INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Many developmental, reproductive, and homeostatic functions are coordinated by steroid hormones via their specific receptors that regulate gene expression (1). The basis for distinct physiological roles of each steroid remains enigmatic, because the highly conserved DNA binding domains (DBD)¹ of several receptors recognize a common hormone response element (HRE). Multiple mechanisms, acting through distinct receptor domains, may influence specificity. Notably, other transcription factors may interact differentially with receptors, in part via dissimilar N-terminal transactivation domains. The DBDs may display differential affinities on nonconsensus HREs of natural promoters. Further, recently discovered co-activators that interact broadly with conserved regions of the ligand binding domain (LBD) may prove to discriminate among receptors.

Steroid receptors interact directly with a variety of transcription factors to elicit precise positive or negative regulation (2-4), and these interactions can be highly context-dependent (5, 6). Nevertheless, no factor has yet been shown to be required by a particular receptor, or to be exclusive in its interactions. The growing number of receptor co-activators and co-repressors (reviewed in Refs. 7 and 8) do not resolve specific response, since they appear to function widely across the nuclear receptor family through conserved protein domains. However, some co-activators more limited in their receptor associations have been described (9).

Some complex regulatory elements from genes that are androgen-specific in vivo retain a specific steroid response when fused to reporter genes, allowing examination of sequence requirements for specificity (10-12). These enhancers require as yet unidentified DNA binding factors for specific response, and characteristically depend on multiple nonconsensus HREs, some of which bind androgen receptor (AR) with higher affinity than glucocorticoid receptor (GR) (13-15). This discrimination among variant HREs may prevent inappropriate hormonal activation, and the difference may be amplified by multiple receptor binding sites.

For AR, activity is modulated by interactions between distinct functional domains of the homodimer. In particular, strong interaction between the amino and carboxyl termini of AR has been implicated in ligand binding, receptor stabilization, and enhanced transactivation (16-20). Physical association between domains requires an intact LBD, within which distinct surfaces are contacted, and multiple regions of the N terminus (17, 19, 20). Within the AR N terminus, a transcriptionally active discrete subdomain akin to the GR 1 (21) has not been defined, but active regions have been revealed by their deletion (18, 22, 23). Large N-terminal deletions render the receptor a trans-dominant repressor of full-length AR (24). Surprisingly, the LBD is most active in in vitro transcription (25), demonstrating that isolated domains can have unexpected behaviors.

For androgen-specific response, the enhancer of the mouse sex-limited protein (Slp) gene depends on a consensus HRE and multiple nonreceptor factor binding sites within a 120-bp DNA fragment (10). GR not only fails to activate this enhancer in transfection, but blocks AR activation, dependent upon an intact GR DBD. Subtle alterations in sequence or position of the HRE, or inclusion of binding sites for other factors, can reduce specificity and permit GR function (13). A chimeric receptor that replaced the AR N terminus with that of GR activated simple HREs but not the androgen-specific enhancer, demonstrating importance of this region in selectivity (10). However, the N terminus of AR before the DBD and LBD of GR was not sufficient to activate the specific enhancer. To determine domain involvements in specificity, we tested an array of chimeric receptors and find that each domain plays a distinct role. The AR DBD escapes suppression of activity that operates on the GR DBD in the context of the specific enhancer. Perhaps more importantly, interactions within and between receptors increase utility of nonconsensus HREs. This provides a mechanism whereby a specific response can be generated by AR cooperativity from multiple weak elements with poor affinity for GR.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Construction of Mutant Receptors-- The mouse androgen receptor (mAR) and rat glucocorticoid receptor (rGR) cDNAs (26, 27) (originally from D. Tindall and K. Yamamoto, respectively) were cloned into the pCMV5 expression vector (28). Chimeric receptors were generated by exchanging the N termini, DBD, or LBD of AR and GR. To create a joint between the DBD and LBD, a BamHI site was introduced by overlap extension PCR-mediated mutagenesis, at Asn⁵³³ for GR and Asp⁶³⁴ for AR. Primers for mutagenesis were (mutated bases are lowercase, BamHI sites are underlined): mAR sense, 5'-CCC ACT GAG GAt CCA TCC CAG A-3' (1891-1913); antisense, 5'-CTG GGA TGG aTC CTC AGT GGG-3' (1912-1891); rGR sense, 5'-CAC TTC GcA ggA TCC TAA CAA AAC-3' (1587-1611); antisense, 5'-GTT TTG TTA GGA Tcc TgC GAA GTG-3' (1611-1587). To retain similar rGR sequence character, Glu⁵³² and Asn⁵³³ were changed to Gln⁵³² and Asp⁵³³. No amino acids were changed in mAR. Primers for the other ends of the PCR products were directed at the translation start codon (ATG), and were altered to include a KpnI site overlapping a Kozak consensus: mAR sense, 5'-ggc ggt acc gcc ATG GAG GTG CAG TTA GGG-3'; rGR sense, 5'-ggc ggt acc gcc ATG GAC TCC AAA GAA TCC-3'; and at the 3'-untranslated region (ending with an XbaI site): mAR antisense, 5'-gcg tct aga GTT ATA TAA CAG GCA GAA G-3'; rGR antisense, 5'-gcg tct aga GTT CAA CTT TCT TTA AGG CAA C-3'.

