Androgen-induced NH 2 - and COOH-terminal Interaction Inhibits p160 Coactivator Recruitment by Activation Function 2*

The androgen receptor undergoes an androgen-spe-cific NH 2 - and COOH-terminal interaction between NH 2 - terminal motif F XX LF and activation function 2 in the ligand binding domain. We demonstrated previously that activation function 2 forms overlapping binding sites for the androgen receptor F XX LF motif and the L XX LL motifs of p160 coactivators. Here we investigate the influence of the NH 2 - and COOH-terminal interac- tion on androgen receptor function. Specificity and relative potency of the motif interactions were evaluated by ligand dissociation rate and the stability of chimeras of transcriptional intermediary factor 2 with full-length and truncated androgen or glucocorticoid receptor. The results indicate that the androgen receptor activation function 2 interacts specifically and with greater avidity with the single F XX LF motif than with the L XX LL motif region of p160 coactivators, whereas this region of the glucocorticoid receptor interacts preferentially with the L XX LL motifs. Expression of the L XX LL motifs as a fusion protein with the glucocorticoid receptor resulted in loss of agonist-induced receptor destabilization and increased half-time of ligand dissociation. The NH 2 - and COOH-terminal interaction inhibited binding and activation by transcriptional intermediary factor 2. We con-clude that the androgen receptor NH 2 - and COOH-ter- minal interaction reduces the dissociation rate of bound androgen, stabilizes the receptor, and inhibits p160 coactivator recruitment by activation function the latter determined by incubating in the presence of a 100-fold excess of unlabeled hormone. Dissociation rate studies per- formed at 35 °C were by adding to the labeling media a 10,000-fold molar excess of unlabeled hormone (0.1 ml/well). carefully washed once 3 ml of phosphate-buffered saline at different time intervals and harvested m M pH 6.8, SDS and glycerol, radioactivity determined scintilla-tion counting. Degradation mutants °C of 5

The androgen receptor (AR) 1 is a member of the steroid receptor family of nuclear receptors that act as ligand-dependent transcriptional regulators. The AR shares with other steroid receptors an overall structural arrangement that includes a COOH-terminal ligand binding domain, central DNA binding region, and a less well conserved NH 2 -terminal region (Fig. 1). Within these domains are two major transactivation regions, activation function 1 in the NH 2 -terminal region and activation function 2 (AF2) in the ligand binding domain. The NH 2 -terminal activation function 1 region, although not well defined, requires androgen binding for transcriptional activity and appears to be critical for AR-mediated gene activation. The AF2 region in the ligand binding domain forms a putative hydrophobic binding site for the LXXLL motifs of p160 coactivators (1)(2)(3)(4)(5)(6), as recently revealed in the crystal structure of the AR ligand binding domain (7,8). The p160 group of transcriptional coregulators includes steroid receptor coactivator 1 (SRC1), transcriptional intermediary factor 2 (TIF2, SRC2), and the SRC3/TRAM1/AIB1/pCIP/ACTR/RAC3 group of activators (9,10), which are associated with histone acetyltransferase activity and can recruit CREB-binding protein, pCAF, and other coactivators required for chromatin modification (11).
The contribution of the AF2 region to AR-mediated transcriptional activity is unclear. Androgen-dependent transcriptional activity of an AR DNA and ligand binding domain fragment (AR-(507-919)) was only observed in cells that overexpressed TIF2 or SRC1 (12), which suggests that the AR AF2 inefficiently recruits p160 coactivators. We also showed recently that AF2 in the AR ligand binding domain can function in addition as a binding site for the AR NH 2 -terminal region. Mutagenesis studies indicated that the androgen-induced interaction between the AR NH 2 -and COOH-terminal (N/C) domains is mediated by two LXXLL-related sequences in the AR NH 2 -terminal region (see Fig. 1). These are FQNLF (FXXLF motif) at residues 23-27 and WHTLF (WXXLF motif) at residues 433-437 (13). In the presence of androgen, the FXXLF motif interacts with the AR AF2 in the ligand binding domain, whereas interaction of the WXXLF motif remains to be characterized (12,13). Most importantly, the N/C interaction is selectively induced by ligands that have AR agonist activity in vivo, such as the high affinity, biologically active androgens testosterone and dihydrotestosterone and the lower affinity anabolic steroids. In striking contrast, the N/C interaction is not induced by ligands that bind the AR and cause its nuclear transport but fail to induce AR-mediated gene activation in vivo (14). The N/C interaction therefore appears to be critical for AR function in vivo as further evidenced by the association of the androgen insensitivity syndrome with single amino acid mutations that disrupt the N/C interaction (12,15).
