Dual function of an amino-terminal amphipatic helix in androgen receptor-mediated transactivation through specific and nonspecific response elements.

Steroid receptors are transcription factors that, upon binding to their response elements, regulate the expression of several target genes via direct protein interactions with transcriptional coactivators. For the androgen receptor, additional interactions between the amino- and carboxyl-terminal regions have been reported. The first amino acids of the amino-terminal domain are necessary for this amino/carboxyl-terminal interaction. Deletion of a FQNLF core sequence in this region blunts the interaction, as does a G21E mutation. We investigated the effect of the aforementioned mutations in the context of the full size androgen receptor on a series of selective and nonselective androgen response elements. Strikingly, the FQNLF deletion strongly reduced the androgen receptor capacity to transactivate through nonselective motifs but did not affect its activity on selective elements. Although the G21E mutation strongly impairs the amino/carboxyl-terminal interaction, it does not significantly influence androgen receptor activity on either selective or nonselective elements. Surprisingly, this mutation leads to an increased binding of the amino-terminal domain to the glutamine-rich region of the steroid receptor coactivator-1 of the p160 family. Taken together, these data suggest that the amino-terminal amino acids of the androgen receptor play a key role in determining its transcriptional activity by modulating the interaction with the ligand-binding domain as well as interaction with p160 coactivators.

Steroid receptors are transcription factors that, upon binding to their response elements, regulate the expression of several target genes via direct protein interactions with transcriptional coactivators. For the androgen receptor, additional interactions between the amino-and carboxyl-terminal regions have been reported. The first amino acids of the amino-terminal domain are necessary for this amino/carboxyl-terminal interaction. Deletion of a FQNLF core sequence in this region blunts the interaction, as does a G21E mutation. We investigated the effect of the aforementioned mutations in the context of the full size androgen receptor on a series of selective and nonselective androgen response elements. Strikingly, the FQNLF deletion strongly reduced the androgen receptor capacity to transactivate through nonselective motifs but did not affect its activity on selective elements. Although the G21E mutation strongly impairs the amino/carboxyl-terminal interaction, it does not significantly influence androgen receptor activity on either selective or nonselective elements. Surprisingly, this mutation leads to an increased binding of the amino-terminal domain to the glutamine-rich region of the steroid receptor coactivator-1 of the p160 family. Taken together, these data suggest that the amino-terminal amino acids of the androgen receptor play a key role in determining its transcriptional activity by modulating the interaction with the ligand-binding domain as well as interaction with p160 coactivators.
Androgens are involved in the differentiation of the male reproductive organs in the embryo and the development and maintenance of secondary sexual characteristics. The biological actions of androgens are mediated by the androgen receptor (AR). 1 The AR is a ligand-activated transcription factor and belongs to the class I subgroup of the nuclear receptor superfamily. Two structural domains are well conserved among the nuclear receptors (NR): the DNA-binding domain (DBD), consisting of two zinc finger modules, and the ligand-binding domain (LBD), containing a transcription activation function (AF-2) (1,2). The amino-terminal domain (NTD) and the hinge region are more divergent. Because of their highly conserved DBD, it is not surprising that the class I steroid receptors (AR, GR, progesterone receptor, and mineralocorticoid receptor) recognize the same palindromic (inverted) repeats of the 5Ј-TGT-TCT-3Ј core sequence, spaced by three nucleotides. The mechanisms that lead to the in vivo steroid specificity of gene regulation are still not completely understood (3)(4)(5)(6). Selective DNA binding by the AR could be one possible mechanism for hormone-specific gene regulation. It has been reported that the AR-DBD does not only recognize response elements organized as inverted repeats of 5Ј-TGTTCT-3Ј sequences with a threenucleotide spacer but can also bind to direct repeats of 5Ј-TGTTCT-3Ј-like sequences, which is proposed to contribute to the AR specificity of transcriptional responses (7)(8)(9)(10)(11)(12)(13)(14).