Construction of Mutant CAT Reporters-- The same primers used to generate the linker scan mutants of C'9 by overlap extension PCR (13) were used to mutate HRE-1 (ls7) in the context of C'9 mutants in which the position of HRE-3 was altered by 10 bp (space 10, s10), to generate s10-ls7.

Cells and Transfection-- CV-1 cells were grown and transfected, and CAT activity assayed as described previously (13, 30). Briefly, 9 µg of reporter DNA and 100 ng of pCMV5 receptor expression plasmid were introduced by DEAE dextran into 10⁶ CV-1 cells in a 10-cm dish. Cells were grown 40 h with or without 10⁷ M dexamethasone, or 10⁸ M dihydrotestosterone (added every 12 h), in Dulbecco's modified Eagle's medium plus 3% charcoal-stripped NuSerum IV (Collaborative Research). Between 10 and 40 µg of cell protein were assayed for CAT activity by the phase extraction method (31). Some transfections as noted were performed by calcium phosphate precipitation of the same amounts of reporter and receptor DNAs, and 2 µg of Rous sarcoma virus-luciferase plasmid. Cells were grown in 5% charcoal-stripped NuSerum IV in Dulbecco's modified Eagle's medium, with or without hormone as before, and glycerol-shocked after 16 h. 24 h later, cells were harvested for CAT activity in the cold to stabilize luciferase.

Western Blot Analysis-- COS-7 cells at 50% confluence were transfected by calcium phosphate precipitation of 20 µg of pCMV5 receptor DNA. 40 h post-transfection, and 4 h after addition of hormone (10⁷ M dihydrotestosterone/10⁷ M dexamethasone), cells were harvested with 0.5 ml/10-cm dish of SDS-TEN buffer (62.5 mM Tris-HCl, pH 6.8, 10% glycerol, 2% sodium lauroyl sulfate; 5% v/v -mercaptoethanol) (32). DNA was sheared by passing the lysate four times through a 26-gauge needle, and proteins were resolved by SDS-polyacrylamide gel electrophoresis. Gels were electroblotted at 15 mA overnight onto BA-85S nitrocellulose (Schleicher & Schuell), which was then blocked with 5% nonfat milk in Tween 20-Tris-buffered saline buffer. Receptors containing the AR N terminus were revealed with a rabbit polyclonal antisera raised against a glutathione S-transferase fusion protein containing mAR residues 133-334 (33), while those with the GR N terminus were detected with FIGR2 monoclonal antibody (34), a generous gift of W. Pratt. Antibody complexes were visualized with the Amersham chemiluminescent kit.

