Targeting the regulation of androgen receptor signaling by the heat shock protein 90 cochaperone FKBP52 in prostate cancer cells

Abstract

Drugs that target novel surfaces on the androgen receptor (AR) and/or novel AR regulatory mechanisms are promising alternatives for the treatment of castrate-resistant prostate cancer. The 52 kDa FK506 binding protein (FKBP52) is an important positive regulator of AR in cellular and whole animal models and represents an attractive target for the treatment of prostate cancer. We used a modified receptor-mediated reporter assay in yeast to screen a diversified natural compound library for inhibitors of FKBP52-enhanced AR function. The lead compound, termed MJC13, inhibits AR function by preventing hormone-dependent dissociation of the Hsp90-FKBP52-AR complex, which results in less hormone-bound receptor in the nucleus. Assays in early and late stage human prostate cancer cells demonstrated that MJC13 inhibits AR-dependent gene expression and androgen-stimulated prostate cancer cell proliferation.

Androgens are a major stimulator of prostate tumor growth, and all current therapies act as classic antagonists by competing with androgens for binding the androgen receptor (AR) hormone binding pocket. This mechanism of action exploits the dependence of AR on hormone activation, but current treatment options become ineffective in castrate-resistant prostate cancer (CRPC), although CRPC remains ligand/AR-dependent. Thus, drugs that target novel surfaces on AR and/or novel AR regulatory mechanisms may provide promising alternatives for the treatment of CRPC (reviewed in ref. 1).

The maturation of cytoplasmic steroid hormone receptors (SHR) to a mature hormone binding conformation is a highly ordered, dynamic process that involves multiple chaperone and cochaperone components (reviewed in ref. 2), all of which present potential opportunities for therapeutic intervention. The final mature complex in which the receptor is capable of high affinity hormone binding includes heat shock protein 90 (Hsp90), a 23 kDa cochaperone (p23), and one of a class of proteins (termed FKBPs) characterized by their Hsp90-binding tetratricopeptide repeat (TPR) domain. The 52 kDa FK506 binding protein (FKBP52) associates with receptor–Hsp90 complexes by way of a C-terminal TPR domain and is a specific positive regulator of AR, glucocorticoid receptor (GR), and progesterone receptor (PR) signaling (3–5). FKBP52 is required for normal male sexual differentiation and development in mice as the fkbp52-deficient mice (52KO) display characteristics of partial androgen insensitivity syndrome including dysgenic prostate (4, 6). FKBP proteins are validated targets of immunosuppressive drugs. FK506 (Tacrolimus) is used clinically to suppress the immune system following organ transplantation. FK506 binds within the peptidyl-prolyl isomerase (PPIase) catalytic pocket of a related family member, FKBP12. The chemical groups of FK506 that project out from the PPIase pocket allow the FKBP12-drug complex to bind tightly to and inhibit calcineurin, which ultimately leads to immunosupression (7). Although FK506 binding to a similar region in FKBP52 does not result in immunosuppression, FK506 does inhibit FKBP52-mediated potentiation of SHR function (3) and the drug inhibits LNCaP prostate cancer cell proliferation (8), suggesting that interference with FKBP52 modulation of AR activity may provide a novel route to develop AR inhibitors with unique characteristics.

FKBP52 association with receptor-Hsp90 complexes results in the enhancement of hormone binding (3, 9, 10); yet the mechanism by which this occurs is unknown. Although FKBP52 binding to Hsp90 is required for FKBP52 regulation of AR, whether or not FKBP52 interacts directly with the receptor within the context of the chaperone complex is unclear. Studies with chimeric receptor proteins have localized FKBP52-mediated effects to the receptor ligand binding domain (LBD) (3). The FKBP52 N-terminal FK506 binding domain (FK1) is required for receptor regulation, and functional domain mapping studies demonstrated that the proline-rich loop overhanging the PPIase catalytic pocket may serve as an interaction surface (10). Although PPIase enzymatic activity is not required for receptor regulation by FKBP52, FK506 may disrupt receptor potentiation through disruption of FK1 interactions with the receptor LBD. Based on this model, we attempted to target the FKBP52 interaction and/or regulatory site on the AR LBD with FKBP52-specific inhibitors. Such compounds would not only serve as valuable pharmacological tools for analysis of FKBP52-receptor interactions, but they also may represent a promising therapeutic approach to inhibit AR transcriptional activity.