In the present study we made use of two strategies to test the effects of the N/C interaction on AR function. We investigated to what extent the AR AF2 recruits p160 coactivators in the presence and absence of the N/C interaction in wild-type and mutant AR. Second, we took advantage of the observation that the agonist-induced N/C interaction (16), which was also reported for estrogen receptor ␣ (17) and the progesterone recep-* This work was supported by NICHD, National Institutes of Health (NIH) Public Health Service Grant HD16910, by cooperative agreement U54-HD35041 as part of the Specialized Cooperative Centers Program in Reproductive Research of National Institutes of Health, and by the International Training and Research in Population and Health Program supported by the Fogarty International Center and NICHD, NIH. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 1 The abbreviations used are: AR, androgen receptor; TIF2, transcriptional intermediary factor 2; AF2, activation function 2; SRC1, steroid receptor coactivator 1; N/C, NH 2 -and COOH-terminal; GR, glucocorticoid receptor; DMEM, Dulbecco's modified Eagle medium; R1881, methyltrienolone; t1 ⁄2 , half-time; CREB, cAMP-response element-binding protein; PCR, polymerase chain reaction; DHT, dihydrotestosterone; MEM, minimum essential medium. tor (18), does not occur in the glucocorticoid receptor (GR) (19). Chimeras were created in which the three LXXLL motif region of TIF-2 was fused to the NH 2 -terminal region of AR and GR. TIF-2 was shown previously to increase the transcriptional activity of nuclear receptors through interaction of its LXXLL motifs with the AF2 region of nuclear receptors (3,5,12,13,20). The effects of an imposed N/C interaction in the TIF2(LXXLL) 3 glucocorticoid receptor chimeras were determined by measuring rates of ligand dissociation and protein degradation. The results indicate that two functional effects of the N/C interaction are agonist-induced receptor stabilization and inhibition of p160 coactivator recruitment.
Transcriptional Assays-Cell lines and transfection methods were selected to optimize transcriptional activity (CV-1 or HeLa cells) or expression levels (COS cells). Among these different cell lines we did not observe qualitative differences in response. Monkey kidney CV1 cells were plated at 4.2 ϫ 10 5 cells/6-cm dish in 5% bovine calf serum in Dulbecco's modified Eagle's medium (DMEM) containing 20 mM Hepes, pH 7.2, penicillin and streptomycin, and 2 mM L-glutamine in a 5% CO 2 incubator at 37°C. The same day 0.1 g/plate of pCMVhAR or pCMVhGR wild-type or mutant expression vectors and 5 g/plate of mouse mammary tumor virus luciferase reporter vector were separated into aliquots for 6 plates/14-ml Falcon tubes and stored at Ϫ50°C overnight. The next day 0.21 ml of H 2 0/plate and 30 l/plate of freshly prepared 2 M CaCl 2 were added to the DNAs followed by 0.24 ml 2ϫ Hepes-buffered saline/plate (0.28 M NaCl, 1.5 mM Na 2 HPO 4 , 0.05 M Hepes, pH 7.2) while vortexing. After 30 min at room temperature to allow for calcium phosphate precipitation, the mixture was briefly vortexed, and 0.475 ml was added to each plate containing 4 ml of 5% bovine calf serum in DMEM. The cells were incubated for 4 h, the media were aspirated, and the cells were incubated for 3 min with 1.5 ml of 15% glycerol in DMEM containing 5% bovine calf serum followed by a 4-ml phosphate-buffered saline wash. Cells were placed in 4 ml of serum-free, phenol red-free DMEM with and without hormones and incubated overnight. The following day, serum-free media with and without hormone were replaced, and the cells were incubated 24 h. The next day cells were washed with 4 ml of phosphate-buffered saline and aspirated dry, 0.5 ml of lysis buffer (25 mM Trizma (Tris base) phosphate, pH 7.8, 2 mM EDTA, 1% Triton X-100) was added, and 0.1 ml analyzed for luciferase activity using a Monolight luminometer.
To determine transcriptional activity induced by the GAL4 DNA binding domain and receptor ligand binding domain fusion proteins, HeLa cells plated at 3 ϫ 10 5 cells/6-cm dish in minimal essential medium (MEM) containing 10% fetal bovine serum, penicillin and streptomycin, and 2 mM L-glutamine in a 5% CO 2 incubator at 37°C. Cells were transfected with 0.25 g each/plate of the GAL4-AR ligand binding domain, GAL4-GR ligand binding domain, and GAL4-progesterone receptor ligand binding domain vectors described above, pSG5TIF2 and G5E1b-luciferase reporter, which contain 5 tandem GAL4 binding sites. The day after plating, medium was replaced with fresh MEM containing 10% fetal bovine serum. DNA was combined with 0.15 ml of EC buffer/plate (Qiagen) and 4 l of enhancer/plate, vortexed, and incubated for 5 min at room temperature. Effectene reagent (Qiagen, 4 l/plate) was added, vortexed for 10 s, and incubated for 10 min. MEM containing 10% serum was added (1 ml/plate) and mixed, and 1 ml of the DNA solution was added to each plate. After incubation overnight at 37°C, cells were washed with 4 ml of phosphate-buffered saline, and 4 ml of serum-free, phenol red-free MEM with and without hormones was added per plate as indicated. The next day cells were washed with phosphate-buffered saline and harvested in 0.5 ml of lysis buffer described above and analyzed for luciferase activity.
