Small Molecule Inhibitors as Probes for Estrogen and Androgen Receptor Action*

Because activated estrogen (ER) and androgen (AR) receptors stimulate cell proliferation in breast and prostate cancer, inhibiting their actions represents a major therapeutic goal. Most efforts to modulate ER and AR activity have focused on inhibiting the synthesis of estrogens or androgens or on the identification of small molecules that act by competing with agonist hormones for binding in the ligand-binding pocket of the receptor. An alternative approach is to implement screens for small molecule inhibitors that target other sites in the pathway of steroid receptor action. Many of these second-site inhibitors directly target ER or AR; others have still unknown sites of action. Small molecule inhibitors that target second sites represent new leads with clinical potential; they serve as novel modulators of receptor action; and they can reveal new and as yet unidentified interactions and pathways that modulate ER and AR action.

Because activated estrogen (ER) and androgen (AR) receptors stimulate cell proliferation in breast and prostate cancer, inhibiting their actions represents a major therapeutic goal. Most efforts to modulate ER and AR activity have focused on inhibiting the synthesis of estrogens or androgens or on the identification of small molecules that act by competing with agonist hormones for binding in the ligand-binding pocket of the receptor. An alternative approach is to implement screens for small molecule inhibitors that target other sites in the pathway of steroid receptor action. Many of these second-site inhibitors directly target ER or AR; others have still unknown sites of action. Small molecule inhibitors that target second sites represent new leads with clinical potential; they serve as novel modulators of receptor action; and they can reveal new and as yet unidentified interactions and pathways that modulate ER and AR action.
Within the large nuclear receptor family, estrogen (ER) 2 and androgen (AR) receptors are unusual in their ability to stimulate cell proliferation. The central roles played by ER␣ and AR in most cases of breast and prostate cancer led to an intense effort to identify agents that modulate receptor activity. The availability of substantial information on the interaction of agonist ligands with ER␣ and AR led to a primary focus on identification of small molecules that act by competing with natural hormones for binding in the ligand-binding pocket of the receptors. This fruitful approach was followed by development of agents that act by inhibiting key enzymes in estrogen synthesis. Although these approaches to development of clinically useful agents remain productive, as shown by recent development of an improved competitive ligand for androgens (1) and an inhibitor of androgen production in prostate tumors (2), their current potential for illuminating novel mechanisms of ER and AR action is limited. Here, we focus on small molecules modulators of ER and AR activity that act outside of the ligand-binding pocket. Some of the sites targeted in attempts to antagonize ER action in breast cancer are illustrated in Fig. 1. Research summarized elsewhere in this minireview series describes some of the array of macromolecular interaction partners that can influence receptor activity (3); these include DNA-binding sites, proteins that tether receptors to DNA, coactivators, corepressors, chaperones, ubiquitin ligases, and diverse modifiers such as kinases, phosphatases, methylases, and acetylases. These interactions provide a wealth of targets that are only beginning to be exploited by screens to identify small molecules that modulate ER and AR activity.
In considering efforts to identify small molecule modulators of ER and AR, it is important to understand why it has been simpler to target binding of natural hormone ligands than to target other sites critical for receptor activity. The natural ligands are relatively small and bind with very high affinity (low nanomolar to subnanomolar) in a discrete binding pocket whose three-dimensional structure is known. In contrast, most other ER and AR interactions involve relatively large, often low affinity, macromolecular interfaces for which little or no structural information is available. Most current noncompetitive small molecule modulators of ER and AR were identified using screens based on known activities of the receptors such as DNA binding or coactivator binding. Although binding sites for many of these small molecules are as yet unknown, they target well defined biological processes.