DNase I Protection Analysis-- DNase I protection was performed as described previously (15). Recombinant receptors were obtained from Spodoptera frugiperda Sf9 cells infected with baculoviruses (kindly provided by O. Janne), harboring full-length (FL) rAR, rAR deleted between residues 46 and 408 (N), or rAR truncated at amino acid 788 (C) (35). Cells were grown and whole cell extracts prepared as described previously (36). After dialysis in binding buffer (below) to remove ammonium sulfate, aliquots were stored at 70 °C. For footprinting, increasing quantities of whole cell extracts were mixed with binding buffer (20 mM HEPES, pH 7.6, 50 mM KCl, 0.2 mM EDTA, 2 mM dithiothreitol, 20% glycerol), 1 µg of poly(dI-dC), 50 nM dihydrotestosterone, protease inhibitors (0.5 mM phenylmethylsulfonyl fluoride, 2 µg/ml leupeptin, 2 µg/ml pepstatin, 2 mM benzamidine, 20 milliunits/ml aprotinin), 0.05% Nonidet P-40, and 10,000 cpm of end-labeled DNA probe. The probes were one or three copies of the consensus HRE-3 cloned into the BamHI site of tkCAT (30), the specific enhancer C'9 (10), or C'9 mutated at HRE-3 (ls9) or at the weak HRE-1 (ls7). Probes were end-labeled with ³²P by polynucleotide kinase at the MluI site and excised with PvuII. After binding the cell extracts to the probe for 3 h on ice, MgCl₂ and CaCl₂ were added to 2 mM each, and varying concentrations of DNase I added for 5 min on ice. The reactions were terminated, deproteinized, concentrated, and resolved on 8% sequencing gels as described previously (13).

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Previous results from our laboratory suggest that the AR N terminus is necessary but not sufficient for androgen-specific gene activation (10). To examine the intrinsic activities of the major receptor domains, as well as interactions between domains, chimeric receptors were constructed from mAR and rGR. Unique restriction sites were introduced at regions of peptide sequence homology to rejoin the cDNAs (see "Materials and Methods"). A NheI site placed within the first zinc finger and a BamHI site within the nuclear localization signal divided the receptors into three cassettes encompassing N terminus, DBD, and LBD (Fig. 1A). Receptors are named by the origin of each domain, with A or G for AR or GR; thus the N terminus of GR fused to the DBD and LBD of AR is the chimera GAA.

View larger version (50K): [in this window] [in a new window]   Fig. 1.   Construction and expression of AR-GR chimeric receptors. A, the N terminus, DBD, and LBD are depicted for mAR and rGR. Positions of the introduced unique NheI (n) and BamHI (b) sites are indicated. B, wild type and mutant receptors were expressed in COS-7 cells for 44 h and detected by Western blotting of 20 µg of whole cell extract. LNCaP and HTC cells express endogenous hAR and rGR, respectively; the COS lane contains extract from untransfected cells. Unmodified mAR and rGR are compared with parent receptors containing the NheI and BamHI sites and modifications of the 5'-untranslated region (ARnb, GRnb). Antibodies directed against N-terminal sequences detected receptors with the AR (left) or the GR N terminus (right) (33, 34). C, activation by chimeric receptors of the androgen-specific enhancer C'9 compared with multimerized simple response elements, 3xHRE-3, was tested by transfection into CV-1 cells. Activity of the linked CAT reporter is expressed as the percent conversion of chloramphenicol to butyrated forms, with the values on C'9 shown in black bars and 3xHRE-3 in white bars; note different scales. The S.E. of four independent assays is shown for C'9. For 3xHRE-3, the extract amount was adjusted to keep assays in the linear conversion range, and values were normalized to AR; S.E. values were within 12% of total activity. Diagrams below show C'9, with the consensus HRE-3, the half-site HRE-1, and a crucial nonreceptor binding site (FPIV), in comparison to 3xHRE-3.

