FXXLF and WXXLF Sequences Mediate the NH2-terminal Interaction with the Ligand Binding Domain of the Androgen Receptor*

The nuclear receptor superfamily members of eukaryotic transcriptional regulators contain a highly conserved activation function 2 (AF2) in the hormone binding carboxyl-terminal domain and, for some, an additional activation function 1 in the NH2-terminal region which is not conserved. Recent biochemical and crystallographic studies revealed the molecular basis of AF2 is hormone-dependent recruitment of LXXLL motif-containing coactivators, including the p160 family, to a hydrophobic cleft in the ligand binding domain. Our previous studies demonstrated that AF2 in the androgen receptor (AR) binds only weakly to LXXLL motif-containing coactivators and instead mediates an androgen-dependent interaction with the AR NH2-terminal domain required for its physiological function. Here we demonstrate in a mammalian two-hybrid assay, glutathione S-transferase fusion protein binding studies, and functional assays that two predicted α-helical regions that are similar, but functionally distinct from the p160 coactivator interaction sequence, mediate the androgen-dependent, NH2- and carboxyl-terminal interaction. FXXLF in the AR NH2-terminal domain with the sequence23FQNLF27 mediates interaction with AF2 and is the predominant androgen-dependent interaction site. This FXXLF sequence and a second NH2-terminal WXXLF sequence 433WHTLF437 interact with different regions of the ligand binding domain to stabilize the hormone-receptor complex and may compete with AF2 recruitment of LXXLL motif-containing coactivators. The results suggest a unique mechanism for AR-mediated transcriptional activation.

Nuclear receptors facilitate ligand-dependent increases of gene transcription by direct interactions with nuclear coactivators. p160 coactivators have histone acetyltransferase activity (1) and interact with nuclear receptors through their ligand binding and NH 2 -terminal regions (2)(3)(4). Additional acetyltransferases p300/cAMP response element binding protein and p300/cAMP response element binding protein associated factor interact with the p160 coactivators (5)(6)(7). Nuclear receptors also interact with multiprotein complexes referred to as thyroid hormone receptor-associated proteins (8), activator-recruited cofactor (9), or vitamin D receptor-interacting protein complex (10). The p160 coactivators and at least one of the thyroid hormone receptor-associated proteins/activator-recruited cofactor/vitamin D receptor-interacting protein subunit 205 interact in a ligand-dependent manner with activation function 2 (AF2) 1 in the ligand binding domain (LBD) of nuclear receptors through the consensus sequence LXXLL, where L is leucine and X is any amino acid (11)(12)(13)(14)(15)(16). Crystal structures of nuclear receptor LBDs have shown that a hydrophobic cleft within a multilayered ␣-helical structure serves as the LXXLL coactivator binding surface AF2 (17,18). In the estrogen receptor, agonist binding positions helix 12 over the binding cavity to complete the AF2 surface (19,20), whereas binding of an antagonist such as 4-hydroxytamoxifen displaces helix 12 (21) causing an LXXLL-like sequence in helix 12 to mimic and thereby block coactivator binding (22).
It has become apparent that the AF2 region overlaps with regions that serve as the binding site for a variety of LXXLLrelated sequences as recently shown for corepressor binding (23)(24)(25). Furthermore, we demonstrated that the AR AF2 region mediates an androgen-dependent NH 2 -terminal/carboxylterminal (N/C) interaction (26). Therefore we investigated the possibility that AF2 in the AR LBD interacts with an LXXLLrelated sequence in the NH 2 -terminal domain. In this report, evidence is presented that sequences similar to but distinct from the LXXLL core sequence mediate a direct interaction between the NH 2 -terminal and carboxyl-terminal regions of AR. The FXXLF core sequence 23 FQNLF 27 in the AR NH 2terminal domain binds AF2 in the carboxyl-terminal region in an androgen-dependent manner. In addition, a second motif in the NH 2 -terminal region WXXLF with the sequence 433 WHTLF 437 binds to a region of the LBD outside of AF2. Interaction of these NH 2 -terminal sequences with the LBD slows the dissociation rate of bound androgen. Sequence specificity was indicated since FXXLF could not be functionally replaced by an LXXLL core sequence.

