ORIGINAL ARTICLE

A positive feedback loop of ER-a36/EGFR promotes malignant growth

of ER-negative breast cancer cells

XT Zhang1, LG Kang1, L Ding2, S Vranic3,4, Z Gatalica3 and Z-Y Wang1,3

1Department of Medical Microbiology & Immunology, Creighton University Medical School, Omaha, NE, USA; 2Departmentof Oncology, Second Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, Zhejiang, PR China; 3Departmentof Pathology, Creighton University Medical School, Omaha, NE, USA and 4Department of Pathology, Clinical Center of theUniversity of Sarajevo, Sarajevo, Bosnia and Herzegovina

It is prevailingly thought that estrogen signaling is notinvolved in development of estrogen receptor (ER)-negative breast cancer. However, there is evidenceindicating that ovariectomy prevents the development ofboth ER-positive and -negative breast cancer, suggestingthat estrogen signaling is involved in the development ofER-negative breast cancer. Previously, our laboratorycloned a variant of ER-a, ER-a36, and found that ER-a36mediated nongenomic estrogen signaling and is highlyexpressed in ER-negative breast cancer cells. In thisstudy, we found that ER-a36 was highly expressed in10/12 cases of triple-negative breast cancer. We investi-gated the role of mitogenic estrogen signaling mediated byER-a36 in malignant growth of triple-negative breastcancer MDA-MB-231 and MDA-MB-436 cells thatexpress high levels of ER-a36 and found that these cellsstrongly responded to mitogenic estrogen signaling bothin vitro and in vivo. Knockdown of ER-a36 expression inthese cells using the small hairpin RNA method dimini-shed their responsiveness to estrogen. ER-a36 physicallyinteracted with the EGFR/Src/Shc complex and mediatedestrogen-induced phosphorylation of epidermal growthfactor receptor (EGFR) and Src. EGFR signalingactivated ER-a36 transcription through an AP1 site inthe ER-a36 promoter, and ER-a36 expression was able tostabilize EGFR protein. Our results, thus demonstratedthat ER-a36 mediates nongenomic estrogen signalingthrough the EGFR/Src/ERK signaling pathway in ER-negative breast cancer cells and suggested that a subset ofER-negative breast tumors that expresses ER-a36, retainsresponsiveness to mitogenic estrogen signaling.Oncogene (2011) 30, 770–780; doi:10.1038/onc.2010.458;published online 11 October 2010

Keywords: EGFR; ER-a36; estrogen signaling; ER-negativebreast cancer

Introduction

Estrogen receptor (ER)-negative breast cancer constitu-tes B30% of all the breast cancers and generally is moreaggressive than ER-positive breast cancer (Lacroixet al., 2004; Simpson et al., 2005). Because of the lackof ER-a expression, it is prevailingly thought thatestrogen signaling is not involved in development andprogression of ER-negative breast cancer. However,several early reports showed that ovariectomy preventsformation of both ER-positive and –negative breastcancers (Nissen-Meyer, 1964; Early Breast CancerTrialists’ Collaborative Group, 1992). Interestingly,BRCA1-mutation-related breast tumors, the vast ma-jority of which are ER-negative, are also effectivelyprevented by prophylactic ovariectomy (Rebbeck et al.,1999; Narod, 2001). These observations suggestedthat ovarian hormones contribute to development ofER-negative breast cancers. Previously, it was reportedthat estrogen activates the PI3K/AKT phosphorylationin ER-negative breast cancer cells (Tsai et al., 2001).Estrogen treatment was reported to stimulate malignantgrowth of ER-negative breast cancer MDA-MB-231cells in immunodeficient mice (Friedl and Jordan, 1994).These results suggested that some ER-negative breastcancer cell lines may retain nongenomic and mitogenicestrogen signaling. However, the molecular mechanismsunderlying these observations are largely unknown.

Previously, we identified and cloned a 36 kDa variantof ER-a, ER-a36, which is mainly expressed on theplasma membrane and mediates nongenomic estrogenicsignaling (Wang et al., 2005, 2006). ER-a36 lacks bothtranscription activation domains AF-1 and AF-2 of the66 kDa full-length ER-a (ER-a66), and possesses analtered ligand-binding domain and an intact DNA-binding domain, consistent with the fact that ER-a36has no intrinsic transcriptional activity but mediatesnongenomic estrogen signaling (Wang et al., 2006).ER-a36 is generated from a promoter located in the firstintron of the ER-a66 gene (Zou et al., 2009), indicatingthat ER-a36 expression is regulated differently fromER-a66, consistent with the findings that ER-a36 isexpressed in specimens from ER-negative patients andestablished ER-negative breast cancer cells that lackER-a66 expression (Wang et al., 2006; Lee et al., 2008;Shi et al., 2009).

Received 1 April 2010; revised 26 August 2010; accepted 1 September2010; published online 11 October 2010

Correspondence: Professor Z-Y Wang, Department of MedicalMicrobiology & Immunology, Creighton University Medical School,Criss III, Room 355, 2500 California Plaza, Omaha, NE 68178, USA.E-mail: zywang@creighton.edu

Oncogene (2011) 30, 770–780& 2011 Macmillan Publishers Limited All rights reserved 0950-9232/11

www.nature.com/onc

We have investigated the contribution of nongenomicestrogen signaling mediated by ER-a36 to malignantgrowth of ER-negative breast cancer cells. Here, wehave demonstrated the existence of a positive feedbackloop between ER-a36 and epidermal growth factorreceptor (EGFR) expression, and a cross talk betweennongenomic estrogen signaling mediated by ER-a36 andthe EGFR/Src/ERK signaling pathway, which pro-motes malignant growth of ER-negative breast cancercells.

Results

Nongenomic estrogen signaling stimulates proliferationof ER-negative breast cancer cellsHere, we examined ER-a36 expression in 12 cases oftriple-negative breast cancer (ER-a66�, PR� and Her2/neu�) and found that 10 out of the 12 cases exhibitedER-a36 expression, predominantly in a cytoplasmic andmembranous pattern (Supplementary Figure 1). Themean percentage of the ER-a36-positive cells was 53%and the majority of the cases showed weak to moderateER-a36 staining. EGFR expression was detected in sixcases, four of which coexpressed ER-a36. These resultssuggested that a subset of triple-negative breast cancerlacks expression of the full-length ER-a (ER-a66) butexpresses a variant of ER-a66, ER-a36.

