Loss of Tgfb Receptor Type 2 Expression Impairs Estrogen Response and Confers Tamoxifen Resistance Susann Busch1, Andrew H

Cancer Tumor and Stem Cell Biology Research

Loss of TGFb Receptor Type 2 Expression Impairs Estrogen Response and Confers Tamoxifen Resistance Susann Busch1, Andrew H. Sims2, Olle Sta l3,Ma rten Ferno€4, and Goran€ Landberg1,5

Abstract

One third of the patients with estrogen receptor a (ERa)- tamoxifen resistance. Functional investigations conﬁrmed that positive breast cancer who are treated with the antiestrogen cell cycle or apoptosis responses to estrogen or tamoxifen in tamoxifen will either not respond to initial therapy or will ERa-positive breast cancer cells were impaired by TGFBR2 develop drug resistance. Endocrine response involves crosstalk silencing, as was ERa phosphorylation, tamoxifen-induced between ERa and TGFb signaling, such that tamoxifen non- transcriptional activation of TGFb, and upregulation of the responsiveness or resistance in breast cancer might involve multidrug resistance protein ABCG2. Acquisition of low aberrant TGFb signaling. In this study, we analyzed TGFb TGFBR2 expression as a contributing factor to endocrine resis- receptor type 2 (TGFBR2) expression and correlated it with tance was validated prospectively in a tamoxifen-resistant cell ERa status and phosphorylation in a cohort of 564 patients line generated by long-term drug treatment. Collectively, our who had been randomized to tamoxifen or no-adjuvant treat- results established a central contribution of TGFb signaling in ment for invasive breast carcinoma. We also evaluated an endocrineresistanceinbreastcancerandofferedevidencethat additional four independent genetic datasets in invasive breast TGFBR2 can serve as an independent biomarker to predict cancer. In all the cohorts we analyzed, we documented an treatment outcomes in ERa-positive forms of this disease. association of low TGFBR2 protein and mRNA expression with Cancer Res; 75(7); 1457–69. 2015 AACR.

Introduction resistance and the identiﬁcation of new therapeutic targets have therefore become increasingly relevant (1–3). The anti-estrogen tamoxifen is the most widely used adjuvant TGFb is a pleiotropic cytokine with potent antimitogenic and endocrine therapy for patients with estrogen receptor a (ERa)- proapoptotic effects (4). Because of its dual role, TGFb is also a positive breast cancer. However, one third of tamoxifen-treated known regulator of cell differentiation, motility, and invasion patients will have a disease recurrence within 15 years as they (5–9). Consequently in cancer, cellular functions mediated by do not respond to initial therapy or acquire drug resistance. TGFb are complex, as blockade and induction of TGFb signal- Deregulation of estrogen signaling is thought to be a common ing in mouse models has been associated with tumor progres- mechanism for endocrine resistance. The activation of escape sion (10–12). pathways provides tumor cells with alternative proliferative or TGFb receptor type 2 (TGFBR2) is the sole ligand-binding survival stimuli including alterations in ERa itself, ERa core- receptor for members of the TGFb family comprising TGFb1, gulatory proteins, cell-cycle regulators, receptor tyrosine kinase -2, and -3. Ligand-induced cell response is mediated through signaling, cell survival, or apoptosis. The discovery of new either canonical, SMAD-dependent, or other noncanonical, biomarkers that classify patient subgroups with potential drug SMAD-independent signaling pathways such as through JNK, Akt, Src, p42/44 (ERK1/2), and p38 MAPK (13). High TGFBR2 expression is a poor prognostic indicator for 1Sahlgrenska Cancer Center, Gothenburg University, Gothenburg, overall survival in ERa-negative breast cancer (14). A TGFb Sweden. 2Applied Bioinformatics of Cancer, University of Edinburgh, response signature using human epithelial cell lines was linked 3 Cancer Research UK Centre, United Kingdom. Department of Clinical to lung metastases in ERa-negative breast cancer, whereas there and Experimental Medicine, Institution of Surgery and Clinical Oncol- ogy, Linkopings€ Universitet, Linkoping,€ Sweden. 4Department of was no correlation with distant metastases for ERa-positive breast Oncology, Clinical Sciences, Lund University, Lund, Sweden. 5Molec- cancer (15). Thus, cells may escape the antitumorigenic effects of ular Pathology, Breakthrough Breast Cancer Research Unit, University TGFb depending on the ERa status. of Manchester, United Kingdom. ERa and TGFb pathways intersect at multiple cross-points Note: Supplementary data for this article are available at Cancer Research including direct protein–protein interactions. ERa blocks TGFb Online (http://cancerres.aacrjournals.org/). signaling in a nongenomic manner by promoting SMAD2/3 Corresponding Author: Goran€ Landberg, Sahlgrenska Cancer Center, Box 425, degradation (16) and activation of MAPK through GPR30 Gothenberg 405 30, Sweden. Phone: 46-31-786-6736; Fax: 46-31-827-194; (17). Furthermore, it has been demonstrated that SMAD3 and E-mail: [email protected] SMAD4 act as coactivators and corepressors, respectively, for ERa- doi: 10.1158/0008-5472.CAN-14-1583 induced gene expression (18, 19). Antiestrogen–induced growth 2015 American Association for Cancer Research. inhibition in breast cancer cells was shown to be mediated

www.aacrjournals.org 1457

through sequential activation of p38 MAPK and TGFb pathway mal cut-off point. Kaplan Meier plots were generated and log (20). Stromal TGFb1 expression stimulated by tamoxifen has rank statistics were calculated using the survival package in R. been reported to contribute to the growth-inhibitory effects of endocrine treatment (21). However, the crosstalk between ERa Cell culture and TGFb pathways in breast cancer is still relatively poorly HEK293T, MCF7, and T47D were cultured in DMEM þ 10% understood and its elucidation may improve our knowledge of FBS in a humid chamber with 5% CO at 37C. For estrogen the acquisition of resistance mechanisms. 2 response studies, breast cancer cells were hormone-starved in In this study, we demonstrate for the first time that loss of phenol red–free media supplemented with 5% charcoal-stripped TGFBR2 expression is a predictor for tamoxifen resistance in ERa- serum (CSS) for 24 hours followed by a 24-hour culture in 1% positive breast cancer. Experimentally, we show that ERa-positive CSS-DMEM before stimulation with either 17-b-estradiol (E2; breast cancer cell lines with shRNA-induced TGFBR2-specific Sigma), 4-hydroxy-tamoxifen (4OHT; Sigma), fulvestrant (Sig- knockdown display altered cell signaling and an impaired ma), recombinant human TGFb1 (rhTGFb1; R&D Systems), response upon estrogen and tamoxifen treatment. In addition, TGFBR1 inhibitor SB431542 (Sigma), or a combination thereof TGFBR2 knockdown abrogates tamoxifen-induced ERa phos- in 1% CSS media. For CSS preparation, 1% (w/v) dextran-coated phorylation and TGFb transcriptional activity and results in charcoal (Sigma) was added to FBS and heated to 55C for 30 elevated levels of multidrug resistance protein ATP-binding cas- minutes in water bath. Suspension was centrifuged at 2,000 g sette subfamily G member 2 (ABCG2). In line, analysis of a well- for 15 minutes before sterilizing it using a 0.22 mm filter. Origin of established long-term treated, tamoxifen-resistant cell line all used cell lines has been confirmed through institutional revealed decreased TGFBR2 expression and deregulated TGFb authentication process, performed by Molecular Biology Core transcriptional activation further highlighting the role of TGFb Facility using a multiplex PCR assay (Applied Biosystems Ampflstr signaling in endocrine response. system). MCF7-derivative, tamoxifen-resistant cell line (TAM-R) was a kind gift of Robert B. Clarke (University of Manchester, Manche- Patients and Methods ster, United Kingdom). Generation of TAM-R has been described Patient and tumor samples previously (27). In brief, MCF7 has been cultured in phenol red– Invasive breast cancer cohort includes 564 premenopausal free RPMI þ 5% CSS þ 0.1 mmol/L 4OHT over 6 months to obtain patients enrolled in a trial from 1986 to 1991 (SBII:2) and TAM-R. For this study, TAM-R were cultured in phenol red–free randomized to either 2 years of adjuvant tamoxifen treatment DMEM þ 5% CSS þ 0.1 mmol/L 4OHT. For epigenetic studies, 0 (n ¼ 276) or no systemic treatment (n ¼ 288). All patients were cells were treated with 0.5 mmol/L 5-Aza-2 -deoxycytidine (Sig- followed up for recurrence-free survival. Recurrence (events) was ma) or 0.1 mmol/L Trichostatin A (Sigma) using DMSO as vehicle defined as local, regional, or distant recurrence and breast cancer– control. specific death, whereas contralateral breast cancer was excluded. Each patient underwent surgery (either modified radical mastec- Lentiviral knockdown tomy or breast conserving surgery) followed by radiotherapy and Transfections of breast cancer cells, MCF7 and T47D, were in a small number of cases adjuvant polychemotherapy (less than performed using lentiviral approach as described by Weinberg 2%). The median postsurgery follow-up time without a breast laboratory (28). Plasmids were a kind gift of Dr. Akira Orimo; cancer event was 13.9 years. Study was approved by Lund and pCMV-VSV-G (Addgene: Plasmid 8454), pCMV-dR8.2dpvr Linkoping€ ethical committee and informed consent was obtained (Addgene: Plasmid 8455), pLKO.1-shRNA-hygro (Addgene: Plas- from all patients participating. Further details of the trial have 0 mid 24150), GFP-shRNA (5 -GCAAGCTGACCCTGAAGTTCA- been previously described (22, 23). 0 0 0 3 ), TGFBR2-shRNA.1 (5 -GATTCAAGAGTATTCTCACTT-3 ), and Representative tumor areas of formalin-fixed and paraffin- 0 0 TGFBR2-shRNA.2 (5 -GAATGACGAGAACATAACACT-3 ). Brief- embedded tissue material were selected for tissue microarray ly, HEK293T cells were transfected with lentiviral plasmid system (TMA) construction. Details regarding TMA assembling and stain- using Fugene 6 (Roche/Promega) over night. Next day, medium ing procedure validity of used antibodies have been reported (23). was replaced and after 24 hours of virus production, supernatant TMA stained for TGFBR2 (Abcam) and phosphorylated SMAD2 was collected, filtered, and supplemented with 5 mg/mL prot- (pSMAD2; Cell Signaling Technology) as previously described amine sulfate (Sigma) before infection of target cells overnight. have been used for this study analyzing the tumor compartment Media was replaced and 24 hours later, breast cancer cells were (24). Scoring of tumor samples was performed under supervision selected in 200 mg/mL Hygromycin (Clontech) containing media. of a pathologist (G. Landberg) without knowledge of pathologic Plasmid pCMV6-XL5 (Origene) containing TGFBR2 cDNA and clinical data. The scoring accounts for intensity of immunos- (accession number NM_001024847) was used for transient, tain-positive tumor cells. None of the examined tumor samples constitutive overexpression of TGFBR2. Control "empty" vector displayed complete absence of TGFBR2 staining. Staining inten- was obtained through NotI digestion to remove TGFBR2-encod- sities of tumor cells were grouped into four categories ranging ing cDNA. Transfected cells were analyzed for RNA expression 24 from weak (score 0), weakly intermediate (score 1), intermediate hours after transfection or subjected to immunohistochemical (score 2) to strong (score 3). staining 48 hours after transfection. Knockdown and overexpression of TGFBR2 was confirmed by Gene expression data quantitative PCR and immunohistochemical staining with Gene expression datasets for tamoxifen-treated patients were TGFBR2-specific antibody (Abcam, 1:50 dilution) of formalin- downloaded from NCBI GEO as previously described (25). The fixed and paraffin-embedded cytospins. Transfected cells were X-Tile software program (26) was used to determine the opti- further analyzed for RNA and protein expression, protein