EXPERIMENTAL PROCEDURES
Transcription Assays and Plasmids-Mammalian two-hybrid N/C interaction assays were performed in Chinese hamster ovary (CHO) * This work was supported by NICHD Grant HD-16910 from the National Institutes of Health, through cooperative agreement U54-HD-35041 as part of the Specialized Cooperative Centers Program in Reproductive Research, and by the International Training and Research in Population and Health Program supported by the Fogerty International Center and the NICHD, National Institutes of Health. 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 (27,28). VPAR-(1-660) codes for the herpes simplex virus VP16 protein transactivation domain residues 411-456 expressed as a fusion protein with AR NH 2 -terminal and DNA binding domain residues 1-660. GALAR-(624 -919) codes for a fusion protein of the Saccharomyces cerevisiae GAL4 DNA binding domain residues 1-147 and AR LBD residues 624 -919. Deletions within VPAR-(1-660) (⌬179 -199, ⌬394 -405, and ⌬429 -439) were created by two polymerase chain reactions (PCR) using oligonucleotide primers with appropriate deletions. The reporter vector G5E1bLuc contained five GAL4 DNA-binding sites and the luciferase reporter coding region (29). CHO cells were plated at 0.425 ϫ 10 6 /6-cm dish and transfected using DEAE-dextran (27,28). Cells were incubated for 24 h in the absence and presence of increasing concentrations of dihydrotestosterone (DHT) as indicated and harvested in lysis buffer (Ligand Pharmaceuticals). Luciferase activity was determined using a Monolight 2010 luminometer (Analytical Luminescence Laboratory, San Diego).
Androgen Binding and Dissociation Assays-Apparent equilibrium binding affinity was determined in whole cell binding assays at 37°C by plating monkey kidney COS1 cells (0.2 ϫ 10 6 /well of 12-well tissue culture plates) and transfecting 0.1 g of pCMVhAR wild-type and mutant DNA using DEAE-dextran (26). Twenty four h after transfection, cells were incubated with increasing concentrations of [ 3 H]R1881 from 0.1 to 5 nM in the presence and absence of a 100-fold excess unlabeled R1881. Cells were incubated at 37°C for 2 h, washed, and harvested in 0.5 ml of 2% SDS, 10% glycerol, and 10 mM Tris, pH 6.8. Radioactivity was determined by scintillation counting. For determination of dissociation rates of [ 3 H]R1881, COS cells were plated at 0.4 ϫ 10 6 cells/well in 6-well plates and transfected with 2 g of pCMVhAR DNA/well using DEAE-dextran. Cells were incubated for 2 h at 37°C with 5 nM [ 3 H]R1881 in the presence and absence of a 100-fold excess unlabeled R1881. Dissociation was started by the addition of 50 M unlabeled R1881, and the cells were incubated for increasing times at 37°C up to 3 h, washed once, and harvested in SDS buffer as described above, with radioactivity determined by scintillation counting.
GST fusion protein binding studies were performed essentially as described previously (26). GST fusion proteins with different regions of the human AR NH 2 -terminal region were expressed from XL1-Blue Escherichia coli cells treated with 0.5 mM isopropyl-1-thio-␤-D-galacto-pyranoside for 3 h during log phase growth. Glutathione-agarose beads (Amersham Pharmacia Biotech) were incubated for 1 h at 4°C with sonicated bacterial supernatants containing the GST-AR fusion proteins. Beads were washed with 0.5% Nonidet P-40, 1 mM EDTA, 0.1 M NaCl, 0.02 M Tris, pH 8.0, and combined with 25 Ci of [ 35 S]methionine (NEN Life Science Products)-labeled human AR LBD (AR amino acid residues 624 -919) using TNT T7 Quick-coupled Transcription/Translation System (Promega) and incubated for 2 h at 4°C in the absence or presence of 0.2 M DHT. Beads were washed, eluted with SDS, and analyzed on 12% acrylamide gels containing SDS.

