The R246S hot-spot p53 mutant exerts dominant-negative effects in embryonic stem cells in vitro and in vivo

Missense mutations in p53 are the most common genetic alterations seen in human cancers (Olivier et al., 2004; Vousden and Lu, 2002). It is thought that these mutations lead to defects in the tumour-suppressive properties of p53 and further contribute to carcinogenesis through either acquiring novel gain-of-function (GOF) properties or through the exertion of dominant-negative (DN) effects over the remaining wild-type allele, as was proposed more than a decade ago (Oren, 1992). In addition, loss of heterozygosity (LOH) of the remaining wild-type p53 allele has been noted in both familial and sporadic cancers containing missense mutations (Nishida et al., 1993; Varley et al., 1997; Forslund et al., 2002; Fenoglio-Preiser et al., 2003). The degree to which both LOH and missense p53 mutations occur in the same tumours varies, and LOH has been suggested to depend on both the mechanism of genotoxicity of the carcinogenic agent and the tissue type (Nishida et al., 1993; Venkatachalam et al., 2001; Forslund et al., 2002; Fenoglio-Preiser et al., 2003). This suggests that total loss of p53 activity, rather than acquisition of GOF or DN properties, might also be more pertinent to tumorigenesis. Nonetheless, evidence points to all three mechanisms being equally important in contributing to the carcinogenic process, although the specific context in which each mechanism operates is unclear.

The phenomenon of DN effects of missense p53 mutants has been intensely studied, mainly through analyzing the properties of the commonly found hot-spot DNA-binding-domain mutations such as R175H, G245S, R248W, R249S, R273H and R282W (Petitjean et al., 2007). Albeit that the DN effects have been extensively described, conflicting data exist to the contrary. Several investigators have used multiple read-outs, including arrest of cellular growth, cellular survival and transactivation ability, the latter using promoter-reporter assays and target gene expression, to evaluate the DN effect of missense mutations. The results have varied, which could be due to the type of cells used as well as to the chosen method of expressing mutant p53, because many initial studies employed transient-transfection protocols (Chan et al., 2004; Davis et al., 1996; Williams et al., 1995). Later studies employed inducible constructs as well as combi-constructs expressing both wild-type and mutant cDNAs in a single vector (Aurelio et al., 2000; Willis et al., 2004). Nonetheless, all the systems thus far have the setback of expressing mutant and wild-type p53 at experimentally high levels, perhaps not precisely reflecting the physiologically relevant situation.

Mice models have also been used to study this phenomenon. For example, the offspring of transgenic mice containing the p53 mutant A135V that were crossed to p53^+/– mice developed a higher incidence of tumours compared with p53^+/– mice without the transgene, highlighting the DN effects of the exogenous mutant p53 over the endogenous wild-type one (Harvey et al., 1995). In addition, recent studies using `knock-in' mice that express physiologically relevant levels of the mutant protein rather than causing its overexpression demonstrated that the mice expressing one allele of the mutant R172H, the human R175H equivalent, had increased rates of metastasis compared with p53^+/– mice, although both had a similar tumour spectrum and survival curve, suggesting that the mutant allele has DN effects (Lang et al., 2004). However, this mutant was not observed as having effective DN properties to promote K-ras-initiated lung adenocarcinomas, in contrast to another hot-spot mutant, R270H, the mouse equivalent of the human R273H mutation, which displayed partial DN activity in this context (Jackson et al., 2005). However, LOH was observed in tumours from both knock-in mice, indicative of residual tumour-suppressive function conferred by the remaining wild-type allele of p53 (Jackson et al., 2005; Lang et al., 2004). Nonetheless, in other cellular contexts, the R270H mutant displayed stronger DN properties: epithelial-specific expression of R270H in the heterozygote state results in an increased incidence of spontaneous and ultraviolet (UV)B-induced skin tumours, affecting latency, multiplicity and progression, although this was not the case with respect to spontaneous tumours in mice expressing the mutant p53 in all tissues (Wijnhoven et al., 2007). Moreover, DN effects were seen in embryonic stem (ES) cells heterozygous for R270H and in primary cells derived from p53^+/R270H and p53^+/R172H mice (de Vries et al., 2002), suggesting that the DN effects might be cell-type and even signal specific.

