A single-nucleotide polymorphism (SNP) in exon 4 results in expression of either arginine (72R) or proline (72P) at codon 72 of p53. We demonstrate that the in vitro response of cells exposed to anticancer agents is strongly influenced by this SNP in wild-type p53. In inducible systems and in cells expressing the endogenous protein, expression of 72P wild-type p53 results in a predominant G1 arrest, with only a minor apoptosis, at drug concentrations causing extensive apoptosis in cells expressing the 72R wild-type variant. The superior apoptosis-inducing activity of the 72R form correlates with more efficient induction of specific apoptosis-associated genes, and is maximal in the presence of serine 46 (S46). In vivo, the outcome of chemo-radiotherapy of squamous carcinomas is more favourable in cancers retaining a wild-type 72R allele, such cases having higher response rates and longer survival than those with wild-type 72P. Together, these results reveal that this SNP is an important determinant of response to anticancer agents in cells expressing wild-type p53. Analysis of complete p53 genotype (mutation and SNP) merits detailed investigation as a simple means for prediction of treatment response and survival in clinical oncology.
‘Activation’ of the DNA-binding function of p53 following exposure to cellular stress leads to cell cycle arrest, senescence or apoptosis (Vogelstein et al., 2000). Activation of p53 involves post-translational modifications such as phosphorylation and acetylation, following which the protein is stabilized and acquires increased DNA-binding affinity. Many p53 target genes have been described (el-Deiry, 1998). These include genes involved in cell cycle arrest such as p21Waf1 and 14-3-3σ and those mediating apoptosis (for example AIP1, Bax, PUMA, Noxa and PIGPC1) (Vousden and Lu, 2002).
p53-dependent apoptosis is an important mechanism through which DNA-damaging anticancer agents exert their biological effects (Lowe et al., 1993, 1994; Johnstone et al., 2002). Serine 46 (46S) is reported to function in the regulation of p53-dependent apoptosis, via proapoptotic genes such as AIP1 (Oda K et al., 2000). However, AIP1 expression is upregulated only at later stages of apoptosis and at high levels of DNA damage, and chromatin immunoprecipitation (ChIP) assays have questioned the candidacy of this gene as a direct effector of p53-dependent apoptosis (Kaeser and Iggo, 2002). More likely candidates may be PUMA (Nakano and Vousden, 2001; Yu et al., 2001) and Noxa (Oda E et al., 2000). Cells lacking these proteins are defective in apoptosis, at least in response to some cellular stresses (Shibue et al., 2003; Villunger et al., 2003). Apoptosis induction is the function of p53 selected against in tumorigenesis (Schmitt et al., 2002) and, accordingly, almost all human-tumour-associated p53 mutants are defective for apoptosis.
A SNP in exon 4 of p53 results in the presence of either arginine (R) or proline (P) at codon 72 (Matlashewski et al., 1987). The polymorphism is balanced, varies with latitude and race, and is maintained at different allelic frequencies across the population of the world (Sjalander et al., 1995). Further, the polymorphism is located in the proline-rich domain, which is important in the apoptosis function of p53 (Baptiste et al., 2002). Taken together, these observations imply biologically important differences between the 72R and 72P variants, which are subject to natural selection pressures. There is experimental evidence that the 72R form of p53 possesses greater apoptosis-inducing potential than the 72P variant (Thomas et al., 1999; Dumont et al., 2003; Pim and Banks, 2004). Mechanistically, this has been linked to differential nuclear/cytoplasmic transport of the 72R and 72P forms and localisation at the mitochondrion (Dumont et al., 2003). We previously showed that the two allelic forms of mutant p53 confer differential cellular resistance to anticancer agents, linking that effect to the outcome of chemotherapy in carcinomas expressing mutant p53 (Marin et al., 2000; Bergamaschi et al., 2003). Of course, the great majority of cells in the body contain wild-type p53, and it is wild-type p53 that mediates the normal physiological function of p53. Using inducible p53 in otherwise isogenic cells and endogenous p53 in human cells, we now show that the 72R and 72P forms of wild-type p53 differ in their ability to cause apoptosis following exposure to drugs from distinct mechanistic classes of agent. Further, we link this to the differential transcriptional effects of p53. To address the in vivo biological significance of the different molecular properties of the 72R and 72P variants, we have analysed the response of a well-defined series of patients with head and neck cancer to chemo-radiotherapy with agents whose cytotoxicity we have shown to be influenced by the SNP.