Ligand Dissociation Rate Studies-Monkey kidney COS cells were plated at 0.4 ϫ 10 6 cells/well in 6-well plates in 3 ml of 10% bovine calf serum in DMEM containing 20 mM Hepes, pH 7.2, penicillin and streptomycin, and 2 mM L-glutamine in a 5% CO 2 incubator at 37°C. Cells were transfected with 2 g/well pCMVhAR or pCMVhGR wild-type or mutant DNA using 0.95 ml/well of 1.08ϫ TBS (TBS: 0.14 M NaCl, 3 mM KCl, 1 mM CaCl 2 , 0.05 mM MgCl 2 , 0.9 mM NaH 2 PO 4 , and 25 mM Tris, pH 7.4) and 0.11 ml/well of 500 mg/ml DEAE-dextran. Media were aspirated, 1 ml of DNA solution was added, and the cells were incubated for 30 min at 37°C. Media were aspirated, and 3 ml of a chloroquinemedium solution (1 ml of 5 mg/ml chloroquine/100 ml DMEM containing 10% bovine calf serum) was added per well. Cells were incubated for 3 h at 37°C. Media were aspirated, and the cells were incubated for 4 min at room temperature with 1 ml of 15% glycerol in DMEM containing 10% bovine calf serum. The glycerol medium was aspirated, and the cells were washed with 3 ml of TBS. Cells were placed with 3 ml of DMEM containing 10% bovine calf serum at 37°C. After 48 h, the medium was aspirated, and 0.6 ml of labeling medium was added containing 5 nM [ 3 H]R1881 (methyltrienolone, 17␣-[methyl-3 H]R1881, 70 -87 Ci/mmol, PerkinElmer Life Sciences) for AR or 8 nM [ 3 H]dexamethasone ([1,2,4,6,7-3 H]dexamethasone, 84 Ci/mmol, Amersham Pharmacia Biotech) for GR and incubated for 2 h at 37°C. Sufficient wells were labeled to allow multiple time points for specific and nonspecific binding, the latter determined by incubating in the presence of a 100-fold excess of unlabeled hormone. Dissociation rate studies performed at 35°C were initiated by adding to the labeling media a 10,000-fold molar excess of unlabeled hormone (0.1 ml/well). Cells were carefully washed once with 3 ml of phosphate-buffered saline at different time intervals and harvested in 0.5 ml of 10 mM Tris, pH 6.8, 2% SDS and 10% glycerol, and radioactivity was determined by scintillation counting.
Immunoblots-Relative expression levels and stability of wild-type and mutant AR and GR were determined by immunoblot analysis. COS cells were plated in DMEM containing 10% bovine calf serum at 1.6 ϫ 10 6 cells/10-cm dish and the next day incubated for 3 h at 37°C with 3 ml/plate containing 10 g of expression vector DNA, 2.85 ml of 1.08ϫ TBS, and 0.33 ml of 5 mg/ml DEAE-dextran solution. The DNA mix was aspirated, and the cells were treated with 3 ml of chloroquine medium (see paragraph above) per well. Cells were incubated for 3 h at 37°C and then with 3 ml of glycerol media as described above, washed once with 8 ml of TBS, and incubated in DMEM containing 10% bovine calf serum at 37°C. The next day cells were placed in phenol red-free, serum-free media with or without 0.5 M DHT or 1 M dexamethasone and incubated for 24 h. Cells were washed in 8 ml of cold phosphatebuffered saline and harvested in 1 ml of phosphate-buffered saline, centrifuged, and solubilized in 0.2 ml of 50 mM Tris, pH 7.5, 0.15 M NaCl, 0.5% Nonidet P-40, 50 mM NaF, 1 mM NaVO 3 , 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, and Sigma protease inhibitor mixture for mammalian cells (P8340). Protein concentrations were determined using the Bio-Rad protein assay with bovine serum albumin as standard. Extracts were separated on 10% acrylamide gels containing SDS and analyzed by immunoblot for GR using rabbit polyclonal anti-human GR antibody (Affinity BioReagents) at 1:2500 dilution. AR was detected on immunoblots using mouse monoclonal AR antibody F39.4.1 raised against human AR peptide residues 302-321 (25) (Biogenex, San Ramon, CA) and used for immunoblots at 1:10,000 dilution or rabbit polyclonal antibody C19 raised against human AR peptide residues 901-919 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) and used at 0.2 g/ml. Secondary antibody goat-anti-mouse IgG or goat anti-rabbit IgG conjugated to horseradish peroxidase (Amersham Pharmacia Biotech) were used for detection by enhanced chemiluminescence (Pierce).
