Distinguishing Androgen Receptor Agonists and Antagonists: Distinct Mechanisms of Activation by Medroxyprogesterone Acetate and Dihydrotestosterone

Abstract

Natural and pharmacological androgen receptor (AR) ligands were tested for their ability to induce the AR NH₂-terminal and carboxyl-terminal (N/C) interaction in a two-hybrid protein assay to determine whether N/C complex formation distinguishes in vivo AR agonists from antagonists. High-affinity agonists such as dihydrotestosterone, mibolerone, testosterone, and methyltrienolone at concentrations between 0.1 and 1 nm induce the N/C interaction more than 40-fold. The lower affinity anabolic steroids, oxandrolone and fluoxymesterone, require concentrations of 10–100 nm for up to 23-fold induction of the N/C interaction. However no N/C interaction was detected in the presence of the antagonists, hydroxyflutamide, cyproterone acetate, or RU56187, at concentrations up to 1 μm, or with 1μ m estradiol, progesterone, or medroxyprogesterone acetate; each of these steroids at 1–500 nm inhibited the dihydrotestosterone-induced N/C interaction, with medroxyprogesterone acetate being the most effective. In transient and stable cotransfection assays using the mouse mammary tumor virus reporter vector, all ligands displayed concentration-dependent AR agonist activity that paralleled induction of the N/C interaction, with antagonists and weaker agonists failing to induce the N/C interaction. AR dimerization and DNA binding in mobility shift assays and AR stabilization reflected, but were not dependent on, the N/C interaction. The results indicate that the N/C interaction facilitates agonist potency at low physiological ligand concentrations as detected in transcription, dimerization/DNA binding, and stabilization assays. However the N/C interaction is not required for agonist activity at sufficiently high ligand concentrations, nor does its inhibition imply antagonist activity.

INTRODUCTION

Androgen receptor (AR) function is required for male sex development in the fetus, virilization at puberty, and maintenance of reproductive function in the adult. Interruption of these processes by pharmacological androgen antagonists or environmental endocrine disruptors can cause incomplete masculinization of the fetus or possibly reduced male fertility later in life (1, 2). Overstimulation of the prostate by androgen agonists may promote prostate cancer (3). Experimental approaches to identify and distinguish AR agonists from antagonists would aid in the classification of environmental and pharmaceutical chemicals since ligand-binding affinity alone does not necessarily reflect biological potency, and transient transcriptional assays can be hampered by the complexity of the systems.

Previous studies from this laboratory identified an AR NH₂-terminal and carboxyl-terminal (N/C) interaction that requires high-affinity androgen binding (4). The androgen-induced N/C interaction is inhibited by the androgen antagonist hydroxyflutamide. These results raised the possibility that an N/C interaction is required for AR agonist activity and that its interruption is a prerequisite for antagonist activity. Similar studies with the estrogen receptor revealed a ligand-dependent N/C interaction that predicted parallel dimerization (5). Recent studies on AR suggest that its N/C interaction is intermolecular and results in the formation of an antiparallel homodimer (6). A feature common to both models is the requirement for high-affinity agonist binding to promote the N/C interaction. In the present report we tested the requirement for the AR N/C interaction in relation to AR dimerization, DNA binding, and transcriptional activity in transient and stable cotransfection assays to distinguish the activities of several natural and pharmaceutical agonists and antagonists. The results suggest that at higher concentrations, certain weak AR agonists such as medroxyprogesterone acetate (MPA) activate AR through a mechanism that does not involve the N/C interaction, although potent agonists capable of AR activation at low ligand concentrations induce the N/C interaction. Furthermore, inhibition of the N/C interaction does not necessarily reflect the activity of an antagonist.

RESULTS

Ligand Binding Affinities

Binding affinities of ligands listed in Table 1 are indicated as apparent equilibrium binding constant (K_d) determined by Scatchard analysis as previously reported, inhibition constant K_i of endogenous AR in rat prostate extracts, or the concentration required for 50% inhibition of [³H]methyltrienolone (R1881) binding to recombinant AR in transfected COS cells (Table 1). Relative competitive binding affinities for [³H]R1881 in transfected COS cells were mibolerone ∼ R1881 > RU56187 > dihydrotestosterone (DHT) ∼ MPA ∼ progesterone > estradiol (E₂) > cyproterone acetate > testosterone ∼ oxandrolone ∼ fluoxymesterone > hydroxyflutamide. Chemical structures of several of the ligands are shown in Fig. 1. Binding of DHT and testosterone was weaker in the COS cell assay when compared with apparent equilibrium binding affinities in tissue cytosols (Table 1) likely resulting from partial metabolism during the 2-h 37 C incubation. Analysis of COS cells transfected with the parent plasmid lacking the AR sequence showed binding only of [³H]progesterone. Lack of binding of [³H]R1881 and [³H]mibolerone in the absence of AR expression suggested that this endogenous binding activity was not due to the progesterone receptor (7–10). Inhibition constants for the anabolic steroids oxandrolone (K_i 62 nm) and fluoxymesterone (K_i 44 nm) were less than that of hydroxyflutamide (K_i 175 nm), but about 100 times greater than the equilibrium binding constant (K_d) for high affinity agonists such as DHT. Competitive binding by MPA and RU56187 were similar to DHT, although the reported K_d values for MPA were slightly greater than that for DHT (Table 1).

Figure 1.

Chemical Structures for MPA, RU56187, DHT, Oxandrolone, and Fluoxymesterone

Table 1.

