Interaction between the Amino- and Carboxyl-terminal Regions of the Rat Androgen Receptor Modulates Transcriptional Activity and Is Influenced by Nuclear Receptor Coactivators*

Identical N-terminal deletions in the wild-type rat androgen receptor (rAR) and a constitutively active rAR (ARΔ641–902) devoid of the ligand-binding domain (LBD) resulted in dissimilar consequences in transcriptional activation: deletion of residues 149–295 abolished wild-type AR activity, but did not influence that of ARΔ641–902. The activity of the N-terminal transactivation domain is thus controlled by the hormone-occupied LBD, suggesting that the N- and C-terminal regions of rAR communicate. Consistent with this idea, a strong androgen-dependent interaction between the N-terminal region and LBD was demonstrated in a mammalian two-hybrid system using GAL4 and VP16 fusion proteins. This interaction can be direct or indirect. Several nuclear receptor coactivators (CBP, F-SRC-1, SRC-1, and RIP140) that interact with other steroid receptors were tested as potential mediators of the N- and C-terminal interaction of rAR using the mammalian two-hybrid system. CBP or F-SRC-1 not only enhanced AR-mediated transactivation, but also facilitated the androgen-dependent interaction between the N- and C-terminal domains, implying that part of the coactivator-dependent transcriptional activation occurs via this mechanism. In contrast, SRC-1, a coactivator for the progesterone receptor, inhibited both AR-mediated transactivation and interaction between the N and C termini. Recruitment of coregulators may involve AR domains other than the LBD, as F-SRC-1 and CBP enhanced, but SRC-1 repressed, the transcriptional activity of ARΔ641–902. Collectively, interplay between the N-terminal region and LBD of rAR results in the formation of a transactivation complex that includes coregulators and that is mandatory for optimal activation of androgen-induced promoters.

The androgen receptor (AR) 1 is a ligand-dependent transac-tivating protein that regulates the differentiation, development, and maintenance of male reproductive functions, along with sexually dimorphic characteristics of nonreproductive tissues (1). AR is a member of the nuclear receptor superfamily that comprises receptors for steroids, thyroid hormones, and retinoids (2,3). The DNA-binding domain (DBD) of AR exhibits a high degree of amino acid sequence identity to other members of the glucocorticoid receptor (GR) subfamily, the progesterone (PR) and mineralocorticoid (MR) receptors, and as a consequence, the four receptors appear to recognize very similar, if not identical, hormone response elements (1)(2)(3). However, the cognate steroids of these receptors elicit biological responses that are hormone-specific, implying that important determinants for steroid-and receptor-specific actions reside outside the DBD.
In the case of other subfamilies of nuclear receptors, different mechanisms have been described to account for cell-and promoter-specific regulation of gene expression, including receptor heterodimerization, variable spacing between hormone response elements, and orientation of hormone response element half-sites (4 -6). Since these mechanisms may not be used by PR, GR, MR, and AR, it is assumed that their highly variable N-terminal regions are responsible for the cell-and steroid-specific regulation of target genes. In support of this notion, induction of mouse sex-limited protein gene (slp) expression has been shown to be mediated by the N-terminal region of murine AR, but not by that of GR (7).
The N-terminal region of GR has been shown to function as a specificity determination domain on the proliferin promoter composite response element (8). Dissimilarities in the transcriptional activities of two human PR forms, PR-A and PR-B, suggest that isoform specificity is associated with structural differences in their N termini (9 -11). Similar to PR, two AR isoforms differing in the length of the N-terminal regions appear to exist in human cells, although functional similarities and/or differences between the two forms have not been elucidated (12).
Several putative coregulators (coactivators and corepressors) that interact in a hormone-dependent fashion with the C-terminal regions of several nuclear receptors have been described, including RIP140 (13), TIF1 (14), ARA 70 (15), SUG1 (16), SMRT (17), N-CoR (18), and TRUP (19) (reviewed in Ref. 20). SRC-1 (21), F-SRC-1 (22), GRIP1 (23), ERAP/p160 (24,25), and TIF2 (26) appear to be members of the same family of coactivator proteins. CBP/p300, a transcriptional coactivator interacting with a variety of transcription factors (27,28), has also been shown to function as a coactivator for nuclear receptors (24,25,29). Only ARA 70 and TIF2 have been reported to enhance AR-mediated transactivation (15,26). Most of the coregulators are relatively promiscuous in that nuclear receptors are capable of associating with many of them, and probably simultaneously. No coactivator or corepressor interacting specif-ically with the N-terminal region of nuclear receptors has been reported. Moreover, it has not been clarified whether some of the factors interact with receptor regions other than the LBD.