1458 Cancer Res; 75(7) April 1, 2015 Cancer Research

phosphorylation, viability, cell cycle, apoptosis, and transcrip- CSS for 96 hours (MCF7) or 8 days (T47D). In case of T47D, tional activity. media were replaced after 96 hours.

Quantitative PCR Cell-cycle analysis Following lentiviral transfection, a fraction of cells was For determination of DNA profile, 4 105 cells per well were pelleted and subjected to RNA isolation (Qiagen). RNA was seeded into 6-well plate. Cells were hormone-starved for 24 hours transcribed using TaqMan Reverse Transcriptase Kit (Applied in 5% CSS before 24-hour treatment in 1% CSS. After treatment, Biosystems) and resulting cDNA was used for real-time quan- cells were harvested with phenol red–free trypsin solution and titative PCR (Applied Biosystems 7900) using Sybr green (Bio- subsequently neutralized with phenol red–free 10% CSS media line) and TGFBR2-specific primers (forward: 50-ACGTGTTGA- and centrifuged at 1,800 rpm at 4 C for 5 minutes. Supernatant GAGATCGAGG-30; reverse: 50-CCCAGCACTCAGTCAACGTC- was removed and cells resuspended in 1 mL ice-cold 70% ethanol 30)andGAPDH-specific primers [forward: GGTCGGAGT- and put in freezer for at least 20 minutes. Afterwards, 5 mL PBS CAACGGATTT-30;reverse50-TGATGGCAACAATATCCACTT-30; were added to cells before centrifugation at 2,000 rpm at 4 C for primer sequences have previously been described (29)]. 5 minutes. Supernatant was removed and 800 mL of Vindelov€ solution (3.5 mmol/L Tris-base pH 7.6, 10 mmol/L NaCl, 0.1% Nonidet P40, 20 ng/mL RNase and 0.005% propidium iodide) Western blot analysis was added. Cells were incubated on ice for 20 minutes before flow After treatment, MCF7 were scraped off in cell lysis buffer (25 cytometry analysis (FACS Calibur BD). mmol/L HEPES, 5 mmol/L EDTA, 30 mmol/L NaPP, 50 mmol/L NaCl, 50 mmol/L NaF, 1% Triton-X, 10% glycerol, pH 7.4) supplemented with protease and phosphatase inhibitor cocktail Apoptosis assay (Roche). Cells were spun down and lysate collected and protein Assessment of cell death was performed using Annexin 5 concentration was determined using BCA assay (Pierce). Samples V-FITC Apoptosis Detection Kit (Abcam). Therefore, 4 10 were denatured in Laemmli buffer (250 mmol/L Tris-HCl pH 6.8, cells per well were seeded into 6-well plate. Cells were hor- 40% glycerol, 8% SDS. 0.01% bromphenol blue, 20% b-mercap- mone-starved for 24 hours in 5% CSS before 24 hours treat- toethanol) and run on 10% SDS-polyacrylamide gel and trans- ment in 1% CSS. After treatment, cells were harvested with ferred onto nitrocellulose membrane (Amersham). Membranes phenol red–free trypsin solution and subsequently neutralized were blocked with either 5% BSA or 5% milk in TBS-T buffer. with phenol red–free 10% CSS media and centrifuged at 1,800 Membranes were incubated with primary antibodies in 1:1,000 rpm for 5 minutes (including cells from culture media). Super- dilution in 3% BSA TBS-T supplemented with 2% blocking natant was removed and cells were washed once in PBS. Cells reagent (Roche): rabbit anti-phospho-SMAD2 (Cell Signaling were resuspended in 0.5 mL binding buffer containing 5 mLof Technology), mouse anti-SMAD2/3 (BD Biosciences), mouse Annexin V-FITC conjugated antibody and 5 mLpropidium anti-phospho (Serine-118) ERa (Cell Signaling Technology), iodide solution for 5 minutes at room temperature before flow rabbit anti-ERa (Thermo Fisher), mouse anti-Cyclin D1 (Dako), cytometry analysis (FACS Calibur BD). rabbit anti-phospho-ERK1/2 (phospho-p44/42 MAPK; Cell Sig- naling Technology), rabbit anti-ERK1/2 (p44/42 MAPK; Cell Reporter assay Signaling Technology), rabbit anti-ABCG2 (Cell Signaling Tech- A total of 4 105 cells per well were seeded into 6-well plate. nology), rabbit anti-tubulin (Cell Signaling Technology) and Following day, cells were cotransfected with 1 mg p(CAGA)12- incubated with secondary horseradish peroxidase-linked anti- MLP-Luc vector (kind gift of Nullin Divecha, University of South- body in 1:5,000 dilution: sheep anti-mouse (GE Healthcare), ampton, Southampton, United Kingdom) or 1 mg pGL4.27-ERE goat anti-rabbit (Dako). Chemiluminescence was detected using (3)-GLuc (Gaussia luciferase; kind gift of Robert B. Clarke) Luminata Forte (Millipore) on X-ray films (Amersham). and 0.1 mg pCMV-RL-TK (Renilla luciferase; kind gift of Robert B. Clarke) using X-treme HP DNA transfectant reagent (Roche). Viability assay Transfection was performed according to manufacturers' instruc- 4 Cell number was determined using Alamar blue (Invitrogen). tions. Next day, cells were trypsinized and 2 10 cells per well Therefore, 20 mL of 20% Alamar blue solution were added per well reseeded into 96-well plate. Cells were cultured for 24 hours in 5% of 96-well plates and after 1 hour incubation fluorescence was CSS DMEM before 24-hour treatment in 1% CSS DMEM. Lucif- read (excitation: 544 nm, emission: 590 nm; BMG Omega Fluos- erase activity was measured using Dual-Glo Luciferase Assay tar). For growth curve analysis, 1,000 cells per well were seeded in System (Promega) according to manual. Luminescence was 96-well plates in 10% FBS-containing media. Viability was mea- detected with microplate reader (BMG Omega Fluostar). sured next day as reference point, media were replaced and for the following 5 days viability was assessed every 24 hours. For TGFb Immunofluorescence inhibition, 10 mmol/L SB431542 was added to the media (DMSO Lentivirally transfected MCF7 cells were seeded in 4-chamber as vehicle control). For estrogen withdrawal analysis, cells were slides (4,000/well). Cells were cultured for 24 hours in 5% CSS hormone-starved for 24 hours in 5% CSS before culture in 1% DMEM followed by 24-hour culture in 1% CSS DMEM. Cells were CSS. For estrogen and tamoxifen/fulvestrant response studies, treated in 1% CSS DMEM for 24 hours for Ki-67 staining and 96 5,000 cells per well were seeded in 96-well plates, hormone- hours for cleaved PARP staining. After treatment, cells were fixed starved for 24 hours in 5% CSS before incubation with different with 4% formalin. Following PBS washing, cells were permeabi- concentrations of E2 (0.1–10 nmol/L) or 1 nmol/L E2 with lized with 0.25% Triton-X100 and blocked with 1% BSA, then increasing concentrations of 4OHT or fulvestrant (1 nmol/L up washed again before incubation with primary antibodies in 1:100 to 10 mmol/L) in absence or presence of 1 ng/mL rhTGFb1in1% dilution overnight at 4C: mouse anti-Ki-67 (Dako) and rabbit