RESULTS
FXXLF in the AR NH 2 -terminal Region-The androgen-dependent interaction between the AR NH 2 -terminal and carboxyl-terminal (N/C) domains occurs in the regions of predicted ␣-helices 3, 4, and 12 that comprise AF2 of the LBD, overlapping with the binding site for p160 coactivators (26). Because LXXLL motifs mediate the interaction of p160 coactivators with AF2 of nuclear receptors (11)(12)(13)(14)(15)(16), it raised the possibility that an LXXLL-like motif (13)(14)(15) in the AR NH 2 -terminal region has a similar function to mediate the N/C interaction. Sequence analysis of the AR NH 2 -terminal regions previously implicated in the N/C interaction (27) revealed four predicted amphipathic ␣-helices that resemble LXXLL core sequences at residues 21-34, 351-359, 395-405, and 432-434 with another predicted outside these regions (33) at residues 177-201. We tested wild-type, deletion, and single amino acid mutations of these ␣-helical regions in a variety of assays. These included the mammalian two-hybrid N/C interaction assay performed in CHO cells and functional assays that included the effects of the mutations on [ 3 H]androgen dissociation rate and transcriptional activation. In vitro domain interactions were also directly tested using E. coli-expressed GST fusion proteins. For the mammalian two-hybrid assay, wild-type and mutant VPAR-(1-660) coding for the AR NH 2 -terminal and DNA binding domains were cotransfected with GALAR-(624 -919) expressing the LBD (Fig. 1). Similar expression levels of wildtype and mutant VPAR-(1-660) vectors were verified by immunoblot analysis using AR52 antibody (data not shown). Also all of the VPAR-(1-660) constructs, when cotransfected with GALAR-(1-503) coding for just the NH 2 -terminal domain, resulted in 2.2-3.2-fold induction of luciferase activity indicative of the NH 2 -terminal/NH 2 -terminal AR interaction previously reported (27), thus confirming similar expression levels of the VPAR-(1-660) vectors.
Wild-type and mutant VPAR-(1-660) coding for the AR NH 2terminal and DNA binding regions were coexpressed in the mammalian two-hybrid assay with GALAR-(624 -919) coding for the AR LBD. The VPAR-(1-660) fragment with wild-type AR sequence induced 38 Ϯ 19-fold luciferase activity relative to the no hormone control (Fig. 1). Deletion of NH 2 -terminal residues 179 -199, 394 -405, or 429 -439 had no significant effect on the interaction (Fig. 1). In contrast, ⌬9 -28 reduced the interaction to only a 1.5-fold increase over the no hormone control and ⌬339 -499 to 10-fold relative to the control ( Fig. 1) (27). These results suggested interactions between the LBD and residues 9 -28 and 339 -499. Because 23 FQNLF 27 lies within the 9 -28 region and resembles the LXXLL core sequence, it was investigated further by mutagenesis. Changing phenylalanine 23 to alanine (F23A) or leucine 26 and phenylalanine 27 to alanine (L26A/F27A) reduced the N/C interaction to 1.6 Ϯ 0.2-fold over background levels, whereas a flanking mutation of glutamine 28 to alanine (Q28A) resulted in an interaction similar to that of wild-type AR (Fig. 1). Changing both phenylalanine residues in 23 FQNLF 27 to leucine (F23L/ F27L) to recreate a consensus LXXLL sequence resulted in only a 2-fold induction of luciferase activity (Fig. 1). These results indicated that an FXXLF motif was required in the N/C inter-action. Specificity of the 23 FXXLF 27 interaction with the AR LBD was indicated by the greatly reduced interaction when FXXLF was substituted by LXXLL.
The requirement for the FXXLF motif in the N/C interaction was also investigated using AR NH 2 -and carboxyl-terminal fragments in transient transfection assays using the MMTVluciferase reporter. As previously reported, the AR DNA binding and ligand binding domain fragment AR-(507-919) had negligible transcriptional activity in the presence of androgen ( Fig. 2) indicative of the lack of AF2 transcriptional activity (32) and weak recruitment of p160 coactivators by the AR AF2 region (26). In cotransfection studies, wild-type NH 2 -terminal domain residues AR-(1-503) interacted with AR-(507-919) to stimulate an 18 -28-fold increase in luciferase activity (Fig. 2). In contrast, NH 2 -terminal fragment AR-(1-503) with residues 14 -150 deleted (⌬14 -150) or with the mutated sequence 22 FQNAA 27 (L26A/F27A) failed to interact with AR-(507-919), supporting an important role for 23 FQNLF 27 in the N/C interaction. Androgen-induced luciferase activity was 6 -12-fold with deletion of NH 2 -terminal residues 142-337 comprising the AR transactivation domain (⌬142-337) which could have resulted from reduced transactivation by AR rather than a decrease in the N/C interaction as previously suggested (27). ⌬339 -499 also reduced the interaction but less effectively than did the L26A/F27A mutation (Fig. 2). The results support the requirement for 23 FQNLF 27 in the N/C interaction and the presence of a second interaction site between residues 339 and 499. Similar results using fusion proteins of the AR NH 2terminal region linked to the VP16 transactivation domain supported the role of these two regions in the N/C interaction (Fig. 2).