Studies using relatively large deletions and truncations revealed that the N-terminal halves of rat and human ARs encompass sequences mandatory for transcriptional activation and repression (30 -33). In a detailed study of hAR, it was shown that this receptor may utilize two different N-terminal transcriptional activation units (34). Although the amino acid sequences of human and rat ARs are highly homologous, the location and length of various poly(amino acid) stretches in the N-terminal region differ considerably (1,35). In this work, we have delineated sequences important for transactivation by rAR and demonstrate that the ability of this receptor to utilize distinct N-terminal regions for promoter activation is modulated by the hormone-occupied LBD. Our results also indicate that coactivators F-SRC-1 and CBP not only enhance AR-mediated transactivation, but also facilitate androgen-dependent association of the N and C termini of rAR. This is not a feature typical of all nuclear receptor coactivators, as SRC-1, which activates PR, GR, and the estrogen receptor (ER), inhibits the functions of rAR.

EXPERIMENTAL PROCEDURES
Materials-pARE 2 tk-CAT contains two copies of the rat tyrosine aminotransferase gene ARE inserted upstream of the thymidine kinase promoter driving the bacterial chloramphenicol acetyltransferase gene, of which the cryptic AP-1 site was deleted from the vector backbone (36). pPREtk-CAT (pAREtk-CAT in this report) is a reporter that contains only one copy of AREs (from Dr. K. Horwitz, University of Colorado Health Science Center, Denver, CO). pMMTV-CAT containing the mouse mammary tumor virus long terminal repeat-driven chloramphenicol acetyltransferase gene was obtained from American Type Culture Collection (Rockville, MD). The pPB(Ϫ285/ϩ32)-LUC construct contains nucleotides Ϫ285 to ϩ32 of the rat probasin promoter (37) in the pGL3-Basic vector (Promega, Madison, WI). pGRE 2 E1b-CAT (pARE 2 E1b-CAT in this report) contains tyrosine aminotransferase AREs inserted upstream of the adenovirus E1b TATA sequence (from Dr. J. Cidlowski, National Institute of Environmental Health Sciences, Research Triangle Park, NC) (38). pARE 2 DS-LUC was generated by inserting a duplicated C3(1)-ARE into the SmaI site of pDS-LUC containing proximal sequences of the murine ornithine decarboxylase promoter (39). pUAS 5 E1b-CAT contains five GAL4-binding sites in front of the adenovirus E1b minimal promoter driving the chloramphenicol acetyltransferase gene (CLONTECH), and pUAS 4 tk-LUC contains four binding sites and the thymidine kinase promoter in front of the luciferase gene ( (22). F-SRC-1 cDNA in pBSIISK was digested with KpnI/XbaI and cloned into mammalian expression vector pcDNA3.1(ϩ) (Invitrogen, Carlsbad, CA). The ␤-galactosidase expression plasmid pCMV␤ was obtained from CLON-TECH, and pSV-␤-gal was from Promega. The following mammalian transactivation and two-hybrid system vectors were obtained from CLONTECH: pM for expressing the DBD of the Saccharomyces cerevisiae GAL4 protein (amino acid residues 1-147) and pVP16 for expressing the transcriptional activation domain of the herpes simplex virus VP16 protein (amino acid residues 411-456); and pM-53 coding for a fusion of the GAL4 DBD to mouse p53 protein and pVP16-T expressing a fusion of the VP16 activation domain to SV40 large T-antigen. The pM2GAL4-DBD fusion vector was a gift from Dr. I. Sadowski (University of British Columbia, Vancouver, Canada) (42). Testosterone was obtained from Makor Chemicals, Ltd. (Jerusalem, Israel). Restriction endonucleases and DNA-modifying enzymes were purchased from Pharmacia Biotech (Uppsala, Sweden). [ 3 H]Acetyl coenzyme A was purchased from NEN Life Science Products. AR antibody (anti-AR peptide 3) was raised against a region corresponding to residues 14 -32 of rat AR (43). Luciferase assay reagent was purchased from Promega.