Summary of Human AR Ligand Binding Affinities, Ligand-Induced N/C Interaction, and Transcriptional Activation and Inhibition

	Kd (nm)	Ki (nm)	Competitive Binding (nm)	Agonist		Antagonist
				N/C	MMTV	N/C	MMTV
				(nm)		(nm)
DHT	0.2–0.5		100	0.1	0.001	–a	–
Mibolerone	0.53		12	0.1	0.001	–	–
Testosterone	0.2–0.5		220	1	0.01	–	–
R1881	0.6		15	1	0.001	–	–
Oxandrolone		62	260	10	1	–	–
Fluoxymesterone		44	300	100	1	–	–
MPA	1.7–3.6		75	–	0.1	1	–
RU56187	0.39		35	–	10	10	100
Estradiol		110	120	–	10	50	500
Progesterone			62	–	100	50	500
Cyproterone acetate			170	–	100	50	100
Hydroxyflutamide		175	460	–	10,000	50	100

The apparent equilibrium binding constant (K_d, nm) is indicated for DHT (62 ), mibolerone (63 ), testosterone (62 ), R1881 (64 ), MPA (28, 42 ), and RU56187 (14 ). The K_i (nm) is indicated for E₂ (1 ), oxandrolone, fluoxymesterone, cyproterone acetate, and hydroxyflutamide. Approximate unlabeled ligand concentrations (nm) that cause 50% competitive inhibition of 5 nm [³H]R1881 binding to recombinant human AR expressed in COS cells were determined as described in Materials and Methods. Ligand concentrations (nm) that induce at least 10-fold agonist activity were determined in the two-hybrid N/C interaction assay in CHO cells (N/C) and transcriptional activation of MMTV-luciferase in CV1 cells (MMTV). Ligand concentrations (nm) for 50% inhibition (antagonist activity) were determined in the two-hybrid N/C interaction assay in CHO cells in the presence of 1 nm DHT (N/C) or with the MMTV-luciferase reporter in CV1 cells in the presence of 0.1 nm DHT (MMTV).

^a

Dashes indicate not detectable.

N/C Interaction

Induction of AR N/C complex formation was determined in a two-hybrid protein assay using Chinese hamster ovary (CHO) cells as previously described (4, 6). Relative ligand potency between 0.1 nm and 1 μm was DHT ∼ mibolerone> testosterone ∼ R1881 > oxandrolone > fluoxymesterone (Fig. 2). No N/C interaction was detected with MPA, RU56187, E₂, progesterone, hydroxyflutamide, or cyproterone acetate up to concentrations of 1 μm. Induction of the N/C interaction did not correlate with relative binding affinities (Fig. 2 and Table 1). The relatively high-affinity ligands, MPA and RU56187, failed to promote the N/C interaction, whereas the lower affinity anabolic steroids, oxandrolone and fluoxymesterone, induced the N/C interaction. The inability of MPA to induce the N/C interaction was not limited to CHO cells, as it was also ineffective in monkey kidney CV1 or COS cells where 5-fold induction of the N/C interaction was observed with 1 nm DHT (data not shown).

Figure 2.

Ligand Dependence of the AR N/C Interaction Determined in a Two-Hybrid Protein Assay in CHO Cells CHO cells were transfected with GALD-H and VPAR1–660 human AR fusion protein expression vectors and the G5E1b-luciferase reporter vector as described in Materials and Methods. Cells were incubated with increasing concentrations of DHT, T (0.1–100 nm), methyltrienolone (R1881), mibolerone (MIB) (0.1–10 nm), oxandrolone (OXAND), fluoxymesterone (FLUOXY) (0.1–100 nm), MPA, RU56187, E₂, progesterone (PROG), hydroxyflutamide (OH-FL), and cyproterone acetate (CA) (0.1–1000 nm) as indicated. Shown are the optical luciferase units, and above the bars, the fold induction relative to the activity determined in the absence of ligand. The data are representative of at least three independent experiments.

Inhibition of the DHT-induced N/C interaction by hydroxyflutamide reported previously (4) raised the possibility that this inhibition may be necessary for and indicative of androgen antagonist activity. Because MPA is a weak AR agonist in vivo (11–13), it was surprising that MPA at concentrations as low as 10 nm blocked the DHT-induced N/C interaction and was about 50 times more potent than hydroxyflutamide as an inhibitor (Fig. 3 and Table 1). RU56187 was a slightly less potent inhibitor of the N/C interaction than MPA, exhibited a high AR equilibrium binding affinity, and is reported to have antagonist activity in vivo (14, 15). Ligands with less inhibitory activity than MPA or RU56187 at concentrations between 50 and 500 nm were hydroxyflutamide, cyproterone acetate, E₂, and progesterone (Fig. 3 and Table 1). The anabolic steroids, oxandrolone and fluoxymesterone, and the potent androgen agonists, DHT, mibolerone, testosterone, and R1881, showed little or no inhibition of the DHT-induced N/C interaction.

Figure 3.

Inhibition of the DHT-Induced AR N/C Interaction Determined in the Two-Hybrid Protein Interaction Assay in CHO Cells CHO cells were transfected with the AR fusion expression vectors and the G5E1b-luciferase reporter vector as described in Materials and Methods and incubated with and without 1 nm DHT or in the presence of 1 nm DHT with increasing concentrations of the indicated ligands (abbreviations as in Fig. 2 legend). Optical luciferase units are shown with the fold induction relative to the activity determined in the absence of ligand indicated above the bars. The data are representative of at least three independent experiments.

To investigate the possibility that MPA induces an AR carboxyl-terminal/carboxyl-terminal (C/C) interaction, we tested GALD-H with VPD-H in the two-hybrid assay. VPD-H contains the VP16 transactivation domain linked as a fusion protein to the AR hinge and steroid-binding domain amino acid residues 624–919 and was used previously to demonstrate lack of a C/C interaction induced by DHT (4). Neither MPA nor DHT induced a C/C interaction in this assay more than 2-fold (results not shown).

Transcriptional Activation

Agonist and antagonist activities were determined in CV1 cells transiently transfected with a mouse mammary tumor virus (MMTV)-luciferase reporter and full-length human AR expression vectors. Ligands with more than 10-fold agonist activity at 0.001 nm were DHT, mibolerone, and R1881 (Fig. 4 and Table 1). Similar induction was achieved by 0.01 nm testosterone, 0.1 nm MPA, and 1 nm oxandrolone or fluoxymesterone. Cyproterone acetate, progesterone, E₂, and RU56187 induced luciferase activity at concentrations between 10 and 100 nm, but transcriptional activity remained low at 100 nm hydroxyflutamide, the latter requiring concentrations of 1–10 μm for agonist activity in this assay (16). Agonist potency, therefore, tended to parallel the ligand-induced N/C interaction. Lack of an N/C interaction induced by MPA is associated with 100-fold higher MPA concentrations necessary for transcriptional activity compared with DHT.

