Prognostic Significance of Mutations to Different Structural and Functional Regions of the p53 Gene in Breast Cancer¹

The p53 tumor suppressor gene encodes a 53-kDa nuclear phosphoprotein whose primary role is to maintain genomic integrity through cell cycle arrest, DNA repair, and apoptosis (1). Inactivation of the gene can occur through mutation, protein sequestration, or allelic loss and may contribute to tumorigenesis (2). p53 gene alterations have been reported to occur in over half of all human tumors and have been associated with poor prognosis in some but not all tumor types (3). Approximately 20% of breast cancers contain a mutation in the exon 5–8 region of the p53 gene. There is now firm evidence from several studies that such mutations are associated with worse patient survival (4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16).

The p53 protein consists of 393 amino acids that can be functionally divided into three domains (17). The NH₂ terminus (amino acids 1–95) controls the transactivational activity of the protein, the central region (amino acids 102–292) the DNA binding activity, and the COOH terminus (amino acids 300–393) is responsible for oligomerization, nonspecific DNA binding, and DNA damage recognition. The majority of mutations to the p53 tumor suppressor gene occur in the central region, where four of the five conserved regions present in the gene are located (17, 18).

The effect of various mutations on the structure of the p53protein has been investigated using monoclonal antibodies directed toward the central portion of the protein (19, 20, 21). The results showed that some p53 mutations in the core region of the protein induced a conformational change in its tertiary structure(denaturing mutations), whereas other mutations had no effect.

The importance of the central region of the p53 protein in DNA binding has been confirmed by crystallography studies (22). This revealed the presence of two β-sheets that act as a scaffold for two loops and a loop-sheet-helix motif within the core domain. Also identified were the four codons involved in binding of a zinc atom. Together these regions form the DNA binding surface of the p53 protein. These regions were also shown to be located within the conserved areas of the p53 protein.

The prognostic significance of mutations in different locations and functional domains of the p53 gene has been investigated by five groups for breast cancer (7, 8, 12, 15, 16). The first study involved a meta-analysis of 119 breast cancer patients with mutation and found that those with a p53 mutation in the zinc binding regions had poorest overall survival (7), a result subsequently confirmed by Gentile et al.(15) and Kucera et al. (16). The study by Bergh et al. (8) involved 69 mutations and found that those within conserved regions II and V were associated with significantly worse prognosis, whereas the study by Berns and colleagues (12) on 66 patients with mutations reported that those mutations affecting the DNA contact domain had the poorest prognosis (12).

We have previously used SSCP³ mutation screening methods to analyze the prognostic significance of p53 mutation in large, consecutive series of node-negative and node-positive (11) or exclusively node-negative breast tumors (14). We have now extended these studies to include a total of 1037 breast tumors, 178 of which had p53mutations. The large majority of these mutations have now been identified by DNA sequencing. This has allowed us to investigate the prognostic significance of different p53 mutation subgroups,including the “hotspot” codons, various exons, conserved and nonconserved regions of the gene, mutations that abolish the proteins’ability to bind to DNA, and those that cause the p53 protein to denature.

Tumor samples were collected from 1037 consecutive primary breast cancer patients undergoing surgery for their disease at the Sir Charles Gairdner and Royal Perth Hospitals in Perth (n = 374)and the Flinders Medical Center in South Australia (n =663). The clinical features of these tumors have been reported previously (11, 23). The median age of the patients was 58 years (range 18–93) and the median follow-up to July 1997 was 65 months (range 1–120 months). At the end of the study period 216 (21%)patients had died of their disease. Information on patient survival was obtained from the Death Registry, Health Department of Western Australia, hospital clinical records, and the South Australian Cancer Registry. Patients that had died of causes other than their disease were censored from the survival analysis at the time of death.

PCR-SSCP analysis for mutations in exons 4–8 of the p53tumor suppressor gene was carried out using a nonisotopic PCR-SSCP minigel system as previously described (11, 24). Tumor samples showing aberrantly migrating bands in two or more independent PCR-SSCP runs were considered to contain a mutation. The majority of mutations (126/155, 81%) detected in exons 5, 7, and 8 were further identified by DNA sequencing. To achieve this, aberrantly migrating bands containing a mutation were excised, and the DNA was eluted from the gel slice and re-amplified in a sequencing reaction as described previously (24). An additional 10 tumor samples suspected of containing one of the mutation hotspots were positively identified by running alongside control tumor DNA known to contain that particular hotspot and comparing the banding profiles.