Effect of the FXXLF Motif on Androgen Dissociation Kinetics-To establish a functional effect of the NH 2 -terminal mutations on AR activity and to obtain additional evidence for the putative second interacting site, we measured the androgen dissociation rate using the synthetic radiolabeled androgen [ 3 H]R1881. These studies were based on previous studies that certain mutations in AF2 of the AR LBD cause androgen insensitivity by disrupting the N/C interaction. Although the equilibrium androgen binding affinity was unaffected by these mutations, the dissociation rate of bound androgen increased FIG. 2. N/C interaction using AR fragments and the MMTVluciferase reporter vector. PCMVhAR-(507-919) (AR507-919, 50 ng/plate) coding for the AR DNA binding domain and LBD was coexpressed in CV1 cells without or with 0.5 or 1 g of pCMVhAR-(1-503) (AR1-503) coding for the AR NH 2 -terminal domain with wild-type sequence or the indicated mutant sequence or with 1 or 5 g of the VP16 activation domain-AR NH 2 -terminal domain fusion protein VPAR-(1-503) with wild-type sequence or with the indicated deletions. Assays were performed in CV1 cells using the MMTV-luciferase reporter (5 g) in the presence and absence of 10 nM DHT. Parent expression vector control pCMV5 (p5) lacked AR sequence. LFAA indicates the L26A/ F27A mutation ( 23 FQNAA 27 ). Fold induction of luciferase activity determined relative to the activity in the absence of DHT is indicated above the bars and is representative of at least three independent experiments. suggesting a corresponding increase in association rate (26,28). The results supported a role for the N/C interaction in slowing the androgen dissociation rate in wild-type AR (31).
Additional evidence that the N/C interaction influences AR ligand binding kinetics is that coexpression of the DNA binding domain and LBD fragment AR-(507-919) with NH 2 -terminal fragment AR-(1-660) slows the dissociation of [ 3 H]R1881 by 2-fold (28). In contrast, no effect was seen by coexpression of the nuclear receptor coactivators transcriptional mediator/intermediary factor 2 (TIF2), vitamin D receptor-interacting protein 205, amplified in breast cancer-1 (AIB1) or protein inhibitor of activated signal transducer and activator of transcription-1 (PIAS1), each of which contains multiple consensus LXXLL motifs (10, 34 -36). Dissociation of [ 3 H]R1881 from the carboxyl-terminal AR-(507-919) fragment (half-time of dissociation t1 ⁄2 of 43 Ϯ 3 min, see Fig. 4) was unchanged with t1 ⁄2 of 42 Ϯ 4 min at 37°C when coexpressed with each of these coactivators (data not shown). The results are consistent with a weak interaction of these coactivators with the AR AF2 region compared with the interaction with the AR NH 2 -terminal domain (26) and suggest a correspondingly higher apparent binding activity of the 23 FXXLF 27 core sequence for AF2 compared with the LXXLL motif.
The functional significance of the NH 2 -terminal FXXLF core sequence was also evident by increased dissociation rates of bound androgen with the introduction of mutations in the FXXLF region in full-length AR. ⌬9 -28, F23A, or L26A/F27A increased the dissociation rate of [ 3 H]R1881 at 37°C to t1 ⁄2 of 72-74 min compared with t1 ⁄2 of 144 Ϯ 24 min for wild-type AR (Fig. 3). In contrast, mutation of the flanking carboxyl glutamine (Q28A) had no effect on dissociation rate (t1 ⁄2 of 162 Ϯ 24 min) (Fig. 3) or N/C interaction (Fig. 1). Furthermore, changing 23 FQNLF 27 to the consensus LXXLL sequence (F23L/F27L) increased the dissociation rate from t1 ⁄2 of 144 Ϯ 24 min for wild-type AR to t1 ⁄2 of 84 Ϯ 9 min (Fig. 3), which was similar to t1 ⁄2 74 Ϯ 5 min for F23A and L26A/F27A, supporting that an LXXLL motif is much less effective than FXXLF in mediating the N/C interaction. None of these mutations or those described below changed significantly the apparent equilibrium binding affinity of [ 3 H]R1881 that ranged from 0.3 to 0.8 nM (data not shown), suggesting a corresponding increase in association rate when the dissociation rate increased. The results indicate that the AR NH 2 -terminal FXXLF motif has an important role in the N/C interaction which results in a reduced rate of androgen dissociation.