Results

Identification of Small Molecule Inhibitors of FKBP52-Enhanced AR Function.

A yeast screen of compound library resulted in the identification of two candidate inhibitors (H7 and H8) that specifically inhibited FKBP52-enhanced AR-P723S function (the P723S mutant increases AR sensitivity to FKBP52 potentiation and was used to increase sensitivity in the assay) (Fig. S1). H8, though functional, was disregarded because it was found to be specific for FKBP52-enhanced AR-P723S but had no effect on wild-type AR. In addition to H7, we assessed 28 additional compounds that represented slight chemical modifications of H7 for effects on FKBP52-enhanced wild-type AR, AR-P723S, and GR signaling in the yeast assay. The results and chemical structures for two of the most promising compounds, MJC01 and MJC13, and the original hit compound H7 are shown in Fig. 1. While H7 produced maximal inhibition at 100 μM (Fig. 1A), MJC01 and MJC13 displayed increased potency with maximal inhibition between 5 and 10 μM (Fig. 1 B and C). All of the FKBP52-specific compounds identified displayed at least some general receptor inhibition at high doses. MJC01 generally inhibited AR signaling in the absence of FKBP52 at 100 μM as evidenced by the upward trend in the dose response curve at that dose (Fig. 1B). In contrast, MJC13 was FKBP52-specific up to 100 μM (Fig. 1C), although it did produce some general receptor inhibition at significantly higher concentrations. Neither MJC13 nor MJC01 affected inducible reporter expression indicating that the inhibition observed is not due to general effects on transcription, translation, or protein stability (Fig. S2). MJC13 was slightly more specific for FKBP52-enhanced AR signaling as compared to GR (Fig. 1C). Although MJC01 has a higher potency than H7 it is less selective for FKBP52 as evidenced by the reduction in FKBP52-specific inhibition at 100 μM. Thus, MJC13 is the most promising lead compound displaying little effect on AR in the absence of FKBP52 and displaying more selectivity for AR than other compounds tested.

Fig. 1.

Identification of inhibitors specific for FKBP52-regulated AR transcriptional activity in yeast. Yeast reporter strains expressing wild-type AR in the absence (control, closed circles) or presence (AR, closed squares) of FKBP52, the AR-P723S point mutant in the absence (control, closed circles) or presence of FKBP52 (AR-P723S, closed triangles), and wild-type GR in the absence (control, closed circles) or presence (GR, closed diamonds) were treated with a range of concentrations of the indicated compounds in the presence of DHT. H7 (A) is the original hit identified from the library screens, MJC01 (B) and MJC013 (C) are the current lead compounds, and flufenamic acid (D) is a known AR inhibitor that associates with the BF3 surface. The structures of the molecules are illustrated above each respective graph. The data were normalized to show only effects on FKBP52-enhanced AR function by calculating the percent reduction in the control strain for each data point and adding that back to each data point for both the control and FKBP52 tester strains. Thus, a hormesis-like effect, as seen for MJC01 at 100 μM (B), indicates that the receptor in the absence of FKBP52 was inhibited at the particular drug dose used (FKBP52-independent inhibition).

Some of the compounds tested, including MJC01 (Fig. 1B), differentially affected wild-type AR and AR-P723S. Interestingly, P723 is within the recently characterized BF3 surface, and the MJC molecules are structurally similar to the BF3-binding fenamic acid derivatives. Fig. 1D demonstrates that the BF3-binding AR inhibitor flufenamic acid displayed no FKBP52-specific effects. Thus, although structurally similar to the fenamic acids, our compounds are functionally distinct. However, the structural similarity to fenamic acids and differential effects on AR and AR-P723S suggest the AR BF3 surface as the possible target site. SHR amino acid sequence alignments identified six amino acid residues (L805, C806, I842, K845, R840, F673) within the AR LBD that are conserved in the FKBP52-regulated receptors, PR and GR, but differ in the FKBP52-insensitive mineralocorticoid receptor (MR). Analysis of the AR LBD crystal structure revealed that these residues comprise a surface region that overlaps with the recently described AR BF3 surface (11) (Fig. S3). Interestingly, multiple residues on this surface, including C806, R840, I842, R846, and P723, have been found mutated in prostate cancer and/or androgen insensitivity syndrome (AIS) patients (McGill Androgen Receptor Gene Mutations Database, http://androgendb.mcgill.ca/). In addition, mutation of P723, within the BF3 surface, results in a receptor that is hypersensitive to FKBP52 potentiation (4). To assess the impact of the additional residues on FKBP52 regulation of AR function we mutated each of the residues and assessed the mutant receptors for their ability to respond to FKBP52 potentiation in yeast reporter assays. We identified two additional mutations, F673P and C806Y, which resulted in AR hypersensitivity to FKBP52 potentiation (Fig. S3). As highlighted in Fig. S3, F673 contacts P723 within the BF3 surface and C806, although not a surface residue, is buried directly below p723 and F673. Thus, the BF3 surface, particularly the region containing F673 and P723, defines a putative FKBP52 interaction and/or regulatory surface.