www.aacrjournals.org Cancer Res; 75(7) April 1, 2015 1459

anti-cleaved PARP (Cell Signaling Technology). Cells were (Mann–Whitney U test; P < 0.001; Table 1). Prognostic recur- washed again and then incubated with secondary antibody in rence-free survival analysis showed no difference between 1:500 dilution for 1 hour at room temperature or at 4C overnight: patients with low or high TGFBR2 expression in the untreated goat anti-mouse AlexaFluor555 (Invitrogen), goat anti-rabbit control group (Fig. 1B; Supplementary Table S1). However, AlexaFluor555 (Invitrogen). After further washing, cells were stratifying ERa-positive patients according to treatment revealed mounted with DAPI-containing media (Sigma) and fluorescence a significantly shortened recurrence-free survival for tamoxifen- was monitored with an inverted light microscope (Olympus treated patients with low TGFBR2 expression (Univariate Cox CKX41) using the EXFO X-Cite illumination system and SPOT regression; HR: 0.312, 95% CI, 0.131–0.742; P ¼ 0.008; Fig. 1C). Imaging Solutions Microscopy Software. A multivariate interaction analysis validated TGFBR2 as a treatment-predictive biomarker independent of known prognostic Custom transcript array markers (Multivariate Cox regression, HR, 0.267; 95% CI, 2 – ¼ Customized RT Profiler PCR Array (SABiosciences) was 0.098 0.731; P 0.010; Table 2). a employed to analyze mRNA expression levels in lentivirally Our group previously reported that ER phosphorylation at a transfected cells in the presence of 1 nmol/L E2 with or without serine-118 (pS118-ER ) is a predictor for tamoxifen sensitivity 0.1 mmol/L 4OHT. A total of 3 105 cells were seeded in a 6-cm (30) and ERK1/2 phosphorylation (pERK) as a marker for dish, next day media were changed to 5% CSS phenol red–free tamoxifen resistance (31). We found that TGFBR2 expression fi a < DMEM. Following day, media were changed to 1% CSS phenol was signi cantly correlated to pS118-ER (Spearman, P a red–free DMEM. Cells were treated for 24 hours in 1% CSS phenol 0.001), but not to pERK (Table 1). S118-ER phosphorylation a red–free DMEM before cell lysis and RNA extraction using RNeasy hasbeenassociatedwithER transactivation in numerous – a Mini Kit (Qiagen). Four hundred nanograms of RNA was sub- studies (32 34) and we noted that cyclin D1, an ER target 2 fi < jected to cDNA synthesis using RT First Strand Kit (SABios- gene, was also signi cantly linked to TGFBR2 (Spearman, P ciences) before PCR run using RT2 SYBR Green qPCR Mastermix 0.001; Table 1). (SABiosciences). PCR reactions and cycling conditions were set up Analyzing four independent publically available gene expres- according to manufacturer's instructions. PCR was run using a sion datasets of tamoxifen-treated breast cancer, we noted signif- real-time PCR thermal cycler (Applied Biosystems 7900). Analysis icantly worse recurrence-free survival in all patient groups with 2 < was performed using SABiosciences' web-based RT Profiler PCR low TGFBR2 expression (log rank, P 0.050; Fig. 1D) supporting fi array data analysis tool (version 3.5). our ndings. To unravel the clinical relevance of canonical TGFb signaling in tamoxifen resistance, we examined pSMAD2 in accordance to the Statistical analyses TGFBR2 study. Images representing each pSMAD2 staining cat- Spearman's rank order correlation coefficient, Kruskal– egory are shown in Supplementary Fig. S2A. In contrast to Wallis and Mann–Whitney U test were performed for evalua- TGFBR2, pSMAD2 was not associated with ERa and PR status tion of clinicopathologic and molecular parameters. The (Supplementary Table S2) or tamoxifen resistance (Supplemen- Kaplan–Meier method was used to estimate recurrence- tary Fig. S2B) but had prognostic qualities (Supplementary Fig. free survival and univariate Cox regression was used to compare S2C), which was significant in multivariate Cox regression anal- recurrence-free survival between different treatment groups or ysis (HR: 2.663; 95% CI, 1.201–5.906; P ¼ 0.016; Supplementary according to marker categories. Cox proportional hazards Table S3). Therefore, TGFBR2 and pSMAD2 as biomarkers may regression was used for relative risk estimation in multivariate possess distinctive clinical significance. analysis. Covariates used for Cox regression included tumor grade,tumorsize,lymphnodestatus,age,Ki-67,ERa status, Targeting TGFb pathway in vitro and treatment interaction. All P values corresponded to two- To further investigate the role of TGFBR2 in tamoxifen response sided tests and P values less than 0.05 were considered statis- in ERa-positive breast cancer cell lines, we utilized two indepen- tically significant. Experimental data are represented as mean dent TGFBR2-specific shRNA constructs (TGFBR2-shRNA.1 and of three to six independent experiments (unless otherwise -shRNA.2) and a pharmacologic inhibitor (SB431542). specified) and error bars indicate SEM and statistical difference Knockdown efficiency was confirmed by quantitative PCR was assessed using two-sided Student t test. with about 75% to 80% decrease in TGFBR2 mRNA expression in MCF7 and about 40% to 50% reduction in T47D (Fig. 2A). TGFBR2 knockdown in MCF7 was confirmed on protein level Results using immunohistochemical staining (Fig. 2B). Recombinant Low TGFBR2 expression is associated with tamoxifen resistance human TGFb1(rhTGFb1) reduced cell growth, an effect that in invasive breast carcinoma was diminished in the TGFBR2 knockdown cells (Supplemen- To delineate the clinical value of the TGFb pathway in breast tary Fig. S3A) signifying loss of responsiveness to TGFb. cancer, we assessed the level of TGFBR2 expression and SMAD2 Although displaying reduced viability in conventional cell phosphorylation (pSMAD2) by immunohistochemical staining culture media (Supplementary Fig. S3B), hormone withdrawal of archived tissue sections of primary tumors using a cohort of experiments revealed that TGFBR2 knockdown cells grow at a 564 premenopausal patients enrolled in a randomized tamoxifen higher rate (Supplementary Fig. S3C) through increased pro- trial (SBII:2), for study design see Supplementary Fig. S1. liferation (cells in S-phase; Fig. 2C) and a lower overall cell Representative images of each staining category in regard to death (apoptotic and necrotic cells; Fig. 2D), although the latter TGFBR2 expression levels are depicted in Fig. 1A. Statistical was not significant. analyses to assess correlations to clinicopathologic markers The chemical compound SB431542 is a potent and selective revealed that TGFBR2 is significantly linked to ERa and PR status inhibitor of TGFb receptor type-1 (TGFBR1) and acts as a