Kaplan-Meier survival analysis was conducted for various p53mutations grouped according to the site of mutation or the possible functional effect of that mutation (see Tables 4,5,6). These groups included:

1. (a) the particular exon in which the p53mutation occurred;

2. (b) mutations within the “hotspot” codons 175, 245, 248,and 273 (17);

3. (c) denaturing mutations that directly affect the stability of the p53 protein: codons 143, 175, 245, 249, and 282 and the zinc binding codons 176, 179, 238, and 242 (19, 20, 21, 22);

4. (d) mutations that affect the p53 proteins’ ability to bind to DNA: codons 120, 241, 248, 273, 276, 277, 280, 281, and 283 (22);

5. (e) mutations that occur within the L2 and L3 loops (codons 163–195 and 236–251) and that affect the stability of the tertiary conformation of the p53 protein (22);

6. (f) mutations that occurred in the evolutionarily conserved regions of the p53 core domain: codons 117–142, 171–180,234–258, and 270–287 (18);

7. (g) mutation type, specifically whether the mutation was a single base substitution or a deletion/insertion; and

8. (h) transition or transversion single base substitutions.

The overall survival in each of these mutation subgroups was compared with the survival of the normal p53 patient group.

The χ² test was used to determine associations between the various prognostic variables of breast tumors and p53 gene mutation. The Mantel-Haenszel test for linear association was used to determine correlation with histological grade,which was treated as a continuous variable. Univariate survival analysis was carried out using the method of Kaplan-Meier and differences between survival curves were compared using the log-rank test. Multivariate analysis was conducted using Cox’s proportional hazard model with stepwise forward selection of independent variables based on the likelihood ratio. All tests were two-tailed and statistical significance was assumed when P ≤ 0.05. Analyses were carried out using the SPSS software package (Chicago,IL).

PCR-SSCP analysis of exons 4–8 of the p53 tumor suppressor gene revealed that 178 (17%) of the 1037 primary breast tumors analyzed displayed aberrantly migrating bands indicative of a p53 mutation. Twelve mutations were detected in exon 4, 67 in exon 5, 11 in exon 6, 42 in exon 7, and 46 in exon 8. The overall frequency of mutations detected in this series was similar to or slightly less than that reported in other molecular studies of p53 mutation in breast cancer (5, 6, 8, 10, 13, 25, 26, 27, 28), but in close agreement with the figure of 18%established by Pharoah et al. (29) in a meta-analysis of breast cancer studies. The relative distribution of mutations within the different exons was also similar to that reported in the IARC database of p53 gene mutations (30)as shown in Fig. 1A, with the exception of a lower proportion of mutations observed in exon 6 in the current study. The major mutation hotspots in breast cancer occur at codons 175, 245, 248, and 273, with minor hotspots at codons 249 and 282. In the present study the frequencies of mutation at the major hotspots (as a percentage of all mutations) were 9% for codon 175, 4% for codon 245, 10% for codon 248, and 5% for codon 273, accounting for 28% of all p53 mutations. A comparison of the mutation hotspot frequency between the present series and the IARC database is shown in Fig. 1 B. A subcohort of 374 tumors was also examined for mutations in exons 9 and 10; however, only one aberrant migration pattern was noted, which was not further characterized.

The associations between p53 gene mutation and various clinicopathological features of the breast tumor series are shown in Table 1. Significant associations were detected between p53 gene mutation and the poor prognostic features of high histological grade, large tumor size, and low hormonal receptor content, but interestingly, not with nodal status. Univariate survival analysis of p53 mutation and of the various clinicopathological features revealed that only patient age was not a significant prognostic factor in the overall and LNP patient groups(Table 2). Histological grade, tumor size, and p53 mutation were significant prognostic factors in LNN patients. Lymph node involvement, tumor size, p53mutation, and estrogen receptor status all retained significance in multivariate survival analysis of the overall and LNP patient groups(Table 3); however, p53mutation was the only factor that retained significance in the LNN subgroup.