Cell Culture and Transfections-CV-1 and COS-1 cells (American Type Culture Collection) were maintained in Dulbecco's minimal essential medium containing penicillin (25 units/ml), streptomycin (25 units/ ml), and 10% (v/v) fetal calf serum. Chinese hamster ovary (CHO) cells (American Type Culture Collection) were maintained in the same medium containing also nonessential amino acids. Cells were transfected using the calcium phosphate precipitation method as described previously (30,45). The cells (1.5 ϫ 10 6 ) were plated on a 10-cm dish 24 h before adding the precipitate with the amounts of expression and reporter vectors depicted in the figure and table legends. ␤-Galactosidase expression plasmids, pSV-␤-gal (4 g/10-cm plate) and pCMV␤ (2 g/ 10-cm plate), were used as internal controls for transfection efficiency. For preparation of whole cell extracts, the indicated amounts of expression vectors/10-cm dish were transfected by electroporation into COS-1 cells (37). An electrical discharge was applied by a Bio-Rad Gene Pulser transfection apparatus (settings, 300 V and 500 microfarads). For domain interaction studies (2.3 ϫ 10 5 cells/35-mm dish), 1.5 g of chimeric expression vectors and 3 g of pUAS 5 E1b-CAT reporter vector were transfected into CHO cells using DOTAP transfection reagent (Boehringer, Mannheim, Germany) according to the manufacturer's instructions. Eighteen hours after transfection, the medium was changed to one containing charcoal-stripped 2% (v/v) fetal bovine serum in the presence or absence of androgen. Chloramphenicol acetyltransferase and ␤-galactosidase activities were assayed as described previously (30,46,47). Protein concentration was determined using Bio-Rad reagents according to the manufacturer's instructions. Luciferase activity was determined with reagents from Promega using a Luminoskan RT reader (Labsystems, Helsinki, Finland) (39). Statistical analyses of the data were carried out using two-tailed Student's t test.
Immunoblotting-Whole cell extracts from transfected COS-1 cells were prepared as described previously and resolved by electrophoresis on polyacrylamide gels under denaturing conditions (43,45). Proteins were transferred onto Immobilon-P membrane (Millipore Corp., Bedford, MA) and processed as described previously (30).

Mapping of the N-terminal Transactivation Domain of Wildtype rAR-
We have previously shown that the AR⌬46 -408 and AR⌬38 -296 mutants of rAR are not able to activate transcription from either an ARE-containing simple promoter or a complex androgen-responsive promoter ( Fig. 1A) (30). On the other hand, the AR⌬40 -147 mutant had transcriptional activity comparable to that of the native receptor. To map the regions that are required for transcriptional activation in more detail, a series of N-terminal internal deletion mutants was constructed.
The activity of the AR⌬149 -295 mutant was only 14% of that of ligand-activated wild-type AR in CV-1 cells (Fig. 1A). Deletion mutants AR⌬212-295 and AR⌬256 -295 both had ϳ50% of the activity of the native receptor. Thus, deletion of 40 amino acids (residues 256 -295) decreased the transcriptional activity of AR by one-half, indicating that part of the core AF1 (activation function-1) is located within this region. Comparison of AR⌬149 -295 and AR⌬212-295 activities (14 Ϯ 1% versus 48 Ϯ 5%) identified a weaker AF1 region located N-terminally from residue 212, in the region of amino acids 149 -212. Very similar results were obtained with a reporter driven by the natural probasin promoter (pPB(Ϫ285/ϩ32)-LUC) ( Table I). In contrast to the results with probasin and thymidine kinase promoters, deletion of amino acids 212-295 (AR⌬212-295) increased the activity of the mutant protein over that of wild-type rAR when a reporter regulated by two AREs in front of the E1b TATA sequence (pARE 2 E1b-CAT) was used (Table I). Furthermore, deletion of residues 256 -295 caused a smaller decrease in the activation function of rAR on the minimal E1b promoter than on the two other promoters studied, suggesting that the region between amino acids 149 and 212 contains the major AF1 needed to activate this minimal promoter. The three latter receptor mutants were also examined in CHO cells using pARE 2 tk-CAT as the reporter. The results were similar, but not completely identical, to those obtained in CV-1 cells (Table  I). Collectively, there are at least two distinct subregions (amino acids 149 -212 and 256 -295) within the N-terminal region of rAR that mediate promoter-and cell-specific contacts in the transactivation process.
Mapping of the N-terminal Transactivation Domain of Constitutively Active rAR-We have previously shown that a partial deletion of the LBD (AR⌬788 -902) yields a receptor that has very low ligand-independent transcriptional activity (30). The influence of the LBD on transcriptional activity was further examined by generating rAR forms devoid of this region (AR⌬641-902 and its derivatives). AR⌬641-902 was functional on every promoter tested, and as expected, its activity was independent of androgen (Table II) (data not shown). This rAR form was up to three times more active than ligand-occupied native rAR when artificial promoters were studied, whereas with more complex promoters (mouse mammary tumor virus and probasin), its activity was similar to or up to five times lower than that of holo-rAR (Table II). These data are interpreted to mean that the LBD functions in synergy with AF1 to activate complex natural promoters such as that of the probasin gene under physiological conditions.