AR, GR, and chimeric receptors were introduced into COS-7 cells, and their expression was compared by Western blotting (Fig. 1B). The receptors were all correct in size, but expression levels varied, despite each having identical 5'-untranslated sequences. AGG accumulated to low levels while GAA and GGA reached high steady states. These disparate levels could result from differences in protein stability or formation of inclusion bodies following overexpression in COS cells (32, 37, 38). However, the functional level of these receptors, as assayed by activation of simple HRE-tkCAT reporters (see below), was more equivalent, and extremes of activity did not correlate with protein expression levels detected by Western blot analysis (see below).

The chimeric receptors were tested for induction of CAT activity from the androgen-specific enhancer C'9, in comparison to multimerized consensus HREs (3xHRE-3), by transfection into CV-1 cells (Fig. 1C). Relative activity is shown rather than fold induction since activity of these reporters in the absence of hormone was negligible. AR activation of C'9-tkCAT was 42% the activity of 3xHRE-3-tkCAT; in Fig. 1C, activation of the two reporters is on different scales to allow direct comparison of specific versus general function of each receptor. The chimeras activated simple HREs at least as well as AR, with AGG and AGA having a fewfold greater activity, except for GGA which was only one-fourth as active as AR. In sharp contrast, only AAG and GAG activated the specific enhancer C'9 to levels approaching that of AR. These two chimeras had in common the AR DBD, indicating a crucial role for this domain. However, the AR DBD may be necessary but was not sufficient for C'9 induction since GAA was only somewhat more active than GR on the specific enhancer. While not all receptors containing the AR DBD could activate C'9, receptors containing both the AR N terminus and DBD could (AAG and AR itself). In fact, GAG activation of the androgen-specific enhancer was unexpected, because DNA selectivity alone is not known to distinguish AR and GR targets. This suggested experiments below to test whether multiple and perhaps alternative functions were involved in specific response.

We have previously shown that AR requires not only the consensus HRE-3 but also a nearby half-site, called HRE-1, to activate the androgen-specific enhancer C'9 (15). To contrast the mechanisms by which AAG and GAG activated the specific enhancer, their use of the weak HRE-1 was tested. Mutation of this site (C'9-ls7) revealed disparate sensitivities to HRE-1 (Fig. 2), whereas all receptors required HRE-3 for activation (not shown). Unlike AR, whose induction was reduced 5-fold in the absence of HRE-1, GAG was at least as active on the mutant as on the wild type enhancer. AAG activity was decreased without HRE-1, but not to the extent of AR. The dependence upon multiple receptor binding sites suggested that activation by AR relied on cooperativity between receptor dimers. That AAG was less affected than AR suggested multiple domains might be involved in cooperative interactions. To explore this further, ARC, which lacks most of the LBD but is half as active as AR on C'9, was tested and, in fact, showed little sensitivity to the weak HRE (Fig. 2). This suggests that the ability of AR to use HRE-1 is not intrinsic to the N terminus or the DBD, but may involve interaction of the AR N and C termini. The intermediate dependence of AAG on two HREs may stem from some functionality provided by the partial homology of the GR LBD.

View larger version (41K): [in this window] [in a new window]   Fig. 2.   Activation by AR depends upon multiple HREs. AR, AAG, GAG, and ARC were tested for activation of C'9 mutated at the HRE-1 half-site (C'9-ls7, depicted below the histogram). Activity ±S.E. for each receptor is relative to activation by that receptor of wild type C'9.

To test further the role of domain interactions and utilization of multiple HREs, a mutant of C'9 was used that supports activation by GR. This variant enhancer has a linker scan disruption within HRE-3 and a copy of the HRE-3 sequence inserted 10-bp distal to the enhancer (see Fig. 3). GR can activate this reporter (s10) about half as well as AR, which was ascribed to escaping repressive effects of juxtaposed nonreceptor factors (13). Mutant receptors that poorly activated C'9 were tested for dependence on a second HRE in derivatives of s10 in which HRE-1 had been inactivated (s10-ls7). AGA, like AR, showed greater induction when both HREs were present (Fig. 3). In contrast, GR, GAG, and ARC were unaffected by mutation of the weak element, while AAG (not shown) was intermediate in sensitivity to AR and GR. The sensitivity by AGA to mutation of the weak HRE was greater even than that of AR. This suggested that both AR N and C termini are involved, via direct or indirect interaction, in cooperative usage of multiple HREs. Further, AGA revealed that the AR DBD is not requisite for activity per se in this context, but may serve to evade repression on the specific enhancer that is directed at the other steroid receptors via their DBDs.