To test for possible functional equivalence between the FXXLF and WXXLF core sequences that might have depended on their position within the AR NH 2 -terminal sequence rather than the binding motif itself, we replaced phenylalanine 23 with tryptophan to thereby replace 23 FXXLF 27 with 23 WXXLF 27 . The two-hybrid N/C interaction of this F23W mutant was only 1.9 -2.2-fold compared with 35-59-fold for the wild-type control at 1 and 100 nM DHT. WXXLF therefore appears to have binding properties distinct from FXXLF such that the two motifs are not functionally equivalent. Whereas FXXLF appears to be the primary NH 2 -terminal interaction site for the AF2 region of the LBD, the core sequence WXXLF seems to contribute to the interaction with the AR LBD by further slowing the androgen dissociation rate.

Distinct Binding Regions of the Two NH 2 -terminal Motifs-
The AR NH 2 -terminal sequences that mediate the N/C interaction were investigated further using E. coli-expressed GST fusion proteins. AR LBD residues 624 -919 were labeled with [ 35 S]methionine and incubated in the presence of 0.2 M DHT and GST-AR NH 2 -terminal fragments of increasing length. Similar expression levels of the GST AR NH 2 -terminal domain fusion proteins were confirmed by Coomassie staining of SDSpolyacrylamide gels (data not shown). AR NH 2 -terminal fragments 1-36, 1-173, and 1-333 bound to the AR LBD in the presence of 0.2 M DHT and was reduced to near background levels by the L26A/F27A mutation (Fig. 5A, lanes 1-7), providing in vitro evidence that 23 FQNLF 27 mediates the N/C interaction. But surprisingly, using the extended NH 2 -terminal residues 1-660, the L26A/F27A mutation had only a minimal effect on the in vitro binding assay (Fig. 5A, lanes 8 and 9). Furthermore, AR NH 2 -terminal fragments 174 -566 and 334 -566 that lacked the 23 FQNLF 27 binding motif showed strong interaction with the AR LBD (Fig. 5A, lanes 10 and 11). The results support the role of the NH 2 -terminal FXXLF in mediating the N/C interaction and provide additional evidence for a second NH 2 -terminal interacting site located between residues 334 and 566.
To determine whether the second site was identical to 433 WHTLF 437 identified in the androgen kinetic studies described above, AR NH 2 -terminal deletions ⌬339 -381, ⌬382-429, and ⌬430 -499 were introduced into GST-AR-(334 -556). As shown in Fig. 5B (lane 5), residues 430 -499 were required for interaction of GSTAR-(344 -566) with the AR LBD. Indeed, deletion of the predicted short ␣-helical region containing 433 WHTLF 437 in ⌬429 -439 greatly decreased the LBD-GSTAR-(344 -566) interaction in the presence of DHT (Fig. 5B,  lane 6). More importantly, introducing the double mutations of the two putative interacting regions FXXLF and WXXLF (L26A/F27A and ⌬429 -439) into AR-(1-660) reduced the N/C interaction to basal levels that were observed in the absence of androgen (Fig. 5B, lane 11). The results support the androgen kinetic studies that both 23 FQNLF 27 and 433 WHTLF 437 in the AR NH 2 -terminal region mediate the N/C interaction. In addition, they indicate that WXXLF interacts with the LBD to a greater extent than was evident in the two-hybrid assay.
The androgen dependence of the interaction of the two NH 2terminal sites with the AR LBD was further investigated in the GST fusion protein in vitro binding assay. Interaction of NH 2terminal fragment GSTAR-(1-333) containing the FXXLF binding motif 23 FQNLF 27 was dependent on the addition of androgen (Fig. 6, lanes 2 and 5) which agreed with the androgen dependence of the N/C interaction. However, surprisingly, binding of GSTAR-(334 -566) containing the 433 WXXLF 437 motif to the 35 S-labeled AR-(624 -919) LBD fragment was independent of the presence of androgen (Fig. 6, lanes 3 and 6). The results raised the possibility that the FXXLF and WXXLF binding motifs interact with different regions of the AR LBD.