One hot-spot p53 mutation that has a very high tissue-type association is the R249S mutation (Petitjean et al., 2007), which is strongly associated with hepatocellular carcinomas (HCC) in areas of high exposure to the dietary aflatoxin B₁ (AFB₁) (Staib et al., 2003). It is interesting to note that HCCs occurring in areas not-exposed to AFB₁ do not carry the R249S mutation, indicating a high level of specificity that is required for this mutation to occur (Staib et al., 2003). Treatment of mice and rats with AFB₁ have failed to recapitulate the R246S (mouse equivalent of human R249S) mutation in p53, even in Hupki `knock-in' mice, which carry the human p53 locus in their germ line, although animals did succumb to HCCs, suggesting the need for yet-undiscovered mechanisms in the generation of this mutation (Ghebranious and Sell, 1998; Hulla et al., 1993; Tong et al., 2006). Nonetheless, human HCCs with the R249S mutation appear to be more aggressive than those without this mutation, highlighting an important role for this mutation in liver carcinogenesis (Oda et al., 1992; Oda et al., 1994). Analysis of R249S-containing HCCs has revealed both a strong correlation and a lack of correlation with LOH, suggesting that both R249S DN-dependent and -independent mechanisms might be at work in HCC formation (Li et al., 1993; Martins et al., 1999; Oda et al., 1992; Peng et al., 1998).

To ascertain the role of R249S in tumorigenesis and its DN activity in vivo in a physiological context, we generated mouse ES cells carrying a `knock-in' allele of the human-R249S-equivalent mutation, R246S, by homologous recombination. Data presented here reveal that this mutation acts in a DN manner when assayed for gene transcription, cell death and tumour formation in vivo.

We generated feeder-independent ES cells expressing the R246S mutation at physiological levels by gene targeting. The targeting construct contained a loxP-site-flanked (floxed) neomycin selection cassette and carried the R246S mutation, which was introduced by site-directed mutagenesis (Fig. 1A). G418-resistant ES cell clones were first screened by PCR using two sets of primer pairs priming either at the 5′ or 3′ end, respectively, of the endogenous p53 locus outside of the targeting construct and on the gene encoding neomycin resistance (data not shown). PCR-positive ES cell clones were further screened by Southern blot hybridization, which gave 17 kb and 8 kb bands for wild-type and mutant alleles, respectively (Fig. 1B, left panel). The homologous-recombination frequency rate was 5.6%. Positive recombinants were further transfected with Cre recombinase to remove the neomycin selection cassette, and the removal was confirmed in G418-sensitive clones by Southern blot hybridization using exon 1 as an external probe (11 kb with and 9 kb without the neomycin cassette) (Fig. 1B, right panel). The presence of the R246S mutation in the genome was confirmed by PCR-RFLP by digestion with BsrBI (data not shown) (please see Materials an Methods for details). Two independently targeted R246S knock-in ES cell clones were analyzed, which gave similar results in all subsequent experiments.

The expression of the mutant allele was first determined by reverse transcriptase (RT)-PCR-RFLP. Similar to other findings (Mendrysa et al., 2003), the presence of the selection cassette strongly suppressed the expression of the targeted allele, which became prominent after removal of the neomycin cassette (Fig. 1C). To confirm the protein expression from the mutant allele, we performed immunoprecipitation analysis using the conformation-specific anti-p53 antibodies, Pab240 (mutant) and Pab246 (wild type) (Sabapathy et al., 1997). Human H1299 cells stably expressing the hot-spot mutant R175H p53 were used as a positive control for the mutant conformation (Fig. 1D, lane 7). Comparison of p53 conformation in p53^+/+ and p53^+/R246S ES cells indicated that p53 adopted a wild-type conformation in both cases (Fig. 1D, compare lanes 2 and 4). To rule out any effects of the wild-type p53 protein over the mutant protein in p53^+/R246S ES cells, we used p53^–/R246S primary embryonic fibroblasts, which do not express any wild-type protein. We noted that the R246S mutant p53 protein adopted the wild-type conformation as well (Fig. 1D, lane 6), consistent with previous reports investigating the conformation of this R246S mutant p53 protein (Ghebranious et al., 1995), indicating that the mutant allele was indeed expressing the mutant protein in a wild-type conformation. Subsequent analysis of the steady-state levels revealed that the level of p53 protein in p53^+/R246S ES cells was higher than in wild-type cells, both before and after doxorubicin treatment (Fig. 1E, compare lane 4 with 1 and 5, and lane 9 with 6 and 10). Furthermore, p53 in p53^+/R246S ES cells localized to the nucleus in both normal and stress conditions, in contrast to in p53^+/+ cells, in which p53 was in the nucleus only after exposure to γ-irradiation (Fig. 1F). The elevated levels and abnormal localization of p53 in p53^+/R246S ES cells is reminiscent of mutant p53 expression found in human cancers (Soussi, 2000), suggesting that the mouse R246S mutant p53 resembles the biochemical characteristics of human cancer-cell-derived mutant p53.