View larger version (32K): [in this window] [in a new window]   Fig. 3.   Utilization of HRE-1 is dependent on interaction between the AR N terminus and LBD. AR, GR, AGA, GAG, and ARC were tested on the s10 mutant of C'9 that is activable by GR due to displacement of HRE-3. The CAT activity ±S.E. is normalized to AR activity on s10, with GAG and ARC on different scales. 50 ng of receptor plasmid were used in transfections, except for ARC which was at 100 ng. The CAT reporters are depicted below, with targeted mutagenesis of either the consensus HRE-3 or the half-site HRE-1 indicated by an X.

Involvement of the AR N Terminus and LBD in Cooperative DNA Binding-- Results thus far indicated that AR activation of the specific enhancer relied on binding to both HRE-1 and -3. This was not solely a function of the DBD, suggesting that distinct domains of receptor dimers at nearby sites might enhance cooperative DNA binding, due to direct or indirect interactions. To obtain physical evidence of this, baculovirus-expressed FL or N- or C-terminally deleted (N, C) rat ARs were used in DNase I protection assays. DNA binding was first compared with a single element versus three tandem copies of the consensus HRE (Fig. 4). FL AR in 5 µg of Sf9 whole cell extract completely protected 3xHRE-3 from DNase I cleavage, but only protected a single HRE to one-fifth that level. Full protection of single HREs was obtained with 20 µg of extract. This cooperativity extends observations of footprints obtained with just the DBD on the probasin promoter (39) or with FL AR or N AR gel-shifts of HREs (36).

View larger version (79K): [in this window] [in a new window]   Fig. 4.   Cooperative DNA binding by full-length AR and AR deletion mutants on consensus HREs. Whole cell extracts from Sf9 cells infected with baculovirus expressing full-length rAR (AR FL), AR-N46-408 (N), or AR-C788-902 (C) were used in DNase I protection assays with the indicated extract protein amounts. DNA with a single copy (1x) or multiple copies (3x) of the consensus HRE-3 from the enhancer was used as a probe. The positions of the HREs are indicated with arrows. The percent protection of the HRE relative to digestion of naked DNA (0) is shown near the lanes with lower extract concentrations, and was derived from selected band intensities measured on a Molecular Diagnostics PhosphoImager, normalized within the lane to a band unaffected by the addition of extract.

View larger version (63K): [in this window] [in a new window]   Fig. 5.   Cooperativity conferred by binding to HRE-1 is diminished in the absence of the N- or C-terminal receptor domains. A, whole cell extract containing full-length rAR was used in DNase I protection on the androgen-specific enhancer C'9, as well as mutants altered in either the weak HRE-1 (ls7), or the strong HRE-3 (ls9). The left lane shows the protection pattern of C'9 without protein (0); positions of HRE-1 and -3 are indicated. The middle panel shows the maximal protection attained with 20 µg of extract. Lanes to the right include percent protection attained for each site with 5 µg of extract when the second site is mutated, in comparison to the center lane with the wild type probe. B, protection of C'9 by N and C at high protein concentrations is shown in comparison to FL AR, revealing poorer protection, especially of HRE-1 by the deleted receptors by the percent protection for each site relative to naked DNA to the right of the respective lane.