Figure 4.

Ligand-Dependent AR-Mediated Transcriptional Activation of the MMTV-Luciferase Reporter Vector in CV1 Cells CV1 cells were transfected with 0.1 μg pCMVhAR full-length human AR expression vector and 5 μg MMTV-luciferase reporter vector using calcium phosphate as described in Materials and Methods. Transfected cells were incubated with increasing concentrations of the indicated ligands (abbreviations as in Fig. 2 legend). Shown is the fold induction of luciferase activity relative to the activity determined in the absence of ligand. The data are representative of at least three independent experiments.

When transcriptional activity was tested in the same cells (CHO cells) used for the N/C two-hybrid assay but using the MMTV-luciferase reporter, 10- to 100-fold higher MPA concentrations were also required relative to DHT. In CV1 cells transfected with the MMTV-luciferase reporter and the pCMV5 parent plasmid lacking AR sequence, there was no induction of luciferase activity by MPA or any other ligand tested, ruling out the possibility that MPA activity was mediated through an endogenous receptor or altered luciferase expression by a nonreceptor mechanism. A luciferase reporter vector with two copies of the MMTV glucocorticoid response element separated by a 29-bp linker derived from pMTV29VTM (17) and cloned into pT81Luc (18) also had greater agonist activity with DHT relative to MPA (data not shown).

Antagonist activity of the ligands was tested in CV1 cells by coincubation with 0.1 nm DHT. Hydroxyflutamide was the most effective antagonist with about 50% inhibition at 100 nm (Fig. 5). Cyproterone acetate was slightly less effective, and RU56187 had some inhibitory activity but decreased in effectiveness at higher concentrations. Antagonist activity was also observed with increasing concentrations of progesterone and E₂, but none was observed with MPA (Fig. 5) or with the high-affinity agonists or the anabolic steroids (results not shown). Thus, except for MPA, at least partial inhibition of DHT-induced transcriptional activity correlated with inhibition of the DHT-induced N/C interaction.

Figure 5.

AR Antagonist Activity in CV1 Cells CV1 cells were transfected with 0.1 μg pCMVhAR and 5 μg MMTV-luciferase reporter vector as described in Materials and Methods and incubated in the presence and absence of 0.1 nm DHT or in the presence of 0.1 nm DHT with increasing concentrations of the indicated ligands (abbreviations as in Fig. 2 legend). Shown is the fold induction relative to activity determined in the absence of DHT. The data are representative of at least three experiments.

The inability of MPA to induce the N/C interaction but have agonist rather than antagonist activity in transient transcription assays and in vivo (11–13, 19) prompted us to investigate whether two AR mutants that cause severe androgen insensitivity might be selectively activated by MPA in transient cotransfection assays. The mutants, V889M and R752Q, each retain high-affinity equilibrium binding of [³H]R1881 (20–22) but are defective in the N/C interaction (6). V889M had a similar blunted response to both DHT and MPA in the MMTV-luciferase reporter assay in CV1 cells, whereas R752Q required about 10-fold higher concentrations of MPA (10 nm) relative to DHT to induce transcription (data not shown). Thus, neither mutant defective in the N/C interaction was efficiently activated by MPA or DHT, suggesting that these regions of the ligand-binding domain are important in AR activation by both steroids.

The weak in vivo AR agonist activity reported for MPA (11–13, 19) is reflected in MMTV-luciferase assays by the requirement for higher MPA concentrations relative to DHT for reporter gene activation. Nevertheless, the agonist activity of MPA was surprising considering that MPA inhibits the N/C interaction better than most antagonists. We therefore tested a CHO cell line in which the MMTV-luciferase reporter and human AR expression vectors were stably integrated in the genome using pcDNA3.1/Zeo vector with the zeocin gene and the human AR-coding sequence (K. Bobseine and W. R. Kelce, unpublished data). Cell lines such as this were used previously to distinguish agonist activities not detected by transient transfection (24). DHT at 0.1 nm stimulated luciferase activity 2-fold while MPA required a 10-fold higher concentration for similar induction (data not shown). A greater overall response to MPA (10-fold) compared with DHT (6-fold) likely resulted from MPA activation of endogenous glucocorticoid receptor since coincubation with 500 nm hydroxyflutamide inhibited MPA-activated gene transcription about 50% and DHT activity by 95%, but had no effect on induction by dexamethasone (data not shown). The AR-mediated MPA response was therefore similar to that of DHT, but required higher steroid concentrations as observed in the transient assays. The chromatin arrangement of the reporter gene seemed to have little effect on the relative concentration-dependent gene activation by DHT and MPA.

Dimerization and DNA Binding

Because binding of baculovirus expressed full-length AR to androgen response element DNA requires exposure of Sf9 cells to androgen and is inhibited by coincubation with the antagonist hydroxyflutamide (25), we tested the activity of these ligands to promote AR DNA binding in vitro. DNA binding of full-length AR was observed with 50 nm DHT, mibolerone, MPA, oxandrolone, fluoxymesterone (Fig. 6A), or 50 nm R1881 or testosterone (Fig. 6B). Concentrations of DHT or R1881 less than 50 nm reduced AR DNA binding probably due to insufficient saturation of baculovirus-expressed AR (data not shown). AR DNA binding was observed at 1 μm RU56187 or hydroxyflutamide but was barely detectable with 1μ m progesterone, E₂, or cyproterone acetate (Fig. 6).

Figure 6.