To evaluate whether the R246S mutant would exert DN effects over the remaining wild-type protein, we first determined the expression of several p53 target genes, such as noxa, p21 and mdm2, before and after genotoxic stress, by quantitative real-time RT-PCR. As expected, the expression of these genes was rapidly induced in a p53-dependent manner in p53^+/+ and p53^+/– ES cells upon doxorubicin treatment and γ-irradiation (Fig. 2A and data not shown). However, the induction of these genes in p53^+/R246S ES cells was similar to that observed in p53^–/– cells, and was much lower compared with wild-type and p53^+/– cells (Fig. 2A), suggesting that the transactivation ability of wild-type p53 was impaired in the presence of the R246S mutant, therefore demonstrating the DN effect over the wild-type protein. It is to be noted that, upon doxorubicin treatment, p53 was abundant in p53^+/R246S cells (Fig. 1E), yet it was unable to result in the activation of target gene expression. These results were further confirmed by northern blot analysis of mdm2 expression (Fig. 2B). Next, analysis of short-term cellular survival rates upon exposure to genotoxic stress revealed that p53^+/R246S ES cells were almost as resistant as p53^–/– cells to doxorubicin (Fig. 2C) and UV (Fig. 2D) treatment over a range of doses, as indicated. However, consistent with previous reports indicating a lack of a role for p53 in the long-term survival of ES cells (Chao et al., 2000), cell counting (supplementary material Fig. S1A) and colony-formation assays (supplementary material Fig. S1B) revealed that p53 status did not affect the long-term survival of ES cells after irradiation. Together, these results demonstrate the DN effect of the R246S mutant protein over its wild-type counterpart in p53^+/R246S ES cells, with respect to target gene activation and short-term cellular survival.

We next investigated whether mutant p53 would also exhibit DN effects in differentiated cells. Molecular analysis of differentiation-related genes confirmed that knock-in cells were as pluripotent as wild-type cells, as determined by the expression of nanog, rex 1, oct3/4, gata4 and pax6 (Fig. 3A, left panel), and morphological analysis showed that cultures had a typical round colony-like morphology (Fig. 3A, right panel). The rate of differentiation, as determined by the changes in nanog and oct3/4 levels, was also found to be similar between wild-type, p53^+/R246S and p53^–/– cells (Fig. 3B). Similar results were obtained upon DMSO-induced differentiation (data not shown), suggesting that the expression of R246S mutant p53 does not affect the pluripotency and differentiation potential of the ES cells; these results are congruent with the lack of differentiation defects in p53^–/– mice and in cells from other p53 hot-spot mutant knock-in mice (Jacks et al., 1994; Lang et al., 2004; Olive et al., 2004). Immunofluorescence analysis revealed that the R246S mutant protein localized to the nucleus of differentiated ES cells even in the absence of any genotoxic stress (Fig. 3C), similar to that noticed in undifferentiated ES cells, indicating that R246S mutant p53 might also exert DN effects over wild-type p53 in the differentiated state. We therefore examined the ability of p53^+/R246S cells to undergo p53-dependent cell cycle arrest upon γ-irradiation, as has been shown for differentiated wild-type ES cells (Chao et al., 2000). As expected, the proportion of S-phase cells dramatically reduced after γ-irradiation in wild-type cultures but not in p53^–/– cultures (% reduction of S-phase cells in p53^+/+ vs p53^–/– cultures: 68.59±1.13% vs 23.34±5.11%) (Fig. 3D). Importantly, p53^+/R246S cultures did not show a significant reduction in the proportion of S-phase cells, similar to that observed with p53^–/– cultures (p53^+/R246S vs p53^–/– vs p53^+/– cells: 29.11±5.47% vs 23.34±5.11% vs 58.74±0.88%), suggesting that cell cycle regulation of p53^+/R246S cells was impaired (Fig. 3D). To further confirm these results, we analyzed the proportion of cells positive for BrdU staining (BrdU⁺), which is a measure of the amount of cells undergoing DNA replication during the S-phase of the cell cycle, before and after doxorubicin treatment. Whereas there was no significant difference among the various untreated cell types, doxorubicin treatment resulted in a significant decrease in the number of BrdU⁺p53^+/+ and p53^+/– cells (BrdU⁺p53^+/+ vs p53^+/– cells after doxorubicin treatment: 9.34±1.48% vs 10.44±0.13%) (Fig. 3E). By contrast, the decrease was subtle and less pronounced in both p53^+/R246S and p53^–/– cells (BrdU⁺p53^+/R246S vs p53^–/– cells after doxorubicin treatment: 18.225±2.96% vs 17.7±0.46%) (Fig. 3E), confirming the DN effect of the mutant allele over the wild-type allele. Together, the data suggest that the R246S mutant protein acts in a DN manner in both undifferentiated and differentiated ES cells.

