Modulation of Androgen Receptor Activation Function 2 by Testosterone and Dihydrotestosterone*

The androgen receptor (AR) is transcriptionally activated by high affinity binding of testosterone (T) or its 5α-reduced metabolite, dihydrotestosterone (DHT), a more potent androgen required for male reproductive tract development. The molecular basis for the weaker activity of T was investigated by determining T-bound ligand binding domain crystal structures of wild-type AR and a prostate cancer somatic mutant complexed with the AR FXXLF or coactivator LXXLL peptide. Nearly identical interactions of T and DHT in the AR ligand binding pocket correlate with similar rates of dissociation from an AR fragment containing the ligand binding domain. However, T induces weaker AR FXXLF and coactivator LXXLL motif interactions at activation function 2 (AF2). Less effective FXXLF motif binding to AF2 accounts for faster T dissociation from full-length AR. T can nevertheless acquire DHT-like activity through an AR helix-10 H874Y prostate cancer mutation. The Tyr-874 mutant side chain mediates a new hydrogen bonding scheme from exterior helix-10 to backbone protein core helix-4 residue Tyr-739 to rescue T-induced AR activity by improving AF2 binding of FXXLF and LXXLL motifs. Greater AR AF2 activity by improved core helix interactions is supported by the effects of melanoma antigen gene protein-11, an AR coregulator that binds the AR FXXLF motif and targets AF2 for activation. We conclude that T is a weaker androgen than DHT because of less favorable T-dependent AR FXXLF and coactivator LXXLL motif interactions at AF2.

The androgen receptor (AR) 2 is a member of the nuclear receptor superfamily of ligand-activated transcription factors. Androgen activation of AR regulates prostate growth, bone and muscle mass, and spermatogenesis and is a predisposing factor in prostate cancer. AR mediates transcriptional activity in response to two biologically active androgens that bind AR with similar high affinity (1). Testosterone (T) is the major circulating androgen secreted by the testis and the active androgen in muscle. Dihydrotestosterone (DHT), the 5␣-reduced metabolite of T, is a more potent androgen required for male reproductive tract development.
AR has a modular structure composed of an NH 2 -terminal transactivation domain, central DNA binding domain, linker hinge region, and carboxyl-terminal ligand binding domain (LBD) (2). Like other nuclear receptor family members (3), AR has two major activation domains. Androgen-induced AR transcriptional activity depends on activation function 1 in the largely unstructured NH 2 -terminal region (2) and activation function 2 (AF2), a highly ordered hydrophobic surface in the LBD that requires androgen binding for its structural integrity (4).
The degree to which AF2 contributes to overall AR activity depends on multiple competing factors. Unlike other nuclear receptors, AR AF2 binds a number of LXXLL-related motifs. Important among these is the AR NH 2 -terminal FXXLF motif 23 FQNLF 27 that binds AF2 in an androgen-dependent and -specific manner. AR FXXLF motif binding to AF2 is the basis for the AR NH 2 -and carboxyl-terminal (N/C) interaction (5-7) that contributes to AR dimerization (8,9) and is critical for AR regulation of androgen-dependent genes (10 -12). The functional significance of the AR N/C interaction in vivo is supported by the effects of several naturally occurring mutations that disrupt AR FXXLF motif binding and cause resistance to androgen without diminishing high affinity androgen binding (13)(14)(15)(16). The androgen insensitivity syndrome results in varying degrees of incomplete masculinization of the external gen-italia in genetic males depending on the extent to which the mutation disrupts AR function (17).
In addition to the AR FXXLF motif, multiple related motifs bind the AR AF2 site with relatively high affinity (18). FXXLF motifs are present in a number of putative AR coregulatory proteins and interact at AF2 (19,20). Within the AR NH 2 -terminal domain is a WXXLF motif that interacts with the AR AF2 site in the presence of androgen but with weaker affinity than the AR FXXLF motif (5,10). Similar to other nuclear receptors, AR AF2 serves as the binding site for steroid receptor coactivator (SRC)/p160 coactivator LXXLL motifs. Crystal structures demonstrate overlapping binding sites for AR FXXLF and coactivator LXXLL motifs (4,18). Based on peptide display screening (18,20,21) and binding affinity measurements (4), AR AF2 preferentially binds FXXLF motifs compared with coactivator LXXLL motifs. In addition, we have shown that androgen-dependent AR FXXLF motif binding to AF2 in the AR N/C interaction competitively inhibits coactivator LXXLL motif binding (19).
The contribution of AF2 to AR transcriptional activity is also influenced by cell-and tissue-specific coactivators, some of which selectively increase accessibility of AF2 to coactivator recruitment. One mechanism proposed to increase AR AF2 activity in prostate cancer is higher levels of SRC/p160 coactivators that compete for the AR N/C interaction and increase AR transcriptional activity through AF2 (10,22). The AR coregulator melanoma antigen gene protein-11 (MAGE-11) of the MAGEA gene family binds the AR FXXLF motif to expose AF2 and increase coactivator recruitment (23). Naturally occurring AR somatic mutations in prostate cancer can increase AR activity by enhancing SRC/p160 coactivator recruitment to AF2 (7). The relative binding affinities and competitive relationships at the AF2 site suggest that high affinity androgen binding triggers sequential interactions of multiple coregulatory proteins.
In this study we provide biochemical and structural evidence that T is a less effective androgen than DHT because of weaker T-dependent FXXLF and LXXLL motif binding at the AR AF2 surface. This conclusion is supported by a prostate cancer somatic mutation AR-H874Y that increases the transcriptional response to T in association with improved FXXLF and LXXLL motif binding at AF2. Crystal structure determination of T-bound WT and H874Y AR LBD provided some insight into the possible differential molecular effects of T versus DHT. Receptor bound with T appears to induce isolated conformational heterogeneity at the AF2 surface. The AR H874Y mutation creates new direct hydrogen (H) bonds between core helix residues that probably contribute to the molecular basis for the described functional rescue by this prostate cancer mutant.