Previously, we showed that certain mutations in the AF2 region of the AR LBD disrupt the N/C interaction without changing the apparent equilibrium binding affinity yet require higher DHT concentrations to induce MMTV-luciferase activity relative to wild-type AR (26). We therefore determined whether one or both of the NH 2 -terminal sites were affected by these mutations in the LBD. In GST interaction assays performed in the presence of 0.2 M DHT, LBD mutations E897K or V716R introduced into 35 S-labeled AR-(624 -919) greatly reduced the interaction between AR-(1-333) and the LBD (Fig. 7, lanes 13  and 18) as well as the interaction with the p160 coactivator fragment, TIF2-M (Fig. 7, lanes 15 and 20). However, interaction of these LBD mutants with AR-(174 -566) which contained the second interacting motif 433 WXXLF 437 was similar to wildtype (Fig. 7, lanes 14 and 19). Similar results were observed using V889M (data not shown), a mutation in the AR LBD that causes nearly complete androgen insensitivity (31,37). On the other hand, 35 S-labeled AR-(624 -919) containing K720A, a mutation that does not affect the N/C interaction or AR transcriptional activity (26), did not decrease the interaction with AR-(1-333) or AR-(174 -566) but eliminated the interaction with TIF2-M (Fig. 7, lanes 7-10). The results suggest that glutamic acid 897 and valine 716 in AF2 and valine 889 preceding helix 12 of the LBD interact with 23 FQNLF 27 but not with 433 WHTLF 437 . Because these residues are directly part (glutamic acid 897 and valine 716) or flanking (valine 889) the AF2 region, the results support the interaction of 23 FQNLF 27 with AF2 in the LBD and suggest that 433 WHTLF 437 interacts with another region of the LBD.
Transcriptional Activation-The influence of mutations in the FXXLF and WXXLF core sequences on AR transcriptional activity by full-length AR was investigated in transient transfection assays using two different luciferase reporter vectors. The 23 FQNAA 27 mutation (L26A/F27A) either alone or combined with the mutation 433 WHTAA 437 (L436A/F437A) reduced the increase in transcriptional activity by at least 50%, whereas ⌬429 -439 alone had less of an effect (Fig. 8A). By using an AR-specific probasin-luciferase reporter (38), mutations in the first or both NH 2 -terminal interacting sites also decreased transcriptional activity by about 50% (Fig. 8B). The results suggest that the N/C interaction facilitates an optimal transcriptional response. DISCUSSION Here we present evidence that the N/C interaction, referred to in the literature also as an interaction between AF1 and the LBD or AF1 and AF2 and initially described for AR (27,39) and the estrogen receptor (40), is direct, androgen-dependent, and mediated predominantly by the core sequence FXXLF in the AR NH 2 -terminal region. Both mammalian two-hybrid interaction and GST fusion protein binding studies indicate that NH 2 -terminal sequence 23 FQNLF 27 mediates interaction with the AF2 region in the LBD in an androgen-dependent manner. This is substantiated by the observation that single amino acid mutations in either the NH 2 -terminal FXXLF motif or the LBD AF2 region eliminate the N/C interaction when the apparent androgen binding affinity is not altered. At least one functional consequence of the N/C interaction previously reported (28,31) and shown here is a decreased dissociation rate of bound androgen. Mutations in 23 FQNLF 27 support this role of the N/C interaction since they increased the androgen dissociation rate.
A surprising finding of the study was that the FXXLF motif was not the only LXXLL-like motif contributing to the N/C interaction. Even though mutation of 23 FQNLF 27 essentially eliminated the N/C interaction in the two-hybrid assay and the interaction between the AR-(1-333) NH 2 -terminal fragment with the AR LBD in GST protein binding studies, mutations in 23 FQNLF 27 had only a marginal effect when the full NH 2terminal region AR-(1-660) was tested in the GST interaction assay. Further studies located the second interaction site between amino acid residues 429 and 439 in the NH 2 -terminal domain. [ 3 H]R1881 dissociation studies implied that 433 WHTLF 437 within this region plays an important role in this interaction. However, although the interaction of 23 FQNLF 27 with the AF2 region was androgen-dependent and largely responsible for the N/C interaction in the two-hybrid assay, interaction with the AR LBD second site 433 WXXLF 437 located within the 429 -439 region was androgen-independent. Furthermore, using the same AF2 LBD mutations that eliminated interaction with FXXLF, it became apparent that the WXXLF motif likely interacts with a region of the LBD outside of AF2. To account for these data and the effect of mutations in both the FXXLF and WXXLF interaction sites on increasing the androgen dissociation rate, it may be that in full-length AR, the second WXXLF site becomes available to bind the AR LBD subsequent to the androgen-dependent interaction of FXXLF with AF2. GST fusion protein interaction assays using AR NH 2 -terminal fragments may circumvent this requirement for androgen binding by releasing structural constraints within the NH 2 -terminal region. It suggests further that the binding region of the AR LBD that interacts with the WXXLF motif is not significantly altered by androgen binding as is the AF2 region in binding the FXXLF motif.