The general tumorigenic potential of p53^+/R246S cells was next investigated. ES cells have been shown to form teratomas, homing to the liver in mice with liver injury caused by carbon tetrachloride (CCl₄) treatment (Yamamoto et al., 2003). We therefore used this model to test whether p53^+/R246S ES cells would display higher tumorigenicity in vivo compared with wild-type p53-containing cells. All ES cell lines were transfected stably with a green-fluorescent protein (GFP) expression vector to monitor their homing capacity. As shown in Fig. 4A, injection of p53^+/+ and p53^+/– ES cells into scid mice resulted in the formation of micronodules in the liver, which were usually not visible to the naked eye, whereas p53^–/– and p53^+/R246S ES cells formed numerous tumour nodules, with an average diameter greater than 5 mm, in all of the four mice tested in each group. We therefore further investigated, by florescence microscopy, whether the tumours were derived from the injected ES cells and found that the teratoma cells were indeed GFP-positive, confirming their ES-cell origin (Fig. 4B). Detailed histological analysis revealed that the injected ES-cell-derived tumours contained heterogeneous cells types, including bone and cartilage cells, columnar epithelial cells, fibroblast-like cells and muscle cells (Fig. 4C). Because it has also been reported that the injected ES cells will also differentiate into hepatocytes, we analyzed the expression of hepatocyte markers, such as α1-antitrypsin, albumin, tryptophan-2,3-dioxygenase, transthyretin and α-fetal protein, and found their expression to be similar among the different genotypes (data not shown), suggesting that the rate of differentiation was not altered in vivo. The data therefore indicate that, although the mutation in p53 does not affect differentiation rates, the tumorigenic potential of p53^+/R246S ES cells is similar to that of p53^–/– ES cells and in contrast to wild-type and p53^+/– ES cells, indicating that the R246S mutant p53 is capable of exerting its DN effect over the wild-type p53 both in vivo and in vitro.

To assess whether the presence of other oncogenic stimuli will affect the growth properties of ES cells of the various p53 genotypes, we generated K-ras^v12-expressing ES cells (supplementary material Fig. S2A). Interestingly, similar to previously published results (Brooks et al., 2001), expression of K-ras^v12 did not promote the growth of unstressed wild-type ES cells (supplementary material Fig. S2B). Similarly, there was no effect of H-ras^v12 expression on cellular growth (supplementary material Fig. S2B) or on colony formation in vitro (data not shown) in ES cells of the various p53 genotypes, indicating that the presence of a further oncogeneic signal does not affect the growth properties of fast-growing ES cells with intrinsic tumorigenic potential.

Taken together, the data presented here demonstrate that the mouse equivalent of the commonly found human R249S hot-spot p53 mutant is capable of DN effects over the wild-type protein both in undifferentiated and differentiated ES cells, as assessed by multiple parameters – including transactivation of target genes, cellular survival, cell-cycle arrest after genotoxic stress and tumorigenic potential – therefore confirming the potency of the mutant protein in supporting tumorigenicity. This mutant, although incapable of transactivation potential both in the context of cultured cells and in transgenic mice (Ghebranious et al., 1995), maintains its capacity to inhibit the wild-type protein, again highlighting the relevance of DN effects of hot-spot mutants in tumorigenesis.