Reporter Gene Assays-The CWR-R1 prostate cancer cell line derived from the CWR22 recurrent human prostate cancer xenograft (24,27) was plated at 1.6 or 2 ϫ 10 5 cells/well in 12-well plates in prostate cell growth medium containing Richter's improved minimal essential medium (Invitrogen) or Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 5 ng/ml selenium, 10 mM nicotinamide, 5 g/ml insulin, 5 g/ml transferrin, and 2% fetal bovine serum and transfected using Effectene (Qiagen). Endogenous CWR-R1 cell AR-H874Y transcriptional activity was detected using 0.1 g/well MMTV-Luc reporter vector. AR AF2 activity was determined in CWR-R1 cells by transfecting 0.1 g of WT or mutant GAL-AR-(624 -919) or GAL-AR-(658 -919) and 0.25 g of 5XGAL4Luc3. For two-hybrid interaction assays, CWR-R1 cells were transfected with 50 ng of VP-TIF2-(624 -1287), WT, or mutant GAL-AR-(624 -919) and 0.1 g of 5XGAL4Luc3. For 12-well plates, DNA was combined with (per well) 45 l of transfection buffer (Qiagen) and 1 l of Enhancer, vortexed, and incubated for 5 min at room temperature. Effectene (1 l/well) was added, vortexed for 10 s, and incubated for 10 min at room temperature. Prostate cell growth medium (0.2 ml) was added and vortexed, and 220 l of DNA solution was added to each well containing 1 ml of medium. The next day cells were washed with phosphate-buffered saline (PBS) and 1 ml of phenol red-free, serum-free basic prostate medium containing Improved minimal essential Zinc Option medium (Invitrogen), and the indicated steroids were added per well. Cells were incubated at 37°C overnight, washed with PBS, and harvested in 0.25 ml of lysis buffer containing 1% Triton X-100, 2 mM EDTA, and 25 mM Tris phosphate, pH 7.8. Cells were rocked at room temperature for 30 min in lysis buffer, and 0.1 ml of cell lysate was analyzed for luciferase activity using a Lumistar Galaxy (BMG Labtech) automated multiwell plate reader luminometer.
Human epithelial cervical carcinoma HeLa cells were maintained in Eagle's minimum essential medium supplemented with 10% fetal bovine serum (Gemini or HyClone), penicillin, streptomycin, and 2 mM L-glutamine. For reporter gene assays, HeLa cells were plated at 5 ϫ 10 4 cells/well in 12-well plates and 24 h later transfected using FuGENE 6 transfection reagent (Roche Applied Science) with 10 ng of pCMVhAR or H874Y mutant and 0.25 g of PSA-Enh-Luc reporter vector to determine androgen-induced AR transactivation. To measure AR AF2 activity, HeLa cells were transfected with 0.1 g of GAL-AR-(658 -919) and the H874Y mutant and 0.25 g of 5XGAL4Luc3. For two-hybrid interaction assays, HeLa cells were transfected with 0.1 g of 5XGAL4Luc3, 50 ng of VP-AR-(1-660) or VP-TIF2-(624 -1287), and 50 ng of GAL-AR-(624 -919), -(640 -919), -(658 -919), or GAL-AR-(624 -919)-H874Y. DNA was added to 43 l of serum-free medium and 0.6 l of FuGENE-6 reagent per well. After a 15-min incubation, 40 l of FuGENE/DNA mixture was added to each well containing 1 ml of medium. The next day cells were washed with PBS, and 1 ml/well serum-free medium lacking phenol red containing the indicated steroids was added and incubated overnight at 37°C. Twenty four hours later cells were washed with PBS and assayed for luciferase activity after harvesting in 0.25 ml of lysis buffer as described above.
Monkey kidney CV1 cells were maintained in Dulbecco's modified Eagle's medium containing 10% bovine calf serum (HyClone), 2 mM L-glutamine, penicillin, streptomycin, and 20 mM HEPES, pH 7.2. Cells (4.2 ϫ 10 5 /6 cm dish) were plated in medium containing 5% bovine calf serum and 24 h later transfected using calcium phosphate DNA precipitation (28). The effect of TIF2 on AR AF2 activity was determined by transfecting 5 g of 5XGAL4Luc3 and 0.1 g of GAL-AR-(624 -919), GAL-AR-(658 -919), or the H874Y mutants in the absence and presence of 2 g of pSG5-TIF2. The effect of MAGE-11 on AR transcriptional activity was determined by transfecting 0.1 g of pCMVhAR or pCMVhAR-H874Y and 5 g of PSA-Enh-Luc in the absence and presence of 2 g of pSG5-MAGE-11 and 2 g of pSG5-TIF2. Cells were incubated overnight with and without the indicated androgens and the next day placed in serum-free medium in the absence and presence of androgen. After 24 h cells were washed with PBS, harvested in 0.25 ml of lysis buffer, and assayed for luciferase activity as described.
