p53 Alterations and Protein Accumulation in Benign Breast Tissue and Breast Cancer Risk: A Cohort Study

The p53 gene product is a multifunctional transcription factor that is involved in regulating cell cycle arrest and apoptosis, facilitating DNA repair, and promoting chromosomal stability (1-3). Disruption of p53 function seems to have a pivotal role in carcinogenesis, and alterations in the p53 gene are observed in more than half of all human cancers (4). Changes in p53 might contribute to carcinogenesis by conferring a proliferative advantage to cells and/or by facilitating the accumulation of additional genetic changes, for example, by allowing aneuploidy, genetic instability, or other mutations to occur (5-8).

Mutations in the p53 gene are among the most common molecular changes detected in breast cancer (9). p53 mutations and p53 protein accumulation (possibly indicative of underlying mutations) have been detected in 13% to 70% of invasive intraductal carcinomas of the breast (9-18) and have also been detected in ductal carcinoma in situ (9, 15-17), in benign breast disease (19-23), and in normal-appearing breast tissue (23). Furthermore, loss of heterozygosity at 17p13.1 (the locus of the p53 gene; ref. 24) has been observed in normal lobules adjacent to breast cancers (25). p53 protein accumulation has also been shown immunohistochemically in the benign breast tissue of women with Li-Fraumeni syndrome (who are at high risk of developing breast cancer; ref. 26) and in the benign tissue adjacent to breast cancer in women with a cancer syndrome distinct from Li-Fraumeni syndrome (27). Overall, these observations suggest that p53 changes can occur before the development of breast cancer and indeed at the earliest detectable stages of progression (25), and they raise the possibility that such changes might be related to the risk of breast cancer development.

In a previous cohort study, we showed that p53 immunopositivity in normal or benign breast tissue was associated with a 2.5-fold increased risk of subsequent breast cancer (28). However, it has been suggested that molecular analysis of p53 should be performed in cancer-related studies rather than immunohistochemical analysis because the latter can neither identify the specific types of underlying mutations that result in immunopositivity nor detect frameshift or nonsense mutations or intronic alterations (29). Indeed, ∼25% to 33% of p53 mutations do not result in positive immunostaining (30-34). Hence, using immunohistochemistry alone may result in underestimation of the prevalence of p53 mutations. In contrast, however, p53 protein accumulation can occur in the absence of underlying p53 mutations (22, 23). For example, post-translational modifications of p53 (e.g., phosphorylation and acetylation) in response to cellular stress can result in its stabilization, accumulation, and activation in the nucleus (35). Given the limited concordance between the results of immunohistochemical and mutation analyses and the multiple points at which p53 function can be disrupted, assessment of the association between p53 and breast cancer risk requires studies that combine both approaches to the detection of p53 changes. Therefore, in the cohort study reported here, in which we expanded our previous nested case-control study population (28), we examined the association between p53 gene alterations and p53 protein accumulation in benign breast tissue, alone and in combination, and risk of developing subsequent breast cancer.

The investigation was undertaken as a case-control study nested within the cohort of 4,888 women in the National Breast Screening Study (NBSS) who received a histopathologic diagnosis of benign breast disease during the active follow-up phase of the NBSS. The NBSS is a multicenter randomized controlled trial of screening for breast cancer in 89,835 Canadian women recruited between 1980 and 1985 and followed actively until 1988 (36, 37). Women were eligible to participate if they were 40 to 59 years old and had no history of breast cancer (in situ or invasive). The NBSS was approved by the appropriate ethics committees, and the study described here involved the analysis of tissue and data from that study in accord with the approved study design.

In the NBSS, when there was clinical or radiological evidence of a breast lesion, patients underwent either a fine-needle aspirate or a surgical biopsy. Each histologic diagnosis was reviewed for study purposes by a reference pathologist. The present study was restricted to subjects who had no evidence of breast cancer (in situ or invasive) on their initial surgical biopsy as determined on review by the reference pathologist. Women with a history of previous benign breast disease were not excluded from participation.

Incident cases of breast cancer were ascertained by record linkage with the Canadian Cancer Registry, and a death clearance was performed by linkage to the Canadian National Mortality Database. These linkages yielded follow-up data to the end of 1993.

In the nested case-control study, cases were the 155 women with a histologic diagnosis of benign breast disease made by a reference pathologist during the active follow-up phase of the NBSS who subsequently developed breast cancer [the median interval between diagnosis of benign breast disease and subsequent breast cancer was 3.6 years (5th and 95th percentiles, 0.6 and 10.7 years, respectively)]. For the purpose of this study, cancer was defined as any form of breast carcinoma; there were 135 cases with invasive carcinoma and 20 cases with ductal carcinoma in situ.