AR N-terminal Subdomains Can Convert the GR N Terminus to an Androgen-specific Activator-- The above experiments suggest that activation of the androgen-specific enhancer by AR relied partly on utilization of multiple HREs, accentuated by interactions involving the AR N terminus and the LBD, and partly on a permissive effect of the AR DBD. To confirm this, we sought to convert the chimeric receptor GAA to an activator of C'9, by adding regions of the AR N terminus that might promote physical association with the C terminus but would not add transactivation functions (17, 18, 20, 22, 40). We tested the first 37 amino acids of AR, fused to the translation start of GAA (A³⁷GAA), and AR sequences just upstream of the DBD, fusing the GR N terminus through residue 391 to AR residue 392 (G³⁹¹aAA). These mutants increased activation of C'9 at least 2-fold over activation by GAA, which was proportionally greater than their increased activation of simple HREs (Fig. 6). These mutant receptors also showed increased sensitivity to mutation of HRE-1 (not shown). A double mutant with both the AR N-terminal addition and the region upstream of the DBD did not show an additive increase in activation (not shown). This modulation of GAA activity by AR N-terminal subdomains, including increased sensitivity to HRE-1, is further evidence that the activation of the specific enhancer, as well as general AR activity, is dependent on interactions between N and C termini.

View larger version (10K): [in this window] [in a new window]   Fig. 6.   Subdomains from the AR N terminus relieve the N-terminal block of GR to transactivation. GAA and variants that included subdomains from the AR N terminus were tested for activation of the specific enhancer C'9, using calcium phosphate to introduce the DNAs into CV-1 cells. CAT activity is expressed relative to that of AR, ± S.E. The receptors are depicted to the left, to indicate N-terminal AR subdomains (black) added to that of GR (white) in GAA. To the right is shown the ratio of CAT attained for each receptor from the reporter driven by C'9 compared with driven by 3xHRE-3, relative to the AR value set to 1.0 (actual value was 0.88 in these experiments).

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

In this study, we have discerned innate features of AR that are involved in functional distinctions in activity from GR. For AR, specific gene activation is problematic since its DNA sequence recognition overlaps with that of GR, which is more widely and abundantly expressed. Therefore androgen specific response mechanisms must emphasize AR activation while minimizing participation by GR. For steroid receptors, like many other transcription factors, a major determinant of specificity arises from interaction with additional DNA binding proteins. For example, precise control by the Drosophila homeoprotein ultrabithorax is attained by its interaction with extradenticle (41, 42). Both in vitro and in vivo, the androgen specificity of the Slp enhancer relies on nonreceptor factors (15, 43). These cooperative interactions may not only restrict DNA binding specificity, but may also allow selective binding to natural sites that are of low affinity. Cooperativity that refines specificity may occur between multimers of the same protein, via domains not directly involved in DNA binding, as shown for STAT4 (44). Results obtained here reveal the importance of AR cooperativity, with itself and with other factors, in accentuating the AR/GR differential response.

The differential ability of AR-GR chimeric receptors to activate the androgen-specific Slp enhancer C'9 reveal that each of the three major domains of the receptor plays a distinct role in specificity. Further, the activity of each domain is modulated by the presence of the others, implying complex interactions within receptor dimers, and between them. Two receptor-intrinsic features in particular enforce the AR specific response. First, the AR DBD evades a context-dependent suppression that operates via the GR DBD. Second, and of more general significance, the presence of both AR N and C termini, whose interaction is well established (16-20), enhances cooperative DNA binding. This allows AR to augment activation by utilizing weak HREs that are ignored by GR.

The domain common to all receptors that activate the androgen-specific enhancer (AR, AAG, GAG) is the AR DBD. This could reflect simply an inability of GR to bind C'9 sequences, or, alternatively, suppressive effects on the DBDs of non-androgen receptors in the C'9 context. Whichever mechanism operates, it is strongly influenced by nonreceptor factors, since this same enhancer is glucocorticoid-responsive in T47D cells (13). GR binds both HRE-3 and the HRE-1 weak site in DNase I footprinting assays, despite HRE-3 having low relative affinity for GR in competitive gel shift analysis (13, 15). Nevertheless, receptors with the GR DBD activate HRE-3 well in other contexts, such as when the site is moved 10 bp distal. Therefore GR can bind HRE-3, but in C'9, binding, or function once bound, is dependent on adjacent sequences and the cohort of cellular factors. The effects of other sequences and factors is likely to influence how the N terminus and LBD constrain the constitutive activity shown by the DBD alone, for both AR and GR (18, 38, 45).

