Introduction
The p53 tumor suppressor protein, encoded by the
TP53 gene, is a transcription factor that when activated as part of the cellular stress response, regulates suites of genes involved in cellular processes including the cell cycle, apoptosis, and senescence [
1]. Mutations in
TP53 are amongst the commonest genetic alterations in human cancer, but unlike other tumor suppressors,
TP53 is not usually inactivated through deletion or truncating mutations [
2]. Instead, it commonly undergoes mono-allelic missense mutations affecting the DNA binding domain, leading to the production of a protein that lacks DNA binding activity but remains capable of binding to, and hence dominantly inactivating, wild-type p53. Some
TP53 mutations also result in the acquisition of new oncogenic properties, and it is unclear whether loss of wild-type function, gain of oncogenic function, or their combined effects, account for the oncogenicity of these mutations [
1,
2]. Interpretation of many experiments addressing this issue is complicated by the recent identification of multiple p53 isoforms, which arise from alternative splicing and the presence of an internal promoter [
3]. The principal isoforms differ in the domain responsible for oligomerization, but are identical through the transactivation and DNA binding domains in which many
TP53 mutations occur.
Wild-type p53 protein is rapidly degraded, except under conditions of cellular stress.
TP53 mutations are often, although not always, associated with the production of a stable protein that is readily detectable by immunohistochemistry (IHC) of cancer cells [
1,
2]. IHC detection of p53 protein is therefore loosely, but imperfectly, associated with mutations in
TP53. Some studies assessing p53 status using either IHC or mutational analysis have concluded that
TP53 mutation is associated with poor prognosis, but other authors have reported no impact of
TP53 mutation on outcome in early breast cancer, and the evidence is not sufficiently strong for p53 status to be recommended as a marker in routine clinical practice [
4]. Studies of women treated with a variety of chemotherapeutic regimens, or hormonal therapy suggest that p53 status may be predictive of response to therapy [
5,
6]. However, until recently few large studies have either considered, or been able to effectively control for, treatment effects. Retrospective analyses of randomized clinical trials, using either IHC measurement of p53 expression (CALB 9344) or
TP53 gene sequencing (BIG 02-98), have identified a significant association with a worse prognosis in patients treated with adjuvant doxorubicin and cyclophosphamide or doxorubin but no significant association with response to taxanes [
7,
8]. Similarly, a prospective clinical trial (EORTC 10994/BIG 1-00) found that
TP53 status was not predictive of response to taxane-based therapy [
9].
The different methods used to assess p53 status have both advantages and disadvantages. Direct identification of mutations by sequencing has the advantage of pinpointing the specific aberration, but is not currently applicable to clinical practice because of its expense, technical difficulty, and likely reduced sensitivity in the routine setting. In addition, it may mis-classify some functionally equivalent genomic alterations, where
TP53 is inactivated indirectly. Functional assessment of p53 activity in yeast offers a distinction between inactivating and other mutations [
10,
11]. Several laboratories have attempted to develop gene signatures of p53 inactivation, which can potentially be measured using a PCR-based test [
12‐
14]. However, these assays are not immediately practical for routine diagnosis, and some of these signatures are potentially measures of subtype rather than p53 status, since mutations in
TP53 are more frequent in tumors lacking estrogen receptor (ER) expression [
15‐
17], in tumors of basal-like, HER2 and luminal B subtypes [
18,
19], tumors showing stem-cell like transcriptional patterns [
20] and in those with high proliferative fraction [
21]. IHC detection of p53 protein mis-classifies some
TP53 mutations, and instances where
TP53 is inactivated through deletion or truncation, or where persistent cellular stress leads to sustained expression of p53 protein. However, it does have the advantage of ready transfer into clinical practice, if studies in large, well-characterized cohorts provide good evidence of potential utility as a biomarker. Since the material available for the present study was limited to tissue micro-arrays, IHC was the most suitable available method.
