Inactivate the remaining p53 allele or the alternate p73? Preferential selection of the Arg72 polymorphism in cancers with recessive p53 mutants but not transdominant mutants

Mutation of the TP53 tumor suppressor gene itself is seen in more than half of all human cancers (1–3), and tumors retaining wild-type p53 often have mutations elsewhere in the p53 pathway or express viral proteins that inactivate p53 (4,5). A common polymorphism at residue 72 [Pro (P) or Arg (R)] has been linked to the ability of p53 to interact with cellular proteins and its susceptibility to degradation by the human papillomavirus E6 protein, possibly through an effect of the polymorphism on the conformation of the p53 protein (6,7). Following the first report of an increased risk of cervical cancer in R72/R72 homozygous individuals (6), many studies have been published showing both positive (8–11) and negative (3,12,13) correlations between polymorphic status and cancer occurrence. At present, the significance of the codon 72 polymorphism remains obscure, both in terms of cancer epidemiology and pathobiology.

Marin et al. recently described a difference in the biochemical activity of codon 72 polymorphic variants and postulated an enhanced pathological role for the R72 form in certain tumors (14). Some p53 mutants undergo a conformational change that promotes binding to the p73 protein and inhibition of p73 activity. The p53 mutant is a more potent inhibitor when it has arginine rather than proline at codon 72. Since p53 mutants with conformational changes are commonly recessive (meaning they do not inactivate wild-type p53 in a transdominant negative manner), we postulated that two functionally distinct pathways of p53 inactivation might exist: an R72-dependent one for recessive alleles and an R72-independent one for dominant alleles. To test this theory we examined whether the mutant p53 allele in our series of human cancers (15–17) was able to inhibit wild-type p53 transactivation in a yeast transdominance assay (18) and analyzed the status of the codon 72 polymorphism by sequencing (19) (Table I). Since the TP53 mutations were originally identified using a yeast p53 functional assay in which a large fragment of p53 cDNA including codon 72 is cloned as a single unit (18), we could readily determine the codon 72 sequence of the mutant p53 allele.

We tested 97 missense mutations and three in-frame deletions found in human cancers (39 brain tumors, eight liver cancers, 11 colon cancers, 17 breast cancers, seven pancreatic cancers, 17 oral cancers and one skin cancer). We excluded chain-terminating mutations (premature stop codon and frameshift mutations) from this study because such mutants cannot interact with the remaining wild-type p53 allele, due to the lack of the C-terminal oligomerization domain at amino acids 324–355. The spectrum of TP53 mutations studied here was essentially identical to that in the IARC database (2), with respect to frequency of hotspot mutations, type of base changes (transition versus transversion) and distribution of the mutations. Codon 72 in the mutant alleles was arginine in 73 tumors and proline in 27. The R72 allele was thus significantly over-represented in the tumors compared with its germline frequency in the Japanese population (R:P = 427:297, P < 0.008, χ² test) (20,21). This is consistent with the observation of Marin et al. that mutant alleles containing R72 are preferentially selected during tumorigenesis (14). Interestingly, examination of transdominance data revealed that this selection for R72 was true only of recessive mutant p53 alleles. There were 23 R72 and 20 P72 alleles in the transdominant group, compared with 50 R72 and only seven P72 alleles in the recessive group. Thus, there is an extreme bias in the selection of alleles between the two types of mutants (Table I; P < 0.0002, χ² test). With transdominant mutants, the frequency of the R72 allele (53%) is close to the germline frequency (59%), suggesting that there is little, if any, selective advantage in retaining the R72 allele when the p53 mutant is transdominant. In contrast, with recessive mutants, the frequency of the R72 allele (88%) is much higher than the germline frequency (59%), suggesting that there is a strong advantage in retaining the R72 allele when the p53 mutant is recessive (P < 0.0002, χ² test). This bias was seen with recessive mutants in every tumor type tested, although the number of samples is too small to reach statistical significance (R72:P72 in brain tumors 23:2; oral cancer 9:1; liver cancer 4:1; breast cancer 11:2; colon and pancreatic cancer, 4:0).

A simple model to explain this finding postulates that at the heterozygous stage in tumorigenesis a cell containing a transdominant mutant has reduced p53 activity, and consequently a growth advantage, irrespective of p73 activity. In contrast, a recessive mutant has near normal p53 activity at this stage, and is hence unlikely to be selected unless it can acquire some growth advantage by blocking p73 activity. Since the transdominant pathway requires only selection for a single variable it biases the p53 mutation spectrum towards transdominant mutants, which partly explains why hotspot mutants are commonly transdominant (22,23) (Table I). Failing that, p53 mutants resort to the subterfuge of hitting p73 in order to reduce the activation of p53 target genes (14). Since recessive mutants tend to be conformationally altered, our model is entirely consistent with the data of Marin et al. (14). Once the cell reaches the homozygous mutant stage by loss of the wild-type allele, a cell with a recessive, conformationally altered mutant has an advantage over its transdominant cousins because it can potentially achieve an even lower level of activation of p53 target genes—through inhibiting p73—than can most classic hotspot mutants. The 175H mutant is the striking exception to the rule that hotspot mutants tend to be conformationally intact (24). This mutant thus has the dual virtues of both inhibiting the wild-type p53 allele and inhibiting p73, perhaps explaining why it is so commonly found in tumors. The known properties of p73 lead us to suggest that the R72 selection in tumors with recessive p53 mutants reflects targeting of p73, but alternative models in which p53 targets other cellular growth regulatory proteins are also compatible with our data.

In conclusion, our finding supports the model proposed by Marin et al. and contributes one more piece to the puzzle of explaining the TP53 mutation spectrum seen in tumors. Since there is sufficient selective pressure to strongly bias the R72/P72 ratio in tumors, it is reasonable to expect that there may be differences in the behavior of tumors with one or other allele. Further studies to determine the biological mechanism and clinical consequences of selection of different codon 72 alleles are thus desirable.

References