A Noncompetitive Small Molecule Inhibitor of Estrogen-regulated Gene Expression and Breast Cancer Cell Growth That Enhances Proteasome-dependent Degradation of Estrogen Receptor α*

The mechanisms responsible for 17β-estradiol (E2)-stimulated breast cancer growth and development of resistance to tamoxifen and other estrogen receptor α (ERα) antagonists are not fully understood. We describe a new tool for dissecting ERα action in breast cancer, p-fluoro-4-(1,2,3,6,-tetrahydro-1,3-dimethyl-2-oxo-6-thionpurin-8-ylthio) (TPSF), a potent small-molecule inhibitor of estrogen receptor α that does not compete with estrogen for binding to ERα. TPSF noncompetitively inhibits estrogen-dependent ERα-mediated gene expression with little inhibition of transcriptional activity by NF-κB or the androgen or glucocorticoid receptor. TPSF inhibits E2-ERα-mediated induction of the proteinase inhibitor 9 gene, which is activated by ERα binding to estrogen response element DNA, and the cyclin D1 gene, which is induced by tethering ERα to other DNA-bound proteins. TPSF inhibits anchorage-dependent and anchorage-independent E2-ERα-stimulated growth of MCF-7 cells but does not inhibit growth of ER-negative MDA-MB-231 breast cancer cells. TPSF also inhibits ERα-dependent growth in three cellular models for tamoxifen resistance; that is, 4-hydroxytamoxifen-stimulated MCF7ERαHA cells that overexpress ERα, fully tamoxifen-resistant BT474 cells that have amplified HER-2 and AIB1, and partially tamoxifen-resistant ZR-75 cells. TPSF reduces ERα protein levels in MCF-7 cells and several other cell lines without altering ERα mRNA levels. The proteasome inhibitor MG132 abolished down-regulation of ERα by TPSF. Thus, TPSF affects receptor levels at least in part due to its ability to enhance proteasome-dependent degradation of ERα. TPSF represents a novel class of ER inhibitor with significant clinical potential.


The mechanisms responsible for 17␤-estradiol (E 2 )-stimulated breast cancer growth and development of resistance to tamoxifen and other estrogen receptor ␣ (ER␣) antagonists are not fully understood. We describe a new tool for dissecting ER␣ action in breast cancer, p-fluoro-4-(1,2,3,6,-tetrahydro-1,3-dimethyl-2-oxo-6-thionpurin-8-ylthio) (TPSF), a potent small-molecule inhibitor of estrogen receptor ␣ that does not compete with estrogen for binding to ER␣. TPSF noncompetitively inhibits estrogen-dependent ER␣-mediated gene expression with little inhibition of transcriptional activity by NF-B or the androgen or glucocorticoid receptor. TPSF inhibits E 2 -
ER␣-mediated induction of the proteinase inhibitor 9 gene, which is activated by ER␣ binding to estrogen response element DNA, and the cyclin D1 gene, which is induced by tethering ER␣ to other DNA-bound proteins. TPSF inhibits anchorage-dependent and anchorage-independent E 2 -ER␣stimulated growth of MCF-7 cells but does not inhibit growth of ER-negative MDA-MB-231 breast cancer cells. TPSF also inhibits ER␣-dependent growth in three cellular models for tamoxifen resistance; that is, 4-hydroxytamoxifen-stimulated MCF7ER␣HA cells that overexpress ER␣, fully tamoxifen-resistant BT474 cells that have amplified HER-2 and AIB1, and partially tamoxifen-resistant ZR-75 cells. TPSF reduces ER␣ protein levels in MCF-7 cells and several other cell lines without altering ER␣ mRNA levels. The proteasome inhibitor MG132 abolished down-regulation of ER␣ by TPSF. Thus, TPSF affects receptor levels at least in part due to its ability to enhance proteasome-dependent degradation of ER␣. TPSF represents a novel class of ER inhibitor with significant clinical potential.
Estrogen receptor ␣ (ER␣) 3 is a well studied member of the steroid/nuclear receptor family of transcription regulators. ER␣ acts in the nucleus to regulate gene expression by binding to estrogen response elements (EREs) and related DNA sequences (1)(2)(3)(4) and through association with transcription factors bound at SP1 and AP-1 DNA binding sites (4 -7). In response to high affinity estrogen binding, ER␣ dimerizes, binds to ERE DNAs, and undergoes a conformational change in the ligand binding domain that facilitates the recruitment of coactivators (8). Bound coactivators promote assembly of a multiprotein complex that enables chromatin remodeling and stabilization of an active transcription complex (9 -11). In contrast, antagonist-occupied ER␣ recruits corepressors (12).
At detection, growth of most human breast cancers depends on 17␤-estradiol (E 2 ) binding to ER␣ (13)(14)(15)(16). Treatment strategies that inhibit estrogen-dependent breast cancer include selective ER modulators such as tamoxifen, which binds in the ER␣ ligand binding pocket, and aromatase inhibitors, which block estrogen production. Nearly half of patients treated with aromatase inhibitors develop resistance (17). The long-term effectiveness of tamoxifen is limited by the development of resistance in nearly all patients with metastatic breast cancer and in ϳ40% of patients with primary breast cancers (18). The development of resistance to current therapies underscores the need to develop new small molecule an-tagonists that act outside the ligand binding pocket of ER␣. We recently described an in vitro high throughput screening strategy to identify small molecule inhibitors of ER␣ binding to DNA. We identified 8-benzylsulfanylmethyl-1,3-dimethyl-3,7-dihydropurine-2,6-dione (TPBM) as a small molecule inhibitor of ER␣ binding to ERE DNA (19). Using a cell-based screen, we evaluated ϳ200 small molecules structurally related to TPBM and identified butyrophenone, p-fluoro-4-(1,2,3,6, -tetrahydro-1,3-dimethyl-2-oxo-6-thionpurin-8-ylthio) (TPSF) as a novel inhibitor of ER␣ Ͼ15-fold more potent than TPBM. Although structurally related to TPBM, TPSF exhibits an entirely different mode of action. Although TPBM inhibits in vitro binding of E 2 -ER␣ to a labeled ERE, TPSF does not. TPSF strongly reduces ER␣ levels in breast cancer cells, whereas TPBM has little or no effect on the level of ER␣. Here we demonstrate the selectivity of TPSF and its ability to inhibit expression of an endogenous ER␣-regulated gene that contains EREs and a gene regulated by tethering of ER␣ through other proteins. We show that TPSF inhibits anchorage-dependent and anchorage-independent growth of tamoxifen-sensitive and tamoxifen-resistant ER␣-containing breast cancer cells and demonstrate that TPSF enhances proteasome-dependent degradation of ER␣.