1460 Cancer Res; 75(7) April 1, 2015 Cancer Research

Figure 1. TGFBR2 expression and its associations to survival outcome in invasive breast cancer. A, immunohistochemical staining of invasive breast carcinomawithin randomized tamoxifen trial cohort depicting representative images for each category according to TGFBR2 antibody–specific staining intensity. Images were taken using Digital Imaging Hub at 20 magnification (scale bar, 50 mm). B, Kaplan–Meier plot illustrating recurrence-free survival regarding TGFBR2 expression in all untreated patients. C, Kaplan–Meier plots portraying recurrence-free survival with regard to TGFBR2 expression and stratified according to treatment arms in ERa-positive patients. (P value, univariate Cox regression; HR, hazard ratio; CI, confidence interval). D, Kaplan–Meier plots for four independent publically available data sets representing recurrence-free survival regarding TGFBR2 gene expression in tamoxifen-treated, ERa-positive patients (P value, log rank).

competitive ATP-binding site kinase inhibitor preventing phos- increase (Fig. 2F), whereas estrogen-mediated effect on apo- phorylation of the downstream targets, SMAD2 and SMAD3 ptosis was unaffected (Fig. 2G). (35). Phosphorylation of SMAD2 (pSMAD2) indicates stimula- Similar observations in regard to estrogen sensitivity using a tion of the canonical TGFb pathway and is therefore a useful TGFBR1 inhibitor (Supplementary Fig. S4B) suggest that manip- readout to monitor activation of TGFb signaling. In MCF7, ulation of the TGFb pathway alters estrogen response, indicative rhTGFb–induced SMAD2 phosphorylation was completely abol- of a TGFb-ERa signaling crosstalk. ished by SB431542 for all treatments (Supplementary Fig. S4A). Transcriptional activity of ERa and TGFb pathway are monitored using ERE-dependent and SMAD-dependent (CAGA TGFBR2 knockdown attenuates estrogen-induced proliferation promoter) luciferase-expressing plasmids, respectively (Supple- Concentration-dependent, estrogen-induced increase in cell mentary Fig. S5A and S5B). Strikingly, basal ERa transcriptional viability was reduced in TGFBR2 knockdown cells (Fig. 2E). activity was elevated in TGFBR2 knockdown cells (Supplemen- The differential estrogen response was even more pronounced tary Fig. S5C), hinting that lack of TGFBR2 results in a loss in the presence of rhTGFb1 (Fig. 2E, right). Flow cytometric cell- of TGFb-mediated ERa inhibition, which may explain the cycle analysis revealed a 5-fold increase in the number of cells observed increased proliferation in hormone deprivation con- in S-phase upon estrogen treatment, which was signiﬁcantly ditions. Estrogen-induced classical ERa transcriptional activity, attenuated in TGFBR2 knockdown cells exhibiting a 2-fold however, was unchanged (Supplementary Fig. S5D).

www.aacrjournals.org Cancer Res; 75(7) April 1, 2015 1461

Table 1. Correlations of TGFBR2 expression to clinicopathologic parameters and molecular markers TGFBR2 01 2 3 Total: 281 n ¼ 7 (2.5) n ¼ 43 (15.3) n ¼ 124 (44.1) n ¼ 107 (38.2) P Tumor size 0.346a 20 mm 0 (0) 17 (39.5) 52 (20.2) 32 (30.2) >20 mm 7 (100) 26 (60.5) 72 (79.8) 74 (69.8) Missing: 1 Tumor type 0.263b Ductal 6 (86.8) 37 (86.0) 102 (85.7) 96 (92.8) Lobular 1 (14.2) 2 (4.7) 6 (5.0) 2 (1.9) Medullary 0 (0) 4 (9.3) 11 (9.2) 6 (5.8) Missing: 8 Lymph node status 0.815a Negative 2 (28.6) 10 (23.3) 38 (30.6) 30 (28.0) Positive 5 (71.4) 33 (76.7) 86 (69.4) 77 (72.0) Missing: 0 Grade (NHG) 0.112c I 0 (0) 3 (7.0) 11 (9.4) 12 (11.7) II 4 (57.1) 12 (27.9) 46 (39.3) 42 (40.8) III 3 (42.9) 28 (65.1) 60 (51.3) 49 (47.6) Missing: 11 Ki-67 0.259a 25% 4 (66.7) 21 (52.5) 82 (70.1) 66 (68.0) >25% 2 (33.3) 19 (47.5) 35 (29.9) 31 (32.0) Missing: 21 ERa <0.001a 10% 2 (33.3) 26 (63.4) 48 (40.3) 20 (18.7) >10% 4 (66.7) 15 (36.6) 71 (59.7) 87 (81.3) Missing: 8 PR <0.001a 10% 4 (66.7) 22 (59.5) 46 (46.9) 21 (22.3) >10% 2 (33.3) 15 (40.5) 52 (53.1) 73 (77.6) Missing: 46 Her2 0.858c Negative (10%) 3 (75) 24 (58.5) 70 (59.8) 54 (56.8) Low 0 (0) 7 (17.1) 15 (12.8) 20 (21.1) Intermediate 0 (0) 3 (7.3) 8 (6.8) 14 (14.7) High 1 (25) 7 (17.1) 24 (20.5) 7 (7.4) Missing: 24 p(S118)-ERa <0.001c 0 1 (33.3) 8 (27.6) 12 (16.7) 5 (6.8) 1 1 (33.3) 6 (20.7) 12 (16.7) 7 (9.6) 2 0 (0) 8 (27.6) 25 (34.7) 17 (23.4) 3 1 (33.3) 7 (24.1) 23 (31.9) 44 (60.3) Missing: 104 Cyclin D1 <0.001c 0 2 (33.3) 21 (52.5) 27 (22.7) 14 (13.7) 1 3 (50.0) 14 (35.0) 47 (39.5) 29 (28.4) 2 1 (16.7) 5 (12.5) 37 (31.1) 39 (38.2) 3 0 (0) 0 (0) 8 (6.7) 20 (19.6) Missing: 14 pERK1/2 0.064c 0 5 (83.3) 27 (64.3) 74 (63.8) 51 (53.1) 1 0 (0) 10 (23.8) 21 (18.1) 23 (24.0) 2 1 (16.7) 3 (7.1) 12 (10.3) 16 (16.7) 3 0 (0) 2 (4.8) 9 (7.8) 6 (6.25) Missing: 21 NOTE: Correlations of TGFBR2 expression to clinicopathologic and molecular markers according to staining categories (score 0–3). Percentages are given in parentheses. Abbreviation: NHG, Nottingham histologic grade. aMann–Whitney U test. bKruskal–Wallis test. cSpearman.

Tamoxifen-induced apoptosis is impaired in TGFBR2 was signiﬁcantly attenuated in TGFBR2 knockdown cells at knockdown cells physiologic relevant concentrations (0.1 and 1 mmol/L 4OHT), In a toxicity assay, the active tamoxifen metabolite 4OHT which was yet more prominent in the presence of rhTGFb1at dose-dependently decreased cancer cell growth. This response an even lower tamoxifen concentration (0.01 mmol/L 4OHT;