Degradation Rate Studies-Degradation rates of AR and mutants were determined at 35°C in the presence of 5 nM DHT by pulse-chase [ 35 S]methionine labeling in transiently transfected COS cells as previously described (19). The full-length AR mutants analyzed included AR-E897K, AR-I898T, and AR-V716R, which are mutations in the AF2 region of the ligand binding domain (12). 23 FQNLF 27 (FXXLF motif) and 433 WHTLF 437 (WXXLF motif) (Fig. 1) disrupt the N/C interaction, which results in an increase in the dissociation rate of bound androgen. As previously reported (13), the dissociation half-time (t1 ⁄2 ) of the radiolabeled synthetic androgen [ 3 H]R1881 measured at 35°C was 158 min for full-length AR and decreased to 89 min by mutating FXXLF to FXXAA. Mutation of WXXLF to WXXAA by itself had no effect on the androgen dissociation rate (13), whereas mutating both FXXLF and WXXLF to FXXAA and WXXAA decreased the t1 ⁄2 to 43 min ( Fig. 2A and Fig. 3). Mutations in both NH 2 -terminal motifs resulted in a ligand dissociation rate equal to that observed for AR-(507-919) (t1 ⁄2 44 min), a mutant that lacks the entire NH 2 -terminal region ( Fig. 2B and Fig. 3).

Specificity of the NH 2 -terminal Motif in the N/C Interaction-Mutations in AR NH 2 -terminal sequences
We used the same strategy to test for the relative effectiveness of LXXLL motifs of TIF2 to interact with AF2 in the AR ligand binding domain. TIF2-AR chimeras were created in which AR amino acid residues 1-171 containing the FXXLF motif were replaced by TIF2 residues 627-780 that include the 3 LXXLL motif region (3). TIF2(LXXLL) 3 AR had a dissociation half-time for [ 3 H]R1881 of 97 min (Fig. 2B, TIFLXL3 and Fig.  3), which suggested that the interaction of the 3 TIF2 LXXLL with the AR ligand binding domain is weaker compared with that of the single AR FXXLF motif. Specificity of the interaction in TIF2(LXXLL) 3 AR was assessed in two control experiments. Mutation of the 3 LXXLL motifs in TIF2 to LXXAA (Fig.  2B, TIFLXA3AR and Fig. 3) and of the second N/C AR interaction domain WXXLF to AXXAA similarly decreased the dissociation half-time of TIF2(LXXLL) 3 AR from 97 min to ϳ60 min, which was ϳ15 min longer than the t1 ⁄2 of 44 min for AR-(507-919) ( Fig. 2B and Fig. 3). The results support a limited interaction between the three LXXLL motifs of TIF2 and the AR ligand binding domain compared with that observed with the single AR NH 2 -terminal FXXLF sequence.
The relevance of ligand dissociation studies with the TIF2/AR chimeras was tested further by creating TIF2/GR chimeras. We chose the GR because we had found previously that deletion of the GR NH 2 -terminal region (residues 1-398) did not change the rapid dissociation rate of [ 3 H]dexamethasone (19), supporting the absence of an N/C interaction in GR. Replacing NH 2 -terminal GR amino acid residues 1-131 with the same (LXXLL) 3  When the TIF2 LXXLL motifs were mutated in the TIF2-GR chimera to TIF2(LXXAA) 3 GR-(132-777), the ligand dissociation half-time was indistinguishable from that of wildtype GR (t1 ⁄2 28 min, Fig. 2C, TIFLXA3GR and Fig. 3). These results demonstrate that it was the LXXLL motifs in the TIF2 fragment that simulated an N/C interaction in GR, causing a dramatic reduction in dissociation half-time of [ 3 H]dexamethasone. The same dependence on the LXXLL motifs was obtained using fusion proteins with the TIF2 (LXXLL) 3 region expressed at the NH 2 terminus of full-length GR (Fig. 3).