Evidence for cross-talk between ER and p53 pathways at several levels suggests that the impact of
TP53 mutation may be affected by the presence of ER. First, ER is a p53 target and conversely estrogen treatment induces p53 expression, although it has also been reported to increase the cytoplasmic localization of p53, thereby functionally inactivating it [
22‐
24]. Second, p53 is required for hormonal protection against carcinogen-induced mammary cancer in rodents [
25,
26]. Finally, the ER and p53 proteins physically interact, leading to repression of their transcriptional activity and protection of p53 from degradation [
27‐
29]. However, other authors have concluded that the dominant interaction is via ER and p53 binding their cognate response elements
in cis to cooperatively regulate p53-responsive genes [
30]. Simultaneous allowance for p53 and ER in survival analysis of patients with early breast cancer, where the presence of ER is a favorable prognostic factor [
31‐
34], and predicts the efficacy of endocrine therapies [
31], has yielded conflicting results. Some authors report independent adverse prognostic significance of ER-negativity and p53 expression [
15,
35] while more recently, different p53 gene signatures have been associated with disparate prognosis and response to cytotoxic therapy in ER-positive and ER-negative disease [
12].
As a step towards addressing some of these issues, we have explored the relationship between ER expression and p53 expression detected by immunohistochemistry on prognosis in the context of large prospective randomized clinical trials. In the present study we have examined available pathological material from International Breast Cancer Study Group (IBCSG) Trials VIII and IX, comparing endocrine adjuvant therapy alone versus combined modality chemo-endocrine therapy in patients with node-negative early breast cancer. This revealed a qualitative interaction between ER and p53 expression, such that p53 expression was prognostically adverse in patients whose tumors expressed ER, but favourable in those whose tumors lacked ER expression.
Discussion
The present study is unusual in demonstrating a significant qualitative interaction between two biological markers. Detectable p53 expression was associated with better prognosis in patients whose tumors did not express ER but with worse prognosis if ER was expressed. This interaction appeared robust in the two trials examined and was independent of other pathologic features and of treatment. However, this and the similar interaction between p53 status and intrinsic subtype (as assessed by IHC) were unexpected and therefore require confirmation in large independent data sets for which p53 status is known to be available, such as that recently described by Lara
et al. for the Cancer and Leukemia Group B [
7], the EORTC 10994/BIG 1-00 trial described by Bonnefoi
et al. [
9], or the BIG 2-98 trial described by Francis
et al. [
48].
If confirmed, such an interaction might at least partially explain the earlier disparate reports of the prognostic significance of p53 staining viewed in isolation. Based on our exploratory analysis of a higher ER cut-point we suggest that any confirmatory study should look primarily at ER-present (IHC
> 1% staining) versus ER-absent staining. Our observation that the interaction is most clearly seen with an ER cut-point reflecting ER expression versus absence of expression may explain the failure to observe such an interaction in earlier small studies which typically used a higher ER cut-point [
15,
16,
35].
If the interaction between ER and p53 as markers of prognosis is biologically real, its basis is currently unclear, although there is evidence for functional relationships between p53 and ER that affect mammary oncogenesis and/or response to tamoxifen. In genetically engineered mice,
p53 heterozygosity leads to increased mammary epithelial proliferation, decreased apoptosis and eventual development of pre-neoplastic mammary lesions [
49]. These responses are all enhanced in the presence of deregulated ER expression [
49]. A direct interaction between p53 and ER has been described to inhibit p53-responsive transcriptional activation in mammospheres, but this response is antagonised by tamoxifen, suggesting a mechanism that could contribute to a better response to tamoxifen in women with wild-type p53 [
50]. Our observation that p53 positivity is associated with a worse prognosis in women treated with tamoxifen in the context of a randomised clinical trial is consistent with this idea.
TP53 mutations are not only more common among breast cancers not expressing ER, and among the basal-like and HER2 molecular subtypes, which typically lack ER expression [
18] but also tend to be different in type [
12]. Mutations are non-randomly distributed along the p53 domains [
5,
51]. Different p53 mutations have been described as carrying differing adverse prognostic significance [
5,
8,
17]. It is possible that p53 staining by IHC seen in tumors not expressing ER reflects mutations that are not disabling, or are less disabling, to cell homeostasis than those responsible for p53 IHC staining in tumors expressing ER.