Small Molecules That Target AR
The central role of AR in both primary and castration recurrent prostate cancer (4 -7) and the limited effectiveness of synthetic AR antagonists such as hydroxyflutamide and bicalutamide/Casodex, which target the ligand-binding pocket of AR, make it an attractive target for development of small molecule inhibitors that target other sites. More than 350 nuclear receptor coregulators have been described (3,8). The first family of coregulators to be described, the steroid receptor coactivators (SRCs), are ϳ160-kDa proteins containing three or four Leu-X-X-Leu-Leu (where X is any amino acid) motifs. The SRCs remain among the most important steroid receptor coregulators. SRC3 and the other SRC coregulators exhibit multiple regulatory functions that go far beyond their interaction with nuclear receptors (3).
In AR and other steroid receptors, agonist binding stabilizes a hydrophobic cleft, AF-2 (activation function 2), above the ligand-binding pocket (9,10). In most steroid receptors bound to agonists, including AR, LXXLL motifs in SRCs bind in this hydrophobic cleft. For AR, N/C-terminal interaction that results from AR N-terminal FQNLF motif binding to AF-2 competes with binding of coactivator LXXLL motifs (11,12). The coregulator MAGE-11 (melanoma antigen gene product 11) specifically binds the AR FXXLF motif, thereby * This work was supported, in whole or in part, by National Institutes of Health Grant RO1 DK071909. This work was also supported by a Bridge grant from the Endocrine Society. This is the fifth article in the Thematic Minireview Series on Nuclear Receptors in Biology and Diseases. This minireview will be reprinted in the 2011 Minireview Compendium, which will be available in January, 2012. 1 To whom correspondence should be addressed. E-mail: djshapir@illinois. edu. 2  increasing accessibility of AF-2 to coactivators. MAGE-11 also directly binds SRC2 and other coactivators, tethering them to AR (13,14). Researchers are only beginning to target the AR-specific interaction surfaces revealed by these studies. The unique ability of AR to bind larger motifs such as FXXLF and WXXVW in phage display libraries (15) suggested an approach to selectively targeting AR. Gunther et al. (16) re-evaluated a library of coactivator-binding inhibitors originally tested on ER (17). Their idea was that pyrimidines containing large aromatic substituents would retain the ability to bind AR but not ER␣. Using a luciferase reporter assay in HEC-1 human endometrial cancer cells, they compared the ability of these compounds to inhibit 17␤-estradiol (E 2 )-ER␣dependent expression of an estrogen response element (ERE)luciferase reporter and dihydrotestosterone-AR-dependent expression of a mouse mammary tumor virus promoter-luciferase reporter. They also evaluated the compounds' activity against the AR T877A mutant found in widely used LNCaP cells and ϳ30% of patients with metastatic prostate cancer treated with the nonsteroidal antagonist hydroxyflutamide (18). Some of the peptidomimetic compounds containing multiple aromatic substituents were highly selective for AR and AR T877A (with IC 50 values as low as 2 and 4 M, respectively) and did not inhibit ER␣-mediated transactivation (16). Thus, an approach based on side chain size provides a system for producing peptidomimetics that selectively target binding of SRCs to AR rather than ER␣.
In an unusual screen, Estébanez-Perpiñá et al. (19) soaked small molecules into crystals of the AR ligand-binding domain (LBD) bound to an SRC fragment and looked for small molecules that disrupted the interaction. They identified a novel hydrophobic binding site, which they called BF-3 (binding function 3). This large site is near AF-2 and is at the junction of helix H1 and the H3-H5 loop (Fig. 2). Binding of small molecules to this site reorganizes amino acid side chains in both BF-3 and AF-2, resulting in loss of coactivator binding.
BF-3 represents a novel allosteric binding site for small molecules that alters AR conformation so that coactivator binding is inhibited. The small molecules identified as binding to BF-3 are quite diverse and include the natural hormone triiodothyronine (T 3 ) (Fig. 2), flufenamic acid, and 3,3Ј,5-triiodothyroacetic acid. Although the compounds exhibited only modest inhibitory potency (IC 50 Ͼ 50 M) in a fluorescence polarization assay, they were more effective (IC 50 ϭ 10 -30 M) in reporter gene assays (19). This study is unusual in that detailed structural data of the inhibitor bound to the receptor are available. Although the concentrations of T 3 that bind BF-3 are probably too high to be encountered in biological systems, it remains possible that more potent and selective naturally occurring small molecules allosterically modify coactivator interaction with AR by binding BF-3.