We did not observe direct interaction between MJC13 and FKBP52. In addition, none of the compounds tested were able to compete with DHT for binding the AR LBD or with SRC2-3 peptide for binding AF2 at relevant concentrations (Fig. S4). In the absence of data directly demonstrating interaction with the AR LBD we performed in silico docking simulations to predict the possible orientation of the molecules on the BF3 surface (Fig. S5). Both MJC01 and MJC13 make extensive nonpolar contacts with residues P723, F673, L830, and Y834 on the BF3 surface. The poses resemble that of flufenamic acid in its AR complex structure (PDB ID code 2PIX). It is clear that the poses shown are one of many that are possible and these simulations should be viewed with caution. However, the poses with the highest docking scores all contained contacts with and/or around the P723 and F673 residues of AR.

Compounds Effectively Target FKBP52-Enhanced AR Signaling in Mammalian Cells.

The compound library screen and subsequent structure activity relationship (SAR) analysis were performed in yeast assays. To assess the effects of the compounds in higher vertebrate model systems, we first tested the compounds for their ability to inhibit AR signaling in MDA-kb2 cells (Fig. 2 A and B). This cell line contains a stable androgen-responsive luciferase reporter and serves as a rapid assay for assessing AR inhibition (12). MDA-kb2 cells were treated with a range of concentrations of the indicated compounds for 20 hr and assessed for both cell viability and AR-dependent expression of the luciferase reporter. The compounds displayed no cellular toxicity as assessed by ATP quantitation (Fig. 2A). However, H7, MJC01, and MJC13 all inhibited AR-mediated expression of the luciferase reporter with half maximal inhibitory concentrations (IC₅₀) of 24.56, 1.79, and 3.60 μM, respectively (Fig. 2B).

Fig. 2.

Effects of the inhibitors on FKBP52-regulated AR function in mammalian cells. (A and B) MDA-kb2 cells expressing a stably transfected AR-responsive luciferase reporter were treated with 0.2 nM DHT and a range of concentrations of the indicated compounds and assessed for cell viability (A) and AR-dependent expression of a luciferase reporter (B). The IC₅₀ values for the compounds are shown in the legend. (C–E) Luciferase reporter assays in 52KO mouse embryonic fibroblast cells in the presence or absence of FKBP52 were performed. Transfected cells were treated with DHT and a range of concentrations of the indicated compounds and assessed for cell viability (C) and AR-dependent expression of a luciferase reporter (D and E). The IC₅₀ values for MJC01 (D) and MJC13 (E) are shown in the legends. (F) Lysates were prepared from 52KO MEF cells transfected with AR and FKBP52 expression vectors after treatment with DHT and a range of MJC13 concentrations (0, 0.05, 0.1, 0.5, 1, 5, 10, 50, and 100 μM) for 24 hr. Lysates were immunoblotted for AR, FKBP52, and GAPDH (loading control).

