Switching Mechanisms of Cell Death in mdm2- and mdm4-null Mice by Deletion of p53 Downstream Targets

p53 maintains genome integrity in response to DNA-damaging agents such as UV radiation (1), γ radiation (2, 3), and chemotherapeutic agents (4), causing cell cycle arrest or apoptosis through transcriptional regulation of its target genes. Consequently, p53 mutations and disruption of the p53 pathway are common events in the genesis of human tumors (5). p53 stimulates the transcription of numerous genes, among them p21 (p21^waf1/cip1), which encodes an inhibitor of the cell cycle (6, 7). Other p53 targets include bax, fas, PERP, Noxa, p53AIP1, Apaf1, and PUMA, encoding proteins that are important in apoptosis (8, 9, 10, 11, 12, 13, 14, 15). p53 also activates the mdm2 gene, which encodes a protein that inhibits p53 function (16, 17, 18, 19).

MDM2 is an E3 ubiquitin ligase, the last link in a complex that ubiquitinates p53, thus tagging p53 for degradation by the proteasome (20, 21, 22, 23). The mdm2 gene is amplified in 36% of human sarcomas (24, 25, 26), as well as in other cancers, including leukemias (27) and malignant gliomas (28, 29). Overexpression resulting from unknown mechanisms is also observed in a myriad of tumors (30). Increased levels of the p53 inhibitor MDM2 in tumors suggest an alternate mechanism of inactivating p53. That MDM2 is a critical inhibitor of p53 in vivo was shown by disruption of mdm2 in the mouse. The embryonic lethality of mdm2-null mice is completely rescued by the loss of p53, indicating that the unregulated expression of active p53 is responsible for the death of mdm2-null embryos (31, 32).

The mdm4 gene is a new member of the mdm2 family (33). mdm4 is amplified in some gliomas that contain wild-type p53 (34). MDM4 also binds p53 and inhibits p53 transcriptional activation (33, 35). That MDM4 is a bona fide inhibitor of p53 came from studies in mice. Inactivation of mdm4 results in an embryonic lethal phenotype by 7.5 days postcoitum (dpc) that is again completely rescued by the concomitant deletion of p53 (36, 37, 38). mdm4^−/− mutant embryos die as a result of the loss of proliferative capacity. Thus, MDM4 is another important regulator of p53 during development.

Because the mdm2- and mdm4-null lethal phenotypes are p53 dependent, they offer a unique system to assay the importance of p53 downstream targets. First, the timing and mechanism of embryonic cell death in mdm2^−/− mice was determined to occur at 3.5 dpc by apoptosis. This phenotype is partially rescued by loss of the p53 target bax. The p53-dependent mdm4-null proliferation defect was not rescued in the absence of p21; however, loss of p21 altered the mechanism of death from cell cycle arrest to apoptosis.

To determine mouse genotypes, we extracted DNA from tail biopsies. PCR analysis was performed using published primer sets for: mdm2 (31), p53 (39), p21 (40), and mdm4 (36). The primers for bax were (a) 5′-gggttgaccagagtggcgtagg-3′, (b) 5′-gagctgatcagaaccatcatgg-3′, and (c) 5′-acccgcttccattgctcagcgg-3′,⁴ of which a and b amplify the wild-type allele, whereas a and c amplify the mutant allele. PCR analysis for mdm2, p53, bax, and p21 was conducted using annealing temperatures of 65°C, and PCR for mdm4 was conducted at 60°C, for 35 cycles with an extension time of 3 min.

Mouse blastocysts were digested in 1× PCR buffer (Boehringer Mannheim, Indianapolis, IN), 1 mg/ml proteinase K, and 0.1% (v/v) Triton X-100 (Sigma Chemical Company, St. Louis, MO) for 1 h at 55°C, followed by proteinase K inactivation by heating at 95°C for 5 min. The entire volume of DNA was subjected to PCR.