Controls were women with benign breast disease who had not developed breast cancer by (but were alive at) the date of diagnosis of the corresponding case. Five controls were selected randomly (with replacement) for each case from those noncases available within strata defined by screening center, NBSS study arm, year of birth (if possible to the nearest year, and mostly within 2 years), and age at diagnosis of benign breast disease (for the latter two variables, attempts were made to match to the nearest year or age; generally, matching was accomplished to within 2 years). Hence, there was a total of 775 controls.

At the time of their enrollment in the NBSS, all participants completed a questionnaire that sought identifying information as well as data on potential breast cancer risk factors, including demographic characteristics, family history of breast cancer, and menstrual and reproductive history.

For the present study, hospitals and clinics storing the paraffin-embedded blocks of benign and malignant tissue were contacted, asked to send one representative block per lesion, and to indicate the type of fixative used and whether the tissue had been frozen before fixation. Sections from the blocks received were reviewed and classified according to the criteria developed by Page and Anderson (38) and without knowledge of the case-control status of the study subjects.

Sections (5 μm) were cut from the paraffin blocks, mounted on aminopropyltriethoxysilane (2%; Sigma Chemical Co., St. Louis, MO)–coated slides, and deparaffinized. The sections underwent antigen retrieval (microwaved in 10 mmol/L citrate buffer (pH 6.0) for 15 minutes at a medium-high setting) and immunostaining was performed as described previously using antibody reactive with p53 (DO-7; monoclonal; dilution, 1:40; Novocastra Laboratories, Newcastle upon Tyne, United Kingdom). Positive controls were sections from a paraffin-embedded breast cancer that was known to have a p53 mutation associated with p53 protein accumulation. The negative controls consisted of replacing the primary antibody either with PBS or with mouse nonimmune serum. Immunostaining and review of the immunostained slides were done without knowledge of the case-control status of the study subjects. Any nuclear staining was considered a positive reaction and cytoplasmic staining was considered nonspecific and interpreted as negative. If only occasional cells showed immunoreactivity, then the sample was considered negative. We have reported previously that there was ∼93% agreement (κ = 0.64) on the presence or absence of p53 immunostaining in the benign specimens re-read by the same reviewer without knowledge of the result of the first reading (28).

The p53 mutation analyses were performed without knowledge of case-control status. For these analyses, 5-μm-thick sections were cut from the paraffin blocks, dewaxed, and stained briefly in hematoxylin. The breast epithelium was microdissected, collected in a microfuge tube, and digested with proteinase K [Life Technologies, Burlington, Ontario, Canada; 0.5 mg/mL in 50 mmol/L Tris-HCl (pH 8.5), 10 mmol/L EDTA, 0.5% Tween 20] for at least 48 hours at 55°C. The proteinase K was inactivated by heating to 95°C for 15 minutes. An aliquot of the digest was amplified using PCR [α³³P]dATP and exon-specific primers under the conditions described previously (39). An aliquot of the reaction product was separated on an 8% nondenaturing polyacrylamide gel and processed for autoradiography. Each sample was run under two conditions (2% and 10% glycerol in the loading buffer). Potential mutations were identified as shifts in band mobility. Samples that showed an abnormal band migration in either one or both gels underwent repeat PCR-single-strand conformation polymorphism (SSCP) under the appropriate percentage of glycerol. If no shift was seen, then the sample was considered wild-type for that exon/portion of intron. If two different patterns were seen in the two PCR-SSCP gels, then another tissue section was cut, microdissected, and underwent PCR-SSCP. For the samples in which clear bands were not obtained for exons 5 and 10, the PCR was repeated using dITP instead of dGTP. For some samples, the exon 4 portion of the exons 3 to 4 partial PCR product was unreadable. For these samples, PCR and sequencing were done with primers that were specific for exon 4 only, in order to read the 3′ portion of exon 4 (39).

If the band shift was confirmed, the band was excised from the SSCP gel and the DNA was eluted into water. The DNA was reamplified by PCR using the same primers and the product was run on a 2% agarose gel. The band was excised and DNA extracted using QIAquick gel extraction kit. The purified DNA was manually sequenced using the Thermosequenase radiolabeled terminator cycle sequencing kit and the sense primer. This was followed by electrophoresis on a 6%, 8.3 mol/L urea, denaturing polyacrylamide gel and autoradiography. Negative controls were included as well as DNA obtained from cell lines with known mutations in p53 where appropriate.