EXPERIMENTAL PROCEDURES
Cell Culture-Unless otherwise indicated, cells were maintained at 37°C in 5% CO 2 in growth medium containing 1% penicillin and streptomycin and fetal bovine serum (FBS) (Atlanta Biological, Atlanta, GA) or calf serum and transferred to phenol red-free medium containing charcoal-dextran (CD)stripped serum at least 2 days before treatment with E 2 , 4-hydroxytamoxifen (OHT), or TPSF. ER␣-positive MCF-7 and ER-negative MDA-MB-231, human breast cancer cells, were cultured in MEM supplemented with 10% calf serum and switched to MEM containing 5% CD-treated calf serum 3 or 4 days before the experiment. The medium was changed on day 2. Tet-inducible MCF7ER␣HA cells were maintained in DMEM supplemented with 1 mM sodium pyruvate, 0.5 g/ml puromycin, and 10% FBS. Four days before the experiment, MCF7ER␣HA cells were switched to the above medium without phenol red containing 10% 6ϫ stripped CD-treated FBS without puromycin (20 -23). ZR-75 human breast cancer cells were maintained in MEM containing 10% calf serum and transferred to medium containing 10% CD-CS 4 days before the experiment. BT474 human breast cancer cells were maintained in improved MEM (iMEM) containing 10% FBS and transferred to phenol red-free iMEM containing 10% CD-FBS 4 days before the experiment. T47D-KBluc breast cancer cells expressing an (ERE) 3 -luciferase reporter gene (24) were maintained in phenol red-free RPMI 1640 containing 2 mM L-glutamine, 1.5 g/liter sodium bicarbonate, 4.5 g/liter glucose, 10 mM Hepes, pH 7.5, 1 mM sodium pyruvate, 10% FBS. Four days before induction with E 2 , cells were transferred to medium without phenol red containing 10% 2ϫ CD calf serum. T47D/A1-2 cells that stably express the glucocorticoid receptor (GR) and contain a mouse mammary tumor virus (MMTV)-luciferase reporter (25) were maintained in MEM supplemented with 10 mM HEPES, pH 7.4, 2 mM glutamine, 5% FBS, and 0.2 mg/ml Geneticin (G418). Four days before the experiment the cells were transferred to the above phenol red-free medium (phenol red-free) containing 10% 2ϫ CD-CS. HeLa-AR1C-PSA-Luc-A6 cells that stably express androgen receptor (AR) and a prostate-specific antigen (PSA)-Luc reporter were maintained in phenol-red free MEM supplemented with 2 mM L-glutamine, 1 mM sodium pyruvate, 10% FBS under selection with 0.1 mg/ml hygromycin B (Roche Applied Science), and 0.5 mg/ml G418. Four days before the experiment, cells were transferred to medium containing 10% 2ϫ CD-CS.
Fluorescence Anisotropy Assays-The fluorescence anisotropy microplate assay for analyzing binding of ER␣ to the fluorescein-labeled consensus ERE was as described (19).
Competitive Radioligand Binding Assays-The relative binding affinity of TPSF for ER␣ and ER␤ was determined in competitive radioligand binding assays using 2 nM [ 3 H]E 2 and a range of TPSF concentrations as described (26,27).
Reporter Gene Assays-Reporter gene assays were performed to compare the ability of about 200 compounds structurally related to TPBM (19) to inhibit estrogen-dependent transcription in T47D-KBluc breast cancer cells stably transfected to express an (ERE) 3 -Luc reporter (24). The ability of TPSF to inhibit AR and GR transcriptional activity was assayed in HeLa AR1C-PSA-Luc-A6 cells that stably express human AR and a PSA-Luc reporter and in T47D/A1-2 cells that stably express GR and MMTV-Luc. Four days before each experiment, cells were switched to medium containing CD-treated serum as described above. HeLa AR-PSA-Luc cells (100,000 cells/well) and T47DA/1-2 and T47D-KBluc cells (200,000 cells/well) were plated in 1 ml of media in 24well plates. After 24 h the indicated concentrations of E 2 , dihydrotestosterone, or dexamethasone in DMSO or DMSO vehicle alone with or without TPSF were added to each well. After 24 h, cells were washed once with phosphate-buffered saline and lysed in 100 l of passive lysis buffer (Promega, Madison WI). Luciferase activity was determined using BrightGlo firefly luciferase reagent from Promega.
Cell Growth and Viability Assays-Cells were maintained in CD-treated serum for at least 4 days before each experiment. To minimize cell aggregation, MCF-7 cells were harvested in 10 mM HEPES, pH 7.4, 1 mM EDTA. Other cell lines were harvested in trypsin-EDTA. To assay anchorage-dependent cell growth, 1000 cells/well were plated in a 96-well plate. For slow-growing ZR-75 cells, 2000 cells were plated/ well. Cells were maintained in medium containing CD-treated serum for 24 h, and the medium was then changed, and E 2 and DMSO vehicle or TPSF in DMSO was added. The medium was replaced after 2 days, except for BT474 cells, whose medium was not changed. After 4 days, cell viability was determined using Promega CellTiter 96 Aqueous One Solution Cell Proliferation Assay (MTS) (Promega).