It was somewhat unexpected that identical N-terminal deletions in the wild-type rAR and AR⌬641-902 proteins influenced transactivation properties in a markedly dissimilar fashion (Fig. 1B). Only the deletion encompassing amino acids 46 -408 attenuated the function of the constitutively active receptor. Even this deletion, which completely extinguished the activity of wild-type rAR (1 Ϯ 1% of control; Fig. 1A), decreased  1. Functional analysis of rat AR mutants. The amino acids deleted from wild-type AR (A) and the receptor devoid of the entire LBD (B) are shown for each mutant, with the numbers after the ⌬ character depicting the first and last residues deleted. Functional activity of the proteins was tested by cotransfecting the receptor expression vector (1 g of DNA), the pARE 2 tk-CAT reporter (5 g of DNA), and pSV-␤-gal (4 g of DNA) into CV-1 cells using the calcium phosphate precipitation method as described under "Experimental Procedures." Eighteen hours after transfection, the cells received fresh medium with (A) or without (B) 10 nM testosterone and were cultured for an additional 30 h. After normalization for transfection efficiency using ␤-galactosidase activity, reporter gene activities (chloramphenicol acetyltransferase (CAT)) are expressed relative to that of wild-type rAR in the presence of androgen, and the means Ϯ S.E. for at least three experiments are given in percentages (pSGrAR plus testosterone ϭ 100%). the activity of AR⌬641-902 on pARE 2 tk-CAT only by one-half. Comparable results were obtained with the reporters driven by the E1b minimal promoter or the probasin promoter (Table III). Thus, the ability of AR to use alternate N-terminal regions for transcriptional activation depends on the presence of the LBD. These results also suggest that the hormone-occupied LBD restricts the function of the N-terminal domain in that, in holo-rAR, region 296 -542 is incapable of interacting with transcription machinery (or coactivators). In the context of AR devoid of the LBD (AR⌬641-902), however, this restriction is relieved. The different transactivation abilities of the receptor mutants examined could not be explained by their dissimilar expression levels, as immunoblot analyses of extracts from transfected COS-1 cells showed that expression levels of mutant proteins were comparable to those of wild-type rAR (Fig. 2) (data not shown).
Transcriptional Activation by GAL4 DBD-rAR Fusion Constructs-To examine the function of the N-terminal region in the absence of other domains, amino-terminal fragments of rAR were cloned in frame to the carboxyl terminus of the S. cerevisiae GAL4 DBD. Chimeric proteins were expressed in CHO and CV-1 cells, and their transactivation function was examined using two reporter genes with GAL4-binding sites as cis-acting elements (pUAS 5 E1b-CAT and pUAS 4 tk-LUC). The GAL4 DBD fused to residues 5-538 of rAR activated the E1b minimal promoter 44-fold in CHO cells (Table IV). The most active N-terminal subregion (amino acids 208 -293) induced a 12-fold increase in the reporter gene activity. Under the same conditions, the hormone-occupied LBD fused to the GAL4 DBD showed ϳ10% of the activity of the longest N-terminal construct in CHO cells (44). The longest N-terminal fusion con-struct (GAL4 DBD-rAR-(5-538)) exhibited much stronger transcriptional activity in CV-1 cells than in CHO cells and activated the reporter gene by 600-fold. The entire N-terminal region was needed for the full activity, as the fragment containing amino acids 144 -293, which had the highest activity of the subdomains studied, stimulated transactivation only by 9-fold in CV-1 cells.
The ability of the N-terminal region of rAR to activate transcription showed distinct promoter specificity (Table IV). Only an 8-fold induction was maximally observed with GAL4 DBD-rAR-(5-538) when a reporter driven by the thymidine kinase promoter was examined. In this case, the most active subregion (residues 295-550) showed 50% of the activity of the longest construct (GAL4 DBD-rAR-(5-538)). Collectively, the subregions that are important for transcriptional activation in the context of native rAR were not able to function as autonomous AF regions as efficiently as the entire N-terminal region.
Hormone-dependent Interaction between the LBD and N-terminal Region-Studies on the transactivation ability of native rAR and the AR⌬641-902 mutant showed that the receptor employs alternate N-terminal activation interfaces, depending on the presence of the LBD. A likely explanation is a physical Immunoblot analysis was performed as described previously (30) using rabbit antiserum raised against a synthetic peptide corresponding to amino acid residues 14 -32 of rAR (ARp3). All samples originated from the same experiment and were processed simultaneously.