Ligand-Dependent DNA Binding of Full-Length Human AR Expressed in Baculovirus Sf9 cells expressing human AR were incubated with increasing concentrations of ligands as described in Materials and Methods. Cells were extracted in high-salt buffer, dialyzed to reduce the salt content, and incubated with ³²P-labeled androgen response element DNA as described in Materials and Methods. Shown is a reproduction of films in which the free ³²P-labeled oligo bands were removed from the bottom. In panels A and B, 50, 250, and 1000 nm of the indicated ligands were added in lanes 1–18 (abbreviations as in Fig. 2 legend). Lane 19 contains extracts from cells left untreated with recombinant virus or ligand. The AR-³²P-oligo complex typically migrates as a double band (indicated with arrows) with the upper more slowly migrating band being predominant. In the absence of ligand the AR-³²P-oligo complex is not detected (25 ).

Dimerization and DNA binding of baculovirus-expressed AR NH₂- and carboxyl-terminal fragments that contain the DNA-binding domain were also shown previously to distinguish androgen agonists and antagonists (25). While the NH₂-terminal and DNA-binding domain fragment AR1–660 (not shown) and the DNA- binding and carboxyl-terminal fragment AR507–919 (Fig. 7, lanes C) each homodimerize and bind DNA independently of hormone, agonists are required for dimerization and DNA binding of the N/C complex and an antagonist such as hydroxyflutamide inhibits this DHT-induced DNA binding (25, 26). Since both fragments contain the DNA-binding domain, dimerization could be mediated by the DNA-binding domain and/or by the N/C interaction. Results of this assay (25) and others (22) nevertheless predicted the AR N/C interaction, which was later confirmed in the two-hybrid interaction assay (4).

Figure 7.

Ligand Dependence of the AR NH₂-Terminal and Carboxyl-Terminal Fragment Dimerization and DNA Binding in DNA Mobility Shift Assays Human AR NH₂-terminal DNA-binding domain fragment AR1–660 (N) and DNA ligand-binding domain-carboxyl terminal fragment AR507–919 (C) were expressed separately in Sf9 cells. Cells expressing C were incubated at the indicated ligand concentrations. After high-salt extraction and dialysis to lower the salt concentration, 20 μg total protein of N or C or 10 μg each of the N- and C-terminal fragments were combined and analyzed in DNA mobility shift assays as described in Materials and Methods. Cells expressing C were incubated with 50 nm DHT, mibolerone (MIB), testosterone (T), and R1881 (shown in panel A) along with 50, 500, and 1000 nm oxandrolone (OXAND) and fluoxymesterone (FLUOXY). R1881 (50 nm) is repeated in panels B and C as a control along with 50, 500, and 1000 nm MPA, RU56187, and cyproterone acetate (CA) (panel B), and the same concentrations of E₂, progesterone (Prog), and hydroxyflutamide (OH-FL) (panel C). Shown are reproductions of films where the free ³²P-oligo band was cut from the bottom. Migration of the N + C and C complexes is indicated by arrows.

Dimerization and DNA binding of the AR fragments were observed at 50 nm DHT, mibolerone, testosterone, and R1881 (Fig. 7A, C+N) but required 0.5–1 μm oxandrolone or fluoxymesterone for similar activity. MPA-induced dimerization and DNA binding of the two AR fragments were somewhat less efficient than DHT but similar to the anabolic steroids (Fig. 7, A and B). Only slightly weaker DNA binding of the N/C hybrids was detected using cyproterone acetate and E₂, whereas 1 μm RU56187 or 1μ m progesterone was required for DNA binding, and essentially no DNA binding was detected with 1 μm hydroxyflutamide (Fig. 7 and Table 2). Binding of the homodimer C fragment alone was most effective with mibolerone, testosterone, and R1881; it was not detected with 50 nm DHT, required 500 nm MPA, and was weak to undetectable with the anabolic steroids and other ligands (Fig. 7, lanes C). Similar high-level AR expression was observed by immunoblot analysis of full-length AR or the AR507–919 fragment after the different hormone treatments and expression levels were independent of the extent of AR dimerization and DNA binding (data not shown).

Table 2.

Summary of AR Stabilization and Half-Times of[ ³H]Ligand Dissociation, AR DNA Binding, and in Vivo Activity of ligands

	Stability (nm)	[3H]Ligand Dissociation Rate (h)	DNA Binding (nm)	In Vivo Activity
DHT	1	3.4	50	Agonist
Mibolerone	1	3.5	50	Agonist
Testosterone	5	1.2	50	Agonist
R1881	1	3.7	50
Oxandrolone	10		500	Agonist
Fluoxymesterone	10		500	Agonist
MPA	100		50	Agonist
RU56187	–a	0.08	1000	Antagonist
E2	–	0.67	1000
Progesterone	1000		1000
Cyproterone acetate	1000		1000	Antagonist
Hydroxyflutamide	–		–	Antagonist

Ligand-dependent AR degradation was determined at 37 °C in COS cells using [³⁵S]methionine-labeled AR as described in Fig. 8. Indicated are the approximate ligand concentrations required to reduce AR degradation to approximately t_1/2 ≥ 5 h at 37 °C. Half-times of [³H]ligand dissociation (h) were determined at 37 °C in COS cells for DHT, testosterone, and R1881 (22 ), and for RU56187 and E₂ as described in Fig. 9. Ligand-dependent AR DNA binding was determined in mobility shift assays as described in Fig. 6. Shown are the ligand concentrations that effected DNA binding of baculovirus-expressed full-length human AR. In vivo agonist and antagonist activities were as previously reported for oxandrolone (45, 46 ), fluoxymesterone (46 ), MPA (11, 12, 13, 19 ), RU56187 (14, 15 ), cyproterone acetate (65 ), and hydroxyflutamide (66, 67 ).

^a

Dashes indicate not detectable up to a concentration of 1 μm.

Figure 8.

Ligand-Dependent AR Stabilization Full-length AR was expressed in COS cells from pCMVhAR as described in Materials and Methods and incubated in the presence of[ ³⁵S]methionine/cysteine for 30 min at 37 C followed by incubation with unlabeled methionine medium for increasing times in the presence of the indicated concentrations of ligand. Samples were extracted in RIPA buffer and AR was immunoprecipitated and analyzed on SDS polyacrylamide gels as previously described (22 ). Scanning the exposed films resulted in optical density readings of the AR bands, which migrated at approximately 114 kDa. Shown are the optical density measurements on a semilog scale. Approximate half-times of AR degradation at 37 C at the indicated ligand concentrations are shown on the figures for RU56187 (panel A), fluoxymesterone (panel B), MPA (panel C), and oxandrolone (panel D).