We compared the relative effects of the (LXXLL) 3 region of TIF2 with those of two other members of the p160 coactivator family. Replacement of the NH 2 -terminal 171 amino acid residues of AR with the (LXXLL) 3 motif regions of SRC1 or AIB1 indicated that these regions of SRC1 (t1 ⁄2 58 min) and AIB1 (t1 ⁄2 51 min) in the chimeras each slowed the ligand dissociation rate to less of an extent than the same region of TIF2 (t1 ⁄2 97 min) and considerably less than the reduction induced by AR FXXLF (t1 ⁄2 158 min, Fig. 3). Taken together, the data of Figs. 1-3 indicate that none of the LXXLL motifs in the 3 p160 coactivators tested was as effective as the single FXXLF motif in AR in slowing the dissociation rate of bound androgen. The results suggest that the AF2 region of AR interacts preferentially with the FXXLF motif. The results raised the question of whether these LXXLL motifs of the p160 coactivators can compete for the AR interdomain N/C interaction and activate the AR through the AF2 region of the ligand binding domain. We tested this in cotransfection assays with increasing amounts of TIF2 DNA.
Influence of the N/C Interaction on TIF2 Activation of AF2-The ability of the N/C interaction to influence AR activation by the p160 coactivators was assessed by measuring transcriptional activation of AR and AR mutants and, in control experiments, of the TIF2/GR chimeras that mimicked the N/C interaction of the AR. In initial studies, we compared the intrinsic AF2 activities of the ligand binding domains of AR, the progesterone receptor, and GR in GAL4-DNA binding domain fusion proteins and their relative activation by TIF2 in the absence of the activation function 1 NH 2 -terminal region. Intrinsic androgendependent AR AF2 activity was not detected as previously reported (12), whereas GAL4-ligand binding domain fusion proteins of progesterone receptor and GR-mediated 6-and 33fold induction, respectively, in the presence of hormone (Fig. 4). TIF2 overexpression resulted in only a 23-fold activation of GAL-AR ligand binding domain compared with the 197-and 193-fold induction of GAL-progesterone receptor ligand binding domain and GALGR-ligand binding domain, respectively, in the presence of hormone (Fig. 4). The results indicate that compared with the progesterone receptor and GR ligand binding domain, the AR ligand binding domain has inherently weak AF2 activity that could be overcome to some extent by TIF2 overexpression.
The high transcriptional activity of the AR NH 2 -terminal activation function 1 region (amino acid residues 142-337, Fig. 1) (26) makes it difficult to measure AF2 activity in the presence of activation function 1. A deletion mutant (AR⌬142-337) was therefore used in which the NH 2 -terminal transactivation domain residues 142-337 were deleted, but the N/C interaction remained intact, as indicated by two hybrid assays (12) and a ligand dissociation rate equivalent to wild-type AR (19). Increasing the amount of transfected pSG5TIF2 expression vector DNA from 0.2 to 5 g was relatively ineffective in activating AR⌬142-337, with only a 12- fold activation detected with 5 g TIF2 DNA (Fig. 5A). In striking contrast, with a construct in which the N/C interaction was weakened by changing the 23 FXXLF 27 motif to 23 FXXAA 27 (AR⌬142-337FXXAA, Fig. 5A), TIF2 was about 100 times more effective in increasing AR-mediated transactivation based on the amount of transfected TIF2 DNA. A similar activation of 12-13-fold was observed using 5 g of TIF2 with AR⌬142-337 or 0.05 g of TIF2 with AR⌬142-337FXXAA. However, TIF2 activation of AR⌬142-337FXXAA was less than that observed when the entire NH 2 -terminal region was deleted (Fig. 5A). The weaker activation by TIF2 of AR⌬142-337FXXAA compared with that with AR-(507-919) likely resulted from the presence of the 433 WHTLF 437 sequence in AR⌬142-337FXXAA, as this WXXLF motif contributes to the N/C interaction (13) and therefore may partially inhibit TIF2 recruitment by AF2.
We also tested transcriptional coactivation of AF2 using a TIF2 mutant in which all three LXXLL motifs were changed to LXXAA. TIF2(LXXAA) 3 did not coactivate with AR⌬142-337, AR⌬142-337FXXAA, or AR-(507-919) above the low intrinsic levels observed with the mutant AR alone (Fig. 5B). Especially striking was the decrease in transcriptional activation with AR⌬142-337FXXAA and AR-(507-919) from 179-and 290-fold with TIF2 to near background levels with the TIF2(LXXAA) 3 mutant. The results of Figs. 5, A and B, suggest that the androgen-induced AR N/C interaction mediated by the FXXLF and WXXLF motifs inhibits p160 coactivator interaction with AF2 in the ligand binding domain. Mutations in the FXXLF region were required to significantly overcome the androgeninduced inhibition imposed by the N/C interaction on p160 coactivator recruitment by AF2.