Using a mammalian two-hybrid screen based on disruption of the interaction of liganded AR with the AR-binding protein gesolin, Joseph et al. (20) carried out a screen of ϳ10,000 small molecules and describe two structurally distinct compounds (D36 and D80) that inhibit the interaction of AR and gesolin. These compounds bind AR at an unknown site outside of the ligand-binding pocket and induce a conformational change that inhibits binding of the synthetic androgen R1881 and recruitment of AR to androgen-responsive genes. D36 and D80 inhibit transcription of luciferase reporter genes and several endogenous androgen-regulated genes and androgen-dependent proliferation in cell-based models for antiandrogen-sensitive and anti-androgen-resistant prostate cancer with IC 50 values of 10 -40 M (20).
To identify small molecule inhibitors of AR, Jones et al. (21) used a conformation-based FRET screen with cyan and yellow fluorescent proteins fused to AR. Two compounds, pyrvinium maoate and harmol hydrochloride, inhibited AR activity in the nanomolar range. Pyrvinium and harmol appear to act at distinct unidentified sites outside of the AR ligand-binding pocket. Because they act synergistically and inhibit different steps in AR action, pyrvinium and harmol appear to have dif-  ferent modes of action. Pyrvinium and harmol inhibited the proliferation of LNCaP cells and did not inhibit growth of HEK-293 cells. In a short-term in vivo experiment, pyrvinium combined with bicalutamide caused a 63% reduction in prostate weight (21).
In related work, Jones and Diamond (22) used the conformation-based FRET screen and high throughput microscopy to assess nuclear accumulation of AR. The most potent inhibitors targeted pathways known to influence the activity of AR and many other proteins (actinomycin (RNA synthesis), cucurbitacin (JAK/STAT3), and radicol (HSP90)) (22).
Efforts to target the DNA response element to which AR binds must contend with two major issues. There is substantial sequence heterogeneity in naturally occurring AR DNA response elements (AREs). Also, there is a great deal of overlap in DNA-binding sites for AR and other receptors. Using a fluorescence polarization/anisotropy microplate assay (23) to analyze binding of purified AR and progesterone receptor (PR) to published DNA-binding site sequences thought to exhibit some specificity for binding AR over other receptors, we were unable to identify any DNA elements that exhibited a strong preference for binding AR over PR. 3 Nickols and Dervan (24) developed and extensively characterized hairpin polyamides that target the minor groove of the proposed consensus ARE (5Ј-AGAACAgcaAGTGCT-3Ј). In LNCaP cells treated with 10 M polyamide, binding of AR to androgen response elements was reduced, and androgen induction of the prostate-specific antigen and FKBP5 genes was inhibited by 60 -70%. Analysis of the effects of the polyamide on the entire set of transcripts in LNCaP cells suggested that the heterogeneity of the AR DNA-binding sites and the existence of AR genes regulated by indirect mechanisms will make it impossible for a single polyamide to inhibit all AR-regulated genes (24). However, these compounds may have considerable potential in studies designed to inhibit genes containing specific AREs or families of closely related AREs.

Small Molecules That Inhibit ER-Coregulator Interactions
Most early efforts to identify small molecules that interfere with ER activity and that act outside of the ligand-binding pocket focused on compounds that interfere with coactivator binding (Fig. 1). Katzenellenbogen and co-workers (25) used a fluorescence polarization assay to identify pyrimidines that inhibit binding of a coactivator peptide from SRC1 to ER␣. To optimize these compounds, they synthesized a larger pyrimidine-based library (17) and evaluated the compounds using time-resolved FRET (26). Further studies showed that these compounds inhibited ER␣-mediated activation of a transiently transfected luciferase reporter gene. A different chemical scaffold based on amphipathic benzenes led to inhibitors with a mean inhibitory concentration of 1.7 M (27).