Anticancer drugs induce higher apoptosis in cells expressing 72R wild-type p53
We investigated whether the reported differences in the apoptosis-inducing potential of the two common polymorphic forms of wild-type p53 result in distinct cellular responses to anticancer drugs. For these studies, we used H1299 cell lines inducibly expressing the 72R and 72P wild-type proteins and immortalized B-cell lines expressing wild-type p53, but having differing genotypes at codon 72. In the H1299 system, we analysed a minimum of six independent 72R-inducible and 72P-inducible clones, to obviate the effect of clonal variation. Using this system, we compared apoptosis resulting from expression of the two polymorphic forms following exposure to clinically important anticancer drugs from different mechanistic classes. The tested drugs comprised camptothecin, cisplatin, doxorubicin, etoposide, 5-fluorouracil (5-FU), melphalan and taxol. Expression of either 72R or 72P p53 resulted in only low levels of apoptosis in the absence of drug treatment (Figure 1a and 2a). Similarly, exposure of noninduced cells to the same agents caused only a small (but reproducible) increase in sub-G1 cells compared to untreated cells (Figure 1a). When induced cells were exposed to anticancer agents, apoptosis was always higher in cells expressing the 72R variant, whereas the predominant effect in cells expressing 72P wild-type p53 was G1 arrest (Figure 1a and 2a). These differences in cell cycle distribution and apoptosis were reproducibly observed in three independent H1299 clones expressing 72R p53 and three expressing 72P p53 (Figure 1c and 2b). The difference in apoptosis between 72R and 72P was as high as eightfold in the case of doxorubicin, but was never less than twofold (for camptothecin) (Figure 1c and 2b).
To verify that similar effects occur in cells expressing endogenous p53, we analysed the response to the same anticancer agents in EBV-immortalized B-lymphoblastoid cell lines. Such lines retain a p53-dependent, DNA damage-inducible pathway, and are thus an appropriate system for these comparative studies (Allday et al., 1995). The difference in apoptosis between 72R and 72P seen in the H1299 system was also observed in lymphoblastoid cell lines. Apoptosis was reproducibly higher in cell lines of 72RR genotype compared to those of 72PP genotype (Figure 2c). Moreover, time course analysis revealed that the onset of apoptosis was more rapid in the 72R cells. From Figure 2c, it can be seen that sub-G1 is clearly detectable 8 h after addition of drugs in cells expressing 72R p53, whereas no increase in apoptosis is seen in cells expressing 72P. Despite the differences in apoptosis, the kinetics of p53 stabilization were indistinguishable following drug exposure (Figure 2d). Together, these studies reveal the differences in apoptosis-inducing activity between the two polymorphic variants of wild-type p53, which are not attributable to differential stability, in cells challenged with anticancer agents from diverse mechanistic classes.
We next determined whether the differences in proapoptotic activity between the polymorphic variants was reflected in different cytotoxicities of anticancer drugs. In colony-survival assays performed in H1299 cells, the cytotoxicities of cisplatin and doxorubicin were invariably higher in 72R-expressing cells than in 72P-expressing cells. These differences were observed in three independent H1299 clones expressing 72R and three expressing 72P (Figure 1e). Similarly, there was reproducibly higher cytotoxocity of etoposide, 5-FU and taxol in cells expressing 72R p53 (data not shown).