It is noteworthy that the effects of mutant p53 are probably not due to its ability to inhibit the other p53 family members, p63 or p73. ES cells do not express high levels of p63 (compared with fibroblasts) (supplementary material Fig. S3A), and inhibition of p73 expression by siRNA-mediated silencing did not affect cell death even in wild-type ES cells (supplementary material Fig. S3B), suggesting that the effect of this hot-spot mutant p53 is indeed through its ability to function in a DN manner over the remaining wild-type allele.

Although the DN activity of mutant p53 has been acknowledged as a contributory mechanism for tumorigenesis, experimental studies both in cultured cells and in mice have resulted in conflicting results. Many reasons have been put forth to explain the lack of DN effects when it was not observed, ranging from tissue-type specificity to the expression levels of the mutant protein in the cell. It is interesting to note that a recent study revealed that at least three molecules of mutant p53 are required to inhibit the function of a wild-type protein, highlighting the requirement for threshold levels for the DN effects of the mutant protein to be observed (Chan et al., 2004). Consistently, the DN effect was not observed in primary tissues from other mutant p53 knock-in mice, in which there was no accumulation of the mutant protein, in contrast to the tumour cells from these mice (Lang et al., 2004; Olive et al., 2004). In this respect, ES cells are different and are known to express extremely high levels of p53 compared with differentiated cell types such as fibroblasts (Sabapathy et al., 1997). Expression of endogenous mutant p53 are higher in these cells (this report), which coincided with the DN effects, noted also in the case of the R270H mutation (de Vries et al., 2002). Therefore, it is plausible that higher steady-state protein level in both normal and stressed conditions, as seen in the context of ES cells, might be essential for the mutant p53 protein to manifest its DN behaviour.

Results from previously generated mutant p53 knock-in mice revealed that there might be a cell-type bias for mutant p53-mediated DN function. Whereas the DN effect was not observed in mouse embryonic fibroblasts (MEFs), it was observed partially in thymocytes and very strongly in the developing brain of these mice (Olive et al., 2004). It is interesting to note that the DN effect is seen in cell types that are not terminally differentiated, such as in thymocytes, which contain primarily double-positive immature T cells, and neural stem cells in developing brain (Brazel et al., 2003; Hayday and Pennington, 2007). p53 levels are elevated upon γ-irradiation in thymocytes, which are known to undergo apoptosis in a p53-dependent manner (Clarke et al., 1993). Similarly, p53 is highly expressed in the neural-stem-cell niche of adult brain, and absence of p53 promotes proliferation, survival and self-renewal of the neural stem cells (Meletis et al., 2006). Therefore, it is not unconceivable that stem cells and less-differentiated cells rely on the activity of p53 more than do terminally differentiated somatic cells, and, hence, the DN effects are more evident in them. Further indirect support for this idea comes from R270H conditional knock-in mice, which only showed partial DN effects in vivo when most of the mutant protein was expressed in the matured epithelial cells by adenoviral induction (Jackson et al., 2005). Similarly, partial DN effects upon UV-irradiation-induced skin carcinogenesis was observed in p53^R270H/K14-Cre double-transgenic mice, in which the mutant protein was expressed in keratinocytes (Wijnhoven et al., 2007). It will therefore be interesting to determine whether expression of the mutant protein in stem-cell niches will lead to stronger DN effects.

In conclusion, the data presented here demonstrate that the commonly found hot-spot p53 mutant, R246S, displays strong DN properties in mouse ES cells, resembling the situation in human tumours. Because cancer is often considered a disease of stem cells, and because mutant p53 is very stable in cancer cells, the ES cell model presented here fulfils both criteria and provides a best-fit to understand mutant p53 properties. It will therefore be interesting to next evaluate whether either or both properties, i.e. pluripotency and high levels of protein, are required for the DN effects of mutant p53 to be manifested, using the knock-in mice for analysis.

