Aims: The majority of all breast cancers are hormone responsive, traditionally defined by the expression of oestrogen receptor (ER) α and/or progesterone receptors. In contrast to ERα, the clinical significance of the relatively recently identified ERβ is still unclear. This study aimed to define the relationship between ERβ and clinicopathological parameters in a mixed cohort of breast cancer and, furthermore, to investigate the impact of ERβ expression on disease outcome.
Methods: The immunohistochemical expression of ERα and ERβ was analysed in tissue microarrays containing a total number of 512 tumours with all incident breast cancers diagnosed at the Malmö University Hospital between 1988 and 1992.
Results: 78% of the tumours were ERα positive and 50% were ERβ positive. ERβ correlated positively with ERα (p = 0.001). In contrast to ERα, ERβ was not associated with any important clinicopathological variables. Furthermore, no overall prognostic significance could be demonstrated for ERβ. In the ERα-positive subgroup, however, a low expression of ERβ correlated with a decreased disease-free survival in patients receiving endocrine treatment (p = 0.003).
Conclusions: Although interrelated, ERα and ERβ seem to be differentially associated to clinicopathological parameters, and this would support the fact that they might have different functions in vivo. Furthermore, ERβ might be a predictive marker of response to endocrine therapy, although this needs to be confirmed in additional studies, preferably randomised trials.
Statistics from Altmetric.com
Although the aetiology of breast cancer is multifactorial, oestrogen hormone signalling pathways are critical for the onset and progression of disease. Oestrogen effects are mainly mediated by oestrogen receptors (ERs) α and β, the latter relatively recently identified1 2 In contrast to ERα, the functions of ERβ in the normal breast as well as in malignant conditions are still unclear. ERα and ERβ share a high degree of homology in the DNA-binding domain3 and would consequently be expected to share similar affinity to classical oestrogen response elements. However, ERα and ERβ differ in the N-terminal activation function 1 domain and the ligand-binding domain,4 and this might explain their different biological functions. In contrast to ERα, ERβ expression is relatively high in normal breast tissue (80–85%) and does not fluctuate with the menstrual cycle.5–7 Studies of ERβ expression in normal breast epithelium, ductal carcinoma in situ and invasive cancer report a decreased expression of ERβ along with tumour progression.8 9 The discovery of ERβ may add further prognostic or predictive information, and provide additional understanding of the varying response to endocrine treatment10 11 In order to gain insight into the clinical significance of ERβ, its expression in breast cancer tissue samples should be compared with well-established prognostic factors such as age, tumour size, lymph node status, Nottingham Histological Grade (NHG) and ERα expression. In the present study, we performed high-throughput tissue analyses of 512 cases of breast cancer in order to investigate the relationship between immunohistochemical ERβ expression and clinicopathological parameters, as well as the expression of ERα, progesterone receptor (PgR), Ki67 and HER2. Clinicopathological data were available for all patients, and treatment and survival data were available for 389 of the patients.
Breast cancer samples were obtained from a consecutive series of unselected patients diagnosed with primary invasive breast cancer at the Department of Pathology, Malmö University Hospital, between 1988 and 1992 (n = 512). Data on treatment and disease outcome had been collected for 389 of the patients, 161 of whom had received endocrine therapy and 228 had not. During that period of time, the criteria for allocating patients to endocrine treatment were not as clearly defined as today. Chemotherapy had been given to 23 patients only, 19 of whom had not received endocrine treatment. The median follow-up for patients without breast cancer event was 106 months (range 0–207).
Tissue microarray construction
Prior to the construction of tissue microarrays (TMAs), all tumours were histopathologically re-evaluated by two breast pathologists (LA and KJ). Two tissue cores (0.6 mm) were taken from areas representative of invasive cancer using a manual arraying device (MTA-1; Beecher, Sun Prairie, Wisconsin, USA) and marker heterogeneity issues were minimised by sampling central as well as peripheral tumour areas,12 except in tumours exhibiting a necrotic centre.
Sections (4 μm) sections were dried, deparaffinised, rehydrated and microwave treated for 2×5 min in a 10 mM citrate buffer (pH 6.0) and processed in an automatic immunohistochemistry (IHC) staining machine (Techmate 500; Dako, Copenhagen, Denmark). The antibodies used were anti-oestrogen receptor β (1:25, EMR02; Novocastra, Newcastle upon Tyne, UK),13 anti-HER2 (1:100, Z4881; Zymed, Carlsbad, California, US) and anti-Ki67 (1:200, MIB-1; Dako). ERα and PgR were analysed using the Ventana Benchmark system (Ventana Medical Systems, Tucson, Arizona, USA) with prediluted antibodies (anti-ER clone 6F11 and anti-PgR clone 16).