TABLE IV Functional analysis of GAL4-AR chimeric constructs
Functional activity of the chimeric constructs was tested by cotransfecting GAL4-AR expression plasmids in amounts that yielded maximal effects (0.2 g of DNA when pUAS 4 tk-LUC was used in CV-1 cells, 2.5 g when pUAS 5 E1b-CAT was used in CV-1 cells, and 5 g when pUAS 5 E1b-CAT was used in CHO cells) and reporter genes (5 g of DNA) into CV-1 or CHO cells using the calcium phosphate precipitation method as described under "Experimental Procedures." The cells were cultured for 30 h. Reporter gene activities are expressed relative to the GAL4 DBD alone (ϭ1%). The mean values for at least two separate experiments are shown. pM3-VP16 was used as a positive control.  interaction between the N-terminal region and LBD. This possibility was investigated using the mammalian two-hybrid system, in which the fusion vectors contained rAR sequences cloned carboxyl-terminal to the GAL4 DBD and to the activation domain of the herpes simplex virus VP16 protein. Chimeric proteins were expressed in CHO cells. pM-53 and pVP16-T vectors expressing mouse p53 protein and SV40 large T-antigen chimeras, respectively, were used as positive controls. When GAL4 DBD-rAR LBD and VP16-rAR-(5-538), a construct containing residues 5-538 fused to the activation domain, were coexpressed in the presence of androgen, strong activation of the reporter gene driven by GAL4-binding sites in front of the E1b minimal promoter (pUAS 5 E1b-CAT) was observed (Fig. 3). This interaction was strictly dependent on the presence of androgen. No activation occurred when either plasmid was coexpressed with parent vectors lacking rAR sequences. Essentially identical results were seen in CV-1 cells (data not shown). Fusion of shorter rAR sequences to VP16, such as fragment 37-265, failed to activate the reporter gene or yielded activity that was Ͻ10% of the longest AR fragment (Fig. 3). Thus, most of the N-terminal region is required to generate a functional interaction interface for the two receptor domains, even though the LBD did interact with the region adjacent to the DBD (residues 295-550) (Fig. 3) encompassing the LBD-restricted activation function (c.f. Fig. 1B).

Modulation of AR-mediated Transactivation and Interaction between the Amino-and Carboxyl-terminal Domains by Nuclear Receptor Cofactors-
The above results indicated that androgen binding promotes the interaction between the aminoand carboxyl-terminal regions of rAR. It is conceivable, however, that the interaction between the N-terminal domain and LBD requires auxiliary proteins such as nuclear receptor coactivators. This possibility was addressed by studies in which the N-and C-terminal domains of rAR were coexpressed with different nuclear receptor cofactors, CBP (24,29), F-SRC-1 (22), SRC-1 (21), and RIP140 (13), known to associate with other steroid receptors. However, the effects of these cofactors on the transcriptional activity of wild-type rAR were first tested using pARE 2 E1b-CAT or pPB(Ϫ285/ϩ32)-LUC reporters in CV-1 cells.
None of the coactivators influenced promoter activity in the absence of hormone. Coexpression of F-SRC-1 or CBP with rAR enhanced the hormone-dependent transcriptional activity from the minimal E1b promoter (Fig. 4A). RIP140 plus rAR also increased promoter activity over that with rAR and androgen alone (Fig. 4A). By contrast, SRC-1, a truncated form of F-SRC-1 (22) that functions as a coactivator for PR and GR in HeLa cells, completely abolished rAR-dependent transcription from the minimal promoter (Fig. 4A). Even a very low amount of SRC-1 expression plasmid (0.1 g/10-cm dish) decreased rAR-mediated transactivation to 45.1 Ϯ 24.7% of that with rAR in the presence of testosterone. To verify that this unexpected effect of SRC-1 on rAR function was not a particular feature of CV-1 cells, the influence of SRC-1 on PR-mediated transactivation was examined in these cells. SRC-1 elicited a 2-fold increase in the ligand-dependent transcriptional activity of hPR-B (192.0 Ϯ 27.7 and 194.6 Ϯ 9.2% of that with hPR-B in the presence of progesterone with 0.1 and 6 g of SRC-1 expression plasmid, respectively), which is in agreement with results reported with HeLa cells (21). Similar to the minimal androgen-responsive promoter, F-SRC-1, CBP, and RIP140 were also able to enhance rAR-dependent transactivation from the probasin promoter (Fig. 4B). Probasin promoter activation was inhibited by coexpressed SRC-1, albeit to a lesser extent than that of the minimal promoter (Fig. 4B).