To assay anchorage-independent cell growth in soft agar, 1%, and 0.7% Select Agar (Invitrogen) was prepared in water and warmed at 40°C before use. 1.5 ml of 0.5% bottom agar diluted in medium was added to each well of a 6-well cell culture plate and allowed to solidify at room temperature. Top agar was prepared by dilution in warm medium containing the various treatments. MCF-7 cells were resuspended in 1.5 ml of 0.35% top agar at 5000 cells/well and plated in 3 wells for each condition. The plate was kept at room temperature for 30 min until the top agar solidified, then 0.5 ml of medium containing the respective treatments was added on top of the agar. Culture medium on top of the agar was changed every 3-4 days. Colonies were visible by 1 week and counted at day 16 using a dissecting microscope. Photographs of colonies were taken using a Zeiss AxioImager2 imaging system at 6ϫ magnification.
Western Blot-MCF-7 cells were plated at 200,000 cells/ well in 6-well plates in MEM containing 5% 1ϫ CD FBS. The medium was changed at day 2, and at day 4 the medium was replaced with fresh medium containing the indicated treatments. Whole cell extracts were prepared after 24 h in 1ϫ radioimmune precipitation assay buffer (Millipore, CA) containing protease inhibitor mixture (Roche Applied Science). Extract (20 g of protein/lane) was run on 10% SDS-PAGE gels and transferred to nitrocellulose membranes. ER was detected using a 1:4000 dilution of ER␣ antibody ER6F11 (Bio Care Medical). The blot was stripped for 25 min and reprobed using a 1:10,000 dilution of ␤-actin monoclonal antibody (Sigma).
Statistical Analysis-Results are expressed as the mean Ϯ S.E. of at least three independent experiments. Student's t test was used for comparison of the means between two groups. Significance was established when p Ͻ 0.05. The comparisons are described in the figure legends and are not shown in the body of the figures.

RESULTS
TPSF Is a Structure-specific Inhibitor That Acts Outside of the Ligand Binding Pocket of ER␣-We evaluated the ability of ϳ200 compounds structurally related to TPBM (Fig. 1A), a first generation ER␣ inhibitor (19), to inhibit E 2 -ER␣-mediated gene expression in ER␣-positive T47D-KBluc human breast cancer cells containing (ERE) 3 -Luc (24). Of the compounds tested, TPSF ( Fig. 1A) was ϳ16-fold more potent than TPBM in inhibiting ER␣-mediated gene expression (Fig. 1B). To test whether TPSF is a structure-specific inhibitor, we compared the ability of TPSF to inhibit ER␣-mediated gene expression to the structurally similar small molecule 99676. TPSF differs from NSC 99676 by having hydrophilic CAO and fluorine substitutions at the phenyl ring (Fig. 1A). The similar, but more hydrophobic, 99676 had an IC 50 of 9 M and was ϳ13-fold less potent than TPSF (Fig. 1B). The results suggest that TPSF is a structure-specific inhibitor of ER␣ and is not simply acting by promiscuous inhibition due to micelle formation.
If TPSF inhibited ER␣ by competing with E 2 for binding in the ligand binding pocket of the receptor, increasing the E 2 concentration should reduce the ability of TPSF to bind ER␣ and block its action. To test this, we varied the E 2 concentration by 500-fold and tested the ability of TPSF to inhibit (ERE) 3 -Luc in T47D cells. Increasing the concentration of E 2 from 0.2 to 100 nM only slightly increased the IC 50 for inhibiting E 2 -ER␣-mediated transcription from 0.4 M (Fig. 1B, filled triangles) to 0.7 M (Fig. 1B, open triangles), suggesting TPSF acts outside the ER␣ ligand binding pocket.
If TPSF is a highly potent ER␣ ligand, it might retain the ability to inhibit ER␣ at 100 nM E 2 . To test whether TPSF binds in the ligand binding pocket of ER␣, the ability of TPSF to compete with radiolabeled E 2 for binding to ER␣ was evaluated. In competitive radiometric binding assays performed across a broad range of concentrations (26,27), TPSF had virtually no ability to compete with E 2 for binding to ER␣. With E 2 set at 100%, TPSF had a relative binding affinity for ER␣ of ϳ0.001%, indicating that it is not a classical ligand that competes with E 2 for binding in the ER␣ ligand binding pocket.
TPSF Is a Specific Inhibitor of ER␣ Transactivation-The ER␣ binding cleft for p160 coactivator LXXLL motifs has been a major target for development of peptide and small molecule inhibitors (29 -33). Although these inhibitors are effective in reporter gene assays in transfected cells, in general they have not been shown to effectively inhibit expression of endogenous ER-regulated genes in breast cancer cells. We, therefore, tested the ability of TPSF to inhibit the expression of the endogenous E 2 -inducible PI-9 gene in MCF-7 cells. The serpin PI-9 is a tumor lethality factor (34 -36) whose induction by estrogens enables breast cancer cells to evade apoptosis induced by the immune cells, cytotoxic T lymphocytes and natural killer cells (21,23,28). PI-9 also inhibits tumor necrosis factor-␣ (TNF-␣) Fas and TRAIL (TNF-related apoptosisinducing ligand)-mediated apoptosis (38,39). Induction of PI-9 results from direct binding of E 2 -ER␣ to EREs and ERE half-sites (19,40,41). TPSF (IC 50 ϭ 0.2 M) potently inhibited E 2 -ER␣-stimulated induction of PI-9 mRNA ( Fig. 2A).
