Cancer Letters 321 (2012) 73–79

Contents lists available at SciVerse ScienceDirect

Cancer Letters

journal homepage: www.elsevier .com/locate /canlet

Different associations of estrogen receptor b isoforms, ERb1 and ERb2,expression levels with tumor size and survival in early- and late-onset breast cancer

Vesna Mandusic a,b, Bogomir Dimitrijevic b, Dragica Nikolic-Vukosavljevic c, Zora Neskovic-Konstantinovic c,Ksenija Kanjer c, Ute Hamann a,⇑a Deutsches Krebsforschungszentrum (DKFZ), Molecular Genetics of Breast Cancer, Im Neuenheimer Feld 580, 69120 Heidelberg, Germanyb University of Belgrade, Institute of Nuclear Sciences ‘‘Vinca’’, Department for Radiobiology and Molecular Genetics, Mike Alasa 12-14, 11000 Belgrade, Serbiac Institute for Oncology and Radiology of Serbia, Department of Experimental Oncology, Pasterova 14, 11000 Belgrade, Serbia

a r t i c l e i n f o a b s t r a c t

Article history:Received 8 December 2011Received in revised form 17 February 2012Accepted 17 February 2012

Keywords:Estrogen receptor betaEstrogen receptor alphaClinical outcomeEarly- and late-onset breast cancersQuantitative RT-PCR

0304-3835/$ - see front matter � 2012 Elsevier Irelandoi:10.1016/j.canlet.2012.02.022

⇑ Corresponding author. Tel.: +49 6221 42 2344; faE-mail address: u.hamann@dkfz-heidelberg.de (U.

Background: In breast cancer, little is known about the consequences of co-expression of ERa with thesecond estrogen receptor, ERb, and its isoforms in light of their joint prognostic value. Previously reportedcorrelations have been based mostly on independent ERa and ERb expression levels in breast tumors.Purpose: To address whether the expression ratio of ERa and ERb and its isoforms may be a more impor-tant parameter than their absolute levels, we analyzed relative mRNA expression ratios of ERb1 to ERb2and ERa in 74 clinical samples of invasive breast cancer including 39 early-onset and 35 late-onset breastcancers. Expression levels were correlated with clinical and histopathological parameters and disease-free interval.Results: A specific correlation of ERb1 expression levels with tumor size was detected in early-onsetbreast cancer patients and of ERb2 levels with tumor size in late-onset patients. Expression of bothERb isoforms inversely correlated with expression of the two estrogen regulated genes, progesteronereceptor and pS2 in both groups. Higher levels of ERb2 than ERb1 isoform were associated with a betteroutcome in late-onset patients.Conclusions: Our results suggest that different isoforms of ERb may be involved in suppression of tumorgrowth in young and elder patients and may have different prognostic values.

� 2012 Elsevier Ireland Ltd. All rights reserved.

1. Introduction transcription via AP1-sites [9]. As homo- or hetero-dimers (when

Due to the endocrine nature of breast cancer, hormone recep-tors and members of their signaling pathways are still among themost promising targets for adjuvant treatment. Estrogen and pro-gesterone receptor (ER and PR) status is a prognostic and predictivemarker for endocrine therapy routinely used in the clinical man-agement of breast cancer.

Estrogens contribute significantly to the development ofhormone-dependent breast carcinomas [1]. The biological effectsof estrogens are mediated through the two estrogen receptors,ERa and ERb, in estrogen responsive normal tissues and in humanbreast cancers [2–4]. Upon binding of estrogens to their receptors,they form homodimers that activate the transcription of targetgenes bearing a classical estrogen response element (ERE) in theirpromoter region or by tethering to other transcription factors, suchas AP1, SP1 or NF-jB [5–7]. In addition to the activation by ERa andERb1 homodimers [8], ERb1, which represents the wild type (wt)isoform of ERb, forms heterodimers with ERa that activate gene

d Ltd. All rights reserved.

x: +49 6221 42 4721.Hamann).

these two receptors are co-expressed in the same cell), they acti-vate expression of unique sets of estrogen responsive genes[10,11]. According to present data, the ERa/ERb ratio changes dur-ing breast carcinogenesis, with ERa being up regulated and ERbbeing down-regulated [12,13]. In addition to the wt ERs, variousisoforms and splice variants exist, especially for ERb and some ofthese have been shown to modulate ERa activity. Isoforms ERb2–ERb5 resulting from alternative usage of exon 8 can inhibit bothERb1 and ERa transactivating activity [12,14,15]. Isoform ERb2(which represents the C-terminal-alternatively spliced isoform,previously designated as ERbcx) does not bind estrogens, exhibitslost or reduced DNA binding ability, but has a dominant-negativeeffect on ERa transcriptional activity on ERE-driven promoters[14–16].

Despite a large amount of data on ERb expression levels inbreast tumors exists, their role in disease progression, responseto hormonal therapy and value in predicting disease outcome isstill unclear. At the protein level, preserved ERb1 expression hasbeen shown to be associated with a favorable prognosis, as itwas associated with tumor parameters indicative of low biologicalaggressiveness including negative lymph node status, low grade,

Table 1Clinical and histopathological characteristics of the breast cancer patients.

Parameter Number of patients (%)

Age at diagnosis (years)640 39 (53)P60 35 (47)

Menopausal statusPremenopausal 39 (53)Postmenopausal 35 (47)

Histological typeInvasive ductal 42 (57)Invasive lobular 24 (32)Other 8 (11)

Tumor size (pT categories)pT1 36 (47)pT2, 3 38 (53)

Number of nodes involved0 33 (45)P1 38 (51)Unknown 3 (4)

ER status (IHC)Negative (IRS 6 2) 11 (15)Positive (IRS P 3) 49 (66)Unknown 14 (19)

PR status (IHC)Negative (IRS 6 2) 7 (9)Positive (IRS P 3) 53 (72)Unknown 14 (19)

TreatmentEarly-onset patientsChemo + TAM 15 (38)Chemo + TAM + LHRHa 12 (31)TAM alone 3 (8)TAM + LHRH 1 (2)Chemo alone 8 (20)

Late-onset patientsChemo + TAM 10 (26)TAM alone 20 (57)Chemo alone 3 (9)No adjuvant therapy 2 (6)

Chemo: Chemotherapy, TAM: Tamoxifen, LHRHa: Luteinizing Hormone-Releasing Hormone analogs.