For evaluation of ERα, ERβ and PgR, the fraction of positively staining nuclei was categorised into negative (0–1%), weak (2–10%), medium (11–75%) or high (76–100%). For statistical analyses, variables were dichotomised into “negative” or “positive” using the clinically established cut-off for ERα and PgR at 10% positive nuclei. Ki67 was evaluated in a similar fashion and categorised into three groups (0–10, 11–25 and >25%). The evaluation of HER2 was performed according to current clinical practice.14
All stainings were evaluated independently twice (SB and KJ) and, in case of discrepancy, a third examination was performed to reach consensus.
Western blot validation
The ERβ antibody used in this study was evaluated on western blot and by IHC using five different breast cancer cell lines. The IHC staining showed clear nuclear reactivity and on western blot: the antibody produced a single band at a molecular mass of ∼59 kDa, corresponding to full-length ERβ, validating the specificity of the antibody (fig 1).
The χ2 test was used to explore the relation between categorised variables, and the Mann–Whitney U test was used for comparison of medians. Associations between linear parameters were calculated using Spearman’s two-tailed significance test. Recurrence-free survival (RFS) was estimated according to the Kaplan–Meier method and the log-rank test was used to compare survival between strata. RFS considered local, regional and distant recurrences. All statistical tests were two-sided and a p value of <0.05 was considered significant. Calculations were performed in SPSS V.13 (SPSS, Chicago, Illinois, USA).
The study was approved by the Ethical Committee at Lund University (LU 613-02).
Clinicopathological characteristics and distribution of IHC markers
Baseline patient and tumour characteristics in the cohort, as well as the distribution of the evaluated IHC markers, are presented in table 1. The study cohort was reduced to 351 cases for which ERβ expression could be assessed. The remaining cases were excluded due to lack of representative tumour or loss during IHC processing of the TMA slides.
ERβ expression in relation to clinicopathological parameters
The relationship between ERβ expression and important clincopathological parameters is demonstrated in table 2. ERβ was strongly associated with the other hormone receptors, but no significant associations between ERβ expression and important clinicopathological parameters could be found. ERα expression was strongly and inversely associated with NHG, tumour size, Ki67 and HER2 status, and positively associated with age at diagnosis. No significant relationship was found between ERα expression and nodal status.
In order to investigate whether different combinations of the ERs correlated with clinicopathological features or disease outcome, combined dichotomised variables were constructed and defined as ERα+/ERβ+, ERα+/ERβ−, ERα−/ERβ+ and ERα−/ERβ−. The distribution of tumours with different combinations of the ERs is also shown in table 2. Analysis of these four subgroups in relation to clinicopathological parameters did not reveal any further significant relationships (supplementary material, table 4).
Recurrence-free survival in relation to clinicopathological parameters and IHC markers
The impact of clinicopathological parameters and IHC markers on RFS was further tested in a Cox regression model (table 3), confirming the value of established important prognostic parameters such as NHG, tumour size, nodal status and proliferation. In contrast to PgR, no significant association to disease outcome could be demonstrated for either ERα or ERβ status, or different combinations thereof. Similar results were seen for overall survival (supplementary material, table 3b).
The prognostic value of ERβ in all patients is illustrated in fig 2A. Stratifying for ERβ status among ERα+ patients did not yield any further prognostic information (fig 2B). Among endocrine treated patients, 134 had ERα+ and 22 had ERα− tumours. The corresponding numbers for untreated patients were 175 and 33 respectively. Based on the outcome in endocrine-treated (fig 3A) and not treated (fig 3B) patients, we compared DFS for different groups in relation to the ERα and ERβ status of their tumours and found that in patients with ERα+/ERβ+ tumours (fig 4A), endocrine therapy had no influence on the outcome, whilst in the ERα+/ERβ− subgroup, disease-free survival was significantly better for patients that had not received endocrine therapy (fig 4B). Similar results were seen for breast-cancer-specific survival (data not shown), whereas overall survival stratified for endocrine treatment was similar in the ERα+/ERβ+ and ERα+/ERβ− subgroups (supplementary material, fig 4aa and fig 4bb).