As well as the well-studied canonical p53 protein, normal breast tissue expresses the p53β and p53γ isoforms, which arise from alternative splicing and differ at the carboxy-terminus [
3]. Most
TP53 mutations will result in changes in all three isoforms, and there are as yet no reagents for IHC that distinguish between them. However, different patterns of isoform expression are apparent in breast cancer: p53β mRNA expression is associated with ER expression, while p53γ mRNA expression is associated with
TP53 mutation [
52]. Each has been detected by PCR in 36 to 37% of breast cancers, but only 19% were found to express both [
52]. Since the antibody used in this study recognises all three isoforms and the p53β protein isoform is more stable than the other isoforms [
53], it is possible that p53 IHC positivity represents a different spectrum of isoforms in cancers expressing, or not expressing ER. The isoforms are differently regulated and may have different functions [
3]. In particular, the physical interaction between ER and p53 occurs via the carboxy-terminal domain, which differs in sequence between the different p53 isoforms [
29]. Although patients expressing mutant p53 but not p53γ have been shown to have a particularly poor prognosis, those expressing mutant p53 and p53γ have a good prognosis, indistinguishable from patients expressing wild-type p53 [
52]. This complexity in the prognostic impact of
TP53 mutations may contribute to the interaction with ER observed here.
Future studies should attempt to clarify this relationship by use of material from patients with p53 staining to ascertain the nature of the p53 mutations involved and examine their prognostic significance. Meanwhile it seems prudent to encourage independent validation of the current data in other large randomized clinical trials and to interpret the prognostic significance of IHC detection of p53 in the context of ER expression.
Acknowledgements
We record with sorrow the recent death of Professor Rob Sutherland who, as Inaugural Director of The Kinghorn Cancer Centre and Director of the Cancer Research Program, Garvan Institute of Medical Research, was the driving force behind this collaboration with the International Breast Cancer Study Group.
The authors would like to thank Ms Joanne Scorer for assistance in preparing the manuscript. We thank the patients who participated in the clinical trials, and the contributions of the physicians, nurses, data managers, and pathologists of the International Breast Cancer Study Group.
This work was supported by the National Health and Medical Research Council (Program Grant 535903 EA Musgrove, RL Sutherland; Fellowship 427601 RL Sutherland), the Cancer Institute New South Wales (Translational Program Grant 10/TPG/1-04 RL Sutherland, EA Musgrove, EKA Millar, SA O'Toole; Sydney Catalyst Translational Research Centre 11/TRC/1-02 RL Sutherland; Fellowships 10/CRF/1-07 SA O'Toole, 11/CDF/3-26 EA Musgrove, 11/CDF/3-25 TJ Molloy), the Australia and New Zealand Breast Cancer Trials Group (RL Sutherland, SA O'Toole), the Australian Cancer Research Foundation (RL Sutherland, EA Musgrove), the Sydney Breast Cancer Foundation (SA O'Toole), the RT Hall Trust (RL Sutherland) and the Petre Foundation (RL Sutherland). Additional support for IBCSG Trials VIII and IX was provided in part by the Swedish Society for Cancer Research; the Frontier Science and Technology Research Foundation; the Swiss Group for Clinical Cancer Research (SAKK); the Swiss Cancer League; and the United States National Institutes of Health (CA-75362).
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
ASC, EKAM, SO'T, EAM and RLS participated in the concept and design of the study. ASC, GV, AG, MC, RDG and BG provided materials for the study. ASC, EKAM, SO'T, TJM, EAM, RLS, MMR and ZS performed data analysis and interpretation for the study. ASC, EKAM, SO'T, EAM and RLS drafted, revised and edited the manuscript. All authors read and approved the final draft of the manuscript.