Mouse embryonic fibroblasts (MEFs) derived from 52KO mice (5) are the only mammalian system that provides an FKBP52-negative background in which to test the compounds for FKBP52-specific effects. Thus, we established AR-mediated luciferase assays in 52KO MEF cells in the presence or absence of an FKBP52 expression vector and assessed the compounds for cellular toxicity and FKBP52-specific inhibition of androgen-dependent luciferase expression. None of the compounds was cytotoxic up to the maximum soluble concentration of 250 μM, as assessed by trypan blue exclusion (Fig. 2C). MJC01 and MJC13 specifically inhibited FKBP52-enhanced AR-mediated expression of the luciferase reporter gene with IC₅₀ values of 0.62 and 0.45 μM, respectively (Fig. 2 D and E). Consistent with the data obtained in the yeast assays, MJC01 displayed significantly higher FKBP52-independent inhibition of AR function (Fig. 2D) as compared to MJC13. MJC13 also produced general AR inhibition in this system at concentrations above 50 μM. To further evaluate compound specificity, the compounds were assessed for effects on constitutive renilla luciferase expression in the 52KO MEF cells (Fig. S2). Neither MJC13 nor MJC01 affected the constitutive expression of renilla luciferase.

Western blots using lysates prepared from the cells in Fig. 2E showed increasing levels of AR and FKBP52 protein that directly correlated with increasing concentrations of MJC13 (Fig. 2F). The degree of stabilization varied between experiments and one of the more dramatic examples is shown in Fig. 2F. Variable stabilization of Hsp90 and p23 protein levels was also observed but to a lesser degree than that seen for FKBP52 and AR.

MJC13 Prevents Receptor–Hsp90 Complex Dissociation and Nuclear Translocation in Cellular Models of Prostate Cancer.

The effects of MJC13 on the stability of AR and associated chaperones is similar to that observed in the presence of nonhydrolyzable ATP analogues or sodium molybdate, which prevent hormone-dependent receptor–Hsp90 complex dissociation. To test the effects of MJC13 on complex formation and/or hormone-dependent complex dissociation we performed coimmunoprecipitations of FKBP52, AR and Hsp90 in lysates of several androgen-responsive prostate cancer cell lines grown in the presence or absence of hormone and MJC13 (Fig. 3). The ability of FKBP52 to bind the AR-Hsp90 complex was unaffected by drug alone, while addition of hormone alone resulted in complex dissociation (as determined by loss of FKBP52, AR, and Hsp90 coprecipitating together). However, complex dissociation in the presence of hormone was abrogated by the addition of MJC13 to LNCaP, LAPC4, and 22Rv1 cells (Fig. 3 A, C, and D, respectively). In addition, Western immunoblots of fractionated lysates prepared from LNCaP, LAPC4, and 22Rv1 cells grown in the presence or absence of hormone and MJC13 revealed that hormone-induced AR translocation to the nucleus was blocked by MJC13 (Fig. 3 B, D, and F, respectively). MJC13 inhibition of AR nuclear translocation was overcome by high hormone concentrations in LNCaP cells (Fig. 3B), which may reflect the lack of receptor dependence on FKBP52 at high hormone concentrations (3).

Fig. 3.

Effects of MJC13 on AR-Hsp90 complex dissociation and AR nuclear translocation in early and late stage prostate cancer cells. The effects of MJC13 on hormone-dependent AR-Hsp90 complex dissociation and AR nuclear translocation were assessed in LNCaP (A and B), LAPC4 (C and D), and CWR22Rv1 (E and F) cells by coimmunoprecipitation and Western blot, respectively. Lysates from cells grown in the presence or absence of the indicated concentrations of hormone and MJC13 for 24 hr were subjected to immunoprecipitation with either an antibody directed against FKBP52 (A) or AR (C and E) and immunoblotted for the indicated proteins. Fetal bovine serum served as the source of hormone in A. Lysates prepared from cells treated with the indicated concentrations of ligand and MJC13 for 24 hr were also fractionated and immunoblotted for AR in both the cytosol and nucleus (B, D, and F).

MJC13 Blocks AR-Dependent Gene Expression and Proliferation in Prostate Cancer Cells.