The AR depends on a strong amino/carboxyl-terminal (N/C) interaction for its full activity (18, 40 -43). These N/C interactions have also been demonstrated in the progesterone receptor (44) and the estrogen receptor ␣ (45,46), but in these receptors the implication in receptor activity is less pronounced. The AR N/C interaction involves LKDIL and FQNLF core sequences, both present in predicted ␣-helical structures within the NTD (18,41). The FQNLF motif is located within the thirty aminoterminal amino acids of the AR and has the ability to bind AF-2 of the carboxyl-terminal domain in an androgen-dependent manner. The LKDIL motif is located between amino acids 178 and 183 and is important for the transcriptional activation function AF-1. In this paper, we describe the involvement of the FQNLF motif and a flanking conserved Gly in the N/C interaction. In addition, we have analyzed the functional role of this interaction on isolated direct and indirect repeat AREs and on AR-specific versus nonspecific enhancers. Our data suggest that the interaction between the AR-NTD and -LBD may be a prerequisite for its transcriptional activity on nonspecific HREs, whereas it is dispensable on specific HREs.

MATERIALS AND METHODS
Plasmid Constructs-The expression vectors pSG5AR 538 -919 (encoding the hAR-DBD-LBD), pSG5AR (expressing FLAG-tagged fulllength hAR), and pSG5SRC-1e and the vector Qr SRC-1 (expressing the Qr domain of SRC-1 fused to the Gal4-DBD) are described elsewhere (14,18,39). The Gal4-DBD and the VP16 fusion constructs hAR 1-529 , hAR 1-529 ⌬FQNLF, and hAR 1-529 L435P were made by a PCRbased method. The construct hAR 1-529 was made by using the following primers: 5Ј-GGCAGATCTCCATGGAAGTGCAGTTAGGGCTGGG-3Ј and 5Ј-GGATCCTCAACGCATGTCCCCGTAAGGTCCGGA-3Ј. As template, the expression vector for the full-length human AR (pSV-AR), a kind gift of Dr. A. O. Brinkmann (Erasmus University of Nijmegen, Nijmegen, the Netherlands) was used. The ⌬FQNLF deletion was generated with the oligonucleotides 5Ј-CCTACCGAGGAGCTCAGAGC-GTCCGCGAAGTG-3Ј and 5Ј-CGCGGACGCTCTGAGCTCCTCGGTAG-GTCT-3Ј. To generate the L435P mutation, the following oligonucleotides were used: 5Ј-CCTGGCACACTCCCTTCACAGCCGAAG-3Ј and 5Ј-CTGGCCTTCTTCGGCTGTGAAGGGAGTGTGCCAGGATGAG-3Ј, in which the mutated base is underlined. These PCR-generated hAR-NTD fragments were inserted in frame with the Gal4-DBD in the BamHI restriction site of pABGal4 and in frame with the VP16 activating domain in the BglII site of pSNATCH-II (18). Expression vectors for the hAR carrying the FQNLF deletion or the L435P point mutation were made by insertion of a Eco52I/Eco81I fragment of the PCR products in the pSG5AR construct, which was cut with the same restriction enzymes. The pSG5 construct expressing the G21E-mutated hAR was generated by coincidence and was detected by sequencing. Restriction and modifying enzymes were obtained from MBI Fermentas GmbH. The luciferase reporter constructs containing the isolated elements TAT-GRE, slp-HRE3, C3 (1)ARE, slp-HRE2, sc-ARE1.2, slp-HRE2 mut-4T-A;ϩ2A-T, and sc-ARE1.2 mut-4T-A;Ϫ2A-T are driven by the thymidine kinase minimal promotor and have been described elsewhere (Ref. 10 and references therein). The reporter construct driven by the E1b promotor and containing two copies of the rat tyrosine aminotransferase-GRE was described previously (47). The thymidine kinase minimal promotor-driven reporter construct containing the slp and sc upstream enhancers and the C3 (1) intronic fragment as well as the pb promotor-driven construct have also been described (10). The reporter construct (Gal4) 5 TATA-Luc, a kind gift from Dr. M. G. Parker (Imperial Cancer Research Fund, London, UK), was used for measuring the AF-1 AR activity. The pCMV-␤-galactosidase vector was obtained from Stratagene.