To determine whether established triple-negativebreast cancer cells that express ER-a36 retain nonge-nomic estrogen signaling, we used breast cancer MDA-MB-231 and MDA-MB-436 cells, both of which aretriple-negative. Western blot analysis showed that bothER-a36 and EGFR are highly expressed in these breastcancer cells, whereas ER-positive MCF7 cells expressedhigh levels of ER-a66 but lower levels of ER-a36 andEGFR (Figure 1a).

To determine whether 17b-estradiol (E2b) inducedphosphorylation of the MAPK/ERK1/2, a typicalnongenomic estrogen-signaling event, in these two celllines, we treated cells with E2b at different concentra-tions and for different time periods. Western blotanalysis with a phospho-specific ERK1/2 antibody wasperformed. Figure 1b shows that E2b elicited ERKphosphorylation in both cell lines in a dose-dependentmanner starting at a extremely low concentration,1� 10�16 M/l. Time course analysis in MDA-MB-231cells revealed that ERK phosphorylation occurredwithin 5min after E2b application, peaked at 15min,declined at 30min and then exhibited another moresustained activation at 60min. However, this double-peak induction pattern of the MAPK/ERK was notobvious in MDA-MB-436 cells (Figure 1b). Conse-quently, E2b was also able to induce expression of thegrowth-promoting genes c-Myc and cyclin D1 in bothcell lines (Figure 1c). These results demonstrated thatthese triple-negative breast cancer cells retained non-genomic estrogen signaling.

We then decided to determine whether estrogenstimulates proliferation of these triple-negative breastcancer cells. As these triple-negative breast cancer cells

express high levels of EGFR, which makes these cellsproliferate at a near-maximal rate in serum-supplemen-ted medium, the stimulating effects of estrogen signalingon proliferation of these cells are, most of the time, toosubtle to detect in vitro in the presence of 10–5% fetalcalf serum (Friedl and Jordan, 1994; Rai et al., 2005). Toalleviate this problem, we devised a new strategy byreducing charcoal-stripped fetal calf serum concentra-tion from 10 to 2.5% and increased estrogen treatmenttime to 12 days. As shown in Figure 1d, the cells treatedwith E2b exhibited a significantly increased growth ratecompared with cells treated with vehicle (Figure 1d).Our data thus demonstrated that mitogenic estrogensignaling stimulates proliferation of these ER-negativebreast cancer cells.

ER-a36 mediates mitogenic estrogen signaling inER-negative breast cancer cellsTo determine whether ER-a36 mediates mitogenicestrogen signaling in these breast cancer cells, wedesigned two small hairpin RNA (shRNA) expressionvectors targeting different regions of the 30UTR of ER-a36 and established two clonal cell lines from MDA-MB-231 cells that express these two different shRNAs.MDA-MB-231 cells transfected with an empty expres-sion vector or an expression vector for shRNA againstfirefly luciferase were used as controls. Both western blotanalysis and reverse transcriptase PCR demonstratedthat ER-a36 expression was knocked-down about 80%in the shRNA-expression vector-transfected cells com-pared with control cells (Figure 2a). E2b treatmentfailed to induce ERK1/2 phosphorylation, expressionof c-Myc and cyclin D1, and cell proliferation inMDA-MB-231 cell lines with knocked-down level ofER-a36 expression (Figures 2b and c). Similar resultswere also observed in MDA-MB-436 cells with knocked-down levels of ER-a36 expression (Figures 2d–g).However, serum was still able to induce ERK activationin MDA-MB-436 cells with knocked-down level of ER-a36 expression (Figure 2e), indicating there was nodefect of the MAPK/ERK signaling in cells with ER-a36 expression knocked down. These results demon-strated that ER-a36 mediates nongenomic and mito-genic estrogen signaling in these ER-negative breastcancer cells.

Previously, ER-negative breast cancer MDA-MB-231cells were found to express ER-b, a subtype of ER (Tonget al., 2002). Western blot analysis showed that MDA-MB-231 cells express higher levels of ER-b comparedwith ER-positive MCF7 cells, whereas MDA-MB-436cells express undetectable levels of ER-b (Supplemen-tary Figure 2A). To determine whether ER-b is involvedin the estrogen effects observed in MDA-MB-231 cells,we knocked-down ER-b expression in MDA-MB-231cells with transient transfection of ER-b small interfer-ing RNA (siRNA). The estrogen effects, such asactivation of the MAPK/ERK and stimulation of cellproliferation, were intact and even with a slight increase(Supplementary Figures 2B–D), suggesting that ER-bmay negatively regulate mitogenic estrogen signalingmediated by ER-a36.

Mitogenic estrogen signaling in ER-negative breast cancerXT Zhang et al

771

Oncogene

To assess the effect of estrogen signaling on thetumorigenicity of these cells, two MDA-MB-231 celllines with knocked-down levels of ER-a36 expression

and a control cell line transfected with the emptyexpression vector were inoculated subcutaneously intothe fat pad of ovariectomized female nude mice. At 5

Mitogenic estrogen signaling in ER-negative breast cancerXT Zhang et al

772

Oncogene

days before inoculation, mice were implanted with E2bor placebo pellets. In two independent experiments,tumors were readily detected at all sites injected withvector-control cells in the absence and presence ofestrogen supplement (Figure 3a). However, the tumorsthat formed in the presence of estrogen developed morerapidly than their counterparts without estrogen.Additionally, MDA-MB-231 cells with knocked-downlevel of ER-a36 expression failed to form tumors in theabsence of estrogen, whereas formed tumors lessefficiently compared with control cells in the presenceof estrogen (Figure 3a). Our results were in goodagreement with the previous report that estrogen wasable to stimulate malignant growth of MDA-MB-231cells in vivo (Friedl and Jordan, 1994).