74 V. Mandusic et al. / Cancer Letters 321 (2012) 73–79

and better course of disease [10,17]. In earlier reports it was sug-gested that co-expression of ERa and ERb may cause resistanceto tamoxifen [18]. At present, however, controversial data exist,which of the two isoforms, ERb1 or ERb2, has an independent pre-dictive value [19,20]. In some studies using expression data ob-tained on the protein level, mainly by immunohistochemistry,ERb1 was of predictive value for tamoxifen response and of inde-pendent prognostic value [17,19,21]. In a recent study, Vinayagamand coworkers [22] reported that the ERb2 mRNA level acts as anindependent predictive marker of a favorable clinical outcome intamoxifen treated patients. It is conceivable that the ratio betweenERa and ERb (and possibly ERa to ERb1 or ERa to ERb2) may be amore important parameter than their absolute levels and that achange in the ‘‘normal equilibrium’’ between ERa and ERb (includ-ing their isoforms) is one of the factors promoting carcinogenesis[13,23]. Taking into account that these receptors form homo- andhetero-dimers, which regulate different sets of downstream targetgenes, their changed equilibrium in malignant cells may lead todifferent expression profiles.

Although the inhibitory effects of ERb isoforms (designated asERb1–5) on ERa transcriptional activity were demonstrated in pre-vious in vitro studies [15,16,24], data about their values as inde-pendent predictors of disease and response to treatment are stillcontroversial. Two retrospective studies showed no associationsof ERb2 expression with clinical outcome among women withinvasive breast cancer and response to treatment [19,25]. In immu-nohistochemistry (IHC)-based studies, ERb1 was of predictive va-lue for tamoxifen response and of independent prognostic value[17,19,21]. On the contrary, other studies found only ERb2 variantto be of predictive value [20,22]. Vinayagam and coworkers [22]reported that ERb2 mRNA expression was independently associ-ated with disease outcome in a cohort of 100 breast cancer pa-tients. Shaaban and coworkers reported an association of nuclearERb2 with overall and DFI [20]. These controversial findings maybe explained by differences in the populations studied, inclusioncriteria, antibodies and cut-off values used for IHC scoring as wellas evaluation criteria. Another cause for this discrepancy may bethe fact that ERb was usually analyzed as an independent marker.

In the present study we determined ERb1 and ERb2 mRNAexpression levels relative to ERa and ERb1 relative to ERb2 in 74clinical samples of invasive breast cancer including 39 early-onsetbreast cancers (diagnosed in women 640 years) and 35 late-onsetbreast cancers (diagnosed P60 years) by real time quantitative RT-PCR. Expression levels were correlated with clinical data.

2. Material and methods

2.1. Patients and tissue specimens

Seventy-four invasive breast tumors were obtained from female patients, whounderwent surgery at the Institute of Oncology and Radiology, Belgrade, Serbia,from January 2005 to December 2007. Thirty-nine patients were diagnosed withbreast cancer under or equal the age of 40 years (early-onset group) and 35 wereequal or older than 60 years (late-onset group). Clinical, histopathological dataincluding histological type, tumor size, lymph node status, tumor grade, ER andPR status and treatment data were collected from medical records and pathologyreports. Clinical follow up for up to 70 months was available for 52 patients. Theclinical, histopathological and treatment data are given in Table 1.

Tumor tissues for RNA isolation were snap frozen in liquid nitrogen and storedat -70 �C. The percentage of tumor cells was estimated by visual inspection ofhematoxylin-eosin stained sections by a pathologist and only regions with at least70% of malignant cells were used for RNA isolation with the majority of samples(69/74) having a tumor cell content of 80–90%. We also included ER-negative breasttumors, because some basal ERa expression below the IHC detection threshold mayexist.

The research protocol was approved by the Ethics Committee of the OncologyInstitute, Belgrade, Serbia and the Ethics Commission of the University of Heidel-berg. Germany. All participants have given informed written consent prior to enroll-ing in the study.

2.2. RNA extraction

Total RNA was extracted from breast tumor tissues by acid–phenol extraction aspreviously described [26]. To avoid genomic DNA contamination, total RNA wastreated with RNase-free DNase using Ambion DNA-free Kit (Ambion, Austin,USA). Following DNase treatment, RNA was precipitated and dissolved in water.The RNA concentration was measured on a ND-1000 spectrophotometer. The RNAintegrity number (RIN) was determined on Agilent 2100. Only high quality sampleswith a RIN greater than eight were used for cDNA synthesis.

2.3. cDNA preparation and quantitative (real-time) RT-PCR

Two micrograms of total RNA were reverse transcribed using Omniscript RT Kit(Qiagen, Hilden, Germany) according to the manufacturer’s instructions. Reversetranscription reactions were performed in the presence of 10 lM random hexamerand 1 lM oligo-dT primers in a final volume of 40 ll. To eliminate ‘‘batch to batch’’variation, a reference RNA sample was included in each RT run (Stratagene QPCRHuman Reference Total RNA, La Jolla, USA). This sample was used for the final cal-ibration of samples from different RT batches in the runs that quantified ERa by therelative standard curve method.

All PCR reactions were performed using TaqMan Pre-designed Gene ExpressionAssays with exon/boundary spanning probes specific for the target sequences in a7900 Sequence Detection System (Applied Biosystems, Foster City, CA). Assaydetails are provided in Supplementary Table 1. PCR reactions were carried out ina final volume of 12 ll containing 1x TaqMan Universal PCR Master Mix (AppliedBiosystems), 1x TaqMan Pre-Designed Gene Expression Assay and cDNA dilutedwith water (1:13).

V. Mandusic et al. / Cancer Letters 321 (2012) 73–79 75

The relative standard-curve method was used to measure expression levels ofpS2 and PR-(A + B), ERb1, ERb2 and ERa as independent variable. Standard curveswere generated by serial dilutions of the sample, which had the highest level ofexpression as determined in a previous pilot experiment. Dilutions were preparedwith RNA isolated from a heterologous insect cell line and reverse transcribed un-der the same conditions as used for reverse transcription of RNA from tumor tis-sues. Results were expressed as N-fold differences in target gene expressionrelative to calibrator (1x sample, Stratagene Human Reference RNA) and normal-ized to 1 lg of total RNA.

In addition, we analyzed quantities of ERb1 and ERb2 mRNA as a relative levelnormalized (N-fold difference) to ERa and calibrated to a reference sample (Strata-gene reference RNA). Values obtained in this way represent a relative portion ofERb1 or ERb2 compared with the transcript pool of ERa in each sample. By usingthe Ct data from the same runs, ERb1 was also expressed relative to ERb2 (ERb1 nor-malized to ERb2) and calibrated to reference RNA. Further, we assessed whetherthese values (expression ratio of ERs) may be associated with clinical and histopa-thological parameters, disease-free interval (DFI) and expression levels of twoestrogen-responsive genes (progesterone receptor (PR) and pS2).