Androgen Dissociation Rate Assays-Ligand dissociation rate studies were performed at 37°C in whole cell binding assays by plating 4 ϫ 10 5 COS cells/well of 6-well plates and transfecting 1 or 2 g of pCMVhAR, pCMVhAR-(507-919), or H874Y mutant per well using DEAE-dextran (28). Transfected cells were incubated for 2 h at 37°C in 0.6 ml of serum-free medium lacking phenol red and containing 5 nM [1,2,6,7-3 H]T (78.5 Ci/mmol) or 3 nM [1,2,4,5,6,7-3 H]DHT (124 Ci/mmol). Nonspecific binding was determined in parallel wells by adding 100fold molar excess of unlabeled androgen. Timed dissociation rates were determined by amending to 50 M T or 17␣-methyltrienolone (R1881) (PerkinElmer Life Sciences) to the labeling media added in 0.1 ml of medium. Cultured cell incubations at 37°C were terminated at different times, and cells were washed with PBS, harvested in 0.5 ml of lysis buffer containing 2% SDS, 10% glycerol, and 20 mM Tris, pH 6.8, and radioactivity was determined by scintillation counting. Dissociation halftimes were determined as the mean Ϯ S.E. time required to reduce specific androgen binding by 50%.
Fluorescence Polarization Measurements of Peptide Binding Affinity-Fluorescence binding studies were performed using WT AR LBD and AR-H874Y LBD expressed in E. coli in the presence of 0.5 mM T or 50 M DHT. Protein was purified as described above except AR LBD-DHT thrombin digestion and overnight dialysis were not performed, and DHT was not read-ded during purification. Protein was concentrated to 0.6 -3 mg/ml using Centriprep centrifugal filter units in buffer containing 10 M T or DHT, 0.15 M Li 2 SO 4 ,10 mM DTT, 0.5 mM EDTA, 10% glycerol, 0.05% ␤-n-octoglucoside, and 25 mM HEPES, pH 7.5. AR FXXLF and TIF2 LXXLL peptide binding affinities were determined by fluorescence polarization at room temperature for 1 h using 5-40 M AR LBD and AR-H874Y LBD purified in the presence of 10 M ligand and assayed with and without addition to 40 M DHT or T and 10 nM AR-(20 -30) fluorescein-RGAFQNLFQSV or TIF2 third LXXLL motif 732-756 fluorescein-QEPVSPKKKENALLRYLLDKDDTKD (Synpep, Dublin, CA). As a control, human estrogen receptor-␤ (ER␤) LBD (residues 257-530) was analyzed in parallel in the presence of 40 M estradiol (E 2 ). Fluorescence polarization values were determined using an Envision (PerkinElmer Life Sciences) fluorescence plate reader with 485 nm excitation and 520 nm emission filters. Binding isotherms were constructed, and K D values were determined by nonlinear least squares fit based on a 1:1 interaction (30).
Crystallization and Data Analysis-Concentrated solutions of purified AR-H874Y LBD and AR LBD complexed with T and amended with 2-3 M excess of AR-(20 -30) FXXLF motif peptide RGAFQNLFQSV or TIF2-(740 -753) LXXLL-III motif peptide KENALLRYLLDKDD were used to obtain diffraction grade crystals. Vapor diffused hanging drops at 22°C with a 1:1 (v/v) ratio of the AR complex and the precipitant solution produced ϳ150 -200 m crystals within 3-15 days. Salt solutions containing 0.6 M sodium-potassium tartrate and Bistris propane, pH 7.0, or Tris, pH 8.5, were used as precipitants. Prior to flash-freezing in liquid N 2 , crystals were transiently mixed with a cryoprotectant solvent consisting of precipitant solution amended to 20% glycerol. X-ray diffraction data were collected at 100 K with an ADSC 210 detector at the IMCA-CAT, sector 17ID, or a MAR225 CCD detector at the SER-CAT, sector 22BM, at the Advanced Photon Source synchrotron. Diffraction data were integrated and scaled with HKL2000 (31).
Structure Determination and Refinement-An initial model for the T-bound AR-H874Y LBD with AR-(20 -30) peptide data was determined by molecular replacement with MolRep (32, 33) and the AR LBD coordinates from the AR-DHT structure (34) (Protein Data Bank access code 1I37). The convincing solution contained a single AR LBD complex in the asymmetric unit that had excellent quality 1.8 Å resolution electron density for T and the respective peptide. Multiple cycles of manual model building with COOT (35) and maximum likelihood restrained refinement with all hydrogens were performed with Refmac (36) in all cases. Initial models for the remaining data sets were determined and refined in a similar manner. Table 3 summarizes the crystallographic and refinement statistics.
Coordinate files with hydrogens were generated with Mol-Probity (37). To eliminate possible slight differences arising from variation in AF2 helix position, backbone heavy atoms for AR chain A LBD residues 680 -890 were used for structure superimposition and performed with the CCP4i LSQAB utility using coordinates without hydrogens. Reported interatomic distances are between heavy atoms unless specified, and the angles with protons when necessary were measured with COOT or PyMol. Structure figures were generated with PyMol from Delano Scientific.

RESULTS
AF2 Activation by T and DHT-To investigate the differential effects of T and DHT on AR AF2 activity, we performed studies using WT AR and a prostate cancer somatic mutant AR-H874Y that has an increased transcriptional response to T (7,24). To optimize detection of AR AF2 activity, we varied the length of the hinge region of several AR LBD-GAL4-DNA binding domain fusion proteins expressed in several cell lines (Fig.  1A). In CV1 cells, T and DHT increased TIF2-dependent GAL-AR-(658 -919) activity to a greater extent than GAL-AR-(624 -919), indicating inhibition by hinge residues that include the AR nuclear targeting signal (38) (Fig. 1C). The H874Y mutation increased androgen sensitivity and overall transcriptional activity but remained dependent on ligand and coexpression of TIF2 (Fig. 1C).
Androgen-dependent AF2 activity of GAL-AR-(658 -919) was stronger in the CWR-R1 prostate cancer cell line and was detected independent of coexpressed TIF2 (Fig. 1E). Activity of GAL-AR-(658 -919) was greater than GAL-AR-(624 -919), and the H874Y mutation increased the response to T. Differences in transcriptional activity did not result from differences in protein expression (Fig. 1B) and was AF2-dependent because charge clamp mutants K720A and E897K eliminated the response (Fig. 1F).