To test whether the tumor-enhancing effects ofestrogen extended to weak tumorigenic breast cancercells, we repeated the above experiment with MDA-MB-436 cells. The vector-control MDA-MB-436 cellsformed palpable mammary tumors in about 6 weeksin the absence of estrogen. In contrast, the vector-control cells in the supplement of estrogen formedtumor with higher efficiency and a significant shorterlatency, arising 2–3 weeks before their counterpartswithout estrogen. MDA-MB-436 cells with knocked-down level of ER-a36 expression did not form tumorsin the absence of estrogen, even in a prolongedincubation (20 weeks), whereas developed tumorswith much less efficiency in the presence of estrogen.These experiments demonstrated estrogen stimulatedmalignant growth of these ER-negative breast cancercells in vivo.

EGFR protein is stabilized by ER-a36 in ER-negativebreast cancer cellsThe finding that ER-a36 downregulation dramaticallysuppresses the tumorigenicity of these ER-negativebreast cancer cells in the absence of estrogen wassurprising as these cells also express high levels ofEGFR, which would promote malignant growth in vivo.To elucidate the underlying mechanisms, western blotanalysis was performed to examine the expression levelsof EGFR protein in the cells with ER-a36 expressionknocked down. Figure 4a shows that expression levels ofEGFR protein were dramatically decreased in MDA-MB-231 cell lines with ER-a36 expression knocked-down compared with control cells. However, we did not

observe significant change in the mRNA levels of EGFRin these cells, suggesting that the steady state levels ofEGFR protein were decreased in ER-a36 knocked-down MDA-MB-231 cells (Figure 4a). A similardestabilization of EGFR protein was also observed inMDA-MB-436 cells with knocked-down levels of ER-a36 expression (Figure 4b). Western blot analysisfurther demonstrated that on treatment with MG132,a proteasome inhibitor, the levels of EGFR protein inthe ER-a36 knocked-down cells were restored to levelscomparable with those of control cells (Figure 4c),indicating that the protein degradation of EGFR wasenhanced in ER-a36 downregulated cells. Thus, ER-a36is involved in regulation of the steady-state levels ofEGFR protein.

EGFR signaling upregulates ER-a36 expressionRecently, we cloned the promoter region of ER-a36 andidentified that ER-a36 promoter harbors several Ap-1binding sites (Zou et al., 2009), suggesting that ER-a36expression may be subjected to regulation of the growthfactor signaling pathways. To determine whether EGFRsignaling influences ER-a36 expression, we treated bothcell lines with the PI3K inhibitor LY294002, EGFRinhibitors BiBx, AG1478 and Gefitinib, and the Srcinhibitor PP2. Figure 5a shows that treatment with theEGFR inhibitors strongly downregulated ER-a36 ex-pression at both protein and mRNA levels in both celllines, whereas the PI3K inhibitor LY294002 had noeffect. Our data thus suggested that ER-a36 transcrip-tion is subjected to positive regulation by EGFRsignaling.

To examine whether EGFR signaling directly upre-gulates ER-a36 promoter activity, we performed co-transfection assays in human embryonic kidney (HEK)293 cells that express no detectable levels of bothER-a36 and EGFR. HEK-293 cells were transientlytransfected with a luciferase reporter plasmid drivenby the ER-a36 promoter (Zou et al., 2009). EGFRco-transfection resulted in an about 3-fold inductionof ER-a36 promoter activity, which was blocked bypretreatment of the EGFR inhibitor AG1478 andGefitinib, and the Src inhibitor PP2, but not by thePI3K inhibitor LY294002 (Figures 5b and c). Whena series of 50 truncated promoter of ER-a36 (Figure 5b)was used, we found that EGFR expression failedto activate the promoter activity of the pER36-513

Figure 1 Nongenomic estrogen signaling stimulates proliferation of ER-negative breast cancer cells. (a) The expression of ER-avariants and EGFR in MCF7, MDA-MB-231 and MDA-MB-436 breast cancer cells. (b) The dose-and time-dependent pattern of E2b-stimulated phosphorylation of the MAPK/ERK1/2 in MDA-MB-231 and MDA-MB-436 cells. Starved cells were treated withindicated doses of E2b or 0.1 nM of E2b for indicated time periods. Western blot analysis was performed to assess induction of ERK1/2phosphorylation. Columns represent the means of three experiments; bars represent the s.e. *Po0.05 for control cells vs cells treatedunder different conditions. The representative results are shown. (c) The dose-dependent induction of c-myc and cyclin D1 by E2b inMDA-MB-231 and MDA-MB-436 cells. Columns represent the means of three experiments; bars represent the s.e. *Po0.05 for cellstreated with vehicle vs cells treated with different concentrations of E2b. The representative results are shown. (d) The effects of E2b onthe proliferation rate of MDA-MB-231 and MDA-MB-436 cells. Cells maintained for 3 days in phenol-red-free Dulbecco0s modiedEagle0s medium plus 2.5% dextran-charcoal-stripped fetal calf serum were treated with indicated concentrations of E2b or ethanolvehicle as a control. The cell numbers were determined using an automatic cell counter after 12 days. Five dishes were used for eachconcentration and experiments were repeated more than four times. The mean cell numbers±s.e. are shown.

Mitogenic estrogen signaling in ER-negative breast cancerXT Zhang et al

773

Oncogene

reporter plasmid (Figure 5b). Close examinationof DNA sequence in the deleted region revealed anAP-1-binding site located between �541 and �551(relative to the transcription initiation site) residues of

the ER-a36 promoter region (Figure 5b). Mutationof this Ap-1 site abrogated induction of ER-a36promoter activity by EGFR co-transfection(Figure 5b), suggesting that EGFR signaling activated

Mitogenic estrogen signaling in ER-negative breast cancerXT Zhang et al

774

Oncogene

ER-a36 promoter activity through the EGFR/Src/ERK1/2/AP-1 pathway.