Transfections-All of the transfections were performed in COS-7 African green monkey kidney cells, obtained from the American Type Culture Collection (Manassas, VA). One day prior to transfection, the cells were seeded in 96-well culture plates in Dulbecco's modified Eagle's medium (Invitrogen) containing 5% dextran-coated charcoalstripped fetal bovine serum at a density of 10 4 cells/well. They were transfected using FuGENE 6 transfection reagent (Roche Molecular Biochemicals) as described by the manufacturer. The amount of luciferase reporter construct was fixed on 100 ng/well, and the amount of pCMV-␤-galactosidase was fixed at 5 ng/well. Where applicable, empty pGEM-T, empty pSG5, empty pSNATCH-II, or pABGal4 vector was added to keep the total amount of transfected DNA constant. After transfection, the cells were incubated for 24 h with medium containing 5% dextran-coated charcoal-stripped fetal bovine serum and supplemented or not with hormones (10 Ϫ8 M). The synthetic androgen R1881 (methyltrienolone) was purchased from PerkinElmer Life Sciences. Finally, the cells were washed twice with phosphate-buffered saline (Invitrogen) and lysed in 25 l of passive lysis buffer (Promega). The luciferase and ␤-galactosidase activities were measured on 2.5 l of the extracts using the assay systems from Promega and Tropix (Westburg, the Netherlands), respectively. The luciferase activity was corrected for transfection efficiency by normalizing it according to its ␤-galactosidase activity. The values shown are the averages of at least three independent experiments performed in triplicate. The error bars are the S.E. values.
Preparation of COS-7 Whole Cell Extracts Containing Full Size AR-COS-7 cells were plated in 10-cm Petri dishes and were transiently transfected as described above with 1 g of expression vectors. At 24 h after transfection, the cells were stimulated for 1 h with R1881 (10 Ϫ8 M). The medium was removed, and the cells were washed twice with icecold phosphate-buffered saline. The cells were collected in 1.5 ml of ice-cold phosphate-buffered saline/dish and pelleted by centrifugation (1 min). The phosphate-buffered saline was removed, and the cells were resuspended in 100 l of ice-cold extraction buffer containing 20 mM Hepes-KOH, pH 7.8, 450 mM NaCl, 0.4 mM EDTA, and 25% glycerol. Immediately before use, dithiothreitol and phenylmethylsulfonyl fluoride are added to a final concentration of 0.5 mM. The samples were freeze-thawed three times. The pellet was collected, and the supernatant was removed and stored at Ϫ80°C. The protein concentrations of the extracts were determined with the Coomassie protein assay kit (Pierce) and were in each case ϳ4 g/l.
Gel Shift Assays and Western Blots-Synthetic complementary oligonucleotides were hybridized, radioactively labeled, and used in band shift assays as described previously (14). In brief, 5 l of total cell extract was preincubated with 1 l of poly(dI-dC) (1 g/l), 10 l of D100 (20 mM Hepes, 5 mM MgCl 2, 0.1 mM EDTA, 17% glycerol, 100 mM NaCl), 1 l dithiothreitol (20 mM), 1 l Triton X-100 (1%), and 1 l of water. Subsequently, the probe is added and incubated for 20 min on ice. Bound probe was separated from the free by an electrophoresis for 2 h at 120 V in a 5% polyacrylamide gel. To obtain supershifts, a rabbit antiserum against human AR was added prior to the probe (48). For Western blotting, equal amounts of protein were separated by SDS-PAGE on an 8% gel and blotted onto polyvinylidene difluoride membranes (Amersham Biosciences). The membranes were probed with a monoclonal M2 anti-FLAG antibody (Stratagene), and immunoreactive proteins were visualized with the chemiluminescence reagent plus (PerkinElmer Life Sciences). For the detection of VP16 fusion proteins, extracts were made in passive lysis buffer. Western blot analysis was performed as described above.

