularly high level of ΔNp73 up-regulation of almost 3,000-fold compared with TAp73 up-regulation plus a mutant p53 status, statistical significance was found (P = 0.014). Taken together, a correlation between tumor-specific up-regulation of ΔNp73 or Ex2Del p73 and wild-type p53 status of the tumor cannot be made with this limited set of tumors, although a trend is present. More tumor samples will need to be analyzed in the future to support the hypothesis that the expression of dominant-negative p73 isoforms alleviates the selection pressure for p53 mutations in tumors.

To further investigate whether tumors up-regulate ΔNp73, we determined ΔNp73 transcript levels in a series of 52 unmatched breast cancers and compared them to 8 available normal breast tissues from unrelated individuals (Fig. 2 B). 16 of 52 breast cancers (31%) overexpressed ΔNp73 levels that were between 6- and 44-fold higher than the average of the 8 normal breast tissues (Fig. 2 B, gray line). An additional 10 tumors showed ΔNp73 up-regulation between two- and sixfold above the normal average. In contrast, four normal breast tissues showed nondetectable levels of ΔNp73, two cases expressed at average level, and only two tissues were elevated two- to fourfold. Next, we used isoform-specific semiquantitative RT-PCR to simultaneously measure ΔNp73 and TAp73 because we previously showed that breast cancers can also overexpress TAp73 (17). Among the 16 cancers with a 6–44-fold increase of ΔNp73, 12 cancers again showed preferential up-regulation of ΔNp73 over TAp73 (unpublished data). Although the data is not complete enough to make strong conclusions, as we had already seen in the gynecological cancers on Table I, our results on breast cancer again suggests that ΔNp73 might selectively be up-regulated during tumorigenesis.

To confirm that tumor-specific up-regulation of ΔNp73 transcripts translate to the up-regulation of proteins, we generated a ΔNp73-specific polyclonal antibody raised against the unique exon 3′. This antibody recognizes ΔNp73 but does not cross react with TAp73α, TAp73β, or p53 (Fig. 2 C). Using this reagent, we determined ΔNp73 protein expression on 10 matched pairs of homogenized tumor/normal tissues from Table I. Tissues were subjected to immunoprecipitations of equal amounts of total protein (2 mg each) with the anti-p73 specific antibody ER15 followed by immunoblotting with polyclonal anti-ΔNp73. Some examples are shown in Fig. 2 D, which represent cases 1, 9, 10, 14, 26, and 31. Tumor-specific up-regulation of ΔNp73 protein was found in all 10 cases as demonstrated by tumors yielding detectable ΔNp73α protein, whereas their respective matched normal tissue showed a complete absence of ΔNp73α protein in nine cases and only a minute amount in case number 10. Moreover, when tumor lysates in these cases (2 mg each) were immunoprecipitated with nonspecific Flag antibody, ΔNp73 protein could not be detected (unpublished data). Also, the immunoprecipitation of cases 9, 14, and 26 with β-specific anti-p73 antibody GC15 did not yield ΔNp73 protein, indicating that in contrast to ΔNp73α, ΔNp73β is not up-regulated to detectable levels in these cases. As expected, no strict correlation between the levels of tumor-associated protein and the ranking in Table I is present because the RT-PCR measurements indicate the relative fold increase of mRNA levels of tumor versus normal rather than absolute values. Interestingly, cases 26 and 31 did not exhibit increased ΔNp73 expression of their tumors at the transcript level, yet clearly do so on the protein level. This notion warrants additional investigation because it might suggest that the prevalence of tumor-associated ΔNp73 up-regulation is somewhat higher than transcript measurements would predict.

To test the hypothesis that human ΔNp73 is a dominant-negative inhibitor of human wild-type p53 and TAp73, we first performed reporter assays with expression plasmids for wild-type p53, TAp73α, TAp73β, and a p53/TAp73-responsive luciferase reporter in the presence or absence of ΔNp73α using p53 null H1299 and SaOs2 cells (Fig. 3, A–C, and unpublished data). Increasing ratios of ΔNp73α expression plasmid were added to a constant amount of p53 or TAp73α and TAp73β plasmids. Fig. 3 B shows that the p53 expression levels in this assay were independent of the amount of ΔNp73α added, and that p53 and TAp73 expression levels were not interfering with the amount of ΔNp73α expressed. Surprisingly, the steady state levels of TAp73β actually increased in proportion to the amount of ΔNp73α added, despite the fact that the same amount of TAp73β plasmid was used in all reactions. Of note, the addition of ΔNp73α exhibited complete suppression of the transcriptional activity of wild-type p53, TAp73α, and TAp73β in a dose-dependent manner (Fig. 3 A and unpublished data). Furthermore, ΔNp73α also efficiently suppresses endogenous target gene products of wild-type p53 and TAp73 (Fig. 3 C and unpublished data). In HeLa and H1299 cells, the transfection of wild-type p53 or TAp73β induces endogenous HDM2, 14-3-3σ, and p21Waf1 compared with basal levels seen with empty vector. However, the concomitant expression of ΔNp73α strongly suppresses each of these response gene products (compare lanes 2 and 3 and lanes 4 and 5). Expression of ΔNp73α alone did not suppress these target genes (compare lanes 6 and 7).