ER-a36 interacts with the EGFR complex and mediatesE2b-induced phosphorylation of Src and EGFRTo elucidate the molecular mechanism underlying ER-a36 functions, we examined whether ER-a36 interactswith the EGFR complex. MDA-MB-231 cells weretransiently transfected with an expression vector forHA-tagged ER-a36 and coimmunoprecipitation assayswere performed with cell lysates from transfected cells.Figure 6a shows that EGFR, Src and Shc coexisted inthe immunoprecipitates of the anti-HA antibody. Wealso noticed that the levels of EGFR in the immuno-precipitates were slightly decreased after E2b treatment,whereas the levels of Shc and Src were increased(Figure 6a). In MDA-MB-436 cells, interaction betweenEGFR and ER-a36 was also observed in the absence ofE2b. After treatment of E2b for 1 h, ER-a36 wasgradually dissociated from the EGFR and associatedwith Src and Shc (Figure 6b).

We also examined changes of the phosphorylationlevels of EGFR with different phospho-specific anti-bodies against different residues of EGFR, includingTyr-845, Tyr-992, Tyr-1045, Tyr-1068 and Tyr-1173 inMDA-MB-231 cells in the presence and absence of E2b.We found that Tyr-845 was the only residue that wasphosphorylated after E2b application (SupplementaryFigure 3) and E2b elicited a transient increase of EGFR-Tyr-845 phosphorylation; started at 5min and declinedat 30min (Figure 6c). The E2b-induced phosphorylationwas totally abrogated by the Src inhibitor PP2 butpartially blocked by the EGFR inhibitor AG1478(Figures 6c and d), consistent with the previous reportthat Src phosphorylates EGFR at Tyr-845 (Biscardiet al., 1999). We also observed that E2b treatmentinduced phosphorylation of Src at Tyr-416 (Figure 6c),which was totally blocked by the Src inhibitor PP2 butpartially by the EGFR inhibitor AG1478 (Figure 6d),consistent with the activation pattern of the MAPK/ERK by E2b (Figure 6d). These results indicatedthat the EGFR/Src/Shc complex is involved in trans-duction of the nongenomic estrogen signaling mediatedby ER-a36.

Discussion

In this report, we found that 10 out of the 12 cases oftriple-negative breast cancer expressed ER-a36, predo-minantly on the plasma membrane and in the cyto-plasm. EGFR expression was detected in six cases, fourof which coexpressed ER-a36, which indicated that asubset of triple-negative breast cancer coexpresses ER-a36 and EGFR. We then used MDA-MB-231 andMDA-MB-436 cells as models to study the effects of the

Figure 2 ER-a36 mediates mitogenic estrogen signaling in ER-negative breast cancer cells. (a) Western blot and reverse transcriptasePCR (RT-PCR) analyses of ER-a36 expression in different MDA-MB-231 cell variants; parental cells (231/P), control cells (231/V,transfected with the empty expression vector; 231/L transfected with a luciferase small hairpin RNA expression vector) and ER-a36expression knocked-down cells (231/Sh36(3-1) and 231/Sh36(1-7)). (b) Western blot analysis of the effects of E2b on thephosphorylation levels of the MAPK/ERK1/2 and expression levels of c-myc and cyclin D1 in different MDA-MB-231 cell variants.Columns represent the means of three experiments; bars represent the s.e. *Po0.05 for cells treated with vehicle vs cells treated withdifferent concentrations of E2b. (c) The effects of E2b on the proliferation rate of different MDA-MB-231 cell variants. Cells weretreated with indicated concentrations of E2b or ethanol (vehicle) as a control. The MAPK inhibitor UO126 (1mM) was included insome experiments as indicated. The cell numbers were determined after 12 days. Five dishes were used for each concentration andexperiments were repeated more than five times. The mean cell numbers±s.e. are shown. (d) Expression levels of ER-a36 in controlMDA-MB-436 cells (436/V, transfected with the empty expression vector) and ER-a36 expression knocked-down MDA-MB-436 cells(436/Sh36) analyzed by western blot and reverse transcriptase PCR (RT-PCR) analyses. (e) E2b or serum induced activation of theMAPK/ERK in 436/V and 436/Sh36 cells. (f) E2b induced expression of c-Myc and cyclin D1 in 436/V and 436/Sh36 cells. Columnsrepresent the means of three experiments; bars represent the s.e. *Po0.05 for cells treated with vehicle vs cells treated with differentconcentrations of E2b. (g) The effects of different concentration of E2b on the proliferation rate of 436/V and 436/Sh36 cells. Theexperiments were repeated three times, and the mean cell numbers±s.e. are shown.

Figure 3 E2 enhances the rate of tumor growth in ER-negativebreast cancer cells in nude mice. Different variants of MDA-MB-231 (a) and MDA-MB-436 (b) cells were implanted into themammary fat pad of the ovariectomized female mice supplementedwith estrogen or placebo pellets. The tumorigenicity was examinedby measurement of tumor size. The experiments were repeatedonce. The data represent the means±s.e. observed in 12 mice ineach group.

Mitogenic estrogen signaling in ER-negative breast cancerXT Zhang et al

775

Oncogene

nongenomic estrogen signaling mediated by ER-a36 onthe malignant growth of ER-negative breast cancer.

Here, we found that ER-negative breast cancer cellsthat lack expression of ER-a66 but express ER-a36exhibited a potent mitogenic estrogen signaling in vitroand in vivo. Previously, other laboratories failed toobserve estrogen-stimulated growth of MDA-MB-231cells in vitro (Friedl and Jordan, 1994; Rai et al., 2005)but found stimulatory effects of estrogen in vivo (Friedland Jordan, 1994). One possible explanation to thediscrepancy between their data and ours is that thein vitro proliferation experiments by these laboratorieswere conducted in phenol-red-free medium supplemen-ted with 10-5% charcoal-treated fetal calf serum andassayed in 7 or 8 days, which made MDA-MB-231 cellsthat express high levels of EGFR, to grow at a rapid ratein serum-supplemented medium. The stimulating effectsof estrogen signaling on proliferation of these cells arenegligible most of the time (Friedl and Jordan, 1994; Raiet al., 2005). To minimize the growth rate of MDA-MB-231 cells, we used phenol-red-free medium containing

2.5% charcoal-treated serum (the minimum concentra-tion that keeps MDA-MB-231 cell viable) and increasedestrogen treatment time to 12 days in our growth assays.Under these conditions, we consistently observed potentgrowth promoting effects of estrogen on proliferation ofthese ER-negative breast cancer cells. It is worth notingthat E2b also stimulated growth of the ER-negativeendometrial carcinoma cells in athymic mice (Friedlet al., 1989), consistent with our recent report that ER-a36 is expressed in ER-negative endometrial carcinomacells (Lin et al., 2010).