Hormone-dependent Functional Interaction between the LBD and the Amino-terminal Domain-Previous studies have
shown that an FQNLF motif at position 23 in the aminoterminal domain of the human AR plays a key role in the androgen-dependent interaction between the NTD and the LBD. Another motif ( 432 WHTLF 436 ) was also postulated as a possible interaction site (40). We have studied the N/C interaction in a mammalian two-hybrid system using a luciferase reporter construct driven by the E1b promotor and containing a tandem repeat of the rat tyrosine aminotransferase-GRE (2ϫTAT-GRE(E1b)-Luc). In these experiments we tested and compared three AR-NTD (AR 1-529 ) mutants (Fig. 1A). The first construct consists of a deletion of the first motif (AR 1-529⌬FQNLF). The second construct has a point mutation immediately amino-terminally of the FQNLF motif; Gly 21 is mutated to a Glu (AR 1-529 G21E). This Gly is outside the FQNLF motif but inside a predicted ␣-helical region and is highly conserved among different species (Fig. 1B). The third construct has a point mutation in the WHTLF motif so that Leu 435 is mutated into a Pro (AR 1-529 L435P). Wild type and mutant forms of the AR-NTD domain were fused to the VP16 transactivation domain and were coexpressed in the mammalian twohybrid assay with pSG5AR 538 -919 , which encodes a DBD-and LBD-containing fragment of the AR (Fig. 1C). Expression levels of the VP16 fusion constructs were similar (Fig. 1D). As expected, the AR 1-529 fragment interacts hormone-dependently with AR 538 -919 . The deletion of the FQNLF motif in AR 1-529 completely abolished the R1881-dependent interaction. These findings are in agreement with previous reports (40,41). Surprisingly, the G21E mutation in the NTD, outside the FQNLF motif, also markedly reduces N/C interaction. In contrast, no difference in the LBD interaction is observed for the AR 1-529 L435P mutation. Therefore, we suggest that the first amino-terminal amino acids of the AR-NTD form the predominant interaction site for the AR-LBD.
Role of the FQNLF Motif in the Transactivation through Different Types of AREs-The effect of the deletion of the FQNLF motif (hAR⌬FQNLF) was analyzed in the context of the full size AR on a series of selective and nonselective AREs (Fig. 2). When using wild type AR, the induction factors are 16.5, 19.9, and 8.9 for the nonspecific AREs TAT-GRE, slp-HRE3, and C3 (1)-ARE, respectively, whereas AR⌬FQNLF showed a markedly decreased induction of these reporter constructs (6.0, 8.7, and 3.1, respectively). However, for slp-HRE2 and sc-ARE1.2, which are androgen-specific HREs, no difference is seen between the wtAR and the AR⌬FQNLF. In contrast, the FQNLF deletion does affect the transactivating abilities of the receptor when tested on the PB-ARE2, which is another known androgen-specific ARE (7,49).
Two mutations were introduced within slp-HRE2 (slp-HRE2 mut-4T-A;ϩ2A-T) and in sc-ARE1.2 (sc-ARE1.2 mut-4T-A; Ϫ2A-T), changing these elements into nonselective ones and leading to a functional loss of AR specificity (10). As shown in Fig. 3, this loss of specificity correlates with a strong decrease in transactivation by the AR⌬FQNLF in comparison with wtAR activity. Although the activity of the ⌬FQNLF mutant on the wild type slp-HRE2 and the sc-ARE1.2 does not significantly differ from that of wtAR, androgen-mediated induction factors of the mutated sc and slp response elements are two or three times lower for the ⌬FQNLF construct compared with the wild type AR.
Role of the FQNLF Motif in Transactivation through Complex Enhancers-We tested four luciferase reporter constructs in transient transfections of COS-7 cells (Fig. 4). One construct is driven by the probasin proximal promotor (pPB-Luc). The three other constructs are driven by the thymidine kinase minimal promotor and contain either the C3 (1) intronic enhancer (pC3 (1)-TATA-Luc), the slp enhancer (pSLP-TATA-Luc), or the sc enhancer (pSC-TATA-Luc). Three of them (pPB- Luc, pSLP-TATA-Luc, and pSC-TATA-Luc) are controlled by androgen-specific elements, whereas pC3 (1)-TATA-Luc is activated by all class I steroid receptors (12). In contrast to the isolated elements tested in Figs. 2 and 3, no difference in the effect of the FQNLF deletion is seen between specific and nonspecific enhancers, because for all reporter constructs, the FQNLF deletion reduces the potency of the AR ϳ2-fold. Analysis of Gly 21 Flanking the FQNLF Core-The point mutation G21E shows a decreased AR N/C interaction similar to the deletion of the FQNLF core in a mammalian two-hybrid assay (Fig. 1C). We tested whether the transcriptional activity of the full size AR is affected by this mutation when specific or nonspecific AREs are used. Surprisingly, the transcriptional activation of the TAT-GRE construct is decreased only marginally when the G21E mutation is introduced in the AR, and no clear effect is seen when the androgen-selective slp-HRE2 was tested. When the nonselective mutant of slp-HRE2 was used, the G21E mutation again had no effect on the functionality of the AR (Fig. 5).