1462 Cancer Res; 75(7) April 1, 2015 Cancer Research

Table 2. Multivariate interaction analysis indicating treatment-predictive reduced in the presence of the TGFBR1 inhibitor SB431542 information of TGFBR2 (Supplementary Fig. S4A). P Variable HR (95% CI) We observed a slight but significant induction of TGFb tran- Grade (NHG) scriptional activity upon tamoxifen treatment compared with I–II 1 III 1.315 (0.829–2.087) 0.244 estrogen treatment alone, the regulation of which was completely Tumor size eradicated in TGFBR2 knockdown cells (Fig. 3I). Altogether, these 20 mm 1 data highlight the role of the TGFb pathway in ERa and ERK >20 mm 1.241 (0.821–1.874) 0.306 activation and its contribution to tamoxifen response. Lymph node status Negative 1 Positive 2.745 (1.647–4.575) <0.001 TGFBR2 knockdown cells overexpress multidrug resistance Age protein ABCG2 Continuous (per year) 0.974 (0.941–1.008) 0.134 A customized mRNA array was employed to investigate differ- Ki-67 ential gene expression of ERa downstream targets in the TGFBR2 25% 1 >25% 1.212 (0.791–1.859) 0.377 knockdown cells. The array included probes for 91 genes involved ERa in proliferation, apoptosis, epithelial–mesenchymal transition, 10% 1 stemness, or are linked to estrogen response (for full gene list, see >10% 0.865 (0.543–1.380) 0.543 Supplementary Table S4). Treatment Tamoxifen treatment resulted in the anticipated downregula- No tamoxifen 1 tion of estrogen-responsive genes such as CCND1, BMP6, and Tamoxifen 1.827 (0.736–4.532) 0.194 TGFBR2 CXCL12 (Table 3). Changes of apoptosis relevant genes were only Low (0–1) 1 observed in control cells. BCL2, which encodes for an antiapop- High (2–3) 1.823 (0.812–4.091) 0.146 totic molecule, was decreased in tamoxifen-treated control, but Interaction not in TGFBR2 knockdown cells confirming an impaired regula- TGFBR2 x tamoxifen 0.267 (0.098–0.731) 0.010 tion of apoptosis in cells lacking TGFBR2. NOTE: Multivariate Cox regression analysis was performed on all patients, TGFBR2 knockdown cells displayed 5- to 8-fold greater mRNA including prognostic markers with regard to TGFBR2 expression. Treatment expression levels for ABC-transporter ABCG2 compared with interaction characterizes contribution of TGFBR2 as a new biomarker to the difference in survival outcome between treatment groups independently of control independent of treatment (Table 3). Overexpression of other markers. ABCG2 in TGFBR2 knockdown cells was confirmed by Western Abbreviation: NHG, Nottingham histologic grade. blot analysis (Fig. 3J). Cells treated with the TGFBR1 inhibitor also exhibited an increased level of ABCG2 (Supplementary Fig. S4D) implying a role for the TGFb pathway in the regulation of ABCG2 Fig. 3A). Similarly, TGFBR2 knockdown in T47D cells (Fig. 3B) expression potentially affecting multidrug resistance. and using a TGFb inhibitor (Supplementary Fig. S4C) impaired the tamoxifen response. Treatment of TGFBR2 knockdown cells with the pure anties- Long-term treated, tamoxifen-resistant cells acquired low trogen fulvestrant also resulted in an altered drug response (Sup- TGFBR2 expression plementary Fig. S6), suggesting that lack of TGFBR2 results in an To further investigate the association between tamoxifen overall desensitization to endocrine treatment. resistance and the TGFb pathway, we compared a well estab- Cell-cycle distribution (Fig. 3C) and proliferation, as moni- lished, TAM-R cell line (27) to tamoxifen-sensitive parental tored by Ki-67 (Fig. 3D) was unaffected between control and MCF7. TGFBR2 knockdown cells upon tamoxifen treatment. Flow cyto- Immunohistochemical staining revealed that TAM-R express metric analysis of Annexin V (Fig. 3E) and immunostaining with lower levels of TGFBR2 than MCF7 (Fig. 4A), although reduction cleaved PARP revealed that the tamoxifen-induced apoptosis is in mRNA levels was less prominent (Fig. 4B) potentially due to reduced in TGFBR2 knockdown cells (Fig. 3F) with statistical additional changes in posttranscriptional regulation of TGFBR2 significance for the immunofluorescence assay. Taken together, expression or receptor recycling or degradation. We further our data demonstrate that the observed tamoxifen resistance in observed increased ABCG2 and decreased ESR1/ERa levels on TGFBR2 knockdown cells is due to a compromised ability to RNA and protein level (Fig. 4B and C). Interestingly, TAM-R cells induce apoptosis. remain estrogen-responsive, regulating proliferation and ERa transcriptional activity (Fig. 4D and E). Whereas TGFb-induced Altered signaling crosstalk in TGFBR2 knockdown cells transcriptional activity and TGFb-mediated inhibition of ERa By Western blot analysis, we proved that TGFb-induced transcriptional activity was unaffected in TAM-R cells (Supple- phosphorylation of SMAD2 is abrogated in TGFBR2 knock- mentary Fig. S7A and S7B), regulation of TGFb transcriptional down cells (Fig. 3G), demonstrating the inability of cells to activity was reversed upon estrogen and tamoxifen treatment activate the canonical TGFb pathway in the absence of the (Fig. 4F). Collectively, reporter assays reveal an altered ERa-TGFb ligand-binding receptor. Furthermore, we revealed that tamox- crosstalk, with aberrant ERa-mediated TGFb signal transduction ifen-induced ERa phosphorylation at serine-118 (pS118-ERa) pathway activation. More importantly, tamoxifen-induced TGFb is abrogated and ERK1/2 phosphorylation (pERK) is complete- signaling-mediated apoptosis may consequently be lost. ly abolished in TGFBR2 knockdown cells (Fig. 3G and H). Recently, it could be demonstrated that long-term tamoxifen Decreased ERK phosphorylation has previously been reported exposure induces epigenetic silencing of estrogen-responsive in siRNA-mediated TGFBR2 knockdown cells (36). Likewise, genes through promoter DNA methylation thereby facilitating estrogen-induced pS118-ERa and TGFb1–induced pERK were endocrine resistance (37). A number of studies have shown that

www.aacrjournals.org Cancer Res; 75(7) April 1, 2015 1463

Figure 2. Hormone withdrawal and estrogen response in ERa-positive cells with TGFBR2-specificknockdown.A,relativeTGFBR2 mRNA expression of lentivirally transfected MCF7 and T47D with two independent TGFBR2-specific shRNA constructs (TGFBR2-shRNA.1 and -shRNA.2) and control construct (GFP- shRNA; , P < 0.05, Student t test). B, immunohistochemical staining of formalin-fixed, paraffin-embedded cytospins with TGFBR2-specificantibodyof transfected MCF7 (scale bar, 100 mm). C–G, cells were grown in serum-reduced, phenol red–free media supplemented with CSS to study effects in a condition excluding exogenous, serum-derived hormones. C, graph depicts cell-cycle distribution of transfected MCF7. Cell-cycle profile was determined by measuring propidium iodide (PI) staining of DNA content in ethanol-fixed cells by flow cytometry (BD Calibur; , P < 0.05, Student t test). D, graph illustrates percentage of apoptotic (PI/AnVþ) and necrotic (PIþ/AnVþ) cells measured by flow cytometry (BD Calibur; AnV, Annexin V-FITC). E, graphs represent the relative viability of transfected MCF7, with increasing concentrations of estrogen (E2) in the absence (left) or presence (right) of rhTGFb1 after 96 hours. F, flow cytometric analysis of cell cycle; graph displays fold change of cell proportion undergoing DNA synthesis (S-phase) after 24-hour treatment with estrogen (, P < 0.05; Student t test). G, fold change of number of apoptotic cells (Annexin V-positive cells) after 96 hours of treatment with estrogen measured by flow cytometry.

the TGFBR2 promoter region, which contains a CpG island, is a R cells (Supplementary Fig. S7C). TGFBR2 overexpression did not target of epigenetic modulation mainly through histone deace- affect estrogen response (Supplementary Fig. S7D) and we tylation rather than DNA methylation (38). To validate this, we observed only a minor improvement in tamoxifen sensitivity in treated MCF7 and TAM-R with a DNA demethylase agent (5-Aza- TAM-R cells (Supplementary Fig. S7E) possibly due to the pres- 20-deoxycytidine) and a histone deacetylase inhibitor (Trichosta- ence of other resistance mechanisms. Reactivation of TGFBR2 tin A). We observed an about 2-fold increase in TGFBR2 expres- expression alone may improve tamoxifen response to some extent sion when cells were treated with Trichostatin A, whereas DNA but does not seem sufficient to fully overcome endocrine demethylation increased TGFBR2 only slightly in tamoxifen- resistance. resistant cells (Fig. 4G) emphasizing the more important role of histone acetylation in TGFBR2 expression. In accordance, tamoxifen response was restored in TAM-R cells with a more pro- Discussion nounced effect using Trichostatin A. In MCF7, DNA demethyla- TGFBR2 is an independent endocrine treatment–predictive tion by 5-Aza-20-deoxycytidine had no effect, whereas Trichosta- marker in invasive breast cancer tin A enhanced tamoxifen sensitivity (Fig. 4H) supporting that Here, we report that low TGFBR2 expression predicts tamox- DNA methylation is a preferred feature in endocrine-resistant ifen resistance in premenopausal invasive breast cancer inde- cells. pendent of other known prognostic factors. Consistent with We further addressed the question whether increased TGFBR2 our finding, four independent gene expression datasets of expression is linked to tamoxifen sensitivity more directly using a tamoxifen-treated, ERa-positive patients showed significant TGFBR2 overexpression plasmid (pCMV5-XL6-TGFBR2). Tran- shorter recurrence-free survival of patients with low TGFBR2 sient plasmid transfection reinstated TGFBR2 expression in TAM- expression.