Differential activation of proapoptotic target genes by p53 variants
To identify the genes differentially expressed between the 72R and 72P proteins, we used a p53 target gene array to compare p53-dependent gene expression in H1299 cell lines induced to express 72R and 72P wild-type p53 (Figure 3a), and then treated with cisplatin or doxorubicin. The mRNA of three genes represented on the array was induced more efficiently in clones expressing 72R p53 than 72P p53. The genes more efficiently upregulated by 72R were PUMA (BBC3), PIGPC1 (PERP) and AIP1, but there was no difference in induction of p21Waf1 or MDM2 (Figure 3b, c and data not shown). Superior induction of these genes by 72R p53 was verified by TaqMan PCR in three independent clones of H1299 cells expressing the 72R and three expressing the 72P variant of wild-type p53 (Figure 3b, c). To verify that the increased induction of these genes was attributable to enhanced specific DNA binding of the 72R protein, we used ChIP and luciferase assays. In the case of the p21Waf1 and MDM2 promoters, there were only small differences in DNA binding and induction of luciferase activity between the two variants (Figure 3d), consistent with TaqMan analyses demonstrating similar induction of mRNA (Figure 3b). In contrast, the 72R variant reproducibly bound the AIP1 and PUMA promoters and activated the promoters of these genes with higher efficiency than the 72P variant (Figure 3d).
The ability of 72R wild-type p53 to more efficiently transactivate AIP1, PUMA and PIGPC1 prompted us to examine, in luciferase assays, whether the promoters of other apoptosis-associated genes were also more efficiently transactivated. Consistent with ChIP assays, the PUMA promoter was more efficiently transactivated by 72R, but no difference was detected in the activation of Bax or Bid (Figure 3e and data not shown). In contrast, the Noxa promoter was reproducibly more efficiently activated in H1299 cells by 72R than by 72P wild-type p53 (Figure 3e).
46S is dispensable for p53-dependent apoptosis, but confers higher activity to the 72R variant
We next sought to determine whether the observed differences between the 72R and 72P variants involved 46S of p53, since this residue is reported to be important in p53-dependent apoptosis (Oda K et al., 2000). We analysed H1299 cells inducibly expressing 46A 72R and 46A 72P p53, and compared this to cells expressing similar levels of the wild-type 46S variants (Figure 3a). In this system, the relatively greater binding of the 46S 72R protein to the AIP1 and PUMA promoters was reduced by substitution of 46S with A (Figure 4a), but there was little difference between the 46S and 46A variants of the 72R and 72P proteins in binding to the p21Waf1 and MDM2 promoters (Figure 4a). Consistent with these observations, p53-dependent induction of PUMA and AIP1 was reduced in cells expressing the 46A variants of wild-type p53 (Figure 5b).
Apoptosis in cells expressing 46A 72P p53 was marginally reduced relative to 46S 72P p53. However, apoptosis in cells expressing 46A 72R p53 was reduced to a level comparable to that in cells expressing 46S 72P and 46A 72P p53 (Figure 4b). This was reproducibly observed in multiple experiments, and was seen following exposure to doxorubicin and cisplatin (Figure 4b and 5a). A similar effect was seen in analysis of etoposide and 5-FU (data not shown). This effect was not related to differences in the stability of the proteins after DNA damage, since levels of p53 protein increased after cisplatin treatment for each of the four p53 variants tested (Figure 4c). Consistent with apoptosis assays, the cytotoxic effect of cisplatin and doxorubicin was diminished in cells expressing 46A 72R p53 relative to 46S 72R, but the sensitivity of cells expressing 46A 72P was similar to that of cells expressing 46S 72P (Figure 5c). These experiments show that p53-mediated apoptosis occurs in the absence of 46S, but indicates that the relatively greater apoptosis- and cytotoxicity-inducing activity of the 72R protein is dependent, at least in part, on the presence of 46S.
Polymorphism in wild-type p53 influences the outcome of cancer chemotherapy
A potentially important implication of our results is that polymorphism in wild-type p53 affects response and outcome in cancer treatment by virtue of its effect on drug-induced apoptosis. We addressed this possibility in 70 patients with inoperable, advanced head and neck squamous cell carcinomas (HNSCC), who received cisplatin-based chemo-radiotherapy. All patients presented with locally advanced, unresectable, TNM stage III/IV HNSCC. In the 43 patients whose cancers retained a wild-type p53 allele and who had an evaluable response to treatment, complete response appeared significantly linked with whether a cancer retained a wild-type 72R p53 allele, wild-type 72P p53 allele or both (P<0.04). Individuals whose cancers retained a wild-type 72R allele had the highest complete response response rate (27 out of 28 patients, 96%). Both overall survival (OS) and progression-free survival (PFS) were significantly longer in patients whose cancers retained a wild-type p53 allele (either 72R or 72P), than in those lacking a wild-type allele (both P<0.0001; Figure 6). Furthermore, there was significantly different OS and PFS between cases retaining a wild-type 72R allele, cases retaining a wild-type 72P allele and cases retaining both wild-type 72R and 72P alleles (P=0.02 and 0.007, respectively), with the best prognosis in cases retaining a wild-type 72R allele (Figure 6).