Previously, ER-negative breast cancer cells werefound to express ER-b receptor (Tong et al., 2002).However, the role of ER-b in ER-negative breast canceris largely unknown. Previous studies showed that ER-binhibited proliferation of breast cancer cells by repres-sing c-myc and cyclin D1 (Lazennec et al., 2001;Paruthiyil et al., 2004; Behrens et al., 2007). In thisstudy, we used ER-b-specific siRNA to knock-downER-b expression in MDA-MB-231 cells and found thatcells with knocked-down levels of ER-b expressionretained a full (or even increased) response to mitogenicestrogen signaling, suggesting that ER-b may negativelyregulate ER-a36-mediated nongenomic estrogensignaling and that the inhibitory activities of ER-bmay be also involved in failure of c-myc and cyclin D1induction and loss of the tumorigenecity observed inMDA-MB-231 cells with knocked-down levels ofER-a36 expression.

In the present study, we also revealed a novel cross-talk mechanism in which EGFR and ER-a36 positivelyregulate each other’s expression, which may play animportant role in malignant growth of triple-negativebreast cancer. We also showed that E2b induced theMAPK/ERK activation through a mechanism thatinvolves ER-a36 and the EGFR/Src/Shc complex. Wenoted that ER-a36 interacted strongly with EGFR inthe absence of estrogen, whereas interaction betweenER-a36 and the Src/Shc was estrogen dependent,suggesting that ER-a36 may dynamically change itspartners during estrogen signaling. We also found thatE2b predominantly induced phosphorylation of theEGFR-Tyr-845 residue but not the major autopho-sphorylation sites of EGFR, such as Tyr-992, -1068 and-1073. EGFR Tyr-845 is a site phosphorylated byactivated Src (Biscardi et al., 1999). Consistent withthis, we found that E2b also induced Src-Tyr-416phosphorylation and the Src inhibitor PP2 blockedE2b-induced phosphorylation of Src-Tyr-416 andEGFR-Tyr-845, whereas the EGFR inhibitor AG1478had less effect. Our results thus indicated that EGFR/Src complex has an integral role in mitogenic estrogensignaling in ER-negative breast cancer cells that expressER-a36.

Clinical evidence established that ER-negative breastcancer is less or non-responsive to antiestrogen therapy,which would be at the odds with our finding thatestrogen signaling is involved in malignant growth ofER-negative breast cancer cells. It is well known thattamoxifen acts as both agonist and antagonist. Recently,we found that ER-a36-mediated agonist action of

Figure 4 EGFR protein is stabilized by ER-a36 in ER-negativebreast cancer cells. (a, b) Western blot and reverse transcriptasePCR (RT-PCR) analyses of EGFR expression in variants ofMDA-MB-231 and 436 cells. (c) Western blot analysis of EGFRexpression in MG132 treated variants of MDA-MB-231 cells. Allexperiments were repeated at least three times, and the representa-tive results are shown.

Mitogenic estrogen signaling in ER-negative breast cancerXT Zhang et al

776

Oncogene

tamoxifen in ER-negative endometrial cancer cells viathe MAPK/ERK and PI3K/Akt pathways (Lin et al.,2010). It is thus possible that ER-a36 may mediatetamoxifen agonist effects in the ER-negative breastcancer cells that express ER-a36. ICI 182780, a ‘pure’

antiestrogen, works by accelerating degradation of ER-a66 protein (Howell et al., 2000). Recently, we reportedthat ICI 182780 failed to induce degradation of ER-a36(Kang and Wang, 2010), presumably because ER-a36has a truncated ligand-binding domain, which lacks thelast four helices (helix 9–12) of ER-a66 (Wang et al.,2005); the helix-12 domain is critical in proteindegradation induced by ICI 182780 (Mahfoudi et al.,1995). This may provide a molecular explanation for theclinical evidence that these antiestrogens failed to inhibitgrowth of ER-negative breast cancer cells that expressER-a36. We also found that these ER-negative breastcancer cells responded to very low concentrations ofestrogen; activation of the MAPK/ERK signaling wasobserved at 10�16 M/l, suggesting that cells expressinghigh levels of ER-a36 are hypersensitive to estrogen.Thus, ER-negative breast cancer cell that expresses highlevels of ER-a36 may be hypersensitive to estrogen,which may render this subset of breast cancer lesssensitive to aromatase inhibitors that usually suppressthe plasma level of E2b to a mean of picomolar range(Geisler et al., 2002).

In summary, we have shown that ER-a36 expressingER-negative breast cancer cells retained mitogenicresponses to estrogen, suggesting that nongenomicestrogen signaling contributes to development andprogression of ER-negative breast cancer cells thatexpress ER-a36. Thus, ER-a36 is a novel player inmitogenic estrogen signaling, which may play importantroles in mammary tumorigenesis and in other types ofestrogen-related tumors as well.