For correlations with PR-(A + B) and pS2 and histopathological parameters, rel-ative levels of ERb1 and ERb2 were determined by the DCt method and expressedas N-fold differences in ERb1 or ERb2 gene expression relative to ERa and calibratedto 1x sample (Stratagene Human Reference RNA). The amplification efficiency of theassays was checked by preparing serial dilutions of the referent sample. The slopes ofthe curves obtained after amplification were calculated; they never differed by morethan 0.1. Real time data were analyzed with ABI software. Finally, the obtained valuefor expression level of ERb1 and ERb2 represents the value normalized to total ERatranscripts and calibrated to 1x sample (Stratagene Human Reference RNA).

All PCR reactions were run in triplicates, and outliers were eliminated from finalanalysis. Final values are means from triplicates or at least duplicates.

2.4. Statistical analysis

Statistical analysis was performed using the SigmaPlot version 10.00 software(Systat Software, Erkrath, Germany). Correlations between relative expression lev-els of ERb1 and ERb2 and pS2 and PR-(A + B) were evaluated by Spearman test. Cor-relations between expression levels and clinical and histopathological parameterswere tested by Mann Whitney test. Two-sided P values of 0.05 or less were consid-ered as statistically significant.

Analysis of DFI was performed using the Log-Rank Analysis and Holm-SidakPost-hoc Test. DFI was defined as the time from primary surgical treatment ofbreast cancer until diagnosis of either distant metastases or local recurrences. The50% median value (0.355) was used as an arbitrary cut-off for ERb1 normalized toERb2 and calibrated to the reference sample. This cut-off value represents a samplewith an about fourfold higher ERb2 than ERb1 expression (based on the differencein uncalibrated Ct values). By this way patients were stratified in a ‘‘lower ERb2’’group (less than fourfold difference in favor of ERb2) and ‘‘Higher ERb2’’ group(more than fourfold difference in favor of ERb2).

Fig. 1. Boxplots of mRNA expression values for ERa, ERb1/ERa and ERb2/ERa bytumor size in early- and late-onset breast tumors. The box represents theinterquartile range; the line across the box indicates the median and the whiskersextend from the box to the highest and lowest values (excluding outliers andextreme points). Black dots represent outliers and extreme points. (a) The ERamRNA levels were determined using the relative standard curve method. Valueswere expressed in arbitrary units, normalized to 1 lg of total RNA and calibrated to1x sample (reference cDNA). (b) ERb1 mRNA levels were expressed in arbitraryunits relative to ERa (N-fold difference to ERa) and calibrated to 1x sample(reference cDNA). (c) ERb2 mRNA levels were expressed in arbitrary units relativeto ERa (N-fold difference to ERa) and calibrated to 1x sample (reference cDNA).

3. Results

3.1. Associations of ERa and relative ERb1 and ERb2 mRNA levels withclinical and histopathological parameters

Table 1 shows the available tumor characteristics. With theexception of tumor size, none of the histopathological parametersshowed an association with ERa and relative ERb1 and ERb2 mRNAlevels. Correlations with histological tumor grade could not be as-sessed because the majority of tumors (n = 69) were of grade G2.

In late-onset breast tumors, ERa mRNA expression was signifi-cantly higher in T2 + 3 tumors (n = 18) compared to T1 tumors(n = 17) (p = 0.01, Mann–Whitney). A similar trend was detectedin early-onset tumors, which however, was not statistically signif-icant (T1, n = 19; T2 + 3, n = 20; p = 0.09, Mann–Whitney, Fig. 1a).

Levels of ERb1 relative to ERa mRNA expression was signifi-cantly lower in T2 + 3 early-onset tumors (n = 20) compared toT1 tumors (n = 19) (p = 0.01, Mann–Whitney). A similar trend(but not significant) was observed in late-onset tumors (T1,n = 19; T2 + 3, n = 20; p = 0.054, Mann–Whitney, Fig. 1b).

Levels of ERb2 relative to ERa mRNA expression was lower inT2 + 3 late-onset tumors (n = 18) compared to T1 tumors (n = 17)(p = 0.02, Mann–Whitney). A similar trend (but not significant)was observed in early-onset tumors (T1, n = 19; T2 + 3), n = 20;p = 0.056, Mann–Whitney, Fig. 1c).

3.2. Distribution of differences in relative ERb1 and ERb2 mRNA levels

The uncalibrated DCt values are shown in Fig. 2. They representdifferences in the threshold cycle between: (a) ERb1 and ERa and(b) ERb2 and ERa, and indicate 2DCt times higher ERa mRNAexpression levels than ERb levels.

Although in the majority of samples ERa mRNA levels overcameERb1 and ERb2 levels for three orders of magnitude (the 75th per-centile DCt was 12.2 for ERb1 and 10.6 for ERb2), in some samples(the 25th percentile DCt values were 8.7 for ERb1 and 6 for ERb2),ERa overcomes ERb1 and ERb2 mRNAs levels for approximatelytwo orders of magnitude.

Fig. 3. Kaplan–Meier curves of time to relapse in late-onset patients. ‘‘Higher ERb2’’refers to a more than fourfold difference in favor of ERb2 (circles), and ‘‘lower ERb2’’refers to a less than fourfold difference in favor of ERb2 (triangles). Log-RankAnalysis of DFI showed a significant difference in time to appearance of recurrencesbetween the two groups (p = 0.021).

76 V. Mandusic et al. / Cancer Letters 321 (2012) 73–79

Eighteen tumors had a DCt value for ERb1/ERa below 8.7 (tu-mors with higher ERb1 to ERa ratio), of which all except one wereT1 tumors. Twenty-four tumors had a DCt value for ERb2/ERa be-low 7 (tumors with higher ERb2 to ERa ratio), of which all exceptfour were T1 tumors.

3.3. Associations of ERb1 relative to ERb2 mRNA levels with DFI

Clinical follow-up for up to 70 months was available for 25patients from the early-onset patient group, and for 27 from thelate-onset group. Most of these women have received adjuvanttamoxifen (22/25 from the early-onset group; 26/27 from thelate-onset group). Log-Rank Analysis of DFI showed a significantdifference in time to relapse depending on the ERb1 to ERb2 ratioin the late-onset patient group (Fig. 3). In the ‘‘Higher ERb2’’ groupthe median DFI was 65.6 months and in the ‘‘Lower ERb2’’ group53.6 months (p = 0.021). No statistically significant difference inDFI was detected in the early-onset patient group, although theresults were in the same direction.