In HeLa cells, the predominant effect of the H874Y mutation was also increased sensitivity to T (Fig. 2B), whereas GAL-AR-(658 -919) and the H874Y mutant responded similarly to DHT. We noted a lack of transcriptional response by GAL-AR-(658 -919) to androstanediol in HeLa cells, which contrasted equivalent activity by androstanediol and T in CWR-R1 cells. Twohybrid studies suggested that the greater activity by androstanediol in CWR-R1 cells resulted from metabolism to an active androgen (data not shown).
Full-length endogenous AR-H874Y in CWR-R1 cells does not activate the PSA-Luc reporter (40) but activates MMTVluciferase in response to T and DHT and higher concentrations of androstenedione and androstanediol (Fig. 3A). Transiently expressed AR and AR-H874Y in CWR-R1 cells activate PSA-Luc in response to T and DHT, and adrenal androgens were more effective with AR-H874Y (data now shown). In HeLa cells, AR-H874Y lacked constitutive activity but increased the response to T with less differential effects by DHT, androstenedione, and androstanediol (Fig. 3B).
The results in both CWR-R1 and HeLa cell lines suggest that the predominant effect of the H874Y mutation is to increase the AF2 response to T. Our ability to detect AR AF2 activity in CWR-R1 cells but not HeLa or CV1 cells without coexpression of TIF2 likely reflects higher endogenous SRC/p160 coactivator levels in CWR-R1 cells that endogenously express the AR-H874Y mutant (40).
Preferential AF2 Activation by MAGE-11-MAGE-11 is an AR coregulator expressed in prostate cancer cell lines and in normal tissues of the human male and female reproductive tracts (23). MAGE-11 binds the AR NH 2 -terminal FXXLF motif and increases AF2 by inhibiting the AR N/C interaction. To gain further evidence that AR-H874Y increases AF2 activity in response to T, we determined the effect of MAGE-11 with and without coexpression of TIF2 using a PSA-luciferase reporter.
Coexpression of TIF2 had minimal effects on AR and AR-H874Y activity (Fig. 4) in agreement with the inhibitory effects of the AR N/C interaction on coactivator recruitment by AF2 (28). Coexpression of MAGE-11 with and without TIF2 preferentially increased AR-H874Y activity in response to T compared with WT AR and AR-H874Y with DHT. AR-H874Y activity induced by T and DHT was nearly equal in the presence of MAGE-11 with or without TIF2. This contrasts WT AR where DHT induced greater activity than T in the presence of MAGE-11 with or without TIF2. Coexpression of MAGE-11 also increased ligand independent activity of AR-H874Y more than WT AR. Differences in transcriptional activity were independent of differences in protein expression levels based on immunoblot analysis (see Fig. 6B). The preferential effects of MAGE-11 on T-dependent AR-H874Y activity suggest that H874Y imparts DHT-like activity to T by increasing coactivator recruitment to AF2.
FXXLF and LXXLL Motif Binding Affinities-Binding isotherms calculated by fluorescence polarization indicate WT AR LBD-DHT binds the AR FXXLF peptide with ϳ2-fold higher affinity than WT AR LBD-T with no significant change by the H874Y mutation ( Fig. 5; Table 1). Similar results were observed for the TIF2 LXXLL peptide except overall binding affinities were weaker than the FXXLF peptide. ER␤ LBD-E 2 bound the TIF2 LXXLL peptide with higher affinity than the FXXLF peptide as reported previously (4). The data suggest a direct differ- ential effect of T and DHT on AF2 motif binding affinity that is not altered by the H874Y mutation.
T and DHT Dissociation Kinetics-Prostate cancer mutation H874Y slows the dissociation rate of synthetic androgen R1881 from AR and AR-(507-919), a carboxyl-terminal fragment that lacks the AR NH 2 -terminal domain (7). To investigate the similar AR-H874Y AF2 activity induced by T and DHT, we determined androgen dissociation half-times from WT and H874Y full-length AR and AR-(507-919) containing the DNA binding domain and LBD, which expressed similarly on immunoblots (Fig. 6B).
T dissociates 3-4 times faster than DHT from full-length WT AR (Fig. 6A; Table 2) (1) but with a similar rate as DHT from AR-(507-919) ( Table 2), suggesting similar steroid contacts in the ligand binding pocket. T dissociates 3-4 times slower from AR-H874Y than from WT AR at a rate similar to DHT dissociation from AR and AR-H874Y. The dissociation rate of T and DHT from AR-(507-919)-H874Y was ϳ2-fold slower than from WT AR-(507-919).
The data indicate that slow dissociation of DHT from fulllength WT AR results predominantly from interactions outside the ligand binding pocket, and H874Y has effects inside and outside the binding pocket. The similar half-time of T dissociation from AR-H874Y to DHT from AR and AR-H874Y parallel AF2 transcriptional activities (see Figs. 1 and 2) and provide further evidence that H874Y imparts DHT-like activity to T by increasing the transcriptional activity of AF2.
