Background
Estrogen actions in tissues are mediated by two structurally related but genetically distinct receptors, estrogen receptor (ER) α and ERβ [
1,
2]. In the normal breast, ERα is expressed in a modest subset of luminal epithelial cells where it mediates proliferation and breast growth. ERα expression is common in breast cancer (BCa), with 75% of tumors being ERα positive. It is an important biomarker for response to anti-estrogen therapy and is widely used in diagnosis and treatment planning. On the other hand, ERβ is more abundant than ERα in normal mammary tissue [
3] and expression is often diminished or lost in BCa [
4‐
6]. ERβ exists as five isoforms, ERβ1-5, but isoforms 2–5 are C-terminally truncated and cannot bind ligands [
7,
8] leaving only ERβ1 as the functional receptor for estrogen ligand action.
Triple negative breast cancer (TNBC) is defined by a lack of expression of ERα, PgR and HER2. These tumors make up 10–15% of all BCa and characteristically recur early with the peak risk of recurrence and the majority of deaths occurring within the first three and five years after the initial treatment, respectively [
9,
10]. They are associated with an inferior prognosis despite their greater sensitivity to cytotoxic chemotherapies in the neo-advujant, adjuvant and later in the metastatic settings. Thus, additional targeted therapies for TNBC are needed.
Tamoxifen is a selective ER modulator (SERM) that competitively binds to ERα and blocks estrogen binding. It is prescribed for the treatment of ERα + BCa due to its ability to inhibit estrogen-stimulated proliferation in cancer cells. There is a small benefit for the rare ERα negative, PR positive cancers and guidelines recommend that these patients be given endocrine therapy. Patients with ERα negative PR negative tumors in general do not benefit from tamoxifen therapy, although a modest proportion (5–10%) show sensitivity to tamoxifen [
11,
12].
Tamoxifen action on signaling targets other than ERα has been proposed as a mechanism to explain sensitivity in ERα negative tumors. Three studies have shown that ERβ1 expression acts as a marker for favorable prognosis in tamoxifen-treated ERα-negative [
13‐
15] and TNBC patients [
14] indicating that ERβ1 is clinically relevant. However, a lack of placebo treated patients for comparison in two studies [
13,
14] prevented robust establishment of whether ERβ1 acts as a general prognostic marker or as a predictor of tamoxifen sensitivity.
Numerous studies have assessed the frequency of ERβ1 expression in TNBC. The spectrum of ERβ1 positivity in these studies ranges from 35 to 75% [
13,
14,
16‐
19]. However, two recent publications have questioned the specificity of many previously employed ERβ antibodies, including the PPG5/10 ERβ1 antibody used in most studies [
20,
21]. In both immunohistochemistry (IHC) and western blotting, the PPG5/10 ERβ1 antibody, which targets the carboxterminal end of ERβ1, demonstrated low specificity, showing positivity in ERβ1-negative control lines [
20,
21]. This calls into question the validity of data from existing studies of TNBC, which have largely used the PPG5/10 antibody.
Nelson et al. used antibody-dependent (IHC and western blotting) as well as antibody-independent (RT-qPCR) analysis to confirm the specificities of multiple antibodies and validated MC10 (targets the N-terminus) and CWK-F12 (targets the ligand binding domain) antibodies as being specific for ERβ1 [
21]. Rapid immunoprecipitation mass spectrometry of endogenous protein (RIME) analysis identifies the specificity and peptide coverage of antibodies, including ERβ1, without the need of another antibody dependent technique such as western blotting, where one must rely on the migration mobility of a band. CWK-F12 also performed very well in the RIME analysis and demonstrated differential IHC nuclear staining of ERβ1 between MDA-MB-231 cells with inducible exogenous ERβ1 expression and control cells [
21].
In view of this, we sought to determine the true percentage of TNBC that express ERβ1 and any relationship with clinical outcome using CWK-F12 antibody [
20‐
22].