Moreover, ΔNp73α is a strong inhibitor of apoptosis induced by wild-type p53 and TAp73 (Fig. 4 A). HeLa and SaOs2 cells undergo wild-type p53- and TAp73-dependent cell death as assessed by Annexin V staining and TUNEL assay. This apoptotic activity is completely abolished by the coexpression of ΔNp73α (Fig. 4 A and unpublished data). The inhibitory action of ΔNp73α is dependent on the presence of transcription-competent wild-type p53 and TAp73 because ΔNp73α alone cannot affect apoptosis (Fig. 4 A, left column). Furthermore, in agreement with the antiapoptotic effect, ΔNp73α is an inhibitor of colony suppression mediated by wild-type p53 and TAp73 (Fig. 4 B and Table II). The reintroduction of wild-type p53 and TAp73 suppresses the growth of SaOs2 cells (2, 26) and this suppression is thought to be largely due to apoptosis (27). In keeping with these results, the transfection of wild-type p53 strongly suppresses macroscopic colony formation of SaOs2 cells compared with many visible colonies with vector backbone alone (4 foci for wild-type p53 vs. 1,778 for vector control; Fig. 4 B and Table II). As expected, the expression of ΔNp73α alone has no growth-promoting effect compared with vector controls (1,383 vs. 1,778 foci), but was actually slightly growth suppressive in these p53 null cells for reasons that are unclear. In contrast, the coexpression of ΔNp73α with wild-type p53 at a 1:1 molar ratio counteracts this effect, leading to a 12.5-fold increase in the number of colonies from 4 to 51 foci. Likewise, TAp73α, although not quite as potent as wild-type p53, suppresses colony formation (82 foci), but the coexpression of ΔNp73α with TAp73α again antagonizes this effect and increases the number of macroscopic colonies by 8.1-fold to 669 foci. Taken together, ΔNp73 is an efficient dominant-negative inhibitor of wild-type p53 and TAp73, although the strong repressive effect of ΔNp73α on p53 and TAp73 seen in transient assays (Fig. 4 A) is more modest in long-term assays (Fig. 4 B). Entirely consistent with this finding were data from a subsequent p53 expression analysis of SaOs2 cell clones that were established from surviving colonies of a duplicate experiment. A complete loss of full-length p53 protein expression was found in five out of six such randomly picked clones that were derived from plates transfected with wild-type p53 alone (Fig. 4 C, left). Although clones 1, 3, and 5 showed no detectable p53 protein at all, clones 2, 4, and 6 expressed only truncated and presumably nonfunctional p53 polypeptides, likely due to chromosomal rearrangement. The functional significance of the very small amounts of full-length p53 protein detectable in clone 6 is unclear, in light of the presence of coexpressed truncated p53. On the whole, this data is in agreement with the fact that wild-type p53 expression is incompatible with the outgrowth of clones in this assay, and rare colonies escape because they have lost wild-type p53 expression (28). In contrast, all six randomly picked colonies derived from plates cotransfected with wild-type p53 and ΔNp73α exhibited only full-length p53 protein (Fig. 4 C, five clones are shown). Upon sequencing the DNA binding domain of three of these clones, all three revealed wild-type p53 genotypes. All clones expressed elevated levels of ΔNp73α protein compared with endogenous levels (Fig. 4 C). Taken together, this data indicates that the coexpression of ΔNp73α neutralizes the growth-suppressive effect of wild-type p53, thereby removing the selection pressure for the deletion or rearrangement of wild-type p53. ΔNp73α is able to effectively counteract p53- and TAp73-induced colony suppression in transformed human cells.