When determined according to standard curve method neitherERb1 nor ERb2 were associated with clinical outcome. No associa-tions of ERb1 and ERb2 to ERa with DFI were found.

3.4. Associations of relative ERb1 mRNA levels with PR-(A + B)and pS2 levels

Relative expression level of ERb1 expressed as N-fold differenceto ERa mRNA was analyzed for its association with PR-(A + B) andpS2 mRNA levels in all breast tumors and in the subgroups of early-and late-onset breast tumors. Relative ERb1 mRNA levels showed a

Fig. 2. Distribution pattern of DCt values for ERb1 and ERb2 relative to ERa. DCtvalues of (a) ERb1 relative to ERa and (b) ERb2 relative to ERa in all breast tumors.

significant inverse correlation with PR-(A + B) mRNA levels in allbreast tumors (n = 72, q = �0.500, p < 0.0005, Spearman, Supple-mentary Fig. 1a). When early- and late onset breast tumors wereanalyzed separately, the inverse correlation with PR-(A + B) mRNAwas detected in both subgroups (early-onset breast tumors: n = 38,q = �0.479, p = 0.002; late-onset breast tumors: n = 34, q = �0.520,p = 0.002; data not shown).

A similar association was found between relative ERb1 and pS2mRNA levels in all breast tumors (n = 71, q = �0.345, p = 0.003,Spearman, Supplementary Fig. 1b) and in the two subgroups(early-onset breast tumors: n = 38, q = �0.442, p = 0.01; late-onsetbreast tumors: n = 34, q = �0.356, p = 0.04; data not shown).

3.5. Associations of relative ERb2 mRNA levels with PR-(A + B)and pS2 levels

Relative expression level of ERb2 mRNA expressed as N-fold dif-ference to ERa mRNA also inversely correlated with PR-(A + B)(n = 73, q = �0.649, p < 0.0005, Spearman, Supplementary Fig. 2a)and pS2 mRNA levels (n = 73, q = �0.451, p < 0.0005, Spearman,Supplementary Fig. 2b) in all breast tumors. Similar correlationswith PR-(A + B) and pS2 mRNA levels were observed in the early-onset group (n = 38, q = �0.648, p < 0.0005, Spearman and n = 38,q = �0.548, p < 0.0005, Spearman, respectively) and in the late-on-set group (n = 35, q = �0.667, p < 0.0005, Spearman and n = 35,q = �0.467, p = 0.005, Spearman, respectively; data not shown).

When ERb2 mRNA expression was analyzed for its associationwith PR immunohistological expression, a similar trend for a weekinverse correlation was observed, which however, did not reachstatistical significance (n = 60, q = �0.234, p = 0.07, Spearman; datanot shown).

4. Discussion

Given the fact that ERb1 and ERb2 can form heterodimers withERa and ERb1 with ERb2, their relative quantities in the totalreceptor pool may be a more important predictor of disease thantheir absolute values. This is in line with a recent report by Bulunand coworkers, who also suggested that the receptor expression

V. Mandusic et al. / Cancer Letters 321 (2012) 73–79 77

ratios may be a more important determinant in ER-mediated sig-naling pathways than their independent values [27]. While previ-ous studies have measured independent ERa and ERb expressionlevels, we determined, for the first time, the mRNA expression ra-tios of ERb1 and ERb2 relative to ERa and ERb1 relative to ERb2 in74 clinical samples of breast cancer including 34 late-onset and 39early-onset cases. Until now early-onset breast cancers have rarelybeen in the focus of clinical studies, which probably is due theirrelative low frequency, and consequently their specific featuresmay have been diluted in the group of the more frequent late-onsetpostmenopausal cases.

We found different associations of relative ERb1 and ERb2mRNA levels normalized to ERa with tumor size in early- andlate-onset breast tumors. An inverse correlation of tumor size withrelative ERb1 mRNA levels was observed in early-onset breast tu-mors and with ERb2 mRNA levels in late-onset tumors. Interest-ingly, neither absolute ERb1, nor ERb2 mRNA levels correlatedwith tumor size in the two subgroups. This suggests that a de-creased amount of these two isoforms relative to ERa may be moreinformative in the context of downstream regulation of targetgenes than their absolute levels (usually determined by IHC stain-ing). Our findings indicate that ERb1 and ERb2 may be differentlyinvolved in breast cancer progression in young and old patients.This is, at least partially, the consequence of loss of transcriptionalrepression of ERa by ERb1 in early-onset breast cancer and by ERb2in late-onset breast cancer, which may lead to an increased estro-gen-stimulated tumor growth.

Omoto and coworkers [28] showed by microarray profiling thatdifferent sets of genes are down-regulated in ERb1- and ERb2-transformed MCF7 cells compared to parental cells. Microarrayprofiling of clinical samples of breast cancer revealed that differentpathways are activated in early- and late-onset disease [29]. Theresults from these two studies imply a potential role of the ERb iso-forms in ‘‘directing’’ ERa:ERb1 and ERa:ERb2 heterodimers to dif-ferent promoter sets or in recruiting different coregulators, someof which may be specifically connected/activated with age [30].These data may at least partially explain the different associationsof tumor size with ERb1 levels in the early-onset and ERb2 levels inthe late-onset patient group.

It is conceivable that physiological differences between youngand old patients, first, estrogen levels may be responsible for thedifferent roles of the two isoforms in tumor growth regulation.Since only ERb1 can bind estrogens, it is possible that under phys-iological conditions of high estrogen levels in young patients, thedecreased portion of this isoform may contribute to tumor growth.In contrast, in older patients with established low estrogen levels,the ERb2 isoform (which does not bind estrogens) may be the ma-jor repressor of ER activity. Thus, the observed effects of ERb1 andERb2 (although measured as mRNA) on tumor growth in youngand elder age may be connected with hormonal status.

While it was previously shown that both, ERb1 and ERb2 inhibitor at least attenuate the activity of ERa on ERE driven promoters[14–16,31], the mechanism by which these two isoforms affectERa is different [28,32]. Omoto and coworkers showed that ERb1inhibits ERa activity mainly by competing for binding sites by for-mation of ERb1:ERb1 homo- or ERa:ERb1 hetero-dimers in stablytransfected MCF7 cells [28]. Similarly, ERb2 repression of ERaactivity was reported to occur via formation of inefficientERa:ERb2 heterodimers [16,28] and by decreasing available ERaprotein through activation of its proteasome degradation [15].We also found an inverse correlation of ERb2 mRNA with ERa pro-tein levels determined by IHC in the analyzed tumors (data notshown), which is in line with the latter mechanism.