T-bound WT AR LBD and AR-H874Y LBD Structures-We determined the crystal structures of WT and H874Y AR LBD bound with T in the presence of the AR-(20 -30) NH 2 -terminal FXXLF motif peptide or TIF2 coactivator peptide TIF2-(740 -753) third LXXLL motif. Crystallographic refinement data are provided in Table 3. Globally, all four structures conform to the canonical nuclear receptor LBD fold (Fig. 7, A and B) and when superimposed are nearly identical based on r.m.s.d. statistics for the xyz displacement relative to the WT AR LBD-T-FXXLF coordinates (0.26 Å for WT AR LBD-T-LXXLL, 0.14 Å for AR-H874Y LBD-T-FXXLF, and 0.27 Å for AR-H874Y LBD-T-LXXLL). Globally, the structures concur with previously reported structures for WT AR LBD bound to DHT and R1881 and prostate cancer mutants AR-T877A LBD and AR-W741L LBD bound to steroid and nonsteroid ligands (4,18,34,(41)(42)(43)(44)(45). Our WT AR LBD-T structures with AR FXXLF (Protein Data Bank access code 2Q7I) or TIF2 LXXLL (Protein Data Bank access code 2Q7J) peptide superimpose to the DHT-bound structures with FXXLF (Protein Data Bank access code 1TR7) or LXXLL (Protein Data Bank access code 1T63) peptide (see Fig. 9) with an r.m.s.d. of 0.27 and 0.3 Å and to WT AR LBD-S-1 (Protein Data Bank access code 2AXA) and AR-W741L LBD-  bicalutamide (Protein Data Bank access code 1Z95) with an r.m.s.d. of 0.37 and 0.31 Å, respectively. Consistent with the AR LBD-R1881 (4) and DHT peptide bound structures (18,42), the LXXLL motif in the T-bound structures is carboxyl-terminally shifted along the helical axis relative to the FXXLF peptide, and  Table 1. The data are the mean Ϯ S.E. expressed as millipolarization units (mP) versus purified receptor LBD concentration. Cells transfected with 10 g of pCMVhAR, pCMVhAR-(507-919), and the corresponding H874Y mutants were incubated in serum-free medium in the absence of hormone. Protein extracts (20 g protein/lane) were separated on a 10% acrylamide gel containing SDS and the transferred protein blot probed using anti-AR antibody AR-52. Leu-745 and Leu-749 lie in register with Phe-23 and Phe-27 (Fig. 7A). The AR-(20 -30) FXXLF peptide H-bonds to conserved charge clamp residues Glu-897 and Lys-720 required for AR AF2 activity (6). The NH 2 terminus of the LXXLL peptide fails to H-bond with Glu-897 and maintains the carboxyl-terminal shift, motif registry, and interaction to Lys-720 as shown for the AR LBD bound to R1881 (4) (Fig. 7, A-C). The T-bound ligand binding pockets are essentially identical to each other and nearly identical to the DHT-bound AR LBD structures (34, 42) (Fig. 7, C and D, and Table 4).

TABLE 1 Androgen-dependent AR LBD and AR-H874Y LBD binding affinities for AR FXXLF and TIF2 LXXLL peptides
WT AR LBD-T-FXXLF and LXXLL-There were no major structural differences to thoroughly account for the noted physiologic differences between T and DHT. For both, the steroid A-ring lies near the side chain of Arg-752, a conserved helix-5 residue required for ligand binding (Fig. 8) (41,43,44). In the T-bound WT AR LBD-FXXLF structure, we observed a 3.0 Å interatomic distance between the steroid 3-keto O and Arg-752 side chain atom N-2 and measured a 126°angle subtended by atoms Arg-752 N-2, H-22, and the T 3-keto O ( Fig. 8A and Table 4). Although the 3.0 Å distance supports the presence of a direct H-bond with Arg-752, the angular displacement is less than the optimal 180°angle and only slightly more favorable than the 110°, 3.0 Å H-bond of DHT-AR LBD (Protein Data Bank access code 1T63) (Fig. 8, B and C).
Also located near the steroid A-ring is conserved structural water HOH1, which ideally could mediate up to four local H-bonds. However Although invoking an HOH1 to T 3-keto O H-bond forms a narrow 80°angle between the 3-keto O, HOH1 O, and Met-745 carbonyl oxygen atoms that violates the ideal tetrahedral water geometry, the superior hydrophilic properties of the T A-ring relative to DHT increase the propensity of T to accept this second H-bond through the fourth coordination of HOH1. Better H-bonding by T appears to also slightly reduce the distance between HOH1 and the Met-745 backbone carbonyl (2.7 Å) relative to DHT (2.9 Å).
On the D-ring of T, the 17␤-hydroxyl group accepts an H-bond from the helix-10 Thr-877 side chain and donates an H-bond to helix-3 Asn-705 O-␦1 (Fig. 7C and Fig. 9) as reported for DHT and R1881 (4,34). Most notable among these is helix-3 Gln-711 near HOH1, the next sequential residue to Leu-712, which contacts i ϩ 1 of the bound peptide motif, and helix-12 Met-895, which lies proximal to Leu-712 but more distal to the i ϩ 1 residue of the bound peptide and the steroid A-ring (Fig. 10). We also observed in each structure, but do not illustrate, a glycerol that derives from the protein buffer solution that binds above the Gln-711 side chain. It is unclear whether these alternate conformers are crystallization artifacts or, as for Gln-711, arise from the presence of glycerol. Others such as Leu-712 or Met-895 may represent conformational freedom arising from a protein-or ligandmediated mechanism.