Kaplan-Meier survival analyses were carried out on patient subgroups classified according to the type of p53 gene mutation as described in the “Materials and Methods.” The analyses were carried out on the entire tumor series (Table 4;Fig. 2,A) as well as on the LNN(Table 5) and LNP patient groups (Table 6). In each case the survival of patients with a particular mutation subtype was compared with that of the wild-type p53 patient group. Mutation in exon 6 was not associated with a significantly different survival to that of wild-type p53 in all three patient groups. The survival of patients with mutations in exon 7 was not significantly different to those with wild-type p53 in the overall and LNP patient groups but was significantly worse in the LNN group. Mutations in exons 4, 5, and 8 were associated with significantly worse survival in each patient group, except for exon 8 in the LNP group, which showed a strong trend. The survival of patients with p53 hotspot mutations was in all patient groups better than that of patients with nonhotspot p53 mutations, although it was worse than that of patients with wild-type p53.

Of the 136 mutations identified by this study (Table 7), 38 (28%) were classified as denaturing mutations and 38 (28%) as DNA contact mutations according to the criteria given in “Materials and Methods.” In all three patient groups (overall, LNN, LNP) denaturing p53 mutations conferred significantly worse prognosis than patients with wild-type p53 (Tables 4,5,6; Fig. 2,B). Patients with non-DNA contact mutations also did significantly worse than patients with wild-type p53 (Tables 4,5,6). In contrast,patients with DNA contact mutations did not have significantly worse survival in any of the three groups (Tables 4,5,6; Fig. 2,B). Forty-one tumors contained mutations within the L2 domain and 36 within the L3 domain, accounting for 23% and 20%, respectively, of all mutations identified. Mutations within the L2 domain were more aggressive than those in the L3 domain in all three patient groups. When mutations in the L2 and L3 domains were combined, survival was significantly worse in the overall (Fig. 2,C) and LNN, but not LNP, patients. Mutations that occurred outside the L2/L3 domain were associated with particularly poor survival in the overall (Fig. 2 C) and LNP patient groups.

Mutations in conserved domains of the p53 gene appeared to be less aggressive than those in nonconserved domains. In all three patient groups, single base substitutions (missense/nonsense) showed significantly worse prognosis, whereas deletion or insertion mutations appeared to be less aggressive. Transversion mutations were associated with a much worse prognosis than transition mutations.

Accurate information on adjuvant therapies was available for 374 of the patients in this study. Due to the relatively small number of patients in each treatment arm, only the “no treatment” and “radiotherapy only” were analyzed for specific associations between mutation type and patient survival (Table 8). Results show no significant relationship with any type of p53mutation in the “radiotherapy only” cohort, but this may merely reflect the small number of mutants in each subgroup. In the cohort subjected to no adjuvant therapy, a significantly poorer prognosis was noted for non-DNA contact mutations. Interestingly,deletions/insertions were also found to be associated with a poorer prognosis in this cohort.

At least seven studies (4, 5, 6, 8, 9, 10, 12, 13) in addition to the two (11, 14) from our laboratory have clearly established that p53 gene mutation is an independent marker of worse survival in breast cancer patients. The increased risk associated with this genetic alteration is in the order of 2–3-fold. If p53 mutation is to find application as a routine molecular prognostic marker, it is important to determine whether these mutations all confer similar properties, or whether some are associated with a more aggressive phenotype than others. Results from in vitro studies suggest that mutations affecting different sites in the p53 gene may result in different effects on the proteins’ activity, notably its transactivational and apoptotic functions (21, 31, 32, 33, 34, 35, 36, 37, 38). Can we also identify “high” or“low” risk p53 mutations in vivo that influence a tumors’ response to adjuvant therapy and/or overall patient survival? The low incidence of p53 mutation in breast cancer and the relatively good survival of these patients,especially those with early stage disease, means that very large numbers of tumors must be examined to answer this question. Using PCR-SSCP analysis we recently screened two series of breast tumors from Perth (11) and Adelaide (14), which together total more than 1000 cases. In each of these series a significant association was noted between the presence of aberrant SSCP profiles(mutations) and a poorer patient prognosis (11, 14), and as such a combined analysis as undertaken here would be appropriate. We have now sequenced the majority of mutations detected in these series,thus allowing us to investigate whether different types of mutation are associated with different effects on overall patient survival. The combined tumor series spans the period in which adjuvant therapies were being widely introduced and is therefore heterogeneous with respect to whether or not patients received these treatments, and the specific adjuvant therapy status was only known for 374 patients. Analysis by adjuvant therapy status was somewhat equivocal, mostly due to the smaller number of mutations in each treatment group. It is possible that a large multicenter study may be required to obtain sufficient data to analyze the effects of different mutations in relationship to adjuvant therapy with any degree of statistical power.

