Introduction
It has been shown that longer exposure to estrogen results in an increased risk of developing breast cancer (BC) and endogenous estrogens are thought to play a major role in BC carcinogenesis [
1]. Moreover, in BC, estrogen receptor-alpha (ER-α) and progesterone receptor (PgR) are well-established biomarkers capable of predicting the likelihood of relapse/progression in response to endocrine therapy. The identification of a second type of ER, named estrogen receptor-beta (ER-β) [
2], has prompted the re-evaluation of the model of estrogen action. To this end, a number of studies have been conducted retrospectively on selected series of invasive BC to evaluate the predictive value of ER-β in patients submitted to endocrine therapy [
3]. Unlike ER-α, antiestrogen-occupied ER-β can activate transcription via nongenomic ER signaling pathways that involve the activation of cytoplasmic signal transduction cascades such as the Src/ERK and the PI3K/Akt pathways [
4]. ER-α and ER-β can mediate the biological effects of estrogens and antiestrogens by modulating the expression of specific target genes. At present, however, limited information is available concerning the differential modulation of gene expression from either ER-α or ER-β, which share a high degree of homology in the DNA-binding domain but differ considerably in the NH
2-terminal region and, to a lesser extent, in the ligand-binding domain. Because of this lack of sequence similarity, it has been suggested that the two receptors might perform distinct functions [
5].
Another important steroid receptor involved in BC progression is the PgR, which plays a pivotal role in the action of progestins in target cells and tissues. In invasive BC, PgR expression is generally regarded as a marker of an intact ER-α signaling pathway [
6]. ER-α and PgR positivities correlate with favorable prognostic features and are predictors of response to hormonal therapy (HT), both in the adjuvant setting and in advanced disease. In contrast, much less is known regarding the contribution of ER-β to estrogen-driven responses [
7] or its prognostic/predictive role in different early BC risk groups treated with different chemotherapeutic/hormonal regimens. This picture has recently been complicated further by the introduction of gene profiling approaches [
8] and by the widespread application of a novel BC classification based on the immunohistochemistry (IHC) phenotypic patterns identified by a few protein biomarkers, namely ER-α, PgR, HER2, epidermal growth factor receptor (EGFR), and low-molecular-weight cytokeratins [
9]. According to the expression of such markers, BC can now be divided into four main subtypes that have distinct behavior in terms of prognosis and response to therapy [
8]: luminal A (LA) and luminal B (LB), characterized by high expression of ER-α; triple-negative (TN), characterized by EGFR and/or by some basal epithelial markers such as cytokeratin 5 positivity; and HER2, characterized by the lack of hormonal receptors. To date, there are no published data concerning the distribution of ER-β among these different molecular subtypes of BC. The aims of the present study were (a) to prospectively evaluate the relationship between ER-β and a number of established biopathological parameters in an observational prospective series of 936 BC patients and (b) to analyze the impact of ER-β expression on clinical outcome and on the response to different therapeutic regimens, taking into account the novel molecular classification.
Discussion
To the best of our knowledge, this is the first study in which the relationship between ER-β expression, established biopathological factors, and patient outcome was investigated in an observational prospective study including a large series of invasive BC, consecutively accrued over a relatively limited and recent period of time (2001 to 2005). While ER-α and PgR expression was significantly associated with HER2, Ki67, p53, Bcl2, T, N, and G, as one would expect in a representative well-balanced cohort of unselected patients with early BC, we observed no significant association between ER-β expression and the classical biopathological parameters. Our findings, while in agreement with other recently published studies [
3,
18], differ from others that have found that ER-β is coexpressed with ER-α and PgR and is associated with nodal status, grade, proliferation rate [
17,
19], and HER2 overexpression [
20]. Inconsistencies among different studies are possibly due to different techniques in determining ER-β expression and lack of validated reagents for IHC. Jarvinen and colleagues [
17] and Umekita and colleagues [
20] used the polyclonal antibody PAI-313 and the MoAb EMR02, respectively, whereas Omoto and colleagues [
19] did not specify the reagents used for their IHC analysis. In our study, we used two well-characterized anti-ER-β MoAbs, PPG5/10 (ER-β1) and 14C8 (total ER-β), which have previously been shown to be the best performing antibodies for IHC staining [
21] with superimposable results. Moreover, further discrepancies could also be related to the selection of different cohorts of patients. In fact most authors included in their study only retrospective series of BC patients dating back to the early 90s, whereas our series is prospective and includes patients treated between 2001 and 2005.