The clinical significance of the relatively recently identified oestrogen receptor (ER) β, is still unclear.
Although strongly interrelated, ERα and ERβ are differently associated to clinicopathological parameters.
In contrast to ERα, ERβ was not associated with any important clinicopathological variables.
No overall prognostic significance could be demonstrated for ERβ.
ERβ may be a predictive marker for response to endocrine therapy.
We analysed the relationship between ERβ expression and well-established clinicopathological parameters in 351 unselected cases of primary breast cancer. The cohort represents a common distribution of tumours and patient characteristics15 and the clinical follow-up time was more than 10 years. The positive association between ERα and ERβ was highly significant and both ERs correlated positively with PgR, which is in line with the findings from several other studies.5 16 In contrast to ERα, ERβ showed no significant correlation to any established clinicopathological parameters, supporting the notion of their independent biological functions and tallying with other reports.10 11 16 19
We used a monoclonal antibody towards a C-terminal epitope (not present in ERβcx) and observed nuclear ERβ staining in MCF-7, T-47D, Cama-1, MB-MDA-468 and MB-MDA-231 cancer cell lines, and with western blot the antibody recognised a single band at the molecular mass corresponding to full-length ERβ. These cell lines have also been documented as ERβ+ by Vladusic et al.18 Existing reports on the applicability of different ERβ antibodies are quite divergent and difficult to compare since a wide range of fixation protocols have been used.16 17 In the high-throughput setting used in this work, no major variation in the staining quality between tissue specimens could be observed. The antibody selected for this study has previously been validated in our laboratory on whole sections of breast cancer diagnosed from the 1960s to 2005, and we found no major difference in the staining quality, despite different fixation protocols and an expected loss of antigenicity in older specimens (n = 30, unpublished data). Moreover, no major heterogeneity in the staining pattern could be observed on the whole sections, validating the TMA-based scores in this study. The specificity of commercially available ERβ antibodies may be disputed, but since the proportion of ERβ-negative tumours reported here (50%) is rather high compared with other studies, even those using a higher cut-off than 10%,16 19 we do not consider this issue a major caveat to our findings.
The prognostic and treatment predictive value of ERβ in breast cancer has recently yielded greater attention, but the published results are contradictory with some studies agreeing on a beneficial effect of ERβ expression in a prognostic and a treatment predictive context,11 20 whereas others found no prognostic value of ERβ alone in an untreated population.10 Our findings are in line with the latter, since we could not prove a survival benefit for ERβ+ patients, either in the entire cohort, or in the subgroup of untreated patients. However, in the subgroup of endocrine-treated patients co-expressing both ERs, a significantly improved disease-free survival was seen compared to ERα+/ERβ− patients. This finding is in agreement with several studies demonstrating an improved response to endocrine treatment in patients with elevated levels of ERβ expression.10 21 22 Gruvberger-Saal et al propose that ERβ may be a predictor of tamoxifen response in ERα− tumours,19 a finding that could not be confirmed in this cohort, since the subgroup of endocrine treated patients with ERα−/ERβ+ tumours was too small (n = 19). In that study, however, only tumours from tamoxifen-treated patients were evaluated, whereas in our study, outcome in relation to ERβ expression was investigated in both treated and untreated patients. It must however be emphasised that this was not a randomised trial and that treatment with endocrine therapy was associated with negative prognostic factors such as a low age, high grade, positive node status and large tumour size (data not shown). Nevertheless, these parameters were equally distributed in the subgroups of patients with ERα+ERβ− and ERα+ERβ+ tumours and an improved outcome upon endocrine treatment could only be observed in the latter, indicating a possible role of ERβ as a modulator of endocrine treatment response. A similar trend was observed for breast-cancer-specific, but not overall, survival.
In conclusion, we have found that, although strongly interrelated, ERα and ERβ are differently associated to clinicopathological parameters, indicating independent biological functions of the different ERs. Furthermore, ERβ may be a predictive marker for response to endocrine therapy, although this finding needs to be confirmed in additional studies, preferably including tumours from patients enrolled in randomised trials.
We thank Ms Elise Nilsson for excellent and skilful technical assistance.
Additional data are published online only at http://jcp.bmj.com/content/vol61/issue2
Funding: We are grateful for financial support from the Swedish Cancer Society, Swegene/Wallenberg Consortium North and Malmö University Research Funds.
Competing interests: None.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.