Figure 9.

Dissociation Rates of [³H]R1881 and[ ³H]RU56187 from Human AR COS cells were transfected with pCMVhAR and incubated with 5 nm [³H]R1881 or 5 nm[ ³H]RU56187 for 2 h at 37 C followed by the addition of a 10,000-fold molar excess of unlabeled ligands as described in Materials and Methods. Cells were harvested at increasing time intervals, and radioactivity was determined. The data are shown on a semilog plot, and the half-time of dissociation is indicated: t_1/2 2.5 ± 0.5 h for[ ³H]R1881 and t_1/2 5 ± 1 min for[ ³H]RU56187.

Thus, MPA, cyproterone acetate, E₂, and progesterone at concentrations between 50 nm and 1 μm induce dimerization and DNA binding of the NH₂- and carboxyl-terminal AR fragments, where both fragments contain the AR DNA-binding domain (Fig. 7B). Yet none of these ligands, except for MPA, promote DNA binding of full-length AR and none induce the N/C two-hybrid interaction (Fig. 7A). The effectiveness of these ligands to induce dimerization and DNA binding of the AR NH₂- and carboxyl-terminal fragments in mobility shift assays is confounded by the presence of the dimerization region in the DNA-binding domain. We therefore attempted to address the contribution of the AR DNA-binding domain in the two-hybrid assay by testing the interaction of a fusion protein comprised of the GAL4 DNA-binding domain and AR DNA and ligand-binding domains (GAL-AR507–919) with VPAR1–660 or VPAR. However, lack of an interaction in this and additional experiments using GAL-AR (GAL4 DNA-binding domain-full-length AR fusion protein) suggests that the presence of two DNA-binding domains in a fusion protein (i.e. from GAL4 and AR) interferes with hybrid formation.

AR Stabilization

One property that has distinguished androgen agonists from antagonists is their ability to stabilize AR against degradation (27). It was therefore of interest to determine the concentration dependence of AR stabilization by MPA, RU56187, and the anabolic steroids. Transfected COS cells were incubated with[ ³⁵S]methionine/cysteine and increasing concentrations of ligands, followed by chase periods with unlabeled methionine for 2–7 h as previously described (27). The results shown for several ligands in Fig. 8 and summarized in Table 2 indicate that more than 100 nm MPA was required to increase the AR degradation half-time to almost 5 h at 37 C (Fig. 8C). A similar degree of AR stabilization was achieved by 1 nm DHT, mibolerone, or R1881, 5 nm testosterone (Table 2), or 10 nm fluoxymesterone or oxandrolone (Fig. 8, B and D). Cyproterone acetate and progesterone stabilized AR only at 1 μm (Table 2), and almost no AR stabilization was observed with 1 μm RU56187 (Fig. 8A), hydroxyflutamide, or E₂ (Table 2). The results tend to parallel the MMTV-luciferase and DNA-binding activities in that ligands that efficiently stabilize AR are more effective agonists. It is noteworthy that the lower affinity anabolic steroids, oxandrolone and fluoxymesterone, promote the N/C interaction and stabilize AR at concentrations of 5–10 nm, concentrations only slightly higher than those required for the high-affinity agonists, DHT, mibolerone, R1881, and testosterone. However, higher concentrations of the anabolic steroids were required for DNA binding of the AR fragments, perhaps reflecting the lower AR binding affinity for these ligands.

Ligand Dissociation Rates

Because only few ligands could be obtained in ³H-labeled form, dissociation rates could not be determined for MPA, oxandrolone, fluoxymesterone, cyproterone acetate, or hydroxyflutamide. Nevertheless, we determined that[ ³H]RU56187 dissociates rapidly from AR with a t_1/2 of 5 min at 37 C (Fig. 9) compared with t_1/2 of 2.5–3.5 h for [³H]R1881 (Fig. 9), [³H]DHT, and [³H]mibolerone (Table 2). A rapid dissociation rate for [³H]estradiol (t_1/2 0.67 h) (Table 2) was also observed. The results raise the possibility that rapid ligand dissociation is associated with the lack of an N/C interaction and with in vivo antagonist activity.

DISCUSSION

Failure of MPA to induce the N/C interaction suggests that during AR dimerization, DNA binding, and gene activation, MPA activates AR by a mechanism different from other agonists. With the other ligands tested, agonist activity correlated with induction of the N/C interaction and antagonist activity with its inhibition. MPA is a weak androgen in vivo (see below) which perhaps relates to its inability to induce the N/C interaction. Induction of the N/C interaction by high-affinity AR agonists appears to contribute to their biological potency at low physiological concentrations. Lack of induction of the N/C interaction may account for the high MPA concentrations required to stabilize AR, which would contribute to a reduced biological potency as an androgen agonist.

The apparent discrepancy in ligand potency between maximal induction in the N/C assay using the GAL4-AR carboxyl-terminal and the VP16-AR NH₂-fragment fusion proteins (0.1–1 nm) vs. agonist potency of full-length AR in the luciferase assay (0.001–0.01 nm) probably reflects deletion of the NH₂-terminal domain. We demonstrated previously that although the AR ligand-binding domain retains high-affinity binding after deletion of the NH₂-terminal region, the ligand dissociation rate increases 5- to 7-fold, and androgen no longer stabilizes this truncated receptor as it does full-length AR (22). Thus, higher ligand concentrations are likely required for the N/C interaction between AR fragments than between monomers of full-length AR.