We show that the response of cells to anticancer agents is strongly influenced in vitro by a SNP in wild-type p53, which modulates apoptosis, at least in part, by its effects on the downstream programme of p53-regulated gene expression. We demonstrate the clinical relevance of these observations by presenting evidence that the superior activity of the 72R wild-type protein is reflected in vivo by significantly higher response rates and survival in patients receiving chemo-radiotherapy for advanced head and neck cancer, whose cancers express the 72R wild-type variant.
We were interested in analysing the effect of the SNP on apoptosis induced by anticancer agents, because studies in other systems have suggested differences in apoptosis potential between the two polymorphic variants (Thomas et al., 1999; Dumont et al., 2003; Pim and Banks, 2004). Further, we had previously observed that the ability of mutant p53 to affect the sensitivity of cells to anticancer agents was influenced by the same SNP (Bergamaschi et al., 2003). Together, these observations suggested that a single SNP might have opposing effects on wild-type and mutant p53 with regard to sensitivity to chemotherapy. In the present study, we have addressed this question in a ‘Tet-on’ system inducibly expressing either 72R or 72P wild-type p53 in an isogenic background, and in nonisogenic B-lymphoblastoid cells expressing endogenous wild-type p53. In these experimental systems, we have shown that exposure of p53-expressing cells to each of the tested drugs results in higher levels of apoptosis in the presence of the 72R variant, whereas a predominant G1 arrest occurs in cells expressing equal amounts of the 72P wild-type protein. These conclusions are in agreement with other recent studies (Pim and Banks, 2004).
Previously, it was shown that the 72R protein more efficiently localizes to the mitochondrion (Dumont et al., 2003). In the present study, we sought evidence of differences in the transcriptional regulatory function between the two polymorphic forms by analysis of p53-dependent gene expression. To permit accurate analysis of the relative induction of each target mRNA, we used TaqMan PCR and observed superior induction of several genes proposed to function in p53-dependent apoptosis. Using ChIP assays, we verified that the higher upregulation of expression by 72R p53 was associated with more efficient binding to the promoters of these genes. One such gene is AIP1. However, ChIP assays have questioned the candidacy of AIP1 as an essential p53 target gene (Kaeser and Iggo, 2002). Our own results also suggest that AIP1 is not an essential gene for p53-dependent apoptosis, because blocking expression by siRNA did not affect apoptosis, despite clear ‘knockdown’ of mRNA levels (data not shown). A second gene more efficiently induced by 72R was PIGPC1, consistent with previous work showing higher induction of PIGPC1 by the 72R variant of a temperature-sensitive p53 mutant (Dumont et al., 2003). However, like AIP1, siRNA-mediated inhibition of PIGPC1 did not markedly affect p53-dependent apoptosis in H1299 cells. As such, the significance of the more efficient induction of AIP1 and PIGPC1 by 72R wild-type p53 awaits further study.
The more efficient transactivation of PUMA and Noxa by the 72R protein may be a more likely partial explanation for the greater apoptosis-inducing activity of the 72R protein. Recent studies using homologous recombination showed that PUMA is an essential gene for p53-dependent apoptosis, cells lacking PUMA failing to undergo p53-dependent apoptosis in response to hypoxia and cisplatin treatment (Yu et al., 2003). In other studies, important roles have been established for both PUMA and NOXA in p53-dependent apoptosis (Shibue et al., 2003; Villunger et al., 2003). These results support the hypothesis that preferential activation of PUMA and Noxa by the 72R variant may contribute to the differences in apoptosis potential between the polymorphic variants of wild-type p53.