To examine the ability of TPSF to inhibit E 2 -ER␣ induction of a gene regulated by tethering of E 2 -ER␣ through DNAbound transcriptional regulators, we tested the effect of TPSF on induction of cyclin D1 mRNA. Cyclin D1 is thought to contribute to the growth of MCF-7 and other breast cancer cells (42)(43)(44) and to tamoxifen-stimulated growth of breast cancer cells (45). As previously reported (43,46), E 2 -ER␣ stimulated a 2-3-fold increase in cyclin D1 mRNA that was blocked by 10 M TPSF (Fig. 2B). The data demonstrate that TPSF inhibits E 2 -dependent gene expression through mechanisms that include direct binding of ER␣ to EREs and through tethering of ER␣ to DNA-associated transcription regulators.
Specificity of TPSF inhibition of E 2 -ER␣-mediated gene expression was evaluated by comparing the ability of TPSF to inhibit gene expression mediated by NF-B (Fig. 2C) and by other steroid receptors (Fig. 2D). The NF-B-regulated IL-8 gene was used to test specificity because many regulators and pathways including IB and other kinases, the ubiquitin/proteasome pathway, and nuclear/cytoplasmic shuttling all influence NF-B activity (47). Analyzing the effect of TPSF on NF-B is a good way to test whether TPSF acts as a promiscuous inhibitor targeting diverse cell proteins and pathways. 30 M TPSF had no effect on the NF-B-mediated induction of IL-8 mRNA by TNF-␣ (Fig. 2C). In the same breast cancer cells where TPSF inhibited E 2 -ER␣ induction of PI-9 mRNA ( Fig. 2A, IC 50 ϭ 0.2 M), a Ͼ100-fold higher concentration of TPSF had no effect on NF-B-mediated induction of IL-8 mRNA by TNF-␣ (Fig. 2C). To further evaluate the specificity of TPSF, we compared the ability of TPSF to inhibit ER␣ to its effect on the AR-and GR-mediated activation of stably transfected reporter genes. AR was assayed in HeLa cells stably transfected to express AR and a PSA-luciferase reporter. TPSF only very weakly inhibited dihydrotestosterone-AR-mediated induction of the PSA-Luc reporter (IC 50 ϭ 33 M, Fig. 2D). GR was assayed in T47D cells stably transfected to express GR and an MMTV-Luc reporter (25). TPSF weakly inhibited GR activation of the MMTV-luc reporter in T47D cells (IC 50 ϭ 10 M). Although T47D cells contain substantial levels of the progesterone receptor (PR), cross-talk between ER␣ and PR makes them unsuitable for assaying inhibitor specificity using PR. The ER antagonist faslodex/fulvestrant/ICI 182,780 inhibited PR induction of the endogenous alkaline phosphatase gene (supplemental Fig. S1). TPSF did not inhibit NF-B, and concentrations of TPSF required to inhibit AR (33 M) and GR (10 M) are far higher than the 0.2 and 0.4 M TPSF required to inhibit the endogenous PI-9 gene and the stably transfected (ERE) 3 -Luc reporter. At low concentrations, TPSF is a relatively specific ER␣ inhibitor.
TPSF Inhibits E 2 and OHT-induced Gene Expression in Tamoxifen-stimulated MCF7ER␣HA Cells-Development of resistance to tamoxifen and other therapeutics that target ER␣ and estrogen production results in treatment failure in both primary and metastatic breast cancer. Recent studies show that tamoxifen-resistant breast cancer cells that retain dependence on ER␣ for growth lose their dependence on SRC3 and other p160 coactivators for E 2 -ER␣-mediated gene transcription (22,48). We explored the ability of TPSF to inhibit E 2 and OHT-dependent gene expression in tamoxifenresistant cells that are less dependent on p160 coactivators for transactivation.
MCF7ER␣HA cells are an MCF-7-breast cancer cell line engineered to increase ER␣ expression in response to Dox (20,22). In Dox-induced MCF7ER␣HA cells overexpressing ER␣, tamoxifen and OHT are potent ER␣ agonists (22,23) and increase ER␣-mediated gene expression independent of SRC3 (22). Because OHT stabilizes ER␣ against degradation whereas E 2 down-regulates ER␣ (21,23), ER␣ levels are ϳ4 times higher in OHT-treated MCF7ER␣HA cells than in cells treated with E 2 . The elevated level of ER␣ in OHT-treated MCF7ER␣HA cells compared with cells treated with E 2 renders OHT more effective than E 2 in inducing PI-9 gene expression and more difficult to inhibit. 10 M TPSF inhibited E 2 -ER␣ (Fig. 3A) and OHT-ER␣ induction of PI-9 mRNA (Fig. 3B). This indicates that TPSF is an inhibitor of both E 2 -ER␣ and OHT-ER␣-mediated gene expression in cells where tamoxifen is a full agonist.

TPSF Inhibits Estrogen-dependent Growth of MCF-7 Cells and Exhibits Low Toxicity in ER␣ Negative MDA-MB-231
Cells-To determine whether TPSF specifically inhibits ER␣dependent growth of breast cancer cells with minimal nonspecific cell toxicity, we tested TPSF inhibition of cell growth in MCF-7 cells and ER␣-negative MDA-MB-231 human breast cancer cells. Compared with MCF-7 cells in estrogendepleted medium, both 1 and 10 pM E 2 stimulated a 4 -5-fold increase in cell number after 4 days (Fig. 4 and data not shown). TPSF elicited a dose-dependent inhibition of estrogen-dependent growth of MCF-7 cells (IC 50 ϭ 2 M) and completely blocked E 2 -dependent growth at 7.5 M (Fig. 4,  filled circles). However, TPSF did not inhibit E 2 -independent cell growth (Fig. 4, compare 7.5 and 10 M TPSF (filled circles) to no E 2 or TPSF (open circle)). TPSF did not inhibit growth of ER␣-negative MDA-MB-231 cells at all concentrations, including 30 M (Fig. 4, filled triangles). To rule out the possibility that MDA-MB-231 cells are unusually resistant to TPSF or other ER␣ inhibitors, we compared the effects of TPSF and OHT on the growth of MDA-MB-231 cells. TPSF was less toxic to ER-negative MDA-MB-231 cells than OHT (supplemental Fig. 2). The results suggest that TPSF specifically inhibits ER␣-mediated growth of breast cancer cells with low nonspecific toxicity in ER␣-negative cells.  Medium was replaced after 2 days, and cells were assayed with MTS after a total of 4 days. Cell number was determined using a standard curve of cell number versus absorbance based on plating a known number of cells and assaying using MTS. Each data point is the average of 8 wells Ϯ S.E. The percentage of cells present after 4 days with E 2 and without TPSF was set equal to 100. By curve-fitting in Sigma Plot, the IC 50 for inhibition of E 2 -dependent growth of MCF-7 cells by TPSF was 2 M.