Odds ratios (OR) and 95% confidence intervals (95% CI) for the risk of breast cancer associated with immunohistochemically detected p53 protein accumulation and p53 changes detected by SSCP and sequencing were estimated using conditional logistic regression models (40). p53 changes were examined overall (i.e., present/absent), according to whether they were exonic or intronic, and according to whether they were mutations or polymorphisms. Risk in association with p53 mutations was also examined after classifying them according to whether they were in conserved, structural (41), or functional (5) domains of the p53 gene, and after classification by biochemical phenotype (41). Adjusted OR estimates were obtained by including terms representing the following potential confounders in the regression models: history of breast cancer in a first-degree relative, age at menarche, age at first live birth, body mass index [weight (kg) / height (m)²], hyperplasia (ductal or lobular, with or without atypia), and menopausal status (premenopausal, perimenopausal, and postmenopausal). (Women who reported having had a menstrual period within the last year were defined as premenopausal as were those who had had a hysterectomy without bilateral oophorectomy and were <45 years old; those who had ceased having menstrual periods at least 12 months earlier without surgical intervention were defined as postmenopausal as were those who had had a bilateral oophorectomy and those who had had a hysterectomy only and were >55 years old; the remaining women were classified as perimenopausal.) All statistical tests were two-sided, and Ps < 0.05 were considered to be statistically significant.

Blocks or sections of paraffin-embedded benign breast tissue were obtained for 109 (70.3%) of the 155 cases and 459 (59.2%) of the 775 controls. This included 1 block for a case for whom no tissue was obtained for the corresponding controls and 50 blocks for controls for whom benign tissue was not obtained for the corresponding cases. Therefore, tissue for 108 cases and 409 matched controls was potentially available for analysis. p53 immunostaining was performed first followed by the p53 mutation analysis. For 28 subjects (4 cases, 24 controls), there was no breast epithelium in the tissue sections, rendering them unsuitable for immunohistochemical analysis. Hence, immunostaining was performed on 104 cases and 385 controls [the earlier report on p53 immunopositivity in relation to breast cancer risk (28) included 71 of these cases and 288 of the controls]. A further 22 cases and 58 controls were excluded from the mutational analyses either because DNA yield was poor or because the DNA that was extracted was too degraded to be analyzed. As a result of these exclusions, 82 cases and 327 controls yielded data on p53 mutations. Subjects for whom tissue was obtained differed little with respect to breast cancer risk factors from those for whom it was not obtained—the main differences were in age at first live birth (mean, 21.7 years for those for whom tissue was obtained versus 19.9 years for those for whom it was not) and body mass index (25.7 versus 24.8 kg/m²). In addition, only small differences were observed between subjects for whom mutational analysis was attempted and those for whom it were not (due to there being insufficient DNA, etc.), and there was no difference between these two groups with respect to their distributions by p53 immunopositivity (data not shown).

p53 immunopositivity in benign breast tissue was associated with a 2-fold increase in the risk of subsequent breast cancer (Table 1), whereas the presence of any type of p53 nucleotide change in the benign tissue was not associated with altered risk. When immunoreactivity and nucleotide changes were examined jointly, however, those subjects with both p53 immunopositivity and a p53 nucleotide change (all changes combined) had a 3-fold increase in breast cancer risk.

A total of 26 cases and 92 controls showed exonic changes (Table 2). The distribution of case subjects [n (%)] showing p53 mutations by exon (exons 2-11) was as follows: 2 (7.7), 0, 7 (26.9), 5 (19.2), 2 (7.7), 6 (23.1), 0, 1 (3.8), 1 (3.8), and 2 (7.7), respectively; for controls, the corresponding figures were 4 (4.3), 1 (1.1), 31 (33.7), 10 (10.9), 17 (18.5), 13 (14.1), 4 (4.3), 2 (2.2), 5 (5.4), and 5 (5.4). Those subjects with any exonic change were not at altered risk of breast cancer. Furthermore, when exonic changes were examined by type, those with nonpolymorphic exonic changes, exonic changes yielding amino acid changes, silent mutations, missense mutations, nonsense mutations, transitions, or transversions were not at altered risk. In contrast, individuals with mutations in the untranslated regions (exons 2 and 11) had a 5.7-fold increase in breast cancer risk of borderline statistical significance. However, only three cases and four controls showed such changes, and the associated 95% CIs were very wide. Overall, intronic p53 changes were not associated with altered breast cancer risk (Table 3). However, there was a statistically significant 2.8-fold increase in risk in association with nonpolymorphic intronic changes. Neither exonic nor intronic polymorphisms were associated with altered breast cancer risk (data not shown).