The results of these experiments implicate p53 polymorphism as an important determinant of cellular sensitivity to anticancer agents in vitro. We therefore determined whether the in vivo clinical activity of such agents is also influenced by this SNP. We report significant differences in clinical response and outcome in vivo according to the codon 72 genotype and mutational status of the tumour, in head and neck cancer patients treated with chemo-radiotherapy regimens based on cisplatin, but also containing 5-FU and taxol. These are all agents whose apoptosis-inducing activity and cytotoxicity are influenced by the SNP in wild-type p53. In this cohort of patients, response rates and survival (both overall and progression-free) were significantly higher in cases retaining a wild-type 72R allele. To the best of our knowledge, this is the first demonstration that clinical outcome in human cancer is influenced by a SNP in wild-type p53. We have previously shown that treatment outcome in advanced HNSCC treated with combined-modality chemo-radiotherapy regimens is less favourable in cases expressing mutant 72R p53 (Bergamaschi et al., 2003). One of the goals of translational oncology is to identify the molecular markers predictive of treatment outcome. Such markers would be of particular value for treatment regimens, such as combined modality therapy, which are associated with severe morbidity and substantial expense, yet produce complete response in only a subset of patients. As such, it is important to find the molecular genetic determinants of treatment outcome to facilitate identification of patients most likely to benefit from such treatments. The data we present in this study suggest that analysis of polymorphism in p53 is one potentially useful marker. It will now be clearly important to verify the effect of the polymorphism in larger series of patients and in other cancers. Together with the previous demonstration that polymorphism in mutant p53 affects the clinical outcome in HNSCC (Bergamaschi et al., 2003), the present results provide further evidence of the potential utility of routine molecular genetic analysis of p53 in prediction of clinical response to specific anticancer agents. Confirmation of the usefulness of analysis of this SNP as a biomarker of treatment outcome should prompt serious consideration of its use in the routine work-up of patients prior to treatment decisions being made.
Wild-type p53 72R and 72 P plasmids were originally obtained from Dr LV Crawford as pArgSP53 and pProSP53 (Matlashewski et al., 1987). The pTet-on regulator plasmid and pTRE were obtained from Clontech. pTRE 72R p53 and pTRE 72P p53 were constructed by subcloning the inserts from pArgSP53 and pProSP53 into pTRE. 46S p53 mutants were constructed in pTRE 72R p53 and pTRE 72P p53 using the Quikchange mutagenesis system, and verified by sequencing. pWWP (p21Waf1) luciferase and PUMA luciferase were generously given to us by Dr B Vogelstein, PERP (PIGPC1) luciferase, AIP1 luciferase and MDM2 luciferase were kindly provided by Drs Tyler Jacks, Silvia Soddu and Moshe Oren, respectively. Noxa luciferase was made by PCR.
Human H1299 p53 −/− cells were maintained in DMEM+10% foetal bovine serum. For establishment of founder Tet-on cell lines, subconfluent cells were cotransfected with pTet-on and pSV2neo. Individual G418-resistant clones were isolated and tested for their ability to support doxycycline-dependent induction of luciferase following transfection with pTRE-luc. The inducible founder clone was then cotransfected with either pTRE 46S 72R p53, pTRE 46S 72P p53, pTRE 46A 72R p53 or pTRE 46A 72P p53, together with pBabePuro. Multiple puromycin-resistant colonies were cloned and screened by Western blotting to define conditions for induction of equal steady-state levels of p53 protein. EBV-immortalized B-lymphoblastoid cell lines were maintained in RPMI1640 medium with 10% FBS.
Drug treatment, apoptosis and cytotoxicity
Flow cytometry was performed essentially as described previously (Bergamaschi et al., 2003). Early log phase cells were treated with anticancer agents at the following concentrations: camptothecin 5 μM; cisplatin 1 and 10 μg/ml; doxorubicin 1 μg/ml; etoposide 10 μM; 5-FU 50 μg/ml; melphalan 10 μg/ml; taxol 500 ng/ml. The cytotoxicity of anticancer agents was determined in colony-survival assays using standard protocols. Briefly, cells were plated at multiple densities, and then exposed to varying concentrations of each drug for 2 h. After drug exposure, cells were re-plated in drug-free medium and grown until the appearance of visible colonies. Each drug exposure was done in triplicate in at least three experiments.