Effects of Mutations within the NTD on DNA Binding-To
analyze the influence of the deletion of the FQNLF motif and the G21E and the L435P mutations on the in vitro interaction of the full size receptor with nonspecific and specific AREs, band shifts assays using wtAR, AR⌬FQNLF, ARG21E, and ARL435P were performed (Fig. 6A). COS-7 cells were transfected with expression vectors for wtAR, AR⌬FQNLF, ARG21E, and ARL435P (1 g). As a specific ARE, we used slp-HRE2, and as nonspecific elements, we used slp-HRE2 mut-4T-A;ϩ2A-T and the TAT-GRE. The gel shift assays showed very similar binding of wtAR and the mutant ARs to different response elements. Western blot analysis of the extracts was performed to assess equal expression levels of the different receptors (Fig. 6B).
Analysis of the AF-1 Activity of the Mutated AR-NTDs and Binding the Qr Region of SRC-1-The AR-NTD contains a strong ligand-independent activation function in the carboxylterminal region of the AR-NTD (Tau-5). To analyze whether the aforementioned mutations influence this constitutively active Tau-5 function, wild type and mutant forms of the AR-NTD domains were fused to the Gal4-DBD and transfected into COS-7 cells. As reporter construct, we used (Gal 4 )5-TATA-Luc (Fig. 7A). Deletion of FQNLF and the G21E mutation resulted in an activation comparable with that of wtAR-NTD, whereas the L435P mutation resulted in a decreased activation.
It has already been proposed that the efficient recruitment of coactivators by the native AR occurs primarily through the NTD (18,(37)(38)(39). SRC-1 is a member of the p160 coactivator family and contains a glutamine-rich region (amino acids 989 -1240) called Qr, needed for interaction with the amino-terminal domain of the AR in a ligand-independent manner (18,(37)(38)(39). We tested whether the mutations AR 1-529 ⌬FQNLF, AR 1-529 -G21E, and AR 1-529 L435P, fused to the VP16 activation domain, are still able to interact with the Qr of SRC-1, fused to the Gal4-DBD (Fig. 7B). A striking characteristic for the point mutation G21E is that, although the NTD shows a strongly reduced interaction with the AR-LBD (Fig. 1C), there is a 2.5-fold better interaction with Qr when compared with wtAR-NTD. This is not seen with AR 1-529 ⌬FQNLF or AR 1-529 L435P (Fig. 7B). To compare the interaction of Qr with AR-NTD and its mutants with the coactivation ability of SRC-1e, the constructs pSG5, pSG5wtAR, pSG5⌬FQNLF, pSG5G21E, and pSG5L435P and the 2ϫTAT-GRE(E1b) luciferase reporter construct were transiently transfected into COS-7 cells and cotransfected with pGEM-T or pSG5SRC-1e (Fig. 7C). Transcriptional activity of ARG21E and ARL435P increased to the same level compared with wtAR in the presence of SRC-1e, whereas   FIG. 3. Transcriptional activation by the wtAR and  AR⌬FQNLF on slp-HRE2, sc-ARE1.2, and their mutated  the net coactivation of AR⌬FQNLF in the presence of SRC-1e is lower. DISCUSSION For the members of the steroid receptor superfamily such as estrogen receptor ␣, GR, and progesterone receptor, a ligandinduced conformational change has been clearly demonstrated to result in a hydrophobic recruitment surface (AF-2) for transcriptional coactivators. These coactivators interact with the LBD via three highly conserved ␣-helical LXXLL signature motifs (35). The AR has been reported to be different from the other steroid hormone receptors in that the AF-2 activity in the LBD is quite low when tested in mammalian cells (2). However, the AR-LBD has the same overall three-dimensional structure as the LBDs of the other nuclear hormone receptors (50).