The incidence, relative distribution (Fig. 1,A) and prognostic significance (Tables 2 and 3) of p53 mutation in this tumor series all closely match those reported by most other studies of breast cancer (4, 5, 6, 7, 8, 9, 10, 12, 13). Furthermore,75% of mutations in the current study were located in the conserved DNA regions compared with 73% reported by Cariello et al.(39), and 48% in the L2/L3 structural regions compared with 53% reported by Borresen et al. (7). We believe these results validate our present analysis of the prognostic significance of various p53 mutation subgroups. The majority of workers to date have screened exons 5–8 for mutation because these contain the major conserved regions. However the results shown in Tables 4,5,6 suggest that exon 4 should also be screened as the mutations located here were associated with very poor prognosis. Exon 4 encompasses codons 33–125 and has recently been reported to contain domains involved in the induction of apoptosis (37, 38). Inactivation of p53-mediated apoptosis might therefore lead to more aggressive tumor behavior and consequently to shortened patient survival. If the maintenance of p53 tertiary structure is required for the induction of apoptosis, it would also explain why denaturing mutations were associated with poor prognosis in this study(Tables 4,5,6).

Five previous studies have attempted to correlate mutations in different domains of p53 with the survival of breast cancer patients (7, 8, 12, 15, 16). Three of these studies found that the survival of patients with mutations in the L2/L3 domains were worse than the survival of patients with mutations outside of these domains (7, 15, 16), although in the study by Kucera and collegues (16) on a cohort of lymph node and steroid receptor-positive patients, the association was of borderline significance. Although both these mutation groups were found in the current study to have worse survival than wild-type p53(Tables 4,5,6), no significant difference in survival between the L2/L3 and non-L2/non-L3 groups was observed in the overall patient series(P = 0.26). This was despite our study having almost twice as many cases with mutations in these groups than with the earlier studies, as well as having a longer patient follow-up period. In the node-positive patient group, non-L2/non-L3 mutant tumors actually showed worse prognosis than the L2/L3 mutant tumors (Table 6),although this did not reach significance (P = 0.13).

In the study by Bergh and collegues (8) mutations in two of the five conserved regions (II and V) were found to be associated with poor survival compared with mutations in nonconserved regions. The total number of tumors with mutations in these two conserved regions was only 10, however. In the present work (Tables 4,5,6) we found no significant difference in survival between patients with mutations in conserved or nonconserved regions (P = 0.50). Further analysis of mutations in conserved regions II (n = 14)and V (n = 27) also revealed no significant differences when compared individually with mutations in the nonconserved region(n = 36).

The study by Berns and colleagues (12) involving 53 patients with mutation found, as in the current study (Table 4), that those within conserved, nonconserved, L2/L3 loops, non-L2/non-L3 loops,and nondirect DNA contact regions all had worse survival than did patients with wild-type p53. However, in contrast to our findings, these workers reported that 10 patients with mutations in the direct DNA contact codons had particularly poor survival. Our study involving almost four times the number of patients found that direct DNA contact mutations were associated with very similar survival to patients with wild-type p53 (Tables 4,5,6). Although our study included two additional amino acids (codons 277 and 281)considered to be involved in direct DNA contact (22) this could not account for the discordant results, because of the 38 mutations in this category only one occurred at either of these two sites.

Reasons for discrepancies between the current results and those of the five other studies cited above may be related to small tumor numbers,the use of adjuvant treatments and differences between node-negative and node-positive patient groups. Our results suggest that all p53 mutation types with the exception of direct DNA contact mutants are associated with worse survival of breast cancer patients. DNA contact mutations accounted for about 25% of all mutations and two-thirds of these were in the hotspot codons 248 and 273. Results from in vitro studies may offer some explanation as to the very good prognosis of patients with these mutations. Mutations in codon 248 have been reported to retain some tumor suppressor activity (31, 33) and do not appear to be as aggressive as the arginine to histidine mutation of codon 175. Furthermore, the arginine to histidine mutation of codon 273 appears to be even less aggressive than mutation of codon 248 (32, 33, 35, 36). The relatively high frequency of codon 248 and 273 mutations suggests that selective pressures in favor of these must still exist, although their impact on patient survival in this cohort is minimal. Cell-specific differences in the biological activities of p53 make it difficult to generalize on the effect of the various mutation types in other tumor types. Nevertheless it will be interesting in future studies to compare the present results obtained in breast cancer with those from other tumor types such as colorectal cancer.