Identification of differentially induced p53 target genes
H1299 cell lines with inducible 72R or 72P p53 proteins were grown in DMEM 10% FCS supplemented with 0.5 mg/ml G418 and 0.5 μg/ml puromycin. Total RNA was prepared from noninduced cells, induced cells expressing equal steady-state levels of 72R and 72P p53, and from cells treated with UV or doxorubicin. This was then used to probe an array of p53 target genes following the protocol provided by the manufacturer (TranSignal human p53 target gene array, Panomics, Redwood City, CA, USA). Each hybridization was performed at least three times and differences in upregulation were then verified by quantitative PCR.
Analysis of differentially induced p53 target genes
H1299 cell lines with inducible 72R or 72P p53 proteins were grown as above. For induction of p53 and drug treatment, cells were plated in 180 cm2 plates in 15 ml medium and treated with doxycycline. After 18 h, cells were exposed to anticancer drugs as indicated above.
For RNA isolation, cells were harvested at various time points and RNA prepared using the Qiagen RNeasy mini kit. Quantitative PCR was performed exactly as described (Kaeser and Iggo, 2002).
For protein extractions, cells were harvested at various time points in lysis buffer (50 mM Tris, 250 mM NaCl, 0.1% Nonidet NP-40, 5 mM EDTA, 50 mM NaF, 1 mM PMSF with protease inhibitor cocktail (Roche)). Antibodies were as follows: p53: DO-1 ascites, 1 : 1000; PCNA: PC-10 ascites, 1 : 1000.
Cells were treated with doxycycline and drugs, as described above. After 24 h, plates were processed for the ChIP assay as described (Kaeser and Iggo, 2002). Each assay included a positive control (Sp1 binding to the DHFR reporter) and a negative control (GAPDH). p53 was immunoprecipitated by anti-p53 DO-1 antibody at 4°C for 1 h. PCR for the AIP1, p21Waf1, MDM2 and PUMA promoters was performed with the primers described elsewhere (Kaeser and Iggo, 2002). Amplification was done using AmpliTaq gold under the following amplification conditions: 10 min at 95°C, 30 s denaturing at 95°C, 1 min annealing at 55°C and 1 min extension at 72°C. TaqMan probes were synthesized as described (Kaeser and Iggo, 2002).
H1299 cells were transfected with pCB6+72R or pCB6+72P wild-type p53 expression plasmids and luciferase reporter constructs, together with pSV2β-gal (control for transfection efficiency). After 24 h, cells were harvested for determination of luciferase and β-galactosidase activities using the Promega system.
The cases were advanced HNSCC. The patient population has been described previously (Bergamaschi et al., 2003). Tissues were obtained as paraffin sections or frozen tissues from patients undergoing diagnostic biopsy prior to commencement of chemo-radiotherapy. In each case, the diagnosis and presence of an adequate proportion of tumour tissue in each sample was confirmed by histopathological analysis.
DNA isolation and p53 analysis
Genomic DNA was purified by proteinase K digestion of 5 μm sticks cut from paraffin sections. When matched normal tissue was not available, this was obtained by microdissection. Mutations and polymorphisms in p53 were identified by direct sequencing of exons individually amplified with Pfx DNA polymerase (Gibco).
χ2 tests and Fisher's exact tests were used to test the frequencies of mutations, LOH and response to treatment rates between the different codon 72 genotypes. Survival times were calculated as the date of diagnosis to the date of death or date of censor if alive. Progression-free survival times were calculated as the date of diagnosis to the earlier of date of progression or death, or date of censor if not progressed or died. Survival curves were constructed using the method of Kaplan and Meier (1958), and the log-rank test (Peto et al., 1977) was used to assess the prognostic ability of retention of wild-type p53 alleles. Statistical analysis was carried out using SAS Statistical Software (SAS Institute, Cary, NC, USA).
We thank Compagnia di San Paolo for supporting parts of this work.