Materials and methods

Chemicals and antibodiesThe 17b-estradiol (E2b) was purchased from Sigma Chemical(St Louis, MO, USA). The MEK1/2 inhibitor U0126, the Srcinhibitor PP2 and the PI3K inhibitor LY294002 were fromTocris Bioscience (Ellisville, MO, USA). The proteasome

Figure 5 EGFR signaling upregulates ER-a36 expression.(a) Western blot and reverse transcriptase PCR (RT-PCR) analysesof the effects of EGFR inhibitors on ER-a36 expression in MDA-MB-231 and MDA-MB-436 cells. (b) Schematic structures ofluciferase reporter plasmid driven by different 50 truncatedpromoters of ER-a36. The �736, �584, and �513 indicate residuesupstream of the transcription initiation site. An AP-1-binding siteis also indicated that was mutated in the pER36-mAP1 plasmid.HEK-293 cells were transfected with different reporter plasmidstogether with an empty expression vector or an expression vectorfor EGFR. The luciferase activities were assayed and normalizedusing a cytomegalovirus-driven Renilla luciferase plasmid. Col-umns represent the means of four independent experiments; barsrepresent the s.e. *Po0.05 for cells transfected with the EGFRexpression vector vs an empty expression vector. (c) HEK293 cellswere transfected with the pER36-736 reporter plasmid with theempty expression vector or the EGFR expression vector and thentreated with vehicle, 10 mM of LY294002, PP2, Gefitinib or AG1478for 24 h. The luciferase activities were then normalized andanalyzed. Results shown in graph are means from four experi-ments; bars, s.e. *Po0.05 for cells treated with vehicle vs underdifferent conditions.

Mitogenic estrogen signaling in ER-negative breast cancerXT Zhang et al

777

Oncogene

inhibitor MG132, and the EGFR inhibitors BiBx and AG1478were purchased from EMD Chemicals (Gibbstown, NJ, USA)and the EGFR inhibitor Gefitinib was from LC Laboratories(Woburn, MA, USA). Phospho-EGFR and -Src antibodies,EGFR, Src and Shc antibodies, anti-phospho-p44/42 ERK(Thr-202/Tyr-204) antibody and anti-p44/42 ERK antibodywere all purchased from Cell Signaling Technology (Boston,MA, USA). Polyclonal anti-ER-a36 antibody was generatedand characterized as described before (Wang et al., 2006).Antibodies of c-Myc, ER-b and cyclin D1 were from SantaCruz Biotechnology (Santa Cruz, CA, USA). ER-a66, HER-2/Neu and progesterone receptor antibodies were obtained fromVentana Medical Systems (Tucson, AZ, USA). ER-b siRNAand control siRNA-A were purchased from Santa CruzBiotechnology (Santa Cruz, CA, USA).

Specimen analysis and immunohistochemistryA total of 12 formalin-fixed paraffin-embedded tumor samplesof triple-negative breast carcinomas were retrieved from thecollection of Clinical Center of the University of Sarajevo afterapproval of the Institutional Review Board. Immunohisto-chemical assay for ER-a36, ER-a66, progesterone receptorand EGFR expression were performed using the commercially

available detection kits and automated staining procedures.Protein expression was scored according to the AmericanSociety of Clinical Oncology/College of American PathologistsGuideline Recommendations (Wolff et al., 2007).

Cell culture, treatment and growth assayMDA-MB-231, MDA-MB-436 cells and human embryonickidney cell line (HEK293) were obtained from American TypeCulture Collection (ATCC, Manassas, VA, USA). All parentaland derivative cells were maintained in Dulbecco0s modiedEagle0s medium and 10% fetal calf serum at 37 1C in a 5% CO2

incubator. For E2b treatment, cells were maintained in phenol-red-free media with 2.5% charcoal-stripped fetal calf serum(HyClone, Logan, UT, USA) for 3 days and then in serum-freemedium for 24 h before experimentation. To test the effects ofdifferent inhibitors, all inhibitors were added 10min before theE2b addition. For the MG-132 treatment, 10mM MG132 wasadded 12 h before cell harvest.To cell growth assays, cells were treated with indicated

concentrations of E2b or vehicle (ethanol) as a control. Thecells were seeded at 1� 104 cells per dish in 60mm dishes andthe cell numbers were determined using the ADAM automaticcell counter (NanoEn Tek Inc., Seoul, Korea) after 12 days.

Figure 6 ER-a36 interacts with the EGFR complex and mediates E2b-induced phosphorylation of Src and EGFR. (a, b) Co-immunoprecipitation and Western blot analyses of HA-ER-a36 and the EGFR complex in MDA-MB-231 (a) and MDA-MB-436 (b)cells. Cells transiently transfected with an expression of HA-tagged ER-a36 were lysed and the cell lysates were immunoprecipitatedwith pre-immune, anti-EGFR and anti-HA antibodies. The immunoprecipitates were blotted by anti-HA, anti-EGFR, anti-Shc andanti-Src antibodies. (c) Western blot analysis of the effects of E2b (1 nM) on the phosphorylation levels of EGFR-845 and Src-846 inMDA-MB-231 and MDA-MB-436 cells. Columns represent the means of three experiments; bars represent the s.e. *Po0.05 for cellstreated with vehicle vs cells treated for different time periods. (d) Western blot analysis of the effects of Src inhibitor PP2, EGFRinhibitor AG1478 and MAPK/ERK inhibitor U0126 on the E2b-stimulated phosphorylation of EGFR, Src and the MAP kinaseinhibitor UO126 in MDA-MB-231 and MDA-MB-436 cells. All experiments were repeated at least three times, and the representativeresults are shown.

Mitogenic estrogen signaling in ER-negative breast cancerXT Zhang et al

778

Oncogene

Five dishes were used for each treatment and experiments wererepeated more than three times.

Establishment of stable cell linesMDA-MB-231 and MDA-MB-436 cells with ER-a36 expres-sion knocked down by the shRNA method were established asdescribed before (Kang and Wang, 2010). MDA-MB-231 cellstransfected with the empty expression vector, a control vectorexpressing shRNA for luciferase or two different ER-a36-specific shRNA expression vector were named as 231/V, 231/lucSh, 231/Sh36(3-1) and 231/Sh36(1-7), respectively. ForMDA-MB-436 cells, cells transfected with the empty expres-sion vector or the ER-a36 shRNA expression vector wereselected for 3 weeks, and more than 20 clones of selected cellswere pooled and named as 436/V and 436/Sh36.

siRNA transfectionMDA-MB-231 cells were seeded at 2� 105 cells/dish in 60mmculture dishes 24 h before transfection. ER-b or control siRNAweighing 1mg was mixed with siRNA transfetion medium andsiRNA transfection reagent (Santa Cruz Biotechnology) andincubated for 30min at room temperature before added intocultured cells. The efficiency of siRNA knock-down wasassessed with western blot analysis.