Similar dose-response studies were performed using TIF2-GR chimeras. Introducing the putative N/C interaction in TIF2-(LXXLL) 3 GR-(132-777) resulted in a reduced response to TIF2 activation compared with that observed with the TIF2(LXXAA) 3 GR-(132-777) mutant (Fig. 5C). A 10-fold higher amount of TIF2 was required to activate TIF2(LXXLL) 3 GR-(132-777) (0.5 g of TIF2, 160-fold) above background levels compared with 0.05 g, the lowest level of TIF2 tested with the LXXAA mutant (142-fold, Fig. 5C). Thus, in agreement with results with AR, the N/C interaction imposed in GR by insertion of the NH 2terminal LXXLL motifs attenuated activation of the receptor by TIF2. It is nevertheless noteworthy that increased TIF2 expression (5 g of pSG5TIF2 DNA) was effective in overcoming the inhibition created by the artificially induced N/C interaction in GR, suggesting that a coregulatory protein with a binding region of similar or greater affinity for AF2 can compete more efficiently for the N/C interaction if it is expressed at sufficiently high levels.
Effect of the FXXLF, WXXLF, and LXXLL Motifs on Receptor Stabilization-An unusual property of the AR is its dramatic stabilization by agonist binding (27), which previous data suggested is mediated by the N/C interaction (15). In contrast, most steroid receptors including the estrogen receptor ␣ (28,29), thyroid hormone receptor (30), GR (31), and progesterone receptor (32) undergo agonist-induced decreases in receptor levels. To further investigate the contribution of the N/C interaction to androgen-induced AR stabilization, we determined the effects of mutations in the NH 2 -terminal FXXLF and WXXLF interaction motifs on AR levels by immunoblot analysis. The addition of 0.5 M DHT to the growth media resulted in a dramatic increase in AR protein (Fig. 6A, lanes 2 and 3), indicating androgen-induced receptor stabilization. In contrast, the FXXAA as well as the FXXAA/AXXAA double mutant AR proteins were detected at similar levels in the absence and presence of DHT (Fig. 6A, lanes 4 -7). We also noted that in the absence of androgen there was a reproducible increase in the levels of these AR mutants relative to wild-type AR. The results support a role of the FXXLF and WXXLF-mediated N/C interaction in ligand-induced AR stabilization.
We made use of the TIF2-GR chimeras to substantiate the role of the N/C interaction in receptor stabilization. Full-length GR undergoes a striking agonist-induced decrease in receptor levels with the addition of 1 M dexamethasone (Fig. 6C, lanes  1 and 2). In contrast, the TIF2-GR chimera TIF2(LXXLL) 3 GR-(132-777), which was shown above to dramatically slow the dissociation half-time of bound [ 3 H]dexamethasone (see Fig. 2C and Fig. 3), exhibited loss of dexamethasone-induced GR destabilization (Fig. 6C, lanes 3 and 4). When the last two leucine residues in each of the three LXXLL motifs were mutated to alanine in TIF2(LXXAA) 3 GR-(132-777), which was shown above to reverse the ligand dissociation half-time to that of wild-type GR, degradation of the TIF2-GR chimera was indistinguishable from that of wild-type GR (Fig. 6C, lanes 5 and 6). Similar results were observed with the TIF2(LXXLL) 3 GR chimeras in which the TIF2 fragment was expressed as a fusion FIG. 4. Intrinsic AF2 activity of the AR, progesterone receptor, and GR ligand binding domains. DNA (0.25 g/plate) for the GAL4 DNA binding domain and ligand binding domains for AR (GALAR-(624 -919)), progesterone receptor (GALPR-(636 -933)) and GR (GALGR-(486 -777)) were transfected into HeLa cells with or without 0.25 g of pSG5TIF2 (TIF2) and 0.25 g of G5E1b-luciferase reporter using Effectene as described under "Experimental Procedures." Cells were incubated in the absence and presence of 50 nM DHT for AR, 50 nM R5020 for progesterone receptor, and 50 nM dexamethasone for GR. Luciferase activity was determined as described under "Experimental Procedures," and the fold induction relative to the activity was determined in the absence of hormone is shown above the bars.
protein with full-length GR, although the extent of stabilization was less pronounced (Fig. 6C, lanes 7-10). Taken together the results indicate that the agonist-induced N/C interaction increases the half-time of ligand dissociation and allows for agonist-induced receptor stabilization and the absence of agonist-induced destabilization that is characteristic of wild-type AR but not GR.
Degradation rates of AR and several AR mutants were determined using [ 35 S]methionine pulse-chase labeling. As summarized in Table I, mutation of the NH 2 -terminal FXXLF and WXXLF motifs resulted in degradation rates intermediate between those of full-length AR and AR-(507-919), as determined in COS cells at 35°C. Increased AR degradation in the presence of 5 nM DHT compared with that of wild-type AR was also observed for AR AF2 mutants E897K, I898T, and V716R. These mutations were shown previously to disrupt the N/C interaction (12). The results support a critical role for the N/C interaction in androgen-induced AR stabilization.