1464 Cancer Res; 75(7) April 1, 2015 Cancer Research

Figure 3. Tamoxifen response, cell signaling, and ABCG2 expression in ERa-positive cells with TGFBR2-specific knockdown. Cells were grown in serum-reduced, phenol red–free media supplemented with CSS. A and B, viability assay using Alamar blue; fluorescence was read at 590 nm (excitation 544 nm). Graphs represent relative viability of untransfected and transfected MCF7 (A) or T47D (B) after 96-hour treatment with estrogen (E2) and increasing concentrations of 4OHT in the absence (left) or presence (right) of recombinant human TGFb1(rhTGFb1; , P < 0.05; Student t test). C, flow cytometric analysis of cell cycle; graph illustrates fold change of cell proportion undergoing DNA synthesis (S-phase) after 24-hour treatment with E2 and 4OHT. Cell-cycle distribution was determined by measuring propidium iodide staining of DNA content in ethanol-fixed cells by flow cytometry (BD Calibur). D, analysis of proliferation marker (Ki-67) by immunofluorescence; graph depicts percentage of positively stained cells. Cells were treated for 24 hours with E2 and 4OHT (, P < 0.05; Student t test). E, fold change of apoptotic cells (Annexin V-positive cells) after 96-hour treatment with E2 and 4OHT measured by flow cytometry (BD Calibur). F, analysis of apoptotic marker (cleaved PARP) by immunofluorescence, graph represents percentage of positively stained cells. Cells were treated for 96 hours with E2 and 4OHT (, P < 0.05; Student t test). G and H, Western blots of transfected MCF7 after 24-hour treatment with E2, 4OHT, and/or rhTGFb1. I, reporter assay measuring relative CAGA-dependent luciferase expression (RLU, relative light units) normalized to Renilla luciferase expression in transfected MCF7 after 24-hour treatment with E2 and 4OHT (, P < 0.05; Student t test). J, Western blot analysis demonstrating ABCG2 expression in untransfected and transfected MCF7.

www.aacrjournals.org Cancer Res; 75(7) April 1, 2015 1465

Table 3. Custom transcript array data of differentially expressed genes E2þ4OHT treatment E2 treatment: knockdown E2þ4OHT treatment: knockdown vs. E2 treatment vs. control vs. control Gene symbol Fold change Gene symbol Fold change Gene symbol Fold change GFP-shRNA PIP 4.0b EPO 2.5b ANK3 2.1 KRT18 2.0 CXCR4 2.0 BCL2 2.1 CCND1 2.1 BMP6 3.5 CXCL12 8.1 TGFBR2-shRNA.1 BID 3.7 PIP 8.2b ABCG2 8.4a PIP 2.9b ABCG2 5.4a PIP 6.0b TWIST1 2.4b JAG1 2.1 CDK6 2.1 HES1 2.2 CDH1 2.0 CCND1 2.1 WNT5A 2.1a CXCR4 2.0 MSI1 2.4 HEY1 2.0 SOX9 2.1 BMP6 2.4a DNER 2.1 SNAI1 2.1 DKK1 2.5 TFF1 2.1 CDH11 2.1b LEF1 2.6 PCNA 2.9 AXIN2 2.1 BCL2 2.8 CDH1 3.0 WNT5A 2.3a TWIST1 3.02 AREG 3.2 BAMBI 2.5 TGFB1 3.1 CCND1 3.6 LEF1 2.6 SNAI2 3.31 TLR3 4.8b AREG 2.7 CDH1 3.3 BMP6 5.4a TGFB1 3.3 TLR3 3.6b CXCL12 7.0 BID 3.3 CDH11 3.7b BAMBI 31.6 SNAI2 3.4a AREG 4.5 BCL2 3.4 BAMBI 128.6 MSI1 3.4 DKK1 3.5 TWIST1 7.9b TGFBR2-shRNA.2 TCF4 2.9b PIP 10.8b PIP 7.7b PIP 2.8b ABCG2 6.3a ABCG2 3.9a ANK3 2.0 ALDH1A1 3.9b WISP1 2.9b MKI67 2.0 TFF1 2.8 HEY1 2.7 TFF1 2.0 HEY1 2.4 TFF1 2.2 PCNA 2.5 WISP1 2.2b LEF1 2.3 CCND1 2.8 SNAI2 2.1 TLR3 2.7b BMP6 3.0 CDH11 2.1b EPO 2.8b ALDH1A1 3.1b WNT5A 2.1a CDH11 4.0b AREG 4.3 AREG 2.3 TGFB1 4.2 CXCL12 6.8 LEF1 2.4 AREG 5.0 TLR3 2.7b MSI1 2.8 TGFB1 3.4 NOTE: Differentially expressed genes (>2-fold changes) in MCF7 with TGFBR2-speciﬁc knockdown (TGFBR2-shRNA.1/2) compared to control (GFP-shRNA) in the presence of 1 nmol/L estrogen with or without 0.1 mmol/L 4OHT or comparing treatments within each cell line. Analysis was performed using SABiosciences' web- based PCR data analysis tool for customized RT2 Proﬁler PCR Array. Data represent two independent experiments. aThis gene's expression is relatively low in one sample and reasonably detected in the other sample, suggesting that the actual fold change value is at least as large as the calculated and reported fold change result. bThis gene's relative expression level is low in both control and test samples. This fold change result may have greater variations.

In accordance, Vendrell and colleagues reported a 47-gene rence-free survival in the untreated patient group. However, signature associated with tamoxifen failure (39) including lower ERa-negative patients with low TGFBR2 expression had a TGFBR2 levels in tamoxifen-resistant tumor samples. Another slightly better prognosis, but patient numbers were relatively study reported the presence of inhibitory TGFBR2 mutations in low (Supplementary Fig. S8A). Examining tumor samples in recurrent breast cancer (40). All tumor samples in this study were regard to high SMAD2 phosphorylation (pSMAD2), there was a derived from patients having undergone adjuvant tamoxifen nonsignificant trend (P ¼ 0.128) for shorter recurrence-free treatment hinting that mutational inactivation of TGFBR2 may survival for ERa-negative, but not for ERa-positive patients contribute to tamoxifen resistance. (Supplementary Fig. S8B). Taken together, our data indicate Low TGFBR2 expression has previously been shown to be that the TGFb pathway may act in a tumor-promotive or associated with longer overall survival in patients with ERa- -suppressive manner depending on the ERa status. Targeting negative breast cancer (14). Stratifying the clinical cohort used the TGFb pathway in ERa-negative breast cancer might therefore in this study according to ERa status, we did not find a prevent tumor progression, whereas this approach may prove statistically significant link of TGFBR2 expression to recur- to be an unfavorable strategy in ERa-positive breast cancer.

1466 Cancer Res; 75(7) April 1, 2015 Cancer Research

Figure 4. Analysis of long-term treated, TAM-R cells in comparison with parental, tamoxifen-sensitive MCF7. A, immunohistochemical staining of formalin-fixed, paraffin-embedded cytospins with TGFBR2-specific antibody (scale bar, 100 mm). B, relative mRNA expression levels of TGFBR2, ABCG2 and ESR1 of MCF7, and TAM-R grown in respective culture media. C, Western blot analysis of MCF7 and TAM-R. D–H, cells were grown in serum-reduced, phenol red–free media supplemented with CSS. D, viability assay using Alamar blue; fluorescence was read at 590 nm (excitation 544 nm). Graph represents relative viability of cells treated with 1 nmol/L estrogen (E2) or 1 ng/mL recombinant human TGFb1(rhTGFb1).EandF,reporterassaymeasuringrelativeERE-dependent (E) or CAGA-dependent (F) luciferase expression (RLU, relative light units) normalized to Renilla luciferase expression in transfected MCF7 and TAM-R cells after 24-hour treatment with E2, 4OHT, or rhTGFb1. G, relative TGFBR2 mRNA levels after 48-hour treatment with 1 nmol/L E2, 0.1 mmol/L 4OHT, and 0.5 mmol/L 5-Aza-20-deoxycytidine (5-Aza; DNA methylase inhibitor) or 0.1 mmol/L Trichostatin A (TSA) (histone deacetylase inhibitor). H, graph represents relative viability of MCF7 and TAM-R cells after 96-hour treatment with 1 nmol/L E2, 0.1 mmol/L 4OHT, and 0.5 mmol/L 5-Aza or 0.1 mmol/L TSA (, P < 0.05; Student t test).