Even though the above cofactors were originally identified through interaction with LBDs of nuclear receptors, it was pertinent to investigate their ability to modulate the function of an rAR form devoid of LBD. AR⌬641-902 was coexpressed with various coactivators with pARE 2 E1b-CAT as the reporter. F-SRC-1, CBP, and RIP140 behaved as efficient coactivators also for AR⌬641-902, whereas SRC-1 abolished the activity of this rAR form (Fig. 5A). F-SRC-1 was, however, incapable of conferring activation upon several other mutant rAR forms, such as AR⌬46 -408 or AR⌬788 -902 (Fig. 5B). Thus, other rAR domains besides the LBD are able to contact nuclear receptor coactivators. In addition to the amino-terminal region and DBD, the LBD-deficient construct used (AR⌬641-902) contains one-half of the hinge region residues 607-659 (35,48).
The possibility that the above cofactors modulate the interaction between the amino and carboxyl termini of rAR was tested using the mammalian two-hybrid system with the fusion vectors GAL4 DBD-rAR LBD and VP16-rAR-(5-538). Coexpression of F-SRC-1 or CBP in CV-1 cells with these fusion proteins in the presence of androgen enhanced the interaction between the N and C termini (Fig. 6). The coactivators did not influence rAR-mediated transcription in the absence of hormone or that of GAL4 DBD-rAR LBD alone in the presence of androgen. Likewise, the activity of the GAL4 DBD-rAR-(5-538)

FIG. 3. Interactions between aminoand carboxyl-terminal regions of rAR.
Combinations of the indicated parent and chimeric expression vectors (1.5 g) and the reporter gene pUAS 5 E1b-CAT (3 g) were cotransfected into CHO cells using DOTAP reagent (see "Experimental Procedures"). Reporter gene activities are expressed relative to that of the interaction between VP16-rAR-(5-538) and GAL4 DBD-rAR LBD constructs in the presence (ϩ) or absence (Ϫ) of 20 nM testosterone (ϭ100%). The means Ϯ S.E. for at least three separate experiments are shown. pm-53 and pVP16-T vectors expressing the mouse p53 protein and SV40 large T-antigen chimeras were used as positive controls. CAT, chloramphenicol acetyltransferase. fusion construct alone was not influenced by cofactors (data not shown). In agreement with the role of CBP as an adapter of Nand C-terminal interactions, the 12 S E1A adenoviral protein, which binds and inactivates CBP (41), abolished the interaction between the N and C termini (Fig. 6). SRC-1 also inhibited the domain interaction, in keeping with its influence on rARmediated transactivation. However, RIP140, which potentiated transactivation by rAR, strongly inhibited the interaction between the amino-and carboxyl-terminal regions of rAR (Fig. 6). DISCUSSION To study the transactivation properties of rAR, a series of deletion mutants was analyzed to delineate amino-terminal regions important for transcriptional activation. The region encompassing residues 147-296 crucial for activating function does not possess significant sequence homology to other members of the nuclear receptor superfamily. This AF1 region is acidic and relatively rich in proline and leucine and contains a long poly(glutamine) stretch between residues 174 and 196 along with several charged amino acids (35,48). Expansion of a comparable, but more N-terminal poly(glutamine) repeat in hAR is associated with Kennedy's disease and late-onset androgen insensitivity (49,50). Almost the entire N-terminal domain (residues 1-485), which contains three glutamine repeats, is required for the full transcriptional activity of hAR (34). The core activation region of hAR encompassing amino acids 101-370 has only one of those glutamine repeats. The core AF1 region of rAR that mediates over half of the maximal activity of the wild-type receptor was located between residues 149 and 212 or residues 256 and 295, depending on the promoter construct examined and the cell line studied, indicating FIG. 4. Modulation of AR-mediated transactivation by putative nuclear receptor coregulators. A, modulation of transactivation by wild-type rAR was tested in CV-1 cells by cotransfecting the receptor expression vector (0.2 g of DNA/10-cm dish) along with the indicated amounts of coregulator expression vectors (in g of DNA), the pARE 2 E1b-CAT reporter (5 g), and pCMV␤ (2 g) using the calcium phosphate precipitation method as described under "Experimental Procedures." The total amount per dish was kept constant by adding empty pcDNA3.1 vector as needed. Sixteen hours after transfection, the cells received fresh 2% (v/v) fetal bovine serum/Dulbecco's minimal essential medium in the presence (ϩ) or absence (Ϫ) of 10 nM testosterone (T) as depicted. After normalization for transfection efficiency using ␤-galactosidase activity, reporter gene activities are expressed relative to those of rAR in the presence of androgen without a coregulator (ϭ100%). The means Ϯ S.E. for at least three separate experiments are shown. B, the experimental conditions were the same as described for A, except that 1 g of wild-type rAR expression plasmid was transfected with the indicated coregulator expression vectors (6 g of DNA) and the pPB(Ϫ285/ ϩ32)-LUC reporter. CAT, chloramphenicol acetyltransferase. that the long poly(glutamine) stretch between residues 174 and 196 in rAR is most likely not mandatory for its activation functions. Quite similar findings on AF1 of rAR using mouse mammary tumor virus as the model promoter in CV-1 cells were recently reported by Chamberlain et al. (51). They mapped AF1 to region 117-326 and described that a core AF1a resides between residues 154 and 167. Another distinct AF1, AF1b, was identified between residues 295 and 359. Promoteror cell-type specificity for AF1a and AF1b was, however, not examined in the study of Chamberlain et al. (51).