DISCUSSION
Several lines of evidence indicate that the agonist-induced N/C interaction between the LXXLL-like sequences 23 FQNLF 27 and 433 WHTLF 437 in the AR NH 2 -terminal region and the AF2 region in the AR ligand binding domain is specific for AR and required for functional activity. The N/C interaction is critical to AR function, because AF2 mutations that disrupt the N/C interaction without affecting the equilibrium ligand binding affinity cause the androgen insensitivity syndrome, whereas mutations that disrupt p160 coactivator binding without affecting the N/C interaction have wild-type activity (12,15). The N/C interaction slows ligand dissociation and increases AR stability yet interferes with p160 coactivator recruitment. Like the LXXLL motifs of p160 coactivators (1,5), the AR FXXLF motif forms an amphipathic ␣-helix that interfaces within the hydrophobic groove of AF2.
Results of experiments with chimeric receptors indicate the AR N/C interaction has greater specificity and potency compared with the interaction of AF2 with the LXXLL motifs of FIG. 5. Effect of TIF2 expression on AR, GR, and TIF2-receptor chimera-mediated transactivation. CV1 cells were transfected using calcium phosphate precipitation as described under "Experimental Procedures." with 5 g of mouse mammary tumor virus luciferase reporter vector and 100 ng of AR or GR expression vector DNA. Shown are the luciferase light units, determined in the absence and presence of 1 nM DHT or 10 nM dexamethasone with fold induction indicated above the bars. In each part, the data are representative of at least three independent experiments. A, cells were transfected in the absence or with increasing amounts of pSG5TIF2 DNA from 0.05 to 5 g together with AR⌬142-337 that lacks AF-1 residues 142-337, AR⌬142-337FXXAA in which the 23 FXXLF 27 motif was mutated to FXXAA and activation function 1 deleted, and the DNA binding domain and ligand binding domain fragment AR-(507-919). B, CV1 cells were transfected using 0.1 g of the AR mutants, 5 g of mouse mammary tumor virus luciferase reporter without or with 5 g of pSG5TIF2 (TIF2) or pSG5TIF2(LXXAA) 3 (TIF2-LXXAA). Cells were incubated in the absence and presence of 1 nM DHT, and luciferase activity and fold induction relative to the activity determined in the absence of DHT are indicated. C, TIF2(LXXLL) 3 GR-(132-777) and TIF2(LXXAA) 3 GR-(132-777) were expressed in CV1 cells in the absence and presence of increasing amounts of pSG5TIF2 DNA as indicated. Cells were incubated in the absence and presence of 10 nM dexamethasone. p160 coactivators. In previous studies, AR/GR chimeras only slightly increased ligand dissociation half-times (19), suggesting that the FXXLF motif interacts less well with the GR AF2 than this region interacts with the TIF2-derived LXXLL motifs. In unpublished studies, 2 glutathione S-transferase affinity matrix assays showed that TIF2 LXXLL motif-containing fragments interact with the estrogen receptor ␣ ligand binding domain, whereas the AR FXXLF fragment does not.
The AR AF2 region only weakly recruits p160 coactivators compared with the AF2 region of the progesterone or glucocorticoid receptor. This relatively weak interaction is further hindered by the N/C interaction. The hinge region of AR (residues 628 -646) was reported to contribute to the low transcriptional activity of AR AF2 (34). However in our unpublished studies, 2 deletion of hinge residues 624 -647 only minimally increased AR AF2 transcriptional activity of a GAL4 fusion protein with the AR ligand binding domain expressed in HeLa cells and resulted in a similar increase in the N/C interaction. The lower transcriptional activity of the AR AF2 region relative to other nuclear receptors more likely results from sequence divergence-induced structural differences and by the N/C interaction.
One functional consequence of the agonist-induced AR N/C interaction may be to present a novel surface to attract ARspecific coactivators. A LIM domain, heart-specific protein FHL2 (35) is a reported AR coactivator that interacts with full-length AR but not with the NH 2 -or COOH-terminal region (36), suggesting it recognizes an N/C interaction-induced conformation. The AR N/C interaction may contribute to the recognition of weaker androgen response elements whose regulation is androgen-specific (37).
Most steroid receptors undergo ligand-induced down-regulation resulting from ubiquitin-mediated proteolysis by the proteosome. Rates of receptor degradation have been correlated with activation potency (38), and activation domains and degradation signals can overlap (39). Proteosome-mediated degradation of estrogen receptor ␣ (29, 40) was linked with coactivator recruitment and transcriptional potency. Mutations in the estrogen receptor ␣ AF2 region at residues critical for coactivator recruitment stabilized the receptor (41), raising the possibility that p160 coactivator interaction with ligand-bound receptors is required for receptor degradation. The thyroid hormone receptor is also rapidly degraded by the proteosome (30), but ligand-dependent degradation of retinoid X receptor did not require transcriptional activity or interaction with p160 coactivators (42).