The AR NH 2 -terminal motifs FXXLF and WXXLF resemble the consensus core sequence LXXLL that mediates p160 coactivator interaction with AF2 in the LBD of nuclear receptors  (26). Specificity for the 23 FXXLF 27 core sequence for interaction with AF2 was evident because substitution with either LXXLL or WXXLF essentially abolished the N/C interaction in the two-hybrid assay. The relatively weak AF2 interaction of the LXXLL motif when substituted for 23 FXXLF 27 is consistent with a weak interaction of the LXXLL-containing p160 coactivators with the AR LBD and with the low inherent AF2 activity of the AR LBD. The results reported here as well as previous evidence from GST fusion protein assays (26) support a direct interaction between the NH 2 -and carboxyl-terminal regions of AR and contrast with previous reports that the interaction is indirect and bridged (41,42) or enhanced (43) by p160 coactivators such as steroid receptor coactivator 1.
Several models for the N/C and coactivator interactions of steroid receptors have been proposed that involve parallel versus antiparallel orientation of monomers and intra-versus intermolecular interactions (12,27,28,40,44). Binding of the AR NH 2 -terminal region to AF2 in the AR monomer or dimer may exclude binding of p160 coactivators as suggested from competition for different coactivator binding (45). The N/C interaction may also influence a proposed temporal sequence of coactivator binding (46) or create a new interacting surface to recruit coactivators (47). p160 coactivators interact with the AR NH 2 -terminal region (26) as reported for other nuclear receptors (33,48). The greater apparent binding affinity of the AR N/C interaction with AF2 compared with p160 coactivator interaction with AF2 is supported by the slower ligand dissociation rate resulting from the N/C interaction but not from p160 coactivator binding. In striking contrast, p160 coactivator binding to the AF2 region slows the ligand dissociation rate from the estrogen receptor (49).
Ligand-dependent N/C interactions have been described for other nuclear receptors including the progesterone receptor, where it facilitates receptor dimerization (44,50). In peroxisome proliferator-activated receptor-␥, N/C interactions were ligand-independent and reduced ligand binding affinity presumably by modifying conformation of the unliganded receptor (51). The functional significance of the AR N/C interaction is supported by several naturally occurring spontaneous mutations that interfere with the N/C interaction and increase the androgen dissociation rate. These include mutations at valine 889 (28) and isoleucine 898 (26) that cause complete or nearly complete androgen insensitivity and at methionine 886 that causes oligospermic infertility (52). Similarly, the N/C interaction is inhibited by site-directed mutagenesis of LBD residues glutamic acid 893 (53), valine 716, and glutamic acid 897 (26). Whereas some of these mutations also interfere with p160 coactivator binding such as TIF2 to the AF2 region (52,53), it became clear that a distinct yet overlapping binding site in AR AF2 acts preferentially as the binding site for the AR NH 2terminal region rather than for p160 coactivators (26). Lysine 720 in human AR corresponds to lysine 366 in the estrogen receptor that was required for AF2 activity (54). Mutations at AR lysine 720, while greatly reducing AR interaction of TIF2, did not decrease AR transcriptional activity significantly (26,33). On the other hand, I898T, which causes complete androgen insensitivity, did not reduce interaction with TIF2 but greatly reduced the N/C interaction. Sequence differences in the AF2 region of AR LBD compared with other nuclear receptors likely contribute to the selective binding of FXXLF versus LXXLL motifs (55). The AR AF2 sequence differs from other steroid receptors at several residues that are otherwise highly conserved. Mutating some of these residues to amino acids of other steroid receptors did not FIG. 7. LBD amino acids involved in interactions with AR NH 2 -terminal fragments. 35 S-Labeled AR-(624 -919) with wild-type sequence (WT) or with mutations K720A, E897K, or V716R were incubated in the presence of 0.2 M DHT with GST0 lacking AR sequence, or with AR NH 2 -terminal fragments GST-AR-(1-333) or GST-AR-(334 -566) as indicated, or with GST-TIF2-M (TIF2 residues 624 -1141) as a positive control (26). Input lanes (I) represent 20% of the total 35 S-labeled AR-(624 -919) used per reaction. The major radiolabeled bands represent wild-type or mutant 35 S-labeled AR-(624 -919).