Equilibrium dissociation constants (MPA, 1.7–2.9 nm; DHT, 0.9–2.6 nm) and saturation binding capacities (MPA, 107–249 fmol/mg protein; DHT, 42–257 fmol/mg protein) for MPA and DHT binding to AR were similar when measured in rat pituitary and hypothalamic extracts (28). However, direct measurement of in vivo bioactivity classifies MPA as a weak androgen. MPA increases the synthesis of β-glucuronidase in mouse kidney but only at 100-fold higher doses relative to testosterone (11, 12). In the androgen- insensitive Tfm mouse, β-glucuronidase activity did not increase, indicating that gene activation in response to MPA is AR mediated (12). MPA doses up to 1000 times higher than testosterone were required to increase ventral prostate weight in castrated rats (13). High-dose (0.9 mg/day) MPA was less effective than low-dose DHT (0.2 mg/day) in stimulating the synthesis of rat prostatic binding protein mRNA, and the effects of MPA were inhibited by flutamide (19), again indicating that its in vivo activity is AR mediated. MPA, rather than a metabolite, was shown to bind AR, and its low in vivo androgenic activity correlated with reduced nuclear uptake (29, 30). MPA was reported to dissociate rapidly from AR (13, 31), although these studies did not account for possible degradation of the MPA-AR complex. X-ray crystal analysis indicates that MPA has an inverted 1β,2α half-chair conformation of the A-ring resulting from steric strain by the 6α-methyl group that restricts side chain flexibility (32). This predicted rigid structure of MPA is in contrast to the flat, flexible structure of methyltrienolone (R1881), which can undergo large shape changes (33). It is conceivable that ligand flexibility facilitates the conformational changes required for the AR N/C interaction.

Acetate derivatives of steroids often have slower metabolic breakdown rates, making them candidates for use in hormone therapy (34). MPA, available for clinical use as Provera or Depo-Provera, has been used in the treatment of sexual precocity; its progestin and weak androgen effects inhibit pituitary gonadotropin secretion and lower gonadal steroid production (35). Stimulation of the growth of pubic hair in female patients without a significant slowdown in skeletal maturation (35) suggested weak androgenic activity of large doses of MPA (200–300 mg every 7–10 days). 21-Hydroxylated metabolites of MPA bind the glucocorticoid receptor and suppress pituitary secretion of ACTH. Because of its progestational effect, MPA was formerly used to prevent spontaneous abortion. However, prenatal exposure to MPA was reported to cause mild clitoral hypertrophy and posterior labial fusion in the female and hypospadias in the male (36–39), contraindicating its use during pregnancy. The virilizing effect of MPA in the female fetus can be explained by its androgenic activity. Antiandrogen effects of MPA in the male fetus could result from competition for DHT binding to AR and subsequent insufficient agonist activity. Anogenital distance, a measure of antiandrogen activity in rodents (1), was lengthened in females and shortened in males exposed to MPA during fetal development. More recently, MPA was approved for use in the United States as an injectable contraceptive based on its effectiveness in suppressing gonadotropin secretion, inhibiting follicular maturation and preventing ovulation. Doses of 150 mg im every 6 weeks to 3 months lack androgen effects in the adult female.

MPA has also been used in breast cancer therapy (40, 41). In an MFM-223 mammary cancer cell line that has high levels of AR, but low levels of estrogen, progesterone, and glucocorticoid receptors, cell proliferation was inhibited by 1 nm DHT or 10 nm MPA (42), indicating the AR agonist effect of MPA inhibits breast cancer cell growth. Response rates of breast cancer patients to MPA therapy correlated with higher AR levels (43). The antiproliferative activity of DHT and MPA on breast cancer cells was attributed to increased 17β-hydroxysteroid dehydrogenase activity, which promotes increased oxidation of estradiol to the weak estrogen, estrone (44).

Agonists vs. antagonist activity is influenced by metabolism, binding affinity, association and dissociation rates, and ligand-induced receptor conformation, stabilization, dimerization, DNA binding, and interactions with associating proteins. Clearly, equilibrium binding affinity is of limited usefulness in predicting in vivo bioactivity unless combined with measurements of ligand dissociation rates. AR binding affinity of RU56187 is similar to that for DHT, yet RU56187 is an antagonist in vivo (14, 15). The anabolic steroids, oxandrolone and fluoxymesterone, have high inhibition constants for binding, yet induce the N/C interaction and stabilize AR at relatively low ligand concentrations and are AR agonists in vivo. Oxandrolone induces male- specific liver P450 enzymes (45). Oxandrolone and fluoxymesterone are structurally related 17α-alkylated synthetic anabolic steroids used clinically to promote weight gain, stimulate growth of the bone matrix, and improve libido and sexual performance (46). In low doses oxandrolone (47, 48) or fluoxymesterone (49) accelerate linear growth in children with constitutional growth delay and Turner’s syndrome (50, 51).

Evidence from crystal structure analysis of the retinoic acid receptor-γ (52), thyroid hormone receptor (53), and estrogen receptor (54) indicates that hormone binding causes helix 12 [helix 11 in retinoid X receptor-α (55)] at the carboxyl terminus to undergo a conformational change closing down over the ligand-binding pocket. For the estrogen receptor, binding of the antagonist raloxifene prevents alignment of helix 12 over the binding pocket (54). MPA binding to AR may distort the position of helix 12 causing an increased rate of ligand dissociation and interference with the N/C interaction. Proper closure of helix 12 might be expected to slow ligand dissociation from the pocket and form a new interface for the N/C interaction. Alignment of helix 12 by MPA binding may differ from that induced by potent agonists or antagonists, a distortion that may account for the high MPA concentrations required to stabilize AR.

Part of the discrepancy between ligand binding affinity and agonist and antagonist activities relates to differences in ligand binding kinetics. Association and dissociation rate kinetics can be fast or slow for high-affinity ligands. Slow dissociation of the most potent AR agonists, DHT, mibolerone, and R1881, is associated with AR stabilization at low ligand concentration. Fast dissociating ligands such as RU56187 fail to stabilize AR and have agonist activity in transcriptional activation assays but are antagonists in vivo. Mutations in the AR hormone-binding domain at valine 889 and arginine 752 cause severe androgen insensitivity, increase the rate of dissociation of bound androgen without altering high-affinity equilibrium binding (22), disrupt the N/C interaction (6), and cause loss of AR stabilization by low ligand concentrations (22). These mutations likely increase the rate of ligand dissociation and AR degradation by preventing helix 12 from closing the binding pocket and interfering with the N/C interaction. Rapid ligand dissociation could reduce in vivo agonist activity and enhance dose-dependent antagonist activity as suggested previously for some antiestrogens and antiandrogens (13).