The AR-NTD not only harbors a strong activation function (AF-1) (16) but also displays a high affinity for the liganded LBD (18, 40 -43). LXXLL-related sequences in the AR-NTD have been investigated for their interaction with the LBD. We  2 and 3, lanes 4 and 5, lanes 6 and 7, and  lanes 8 and 9, respectively). For each probe, no protein was added in lane 1 as a negative control. The supershifts were performed in the presence of specific antibodies against AR (␣-AR, lanes 3, 5, 7, and 9). Free probe, shifted, and supershifted complexes are indicated on the right by an open arrow, an asterisk, or a black arrow, respectively. B, immunoblot analysis of the AR and AR mutants. Equal amounts of COS extracts containing AR⌬FQNLF, ARG21E, and ARL435P as used in the supershift assay and as indicated at the top were used as to perform Western blot. The extracts were resolved on 8% SDS-polyacrylamide gel and subjected to immunoblotting with monoclonal M2 anti-FLAG antibody.

FIG. 7. Tau-5 activation function of wtAR-NTD and mutated
NTDs, interaction with QrSRC-1 and coactivation of wtAR and AR mutants by SRC-1e. A, study of the constitutive activation domain Tau-5. COS-7 cells were transfected with 50 ng of the AR-NTD and the mutated NTDs, fused to the Gal4-DBD, together with the luciferase reporter construct p(Gal 4 )5-TATA-Luc. The activities are depicted relative to the activity of the wtAR-NTD construct, which was set to 100. B, Interaction of AR-NTD and the mutated NTDs with QrSRC-1. pABGal4-DBDQrSRC-1e (989 -1240) (50 ng/well) was coexpressed in COS-7 cells with 50 ng of empty pSNATCH-II, pSNATCH-IIAR 1-529 , pSNATCH-IIAR 1-529 ⌬FQNLF, pSNATCH-IIAR 1-529 G21E, or pSNATCH-IIAR 1-529 L435P. The assays were performed using the (Gal 4 )5-TATA-luciferase reporter (100 ng). The activities are depicted relative to the activity of the wtAR-NTD construct, which was set to 100. C, coactivation of wtAR and AR mutants by SRC-1e. COS-7 cells were transfected with 20 ng of empty vector and pSG5 with wtAR, AR⌬FQNLF, ARG21E, or ARL435P and cotransfected with 100 ng pGEM-T or pSG5SRC-1e. As reporter construct, 100 ng of 2ϫTAT-GRE (E1b)-Luc was used. The activities are depicted relative to the activity of the pSG5wtAR construct in the presence of androgens, which was set to 100. Luciferase activities were adjusted according to the ␤-galactosidase activity in the same sample. The results are the averages of at least three separate experiments performed in triplicate; the error bars indicate the S.E. values. already reported the involvement of one LXXLL-like motif, 179 LKDIL 183 (18), the mutation of which (I182A/L183A) impairs the N/C interaction and the activity of the AR. Two other candidate motives have been described: 23 (40,41). In this study, we performed a functional analysis of these latter motifs. Deleting the FQNLF motif destroys the ability of the AR-NTD to interact with the AR-LBD. The mutation of Gly 21 flanking the FQNLF core to Glu strongly attenuates the N/C interaction. In contrast to another report (40), LBD binding by the L435P mutated NTD is comparable with that observed for the wtAR-NTD (Fig. 1C).
To investigate the role of the N/C interaction in AR functionality in more detail, we analyzed the activity of the mutated receptors on different responsive constructs. Next to the classical nonselective class I elements, there is an additional group of AREs that are only recognized by the AR. Examples of such androgen-specific elements are the PB-ARE2 from the rat probasin promotor (7,49,51), the SC-ARE present in the first exon of the human secretory component (SC) gene (13,52), the sc-ARE1.2 in the far upstream enhancer of the human SC gene (8,10), the slp-HRE2 from the upstream enhancer of the mouse sex-limited protein (53), and the Pem ARE-1 and ARE-2 in the proximal promotor of the murine pem gene (54). Elsewhere, we hypothesized that an alternative dimerization mechanism of the DNA-binding domain would be responsible for the androgen specificity (11). We demonstrated indeed that the AR is the only steroid receptor which binds direct repeats of 5Ј-TGTTCT-3Ј-like sequences (9 -11, 13), indicating that the AR could bind the specific elements in a head-to-tail configuration and not in the conventional head-to-head configuration. In this study, we compared the transcriptional activity of the constructs AR⌬FQNLF, ARG21E, and ARL435P on specific versus nonspecific response elements and enhancers (Figs. [3][4][5]. For the mutation L435P in the WHTLF motif and the mutation I182A/ L183A in the LKDIL motif, no difference on selective versus nonselective elements was seen. In fact, the L435P mutation did not affect the transcriptional activation by any of the tested motifs (data not shown).