TPSF Inhibits Anchorage-independent Growth of MCF-7
Cells-The capacity for anchorage-independent growth is a hallmark of cancer cells. Growth in soft agar is often used to evaluate anchorage-independent growth of human breast cancer cells. We tested the ability of TPSF to inhibit E 2 -stimulated growth of MCF-7 cells in soft agar. MCF-7 cells grown in medium containing E 2 formed large colonies after 16 days (Fig. 5, E 2 ). The addition of 10 M TPSF completely inhibited growth of MCF-7 cells in soft agar (Fig. 5, E 2 ϩTPSF). When colonies in equal areas of the soft agar plate were counted, the E 2 -treated plate had 33 colonies Ͼ0.5 mm in diameter, whereas there were no colonies Ͼ0.5 mm in diameter in the E 2 and TPSF-treated plate. The data indicate that TPSF inhibits estrogen stimulation of anchorage-dependent (Fig. 4) and anchorage-independent (Fig. 5) growth of breast cancer cells.
TPSF Inhibits E 2 -ER␣-dependent Growth of Tamoxifenresistant Breast Cancer Cells-The ability of TPSF to inhibit E 2 -ER␣-dependent cell growth was tested using human cell models of tamoxifen-resistant breast cancer. ZR-75 cells are usually reported as partially tamoxifen and OHT-resistant (49 -51), whereas BT474 cells are fully tamoxifen-resistant and contain amplified expression of HER2 and AIB1 (52,53). TPSF inhibited E 2 -ER␣-dependent growth of BT474 and ZR-75 cells with near maximal inhibition at 5 M TPSF (Fig.  6). Because TPSF has minimal nonspecific toxic effects, cell numbers after TPSF treatment were not 0 and represented cells plated at day 0 plus E 2 -ER␣-independent cell growth over the 4 days. TPSF IC 50 values were 0.9 M for slow-growing ZR-75 cells and 1.6 M for BT474 cells. The lower levels of ER␣ in ZR-75 compared with MCF-7 cells (54) may be responsible for the greater potency of TPSF in ZR-75 cells. Some tamoxifen-resistant breast cancers regress after tamoxifen withdrawal, suggesting that tamoxifen stimulates tumor growth (45,(55)(56)(57). The MCF7ER␣HA cell line is a model for tamoxifen-stimulated breast cancer, where tamoxifen and OHT act as full agonists (Fig. 3) (21, 23). In MCF7ER␣HA cells treated with Dox, overexpression of ER␣ increased E 2independent ER␣-mediated cell growth, which was modestly increased by 1 pM E 2 with and without 5 M OHT and was inhibited by 5 M TPSF (supplemental Fig. 3). Table 1 summarizes the effect of TPSF on gene expression and cell growth.
TPSF and TPBM Have Different Modes of Action-Our data show that TPSF is a potent and selective inhibitor of ER-stimulated gene expression and breast cancer cell growth. We, therefore, began to assess how TPSF might exert its actions. We used our fluorescence anisotropy microplate assay (19,58,59) to compare the ability of TPSF and TPBM to inhibit binding of purified ER␣ to a fluorescein-labeled consensus ERE (flcERE). When polarized light excites the flcERE, most of the emitted light is depolarized because of rapid rotational diffusion of the flcERE that results in its position being largely randomized at the time of emission. Binding of the larger ER␣ protein to the flcERE slows rotation of the flcERE, increasing the likelihood that the complex is in the same plane at emission and excitation. Interaction of ER␣ with the flcERE increases fluorescence polarization/fluorescence anisotropy.   50 values for TPSF inhibition of cell growth were calculated by curve-fitting using Sigma Plot. Although some portion of ZR-75 cell growth is likely ER␣-independent, to calculate the IC 50 using Sigma Plot, we used the conservative assumption that all cell growth beyond the 2000 ZR-75 cells plated was E 2 -ER␣-dependent growth and, thus, subject to inhibition by TPSF.

TABLE 1 Summary of IC 50 values (M) for inhibition of gene expression and growth of ER␣-positive and ER␣-negative human breast cancer cells
IC 50 values were calculated from data in Figs. 1, 2, 4, and 6 by curve-fitting using Sigma Plot.
We compared the ability of TPBM and TPSF to inhibit binding of ER␣ to the flcERE. Consistent with our recent report (19), TPBM inhibited binding of E 2 -ER␣ to the ERE (Fig.  7A). Surprisingly, even at 30 M, TPSF had no effect on binding of E 2 -ER␣ to the flcERE (Fig. 7A). Thus, in a direct in vitro assay containing only E 2 -ER␣, the flcERE, TPSF did not inhibit binding of ER␣ to an ERE.