AR H874Y LBD-T-FXXLF and LXXLL-We noted a more definitive structural change in our analysis of the AR-H874Y mutant LBD bound to T and AR FXXLF (Protein Data Bank access code 2Q7K) or TIF2 LXXLL (Protein Data Bank 2Q7L) peptide. Side chains of exterior helix-10 WT residue His-874 (Fig. 10A) and H874Y mutant residue Tyr-874 (Fig. 10B) occupy space in a second shell of residues that surround Met-742, a first shell interior helix-5 residue that contributes to the hydrophobic core and whose side chain lies adjacent to the steroid C-ring in the binding pocket. Side chains for Met-742 and third shell AF2 helix residues Val-903, Ile-906, and Leu-907 located above residue 874 are virtually superimposed atom for atom in the WT and H874Y structures (Fig. 10D), and the Met-

Dissociation rates of [ 3 H]T and [ 3 H]DHT from AR, AR-(507-919)
, and the corresponding H874Y mutants expressed from pCMV5 were determined in transfected COS cells at 37°C as described under "Experimental Procedures." Dissociation half-times were calculated as the mean Ϯ S.E. from at least three independent assays. 742 side chain clearly conforms to a single orientation for WT and H874Y AR. This overall WT AR configuration allows structural HOH3 to mediate an H-bond network from the His-874 side chain (N-⑀2) to the Met-742 backbone carbonyl and continues through HOH4 to the helix-4 Tyr-739 backbone carbonyl (Fig. 10A). Despite the bulkier phenyl hydroxyl group, Tyr-874 in the H874Y mutant appears easily accommodated with no major rearrangement of neighboring helices or side chains (Fig. 10, B and C). Tyr-874 supplies a larger side chain that extends more than 2 Å further toward helix-5 and displaces HOH3 with its phenolic hydroxyl group and presents a definitive change in H-bonding scheme. The helix-10 Tyr-874 phenolic oxygen can accept a direct 3.4 Å H-bond from the backbone amide of helix-5 Met-742 at a favorable angle of 120°(Tyr-874 C-, Oto Met-742 N) that closely aligns the Met-745 amide N-H bond vector to the assumed 120°sp 2 electrons of the Tyr-874 Oatom. In turn the Tyr-874 hydroxyl proton donates a 2.8 Å direct H-bond to the backbone carbonyl of helix-4 Tyr-739 at a favorable angle of 121°(Tyr-874 O-to Tyr-739 O, C) that closely aligns with the assumed 120°sp 2 electrons of the Tyr-739 carbonyl O atom. It is noteworthy that helix-4 residue Tyr-739 is adjacent to Gln-738, a residue whose side chain lies adjacent to the i ϩ 1 residue and displays different conformations with induced fit binding to the FXXLF or LXXLL motif and can participate in an H-bond network that links the helix-4 Met-734 CO to Lys-905 through the Gln-738 and Gln-902 side chains (Fig. 10D). The Tyr-874 phenyl presents more favorable side chain chemistry to engage C-␦2 and C-⑀1 in hydrophobic interactions with Val-903 C-␥2 (3.5 Å) and Ile-906 C-␦1 (3.4 Å) than the heterocyclic imidazole ring of WT His-874 (4.0 Å from C-␦2 to Val-903 C-␥2 and 3.7 Å from C-⑀1 to Ile-906 C-␦1) (Fig.  10D). The nearly identical T-bound crystal structures reveal that restored activation by T-bound AR-H874Y is not directly ligand-mediated or accompanied by T-induced structural rearrangement and must be driven by the H874Y mutation.

DISCUSSION
AR Activation by T and DHT-AR is unique among the family of steroid hormone receptors by having two biologically active high affinity hormones that differ in physiological potency. DHT is a morphogen required for male sexual developmental, whereas T is the major androgen in muscle and is anabolic at puberty. Normal levels of T without conversion to DHT fail to stimulate complete male genital development of the human fetus. This is evident from the human 5␣-reductase syndrome caused by a genetic defect in the enzyme that converts T to DHT (46). Activity differences between T and DHT cannot be explained by differences in transcription targets because there is no compelling evidence for differentially regulated gene sets, nor are they explained by the often reported different AR binding affinities for T and DHT. True equilibrium binding conditions may not be uniformly established, because the ligand-free AR and AR bound to T are more susceptible to degradation than AR bound to DHT leading to overestimation of T binding affinity. By measuring association and dissociation rate constants and accounting for AR instability in the absence and presence of ligand, T and DHT equilibrium binding affinities are similar (1). Nevertheless, an ϳ10-fold higher concentration of T is required to achieve the AR mediated transcriptional effects of DHT (47). A DHT-like transcriptional response by higher concentrations of T is supported by the 5␣-reductase gene knock-out mouse where a compensatory rise in circulating T levels results in masculinization at birth (48). Masculinization in humans with 5␣-reductase deficiency occurs at puberty when circulating T levels increase (49).
In this study we sought to elucidate the molecular basis for the different activities of T and DHT. Our biochemical data show that relative to DHT, T is a less potent androgen because of weaker FXXLF and LXXLL motif interactions at AF2 that are increased by the H874Y mutation. T and DHT dissociate with a R sym ϭ ¥͉I avg Ϫ I i ͉/¥I i is the data consistency, where I avg is the mean observed intensity and I i is the observed intensity. b R factor ϭ ¥͉F obs Ϫ F calc ͉/¥F obs , where F obs and F calc are the observed and calculated structure factors; R free is calculated from 3.2% of randomly selected reflections excluded in refinement and R factor calculations. c Reported as the r.m.s.d. from ideal geometry. d Structure coordinates and structure factor files are available at the Protein Data Bank web site. similar rates from AR-(507-919), but T dissociates ϳ3 times faster than DHT from WT AR and considerably slower from AR-H87Y. These results indicate that weaker AR FXXLF motif binding to AF2 results in the more rapid dissociation of T. Conversely, stronger AR FXXLF motif binding slows DHT dissociation from WT AR and T and DHT dissociation from AR-H874Y. Weaker interactions at AF2 thus appear to explain the reduced androgenic activity of T.