We observed a decrease in the transcription activating capacity of AR⌬FQNLF compared with wtAR on nonspecific HREs, whereas no changes were observed on the specific slp-HRE2 and sc-ARE1.2 elements. However, when these AREs are mutated in their left and/or right half-sites (slp-HRE2 mut-4T-A;ϩ2A-T and sc-ARE1.2 mut-4T-A;Ϫ2A-T) to become partial palindromic repeats, AR transactivation was significantly affected by the FQNLF deletion. A possible explanation for why no decrease in AR transcriptional activity is observed on the two specific AREs is that AR dimers, in which the DBD is dimerized in a head-to-tail conformation, might be less dependent on the N/C interaction for transactivation. Whether other factors or cofactors play a role in these differences remains to be elucidated.
Nonspecific HREs seem to be more dependent on the N/C interaction compared with the direct repeat AREs (Figs. 2 and  3). One exception is the AR-specific PB-ARE2, although this might be explained by the fact that PB-ARE2 (5Ј-GGTTCT-nnnAGTACT-3Ј) can be considered as a direct as well as a palindromic repeat. It is clear from Fig. 6A that there are no obvious differences between the wild type and the mutant receptors in their binding characteristics for the different DNA elements.
The G21E mutation strongly reduces N/C interaction (Fig.  1C). It is therefore clear that not only the core FQNLF motif but also some flanking residues contribute. This correlates with the stronger conservation of this region among the AR of different species (Fig. 1B) (41,55). Surprisingly, the differences seen on AR transcriptional activity at specific versus nonspecific elements by the deletion of FQNLF were not observed with this point mutation. This might be explained by the residual AR-NTD G21E /AR-LBD interaction, observed in Fig. 1C, which could be sufficient for AR activity on the nonspecific HREs.
The differences we observed for AR⌬FQNLF transcriptional activity at specific versus nonspecific elements are limited to isolated response elements because no effect is seen on specific and nonspecific enhancers. This could be due to the fact that hormone response elements form part of more complex enhancers, whose activity is also tightly controlled by additional transcription factors (12). The slp enhancer, for example, contains three HREs: HRE-1, HRE-2, and HRE-3, and other transcription factors binding to the slp-ARU were identified as NFB, octamer transcription factor, and AML3/CBF␣1 (56).
A second interesting feature for the G21E mutation is an almost 3-fold increase in the affinity for the Qr of SRC-1 as measured in double-hybrid assays (Fig. 7B). This increase in Qr binding is not seen for the AR-NTD⌬FQNLF nor for the L435P mutated AR-NTD. It should be noted that deletion of the FQNLF core and the G21E mutation does not affect the intrinsic activity of the NTD, in contrast to L435P mutation, which reduces the activity 2-fold (Fig. 1C). Possibly, the lowered affinity of the NTDG21E for the LBD and its effect on transactivation by the AR is compensated by the increased affinity of the NTD for the Qr of SRC-1. SRC-1e coexpression seems to rescue the ARG21E activity to the same level as wtAR and ARL435P (Fig. 7C).
Our data do not exclude the possibility of a stabilization by the NTD of the conformation of the LBD, which would lead to an enhanced recruitment of coactivators as postulated earlier (57). However, the effects of the G21E mutation on Qr recruitment by the isolated NTD (Fig. 7B) leads us to propose a more direct role in the p160 recruitment.
In summary, our findings demonstrate a dual role of the first amino acids of the amino-terminal domain of the AR, which are essential for the N/C interaction and which seem to have an additional function in interaction with Qr of SRC-1. These results contribute to the unraveling of the mechanisms involved in the AR-N/C interactions and the mechanisms of SRC-1 recruitment. Recently, it has been described that the relative contribution of the two interaction sites of SRC-1 with the AR, the LXXLL motifs and the Qr region, depend on the nature of the enhancers (39). In addition, we provide the first indication that transactivation by the AR through some selective AREs (slp-HRE2 and sc-ARE1.2) is structurally different from transactivation through other classical AREs.