RNA purification and reverse transcriptase–PCRTotal RNA was prepared with the ‘TRIzol’ RNA purificationreagent. Total RNA weighing 1 mg was reversely transcribedusing the ProtoScript II RT–PCR kit (New England Biolab,Ipswich, MA, USA). Reverse transcriptase PCR analysis ofER-a36, EGFR and b-actin was performed using gene specificprimers as the following. ER-a36—forward primer: 50-CAAGTGGTTTCCTCGTGTCTAAAG-30, reverse primer: 50-TGTTGAGTGTTGGTTGCCAGG-30; EGFR—forward primer:50-CGTCCGCAAGTGTAAGAA-30, reverse primer: 50-AGCAAAAACCCTGTGATT-30; b-actin—forward primer: 50-TGACGGGGTCACCCACACTGTGCCCATCTA-30, reverseprimer: 50-CTAGAAGCATTTGCGGTGGACGATGGAGGG-30. PCR procedure was carried out as described before (Zouet al., 2009). PCR products were analyzed by electrophoresis ina 1.5% agarose gel and visualized by ethidium bromidestaining under ultraviolet illumination.

DNA mutagenesisTo mutate the AP-1 consensus-binding site from 50-AGAGTCA-30 to 50-AGctTCA-30, the mutated primers ER-a36P-AP1m forward: 50-GCAGCCCGCGCTGCGTTCAGctTCAAGTTCTCTCGCCGGG-30 and reverse: 50-CCCGGCGAGAGAACTTGAagCTGAACGCAGCGCGGGCTGC-30 wereused. The mutagenesis was performed using the QuikChangeII site-directed mutagenesis kit (Stratagene, La Jolla, CA,USA) according to the manufacturer’s protocol. The mutationwas verified by DNA sequencing.

DNA transfection and luciferase assayHEK-293 cells were transfected using FuGene 6 transfectionreagent (Roche Applied Science, Indianapolis, IN, USA) withthe pER36-736-Luc, pER36-584-Luc, pER36-513-Luc orpER36-296-Luc reporter plasmids as described before (Zouet al., 2009) and an empty expression vector or the expressionvector for EGFR (a kind gift from Dr Laura Hansen atCreighton University). Cells were co-transfected with acytomegalovirus-driven Renilla luciferase plasmid, pRL-CMV (Promega, Madison, WI, USA) to establish transfectionefficiency. At 24 h after transfection, cells were treated with

vehicle, 10mM of U0126, PP2, or LY294002 for 24 h. At 48 hafter transfection, cell extracts were prepared and luciferaseactivities were determined and normalized using the Dual-Luciferase Assay System (Promega).

Immunoprecipitation and immunoblot analysisFor immunoprecipitation assays, cells were washed twice withice-cold phosphate buffered saline and lysed with the lysisbuffer (150mM NaCl, 20mM Tris–HCl, pH 7.4, 0.1% NP-40)supplemented with protease and phosphatase inhibitors(Sigma Chemical). Cell lysates were then incubated withindicated primary antibodies or pre-immune serum andimmunoprecipitated with protein A/G plus agarose. Theprecipitates were then washed, separated on SDS–polyacryl-amide gel electrophoresis and analyzed with western blotanalysis as described before (Kang and Wang, 2010).

Tumor formation in nude miceTumor formation was assayed using ovariectomized femalenude mice (5- to 6-week old, strain CDI nu/nu, Charles RiverLaboratories International, Wilmington, MA, USA). ER-negative breast cancer cells maintained in phenol-red-freemedia with 2.5% charcoal-stripped fetal calf serum for 3 dayswere washed with 0.025% edetate sodium (Versene Dow,Midland, MI, USA) and 0.05% trypsin in a Ca2þ - and Mg2þ -free PBS before inoculation. For MDA-MB-231 cells, a totalof 1� 106 cells for each clone were resuspended in 0.1ml ofPBS and inoculated subcutaneously into the mammary fat padof ovariectomized female nude mice 5 days after subcutaneousimplantation of 1.7mg/60-day release of E2 (treated; 12 mice)or placebo (control; 12 mice) pellets (Innovative Research ofAmerican, Sarasota, FL, USA). Tumor growth was monitoredby measuring two perpendicular diameters with verniercalipers. For MDA-MB-436 cells, 1� 106 cells for each cellline were resuspended in 0.1ml Matrixgel Basement Mem-brane Matrix (BD Biosciences, San Jose, CA, USA) andinoculated into the fat pad of ovariectomized female nudemice. The estrogen and placebo pellets were replaced after 60days. Tumor growth was measured weekly. Tumor volumebased on caliper measurements were calculated by the formula:tumor volume¼ 1/2(length�width2).

Statistical analysisData were summarized as the mean±s.e. using the GraphPadInStat software program (GraphPad Software, La Jolla, CA,USA). Tukey–Kramer multiple comparisons test was alsoused, and the significance was accepted for Po0.05.

Conflict of interest

The authors declare no conflict of interest.

Acknowledgements

This work was supported by National Institute of HealthGrant DK84328 and by the Nebraska Tobacco SettlementBiomedical Research Program Award (LB-595 and LB692) toZY Wang. Dr Semir Vranic was a research fellow at CreightonUniversity Medical School, Omaha, NE, USA and had beensupported by a UICC American Cancer Society BeginningInvestigators Fellowship (ACSBI) (ACS/08/004) funded by theAmerican Cancer Society.

Mitogenic estrogen signaling in ER-negative breast cancerXT Zhang et al

779

Oncogene

References

Behrens D, Gill JH, Fichtner I. (2007). Loss of tumourigenicity ofstably ERb-transfected MCF-7 breast cancer cells. Mole Cell

Endocri 274: 19–29.Biscardi JS, Maa MC, Tice DA, Cox ME, Leu TH, Parsons SJ. (1999).

c-Src-mediated phosphorylation of the epidermal growth factorreceptor on Tyr845 and Tyr1101 is associated with modulation ofreceptor function. J Biol Chem 274: 8335–8343.