The genomic clones of murine p53 (G2, exon 2-6 and G10, exon 7-11) were kindly provided by A. de Vries (National Institute of Public Health and the Environment, The Netherlands) (de Vries et al., 2002), and the floxed neomycin-resistant gene cassette (pKSloxpNT) and diphtheria toxin A (DTA) expression cassette were gifts from M. Sibilia (University of Vienna, Austria). The DTA cassette was PCR amplified with a pair of primers containing the XhoI site and was sub-cloned into the XhoI site at the 3′ end of the genomic clone G10 to generate the intermediate clone, G10-DTA. The floxed neomycin-resistance gene was released by EcoRI-KpnI digestion and G10-DTA was linearized by XbaI. Blunt ends of all fragments were generated using Klenow and the floxed neomycin-resistant gene cassette was sub-cloned into the XbaI site by blunt-end ligation to generate the second intermediate clone, neo-G10-DTA. The R246S mutation (CGA to TCT) was generated by site-directed mutagenesis on the neo-G10-DTA clone by QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). The mutation creates a novel BsrBI cut site, which was used for screening constructs and ES cells. A 4 kb BssHII fragment of G2 was sub-cloned into cloning vector pSL1180 (Amersham Biosciences, Buckinghamshire, UK) to obtain an additional HpaI and NotI site at the 5′ end of the fragment (pSLG2). The 4-kb G2 fragment from pSLG2 was released by NotI digestion and neo-G10(R246S)-DTA was linearized with NotI. These fragments were ligated together to generate the targeting construct (Fig. 1A). All the exons and the splice junctions in the targeting construct have been verified by sequencing to ensure that no additional unwanted mutations were introduced during cloning.

CCE (p53^+/+), p2.4 (p53^+/–) and p1.1 (p53^–/–) ES cells were cultured and differentiated with 0.1 μM retinoic acid for 6 days as described previously (Lee et al., 2005). 30 μg of HpaI-linearized construct was electroporated into exponentially growing ES cells and selected with 250 μg/ml G418 for 10-14 days. Homologous recombinants were identified by PCR and Southern blot screening, and were transiently transfected with pCAG-Cre to remove the neomycin selection cassette. G418-sensitive ES cells clones were screened to confirm the removal of selection cassette. The presence of R246S mutant allele was determined by PCR/RFLP and the expression of mutant allele was confirmed by RT-PCR-RFLP.

The p53^R246S/– MEFs were obtained from 13.5 dpc embryos by mating p53^+/R246S and p53^+/– mice. R175H-expressing H1299 cells were generated in our laboratory as described previously (Vikhanskaya et al., 2007). The p63^–/–p73^–/– mouse embryonic fibroblasts were a kindly provided by E. Flores (MD Anderson, TX).

Undifferentiated or differentiated ES cells were treated with 0.1 μg/ml doxorubicin for 24 hours prior to BrdU addition (to a final concentration of 10 μM), incubated for a further 1 hour and fixed in 70% ethanol overnight.

For gene silencing experiments, 1×10⁵ undifferentiated ES cells were transfected with scrambled or p73 siRNA for 36 hours by RNAiFect, following the manufacturer's protocol (Qiagen, Germany), and treated with 1 μg/ml doxorubicin for 12 hours prior to determination of cell viability by flow cytometry.

2×10⁴ undifferentiated ES cells were plated onto six-well plates and γ-irradiated at various doses. Irradiated cells were cultured for a further 8 days and the total number of surviving cells was counted. For Ras-expressing cells, 1×10⁵ cells were plated onto six-well plates and counted daily. Independent experiments were performed, in duplicate, at least thrice and data represent mean ± s.e.m.

Long-term survival of ES cells was assayed by colony-formation assays. Essentially, 1×10³ undifferentiated ES cells were irradiated as mentioned above and cultured for 8-10 days. To determine the effects of H-ras expression, undifferentiated ES cells were transfected with linearized pBabe-Ras^V12 plasmid as described previously (Lee et al., 2005) and selected with 1 μg/ml puromycin for 8-10 days. Surviving cells were pooled and maintained in 0.5 μg/ml puromycin at all times. For colony-formation assay, 100 cells were plated and allowed to grow for 10 days. Surviving colonies were stained with crystal violet solution (MERCK, Whitehouse Station, NJ) as described previously (Vikhanskaya et al., 2007).

10 μg of ES cell genomic DNA was EcoRI-digested, separated by 0.7% agarose gel electrophoresis and transferred to positively charged nylon membrane (Amersham Biosciences). PCR-amplified exon 1 and 11 probes were used for Southern blot hybridization as described (Luo et al., 2001; Olive et al., 2004).