The most reliable in vitro indicators of in vivo AR antagonist activity therefore appear to be failure of a ligand to stabilize AR against degradation at steroid concentrations of 500 nm or more and an inability to induce AR DNA binding. DNA binding itself, however, appears to be a poor indicator of agonist potency. In vivo agonist activity is best reflected by a slow dissociation rate of bound ligand, AR stabilization at low ligand concentrations (≤ 10 nm), and induction of the N/C interaction. Concentrations at which a ligand activates AR in MMTV-luciferase assays can indicate agonist potency. MPA is an agonist in transient transcription assays but requires 100-fold higher concentrations than DHT. A similar shift in in vitro sensitivity to DHT results from certain AR missense mutations that cause partial or complete androgen insensitivity (56), indicating the critical importance of AR activation by low ligand concentrations. However, even though acetate derivatives of steroids have increased metabolic half-lives, it cannot be ruled out that the in vivo pharmacology of MPA limits its bioavailability to the AR.

A model for androgen-induced AR dimerization suggests an antiparallel orientation of monomers interacting through the DNA-binding domain and a ligand-dependent N/C interaction (4). Similar studies with the estrogen receptor (5) and AR fragments expressed in yeast (57) predict a parallel interaction model, and studies on solution dimerization of the human progesterone receptor favor a parallel model (58). More recent studies on AR made use of androgen insensitivity mutations in the steroid-binding domain that do not interfere with high-affinity equilibrium binding of androgen but increased the dissociation rate of bound androgen and disrupted the N/C interaction. Placement of the mutations in different AR fragments allowed assessment of directional dimerization in association with AR transcriptional activation, and the results were consistent with an antiparallel activated AR dimer model (6). Lack of an interaction between the ligand-binding domains bound to MPA or DHT argue against a parallel dimer model for AR with either ligand. Taken together the results suggest that the N/C interaction is required for potent in vivo agonists to be effective at low concentrations, but is not required for AR DNA binding in vitro or weak in vivo agonist activity at higher ligand concentrations. Formation of the N/C interaction likely contributes to in vivo potency by stabilizing AR at low ligand concentrations.

MATERIALS AND METHODS

Ligand Binding and Dissociation

Reagents were obtained as previously reported (22) with MPA and other steroids from Sigma Chemical Co. (St. Louis, MO) and RU56187 from Roussel Uclaf. Relative equilibrium binding was determined in COS cell competitive binding assays using [³H]R1881. Monkey kidney COS cells (3.5 × 10⁵ cells per well of six-well plate) were transiently transfected using diethylaminoethyl (DEAE)-dextran and 1 μg pCMVhAR full-length AR expression vector per well. Cells were maintained in 10% calf serum and DMEM for 36 h and labeled for 2 h at 37 C with 5 nm[ ³H]R1881 in the presence and absence of increasing concentrations of unlabeled ligands. Cells were washed with PBS and harvested in 2% SDS, 10% glycerol, and 10 mm Tris, pH 6.8, and radioactivity was determined by scintillation counting. Dissociation rate kinetics were determined in COS cells transfected as described above using 3 μg pCMVhAR/well. Transfected cells were incubated with 5 nm [³H] ligand for 2 h followed by the addition of a 10,000-fold molar excess of unlabeled ligand. After increasing times, cells were washed with PBS and harvested in 0.5 ml of the SDS buffer above. Radioactivity was determined by scintillation counting. Apparent inhibition constants (K_i) for hydroxyflutamide, oxandrolone, and fluoxymesterone were determined using rat prostate cytosols prepared from tissue obtained 24 h after castration, extracted in binding buffer as previously described (1, 59), and incubated for 20 h at 4 C with 0.5–20 nm [³H]R1881 with or without increasing concentrations of unlabeled ligands between 0.1 and 1μ m. Apparent K_i values were determined using double reciprocal plots and slope-replot analysis.

N/C Luciferase Assay

Recombinant fusion proteins included GALD-H which contained the Saccharomyces cerevisiae GAL4 DNA-binding domain amino acid residues 1–147 linked in frame with human AR steroid-binding domain amino acid residues 624–919. VPAR1–660 contained the herpes simplex virus VP16 transactivation domain amino acid residues 411–456 linked in frame with AR NH₂-terminal and DNA-binding domain amino acid residues 1–660 (4). CHO cells (0.4 × 10⁶ cells per 6-cm dish) were transfected using DEAE-dextran and 1 μg GALD-H, 1μ g VPAR1–660, and 5 μg G5E1b-luciferase per plate, the latter containing five GAL-4 DNA-binding sites (60). DNA was added to 0.42 ml H₂0 plus 0.5 ml 2×TBS (0.14 m NaCl, 3 mm KCl, 1 mm CaCl₂, 0.5 mm MgCl₂, 0.9 mm NaH₂PO₄, and 25 mm Tris-HCl, pH 7.4), and then 0.11 ml DEAE-dextran (0.5%) was added, after which the mixture was added to the aspirated plates and incubated for 1 h at 37 C. Plates were aspirated and 4 ml α-MEM containing 10% calf serum, penicillin/streptomycin, and 20 mm HEPES, pH 7.2, were added and incubated at 37 C for 3 h followed by a 4-min 15% glycerol shock in α-MEM. Cells were washed twice with 4 ml TBS, and 4 ml 0.2% calf serum-α- MEM media were added. The medium was changed 24 and 48 h later to serum-free medium and ligands were added. Cells were washed 4 h after the last addition with 4 ml PBS and harvested in 0.5 ml lysis buffer (Ligand Pharmaceuticals Inc., San Diego, CA). Luciferase light units were measured on a Monolight 2010 luminometer (Analytical Luminescence Laboratory, San Diego, CA). To test for inhibitory activity, cells were incubated with 1 nm DHT in the presence of increasing concentrations of ligands. Decreases due to toxicity of up to 1 μm ligand were monitored using an AR fusion plasmid GAL-A1 (4) coding for a constitutively active fusion protein containing AR NH₂-terminal residues 1–503 linked to the GAL-4 DNA-binding domain. Decreases in constitutive transcriptional activity after exposure were minimal.

