The Impact of p53 Protein Core Domain Structural Alteration on Ovarian Cancer Survival¹

Ovarian carcinoma is the deadliest gynecologic malignancy. In 2003, there will be an estimated 25,400 new cases and 14,300 deaths in the United States alone (1). The mortality rate remains high because of the delay in diagnosis that characterizes ovarian cancer. Because of this poor prognosis, many attempts have been made to identify molecular prognostic markers that may prove to be useful both diagnostically and therapeutically on an individual patient basis.

p53 is a tumor suppressor gene that encodes a M_r 53,000 nuclear phosphoprotein involved in Gap 1 (G₁) cell cycle arrest (2). p53 dysfunction is one of the most common genetic alterations found in cancer, occurring in up to 50% of all human malignancies (3). We and others (4, 5, 6, 7, 8) have shown previously that p53 is mutated in ∼60% of ovarian malignancies. In lung and ovarian cancers, p53-null mutations are independent molecular predictors of compromised survival (5, 9). In contrast, survival with p53 missense mutation is highly variable. Unlike colon, lung, or breast cancer, p53 mutation in general was not found to compromise overall survival of ovarian cancer (5, 10, 11, 12). However, p53 protein truncating or null mutations were associated with early, distant metastases (13) and, thus, not surprisingly, with compromised survival relative to individuals whose cancer harbored p53 missense mutations(5). These findings suggest a hypothesis that not all missense mutations should be expected to be functionally equivalent.

Indeed, p53 mutations at highly conserved residues, DNA binding residues, or important residues that modify protein structure could be predicted to compromise p53 function more like a null mutation and be expected to compromise survival relative to missense mutations occurring at nonvital residues. This model, relating p53 structure to function has proven useful to prognostic outcome of colon (14), breast (15), non-small cell lung (16), as well as head and neck cancer (17).

As an alternative hypothesis, the mechanisms that give rise to p53 mutation (slippage mismatch effecting insertion/deletion at primarily pallindromic regions, as opposed to point mutations such as seen with truncation mutations at CpG residues; Ref. 18) may be a generalized process reflective of disease aggressiveness. According to this hypothesis, one would predict that cancers rendered p53 null on the basis of point mutation might behave differently from those rendered null by frameshift mutation. To test these hypotheses and to obtain a better understanding of the impact of p53 mutation as a prognostic indicator for ovarian cancer survival, we have carried out a detailed p53 structure function analysis based upon protein modeling. In addition, we tested the hypothesis that the poor prognosis associated with p53 null mutation is independent of the origin of mutation by stratifying the null mutations based upon point versus frameshift mutations.

A total of 267 patients were seen either initially or in consultation for the diagnosis of primary, invasive epithelial ovarian cancer at The Holden Comprehensive Cancer Center of The University of Iowa. Diagnosis was between January 1, 1990, and December 31, 1998. The study was carried out in accordance with the standards of the Institutional Human Subjects Protection Review Board.

Tumor was obtained from either fresh, snap-frozen specimens (n = 133) or paraffin-embedded samples (n = 134). DNA or cDNA isolation and p53 gene sequencing techniques have been well described in previous publications (4, 5). p53 mutations were characterized as null (frameshift or nonsense) and missense. We did not attempt to distinguish between expressed wild-type p53 sequence and potentially unexpressed sequences.

Because of the infrequency of p53 gene mutations in exons 2, 3, and 11 previously reported, only exons 4–10 were studied (5, 19).⁴ Exons 4–10 were directly sequenced using laser identification of fluorescent-labeled dideoxy chain-terminating nucleotides. We were unable to completely sequence some samples because of poor PCR amplification of DNA extracted from paraffin-embedded tissue. In these cases, SSCP⁵ gels were performed. The more sensitive SSCP gels require smaller concentrations of clear template to identify abnormal band migration, allowing SSCP to serve as a good screening method for p53 mutation. Forty-nine percent of all samples, both paraffin and fresh, had one or more exons examined by SSCP. In the case of abnormal SSCP gel migration pattern, DNA was re-extracted, and the exon in question was sequenced to confirm the presence or absence of mutation. Overall, we were no more likely to find a mutation from paraffin-embedded tissue than snap-frozen, indicating that artifacts were unlikely. Although some initial erroneous mutations were found in both paraffin-embedded and snap-frozen tissue, control with bi-directional sequencing and repeating individual template reactions enabled us to eliminate these false positives.