Embryos at 5.5–9.5 dpc were fixed overnight in 10% (v/v) phosphate-buffered formalin and were embedded in paraffin. Sections, 7 μm thick, were mounted on Superfrost Plus slides (Fisher Scientific, Pittsburgh, PA) and were stained with H&E. Proliferating cell nuclear antigen (PCNA) was detected in 6.5-dpc embryos using the PCNA staining kit (Zymed Laboratories, San Francisco, CA).

Blastocysts were collected by flushing from the uterus using DMEM with 10% FCS and 1 mm HEPES. TUNEL (41) was carried out on blastocysts using the in situ cell death detection kit POD (Boehringer Mannheim, Indianapolis, IN). For older embryos, sections were treated with 20 μg/ml proteinase K for 15 min., washed, and then incubated in 3% (v/v) H₂O₂ in methanol for 5 min to inactivate endogenous peroxidases. After a 2-min incubation in deoxynucleotidyltransferase buffer [30 mm Tris HCl (pH 8.0), 140 mm sodium cacodylate, and 1 mm CoCl₂], sections were labeled with biotin-conjugated dUTP (1:200) using terminal deoxytransferase (1:400) for 45 min at 37°C. Labeling was detected using 3,3′-diaminobenzidine (Vector Labs, Burlingame, CA).

The statistical significance of differences in apoptosis in blastocysts from mdm2^+/− × mdm2^+/− crosses was determined using the test for equality of proportions. The statistical significance of the differences in partial rescue of the mdm2-null phenotype in a bax-null background was determined using the χ² test.

To determine the mechanism of mdm2-null lethality and the role of p53 downstream genes, we first attempted to discern whether the few remaining cells in the empty decidua (5.5 dpc) were of embryonic origin. We, therefore, crossed a female mouse heterozygous for mdm2 with a Rosa 26 male mouse that carried a ubiquitously expressed β-galactosidase gene inserted into chromosome 6 (42). Male mice heterozygous for both β-galactosidase and mdm2 were then crossed with female mice heterozygous for mdm2. These females were sacrificed at 5.5 dpc and the deciduae treated with 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-gal). Of 30 deciduae analyzed, 25% were empty and none contained blue-stained cells, suggesting that, at 5.5 dpc, the empty decidua is truly devoid of embryonic tissue (data not shown). One-half of the phenotypically normal embryos stained blue as a positive control.

To examine embryos earlier in development, mdm2 heterozygous mice were mated and embryos were collected at the blastocyst stage (3.5 dpc) before implantation. Of 84 blastocysts genotyped, 21 were mdm2-null (data not shown), the expected Mendelian ratio. To determine whether mdm2-null embryos demonstrated a phenotypic abnormality at this stage, we examined blastocysts microscopically, categorized them by developmental stage, and then genotyped them. Interestingly, of the 28 blastocysts analyzed in this experiment, none that were null for mdm2 had hatched (Fig. 1). Hatching is the process by which the developing embryo erupts from the zona pelucida, a membrane that surrounds the early embryo (43). This lack of hatching suggested that the mdm2-null phenotype actually appears as early as the blastocyst stage.

Because the mdm2-null phenotype is rescued by deletion of p53, the death of the embryo is due to unregulated p53 activity. Because p53 induces cell cycle arrest or apoptosis depending on factors such as cell type and cell cycle stage, we next determined which of these two mechanism caused mdm2-null lethality. To this end, mdm2 heterozygous mice were mated and the females sacrificed at 3.5 dpc. Blastocysts were then tested for apoptosis using the TUNEL assay, which stains the nuclei of cells committed to apoptosis. Because up to 10% of wild-type cells from a blastocyst can undergo apoptosis (44), crosses between wild-type mice were analyzed to determine whether this apoptosis could be detected by the TUNEL assay. Of the 16 wild-type embryos tested, a maximum of 5 of the 60–70 cells in the blastocyst were TUNEL positive, indicating that the normal apoptotic number was less than 10% (Fig. 1). Of 73 blastocysts analyzed from mdm2 heterozygous crosses, 17 (23.2%) demonstrated excess apoptosis (Table 1). The affected blastocysts generally displayed a large number of apoptotic nuclei, ranging from 10 to 30 positive nuclei in any one blastocyst (Fig. 1). Several attempts were made to genotype the blastocysts after TUNEL without success. However, only the population containing mdm2-null embryos had abnormal levels of apoptosis, suggesting that the loss of mdm2 initiated a p53-dependent apoptotic cell death.