FIG. 8. Transcriptional activation of wild-type AR and NH 2 -terminal domain mutants using luciferase reporter vectors. Transcriptional activity of full-length wild-type (WT) AR or with the indicated AR amino acid or deletion mutations were determined in CV1 cells at increasing DHT concentrations as described under "Experimental Procedures" using MMTV-luciferase (A) or probasinluciferase reporter vectors (B). pCMVhAR DNA (25 ng) was transfected with 5 g of reporter vector. Shown above the bars is the fold induction determined relative to the activity in the absence of DHT. The data are representative of at least three independent experiments. enhance p160 coactivator binding (26). The sequence differences unique to the AR LBD may redirect the function of AF2 toward higher binding affinity for the FXXLF binding motif in the AR NH 2 -terminal domain rather than for the LXXLL motifs in the p160 coactivators. Sequences flanking the LXXLL motifs of p160 coactivators (56,57) apparently contribute to selective binding to AF2 of other nuclear receptors (45). The use of combinatorial phage display screening showed that in some cases sequences flanking the LXXLL motifs contribute to specific interactions with nuclear receptors (56). However, our study of 23 FQNLF 27 and 433 WHTLF 437 indicates that of the flanking AR NH 2 -terminal sequences tested, none had a predominant role in the N/C interaction. Mutation of glutamine 28 carboxyl-terminal to 23 FQNLF 27 or serine 432 NH 2 -terminal to 433 WHTLF 437 did not increase the androgen dissociation rate, an indicator of the N/C interaction.
Evolution of amino acid sequence in the AR NH 2 -terminal region and LBD seems to favor conservation of the N/C interaction sites. In the NH 2 -terminal region, both 23 FQNLF 27 and 433 WHTLF 437 and their flanking sequence are fully conserved among the primate AR even though other regions of the NH 2terminal domain are not conserved (58). In the rainbow trout AR ␣ and ␤ forms, which have the strikingly low homology of 19% in the NH 2 -terminal amino acid sequence compared with human AR, structural conservation occurs at the androgen-dependent interaction site with the trout sequence 22 FQNVF 26 compared with 23 FQNLF 27 in human AR (59). Furthermore, whereas the primate LBD is fully conserved with respect to human amino acid sequence, rainbow trout AR ␣ and ␤ forms have only 65 and 68% sequence similarity in the LBD, respectively (59), yet residues critical for the N/C interaction are conserved. Natural selection therefore seems to favor residues in AR domains that mediate the N/C interaction supporting their critical role in AR function.
The functional significance of the N/C interaction remains to be fully established. The N/C interaction appears to be required for AR function in vivo as suggested by androgen insensitivity mutations. It is, however, apparently not an absolute requirement for AR induction of MMTV reporter gene activity in transient transfection assays. The AR NH 2 -terminal and DNA binding domain fragment that lacks the LBD has strong constitutive transcriptional activity with the MMTV-luciferase reporter gene (32,60), and only a 50% reduction in activity was observed with mutations in full-length AR that disrupt the N/C interaction using either the MMTV-or probasin-luciferase reporter genes. The N/C interaction may nevertheless contribute to androgen-specific gene induction since the AR shares the ability with other steroid receptors to bind simple consensus DNA response elements but has distinctly different physiological effects (61). Furthermore, ligands such as medroxyprogesterone acetate (Provera) that fail to promote the N/C interaction are weak androgens in vivo but strong androgen agonists in transient transfection assays (30). These observations support that the N/C interaction is important in mediating AR function in vivo and that transient transcriptional assays may not provide a good reflection of this in vivo requirement. The N/C interaction may increase the sensitivity of AR to low circulating androgen concentrations by slowing the dissociation rate of bound androgen. This function might be especially important during male sexual development in utero.