The guanylhydrazone ERI-05 was identified using high throughput screening for inhibitors of ER-coactivator interaction. In mammalian two-hybrid assays, ERI-05 inhibited the interaction of the Gal4 DNA-binding domain-ER␣ LBD fusion and SRC1-, SRC2-, and SRC3-VP16 fusions with an IC 50 of 6 M. However, the 20 M concentration of ERI-05 required to inhibit ER␣ induction of the endogenous pS2 gene in MCF-7 human breast cancer cells approaches the concentration that is toxic to cells (28). Development of these compounds continues, and some show improved potency relative to ERI-05 (29).
In the classical model for antagonist action, binding of 4-hydroxytamoxifen (OHT) to ER results in a conformation in which a segment of helix 12 of ER occupies the coactivatorbinding groove (30). In an intriguing structural study of the ER␤ LBD, a second molecule of OHT was bound to the receptor and occupied the hydrophobic groove of the coactivatorbinding surface (31). Whether this second binding site plays a direct role in OHT antagonism of coactivator binding and the level of OHT required to occupy this site remain to be established.
Although they are not small molecules, coactivator peptide inhibitors display a novel mechanism for inhibition of ER. A nona-arginine TAT peptide tag linked to the SRC2/TIF2 box 2 peptide (to improve cell permeability) resulted in an inhibitory peptide that at 70 M completely blocked E 2 -ER␣ induction of pS2 mRNA in MCF-7 cells (32). In addition to facilitating passage across the plasma membrane, linking peptides to oligoarginine results in their accumulation in the nucleolus. Using an ER␣-cyan fluorescent protein fusion transfected into U2OS cells enabled Carraz et al. (32) to show that binding of these peptides to ER␣ results in sequestration of ER␣ in the nucleolus. This is an example of a second-site inhibitor that acts in part by altering subcellular localization of a steroid receptor.

Small Molecules That Target Binding of ER␣ to DNA
ER␣ binds to its consensus palindromic binding site, 5Ј-aGGTCA-nnnTGACCt-3Ј, and related sequences with relatively high specificity. However, direct repeats containing GGTCA half-sites are found in binding sites for numerous nuclear receptors. Gearhart et al. (33) reported pyrrole-imidazole polyamides that bind to the minor groove of an ERE with a K d of 1 nM. However, these compounds may be more selective for estrogen-related receptors that can bind an extended ERE half-site with high affinity. The DNA topoisomerase inhibitor XR5944 intercalates at 5Ј-TpG:CpA in duplex DNA. XR5944 inhibited ER␣-mediated expression of a transfected luciferase reporter gene in MCF-7 cells with an IC 50 of ϳ1 nM. At the low nanomolar concentrations tested, XR5944 did not inhibit activation of an SP1-regulated reporter (34).
The DNA-binding domain of ER␣ contains two different Cys 4 zinc fingers. 2,2Ј-Dithiobisbenzamidine (DIBA), which was originally identified as releasing zinc from a retroviral Cys 3 -His zinc finger protein, was shown by Wang et al. (35) to cause release of zinc from purified ER␣. MCF-7 xenografts in immune-suppressed mice treated with DIBA showed a dosedependent decrease in growth, with a high dose (30 mg/kg) completely blocking tumor growth (35). Importantly, DIBA was reported to enhance the effectiveness of tamoxifen in tamoxifen-resistant cell lines (36). ChIP showed that DIBA reduced binding of ER to EREs of responsive genes but had no effect on the ability of E 2 or OHT to induce genes that are regulated by tethering of ER to DNA by proteins bound to AP-1 sites. DIBA treatment was also reported to alter the balance between coactivator and corepressor recruitment to DNA-ER complexes, with decreased recruitment of SRC3/ AIB1 (amplified in breast cancer 1) and increased corepressor recruitment (36). Whether DIBA influences ER␣ levels was not reported.