In both early- and late-onset breast tumor samples in this study,decreased ERb1 and ERb2 mRNA levels correlated with increasedlevels of PR and pS2 mRNA. These findings indicate that modula-

tory effects of ERb1 and ERb2 on promoters other than PR andpS2 and regulated by ERa may be responsible for the differentassociations with tumor size in early- and late-onset breast cancerpatients. In line with our results, Saji and coworkers found an in-verse association of ERb2 and PR protein expression in preopera-tive needle biopsies of breast tumors [33]. In another study, Parkand coworkers reported a decreased ERb2 mRNA expression inbreast tumor tissues compared to matched normal tissue in pre-and post-menopausal patients [12]. These results are not contra-dictory to our findings of similar absolute median ERb1 and ERb2mRNA levels in T1 and T2 + 3 breast tumors in both age groups(data not shown), with only their levels expressed relative to ERavarying in small and large tumors. It is possible that after an initialdecrease of ERb1 and ERb2 in early stages of carcinogenesis, thenext or parallel event is a continuous increase of ERa. This increasemay be a major cause of the elevated ERa:ERb ratio during furthertumor progression. This is in agreement with data showing thatERa protein levels (fmol/mg protein) increase with age in breastcancer patients [30,34,35]. In our study we have found a signifi-cantly higher median ERa mRNA expression in T2 + 3 tumors com-pared to T1 tumors in late-onset patients and a similar trend inearly-onset patients. Therefore, the increased ERa:ERb ratio duringtumor progression appears to be the consequence of elevated ERa,rather than decreased ERb. During tumor growth, different mecha-nisms leading to ERa overexpression may be operative, like (i)demethylation of its promoter [36], (ii) presence of cis acting pro-moter elements [37,38] and/or (iii) gene amplification [39]. Gener-ally, the ERa:ERb ratio seems to be the most critical factor in theability of ERb to modulate ERa during tumor growth.

In our cohort, different Ct values were detected for ERa andERb1 ranging from 2.9 to 12.6, which indicate about 10- to over6000-fold higher ERa than ERb1 mRNA levels. Interestingly, theERa:ERb2 ratio covered an even larger range but with overall levelsof ERb2 being higher than those of ERb1 (data not shown). In sixcases, ERb2 mRNA levels almost reached ERa levels and in threecases, ERb2 levels were even higher than ERa levels.

The involvement of ERb1 and ERb2 in the regulation of differentestrogen regulated genes, like pS2, TGF-a and p21 has been dem-onstrated using in vitro models [40]. In cell culture experiments,the approximate levels of transfected isoforms have been deter-mined, and for most cases it was shown that the inhibitory effectof ERb isoforms on estrogen-signaling is dose-dependent. Similareffects should be observed in clinical samples, but so far, the ratioof ERb1 and its isoforms compared to ERa was not investigated indepth. Although it is known that levels of ERb including its iso-forms are significantly lower in homogenates of breast cancer tis-sue [41] than ERa levels, it is important to note that the expressionratio of ERa and ERb receptors can vary between individuals.

Our findings may have some important implications. Firstly,they suggest that higher expression of ERb isoforms may lead tothe generation of ER-positive/PR-negative breast cancer, in addi-tion to the several other proposed mechanisms (reviewed in[42]). This is in line with data from our previous study [43], inwhich an inverse correlation of mRNA expression levels of anotherERb isoform, splice variant ERbD5 (also considered as an inhibitoryvariant) with PR protein levels was reported. It may be possiblethat the ERa:ERb1 (or ERb2) ratio may be useful in predicting re-sponse to tamoxifen, since it was proposed that the co-expressionof ERa and ERb1 is related to tumor characteristics and indicatespoor prognosis and resistance to tamoxifen [17,18,44]. Ifthis would be true, a dual role of ERb1 and ERb2 isoforms inERa-positive breast cancer cells may be substantiated: They bothattenuate estrogen-stimulated tumor growth, but simultaneouslymay cause tamoxifen resistance. In particular, this may be the casefor ERb1, which acts as a tamoxifen agonist according to in vitrodata [45]. In future studies, the ratio of ERa and ERb proteins

78 V. Mandusic et al. / Cancer Letters 321 (2012) 73–79

should be systematically tested for its transcriptional activity on awide range of promoters in cell culture experiments and in patientcohorts. Currently, the main limitation for further exploration isthe absence of a fully quantitative protein assay for ERb1 andERb2. In parallel, reducing the large linear range of ERa and ERblevels (which actually represent the continuous variable) to ± cat-egories or to few IHC scoring units may introduce limitations indefining the combined role of ERb and ERa in breast cancer as sug-gested earlier [46].

Despite the observed associations of ERb1 and ERb2 to ERaexpression ratios with tumor size, no associations were found withDFI. Interestingly, DFI was significantly associated with the ERb1 toERb2 expression ratio. By using the arbitrary cut-off of four-timesmore ERb2 than ERb1 (based on differences in Ct values) weshowed that higher ERb2 than ERb1 mRNA expression was an inde-pendent predictor of better outcome in late-onset patients.

Our study has a few limitations. Firstly, the study size with 79breast tumors is rather small. Secondly, the measurement of ERbexpression on the mRNA level may affect the potential impact ofour work since no correlation of ERb mRNA with protein expres-sion or ER status has been reported in some studies [47,48]. How-ever, in contrast to these studies, various other studies includingthe present one have measured ERb expression levels on the mRNAand not on the protein level and have reported significant correla-tions of ERb2 mRNA expression levels with clinical outcome[12,49,50]. Moreover, in studies in which ERb was determined byIHC and RT-PCR [22,33,51] ERb2 mRNA has also been shown tobe a predictor of good clinical outcome. The lack of correlationsof ERb mRNA with protein levels may be at least in part due tothe use of different measuring strategies with proteins being usu-ally determined by IHC and mRNA from tissue extracts by RT-PCR.In addition, the quantitative range of IHC is limited to few catego-ries depending on the scoring system and results may be affectedby the choice of available antibodies that often vary in their effi-ciency and specificity [52,53]. Finally, there may be differences inthe patient cohorts, especially with respect to the ERa status[54]. Thirdly, in our study mRNA was isolated from micro-dissected tumor regions with a tumor cell content of 80–90% afterH&E sections have been inspected by a pathologist. However, sinceERb is widely expressed in many different cell types [55–57], itmay be possible that the levels of ERb are overestimated becauseof the remaining 10–20% of cells within the tumors.