The effects of MJC13 on AR-dependent gene expression were assessed by analysis of prostate specific antigen (PSA) expression in LNCaP and VCaP cells. We also assessed the impact of MJC13 on expression of the 51 kDa FK506 binding protein (FKBP51) in these cell lines. FKBP51 has emerged as a potential hormone-dependent cancer biomarker (13) due to its hormone-inducible expression (14). However, unlike PSA, FKBP51 is a component of SHR–chaperone complexes (15) and has recently been shown to promote AR function in LNCaP cells similar to FKBP52 (16). Thus, inhibition of FKBP51 expression by MJC13 may have therapeutic implications. ELISA analysis of PSA secretion from LNCaP cells (Fig. 4A) and VCaP cells (Fig. 4B) demonstrated that MJC13 effectively inhibits PSA secretion from both cell lines. Inhibition of hormone-stimulated PSA secretion from LNCaP cells was more potent as compared to hormone-independent secretion from VCaP cells. However, in the presence of hormone MJC13 inhibited PSA secretion from VCaP cells to a similar degree as in LNCaP. The effects of MJC13 on endogenous levels of PSA and FKBP51 in LNCaP and VCaP cells were assessed by Western immunoblot and densitometry (Fig. 4 C and D). Representative blots for FKBP51, PSA, and the loading control GAPDH are shown (Fig. 4 C and D, Upper). The normalized and averaged densitometry data from three independent experiments demonstrate that MJC13 reduced endogenous FKBP51 and PSA expression in a dose-dependent manner (Fig. 4 C and D, Lower). We also assessed endogenous PSA gene expression and expression of the AR-responsive gene TMPRSS2 by quantitative real time PCR (Q-PCR) in LNCaP and 22Rv1 cells. MJC13 effectively abrogated constitutive expression of both AR-driven genes (Fig. 4E, Left). In contrast, in 22Rv1 cells, MJC13 was only able to block androgen-induced (but not constitutive) PSA expression (Fig. 4E, Right). These data are consistent with the fact that 22Rv1 cells express both full-length AR and a constitutively active, truncated AR lacking the LBD (and so predicted to be insensitive to androgen, FKBP52, and MJC13).

Fig. 4.

Effects of MJC13 on AR-dependent gene expression in early and late stage prostate cancer cells. (A and B) ELISA assays to measure PSA secretion were performed in LNCaP (A) and VCaP (B) cells. Cells were treated with the indicated MJC13 concentrations in the presence (A) or absence (B) of DHT for 24 hr, and PSA levels in the media were quantified. (C and D) Western blots to measure AR-dependent expression of PSA and FKBP51 were performed in LNCaP (C) and VCaP (D) cells. Cells were treated with the indicated concentrations of MJC13 in the presence (C) or absence (D) of DHT for 24 hr, lysed, and lysates were electrophoresed and immunoblotted for FKBP51, PSA, and GAPDH (loading control). The upper panels show representative Western blots. The lower panels represent averaged densitometry data from at least three independent experiments. (E) Left panel: PSA and TMPRSS2 gene expression in LNCaP cells was assessed by Q-PCR. Cells were treated for 24 hr with increasing concentrations of MJC13 in the presence of 10% fetal bovine serum. Data are displayed as expression relative to that of 18S rRNA; right panel: R1881-dependent and independent PSA gene expression in 22Rv1 cells was assessed by Q-PCR. Cells (in the presence of charcoal-stripped serum) were untreated, treated for 24 hr with MJC13 alone, or with 0.5 nM R1881 in the presence and absence of 30 μM MJC13. Data are displayed as PSA mRNA expression relative to that of 18S rRNA.

The effect of these compounds on androgen-dependent prostate cancer cell proliferation was assessed by tritium thymidine incorporation in LNCaP, LAPC4, and 22Rv1 cells (Fig. 5). MJC13 inhibited androgen-dependent cell proliferation at concentrations consistent with those observed to be effective in reporter assays. For comparison, the effect of a known AR antagonist, bicalutamide, which interacts with the hormone binding pocket, was assessed. MJC13 was more potent than bicalutamide in these assays. None of the compounds affected cell proliferation in the absence of hormone.

Fig. 5.

MJC compounds effectively inhibit androgen-dependent prostate cancer cell proliferation. Tritium (tritiated thymidine) incorporation assays were performed on LNCaP (A), LAPC4 (B), and 22Rv1 (C) cells treated with a range of compound concentrations in the presence (closed symbols) or absence (open symbols) of 0.5 nM DHT. The known AR antagonist bicalutamide (circles) was included for comparison with MJC13 (squares). All data are expressed as a percentage with the level of tritiated thymidine incorporation in the absence of compound for each condition set to 100%.