The effect of individual missense mutations on p53 structure was analyzed with the aid of the International Agency for Research p53 Mutational Database, which contains the results of a systematic automated analysis of the effects of p53 mutations on p53 core domain structure. This analysis was developed and reported by Martin et al. (20) and was used with the author’s permission. Mutations in the p53 core domain were classified as either those that could be explained or not explained in structural or functional terms by their predicted effects on protein folding or on protein-DNA contacts. The source of these effects were classified as perturbations of hydrogen bonding, mutations to proline, mutations from glycine, residue clashes, mutations in DNA and zinc-binding domains, as well as those occurring in highly conserved regions of the p53 gene. The specific codons in question that would lead to significant structural alterations in the protein were identified based upon crystal structure examination. There were 63 residues that are 100% conserved throughout all species of p53 sequences that were analyzed by Martin’s group. These 63 residues were classified as conserved for our analysis and include: 98, 113, 120, 121, 122, 125, 127, 130, 132, 137, 139, 142, 151, 152, 158, 159, 164, 172, 173, 175, 177, 178, 179, 196, 198, 199, 205, 208, 215, 216, 218, 219, 220, 221, 223, 230, 239, 240, 241, 242, 243, 244, 245, 247, 249, 251, 253, 257, 262, 265, 266, 267, 270, 271, 272, 275, 276, 277, 278, 279, 280, 281, and 282. Mutations in DNA binding residues were defined as those in which there was at least a 5% change between the complex form of p53 observed in the crystal structure and the same structure of p53 but with the DNA removed. These were identified as Ala¹¹⁹, Lys¹²⁰, Ser²⁴¹, Met²⁴³, Asp²⁴⁷, Arg²⁴⁸, Cys²⁷⁵, Ala²⁷⁶, Cys²⁷⁷, Arg²⁸⁰, and Arg²⁸³. Residue clashes were defined as substituted residues that resulted in three or more bad contacts with surrounding atoms in the best side chain orientation, leading to distortion of the structure and possibly incorrect protein folding. Mutations classified as those that could be explained, consisted of mutations found by analysis to significantly alter protein structure as well as those occurring in highly conserved or DNA binding codons. Mutations classified as structurally explained consisted only of the mutations found on analysis to significantly alter the protein structure. Null mutations were stratified by their mechanism of origin: point or frameshift mutations.

Frequency tables and descriptive statistics were generated to summarize the data. The Pearson χ² test was used to measure the association between structural alteration and the remaining predictor variables. Kaplan-Meier plots were constructed to provide estimates of the survival functions. The multivariable effect of structural perturbation upon survival was modeled using Cox proportional hazards regression. Analyses were performed with the SPSS statistical software package (version 10.0.1; SPSS, Inc., Chicago, IL).

Among the 267 tumors analyzed, 82 had missense, whereas 43 contained null p53 gene mutations. Patient follow up was last surveyed in August 2002; and at that time, median follow-up of survivors was 6.5 years. To date, 52 (63%) of 82 patients with tumor p53 missense mutations are dead of disease. Thirty-five (81%) of 43 patients with tumor p53-null mutations are dead of disease. Patients who died from causes unrelated to ovarian cancer were censored appropriately. The pathological characteristics of the study cancers and their correlation with the ability to explain the effects of specific missense mutations is listed in Table 1. The effects of 62 (76%) of the missense mutations could be explained by alterations in the p53 protein, whereas 20 (24%) could not. There were no relationships between stage, grade, residual disease, or nonserous histology and explained mutations. The breakdown of the individual missense mutations and their specific structural alterations is listed in Table 2. Twenty-three (28%) of the explained mutations occurred in highly conserved regions of the p53 core protein. Kaplan-Meier survival plots were constructed based upon the structural alteration and years of follow up. There was no survival difference between patients with tumor p53 missense mutations that could be explained (median = 4.4 years) and those that were not explained (median = 3.6 years; P = 0.43). There was no survival difference between patients with tumor missense mutation that could only be structurally explained (median = 4.4 years) and those that could not be explained structurally (median = 4.8 years; P = 0.98). Similarly, there was no survival difference when missense mutations occur at highly conserved residues of the p53 core domain (median = 4.8 years) as opposed to residues that are not highly conserved (median = 4.4 years; P = 0.41). There was no difference between missense mutations that occur at DNA binding residues (median = 4.0 years) versus those that occur at non-DNA binding regions (median = 4.8 years; P = 0.22). In a Cox multivariable analysis controlling for age at diagnosis, stage, grade, serous versus nonserous histology, and cytoreduction, explained mutations offered no prognostic significance (Table 3).

Finally, because survival of individuals with tumor p53-null mutations is compromised relative to those with missense mutation, we investigated survival based upon the type of p53-null mutation sequenced. Median survival when a mutation resulted from a point mutation (2.0 years) was identical to that when the null mutation resulted from a frameshift mechanism (1.9 years; P = 0.62). Overall, the presence of any null mutation significantly compromised survival relative to the presence of any missense mutation (median survival 4.0 versus 1.9 years; P = <0.01).