In summary, in the present study we found an inverse correla-tion of tumor size with relative ERb1 mRNA levels compared toERa in early-onset breast tumors and with relative ERb2 mRNAlevels in late-onset tumors, while no associations with absoluteERb1 and ERb2 mRNA levels were detected. In addition, moreERb2 compared to ERb1 seemed to be a predictor of better outcomein the late-onset patient group. The associations of ERb2 with both,tumor size and DFI in elder patients imply a role of ERb2 in thedevelopment of late-onset breast cancer. Our findings show thatthe ERa:ERb1, ERa:ERb2 and ERb1:ERb2 expression ratios appearto influence tumor progression and course of disease rather thantheir absolute levels. Moreover, they suggest that some biologicalmarkers may have different prognostic values in young and elderpatients. Given the small size of the present study, our data haveto be considered preliminary. For confirmation, larger studies onearly- and late onset breast tumors are warranted.

Funding

This work was supported by a UICC American Cancer SocietyBeginning Investigators Fellowship funded by the American CancerSociety (ACSBI Award ACS/07/014 2007), the DeutschesKrebsforschungszentrum, Heidelberg and the Ministry for Science

and Technology Development of Serbia (Grants 173049 and175068).

Acknowledgment

We thank Rodney J. Scott for critical reading of the manuscript.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.canlet.2012.02.022.

References

[1] B.E. Henderson, R. Ross, L. Bernstein, Estrogens as a cause of human cancer: theRichard and Hinda Rosenthal foundation award lecture, Cancer Res. 48 (1988)246–253.

[2] S. Green, P. Walter, V. Kumar, A. Krust, J.M. Bornert, P. Argos, P. Chambon,Human oestrogen receptor cDNA: sequence, expression and homology to v-erb-A, Nature 320 (1986) 134–139.

[3] G.L. Greene, P. Gilna, M. Waterfield, A. Baker, Y. Hort, J. Shine, Sequence andexpression of human estrogen receptor complementary DNA, Science 231(1986) 1150–1154 (New York, NY).

[4] G.G. Kuiper, E. Enmark, M. Pelto-Huikko, S. Nilsson, J.A. Gustafsson, Cloning of anovel receptor expressed in rat prostate and ovary, Proc. Nat. Acad. Sci. USA 93(1996) 5925–5930.

[5] B. Saville, M. Wormke, F. Wang, T. Nguyen, E. Enmark, G. Kuiper, J.A.Gustafsson, S. Safe, Ligand-, cell-, and estrogen receptor subtype (alpha/beta)-dependent activation at GC-rich (Sp1) promoter elements, J. Biol. Chem.275 (2000) 5379–5387.

[6] P.J. Kushner, D.A. Agard, G.L. Greene, T.S. Scanlan, A.K. Shiau, R.M. Uht, P. Webb,Estrogen receptor pathways to AP-1, The J. Steroid Biochem. Mol. Biol. 74(2000) 311–317.

[7] S. Chu, Y. Nishi, T. Yanase, H. Nawata, P.J. Fuller, Transrepression of estrogenreceptor beta signaling by nuclear factor-kappab in ovarian granulosa cells,Mol. Endocrinol. 18 (2004) 1919–1928 (Baltimore, Md).

[8] K. Pettersson, K. Grandien, G.G. Kuiper, J.A. Gustafsson, Mouse estrogenreceptor beta forms estrogen response element-binding heterodimers withestrogen receptor alpha, Mol. Endocrinol. 11 (1997) 1486–1496 (Baltimore,Md).

[9] S. Maruyama, N. Fujimoto, K. Asano, A. Ito, Suppression by estrogen receptorbeta of AP-1 mediated transactivation through estrogen receptor alpha, The J.Steroid Biochem. Mol. Biol. 78 (2001) 177–184.

[10] C.Y. Lin, A. Strom, S. Li Kong, S. Kietz, J.S. Thomsen, J.B. Tee, V.B. Vega, L.D.Miller, J. Smeds, J. Bergh, J.A. Gustafsson, E.T. Liu, Inhibitory effects of estrogenreceptor beta on specific hormone-responsive gene expression and associationwith disease outcome in primary breast cancer, Breast Cancer Res. 9 (2007)R25.

[11] E.C. Chang, J. Frasor, B. Komm, B.S. Katzenellenbogen, Impact of estrogenreceptor beta on gene networks regulated by estrogen receptor alpha in breastcancer cells, Endocrinology 147 (2006) 4831–4842.

[12] B.W. Park, K.S. Kim, M.K. Heo, W.I. Yang, S.I. Kim, J.H. Kim, G.E. Kim, K.S. Lee,The changes of estrogen receptor-beta variants expression in breastcarcinogenesis: decrease of estrogen receptor-beta2 expression is the keyevent in breast cancer development, J. Surg. Oncol. 93 (2006) 504–510.

[13] A. Bardin, N. Boulle, G. Lazennec, F. Vignon, P. Pujol, Loss of ERbeta expressionas a common step in estrogen-dependent tumor progression, Endocr. Relat.Cancer 11 (2004) 537–551.

[14] B. Peng, B. Lu, E. Leygue, L.C. Murphy, Putative functional characteristics ofhuman estrogen receptor-beta isoforms, J. Mol. Endocrinol. 30 (2003) 13–29.

[15] C. Zhao, J. Matthews, M. Tujague, J. Wan, A. Strom, G. Toresson, E.W. Lam, G.Cheng, J.A. Gustafsson, K. Dahlman-Wright, Estrogen receptor beta2 negativelyregulates the transactivation of estrogen receptor alpha in human breastcancer cells, Cancer Res. 67 (2007) 3955–3962.

[16] S. Ogawa, S. Inoue, T. Watanabe, A. Orimo, T. Hosoi, Y. Ouchi, M. Muramatsu,Molecular cloning and characterization of human estrogen receptor betacx: apotential inhibitor ofestrogen action in human, Nucleic Acids Res. 26 (1998)3505–3512.

[17] S.K. Gruvberger-Saal, P.O. Bendahl, L.H. Saal, M. Laakso, C. Hegardt, P. Eden, C.Peterson, P. Malmstrom, J. Isola, A. Borg, M. Ferno, Estrogen receptor betaexpression is associated with tamoxifen response in ERalpha-negative breastcarcinoma, Clin. Cancer Res. 13 (2007) 1987–1994.