The similar LBD crystal structures of T-bound WT AR LBD with AR FXXLF or TIF2 LXXLL peptide to that of DHT (18,34,42) provide only subtle clues how these chemically similar ligands transmit different signals to the AF2 surface. Our structural data suggest that differences in A-ring H-bonding alter the conformational freedom of neighboring AF2 floor residue Leu-712. Structures of T-bound AR-H874Y indicate a gain-of-function arising not from chemical differences between T and DHT or altered motif binding affinity, but from replacement of a water-mediated H-bond network with direct H-bonds between external helix-10 Tyr-874 side chain and internal helix-4 Tyr-739 and helix-5 Met-742 backbone atoms. For both WT AR and AR-H874Y, small changes in H-bonding have measurable effects on motif binding at AF2 and ultimately AR transcriptional activity.
Chemical Properties of T and DHT-T is the major circulating male hormone and like DHT has 19 carbons and differs only by a ⌬4,5 double bond in ring A. With two fewer protons than the saturated ring A of DHT, the ⌬4,5 double bond of T polarizes the region and increases the negative charge at the 3-keto O and positive charge at carbon 5. Based on Coulomb's law for simple electrostatic interactions (50), these properties of T impart greater H-bonding potential that accounts for its 10-fold greater water solubility than DHT. However, water solubility and inherent hydrophilicity and hydrophobicity do not explain androgen retention times in the binding pocket because T, DHT, and R1881 dissociate with similar rates from an AR fragment containing the LBD, and R1881 dissociates from AR at a rate intermediate between T and DHT (7). Water solubility and dissociation rate of R1881 are further influenced by a 17-methyl group on ring D that introduces more hydrophobic character to a hydrophobic pocket near Met-780, Leu-704, and Leu-701.
In the same sense, the saturated nonpolar ring A of DHT is more chemically compatible with the hydrophobic environment of proximal ligand binding pocket residues than the unsaturated polar A-ring of T. Phe-764 is a conserved hydrophobic residue among steroid receptors (41) along with Val-746, Met-749, Leu-704, and Leu-707 that contact the bound ligand. Compared with T, the greater hydrophobic character and complementary shape of the DHT A-ring cannot explain the slower dissociation rate of DHT from full-length AR because T and DHT dissociate with similar rates from AR-(507-919). On the other hand, the saturated A-ring of DHT may more effectively increase AR AF2 activity by stabilizing the LBD core for higher affinity motif interactions.
Counterintuitive Hypothesis for T and DHT Activity-Modulation of FXXLF or LXXLL motif binding at AF2 by T and DHT relies on a conserved H-bond between Arg-752 N-2 and the steroid 3-keto O which is influenced by the ⌬4,5 double bond in T that imposes a more planar nature to the double bond side of ring A. Our T structures display a 3.0 Å H-bond heavy atom distance from the ring A 3-keto O to Arg-752 the same as DHT, but an angle from the more planar A-ring of T (126°) that is slightly more favorable for H-bonding than the chair configuration in DHT (110°) (Fig. 8, A-C). This contrasts a recent report indicating a better Arg-752 to 3-keto O H-bond for DHT than T (51).
Ring A chemistry also influences a network of H-bonds through structural water HOH1 that is centrally positioned between ring A of T and key residues in AF2 backbone helices 3 and 5. HOH1 can accept a proton from side chains of helix-5 Arg-752 (3.0 Å) and helix-3 Gln-711 (2.6 Å) and donate a proton to the backbone carbonyl of helix-5 Met-745 (2.7Å). 3 Based on the distance and angular relationships, these H-bond interactions for bound T and DHT satisfy 3 of 4 possible tetrahedral coordinates to water (52). T appears more likely than DHT to also accept an H-bond from HOH1 because of the more planar Our results suggest the counterintuitive hypothesis that greater H-bonding by T is detrimental to agonist activity. With shorter distances between HOH1 and the 3-keto O and Met-745 carbonyl and the negative charge character of its 3-keto O, the polarized A-ring of T may over-constrain the geometry and introduce unfavorable hydrophilic character into the hydrophobic environment of the binding pocket. The T ⌬4,5 polarized double bond is located in a hydrophobic region bounded by Met-745, Phe-764, Met-749, and Val-746 within 5 Å of the Met-745 to Val-746 amide bond. This introduces a polar atom mismatch with the 3-keto O being 3.8 Å to the Met-745 O and the presumably positive T C-5 atom within 5 Å of the Val-746 NH. In contrast, through changes also not evident in the crystal structures, it was recently suggested that a novel high affinity nonsteroidal AR modulator may influence AF2 activity by engaging more favorable hydrophobic -bonding to Phe-764 and alternative H-bonding to backbone residues in helices 3 and 5 (45).
For DHT, the nonpolar saturated boat-configured A-ring provides a neutral 3-keto O and increases the distance between HOH1 and the 3-keto O to 3.5 Å, which weakens or eliminates a second HOH1-mediated H-bond. More relaxed A-ring geometry of DHT is further evident in the 0.2 Å longer distance between HOH1 and the Met-745 O. The saturated A-ring of DHT eliminates the polarized atom mismatch and provides better hydrophobic interactions with neighboring residues listed above. Just exactly how T and DHT transmit these different signals to the AF2 surface is not clear, but in both cases the side chain of Met-745 lies above the steroid A-ring and projects toward Leu-712, a proximal residue that lies in the floor of AF2.