Low TGFBR2 expression in ERa-positive breast cancer cells is nism enabling crosstalk between signaling pathways, and several associated with tamoxifen resistance and deregulated ERa and phosphorylation sites have been linked to endocrine response TGFb signaling (42, 43). Serine-118 ERa phosphorylation, described as a pre- Analysis of an established TAM-R cell line with downregulated dictor for tamoxifen sensitivity (30), was linked to TGFBR2 expression levels of TGFBR2 signifies that loss of TGFBR2 plays a expression in the clinical cohort and in our experimental settings. role in the acquisition of endocrine resistance. In the current Various studies have described a relationship between anti- study, we reveal that shRNA-induced loss of TGFBR2 expression estrogens and the TGFb pathway. The clinical efficacy of tamox- in antiestrogen sensitive ERa-positive breast cancer cells gives rise ifen in breast cancer has been attributed to growth arrest and to an impaired response toward estrogen and tamoxifen through induction of apoptosis (44, 45) also via TGFb signaling (46, 47). alterations in proliferation and induction of apoptosis. Our data support that tamoxifen-induced apoptosis seems Interestingly, TGFBR2-specific knockdown and TAM-R cells to be at least partly mediated by the canonical TGFb pathway. maintain estrogen responsiveness, however, with considerably However, deregulated TGFb transcriptional activation, as elevated ERa transcriptional activity. It has been reported previ- observed in TGFBR2 knockdown and TAM-R cells, is potentially ously that tamoxifen-resistant breast cancer exhibit enhanced ERa impeding tamoxifen-induced apoptosis, thus contributing to binding including additional ERa binding regions and that drug resistance. We further noted increased expression level of enhanced ERa-binding capacity was associated with worse clin- ATP-binding cassette (ABC) efflux transporter ABCG2, also ical outcome (41). referred to as breast cancer resistance protein (BCRP), in TGFBR2 Furthermore, site-specificERa phosphorylation has been knockdown and TAM-R cells. ABCG2 as a multidrug resistance implicated as an important posttranslational regulatory mecha- protein confers cells with the capability to pump out large

www.aacrjournals.org Cancer Res; 75(7) April 1, 2015 1467

hydrophobic molecules including therapeutic agents. Recently, it represent a novel therapeutic approach in these patients; however, has been reported that ABCG2 is able to recognize and efflux they may prove challenging due the dual nature of TGFb. tamoxifen (48). Cells with elevated levels of ABCG2 may therefore become insensitive to tamoxifen due to limited drug expo- Disclosure of Potential Conflicts of Interest fl sure. ABCG2 expression has been shown to be negatively regu- No potential con icts of interest were disclosed. lated by the TGFb and ERK pathway (49). Upregulation of ABCG2 Authors' Contributions may thus be mediated through the abrogation of activated ERK Conception and design: S. Busch, O. Stal, M. Ferno, G. Landberg signaling as observed in TGFBR2 knockdown cells. Development of methodology: S. Busch It has been demonstrated that long-term tamoxifen exposure Acquisition of data (provided animals, acquired and managed patients, induces epigenetic silencing of estrogen-responsive genes that provided facilities, etc.): S. Busch, G. Landberg negatively regulate proliferation (37). Reactivation of identified Analysis and interpretation of data (e.g., statistical analysis, biostatistics, genes was achieved through inhibition of promoter DNA computational analysis): S. Busch, A.H. Sims, G. Landberg methylation, which restored estrogen-dependent cell prolifer- Writing, review, and/or revision of the manuscript: S. Busch, A.H. Sims, O. Stal, M. Ferno, G. Landberg ation and tamoxifen sensitivity. Using epigenetic regulators, we Study supervision: M. Ferno, G. Landberg noted an upregulation of TGFBR2 mRNAexpressionintamox- ifen-resistant cells concomitantly with a resensitization to tamoxifen treatment. Acknowledgments The authors thank Elise Nilsson and Ylva Magnusson for excellent technical In summary, our study provides further evidence for a crosslink a b assistance. Further, the authors would like to thank Robert B. Clarke for the kind between the ER and TGF pathway and its implication in gift of reporter plasmids and TAM-R cells, Nullin Divecha for reporter plasmids, endocrine resistance. We demonstrate that TGFBR2 is an inde- and Akira Orimo for lentiviral plasmids. pendent treatment-predictive biomarker in breast cancer and hypothesize that loss of TGFBR2 contributes to the acquisition Grant Support of resistance mechanisms through (i) deregulation of ERa sig- This study was supported by the Swedish Cancer Society and Breakthrough naling (activation, phosphorylation), (ii) impaired induction of Breast Cancer UK. The project was further supported by BioCARE—a National apoptosis via TGFb signaling, and (iii) ABCG2-mediated drug Strategic Research Program at University of Gothenburg. desensitization. The costs of publication of this article were defrayed in part by the payment a of page charges. This article must therefore be hereby marked advertisement The acquisition of tamoxifen resistance in some ER -positive in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. breast cancer patients may be a consequence of epigenetic silencing of TGFBR2 expression, which has been reported to be an early Received May 28, 2014; revised November 6, 2014; accepted December 20, event in breast tumorigenesis (50). Thus, epigenetic therapies may 2014; published online April 1, 2015.

References 1. Musgrove EA, Sutherland RL. Biological determinants of endocrine resis- 13. Derynck R, Zhang YE. Smad-dependent and Smad-independent path- tance in breast cancer. Nat Rev Cancer 2009;9:631–43. ways in TGF-beta family signalling. Nature 2003;425:577–84. 2. Osborne CK, Schiff R. Mechanisms of endocrine resistance in breast cancer. 14. Buck MB, Fritz P, Dippon J, Zugmaier G, Knabbe C. Prognostic signi- Annu Rev Med 2011;62:233–47. ﬁcance of transforming growth factor beta receptor II in estrogen receptor- 3. Massarweh S, Schiff R. Unraveling the mechanisms of endocrine resistance negative breast cancer patients. Clin Cancer Res 2004;10:491–8. in breast cancer: new therapeutic opportunities. Clin Cancer Res 2007;13: 15. Padua D, Zhang XH, Wang Q, Nadal C, Gerald WL, Gomis RR, et al. 1950–4. TGFbeta primes breast tumors for lung metastasis seeding through angio- 4. Siegel PM, Massague J. Cytostatic and apoptotic actions of TGF-beta in poietin-like 4. Cell 2008;133:66–77. homeostasis and cancer. Nat Rev Cancer 2003;3:807–21. 16. Ito I, Hanyu A, Wayama M, Goto N, Katsuno Y, Kawasaki S, et al. Estrogen 5. Moustakas A, Pardali K, Gaal A, Heldin CH. Mechanisms of TGF-beta inhibits transforming growth factor beta signaling by promoting Smad2/3 signaling in regulation of cell growth and differentiation. Immunol Lett degradation. J Biol Chem 2010;285:14747–55. 2002;82:85–91. 17. Kleuser B, Malek D, Gust R, Pertz HH, Potteck H. 17-Beta-estradiol inhibits 6. Pardali K, Moustakas A. Actions of TGF-beta as tumor suppressor and transforming growth factor-beta signaling and function in breast cancer pro-metastatic factor in human cancer. Biochim Biophys Acta 2007; cells via activation of extracellular signal-regulated kinase through the G 1775:21–62. protein-coupled receptor 30. Mol Pharmacol 2008;74:1533–43. 7. Massague J. TGFbeta in cancer. Cell 2008;134:215–30. 18. Matsuda T, Yamamoto T, Muraguchi A, Saatcioglu F. Cross-talk between 8. Ikushima H, Miyazono K. TGFbeta signalling: a complex web in cancer transforming growth factor-beta and estrogen receptor signaling through progression. Nat Rev Cancer 2010;10:415–24. Smad3. J Biol Chem 2001;276:42908–14. 9. Rahimi RA, Leof EB. TGF-beta signaling: a tale of two responses. J Cell 19. Wu L, Wu Y, Gathings B, Wan M, Li X, Grizzle W, et al. Smad4 as a Biochem 2007;102:593–608. transcription corepressor for estrogen receptor alpha. J Biol Chem 2003; 10. Muraoka RS, Koh Y, Roebuck LR, Sanders ME, Brantley-Sieders D, Gorska 278:15192–200. AE, et al. Increased malignancy of Neu-induced mammary tumors over- 20. Buck MB, Pﬁzenmaier K, Knabbe C. Antiestrogens induce growth inhibi- expressing active transforming growth factor beta1. Mol Cell Biol 2003; tion by sequential activation of p38 mitogen-activated protein kinase and 23:8691–703. transforming growth factor-beta pathways in human breast cancer cells. 11. Yang L, Huang J, Ren X, Gorska AE, Chytil A, Aakre M, et al. Abrogation of Mol Endocrinol 2004;18:1643–57. TGF beta signaling in mammary carcinomas recruits Gr-1þCD11bþ 21. Butta A, MacLennan K, Flanders KC, Sacks NP, Smith I, McKinna A, et al. myeloid cells that promote metastasis. Cancer Cell 2008;13:23–35. Induction of transforming growth factor beta 1 in human breast cancer in 12. Forrester E, Chytil A, Bierie B, Aakre M, Gorska AE, Sharif-Afshar AR, et al. vivo following tamoxifen treatment. Cancer Res 1992;52:4261–4. Effect of conditional knockout of the type II TGF-beta receptor gene in 22. Ryden L, Jonsson PE, Chebil G, Dufmats M, Ferno M, Jirstrom K, et al. Two mammary epithelia on mammary gland development and polyomavirus years of adjuvant tamoxifen in premenopausal patients with breast cancer: middle T antigen induced tumor formation and metastasis. Cancer Res a randomised, controlled trial with long-term follow-up. Eur J Cancer 2005;65:2296–302. 2005;41:256–64.