We used a fluorescence polarization screen (23) for small molecule inhibitors of binding of ER␣, PR, and AR to their respective fluorescein-labeled DNA response elements. One small molecule, 8-((benzylthio)methyl)theophylline (TPBM; 8-benzylsulfanylmethyl-1,3-dimethyl-3,7dihydropurine-2,6-dione) (Fig. 3A), was selected for further study (37). TPBM preferentially inhibited binding of ER to the consensus ERE (cERE; IC 50 values for binding to ER␣, PR, and AR of 3, 10, and 8 M, respectively). Because we found that increasing the concentration of E 2 in the binding assays to 10 M had no effect on the ability of TPBM to inhibit binding of ER␣ to the fluorescein-labeled cERE, TPBM does not act by binding in the ligand-binding pocket of ER␣. Unlike DIBA, TPBM does not act as a zinc chelator. In stably transfected cell lines expressing a luciferase reporter linked to EREs or to the PR-and GR-regulated mouse mammary tumor virus progesterone response element, TPBM inhibited expression of the estrogen-responsive reporter with an IC 50 of 12 M and did not inhibit PRand GR-regulated transcription (37).
To analyze the intracellular action of TPBM, we examined its ability to inhibit expression of the estrogen-inducible PI-9 (proteinase inhibitor-9) gene. Induction of the serpin and tumor lethality factor PI-9 results from direct binding of E 2 -ER␣ to EREs and ERE half-sites (38,39). Estrogen induction of PI-9 enables breast cancer cells to evade apoptosis induced by immune cells (40,41). We used MCF7ER␣HA cells, which overexpress ER␣ (42). In these cells, tamoxifen and OHT are potent agonists (40,43). TPBM inhibited both E 2 induction of PI-9 mRNA (IC 50 ϭ 9 M) and OHT induction of the PI-9 gene. Using quantitative RT-PCR to measure the PI-9 mRNA level and semiquantitative ChIP to measure occupancy of the PI-9 estrogen-responsive region, we observed a good correlation between the extent to which TPBM inhibits induction of PI-9 mRNA and the extent to which TPBM reduces E 2 -ER␣ occupancy at the PI-9 gene (37). Thus, the primary mechanism by which TPBM exerts its intracellular action is by decreasing the interaction of ER␣ with regulatory regions of estrogen-responsive genes. It is likely that the interaction of TPBM with ER␣ induces a conformational change in the receptor and that one result of this conformational change is decreased association of E 2 -ER␣ with EREs. TPBM is a useful research tool to probe the role of DNA binding in ER-mediated processes. TPBM has been used to inhibit binding of ER to EREs in several genes (44,45) and to analyze the contribution of DNA binding to stabilization of ER dimers (46).

Down-regulation of ER␣ Levels by a Small Molecule Inhibitor
In testing using ER␣-positive T47D human breast cancer cells stably transfected to express an (ERE) 3 -luciferase reporter, TPSF with an IC 50 of 0.7 M was ϳ15-fold more potent than TPBM. Competitive radiometric binding assays and cell-based inhibition studies done at E 2 concentrations that varied by 500-fold demonstrated that TPSF is not a classical ligand that competes with E 2 for binding in the ER␣ ligandbinding pocket.
In MCF-7 cells, TPSF potently inhibits E 2 -ER␣-mediated induction of the PI-9 gene (IC 50 ϭ 200 nM), which is activated by direct binding of ER␣ to an ERE DNA and the cyclin D 1 gene, which is thought to be induced by tethering ER␣ to other DNA-bound proteins. Because 30 M TPSF had no effect on the NF-B-mediated induction of IL-8 mRNA in MCF-7 cells and because very high concentrations of TPSF were required to inhibit GR-and AR-mediated transcription, TPSF is a selective inhibitor of ER␣ (47).