Because we have shown that ΔNp73 is an efficient inhibitor of exogenous wild-type p53-mediated apoptosis and growth suppression, we next tested whether ΔNp73α also inhibits endogenous wild-type p53-mediated apoptosis after DNA damage. RKO cells, a human colon cancer line harboring wild-type p53 and TAp73, were transfected with empty vector, an irrelevant control plasmid (LcRel), or with ΔNp73α expression plasmid and treated with either 5 μM camptothecin or left untreated (Fig. 4 D). Camptothecin is a topoisomerase inhibitor generating single and double strand DNA breaks, which induce an increase of p53 and TAp73 protein levels (29). As expected, control cells treated with camptothecin and transfected with either empty vector or LcRel underwent marked apoptosis with 44 and 39% of cells dying after 24 h. In contrast, ΔNp73α-expressing camptothecin-treated cells were strongly protected from drug-induced apoptosis, causing only 6% of cells to die. Thus, ΔNp73α protects human cancer cells from p53- and TAp73-induced cell death mediated by a chemotherapeutic agent.

One explanation for the observed dominant-negative effect is a direct physical interaction between ΔNp73 and either wild-type p53 or TAp73 proteins, analogous to the dominant-negative mode of action of mutant p53 proteins toward wild-type p53. To test this hypothesis directly, lysates prepared from p53 null SaOs2 cells cotransfected with wild-type p53 and ΔNp73α were immunoprecipitated with monoclonal antibody ER15, which recognizes ΔNp73α. Immunoblot analysis with an antibody specific for p53 (CM1) revealed a complex of the two proteins (Fig. 5 A, lane 3). Of note, TAp73 isoforms are unable to form a protein complex with wild-type p53 (Fig. 5 A, lane 7, B, lane 3, and C, lane 1; references 18 and 30–32), excluding the possibility that the observed p53 band was coimmunoprecipitated via the endogenous TAp73 protein of SaOs2 cells. As control, no such complex was seen in SaOs2 cells transfected with either empty vector, TAp73α, or ΔNp73α alone (Fig. 5 A, lanes 1, 2, and 4), nor was a complex seen in SaOs2 cells transfected with p53 alone or with p53 plus TAp73α (Fig. 5 A, lanes 6 and 7), indicating the specificity of the p53–ΔNp73α complex. Moreover, a similar complex was seen in wild-type p53-expressing human U2OS cells after transfection with ΔNp73α alone. Fig. 5 B shows a specific complex between endogenous wild-type p53 and ectopic ΔNp73α that was immunoprecipitated by ER15 from U2OS cells (lane 1). No such complex was seen when an irrelevant monoclonal antibody against GFP was used (lane 2), or when TAp73α or empty vector were expressed (lanes 3 and 4). The same specific complex can again be immunoprecipitated from U2OS cells using a monoclonal antibody specific for p53 (421) and immunoblotted with the polyclonal antibody specific for ΔNp73 that does not cross react with any TAp73 proteins (Fig. 5 C, lane 2). Again, no such complex is found with preimmune mouse IgG (Fig. 5 C, lane 3) or when U2Os cells were transfected with TAp73α (lane 1) or vector (unpublished data). The lysate lane in Fig. 5 A represents 5% of the immunoprecipitation input and the lysate lane in Fig. 5 C represents 2–10% of the immunoprecipitation input, depending on the experiment. Together, this data indicates that a small but highly reproducible portion of ΔNp73 forms a stable, specific complex with wild-type p53 in cells. To confirm the existence of this complex in vivo in tumor tissues, we coimmunoprecipitated the ΔNp73–p53 complex from an ovarian carcinoma and an endometrial carcinoma with wild-type p53 status (see example in Fig. 5 D, lane 1). Of note, this mixed complex was only seen in the patients' tumors but not in the corresponding normal ovarian tissues (Fig. 5 D, lane 2). Moreover, no specific complex was seen when the immunoprecipitating anti-p53 antibody beads were substituted by protein G beads (lanes 3 and 4).