We next compared the effects of TPSF and TPBM on the intracellular levels of E 2 -ER␣ in MCF-7 cells. TPBM at 5-20 M had little or no effect on the level of E 2 -ER␣. In contrast, TPSF elicited a concentration-dependent reduction in E 2 -ER␣ levels, with 10 M TPSF decreasing the level of E 2 -ER␣ by ϳ4-fold (Fig. 7B). TPSF also reduced E 2 -ER␣ levels in T47D breast cancer cells (Fig. 7C). Because TPSF had very little or no effect on the levels of AR and GR (supplemental Fig. 4), TPSF selectively down-regulates the level of ER␣. The results indicate that TPBM and TPSF have distinct modes of action and that TPSF is not simply a more potent version of TPBM.
TPSF Does Not Alter the Level of ER mRNA-TPSF might reduce ER␣ levels by decreasing transcription or by destabilizing ER␣ mRNA. To test for the effects of TPSF at the mRNA level, we examined the effect of TPSF on ER␣ levels in HeLa cells that stably express ER␣ mRNA from a CMV promoter. TPSF retained the ability to down-regulate ER␣ protein from the CMV promoter and from the 2-kb ER␣ mRNA coding region that lacks ϳ4 kb of 5Ј-and 3Ј-untranslated region (Fig.  8A). TPSF had no effect ER␣ mRNA levels in MCF-7 cells (Fig. 8B). Taken together the results suggest that TPSF downregulates ER␣ protein levels through mechanisms that are independent of the level of ER␣ mRNA.
The Proteasome Inhibitor MG132 Blocks the Down-regulation of ER␣ by TPSF-To further examine the effects of TPSF, we determined the time course of TPSF down-regulation of ER␣. Consistent with a TPSF-induced increase in proteasome-dependent degradation of ER␣, there was a progressive decrease in ER␣ protein levels after 6 -8 h (Fig. 9A). To examine this further we tested the ability of the proteasome inhibitor MG132 to block the effects of TPSF. Compared with E 2 alone, TPSF reduced ER levels, and the TPSF-mediated reduction in ER␣ levels was completely blocked by MG132 (Fig.  9B). Efforts to examine the effect of TPSF on ubiquitination of ER␣ in MCF-7 cells were complicated by the use of endogenous untagged endogenous ubiquitin and because E 2 downregulates ER and influences its degradation. The data suggests that much of TPSFs effectiveness as an ER inhibitor resides in its ability to enhance proteasome-mediated degradation of ER␣.
TPSF Does Not Enhance ER␣ Degradation by Reducing the Level of the Muc1 Oncoprotein-Kufe and coworkers (60) reported that the cytoplasmic domain of the Muc1 oncoprotein binds ER␣ and stabilizes ER␣ against degradation, which contributes to enhanced ER␣ transactivation and the estrogendependent growth of breast cancer cells. Although the mechanism by which Muc1 influences ER degradation is unknown, knockdown of Muc1 with RNAi enhanced degradation of ER␣ and inhibited ER␣-mediated transactivation and growth of ER-positive breast cancer cells (60). Because the effects of TPSF and RNAi knockdown of Muc1 protein are similar, we tested whether TPSF influenced the level of the Muc1 cytoplasmic domain. Using the same antibody used by Kufe and coworkers (60), TPSF did not alter the level of the Muc1 cytoplasmic domain (supplemental Fig. 5), indicating that the re-FIGURE 7. Different modes of action of TPSF and TPBM. A, TPSF does not inhibit binding of E 2 -ER␣ to the flcERE. Fluorescence anisotropy microplate assay was performed as described (19) in the presence of increasing concentrations of TPSF (solid bars) and 5 M TPBM (hatched bar). Consistent with our detailed dose-response study (19), 5 M TPBM inhibited binding of TPBM to the flcERE by ϳ60%. Data were plotted with the change in anisotropy for binding of E 2 -ER␣ to the flcERE in the absence of small molecule inhibitors (open bar) set to 100% (actual anisotropy: flcERE, 44 mA units; E 2 -ER␣-flcERE, 81 mA units). Data are the average ϩ S.E. of four experiments. The difference between 5 M TPSF and the control (no inhibitor) was not significant (p Ͼ 0.05). The data for 5 M TPBM were significantly different from both the control and from 5 M TPSF (p Ͻ 0.01 using Student's t test) B, TPSF decreases ER␣ levels. MCF-7 cells were cultured in 5% CD calf serum for at least 2 days and maintained in the absence or presence of E 2 and the indicated concentrations of TPSF or TPBM for 24 h and analyzed for ER␣ by Western blot using 8 g of protein/lane with actin as internal standard. Data are from the Western blot shown and two additional Western blots from independent experiments and are presented as the mean Ϯ S.E. Quantitation of ER␣ and actin was by PhosphorImager analysis. The value for ER␣/actin in the absence of E 2 was set equal to 1. C, T47D cells were maintained as described under "Experimental Procedures," maintained in the absence or presence of E 2 and the indicated concentrations of TPSF, harvested, and analyzed by Western blot as described for panel B. duction in ER␣ levels elicited by TPSF was likely independent of the level of Muc1.

DISCUSSION
Specificity and Toxicity of TPSF-An optimal small molecule inhibitor of E 2 -ER␣ action and growth of breast cancer cells will exhibit specificity for ER␣ and low nonspecific toxicity. Independent testing of TPSF at concentrations up to 10 M against a panel of 60 cancer cell lines at the National Cancer Institute Developmental Therapeutics Program demonstrated that TPSF is generally not toxic to cancer cells (testing was terminated because Ͻ6 of the 60 cell lines showed 50% inhibition of cell growth at 10 M TPSF). In agreement with this, we provide evidence that TPSF selectivity targets E 2 -ER␣-dependent cell growth, with little effect on ER␣-independent cell growth. After 4 days of treatment, E 2 increased MCF-7 cell numbers by ϳ4-fold, which corresponds to a doubling time of ϳ1 day with E 2 and ϳ2 days without E 2 . The number of cells treated with 7.5 and 10 M TPSF was similar to that seen without E 2 . In addition, studies using ER-negative MDA-MB-231 cells showed 30 M TPSF was not toxic. The ability of 10 -20 M OHT to induce apoptosis of MDA-MB-231 cells suggests that these cells are not especially resistant to nonspecific toxic effects. Over several decades tamoxifen has displayed an excellent safety profile in humans. The toxicity of OHT is used only to demonstrate that MDA-MB-231 cells remain susceptible to cell death and that the failure of TPSF to damage the cells is therefore due to low toxicity rather than resistance of these cells to cell death.