[18] V. Speirs, A.T. Parkes, M.J. Kerin, D.S. Walton, P.J. Carleton, J.N. Fox, S.L. Atkin,Coexpression of estrogen receptor alpha and beta: poor prognostic factors inhuman breast cancer?, Cancer Res 59 (1999) 525–528.

[19] N. Honma, R. Horii, T. Iwase, S. Saji, M. Younes, K. Takubo, M. Matsuura, Y. Ito,F. Akiyama, G. Sakamoto, Clinical importance of estrogen receptor-betaevaluation in breast cancer patients treated with adjuvant tamoxifentherapy, J. Clin. Oncol. 26 (2008) 3727–3734.

[20] A.M. Shaaban, A.R. Green, S. Karthik, Y. Alizadeh, T.A. Hughes, L. Harkins, I.O.Ellis, J.F. Robertson, E.C. Paish, P.T. Saunders, N.P. Groome, V. Speirs, Nuclearand cytoplasmic expression of ERbeta1, ERbeta2, and ERbeta5 identifies

V. Mandusic et al. / Cancer Letters 321 (2012) 73–79 79

distinct prognostic outcome for breast cancer patients, Clin. Cancer Res. 14(2008) 5228–5235.

[21] N. Honma, S. Saji, R. Kurabayashi, J. Aida, T. Arai, R. Horii, F. Akiyama, T. Iwase,N. Harada, M. Younes, M. Toi, K. Takubo, G. Sakamoto, Oestrogen receptor-beta1 but not oestrogen receptor-betacx is of prognostic value in apocrinecarcinoma of the breast, APMIS 116 (2008) 923–930.

[22] R. Vinayagam, D.R. Sibson, C. Holcombe, V. Aachi, M.P. Davies, Association ofoestrogen receptor beta 2 (ER beta 2/ER beta cx) with outcome of adjuvantendocrine treatment for primary breast cancer – a retrospective study, BMCCancer 7 (2007) 131.

[23] J. Matthews, J.A. Gustafsson, Estrogen signaling: a subtle balance between ERalpha and ER beta, Mol. Interv. 3 (2003) 281–292.

[24] K. Pettersson, F. Delaunay, J.A. Gustafsson, Estrogen receptor beta acts as adominant regulator of estrogen signaling, Oncogene 19 (2000) 4970–4978.

[25] M. Esslimani-Sahla, J. Simony-Lafontaine, A. Kramar, R. Lavaill, C. Mollevi, M.Warner, J.A. Gustafsson, H. Rochefort, Estrogen receptor beta (ER beta) levelbut not its ER beta cx variant helps to predict tamoxifen resistance in breastcancer, Clin. Cancer Res. 10 (2004) 5769–5776.

[26] P. Chomczynski, N. Sacchi, Single-step method of RNA isolation by acidguanidinium thiocyanate-phenol-chloroform extraction, Anal. Biochem. 162(1987) 156–159.

[27] S.E. Bulun, Y.H. Cheng, M.E. Pavone, Q. Xue, E. Attar, E. Trukhacheva, H.Tokunaga, H. Utsunomiya, P. Yin, X. Luo, Z. Lin, G. Imir, S. Thung, E.J. Su, J.J. Kim,Estrogen receptor-beta, estrogen receptor-alpha, and progesterone resistancein endometriosis, Semin. Reprod. Med. 28 (2010) 36–43.

[28] Y. Omoto, H. Eguchi, Y. Yamamoto-Yamaguchi, S. Hayashi, Estrogen receptor(ER) beta1 and ERbetacx/beta2 inhibit ERalpha function differently in breastcancer cell line MCF7, Oncogene 22 (2003) 5011–5020.

[29] C. Yau, V. Fedele, R. Roydasgupta, J. Fridlyand, A. Hubbard, J.W. Gray, K. Chew,S.H. Dairkee, D.H. Moore, F. Schittulli, S. Tommasi, A. Paradiso, D.G. Albertson,C.C. Benz, Aging impacts transcriptomes but not genomes of hormone-dependent breast cancers, Breast Cancer Res. 9 (2007) R59.

[30] J. Quong, S. Eppenberger-Castori, D. Moore 3rd, G.K. Scott, M.J. Birrer, W.Kueng, U. Eppenberger, C.C. Benz, Age-dependent changes in breast cancerhormone receptors and oxidant stress markers, Breast Cancer Res. Treat. 76(2002) 221–236.

[31] S. Ogawa, S. Inoue, T. Watanabe, H. Hiroi, A. Orimo, T. Hosoi, Y. Ouchi, M.Muramatsu, The complete primary structure of human estrogen receptor beta(hER beta) and its heterodimerization with ER alpha in vivo and in vitro,Biochem. Biophys. Res. Commun. 243 (1998) 122–126.

[32] S.I. Hayashi, H. Eguchi, K. Tanimoto, T. Yoshida, Y. Omoto, A. Inoue, N. Yoshida,Y. Yamaguchi, The expression and function of estrogen receptor alpha and betain human breast cancer and its clinical application, Endocr. Relat. Cancer 10(2003) 193–202.

[33] S. Saji, Y. Omoto, C. Shimizu, M. Warner, Y. Hayashi, S. Horiguchi, T. Watanabe,S. Hayashi, J.A. Gustafsson, M. Toi, Expression of estrogen receptor (ER)(beta)cx protein in ER(alpha)-positive breast cancer: specific correlation withprogesterone receptor, Cancer Res. 62 (2002) 4849–4853.

[34] S. Eppenberger-Castori, D.H. Moore Jr., A.D. Thor, S.M. Edgerton, W. Kueng, U.Eppenberger, C.C. Benz, Age-associated biomarker profiles of human breastcancer, Int. J. Biochem. Cell Biol. 34 (2002) 1318–1330.

[35] D. Nikolic-Vukosavljevic, N. Vasiljevic, M. Brankovic-Magic, D. Polic, Variationsin steroid hormone receptor content throughout age and menopausal periods,and menstrual cycle in breast cancer patients, Neoplasma 43 (1996) 163–169.

[36] T. Yoshida, H. Eguchi, K. Nakachi, K. Tanimoto, Y. Higashi, K. Suemasu, Y. Iino,Y. Morishita, S. Hayashi, Distinct mechanisms of loss of estrogen receptoralpha gene expression in human breast cancer: methylation of the gene andalteration of trans-acting factors, Carcinogenesis 21 (2000) 2193–2201.