Methods
Tissue microarrays (TMA) of TNBC patients
To analyse the frequency of ERß1 in TNBC samples, protein levels were determined using two independent TNBC TMAs. Cases that were included were all TNBC cases (stage 1–3) that were coming through the clinic and were pathologically ER negative, PR negative and HER2 negative. All cases were negative for ER-alpha (0% staining) except for 2 cases which were < 1% weak staining. Thus, all are considered negative under the historical guidelines (2000–2010; < 10% staining) and revised 2010 ASCO/CAP guidelines (< 1% staining). The first TMA was obtained from the Peter MacCallum Cancer Centre (PMCC) and contained 1 mm cores from 70 human primary TNBC tumors. Tissue samples were obtained from the PMCC, Royal Melbourne Hospital, St Vincent’s Hospital and Monash Health from women undergoing breast surgery between 2004 and 2011. The median age at diagnosis was 60 years and patients had a median follow-up of 72 months (range 0.2–137 months). The second TMA was obtained from Perth, Western Australia, containing 1 mm cores from 97 primary breast tumors. Tissue samples were retrieved from Sir Charles Gairdner Hospital (SCGH) from women undergoing breast surgery between 2005 and 2013. Of these, 9 cases did not have successful staining (no core after IHC, or not enough tumour in the core) and were removed from further analysis. We also removed another 2 samples as followup was too short or one core was a replicate of an exisiting core. The median age at diagnosis of the remaining 56 samples was 59 years with a median follow-up of 84 months (range 5–155 months). When both cohorts combined, the median age at diagnosis was 59 years and the median follow-up was 78 months (range 0.2–155 months). Table
1 shows the demographics and characteristics for the cohorts. Our cohorts precede the more widespread utilization of newer agents such as checkpoint immunotherapy and Sacituzumab for TNBC patients. Information on breast cancer recurrence and death came from medical records and the respective state cancer registries for Victoria and WA. For the combined cohort (
n = 156), 36% (55/156) had a recurrence. The percentage of patients that died was 35% (54/156) and this was mainly death due to BCa (42/54) rather than other causes (11/54) or unknown (1/54).
Table 1
Patient demographics and characteristics according to ERβ expression
Total Patients
|
Total | 81 | 75 | < 0.00001 |
PMCC | 51 (63) | 19 (25) |
WA | 30 (37) | 56 (75) |
Age
|
Median age | 62 | 56 | 0.0073 |
< = 50 | 14 (17) | 30 (40) |
> 50–70 | 44 (54) | 28 (37) |
> 70 | 23 (28) | 17 (23) |
Grade
|
1 | 0 (0) | 0 (0) | 0.659 |
2 | 4 (5) | 5 (7) |
3 | 77 (95) | 70 (93) |
Tumour Size
|
Median size | 25 | 25 | 0.613 |
T1 (1–19) | 30 (37) | 24 (32) |
T2 (20–49) | 45 (56) | 46 (61) |
T3 (50–99) | 5 (6) | 3 (4) |
T4 (100 +) | 0 (0) | 0 (0) |
unknown | 1 (1) | 2 (3) |
LN status
|
NO | 27 (33) | 27 (36) | 0.947 |
N1 + | 44 (64) | 43 (57) |
unknown | 10 (12) | 5 (7) |
Stage
|
1A | 22 (27) | 19 (25) | 0.931 |
IIA | 30 (37) | 28 (37) |
IIB | 12 (15) | 14 (19) |
III/IV | 13 (16) | 11 (15) |
Unknown | 4 (5) | 3 (4) |
Recurrence
|
No | 47 (58) | 54 (72) | 0.089 |
Yes | 34 (42) | 21 (28) |
Mortality
|
Yes (Death) | 29 (36) | 25 (33) | 0.834 |
No (Alive) | 52 (64) | 50 (67) |
Ethics approval for human samples
The PMCC (03/90, 00/81) and SCGH cohorts received ethics approval from their local ethical review boards to collect and share samples and clinical data. Patients had either given broad written consent to future research with their samples and data, or waivers of consent were in place. The research assessing estrogen receptor beta was approved by the Peter MacCallum Human ethics committee (10_16 and 21_76). The study was conducted in accordance with the Australian National Health and Medical Research statement on ethical conduct in human research. The study was performed in accordance with the Declaration of Helsinki.