Several lines of evidence support the specificity of TPSF for ER␣. For example, NF-B is regulated by a variety of signaling mechanisms that include the ubiquitin/proteasome pathway, nuclear/cytoplasmic shuttling, IB, and other kinases and acetylases (47). In MCF-7 cells, TNF-␣ activation of NF-B in MCF-7 cells increases IL-8 mRNA levels by ϳ50-fold. The absence of an effect of 30 M TPSF on TNF-␣ induction of IL-8 mRNA suggests that TPSF does not exhibit nonspecific effects on these diverse cell pathways.
TPSF specificity for ER␣ was also demonstrated relative to other steroid receptors. TPSF strongly down-regulated the level of ER␣ but had very little or no effect on the levels of AR and GR. TPSF was a more potent inhibitor of transactivation by ER␣ than by AR or GR. We have identified other compounds that inhibit AR and GR under the same assay conditions (data not shown), suggesting that the failure of low concentrations of TPSF to inhibit transactivation by AR and GR was not due to assay conditions. The partial inhibition of GR and AR by higher concentrations of TPSF will not impede future animal or human studies. Two recently described AR inhibitors being tested for prostate cancer therapy, harmol and pyrvinium (61), were strong inhibitors of GR but were used with some success as AR inhibitors in studies in mice (61). Mifepristone (RU-486), a classical PR antagonist that also inhibits GR, has been used in long-term clinical studies without significant GR-related pathology (62,63).
Our initial ER inhibitor, TPBM, has proven useful as a selective inhibitor of the binding of E 2 -ER␣ to cellular genes (64,65). The identification of a new coactivator binding surface on AR using moderate potency (IC 50 ϳ50 M) small molecule inhibitors of AR selected by screening (66) that are unrelated to TPSF also supports the utility of small molecule inhibitors as probes to understand the mechanisms of steroid receptor action.
Inhibition of Gene Expression by TPSF-ER␣ activates gene expression by direct binding to ERE-related DNA sequences and by tethering to DNA-associated transcription factors.
Our studies indicate that both of these mechanisms are inhibited by TPSF. TPSF inhibited the induction of PI-9 mRNA by E 2 -ER␣ and by OHT-ER␣. PI-9 gene expression induced by E 2 -ER␣ results from binding to two adjacent ER binding sites in the PI-9 promoter region (40,41). Induction of PI-9 may be a mechanism by which estrogens enable breast cancers to evade immune surveillance and apoptosis (21,23). PI-9 inhibits granzyme B and cytotoxic T lymphocyte and natural killer and cell-mediated apoptosis of target cancer cells (21,23,28) and caspase 8-dependent apoptosis induced by TNF-␣ family members (38,39). Expression of PI-9 is associated with a poor prognosis in some cancers (34 -36, 67).
Cyclin D1 plays a key role in cell cycle progression and is induced by tethering E 2 -ER␣ to transcription factors bound at  . The proteasome inhibitor MG132 blocks degradation of ER␣ by TPSF. A, shown is the time course of the effect of TPSF on ER levels. MCF-7 cells were plated as described under "Experimental Procedures." After 4 days in MEM ϩ 5% 1ϫ CD-FBS, the medium was replaced with medium containing 10 nM E 2 with or without 10 M TPSF. Cells were harvested at the indicated times, extracts prepared, and ER␣ protein levels were determined by Western blot as described under "Experimental Procedures." B, MG132 reverses the down-regulation of ER by TPSF. Cells were treated as in panel A and maintained for 24 h in medium containing 10 nM E 2 with or without 10 M and 10 M MG132. Preparation of cell extracts and Western blotting were as described under "Experimental Procedures." SP1 sites (46). Cyclin D1 induction is proposed to play a role in estrogen-dependent growth of breast cancer cells (42,43,45). Consistent with TPSF as a broad spectrum ER␣ inhibitor, 10 M TPSF abolished E 2 induction of cyclin D1 mRNA but did not reduce the level of cyclin D1 mRNA much below the basal (ϪE 2 ) level. Inhibition of E 2 -ER␣-dependent MCF-7 cell growth by 10 M TPSF is consistent with a role for cyclin D1 in estrogen-stimulated growth of breast cancer cells. Our work extends earlier studies demonstrating that nearly complete loss of cyclin D1 after RNAi knockdown reduces growth of MCF-7 cells in medium containing estrogen (68). Because TPSF specifically targets the ER␣-regulated component of target gene expression and did not influence basal gene expression, TPSF may be a promising new probe to help clarify the role of ER␣ regulation of specific genes in growth of breast cancer cells.
TPSF Is Effective in Tamoxifen-stimulated and Tamoxifenresistant Breast Cancer Cells-Development of resistance to tamoxifen and other endocrine therapies is a multifactorial clinical challenge in the treatment of breast cancer. A therapeutically useful small molecule inhibitor of ER␣ should inhibit the growth of a primary tumor as do tamoxifen and its active metabolite OHT and also inhibit growth of tumor cells with acquired resistance to tamoxifen. Tamoxifen-resistant tumors can be grouped into three broad classes. Some tumors become independent of ER␣ for growth and may be unaffected by therapies that target ER␣. Others tumors remain dependent on E 2 and ER␣ for growth. A third group loses estrogen dependence but requires ER␣ for growth. The mechanisms involved in resistance to endocrine therapy are diverse. For example, ER␣ transactivation in tamoxifen-resistant cell lines may depend on as yet unidentified coregulators or may be independent of the p160 coactivators (22,48).