This is the first study to evaluate ovarian cancer survival based upon the structural alterations effected by p53 mutation. The many sequenced p53 mutations in ovarian cancer and the recent publication of the most systematic and complete analysis of the structural effects of p53 point mutations (20) afforded us the opportunity to evaluate ovarian cancer survival within these parameters. Although the role of p53 null mutation as a poor prognostic factor for lung (9) and ovarian cancer has been previously established, the majority of ovarian cancers contain p53 missense mutations (5, 7).⁴ Because of the highly variable course of disease when these cancers contain a missense mutation, it has been theorized that the difference may be secondary to the structural effects that an individual point mutation has upon the p53 protein. This variable hypothesis is quite plausible because it is well known that specific point mutations may cause significant changes to the three-dimensional properties of proteins, whereas others may have little, if any, effect. Support for this hypothesis is seen from the work of Reles et al. (21), who found decreased survival among patients with p53 mutations in highly conserved domains of the p53 gene, as opposed to mutations in nonconserved domains. In addition, some mutations have been shown to be flexible in that tertiary structure may vary between wild-type and mutant conformations (22). Indeed, the profiles of expression of p53 response genes can vary depending upon specific p53 missense mutations (23).

Support for the concept that p53 structural alteration is important for cancer prognosis is obtained from the work of Webley et al. (24), who examined p53 in 36 primary colorectal tumors by immunoprecipitation. Their investigations characterized tumor p53 as either wild-type, mutant, or flexible using the Pab240 antibody, which recognizes an epitope displayed by mutant but not wild-type p53 (25). Apparently, cancers with flexible p53 mutation demonstrated a trend toward more aggressive tumor behavior such as distant metastases and poor cellular differentiation (24).

Taking a different approach, the impact of p53 mutation at DNA contact residues has been studied in squamous cell carcinomas of the head and neck (17). Erber et al. (17) reported that p53 mutations at these sites were associated with tumor progression and resistance to therapy when compared with structural mutations and mutations outside of the core domain. Similarly, Berns et al. (15) examined clinical outcomes in 66 patients whose breast cancers contained p53 mutations. Mutations in codons directly involved with DNA binding displayed the poorest relapse-free and overall survival. We could not confirm the relevance of this observation to ovarian cancer.

Yet another critical p53 site is the L3 zinc-binding domain. Mutation at this site has been shown to predict poor survival both in patients with colorectal (14) and breast cancers (12). However, in the Borreson-Dale group study of colorectal cancer, all residues in the L3 zinc-binding domains were considered for analysis rather than focusing on the particular residues determined by crystallography to result in nonfunctional p53. Thus, overall, the adverse impact of p53 structural change is clearly supported in colon, breast, and squamous cell head and neck cancer but not in ovarian cancer.

In contrast, although we were able to show that a significant number (24%) of p53 missense mutations in ovarian cancer could not be explained either by critical residue mutations or resultant effect upon p53 protein structure, this finding did not designate a subset of missense mutation with a better prognosis. We have not only taken into account mutations that occur in critical areas involving conserved regions, DNA and zinc binding, but also have used a systematic automated analysis of how these mutations will effect the three-dimensional structure of the p53 core domain. Past studies have compared all residues involved in critical areas (including p53-null mutations) to those not involved, whereas we have limited the residues to those found to predict structural alteration based upon the structural examination used by Martin’s model. We have concluded from this detailed analysis that the structural alteration effected by specific missense mutations does not impact ovarian cancer survival.

This unexpected result was inconsistent with our working hypothesis. However, given the clearly adverse impact of p53-null mutation (median survival = 1.9 years) versus any p53 missense mutation (median survival = 4.0 years), an alternate explanation must be sought as to why a mutation at, for instance, a DNA contact residue, is not functionally equivalent to a p53-null mutation. After all, the contact site mutation can be shown to clearly alter protein structure and, thus, should alter transcriptional events. Our alternate hypothesis was that rather than the specific mutation as the factor impacting survival, p53 mutation may simply be a surrogate downstream event in ovarian carcinogenesis. This hypothesis was testable by looking at clinical outcome in the cancers rendered p53 null either by point mutation (analogous to the mechanism giving rise to missense mutations) or by a frameshift error resulting from either insertion or deletion, which is thought to reflect a slippage mismatch mechanism. Unfortunately, we must reject our alternate hypothesis as well. This leaves us with the conclusion that aside from ovarian cancer rendered p53 null, we cannot explain differential clinical outcomes in the majority of advanced ovarian cancers with missense p53 mutations found by any conventional methods of structure-function relationships, including the sophisticated Martin model.

We thank Melanie Hatterman-Zogg and Lisa Blake for technical assistance.