For RT-PCR-RFLP, total RNA from ES cells was used for first-strand cDNA synthesis. p53 cDNA and p53 exon 7 from genomic DNA were PCR amplified as described previously (Lee et al., 2005), and were digested with BsrBI for RFLP analysis.

Quantitative real-time PCR was performed using gene-specific primers and Quantitect real-time PCR reagent (Qiagen) in Cobett real-time PCR machine (Cobett Research, Sydney, Australia) as described previously (Lee et al., 2005). mRNA expression of target genes were normalized with gapdh expression and fold induction was calculated with reference to untreated samples. For semi-quantitative RT-PCR, 1 μl of cDNA was used to amplify the gene of interest as described previously (Lee et al., 2005).

The expression of mouse p63 and p73 genes was analyzed by semi-quantitative RT-PCR using p63-specific primers (forward, 5′-CACAGAATAGCGTGACGGCGCC-3′ and reverse, 5′-CTCTGCCTTCCCGTGATAGGATC-3′) and p73-pecific primers (forward, 5′-GAGCACCTGTGGAGTTCTCTAGA-3′ and reverse, 5′-GTGACAGGGTCATCCACGTACTGG-3′). p73 siRNA has been described (Vikhanskaya et al., 2007).

Expression of Ras was determined by semi-quantitative RT-PCR (Ras forward primer, 5′-AGAAGGCATCCTCCACTCC-3′ and reverse primer, 5′-CCATCAACCAACACCCAAG-3′).

20 μg of total RNA was used for northern blot analysis. Probe was labelled and purified as described above.

Whole-cell lysates from ES cells were used for western blotting performed as described previously (Lee et al., 2005), with either anti-p53 antibody (CM5; Novocastra, Newcastle, England) or anti-actin antibody (Sigma, St Louis, MO).

500 μg of whole-cell protein lysate was used for immunoprecipitation with the following conformation-specific anti-p53 antibodies: Pab240 (Calbiochem, San Diego, CA), which recognizes the denatured and mutant conformation, and Pab246 (Calbiochem), which recognizes the native wild-type conformation, followed by anti-mouse IgG agarose beads. Immunoprecipitates were then separated by 10% SDS-PAGE and western blot analysis was performed with a mixture of anti-p53 antibody 1C12 (Cell Signaling, Danvers, MA) and DO-1 (Santa Cruz, CA).

For immunocytochemistry, ES cells were fixed in 4% paraformaldehyde 3 hours after irradiation, stained with anti-p53 antibody Pab240 (Calbiochem) followed by Alexa-Fluor-488-conjugated anti-mouse IgG antibody (Molecular Probes, Eugene, OR) and propidium iodide. Fluorescence confocal microscopy was performed using the LSM 510 laser scanning confocal microscope (Zeiss, Jena, Germany).

Apoptosis was measured by annexin-V binding assay and flow cytometry according to the manufacturer's protocol (BD Biosciences, Franklin Lakes, NJ). For cell cycle analysis, cells were fixed in 70% ethanol at 4°C overnight, washed with PBS once and incubated with 0.2 mg/ml RNase and 50 μg/ml PI in PBS for 20 minutes, prior to flow cytometric analysis with ModFit cell cycle analysis software. The percentage of BrdU⁺ cells was determined by flow cytometry after staining with FITC conjugated anti-BrdU antibody (BD Biosciences), as per the manufacturer's instruction.

ES cells of various p53 genotypes were first transfected with GFP-encoding plasmid pRNAT-U6.1/Hygro (GenScript Corp., Piscataway, NJ), and hygomycin-resistant ES cells were used to induce teratoma as described previously using 6-week-old scid mice treated with CCl₄ (Yamamoto et al., 2003). All mice were sacrificed 3 weeks later and livers were fixed in 4% neutral-buffered formaldehyde, followed by dehydration and paraffin embedding. Histological analysis was carried out on 5-μm sections stained with haematoxylin and eosin (Sigma). All animal experiments were carried with the approval of the Institutional Animal Care and User Committee

We are grateful to A. De Vries, M. Sibilia and E. Flores for constructs and cells. L.M.K. was partially supported by a Singapore Millennium Foundation fellowship. We thank the National Medical Research Council of Singapore for their generous funding and support to K.S.