We also examined the second possibility, namely direct competition for specific DNA binding between ΔNp73α and wild-type p53 at the promoter level. Gel shift assays were performed with two different p53-specific DNA binding sequences as probes, using nuclear extracts of H1299 or U2OS cells transfected with expression plasmids for p53 alone or in combination with ΔNp73α. Although nuclear extract transfected with empty vector showed no specific DNA complex (Fig. 6 A, lane 1), p53 formed a specific complex (lane 2) that could be supershifted with the p53-specific antibody PAb 421 (lane 3) and competed off with 50-fold excess of unlabeled specific probe (lane 4). Likewise, p53 only binds to the specific probe (Fig. 6 A, lane 9), but not to a scrambled version of the specific probe (lane 11). The specific p53/DNA band (Fig. 6 A, lane 9) is supershifted by antibody 421 (lane 10), but 421 cannot produce a “false” supershifted band when p53 fails to bind to the scrambled probe (lane 12). Extracts from cells transfected with ΔNp73α alone failed to bind to both probes despite high levels of ΔNp73α expression on immunoblots (Fig. 6 A, lane 8, and unpublished data). However, nuclear extracts from cells cotransfected with p53 and ΔNp73α at a 1:4 ratio showed a slight decrease in p53–DNA complex formation (Fig. 6 A, compare lanes 2 and 5), which became more pronounced at a 1:20 ratio of p53 to ΔNp73α (compare lanes 6 and 7). Moreover, in a p53 reporter assay using PG13, a tetramerization-incompetent but DNA binding–competent mutant of ΔNp73α, called ΔNp73 L322P (corresponding to L371P in TAp73α; reference 4), showed a significant but incomplete reversal of the dominant-negative inhibition of p53 transactivation by ΔNp73α (Fig. 6 B). These results support the notion that competition at the promoter site is another mode of inhibition. Taken together, the formation of inactive heterooligomers between p53 and ΔNp73 appear to play a role in the dominant-negative inhibition of p53 by ΔNp73. Such complexes occur naturally in primary tumors and in cultured cells. On the other hand, competition between ΔNp73 and p53 at the level of the promoter, where the presence of ΔNp73 interferes with the ability of p53 to bind effectively to its cognate binding site, might be a second mechanism that is physiologically relevant. Using nuclear extracts of ΔNp73-transfected cells, however, we failed to obtain clear evidence that ΔNp73α alone can directly bind to p53 cognate sites. This issue justifies additional studies.

Our studies demonstrate a dominant-negative action of ΔNp73 toward the apoptosis, focus suppression, and transactivation functions of p53 and TAp73 within the context of forced ΔNp73 expression. To test whether ΔNp73 would have the same inhibitory effect in cells when present at endogenous levels, we used antisense oligonucleotides to suppress endogenous ΔNp73 expression in wild-type p53-harboring RKO cells that undergo apoptosis after DNA damage. RKO cells treated with camptothecin alone underwent a certain level of apoptosis (Fig. 7 A, left bar). In contrast, the preincubation of RKO cells with antisense oligonucleotides directed against the unique exon 3′ of ΔNp73 showed a significant enhancement of p53-mediated apoptosis after camptothecin stress (Fig. 7 A, center bar), whereas camptothecin-stressed RKO cells pretreated with the sense version of the same oligonucleotide did not (Fig. 7 A, right bar). To further confirm that the gain in apoptotic ability seen after down-regulation of endogenous ΔNp73 was in fact due to specific derepression of p53-mediated apoptosis, we modified the above experiment by directly transfecting p53. RKO cells were transfected with wild-type p53 expression plasmid before incubation with antisense or sense oligonucleotides. As already seen in Fig. 7 A, p53-expressing cells treated with antisense oligonucleotides showed significantly enhanced apoptosis compared with the same cells treated with sense oligonucleotides (Fig. 7 B). To confirm that our antisense strategy works, we determined ΔNp73 protein levels. As shown in Fig. 7 C, transfection of the antisense oligonucleotide clearly down-regulated endogenous ΔNp73 protein levels compared with the sense control oligonucleotide. Taken together, these studies clearly indicate that endogenous ΔNp73 exerts a significant transdominant inhibition of p53 function. Down-regulation of endogenous ΔNp73 levels alleviates its suppressive action on p53-mediated apoptosis after DNA damage.

p53 controls a powerful stress response by integrating signals from many types of DNA damage or inappropriate oncogenic stimulation. Activation of p53 elicits a cellular response of apoptosis, cell cycle arrest, or senescence, thereby preventing tumor formation. Therefore, direct or indirect loss of p53 function is a preeminent defect in most human cancers whether via intragenic mutation of p53 itself (33), lack of p53 nuclear retention (34), loss of its upstream activator p14ARF (35), or inhibition by its antagonist HDM2 (36, 37). Here we show that the human TP73 gene, a homologue of p53, produces an NH₂ terminally truncated isoform driven by an alternative internal promoter in intron 3. Human ΔNp73 starts with a unique exon 3′ that is highly conserved from mouse exon 3′ and is in frame with exon 4. ΔNp73 lacks the transactivation domain and is therefore predicted to function as a transdominant inhibitor of p53. In physical and functional assays, we demonstrate dominant-negative interactions between human ΔNp73 and wild-type p53 or transcription-competent TAp73. ΔNp73 is a potent inhibitor of wild-type p53 and TAp73 with respect to their transcriptional activation, apoptotic ability, and growth suppressor function. Of note, ΔNp73-mediated interference with the biological p53 response occurs at endogenous ΔNp73 levels, because antisense-directed down-regulation of endogenous ΔNp73 expression leads to a derepression of the p53 response.