Similar to AR, the major AF1 of human GR has been mapped to a 185-amino acid segment (residues 77-262, tau-1) in the central part of the N-terminal region (52). A region composed of 41 amino acids near the C terminus of tau-1 is reported to possess a core AF1 (53). Certain aromatic amino acid residues may be important for AF1 function (54). However, the AF1 regions that were identified for rAR in our and other (51) experiments do not bear amino acid sequence identity to the core AF1 of human GR besides having a similar amino acid composition. Interestingly, the localization of AF1 in hPR, as determined using GAL4-hPR fusion constructs, differs considerably from those in AR and GR and maps to a 91-amino acid stretch immediately adjacent to the DBD (residues 456 -546) (11).
Our results demonstrating that AF1 exhibits promoter context-and cell-type specific functions (Table IV) suggest that this region is capable of contacting different coactivator proteins, some of which may be cell-specific. Similar phenomena have been reported for human PR and ER (9,11,(55)(56)(57)(58). AF1 of ER efficiently activated transcription in chicken embryo fibroblasts, whereas it behaved as a poor transactivator by itself in HeLa cells, but could synergize with AF2 to activate some promoters (55)(56)(57). The manifestation of AF1 and AF2 of human ER also depends strongly on the promoter context; on certain promoters, both activation functions are required for full transcriptional activity, whereas on others, AF1 and AF2 may function independently (58).
Deletion of AF1 did not reduce the transcriptional activity of AR⌬641-902, and the largest deletion (residues 46 -408) caused reduction of its activity only to one-half, indicating that dissimilar N-terminal regions are employed for transactivation depending on the presence of the LBD. The amino-terminal region adjacent to the DBD (residues 360 -528) of hAR was sufficient to activate transcription by the constitutively active human receptor (34). Taken together, these data imply that the ligand-induced conformational change in the AR LBD allows for the interaction of the LBD with the N-terminal region in such a fashion that the domain adjacent to the DBD (residues 408 -542 in rAR) is not accessible for a productive interaction with auxiliary proteins.
When the N-terminal region and LBD were expressed as separate fusion proteins, they showed a strong androgen-dependent interaction in intact cells. This finding is consistent with the notion that intramolecular interactions between the N-terminal AF1 and LBD modulate the transcriptional activity of AR. Ligand-dependent interaction between the corresponding domains of hAR has been reported previously (59,60). In agreement with our findings on rAR, studies on the human receptor, performed also in yeast, have indicated that most of the N-terminal domain of hAR was required to generate an efficient interaction interface for the ligand-occupied LBD (59,60). In particular, the first N-terminal residues may be important for the interaction interface, as judged by the poor activity of fragment 37-265 in mammalian two-hybrid assays. Functional interaction of AF1-and LBD (AF2)-containing regions of ER, expressed as separate polypeptides in mammalian cells, has also been demonstrated to occur in response to estrogen and antiestrogen binding (61,62). Hormone-dependent association of the AF1 and AF2 regions of steroid receptors may thus be a general requirement for their maximal ability to activate transcription.