We tested TPSF inhibition of cell growth in ER␣-positive MCF-7 cells and in ER␣-negative MDA-MB-231 human breast cancer cells. TPSF elicited a dose-dependent inhibition of the estrogen-dependent growth of MCF-7 cells with an IC 50 of 2 M and completely blocked estrogen-dependent growth by 7.5 M. In ER-negative MDA-MB-231 cells, TPSF did not inhibit growth at all concentrations, including 30 M. The capacity for anchorage-independent growth is a hallmark of cancer cells that is often evaluated by growth in soft agar. MCF-7 cells grown in medium containing E 2 formed large colonies. The addition of 10 M TPSF completely blocked formation of MCF-7 cell colonies (47). Thus, TPSF inhibits estrogen stimulation of both anchorage-dependent and anchorage-independent growth of breast cancer cells.
TPSF also inhibits ER␣-dependent cell growth in three models for tamoxifen resistance: OHT-stimulated MCF7ER␣HA cells, which overexpress ER␣; fully tamoxifenresistant BT474 cells, which have amplified HER-2 and AIB1; and partially tamoxifen-resistant ZR-75 cells. Thus, TPSF is effective in cells that become tamoxifen-resistant through different mechanisms.
Although TPBM and TPSF are very similar structurally (Fig. 3A), they have very different effects on ER␣. Whereas TPBM inhibits in vitro binding of E 2 -ER␣ to a labeled ERE in vitro, TPSF does not (Fig. 3B). TPSF strongly reduces ER␣ levels in breast cancer cells, whereas TPBM has little or no effect on the level of ER␣ (Fig. 3C). Because TPSF has very little or no effect on the levels of AR and GR, TPSF is highly selective for down-regulation of ER. The proteasome inhibitor MG132 abolished down-regulation of ER␣ by TPSF (47). Thus, TPSF influences receptor levels at least in part due to its ability to enhance proteasome-dependent degradation of ER␣.
How might we account for the different modes of action of these two closely related small molecules? ER␣ and other steroid receptors exhibit a high level of conformational flexibility, and 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 results in functionally distinct agonist and antagonist conformations (30). Thus, binding of TPBM and the more potent TPSF may cause distinct ER␣ conformations that are associated with different modes of action. Our observation that TPSF down-regulates E 2 -ER␣ more effectively than unliganded ER␣ is consistent with an important role for receptor conformation in TPSF action.
Although several pathways modulate ER␣ degradation, mechanisms by which a small molecule might enhance degradation of ER␣ are poorly defined. Activation of the aryl hydrocarbon receptor by 2,3,7,8-tetrachlorodibenzo-p-dioxin stimulates proteasome-dependent degradation of ER␣ (48). Knockdown of the oncoprotein Muc1, which binds to and stabilizes ER␣, increases ER␣ degradation and inhibits ER␣mediated transactivation and growth of breast cancer cells (49). Fulvestrant/Faslodex/ICI 182,780 is a high affinity antagonist ER ligand that is used therapeutically to treat advanced breast cancer and that enhances degradation of ER␣ (50,51). Recent structural studies suggest the 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.

Future Prospects
The use of small molecules identified by physical or virtual screening to illuminate the actions of steroid receptors is gaining momentum. Although most studies have focused on small molecules that act at known well defined sites in steroid receptor action, an intriguing approach, which is still in its infancy, is to use cell-and organism-based high throughput screens based on inhibition of receptor activity or cell growth. Cell-and organism-based screens can potentially identify any small molecule whose actions influence receptor-mediated transactivation or cell growth. This type of unbiased screen asks the cell to tell us what interactions and pathways are susceptible to targeting by small molecules with a readout of altered receptor-mediated transactivation or cell proliferation. Small molecule inhibitors identified using cell-based screens have the potential to identify novel pathways and interactions that influence receptor activity. By their very nature, it is likely to be both challenging and rewarding to identify these novel sites of inhibitor action.