Mutations in the conserved domains of the p53 gene (domains II-V) were not associated with substantial alterations in breast cancer risk (Table 4). Although mutations in domain III were associated with a 75% increase in risk and mutations in domain IV were associated with a 57% reduction in risk, neither of these point estimates was statistically significant, and their associated 95% CIs were very wide. The risk of breast cancer in association with a mutation in any of the conserved domains was 0.80 (95% CI, 0.26-2.48).

The DNA-binding domain of the p53 gene involves the loop-sheet-helix motif (which includes the L1 loop, the S2-S2′ β-strand, amino acids 275-277, and the H2 helix) and the L3 loop (41). Changes in these components of the binding domain were not associated with altered breast cancer risk (data not shown). Furthermore, mutations neither in the DNA-binding domain as a whole nor within the region of the p53 gene that encodes the part of the p53 protein in direct contact with DNA were associated with risk (Table 5). Similarly, breast cancer risk was not altered in association with mutations in the zinc-binding domain (which involves the L2 and L3 loops), the region of the zinc-binding domain that encodes amino acids that bind directly to zinc, or the tetramerization domain (which includes a β-strand, a turn, and an α-helix; Table 5).

Although there was some suggestion that mutations in the oligomerization domain and in the COOH-terminal domain were associated with 80% to 90% increases in breast cancer risk (Table 6), neither of these associations was statistically significant, reflecting the small numbers of subjects showing these changes. Changes in other p53 gene functional domains were not associated with altered risk.

Table 7 shows breast cancer risk in association with biochemical phenotypes of p53 changes. Missense mutations inside the conserved domains, DNA contact domains, or zinc-binding domains, excluding the portion of these domains that encode amino acids involved in DNA contact or zinc binding, were not associated with altered breast cancer risk. No cases and only one control showed a missense mutation in the region of the gene that encodes amino acids directly involved in DNA contact or zinc binding.

The patterns described above were similar after exclusion of case subjects (and their matched controls) whose cancer was diagnosed within 1 year of their diagnosis of benign breast disease, after restriction of the case group to those with invasive cancer (i.e., after exclusion of those case subjects who had carcinoma in situ), after restriction of the analyses to those cases whose benign and malignant lesion occurred in the same breast, and (for analyses focusing on p53 immunopositivity) after exclusion of those subjects for whom slides rather than blocks of tissue were obtained and after varying the definition of p53 immunopositivity to require immunostaining in at least 10% or at least 50% of cells. However, some of the point estimates did differ somewhat from those obtained from the full data set, reflecting the relatively small number of subjects with p53 mutations at specific locations. The results were also similar when the matching was broken and the data were analyzed using unconditional logistic regression with adjustment for the matching variables, thereby allowing inclusion of cases and controls for which no corresponding control or case was available.

In the prospective study reported here, in which we enlarged our previous nested case-control study (28), we observed that p53 immunopositivity in normal or benign breast tissue was associated with a 2-fold increase in the risk of subsequent breast cancer. This estimate, although slightly lower than the one that we reported earlier, was more precise. In the current study, we also used DNA extracted from breast tissue to determine whether there were p53 gene changes. Overall, individuals with any p53 nucleotide change (exonic or intronic) were not at altered risk of breast cancer. However, mutations in the untranslated regions of the p53 gene (exons 2 and 11) were associated with a 5.7-fold increase in breast cancer risk (of borderline statistical significance), and nonpolymorphic intronic changes were associated with a statistically significant 2.8-fold increase in risk. In contrast, mutations in conserved domains of the p53 gene, and in domains defined structurally, functionally, or biochemically, were not associated with altered risk. Similarly, breast cancer risk was unaltered in association with polymorphisms, nonpolymorphic exonic changes, exonic changes yielding amino acid changes, silent mutations, missense mutations, nonsense mutations, transitions, or transversions. Examination of risk in association with combined levels of p53 immunopositivity and p53 nucleotide changes revealed that individuals who were positive for both had a 3-fold increase in breast cancer risk.

There have been three previous follow-up studies of p53 protein accumulation in benign breast tissue in relation to risk of subsequent breast cancer: one showed no association between p53 protein accumulation and risk of breast cancer (20); another (42) suggested that p53, when evaluated as part of a group of biomarkers, might be associated with risk; and in a previous study based on a subset of the subjects included in the present report, we showed that p53 protein accumulation in benign breast tissue was associated with a 2.5-fold increase in the risk of subsequent breast cancer (28). However, the first two of these studies were small and adjusted for confounding either incompletely (42) or not at all (20). Furthermore, all of these studies used immunohistochemistry to detect p53 protein accumulation, which may have resulted in underascertainment of p53 changes.