The association between the ligand-occupied LBD and AF1 can be direct or indirect. In view of this, some recently identified nuclear receptor cofactors such as SRC-1, F-SRC-1, RIP140, and CBP (13,21,22,24,29), which have been shown to associate with hormone-bound LBDs of some other steroid receptors, were examined as potential adapters of AF1 and LBD association. Hormone-induced interaction between the N-terminal transactivation domain and LBD of rAR, when expressed as separate polypeptides in mammalian cells, was facilitated by the coactivators F-SRC-1 and CBP. These coactivators also enhanced the transcriptional activity of wild-type rAR. In vitro binding studies have previously demonstrated that F-SRC-1 binds to ER, the thyroid hormone receptor, the retinoid X receptor, and the retinoid acid receptor in a liganddependent fashion (22). It may also associate with subregion(s) other than those in AF2 of some nuclear receptors (22). SRC-1 is highly homologous to F-SRC-1, but lacks a 380-amino acidlong region in the N terminus. This additional sequence exhibits strong homology to a PAS domain typical of the PAS/helixloop-helix proteins, which mediates dimerization between PAS domain-containing proteins as well as between PAS and non-PAS proteins (24,25).
SRC-1 functions as a coactivator for various nuclear receptors (thyroid hormone receptor, retinoid X receptor, ER, PR, and GR) in HeLa cells (21). It was therefore unexpected that it was a potent inhibitor of both transactivation and interaction between the amino-and carboxyl-terminal domains of rAR. Interaction of SRC-1 with rAR appears to form transcriptionally abortive complexes, whereas F-SRC-1, which contains the PAS domain (22), both binds to rAR and forms transcriptionally active complexes. Squelching of other transcription factors is not a likely explanation for the inhibitory action of SRC-1, as the inhibition of transactivation by rAR was detectable already with very low amounts of SRC-1 expression plasmid (0.1 g/10-cm plate).
Although the transactivation properties of AR, GR, and PR are in many aspects similar, some of their cross-modulatory properties with AP-1 and RelA family members are distinguishable (31,37). We have previously suggested that the cross-talk between AR and RelA involves, at least in part, competition for coactivators of transcription (37). Moreover, wild-type rAR, which is clearly a less potent transactivator than GR and PR, has dominant-negative action on GR-and PR-mediated transcription, suggesting differential usage of cofactor proteins by steroid receptors (63). Dissimilar coactivator and/or corepressor requirements by members of the GR subfamily recognizing similar DNA elements may have some bearing on the acquisition of hormone-specific responses when a cell contains concomitantly multiple receptors and their ligands.
CBP/p300, originally identified as a cAMP response elementbinding protein-binding protein (40), interacts with a wide variety of transcription factors (reviewed in Ref. 27), contains intrinsic histone acetyltransferase activity (64,65), and associates with RNA polymerase II and transcription factor IIB, suggesting that this coactivator stimulates transcription, at least in part, through the recruitment of RNA polymerase II to target gene promoters. CBP has been shown to interact physically with several members of the nuclear receptor superfamily and enhance their transcriptional activity (24,29,65,66). Consistent with the role of CBP as an adapter in the N-and C-terminal interaction of rAR, the 12 S E1A adenoviral protein, which binds and inactivates CBP (41), abolished this interaction. Thus, a potential mechanism by which the coactivators F-SRC-1 and CBP facilitate the transcriptional activity of rAR is the stabilization of amino-and carboxyl-terminal interactions.
RIP140 has been shown to interact with the LBDs of ER, the thyroid hormone receptor, and retinoid receptors (13,67). In our experiments, RIP140 enhanced AR-mediated transactivation, but prevented the association of N and C termini. This contrasting behavior of RIP140 in the domain interaction experiments and transactivation studies could be explained by the possibility that rAR is capable of recruiting interaction of RIP140 only when the receptor is bound to AREs in a dimeric form.
Many of the above coregulators were originally identified on the basis of their interaction with LBDs (plus hinge regions) of nuclear receptors (20). In view of this, it was interesting that the activity of the rAR form (AR⌬641-902) devoid of the LBD and one-half of the hinge region residues, but not those forms lacking the N-terminal AF1, was modulated by coactivators. However, the activity of the N-terminal transactivation domain (residues 5-538) as a GAL4 DBD fusion protein was not influenced by coactivators, suggesting that the DBD plus the hinge region and/or dimerization of the receptor plays a role in the coregulator-rAR interaction. Since in vitro interactions between purified AR proteins and coactivators were not studied in this work, we cannot rule out the possibility that coactivator recruitment by rAR involves auxiliary proteins.
Nuclear receptor coactivators and corepressors, some of which communicate with other signaling systems as well, provide another means for cross-talk among the receptors. The availability of these coregulators may modulate transcription in cells where several nuclear receptors are simultaneously expressed. Dissimilar expression of coactivators and/or corepressors may also contribute to differences in transcriptional activation observed among various steroid receptors in certain target cells. The physiological significance of CBP and F-SRC-1 in the development of androgen-responsive phenotype and androgen insensitivity warrants further genetic studies.