In our unpublished studies 2 AR degradation is mediated by the proteosome; however, in contrast to most nuclear receptors, AR and the vitamin D receptors (43) undergo agonist-induced stabilization. For the vitamin D receptor, inhibition of ubiquitin-proteosome-mediated degradation amplified the transcriptional response (43)(44)(45). For AR, it remains to be established whether in vivo transcriptional activity at certain androgen response elements requires the N/C interaction or the resulting agonist-induced increase in AR stability. The 5-fold reduced dissociation half-time of bound dexamethasone and reversal of the dexamethasone-induced decrease in GR levels in the TIF2-GR chimeras dramatically demonstrated the influence of the N/C interaction on receptor stabilization. This artificial N/C interaction in GR resulted in ligand dissociation and stability properties similar to those of wild-type AR.  Dose-response transcription assays where TIF2 is transiently overexpressed as shown here indicate that the AR N/C interaction blocks AF2 recruitment of p160 coactivators. Transcriptional inhibition is predicted for other p160 coactivators such as SRC1 that interact with AF2 through LXXLL motifs. It is not known, however, to what extent inhibition of p160 coactivator recruitment by the N/C interaction limits the activity of these coactivators in vivo. Neither is it known how the interaction of other coactivators such as p300/CREB-binding protein or ARA70 with the AR is affected by the N/C interaction. Previous studies indicated that mutation of lysine 720 reduced the interaction of p160 coactivators with AF2, but this mutant AR retained full transcriptional activity (12). The corresponding lysine in mouse estrogen receptor ␣ (lysine 366) was required for p160 coactivator recruitment and estrogen receptor ␣ transcriptional activity (46). p160 coactivator overexpression in transient transfection experiments (12) and possibly in vivo in some disease states (47) may increase steroid receptor transcriptional activity. This was shown for AR where coactivation by TIF2 was mediated by the AR NH 2 -terminal and COOH-terminal regions (12,48). However these types of studies do not address whether p160 coactivators increase AR activation when coactivators are present at normal physiological levels. We showed recently that TIF2 and SRC1 are almost undetectable in human benign hyperplastic prostatic tissue compared with greatly increased levels in recurrent prostate cancer (47). In recurrent prostate cancer, overexpression of p160 coactivators may effectively compete for the N/C interaction through their interaction with the NH 2and COOH-terminal domains.
N/C interdomain interactions have been reported for other steroid and nuclear receptors; however, the potency and functional consequences differ. Intracellular phosphorylation of the A/B NH 2 -terminal domain of PPAR-␥ reduced ligand binding affinity (49). Interactions between the hinge-amino-terminal regions of the progesterone receptor contributed to dimerization and increased activation (50), whereas GR seems to lack an N/C interaction based on ligand dissociation (19) and direct interaction assays (51). Functional synergism between the NH 2 -and COOH-terminal domains was also reported (52,53). For estrogen receptor ␣ and ␤, agonist-induced synergism was mediated by p300/CREB-binding protein and TIF2 binding to the NH 2 -terminal region (54,55). Similarly, in unpublished studies 2 we could not demonstrate direct NH 2 -and COOHterminal domain interactions for estrogen receptor ␣ in glutathione S-transferase affinity matrix studies. CREB-binding protein and SRC1a were reported to enhance the AR N/C interaction through an indirect mechanism where coactivators act as adapters between activation function 1 and AF2 (56). The glutamine-rich region of SRC1e at residues 1053-1123 interacted with AR NH 2 -terminal residues 360 -494, suggesting SRC1 acts as a bridging molecule to mediate the N/C interaction (33). p160 coactivators also reportedly increased the AR N/C interaction, suggesting the N/C interaction is indirect. However, the identification of FXXLF and WXXLF motifs in the AR NH 2 -terminal region (13) that bind the AF2 region of the ligand binding domain (12) provides strong evidence that the AR N/C interaction is direct. Moreover, transcriptional assays indicate that the agonist-induced N/C interaction interferes with the recruitment of p160 coactivators.
The AR FXXLF-AF2 N/C interaction is specific for the biologically active androgens testosterone and DHT and for anabolic steroids. However, the identification of the androgeninduced FXXLF-AF2 N/C interaction suggests that AR antagonists can be identified that inhibit or agonists that promote interactions with other FXXLF or LXXLL motif-contain-ing proteins. Pharmaceutical ligands that bind AR with moderate or high affinity may promote interactions with related peptide sequences present in coregulatory proteins. Whether the FXXLF motif or related sequences occur in AR-specific coactivators that have sufficient affinity to compete for the agonist-induced N/C interaction or are induced to interact by other ligands remains to be established. In the presence of such ligands, coregulatory proteins might inhibit or compete for the androgen-induced N/C interaction to regulate tissue-selective AR-mediated gene activation.