AF2 Residue Leu-712-Leu-712 establishes key hydrophobic contacts with i ϩ 1 Phe-23 of the AR F iϩ1 XXLF motif and i ϩ 1 Leu-745 of the TIF2 LXXLL motif. Physiological relevance for Leu-712 in AR activity is established by the L712F mutation that causes grade 3 partial androgen insensitivity without altering equilibrium androgen binding affinity (7,54). Increased bulk by Phe-712 in AR-L712F may interfere at the i ϩ 1 motifbinding site. Low intensity difference map electron density

Hydrogen bond distances and angles for WT AR LBD bound with T and AR FXXLF peptide or DHT and LXXLL peptide
Gln-711 N is oriented up for WT AR LBD-T with AR-(20 -30) FXXLF peptide (Protein Data Bank code 2Q7I) and WT AR LBD-DHT with ARA70 FXXLF peptide (Protein Data Bank code 1T63) (42) consistent with the original AR LBD-DHT structure (34). Distances and angles were measured by PyMol from heavy atom (H), proton (P), carbonyl oxygen (CO), carbonyl oxygen, or carbon (C), heavy atom to heavy atom (H2H), heavy atom to proton (H2P). Ideal water geometry is tetrahedral with ϳ109°angles.
indicates the Leu-712 side chain is equally positioned in two conformations in all of our T-bound crystal structures. Two conformers are also seen with Met-895, an AF2 helix-12 residue within ϳ4 Å of Leu-712, but more distant to the i ϩ 1 and steroid A-ring binding sites. Notably, there were single conformers of Leu-712 and Met-895 for WT AR LBD bound to DHT with FXXLF or LXXLL peptide (18,34,42).
We cannot rule out the possibility that the two conformations of Leu-712 and Met-895 are crystallographic artifacts. On the other hand, the better interactions between the 3-keto O of T to HOH1 and HOH1 to Met-745 and the polarity mismatch in the hydrophobic region near steroid carbons C-4 and C-5 suggest a mechanism not directly discerned from the structure. The effects of T appear to transmit through Met-745 to nearest AF2 floor residue Leu-712, which contacts the i ϩ 1 residue of the bound peptide. Based on proximity to the A-ring and HOH1, the signaling conduit transmits through Gln-711 and/or Met-745 to Leu-712. Of these, Met-745 is most likely because it lies above the T A-ring ⌬4,5 double bond to directly transmit A-ring chemistry to the side chain position of Leu-712 (Fig. 8). Gln-711 (adjacent to Leu-712) is less likely because two conformers of Gln-711 were in only two of four T-bound structures and was possibly influenced by a spuriously bound bufferderived glycerol. More importantly, two conformers of Gln-711 were reported for WT AR bound to DHT and FXXLF peptide (18). Two conformations of Met-895 may be a more indirect contributor to AF2 or have a cross-helix influence from Leu-712. In contrast, DHT appears to impart greater structural integrity to Leu-712 and AF2 helix-12 Met-895 in and near AF2 allowing near maximum motif binding and AR transcriptional activity. Elimination of conformational heterogeneity in Leu-712 and Met-895 by DHT may be required for optimal AF2 activity.
There are examples where side chain conformations of ligand binding pocket residues are strongly influenced by chemical architecture of the bound ligand. The C-19 methyl group of T (shown here) and DHT (34) direct the side chains of Met-745 and Trp-741 into identical positions relative to the binding pocket. The Trp-741 nitrogen in the T and DHTbound structures may establish interactions with structural HOH3. In contrast, R1881 lacks a C-19 methyl group, which allows Trp-741 the conformational freedom 4 to adopt a position where the indole nitrogen is rotated away from HOH3 (4, 7) as shown in Fig. 11. Nonsteroid ligands such as bicalutamide (44) or its S-1 agonist analog (43) extend ether linked parasubstituted phenyl groups into an open channel between the  Table 4). Arrowhead with black dashed lines indicate the direction of donated H-bonds and orange dashed lines designate potential interactions with neighboring polar atoms of WT AR LBD bound to T and AR-(20 -30) FXXLF peptide (tan) (A); WT AR LBD bound to DHT and GRIP-1-(740-752) LXXLL peptide (green) (42) (B); and the superimposition of A and B (C). Superior hydrophilic properties and a shorter distance are thought to enhance the HOH1 to T 3-keto O H-bond over that in DHT. H-bonds between protein groups and buried water molecules can stabilize structure through compensatory changes in enthalpy and entropy (65) as might occur for the H-bonding projections of Gln-711, Arg-752, Met-745, His-874, and Tyr-874. Such water-mediated H-bonds can provide favorable enthalpy but less favorable entropy than direct H-bonds in protein-ligand interactions (64,66). Deeply buried structured water molecules engaged in multiple H-bonds increase protein flexibility and vibrational entropy, whereas direct protein-mediated H-bonds provide a more rigid structure with fewer degrees of freedom compared with water-mediated H-bonds (67). Replacement of structured water by direct H-bonds in AR-H874Y may reduce vibrational entropy and stabilize AF2 helix-12 for improved FXXLF and LXXLL motif binding (5,7). Increased stabilization of AF2 helix-12 by direct H-bonding in AR-H874Y could increase the activity of T and weaker adrenal androgens. Nonsteroidal ligands such as bicalutamide or the S-1 analog have an extended para-phenyl substituent that binds in a channel between AF2 and His-874 that can directly H-bond to the same HOH3 through a fluorine atom (43,44).
We conclude that T and DHT differentially modulate AR activity by altering the AR AF2 surface response toward AR FXXLF and coactivator LXXLL motif binding mediated through a network of water-mediated H-bonds and hydrophobic interactions. T-bound WT AR acquires subtle conformational instability arising from the increased polarity of T, which decreases the effectiveness of AF2 to serve as an FXXLF and LXXLL motif-binding site. DHT-bound WT AR and prostate cancer mutant AR-H874Y bound to T or DHT fully engage the FXXLF and LXXLL motifs for maximal AR transcriptional activity. The biologically active androgens T and DHT are examples of agonist-dependent modulation of AF2 transcriptional activity that has profound physiological consequences in vivo.