Early Breast Cancer Trialists’ Collaborative Group (1992). Systemictreatment of early breast cancer by hormonal, cytotoxic, or immunetherapy. 133 randomised trials involving 31 000 recurrences and24 000 deaths among 75 000 women. Lancet 339: 1–15.

Friedl A, Gottardis MM, Pink J, Buchler DA, Jordan VC. (1989).Enhanced growth of an estrogen receptor-negative endometrialadenocarcinoma by estradiol in athymic mice. Cancer Res 49:4758–4764.

Friedl A, Jordan VC. (1994). Oestradiol stimulates growth ofoestrogen receptor-negative MDA-MB-231 breast cancer cells inimmunodeficient mice by reducing cell loss. Eur J Cancer 30A:1559–1564.

Geisler J, Haynes B, Anker G, Dowsett M, Lonning PE. (2002).Influence of letrozole and anastrozole on total body aromatizationand plasma estrogen levels in postmenopausal breast cancer patientsevaluated in a randomized, cross-over study. J Clin Oncol 20:751–757.

Howell A, Osborne CK, Morris C, Wakeling AE. (2000). ICI 182 780(Faslodex): development of a novel, ‘pure’ antiestrogen. Cancer 89:817–825.

Kang L, Wang ZY. (2010). Breast cancer cell growth inhibition byphenethyl isothiocyanate is associated with downregulation ofestrogen receptor-alpha36. J Cell Mol Med 14: 1485–1493.

Lacroix M, Toillon RA, Leclercq G. (2004). Stable ‘portrait’ of breasttumors during progression: data from biology, pathology andgenetics. Endocr Relat Cancer 11: 497–522.

Lazennec G, Bresson D, Lucas A, Chauveau C, Vignon F. (2001). ERbinhibits proliferation and invasion of breast cancer cells. Endocr 142:4120–4130.

Lee LM, Cao J, Deng H, Chen P, Gatalica Z, Wang ZY. (2008). ER-alpha36, a novel variant of ER-alpha, is expressed in ER-positiveand -negative human breast carcinomas. Anticancer Res 28:479–483.

Lin SL, Yan LY, Zhang XT, Yuan J, Li M, Qiao J et al. (2010). ER-alpha36, a variant of ER-alpha, promotes tamoxifen agonist actionin endometrial cancer cells via the MAPK/ERK and PI3K/Aktpathways. PLoS One 5: e9013.

Mahfoudi A, Roulet E, Dauvois S, Parker MG, Wahli W. (1995).Specific mutations in the estrogen receptor change the propertiesof antiestrogens to full agonists. Proc Natl Acad Sci USA 92:4206–4210.

Narod SA. (2001). Hormonal prevention of hereditary breast cancer.Ann NY Acad Sci 952: 36–43.

Nissen-Meyer R. (1964). ‘Prophylactic’ ovariectomy and ovarianirradiation in breast cancer. Acta Unio Int Contra Cancrum 20:527–530.

Rai D, Frolova A, Frasor J, Carpenter AE, Katzenellenbogen BS.(2005). Distinctive actions of membrane-targeted versus nuclearlocalized estrogen receptors in breast cancer cells. Mol Endocrinol

19: 1606–1617.Rebbeck TR, Levin AM, Eisen A, Snyder C, Watson P, Cannon-

Albright L et al. (1999). Breast cancer risk after bilateralprophylactic oophorectomy in BRCA1 mutation carriers. J Natl

Cancer Inst 91: 1475–1479.Shi L, Dong B, Li Z, Lu Y, Ouyang T, Li J et al. (2009). Expression of

ER-{alpha}36, a novel variant of estrogen receptor {alpha}, andresistance to tamoxifen treatment in breast cancer. J Clin Oncol 27:3423–3429.

Simpson PT, Reis-Filho JS, Gale T, Lakhani SR. (2005). Molecularevolution of breast cancer. J Pathol 205: 248–254.

Paruthiyil S, Parmar H, Kerekatte V, Cunha GR, Firestone GL,Leitman DC. (2004). Estrogen receptor b inhibits human breastcancer cell proliferation and tumor formation by causing a G2 cellcycle arrest. Cancer Res 64: 423–428.

Tong D, Schuster E, Seifert M, Czerwenka K, Leodolter S, ZeillingerR. (2002). Expression of estrogen receptor beta isoforms in humanbreast cancer tissues and cell lines. Breast Cancer Res and Treat 71:249–255.

Tsai EM, Wang SC, Lee JN, Hung MC. (2001). Akt activation byestrogen in estrogen receptor-negative breast cancer cells. Cancer

Res 61: 8390–8392.Wang Z, Zhang X, Shen P, Loggie BW, Chang Y, Deuel TF. (2005).

Identification, cloning, and expression of human estrogen receptor-alpha36, a novel variant of human estrogen receptor-alpha66.Biochem Biophys Res Commun 336: 1023–1027.

Wang Z, Zhang X, Shen P, Loggie BW, Chang Y, Deuel TF.(2006). A variant of estrogen receptor-{alpha}, hER-{alpha}36:transduction of estrogen- and antiestrogen-dependent membrane-initiated mitogenic signaling. Proc Natl Acad Sci USA 103:9063–9068.

Wolff AC, Hammond ME, Schwartz JN, Hagerty KL, Allred DC,Cote RJ, et al., American Society of Clinical Oncology, College ofAmerican Pathologists. (2007). American Society of ClinicalOncology/College of American Pathologists guideline recommenda-tions for human epidermal growth factor receptor 2 testing in breastcancer. J Clin Oncol 25: 118–145.

Zou Y, Ding L, Coleman M, Wang Z. (2009). Estrogen receptor-alpha(ER-alpha) suppresses expression of its variant ER-alpha 36. FEBS

Lett 583: 1368–1374.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Mitogenic estrogen signaling in ER-negative breast cancerXT Zhang et al

780

Oncogene