Importantly, in this study we provide the first clinical evidence that ΔNp73 is frequently up-regulated in a variety of primary cancers including cancers of the breast and the female genital tract. We show that tumor-specific up-regulation of ΔNp73 occurs at the mRNA and protein level in primary tumors. In a rigorous comparison of tumor/normal tissue pairs from 37 patients, 73% show ΔNp73 up-regulation specifically in their tumor tissues but not in their respective normal tissues of origin. In addition, a subset of those tumors also up-regulate the previously described Ex2Del p73, another transdominant inhibitor of p53, but one that is generated from the P1 promoter of TP73 via alternative splicing. Taken together, 81% of tumor pairs exhibited tumor-specific up-regulation of ΔNp73 and/or Ex2Del p73. Of note, among the tumors with up-regulation of any one or all three TP73 transcripts (ΔNp73, Ex2Del p73, and TAp73), 71% exhibited preferential up-regulation of ΔNp73 or Ex2Del p73. Moreover, two thirds of the latter group of tumors showed exclusive up-regulation of ΔNp73 and/or Ex2Del p73 without any concomitant rise in TAp73. Furthermore, among tumors with preferential up-regulation of ΔNp73 and Ex2Del p73, 71% harbored wild-type p53. In contrast, among the (small) group of tumors with preferential TAp73 up-regulation, 66% harbored mutant p53. In summary, given the trend between tumor-specific up-regulation of ΔNp73 and/or Ex2Del p73 on the one hand and wild-type p53 status on the other, it is tempting to speculate that the expression of dominant-negative p73 isoforms alleviates the selection pressure for p53 mutations in tumors by functionally inactivating the suppressor action of p53. More tumor samples need to be analyzed to clarify this question. If confirmed, however, this mode of p53 inactivation would be functionally equivalent to the amplification of the HDM2 gene in some human sarcomas and leukemias (36, 37). While this work was in progress, Grob et al. (38) reported that p53 and TAp73 can directly induce ΔNp73 via a p53-responsive element in the ΔNp73 promoter in cultured cells. Moreover, Nakagawa et al. (39) confirmed that TAp73 directly transactivates ΔNp73, but such an activity for p53 could not be found by this group. Nevertheless, these results further fortify the analogy to the p53-HDM2 autoregulatory feedback loop. It will be interesting to determine whether p53 and/or TAp73 itself drive the overexpression of ΔNp73 in primary tumors as well.

We provide evidence that physical interaction between ΔNp73α and wild-type p53 is one of the possible functional mechanisms of this inhibition. We showed that this mixed complex occurs naturally in cultured human tumor cells and tumor tissues. While this work was under review, Nakagawara et al. (39) also reported that ΔNp73α and ΔNp73β form stable mixed complexes with endogenous and ectopic wild-type p53, as well as with TAp73α. Similarly, Pozniak et al. (18) reported a direct physical interaction of mouse ΔNp73α and ΔNp73β proteins with mouse wild-type p53, both of which inhibited the apoptosis-promoting activity of p53. Thus, heterocomplexes appear to interfere with the specific DNA binding activity of wild-type p53. In contrast to ΔNp73, TAp73 isoforms are unable to form a protein complex with wild-type p53 (Fig. 5, B and C; references 18 and 30–32). Therefore, the existence of mixed complexes suggests that the unique exon 3′ at the NH₂ terminus of ΔNp73, in conjunction with the missing transactivation domain, induces a conformational change in ΔNp73 that allows the complex to form. In contrast, another recently described isoform named ΔTA-p73α, which is encoded by p73 (exons 4–14) but missing the unique exon 3′, does not appear to form a complex with wild-type p53 (40). Heterooligomeric complexes mirror the ability of many missense p53 mutants in heterozygous tumors to abrogate the function of the remaining wild-type p53 allele (41, 42). Of note, a stoichiometry as low as 1:3 of mutant to wild-type p53 molecules already abrogates wild-type p53 function, suggesting that a similar stoichiometry might be effective for ΔNp73 as well. An additional mechanism of p53 inhibition might be direct promoter competition, with ΔNp73 displacing p53 from the DNA binding site (43). Such a phenomena has been seen in gel shift assays with high concentrations of in vitro translated p73 (exons 4–14) protein (ΔTA-p73α; reference 40). In nuclear extracts of H1299 and U2OS cells cotransfected with p53 plus ΔNp73, which more closely mimics the physiological situation than in vitro translated proteins, we also observed a decrease in p53–DNA complex formation in the presence of ΔNp73α, supporting the notion of promoter competition. Under these experimental conditions, however, we failed to obtain clear evidence that ΔNp73α alone can directly bind to p53 cognate sites. Although both inhibitory mechanisms are likely, this point needs additional study.