Discussion

Functional domain mapping in yeast suggests that the FKBP52 FK1 domain is critical for regulating SHR function through interaction with the receptor LBD (10). In support of this proposal, we have identified a surface region on the AR LBD that, when mutated, displays increased functional dependence on FKBP52 (Fig. S3). This surface directly correlates with the recently identified BF3 surface (11). Although we do not provide direct evidence for FKBP52 interaction with BF3, the data presented here indicate, at the very least, that FKBP52 can indirectly influence receptor function through this surface. We demonstrate that FKBP52 regulation of receptor function can be blocked by small molecules that are predicted to bind the BF3 surface. We demonstrate that the lead molecule, MJC13, specifically inhibits FKBP52-enhanced AR activity in both yeast and mammalian cell lines (Figs. 1 and 2). MJC13 prevents hormone-induced AR-Hsp90 complex dissociation in the presence of FKBP52, which ultimately results in less receptor translocation to the nucleus (Fig. 3). As a consequence, AR-dependent gene expression and androgen-stimulated proliferation in prostate cancer cell lines are inhibited (Figs. 4 and 5). These data suggest that FKBP52 regulates AR function through the BF3 surface and that FKBP52-mediated receptor potentiation can be inhibited by targeting the BF3 surface with small molecules. FKBP52 potentiation of AR signaling may be of particular relevance in CRPC, where androgen levels are markedly reduced but still effectively stimulate the AR (13).

Although our data suggest that MJC13 binds the AR BF3 surface, efforts to provide direct evidence of this interaction through SPR analysis and cocrystallization have not been informative. These difficulties are common to other molecules known or thought to bind BF3. Providing direct evidence of BF3 binding for some of the fenamic acid-derived AR inhibitors has also proven difficult. In addition to the fact that some of these molecules only weakly bind to BF3 (11), we postulate that these molecules can associate weakly with multiple sites on the LBD at high concentrations. It is also possible that in the absence of the dynamic chaperone-assisted folding cycle the BF3 surface on the purified AR LBD is not in an optimal conformation for MJC13 binding. Nevertheless, multiple lines of evidence suggest that MJC13 inhibits FKBP52-mediated AR function through binding BF3. First, we demonstrated that FKBP52 influences at least a portion of the BF3 surface (Fig. S3), and the molecules identified in our screen that inhibit FKBP52 regulation of receptor function are structurally similar to known BF3 binding molecules. Second, many of the compounds tested in the SAR studies differentially affected AR-P723S as compared to wild-type AR and a few of the molecules were specific for AR-P723S (Fig. 1). Thus, mutations within the putative BF3 binding surface accentuate inhibitor activity. Third, MJC13 effectively blocked hormone-dependent PSA expression but failed to block hormone-insensitive PSA expression in 22Rv1 cells (Fig. 4E, Right). Because MJC13 is predicted to target the BF3 surface of the AR LBD, MJC13 would not be expected to affect the expression of PSA in these cells resulting from a constitutively active AR protein lacking the LBD. Finally, in silico docking simulations support the notion that MJC13 binds BF3 (Fig. S5).

In summary, we have identified a surface region on the AR LBD that, when mutated, results in a greater functional dependence on FKBP52. This motif overlaps with the recently characterized BF3 surface. We have developed a series of small molecules that inhibit FKBP52 regulation of AR function. These agents are predicted to act via binding to the BF3 surface in the AR LBD. The most promising compound, MJC13, inhibits hormone-induced AR-chaperone complex dissociation and nuclear translocation and effectively blocks AR-dependent gene expression in cellular models of prostate cancer. Further studies to characterize the MJC13 binding site, improve compound efficacy, and improve receptor specificity are needed. MJC13 is a novel example of an inhibitor that specifically targets the regulation of SHR function by an Hsp90-associated cochaperone and thus serves as an excellent starting point for development of FKBP52-specific inhibitors to treat hormone-dependent diseases.

Materials and Methods

Yeast Strains and Assays.

β-galactosidase reporter assays (3, 17) were used as a quantitative indicator of AR activity for the yeast compound library screens and for the other experiments described in this study. The basic reporter strains used for wild-type AR, the indicated AR point mutant, and GR assays were based on a W303a genetic background (MATa leu2-112ura3-1 trp1-1 his3-11, 15 ade2-1 can1-100 GAL SUC2) and all contained a URA3-marked steroid receptor-mediated β-galactosidase reporter plasmid (pUCΔs-26X, a kind gift from Brian Freeman, University of Illinois, Urbana-Champaign, IL). Refer to SI Methods for more detailed information on the yeast reporter assays.

Compound Library Screen.