[37] S. Hayashi, K. Imai, K. Suga, T. Kurihara, Y. Higashi, K. Nakachi, Two promotersin expression of estrogen receptor messenger RNA in human breast cancer,Carcinogenesis 18 (1997) 459–464.

[38] K. Tanimoto, H. Eguchi, T. Yoshida, K. Hajiro-Nakanishi, S. Hayashi, Regulationof estrogen receptor alpha gene mediated by promoter B responsible for itsenhanced expression in human breast cancer, Nucleic Acids Res. 27 (1999)903–909.

[39] F. Holst, P.R. Stahl, C. Ruiz, O. Hellwinkel, Z. Jehan, M. Wendland, A. Lebeau, L.Terracciano, K. Al-Kuraya, F. Janicke, G. Sauter, R. Simon, Estrogen receptor

alpha (ESR1) gene amplification is frequent in breast cancer, Nat. Genet. 39(2007) 655–660.

[40] G. Lazennec, D. Bresson, A. Lucas, C. Chauveau, F. Vignon, ER beta inhibitsproliferation and invasion of breast cancer cells, Endocrinology 142 (2001)4120–4130.

[41] J.P. Brouillet, M.A. Dujardin, D. Chalbos, J.M. Rey, J. Grenier, P.J. Lamy, T.Maudelonde, P. Pujol, Analysis of the potential contribution of estrogenreceptor (ER) beta in ER cytosolic assay of breast cancer, Int. J. Cancer 95(2001) 205–208.

[42] X. Cui, R. Schiff, G. Arpino, C.K. Osborne, A.V. Lee, Biology of progesteronereceptor loss in breast cancer and its implications for endocrine therapy, J. Clin.Oncol. 23 (2005) 7721–7735.

[43] V. Mandusic, D. Nikolic-Vukosavljevic, N. Tanic, K. Kanjer, Z. Neskovic-Konstantinovic, D. Celeketic, B. Dimitrijevic, Expression of estrogen receptorbeta wt isoform (ERbeta1) and ERbetaDelta5 splice variant mRNAs in sporadicbreast cancer, J. Cancer Res. Clin. Oncol. 133 (2007) 571–579.

[44] V. Speirs, C. Malone, D.S. Walton, M.J. Kerin, S.L. Atkin, Increased expression ofestrogen receptor beta mRNA in tamoxifen-resistant breast cancer patients,Cancer Res. 59 (1999) 5421–5424.

[45] K. Paech, P. Webb, G.G. Kuiper, S. Nilsson, J. Gustafsson, P.J. Kushner, T.S.Scanlan, Differential ligand activation of estrogen receptors ERalpha andERbeta at AP1 sites, Science 277 (1997) 1508–1510 (New York, NY).

[46] S.J. Schnitt, Estrogen receptor testing of breast cancer in current clinicalpractice. What’s the question?, J Clin. Oncol. 24 (2006) 1797–1799.

[47] R. Cullen, T.M. Maguire, E.W. McDermott, A.D. Hill, N.J. O’Higgins, M.J. Duffy,Studies on oestrogen receptor-alpha and -beta mRNA in breast cancer, Eur. J.Cancer 37 (2001) 1118–1122.

[48] P.A. O’Neill, M.P. Davies, A.M. Shaaban, H. Innes, A. Torevell, D.R. Sibson, C.S.Foster, Wild-type oestrogen receptor beta (ERbeta1) mRNA and proteinexpression in tamoxifen-treated post-menopausal breast cancers, Br. J.Cancer 91 (2004) 1694–1702.

[49] I. Poola, S.A. Fuqua, R.L. De Witty, J. Abraham, J.J. Marshallack, A. Liu, Estrogenreceptor alpha-negative breast cancer tissues express significant levels ofestrogen-independent transcription factors, ERbeta1 and ERbeta5: potentialmolecular targets for chemoprevention, Clin. Cancer Res. 11 (2005) 7579–7585.

[50] M.P. Davies, P.A. O’Neill, H. Innes, D.R. Sibson, W. Prime, C. Holcombe, C.S.Foster, Correlation of mRNA for oestrogen receptor beta splice variantsERbeta1, ERbeta2/ERbetacx and ERbeta5 with outcome in endocrine-treatedbreast cancer, J. Mol. Endocrinol. 33 (2004) 773–782.

[51] H. Sugiura, T. Toyama, Y. Hara, Z. Zhang, S. Kobayashi, Y. Fujii, H. Iwase, H.Yamashita, Expression of estrogen receptor beta wild-type and its variantER{beta}cx/{beta}2 is correlated with better prognosis in breast cancer, Jpn. J.Clin. Oncol. (2007).

[52] G.P. Skliris, A.T. Parkes, J.L. Limer, S.E. Burdall, P.J. Carder, V. Speirs, Evaluationof seven oestrogen receptor beta antibodies for immunohistochemistry,western blotting, and flow cytometry in human breast tissue, J. Pathol. 197(2002) 155–162.

[53] G.E. Weitsman, G. Skliris, K. Ung, B. Peng, M. Younes, P.H. Watson, L.C. Murphy,Assessment of multiple different estrogen receptor-beta antibodies for theirability to immunoprecipitate under chromatin immunoprecipitationconditions, Breast Cancer Res. Treat. 100 (2006) 23–31.

[54] G.P. Skliris, E. Leygue, P.H. Watson, L.C. Murphy, Estrogen receptor alphanegative breast cancer patients: estrogen receptor beta as a therapeutic target,The J. Steroid Biochem. Mol. Biol. 109 (2008) 1–10.

[55] E.V. Jensen, G. Cheng, C. Palmieri, S. Saji, S. Makela, S. Van Noorden, T.Wahlstrom, M. Warner, R.C. Coombes, J.A. Gustafsson, Estrogen receptors andproliferation markers in primary and recurrent breast cancer, Proc. Nat. Acad.Sci. USA 98 (2001) 15197–15202.

[56] V. Speirs, G.P. Skliris, S.E. Burdall, P.J. Carder, Distinct expression patterns of ERalpha and ER beta in normal human mammary gland, J. Clin. Pathol. 55 (2002)371–374.

[57] M. Cummings, J. Iremonger, C.A. Green, A.M. Shaaban, V. Speirs, Geneexpression of ER beta isoforms in laser microdissected human breast cancers:implications for gene expression analyses, Cell Oncol. 31 (2009) 467–473.