Responsiveness to oncogenes and selected forms of DNA damage might suggest a putative tumor suppressor role of the TP73 gene analogous to TP53. However, tumor-associated overexpression of (total) p73 and in some cases TAp73 rather than loss of expression, is the most commonly observed tumor-specific abnormality of the TP73 gene. This fact, in conjunction with a conspicuous lack of p73 mutation in human tumors and a lack of a cancer phenotype in p73-deficient mice, provides clear evidence that the TP73 gene is not a classic Knudson-type tumor suppressor in vivo. Instead, our finding that a significant percentage of human tumors specifically select for dominant-negative p73 isoforms strongly argues for their oncogenic role during tumorigenesis in vivo. Preferential up-regulation of ΔNp73 and Ex2Del p73 in tumors might bestow oncogenic activity upon the TP73 gene that specifically interferes with the tumor suppressor functions of wild-type p53 and TAp73. In fact, this scenario suggests that the TP73 gene embodies the “two genes in one” idea with products that play opposing roles within the family circuitry. Their impact on tumor formation in vivo might therefore depend on the balance between tumor-promoting and -suppressing family members. The existence of this inhibitory family network might explain the paucity of p73 mutations in human tumors. In the developing mouse brain, ΔNp73 is the predominant form of TP73 and a powerful inhibitor of p53 (18). In vivo and in vitro studies showed that ΔNp73 plays an essential antiapoptotic role required to counteract p53-mediated neuronal death during the “sculpting” of the developing brain. This explains why p73^−/− mice, missing all forms of p73 including protective ΔNp73, underwent accelerated neuronal death in sympathetic ganglia (18). Thus, in keeping with a common theme in cancer, a developmental inhibitory network that is essential for normal physiology resurfaces in cancer, but in a corrupted and derailed fashion. In support of the idea that ΔNp73 can act as an oncogene in vivo, overexpression of ΔN isoforms of p63, a p73 homologue, is found in human cancers including lung cancer, squamous cell carcinoma of the head and neck (44), bladder cancer (45), and nasopharyngeal carcinoma (46). Importantly, a specific ΔNp63 isoform is oncogenic in nude mice and in the Rat 1a focus formation assay (44).

Evidence from some human cancers and mouse models indicates that the mutational status of p53 is an important clinical variable for prognosis as well as for guiding the aggressiveness of anticancer therapy. However, currently, p53 mutational status is not widely accepted as a clinical indicator because many studies show that its prognostic power for survival and its predictive value for therapeutic response is poor (for review see reference 47). From a practical standpoint, the discovery of a ΔNp73-based p53–p73 interference network suggests that the p53 status of a tumor should no longer be considered in isolation. The existence of this inhibitory network might account for the inconsistencies of clinical studies that sought to use p53 status as a predictor of outcome. It mandates that we do a careful analysis of the functional consequences of this network in vivo. The establishment and clarification of an inhibitory p53–p73 network would have a major clinical impact ranging from fine tuning the prognostic power of p53 mutation status to rational p53–p73-targeted drug design.

In conclusion, we report here the cloning of human ΔNp73 and show that ΔNp73 mediates a novel inactivation mechanism of wild-type p53 and TAp73 via dominant-negative interference. Deregulated expression of ΔNp73 can bestow oncogenic activity upon the TP73 gene by functionally inactivating the suppressor action of p53 and TAp73. This trait might be frequently selected for in a variety of human cancers.

We thank T. Shirangi for technical assistance.

This study is supported by grants from the National Cancer Institute and the United States Army Medical Research Command.