One proposed mechanism for tumor resistance to antagonists is the overexpression of steroid receptors. Overexpression of AR was suggested as an important mechanism of resistance to endocrine therapy in castration-recurrent prostate cancer (69). A subset of breast cancers that contain high levels of ER␣ are often refractory to tamoxifen therapy (70 -72). In MCF7ER␣HA cells that overexpress ER␣, tamoxifen and OHT are full agonists and induce PI-9 expression. In MCF7ER␣HA cells maintained in the presence of OHT, levels of ER␣ are Ͼ10 times higher than in wild-type MCF-7 cells maintained in the presence of E 2 (21). In cells expressing high levels of ER␣, 10 M TPSF inhibited both E 2 -ER␣ and OHT-ER␣ induction of PI-9 mRNA.
Clinical specimens of tamoxifen-resistant metastatic breast cancer can be difficult to obtain (73). We and others (49) have, therefore, evaluated ER␣ inhibitors using stable breast cancer cell lines resistant to tamoxifen. TPSF inhibited E 2 -ER␣-dependent growth of ZR-75 human breast cancer cells (IC 50 ϭ 0.9 M), whose slow growth is only weakly stimulated by E 2 and are partially resistant to tamoxifen and OHT (49,74). In contrast to OHT, TPSF blocked the growth of MCF7ER␣HA cells that are tamoxifen-resistant because they overexpress ER␣. TPSF inhibited E 2 -ER␣-dependent growth of BT474 cells, which contain amplified HER2 and AIB1 and are fully tamoxifen-resistant in cell culture (49) and in xe-nograft studies (52). Thus, TPSF is effective in cells that become tamoxifen-resistant through different mechanisms.
Small Molecules Inhibitors of ER␣-TPSF is structurally distinct from disulfide benzamide, a zinc chelator that acts outside the ER␣ ligand binding pocket. Disulfide benzamide promotes an ER␣ conformation conducive to the antagonist activity of OHT in tamoxifen-resistant cell lines. However, 5 M disulfide benzamide inhibited the growth of ZR-75 cells by ϳ20% but did not inhibit the growth of tamoxifen-resistant BT474 cells (49). In contrast, growth of both ZR-75 cells and BT474 cells was inhibited by TPSF (IC 50 values ϭ 0.9 and 1.6 M, respectively). Because TPSF does not compete with E 2 for binding to ER in a direct binding assay or in transactivation, TPSF is not a classical antagonist ligand and is distinct from known ER␣ inhibitors.
Fulvestrant is a high affinity ER ligand with nearly pure antagonist activity. Although fulvestrant is used therapeutically to treat advanced breast cancer, its use is limited by the fact that it can require several months for fulvestrant to reach a therapeutic level in serum (75). It has been known for many years that fulvestrant and related compounds, such as ICI 162,380, enhance the degradation of ER␣ (37,76), although the mechanisms are not known. Recent solution of the structure of fulvestrant bound to the ligand binding domain of ER␣ suggests that fulvestrant binding may distort ER␣ structure so that a few hydrophobic amino acids are exposed near the surface, perhaps triggering recognition of ER␣ as a misfolded protein and rapid degradation (8). Although this is an attractive hypothesis, this idea has not been tested in experiments.
Structurally Related TPBM and TPSF Elicit Different Effects-Because ER␣ and other steroid receptors exhibit a high level of conformational flexibility, small molecules can elicit quite different conformations when they interact with ER␣. For example, binding of E 2 or OHT in the ER␣ ligand binding pocket resulted in functionally distinct agonist and antagonist conformations (8). Thus, binding of structurally related ER␣ inhibitors TPBM (19) and the more potent TPSF may cause distinct ER␣ conformations that are associated with different modes of action. The different actions of TPBM and TPSF are illustrated in Fig. 10. TPBM inhibited E 2 -ER␣ binding to ERE DNA in vitro but had no effect on the intracellular level of ER␣. In contrast, TPSF had no effect on binding of E 2 -ER␣ to ERE DNA but elicited a concentration-dependent reduction in ER␣ protein levels. TPSF also reduced the ER␣ protein level in our HeLa-ER␣ cells that stably express FLAG-tagged ER␣ from a cytomegalovirus promoter that drives transcription of the ϳ2000 nucleotide ER␣ cDNA (76) and did not reduce the level of ER␣ mRNA. Thus, at least part of the inhibitory effect of TPSF appears to reflect its ability to downregulate ER␣ protein. E 2 -ER␣ induction of PI-9 mRNA and of the stably transfected (ERE) 3 -Luc reporter is inhibited by TPSF with IC 50 values of 0.2 and 0.7 M, respectively, with only a modest effect on ER␣ levels. It is possible that regulation of some genes is more sensitive to small changes ER␣ levels. Another possibility is that low concentrations of TPSF did not saturate ER␣. Under these conditions, ER␣ may assume a transient TPSF-induced conformation sufficient to alter ER␣ function and inhibit E 2 -ER␣-mediated transactiva-tion at PI-9 and (ERE) 3 -Luc, whereas higher concentrations of TPSF may be required for an effect on levels of ER␣.
In conclusion, TPSF is a potent and specific small molecule inhibitor of ER␣ that blocks ER␣-mediated gene expression and estrogen-dependent growth of tamoxifen-sensitive and tamoxifen-resistant human breast cancer cells. TPSF inhibition of ER␣ is consistent with a role for estrogen induction of cyclinD1 in triggering the growth of breast cancer cells. TPSF represents a novel compound with potential for treating breast cancer and for probing the mechanisms of ER action.