The yeast β-galactosidase reporter assays used to screen the library of compounds were modified to a 96-well plate format (see Fig. S1). The parent strain was deleted for the pleiotropic drug resistance 5 (PDR5) gene to prevent the potential transport of compounds out of the yeast. The control strain contained a wild-type human AR expression plasmid and an empty plasmid, and the tester strain contained an AR-P723S expression plasmid and a human FKBP52 expression plasmid. The assay protocol was designed to identify compounds that specifically reduce signaling from the tester strain and not the control strain (FKBP52-specific inhibition). The use of the AR-P723S mutant in these assays increased the sensitivity of detection as signaling at the hormone doses used in the tester strain depends entirely on the presence of FKBP52. The DHT concentrations used correlated with the EC₅₀ values for DHT in both the control and tester strains. The Diversity Set Library from the Developmental Therapeutics Program of the National Cancer Institute (http://dtp.nci.nih.gov) was used for the screen. This library contains approximately 2,000 structurally characterized compounds representative of a diverse chemical space and derived from a collection of almost 140,000 compounds. Library compounds were assayed in the control and tester strains at an initial concentration of 50 μM. See SI Methods for a more detailed description of the compound library screens.

Mammalian Cell Lines.

LNCaP, LAPC4, and VCaP prostate cancer cells all express endogenous AR and are sensitive to androgens. LNCaP cells are characterized by the presence of the AR T877A mutation. Both AR alleles are wild type in LAPC4 cells. VCaP cells are characterized by endogenous AR gene amplification and, although they can respond to androgens, are also capable of androgen-independent growth. The 22Rv1 cell line was derived from a xenograft that was serially propagated in mice after castration-induced regression and relapse of the parental, androgen-dependent CWR22 xenograft. The AR mutation occurred during the progression to androgen independence. Full-length AR in 22Rv1 is characterized by an in-frame tandem duplication of exon 3 that encodes the second zinc finger of the AR DNA-binding domain (18, 19). 22Rv1 cells also express a constitutively active truncated form of AR lacking the C-terminal hormone binding domain. As a result, 22Rv1 cells can both respond to hormone and display hormone-independent growth. Refer to SI Methods for more detailed information on the cell lines and standard growth conditions.

Luciferase Reporter Assays.

Plasmid transfections and luciferase assays were performed according to standard procedures. Refer to SI Methods for a more detailed description of the transfections and reporter assays.

Cellular Toxicity Assays.

Cellular toxicity was determined using trypan blue (Invitrogen) exclusion or ATP measurements (CellTiter-Glo; Promega). For a more detailed description of the cellular toxicity assays refer to SI Methods.

Cell Proliferation Assays.

Cells were plated in a U-shaped 96-well plate at a density of 3 × 10³ cells/well. After the cells attached they were treated with inhibitor for 1 h followed by the addition of 500 pM DHT. The wells were treated with 20 μl of tritiated thymidine (Isotype- from Perkin Elmer) for 18 hr. Cells were lysed using a Cell Harvester (Micro96 Harvester, Skatron Instruments) and lysates were transferred to filter paper (FilterMAT Cat # 11731, Skatron Instruments) and incubated for 1 hr at room temperature. Samples were diluted in 3 ml of scintillation fluid (Scinti SAFE Econo F (LSC Cocktail) SX-22-5, Fisher) and subjected to scintillation counting. All experimental measurements were performed in triplicate.

ELISA Assays for PSA.

ELISA assays were performed according to manufacturer’s instructions (Alpha Diagnostic International). Refer to SI Methods for a more detailed description of the ELISA assays.

Immunoblots and Coimmunoprecipitations.

Western blots and coimmunoprecipitations were performed according to standard procedures. Refer to SI Methods for a detailed description. The mouse monoclonal antibody directed against FKBP52 (HI52D) was previously described (20). The mouse monoclonal antibody directed against Hsp90 (H90-10) was generously provided by David Toft (Mayo Clinic, Rochester, MN). The polyclonal rabbit antibody directed against AR (N-20; Santa Cruz Biotechnology), the mouse monoclonal antibody directed against glyceraldehyde phosphate dehydrogenase (6C5; Biodesign International), and alkaline phosphatase-conjugated anti-rabbit and anti-mouse secondary antibodies (Southern Biotechnology Associates) were all obtained commercially.