1468 Cancer Res; 75(7) April 1, 2015 Cancer Research

23. Holm C, Rayala S, Jirstrom K, Stal O, Kumar R, Landberg G. Association 36. Panno ML, Giordano F, Rizza P, Pellegrino M, Zito D, Giordano C, et al. between Pak1 expression and subcellular localization and tamoxifen Bergapten induces ER depletion in breast cancer cells through SMAD4- resistance in breast cancer patients. J Natl Cancer Inst 2006;98:671–80. mediated ubiquitination. Breast Cancer Res Treat 2012;136:443–55. 24. Busch S, Acar A, Magnusson Y, Gregersson P, Ryden L, Landberg G. TGF- 37. Stone A, Valdes-Mora F, Gee JM, Farrow L, McClelland RA, Fiegl H, et al. beta receptor type-2 expression in cancer-associated fibroblasts regulates Tamoxifen-induced epigenetic silencing of oestrogen-regulated genes in breast cancer cell growth and survival and is a prognostic marker in pre- anti-hormone resistant breast cancer. PLoS ONE 2012;7:e40466. menopausal breast cancer. Oncogene 2013;34:27—38. 38. Chowdhury S, Ammanamanchi S, Howell GM. Epigenetic targeting of 25. Taylor KJ, Sims AH, Liang L, Faratian D, Muir M, Walker G, et al. Dynamic transforming growth factor beta receptor II and implications for cancer changes in gene expression in vivo predict prognosis of tamoxifen-treated therapy. Mol Cell Pharmacol 2009;1:57–70. patients with breast cancer. Breast Cancer Res 2010;12:R39. 39. Vendrell JA, Robertson KE, Ravel P, Bray SE, Bajard A, Purdie CA, et al. A 26. Camp RL, Dolled-Filhart M, Rimm DL. X-tile: a new bio-informatics tool candidate molecular signature associated with tamoxifen failure in primary for biomarker assessment and outcome-based cut-point optimization. breast cancer. Breast Cancer Res 2008;10:R88. Clin Cancer Res 2004;10:7252–9. 40. Lucke CD, Philpott A, Metcalfe JC, Thompson AM, Hughes-Davies L, 27. Knowlden JM, Hutcheson IR, Jones HE, Madden T, Gee JM, Harper ME, Kemp PR, et al. Inhibiting mutations in the transforming growth factor et al. Elevated levels of epidermal growth factor receptor/c-erbB2 hetero- beta type 2 receptor in recurrent human breast cancer. Cancer Res dimers mediate an autocrine growth regulatory pathway in tamoxifen- 2001;61:482–5. resistant MCF-7 cells. Endocrinology 2003;144:1032–44. 41. Ross-Innes CS, Stark R, Teschendorff AE, Holmes KA, Ali HR, Dunning MJ, 28. Stewart SA, Dykxhoorn DM, Palliser D, Mizuno H, Yu EY, An DS, et al. et al. Differential oestrogen receptor binding is associated with clinical Lentivirus-delivered stable gene silencing by RNAi in primary cells. RNA outcome in breast cancer. Nature 2012;481:389–93. 2003;9:493–501. 42. Murphy LC, Seekallu SV, Watson PH. Clinical significance of estrogen 29. Yamashita S, Takahashi S, McDonell N, Watanabe N, Niwa T, Hosoya K, receptor phosphorylation. Endocr Relat Cancer 2011;18:R1–14. et al. Methylation silencing of transforming growth factor-beta receptor 43. de Leeuw R, Neefjes J, Michalides R. A role for estrogen receptor phos- type II in rat prostate cancers. Cancer Res 2008;68:2112–21. phorylation in the resistance to tamoxifen. Int J Breast Cancer 2011;2011: 30. Kok M, Holm-Wigerup C, Hauptmann M, Michalides R, Stal O, Linn S, et al. 232435. Estrogen receptor-alpha phosphorylation at serine-118 and tamoxifen 44. Mandlekar S, Kong AN. Mechanisms of tamoxifen-induced apoptosis. response in breast cancer. J Natl Cancer Inst 2009;101:1725–9. Apoptosis 2001;6:469–77. 31. Svensson S, Jirstrom K, Ryden L, Roos G, Emdin S, Ostrowski MC, et al. ERK 45. Ellis PA, Saccani-Jotti G, Clarke R, Johnston SR, Anderson E, Howell A, et al. phosphorylation is linked to VEGFR2 expression and Ets-2 phosphoryla- Induction of apoptosis by tamoxifen and ICI 182780 in primary breast tion in breast cancer and is associated with tamoxifen treatment resistance cancer. Int J Cancer 1997;72:608–13. and small tumours with good prognosis. Oncogene 2005;24:4370–9. 46. Perry RR, Kang Y, Greaves BR. Relationship between tamoxifen-induced 32. Kato S, Endoh H, Masuhiro Y, Kitamoto T, Uchiyama S, Sasaki H, et al. transforming growth factor beta 1 expression, cytostasis and apoptosis in Activation of the estrogen receptor through phosphorylation by mitogen- human breast cancer cells. Br J Cancer 1995;72:1441–6. activated protein kinase. Science 1995;270:1491–4. 47. Yoo YA, Kim YH, Kim JS, Seo JH. The functional implications of Akt activity 33. Chen D, Riedl T, Washbrook E, Pace PE, Coombes RC, Egly JM, et al. and TGF-beta signaling in tamoxifen-resistant breast cancer. Biochim Activation of estrogen receptor alpha by S118 phosphorylation involves a Biophys Acta 2008;1783:438–47. ligand-dependent interaction with TFIIH and participation of CDK7. Mol 48. Selever J, Gu G, Lewis MT, Beyer A, Herynk MH, Covington KR, et al. Dicer- Cell 2000;6:127–37. mediated upregulation of BCRP confers tamoxifen resistance in human 34. Cheng J, Zhang C, Shapiro DJ. A functional serine 118 phosphorylation site breast cancer cells. Clin Cancer Res 2011;17:6510–21. in estrogen receptor-alpha is required for down-regulation of gene expres- 49. Imai Y, Ohmori K, Yasuda S, Wada M, Suzuki T, Fukuda K, et al. Breast sion by 17beta-estradiol and 4-hydroxytamoxifen. Endocrinology 2007; cancer resistance protein/ABCG2 is differentially regulated downstream of 148:4634–41. extracellular signal-regulated kinase. Cancer Sci 2009;100:1118–27. 35. Inman GJ, Nicolas FJ, Callahan JF, Harling JD, Gaster LM, Reith AD, et al. 50. Hinshelwood RA, Huschtscha LI, Melki J, Stirzaker C, Abdipranoto A, Vissel SB-431542 is a potent and specific inhibitor of transforming growth factor- B, et al. Concordant epigenetic silencing of transforming growth factor-beta beta superfamily type I activin receptor-like kinase (ALK) receptors ALK4, signaling pathway genes occurs early in breast carcinogenesis. Cancer Res ALK5, and ALK7. Mol Pharmacol 2002;62:65–74. 2007;67:11517–27.

www.aacrjournals.org Cancer Res; 75(7) April 1, 2015 1469

Downloaded from cancerres.aacrjournals.org on September 30, 2021. © 2015 American Association for Cancer Research. Loss of TGFβ Receptor Type 2 Expression Impairs Estrogen Response and Confers Tamoxifen Resistance

Susann Busch, Andrew H. Sims, Olle Stål, et al.

Cancer Res 2015;75:1457-1469.

Updated version Access the most recent version of this article at: http://cancerres.aacrjournals.org/content/75/7/1457

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2015/06/18/75.7.1457.DC1

Cited articles This article cites 50 articles, 18 of which you can access for free at: http://cancerres.aacrjournals.org/content/75/7/1457.full#ref-list-1

Citing articles This article has been cited by 2 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/75/7/1457.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at Subscriptions [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/75/7/1457. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.