After evading the context-dependent suppression operating on other receptors via the DBD, AR activation is greatly enhanced by cooperation from multiple response elements. The importance of the AR N-C domain interaction in cooperativity is underscored by comparing the behaviors of the chimeras GAG and AGA. GAG activates the specific enhancer, having escaped suppression via the DBD, but does not require the weak HRE, as demonstrated by its insensitivity to ls7 mutations. In contrast, the chimera AGA, which functions when DBD repression is abrogated on the s10 enhancer, gains greatly in activity when HRE-1 is available. In fact, the extreme sensitivity of AGA to ls7 may indicate that the AR DBD has a more direct influence on the transcriptional response, rather than a simply permissive role, perhaps by affecting efficacy of the N-C interaction.

A notable exception to the permissive effect of the AR DBD is shown by GAA, whose inactivity on C'9 demonstrates that the AR DBD is a necessary, but insufficient, determinant of specificity. Importantly, this receptor can be partially cured of its dysfunction by inclusion of AR N-terminal sequences that allow association with the AR LBD (17), or by removing the AR LBD entirely (Fig. 6). This implies that the AR LBD restrains activity of the GR N terminus, in a manner similar to the transdominant repression imposed by ARN against full-length AR (24). This inhibitory effect occurs with the AR but not the GR LBD, since AAG activates C'9 efficiently. Along with the N terminus, the AR LBD itself aids recognition of, or stabilizes binding to, weak HREs (Fig. 5). Thus when N-terminal AR sequences are added to GAA, in addition to circumventing LBD repressive effects, they confer an enhanced ability to utilize HRE-1.

The precise physical role of the N terminus and LBD in cooperative receptor interactions is not yet clear. However, absence of either domain lessens cooperative binding to multimerized consensus sites and is particularly detrimental to binding to HRE-1 (Figs. 4 and 5). The importance of the LBD is apparent in poor cooperativity of C even on consensus elements. While N shows moderate cooperativity on multimerized HREs, it binds HRE-1 poorly, like C. This suggests that the N terminus aids use of weak HREs, at either the level of sequence recognition or of binding stability. Communication between the N and C termini during the process of DNA binding may occur via conformational changes in the DNA-bound DBD, by direct contacts within the dimer, or through contacts between dimers on adjacent HREs. In contrast to intra-receptor interactions of ROR (46), the major transactivation region of the AR N terminus does not appear to be involved in HRE discrimination, however sensitive it might be to hormone occupation of the LBD. It is appealing to consider that modes of communication already well characterized within the AR dimer (17, 18, 20) may also exist between dimers.

Activation reliant on multiple and often nonconsensus HREs appears to be a hallmark of androgen-dependent enhancers. Conforming to this mode are the prostate-specific probasin promoter, which like the Slp enhancer retains androgen specificity in vitro (11), and complex regulatory elements from genes for prostate specific antigen (23), prostate C3 (47), and mouse aldose reductase (48). In contrast, other androgen-dependent genes manifest only single or less obvious HREs in regions required for hormonal response (12, 49). As a consequence, nonreceptor factors are crucial to regulation of these latter promoters. Even on promoters with multiple HREs, however, AR requires additional factors to generate a specific response, both to increase intrinsic activation and to exclude other receptors. Multiple weak binding sites are commonly exploited by nuclear receptors (50-52), but AR in particular seems to rely on this function. Our data suggests that the strong interaction between the N terminus of AR and its LBD underlies this heightened cooperativity. Coupled with the ability of the AR DBD to evade context-dependent suppression, these intrinsic features of receptor work in concert with nonreceptor factors to refine the differential between AR and GR response. In this manner, precise specificity can be attained by conjoining several mechanisms, no one of which by itself is stringently specific.