Immunohistochemical staining and scoring
The level of ERß1 was analysed using the CWK-F12 ERß1 antibody (
Developmental Studies Hybridoma Bank, DSHB) using IHC. As discussed, validation using RIME showed this antibody to be ERß1 specific [
21] in addition to which we have previously validated its specificity using IHC [
22]. ERß1 IHC was performed on an automated IHC slide staining system, Ventana BenchMark Ultra (Roche Diagnostics, USA). Briefly, 3 µm thick FFPE sections mounted on coated slides (Series 2 Adhesive, Trajan Scientific Australia) were de-waxed and antigen retrieved in ULTRA Cell Conditioning Solution 2 (CC2, Roche Diagnostics) for 40 min at 97 °C. Following incubation in the OptiView Peroxidase Inhibitor (Roche Diagnostics, USA) for 5 min at room temperature, the sections were incubated in the ERß1 antibody, CWK-F12 (DSHB Hybridoma) at 0.14 µg/ml (1:320) for cell pellets or at 1.1 µg/ml (1:40) for tissue sections for 60 min at room temperature. On-board detection system, OptiView Universal DAB Detection Kit (Roche Diagnostics, USA), was used in a visualization step in accordance with the manufacturer’s instructions.
For exploratory analysis allowing comparison to previous work, we also assessed expression of ERß1 using the PPG5/10 antibody. The DAKO EnVision FLEX high pH kit was used with the ERß1 PPG5/10 (GeneTex) antibody diluted at 1:15. Scoring was performed by a breast pathologist (PA) and independently confirmed by a second scorer (KB). Both scorers were blinded to the clinical characteristics of the tumor samples. Cores were scored for the percentage of ERß1 positive tumor cells, as well as the intensity of staining to generate an Allred score that incorporates both aspects. Intensity was scored as negative = 0, weak = 1, moderate = 2 or strong = 3, and the percentage of positively stained tumor cells was classified as: 0% = 0; < 1% = 1, 1–10% = 2, 11–33% = 2, 34–66% = 4, 67–100% = 5. Scores were added to form a maximum score of 8. ERß1 positive was defined as those tumors with 40% or more ERβ1 + cells of any intensity. To determine the most appropriate cut-off of expression for our analysis with clinicopathological features, we assessed the distribution data from Allred scores and ERβ1 expression percentages (vs frequency) as recommended from past studies [
23]. A mixture model of two Gaussian distributions is fitted to the histogram of the expression using the
flexmix function in R. The optimal cutoff is determined as the value where the probability density functions of the mixing distribution coincide.
Statistical analysis
Survival analysis was performed in RStudio (v1.1.453, running R v4.0.3). Descriptive statistics were used to assess the proportion of TNBCs that were ERß1 positive and the association of ERß1 with BCa recurrence and survival. BCa recurrence was defined as locoregional and distant recurrence). Overall and breast cancer-specific survivals were assessed. Cox proportional hazards model was used to determine the association of variables with survival (survival::coxph), and univariate models were visualised in a Kaplan–Meier plot (unadjusted, survminer::ggsurvplot). Multivariate analyses included age, tumor grade and tumor size as continuous variables, axillary lymph node status chemotherapy and ERß1 status as categorical variables and cohort as a stratifying variable. Lymph node was assessed categorically as we did not have continuous data for all patients. survival::cox.zph was used to test the assumptions of the multivariate Cox proportional hazards test, which were visualised by ggcoxzph. Akaikie Information Criterion (AIC) was also used for step-wise model selection using MASS:stepAIC.