We took this kind of analysis one step further and also examined the distribution of ER-β among different, molecularly distinct, BC subtypes. Gene expression profiling [
8] has, in fact, led to the identification of subtypes of invasive BC with different outcomes, namely LA, LB, TN, and HS. Such classification has since been translated into routine clinical practice by combining a limited set of markers (ER, PgR, HER2, and basal cytokeratins) that can be assessed by IHC [
9]. We stratified our 936 BC patients according to these molecular subtypes and found that ER-β evenly distributes across the four subtypes, as recently reported by other authors [
22]. Such results were confirmed by MCA [
23,
24], an alternative method for analyzing multiple categorical variables by graphically visualizing their interrelationships [
25], which showed that ER-β expression presents a limited dispersion around the origin, regardless of the method used to classify all of the variables, that is, by clustering them into discrete subgroups (Figure
3a) or considering them individually (Figure
3b). These findings strongly support the lack of correlation between ER-β expression and the other biopathological parameters considered, further validating the hypothesis that ER-β has functions that are distinct from those of ER-α [
26].
The lack of association between ER-β and other classical prognostic factors makes it an even more attractive candidate as a prognostic/predictive biomarker. The impact of ER-β expression on disease outcome (in terms of DFS) was therefore studied in a subset of 728 patients with a median follow-up of 50 months. Using a nonparametric statistical procedure, C&RT analysis, we were able to identify ER-β as a discriminating factor in two very interesting subgroups of patients: (a) node-positive patients, in whom ER-β
+ appears to convey a higher risk of relapse, particularly when coupled with PgR negativity, and (b) node-negative patients, in whom ER-β
+ appears to predict a favorable response to endocrine therapy. These results were substantially validated by conventional statistical procedures, such as Kaplan-Meier analysis of DFS curves and univariate and multivariate Cox regression analysis, both of which seem to indicate a divergent role of ER-β expression as a positive predictive factor in node-negative patients subjected to HT as well as a negative prognosticator in node-positive patients which does not predict the response to any therapeutic regimen. Though based on a limited number of DFS events, the finding of a positive influence of ER-β expression on the outcome of node-negative BC patients treated exclusively with HT is supported by several other reports in which the predictive value of ER-β, detected by mRNA or IHC staining, was investigated in BC patients undergoing endocrine therapy [
3,
27,
28]. In these studies, positive ER-β protein staining was invariably almost associated with a favorable response to antiestrogen treatment, consistent with its antiproliferative and anti-invasive properties observed in ER-β-expressing cell lines [
29]. Conversely, to the best of our knowledge, this is the first study in which ER-β expression is unexpectedly found to be significantly associated with an unfavorable prognosis in node-positive in an observational prospective series of BC patients. This is in agreement with data reported in prostate cancer providing evidence that ER-β
+ tumors had a higher rate of relapse and a small but significant decrease in relapse-free survival compared with those in which ER-β expression had been lost [
30]. One likely explanation for our findings in BC is that all of the previous studies that have measured ER-β in BC have focused on response to tamoxifen therapy in either adjuvant or metastatic settings [
26,
31], while our subset of node-positive BC patients mostly received adjuvant CHT (with or without ACs). It is interesting to note that the divergent role of ER-β expression is maintained even when established pathological factors are clustered together into distinct molecular subgroups in the context of a widely used clinical translation of gene expression profiling studies (LA, LB, TN, and HS). Indeed, depending on nodal status, ER-β expression might usefully complement the prognostic assessment of patients in those subgroups (LA and LB) where further risk stratification by gene expression analysis is needed to accurately predict prognosis [
2,
20]. In this context, ER-β positivity might signal responsiveness to hormonal treatment in node-negative LA patients, on one hand, and a more aggressive clinical course, requiring
ad hoc tailored therapeutic interventions, in node-positive LB patients, on the other, thereby possibly contributing to the implementation of individualized therapeutic strategies.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
FN and M Milella were responsible for the study design; produced, acquired, analyzed, and interpreted the data; and drafted the manuscript. FN and M Milella contributed equally to this work. EM and SB provided assistance in data acquirement and interpretation and in manuscript drafting. IS was responsible for the database set-up and for the statistical analyses. LP and RP-D revised all the slides submitted to immunohistochemical staining and confirmed the histological diagnosis. IV, CN, AF, and ADB provided clinical data of the patients, including treatment schedule and follow-up. PGN critically revised the manuscript for important intellectual content. M Mottolese was responsible for the study concept and design and for the interpretation of results, helped in data discussion, critically revised the manuscript for important intellectual content, and obtained funding for the study. All authors read and approved the final manuscript.