p53 initiates apoptosis through the activation of its downstream target genes. One of these targets, bax, is a member of the bcl-2 gene family that positively regulates apoptosis (19, 45, 46). To determine whether the mdm2-null phenotype could be rescued in a bax-null background, crosses between mdm2^+/− bax^+/− males and mdm2^+/− bax^−/− females were analyzed (bax-null male mice are sterile). All of the crosses were performed in a mixed C57BL/6 and 129Sv backgrounds. From these matings, one of eight pups is expected to be null for both bax and mdm2. Of 67 mice analyzed, none were null for both genes, suggesting that the absence of bax, unlike the absence of p53, did not rescue the mdm2-null lethality.

Although the absence of bax could not fully rescue the mdm2-null lethality, the possibility remained that the absence of bax could partially rescue the phenotype. To explore this possibility, we crossed mdm2^+/− bax^+/− males with mdm2^+/− bax^−/− females and analyzed the embryos at various stages of development. Embryos at 5.5 and 6.5 dpc were too small to be accurately genotyped and, thus, were examined histologically. Of 109 deciduae examined at 5.5 and 6.5 dpc, we observed only one-half of the expected number of empty deciduae, which suggested a partial rescue of the mdm2-null phenotype (Table 2). By 7.5 dpc, we observed the expected normal:abnormal embryo ratio. At 7.5 dpc, genotypes were determined for 42 of the embryos, none of which was null for bax and mdm2 (data not shown). These data indicated that lethality in embryos that were null for both bax and mdm2 occurred between 6.5 and 7.5 dpc, instead of at 3.5 dpc as seen in embryos null for mdm2 alone.

To determine whether these embryos were still dying by apoptosis, we performed the TUNEL assay on sections of 6.5-dpc embryos from the above cross. Of 19 embryos examined, none revealed TUNEL-positive cells. As a positive control, mouse intestine that was fixed and sectioned by the same method stained positively in the crypt cells (data not shown). If apoptosis had been the cause of lethality in the mdm2-null embryos at 6.5 dpc, two or three embryos should have been abnormal by the TUNEL assay. To assess whether mdm2-null embryos were dying by cell cycle arrest in a bax-null background, embryo sections were stained for proliferating cell nuclear antigen (PCNA), a marker for DNA synthesis. Normal embryos at 6.5 dpc were proliferating and stained strongly for PCNA (Fig. 2). Of the 28 embryos tested from a mdm2^+/− bax^+/− × mdm2^+/− bax^−/− cross, 22 stained strongly for PCNA, 5 stained only weakly, and 1 did not stain at all (Fig. 2). The number of weakly staining embryos corresponded roughly to the number of embryos expected to be null for both bax and mdm2. This weak staining may have been due to the long half-life of PCNA (47). Maternal cells in the proliferating zone of the deciduae of all of the embryos consistently exhibited strong staining, providing an internal control for the experiment. These data suggest that cell cycle arrest and not apoptosis is the cause of mdm2^−/− lethality in a bax-null background.

p53 also initiates cell cycle arrest in response to DNA damage through transcriptional activation of p21 (p21^waf1/cip1). mdm4-null embryos die by cell cycle arrest in a p53-dependent manner. To determine whether the mdm4-null phenotype could be rescued by the deletion of p21, we generated mdm4^+/− p21^−/− mice. These mice were mated to each other and the offspring was genotyped. We expected one of four mice to be double null for p21 and mdm4. Of 57 mice analyzed, none were null for both genes. These data suggest that the mdm4 lethal phenotype cannot be rescued by the absence of p21 as it was by the absence of p53.

To determine whether a partial rescue of the phenotype occurred, mdm4^+/− p21^−/− mice were crossed with each other and pregnant females were sacrificed at different stages of development. Table 3 shows the genotyping of 8.5 and 9.5 dpc embryos. All of the abnormal embryos were mdm4^−/− p21^−/− (data not shown), as were mdm4^−/− embryos, which suggests that no partial rescue of the mdm4-null phenotype occurred.