ERß expression in TNBC subtypes
Lehmann and colleagues compiled 587 TNBC gene expression profiles from 21 studies (training set = 386 and validation set = 201. They used k-means and consensus clustering of the tumor profiles to reveal that TNBC is composed of six stable subtypes. These were Basal-like 1 (BL1), basal-like (BL2), immunomodulatory (IM), mesenchymal (M), mesenchymal stem–like (MSL), and luminal androgen receptor (LAR) subtype [
24]. They later refined their classification to 4 consistent classes (TNBCtype-4)-BL1, BL2, M, and LAR [
25]. We accessed the same GSE files (apart from GSE-28821, GSE-28796, GSE-22513 and GSE-18864 which were not detailed in their supplemental data) to assess the relationship of the different TNBC subtypes with ERß1 (gene symbol ESR2). Raw data (Affymetrix CEL files) were downloaded from public data repositories (GSE7390, GSE2603, MDA133, GSE3494_hgu133a, GSE2990, GSE2034, GSE11121, GSE1561, GSE7904, GSE1456_hgu133a, GSE5847, GSE20194, GSE19615, GSE5327, GSE16446, GSE12276) and files read into R and normalized using robust multi-array normalisation with the affy R package (version 1.64.0) [
24] and then log-normalized. For each array dataset, probes were matched to gene symbols using the AnnotationDbi R package (version 1.48.0) and expression values were collapsed to gene-level by taking the probe for each gene with the highest interquartile range of log-expression. All arrays were then quantile normalised together using the normalize.quantiles function from the preprocessCore R package (version 1.48.0) and batch effects were removed using the removeBatchEffect function from the limma R package (version 3.42.2) [
26]. TNBC samples were then identified based on the 2-component Gaussian mixture distribution model of Lehmann and colleagues [
25].
P values and confidence intervals for the differences of the means of each gene for each pairwise comparison between the TNBC subtypes were calculated using Tukey’s range test.
Expression of ERß and downstream targets TNBC
The TCGA Breast Invasive Carcinoma study data was utilised, specifically the 1084 samples that have been contributed to the PanCancer Atlas study. The clinical data was downloaded from cBioPortal on 8thNovember 2022. The RNAseq data was downloaded from the ICGC Data Portal on the 17th November 2022. The RNAseq RSEM raw count data was filtered for lowly expressed genes and TMM normalized to generate log CPM data using the edgeR package. The basal subtype samples (n = 173) were evaluated for ESR2 and downstream gene expression. As ESR2 RNA expression levels varied across the cohort, each sample was classified based on their ESR2 expression level as high (top quartile), moderate (middle two quartiles) and low (bottom quartile). Pearson correlations were performed on the high ESR2 expressing group between ESR2 and its downstream genes expression level.
Discussion
In our series of TNBCs nuclear ERβ1was expressed in 72% of cases. This is the first study to assess the expression of ERβ1 in TNBC since the robust validation of the existing CWK-F12 ERβ1 antibody by western blot and RIME analyses [
21] as well as our IHC validation [
22]. The high proportion of ERβ expression in TNBC in our work is similar to Yan and colleagues who showed that approximately 75% of TNBC were ERβ1 positive, but used the PPG5/10 antibody which has been shown to be non-specific via western blot and RIME analysis. Gruvberger-Saal and colleagues [
13] used a cocktail of ERβ antibodies (PP65-10 and 14C8), the former non-specific and the latter specific by western blot, and found a high percentage of ERα negative tumors to be ERβ1 positive (~ 55%). Three other studies have assessed the levels of ERβ1 protein and found 25% of ERα-negative [
17], 35.5% of TNBC [
16] and 83% of TNBC [
14] to be ERβ1 positive. Two of these studies used the PPG5-10 antibody [
14,
17]. The polyclonal rabbit antibody used by Guo and colleagues (#BY-02101; Shanghai Yueyan Biological Technology, Co., Ltd., Shanghai, China) has not been assessed for specificity [
16]. Regardless, our results shows that a high proportion of TNBC express nuclear ERβ1.