We are not aware of any previous prospective studies of p53 mutations in relation to breast cancer risk. In the present study, risk was increased noticeably in association with mutations in the untranslated regions of the p53 gene and in association with nonpolymorphic intronic changes. However, only the latter of these associations was statistically significant, and given the many associations that were examined, it is conceivable that it represents a chance finding. Nevertheless, there is increasing evidence that changes in noncoding DNA may affect disease susceptibility (43). In particular, cis-acting regulatory sequences (i.e., those portions of noncoding DNA that bind to transcription factors that activate or repress gene expression) seem to play a critical role in transcriptional regulation. In addition, p53 intronic changes have been detected by others in a variety of tumors, such as malignant pheochromocytomas (44) and the blast phase of chronic myeloid leukemia (45). In some tumors, intronic p53 mutations have been shown to alter cell functions, such as apoptosis and proliferation (46), and have been associated with stabilization of the p53 protein (44). Alterations in introns may affect gene splicing by introducing cryptic splicing sites, resulting in exon skipping, and they may also affect transcription, because it has been shown that transcriptional regulatory elements can be found near exon/intron boundaries (46). In the absence of functional studies, it is not possible to know whether the changes that we observed had an effect on the p53 protein product.

The combination of positive p53 immunostaining and the presence of p53 gene alterations was associated with an even greater risk of developing breast cancer than the presence of either alone. This combination may identify a group of individuals with changes that have substantial functional consequences. The relatively small study size precluded examination of specific p53 changes in combination with p53 immunostaining, but we are currently conducting a much larger cohort study in which such analyses should be possible.

Several limitations of this study have been discussed previously (28). In brief, although paraffin-embedded blocks of benign breast tissue were not obtained for all subjects, it is unlikely that tissue availability was related to p53 status; differential bias in the assessment of p53 status seems unlikely given that the assays were done without knowledge of the case-control status of the study participants; factors, such as fixative type and duration, limited sensitivity of the antibody, and, for a few subjects, access to slides only rather than tissue blocks, are most likely to have induced false-negative immunohistochemical results distributed nondifferentially between cases and controls with consequent biasing of point estimates toward the null; and the study was relatively small, which compromised its statistical power to detect small effects.

In addition to these issues, the limitations of the tissue samples and of PCR-SSCP and sequencing need to be addressed. With respect to the former, for some samples, it was not possible to sequence all of the exons because of the DNA degradation that results from formalin fixation and paraffin embedding. Furthermore, it is possible that too few cells had gene changes, which therefore could not be detected because most cells were wild-type. In this regard, it is of note that subjects who had positive immunostaining for p53 most commonly showed protein accumulation in <10% of cells. With respect to the latter, SSCP was a commonly used method to screen for genetic alterations when this study was initiated (47). Compared with direct sequencing, it is less labor intensive (48), but it can miss mutations. For example, it may miss cytosine-to-thymidine changes because they do not result in easily detectable conformational changes (49). The sensitivity of SSCP for detecting mutations can be as low as 62% in DNA extracted from formalin-fixed, paraffin-embedded tissue (50), so it is possible that by using this method we underestimated the gene changes, with consequent biasing of the OR estimates toward the null. We attempted to improve the sensitivity by varying the gel conditions, but the number of variations that can be used is tempered by the need to conserve DNA, a problem when small samples of tissue are being analyzed, as was the case in this study. Direct sequencing only detects mutations when ∼20% of cells have the gene change, which also may have contributed to underestimation of the p53 changes. However, we did repeat PCR and SSCP on 20 subjects who were negative for p53 mutations on first assessment and they all remained negative on repeat analysis. To minimize the number of false positives, each SSCP analysis that showed an abnormal band migration was confirmed by repeating the analysis using product from a different PCR. Nevertheless, the possibility of false positives cannot be excluded.

In conclusion, the results of this study suggest that p53 changes (e.g., p53 immunopositivity and nonpolymorphic intronic changes) detected in normal or benign breast tissue are associated with increased risk of subsequent breast cancer. Future studies should be large (which would allow assessment of effect modification by factors, such as menopausal status), should assess both p53 immunopositivity and mutation status, and should use state-of-the-art methods for p53 mutation detection. Although the present report involved the use of paraffin-embedded breast tissue, the analyses described herein can also be done on tissue obtained from needle biopsies and on cells obtained from fine-needle aspirates of the breast, both of which are done increasingly commonly nowadays (42, 51-53). If confirmed, the findings reported here might facilitate the development of new approaches both to the detection of women at increased risk of breast cancer and to their clinical management [e.g., by enabling chemoprevention or therapy to restore p53 function (54) to be targeted to specific subgroups of women].

We thank Taj Bhardwaj for his excellent technical assistance.