MMTV-Luciferase Assay

Monkey kidney CV1 cells (0.4 × 10⁶ cells per 6-cm dish) were transfected 24 h after plating using calcium phosphate with 100 ng human AR expression vector pCMVhAR and 5 μg MMTV-luciferase reporter vector per plate. DNA is added to 0.28 m NaCl, 1.5 mm Na₂HPO₄, 0.05 m HEPES, pH 7.11 (0.14 ml/plate; 0.28 m NaCl, 1.5 mm Na₂HPO₄, 50 mm HEPES, pH 7.12); 0.25 m CaCl₂ (0.14 ml/plate) was added dropwise with vortexing and incubated 5 min, followed by the addition of DMEM-H media containing 10% calf serum, penicillin, streptomycin, and 20 mm HEPES, pH 7.2 (0.8 ml/plate) to minimize particle size, and then incubated for 15 min. The mixture was added to the aspirated plates followed by the addition of 3 ml DMEM-H containing 10% calf serum and incubated for 4 h at 37 C. Cells are washed twice with TBS, and 4 ml phenol red-free medium containing 0.2% calf serum was added. Ligands were added and cells were harvested and assayed as described above for the N/C luciferase assay.

DNA Mobility Shift Assay

Spodoptera frugiperda (Sf9) cells plated at 3.5× 10⁶ cells per 6-cm dish or 1 × 10⁷ cells per 10-cm dish were infected for 45 h at 27 C at multiplicity of infection of 1–5 with AR recombinant baculovirus in Autographa californica nuclear polyhedrosis virus (AcMNPV) coding for full-length human AR, AR1–660 coding for the NH₂-terminal, DNA-binding, and hinge regions (amino acid residues 1–660), and AR507–919 coding for the DNA- and steroid-binding domains (amino acid residues 507–919) (25). The indicated concentrations of ligands were added 24 h and again 4 h before cell harvest. Cells were washed once in PBS at 4 C, pelleted, and resuspended in 0.15 ml/6-cm dish or 0.4 ml/10-cm dish in high-salt extraction buffer containing 0.5 m NaCl, 1 mm EDTA, 1 mm dithiothreitol, 10% glycerol, 10 mm Tris, pH 7.4 with protease inhibitors, 40μ m leupeptin, 5 μm aprotinin, 10μ m pepstatin A, 2 mm Pefabloc, 5 mm benzamidine, and 10 mmε -amino-n-caproic acid. Cells were frozen and thawed three times, incubated on ice for 40 min, and microfuged for 15 min. Supernatants were dialyzed against the above buffer except containing 25 mm KCl and 0.5 mm phenylmethylsulfonyl fluoride as the only protease inhibitor. The reaction mixture contained approximately 20 μg total cell protein of either full-length AR, AR1–660 (N), or AR507–919 (C), or when combined, 10 μg total protein each for extracts of N and C. The reaction mix also contained 4μ g of poly dI-dC, 80 μg BSA, and DNA-binding buffer (25 mm KCl, 10% glycerol, 0.2 mm EDTA, 1 mm dithiothreitol, 10 mm Tris-HCl, pH 7.5) to a final volume of 20 μl. ³²P-labeled oligonucleotides (0.2–0.3 ng, 20,000–25,000 cpm) were added and incubated for 1 h on ice. Annealing oligos 5′-CGACCAGAGTACGTGATGTTCTCAGG-3′ with AccI-5′ compatible end and 5′-GATCCCTGAGAACATCACGTACTCTGGT-3′ with 3′ BamHI compatible end were ³²P-labeled using the Klenow fragment of DNA polymerase. The androgen response element (underlined) derives from the 0.5-kb first intron fragment of the rat C3 prostatein gene (61). Before electrophoresis, 2 μl 0.2% bromophenol blue were added, and the 5% nondenaturing acrylamide gel was preelectrophoresed at 100 V for 30 min at 4 C. Samples are electrophoresed at 150 V for 4 h at 4 C. Gels are dried under vacuum at 80 C for 1 h and exposed to Biomax MR x-ray film (Eastman Kodak, Rochester, NY) at −80 C.

AR Stabilization

Full-length AR was expressed from pCMVhAR (8 μg) in COS cells (1.2 × 10⁶ cells/10-cm dish) transfected using DEAE-dextran. After 48 h, cells were incubated in methionine-free medium for 20 min followed by the addition of methionine-free medium containing 100 μCi [³⁵S]L-methionine/cysteine (PRO-MIX, Amersham, >1000 Ci/mmol) in vitro labeling mix. Cells were incubated for increasing times in the presence of the indicated concentrations of ligands, washed twice with PBS, and harvested in RIPA buffer and immunoprecipitated using AR52 antipeptide AR antibody as previously described (22).

Acknowledgments

We are grateful for the technical assistance of K. Michelle Cobb, Christy Lambright, and De-Ying Zang; to Frank S. French for helpful discussions and reading the manuscript; and D. Gallet and D. Martini at Hoechst Marion Roussel (Roussel Uclaf) for labeled and unlabeled RU56187.

This work was supported by Grants HD-16910 and IU54-HD-35041 from the National Institute of Child Health and Human Development Center for Population Research, and by ES-08265 from the National Institute of Environmental Health Sciences.

^*

Present address: Departamento de Biotecnología, Instituto de Investigaciones Biomédicas, Universidad Nacional Autónoma de México, México, D. F. México.

^†

Present address: Monsanto Company, 645 South Newstead Avenue, St. Louis, Missouri 63110.