We have demonstrated that the expression of ERβ1 using the CWK-F12 ERß1 antibody was not associated outcome in TNBC. Guo and colleagues found that survival in TNBC cancer patients (
n = 107) was inferior in those with ERβ expression (χ
2 = 5.330,
p = 0.021) [
16]. We found a trend for inferior recurrence, but no effect on survival. The discrepancy in the data may be due to the differences in the numbers of patients assessed, and potentially the clinical treatment of the patients. We did not find an effect on overall survival. Our results demonstrate that ERβ1 is not prognostic for recurrence or survival. To assess if downstream ERβ1 specific transcriptional programs were present in
ESR2 + TNBC cancers we performed Pearson correlation analysis between the high
ESR2 expressing group in TNBC and downstream genes known to be upregulated or downregulated following
ESR2 activation [
29].
PROM6,
FN1 and
SLC16A6 were significantly positively correlated with
ES2R.
ANKRD35,
ASB9 and
SELENBP1 were negatively correlated. Whilst this in encouraging and indicates that ERβ1 is driving downstream estrogen actions in TNBC, further work is needed to define what this means functionally for the breast cancer cells.
To allow some comparison of our data with the previously published results using non-validated antibodies, we also stained our cohort with the PPG5-10 antibody and assessed the relationship between ERβ1 expression and prognosis. High ERβ1 (> 40%) expression (compared to low) was not associated with recurrence but an Allred score of > 5 was associated with less recurrence compared to an Allred of 1–5. High ERβ1 (> 40%) expression (compared to low) was associated with better overall survival, as was an Allred of > 5 compared to 1–5. This supports previous reports with this antibody showing that in ERα negative tumors, ERβ1 expression is associated with good prognosis [
13,
14]. However, we and others believe this antibody is non-specific and thus do not draw conclusions from its expression and recurrence or survival when tested on the same patient cohort. If further work on this antibody does indicate it is specific, we acknowledge that our work may indicate that full length ERβ, but not splice variants associate with BCa prognosis. At present however, our work cautions against the interpretation of previous data which has used a non-specific ER antibody for staining.
Currently only those women with ERα positive tumors are treated with endocrine therapies such as tamoxifen. TNBC patients lack defined drug targets, and so would benefit greatly from the identification of new targeted therapeutics. There is some indication that ERβ expression may act as a biomarker of tamoxifen sensitivity. In a small cohort of TNBC patients (
n = 50) who were treated with tamoxifen for two or more years, Honma and colleagues [
14] reported that those whose tumors expressed ERβ had significantly longer survival. However, this study used the non-specific PPG5-10 antibody, and did not include a control cohort without tamoxifen treatment. Similarly, Gruvberger-Saal and colleagues found that expression of ERβ was associated with increased survival (distant disease-free and overall survival) in tamoxifen-treated ERα-negative patients but not in the ERα-positive subgroup [
13], but again did not evaluate an untreated cohort. In a study assessing ERβ1 expression in tissue microarrays from a randomized, placebo-controlled trial of tamoxifen therapy (NCIC-CTG-MA12), high ERβ1 expression in ERα negative patients was associated with longer recurrence free survival in tamoxifen-treated patients compared to placebo [
15]. This study used a polyclonal, GC17/385P, Biogenex ERβ antibody that was shown previously to be specific [
30], but was not tested in the recent antibody validation studies. Interestingly, this study demonstrated that in ERα-negative patients, ERβ1-high tumors were associated with a worse outcome (52% 5-year recurrence free survival), which could be improved to a level of survival (77% 5 year survival) very similar to patients with ERβ1-low tumors (75–76%) by tamoxifen treatment.
Our study represents the first essential step towards determining whether ERβ1 expression should be routinely assessed in TNBC in general as a prognostic factor. We find that the often-used PPG5/10 antibody did show associations with recurrence and survival, however these trends were not observed when we used a validated ERβ1 antibody. This highlights the need to re-assess the relationship between ERβ1 expression and tamoxifen sensitivity in TNBC patients that have been treated with Tamoxifen [
13‐
15]. This would then allow the field to determine if ERβ1 has any potential therapeutic target in TNBC as indicated in fulvestrant treated ERβ1 positive TNBC [
31].
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit
http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (
http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.