To address the mechanism of the mdm4^−/− embryo lethality in a p21-null background, we analyzed embryos at 7.5 dpc, the time at which mdm4^−/− embryos show proliferative arrest. Embryo sections were stained for PCNA, and normal embryos (Fig. 3,B), as well as abnormal embryos null for mdm4 and p21 (Fig. 3,E), were strongly positive for PCNA, whereas mdm4^−/− mutant embryos stained weakly at best (Fig. 3,H). These data indicated that the mdm4^−/− embryos were not dying by cell cycle arrest in the absence of p21. To determine whether mdm4^−/− p21^−/− embryos were dying by apoptosis, embryo sections were stained using the TUNEL assay. Normal embryos and mdm4^−/− mutant embryos were negative for apoptosis (Fig. 3, C and I). However, the abnormal embryos from a cross between mdm4^+/− p21^−/− mice were TUNEL positive (Fig. 3 F). These data indicated that in contrast to mdm4^−/− embryos, mdm4/p21 double-null embryos were dying by apoptosis and not by cell cycle arrest.

Loss of either mdm2 or mdm4 results in an embryo lethal phenotype as a result of deregulated expression of active p53. As such, these mouse models represent in vivo systems to dissect the importance of components of the p53 pathway. In this study, we first observed that death in mdm2-null embryos was due to apoptosis. These data have been recapitulated in tissue culture in which cells null for both mdm2 and p53 (mdm2-null cells never grow unless p53 is also missing) died by apoptosis on reintroduction of p53 (48). Normally mouse embryo fibroblasts need to be sensitized to apoptosis by transfection with adenovirus E1A, but in this scenario, simple loss of mdm2 resulted in p53-dependent apoptosis. To determine the importance of bax, a p53 target gene, we crossed the mdm2 and bax mice to each other. Deletion of bax allowed mdm2^−/− embryos to live 3 days longer and remarkably altered the mechanism of cell loss to that of proliferative arrest (the mouse phenotypes are summarized in Table 4). It is unlikely that we missed an apoptotic response in these embryos because later time points result in complete absence of an embryo.

mdm4-null embryos, on the other hand, undergo a p53-dependent proliferative arrest during embryogenesis (36). In our study, we showed that the lethal phenotype was not delayed by the absence of the p21, a p53 target gene that initiates cell cycle arrest (6, 7). However, we found that in the absence of p21, the mechanism of cell death changed from cell cycle arrest to apoptosis. These data complement findings in human tumor cell lines, showing that increased levels of p21 protect against apoptosis (49, 50). Therefore, p21^waf1/cip1 is an important component of the p53 cell cycle arrest pathway in this in vivo model.

How does deletion of different p53 inhibitors alter the pathway chosen by p53? One possibility is that the p53 inhibitors, MDM2 and MDM4, bind different modified versions of p53. MDM2 binds a p53 poised to initiate apoptosis, and, thus, the loss of mdm2 results in that p53 initiating apoptosis. Similarly, MDM4 binds a p53 modified to initiate cell cycle arrest, and the loss of mdm4 results in proliferative arrest. This hypothesis implies that the ratio of MDM2:MDM4 determines which p53 downstream event is initiated.

The data provided in this study suggest that p53 is very versatile and that, in the absence of one of its downstream targets, it chooses another pathway. Because the promoters of bax and p21 are intact in the respective knockouts (51, 40), loss of p53 DNA-binding sites does not contribute to this decision. It implies a feedback mechanism in which p53 senses the lack of a cellular response and chooses another alternative. Thus, tumor cells in which wild-type p53 has been reintroduced are likely to respond by whatever mechanism to an active p53. To date, deletion of any of the known p53 targets has not halted the p53-dependent response, which suggests that p53 has multiple alternatives to achieve its ends.

We thank R. Behringer (University of Texas M.D. Anderson Cancer Center, Houston, TX) for Rosa 26 mice.