Differential mechanisms of nuclear receptor regulation by receptor-associated coactivator 3.

Steroid and nuclear receptor coactivators (NCoAs) have been implicated in the regulation of nuclear receptor function by enhancing ligand-dependent transcriptional activation of target gene expression. We have previously isolated receptor-associated coactivator 3 (RAC3), which belongs to the steroid receptor coactivator family. In this study, we investigated the differential mechanisms by which RAC3 interacts with and modulates the transcriptional activity of different nuclear receptors. We found that the vitamin D receptor (VDR) and estrogen receptor beta interact with different alpha-helical LXXLL motifs of RAC3. Peptides corresponding to these motifs have diverse affinities for the VDR and estrogen receptor beta, and mutation of specific motifs differentially impairs the ability of RAC3 to interact with these receptors in vitro. Consequently, these mutations inhibit the enhancement of transcriptional activation by these receptors in vivo. Furthermore, we found that the activation function-2 (AF-2) domain of the retinoid X receptor interferes with RAC3 binding to a DNA-bound VDR/retinoid X receptor (RXR) heterodimer, whereas the VDR AF-2 domain is required for this interaction. These results suggest a receptor-specific binding preference for the different LXXLL motifs of RAC3, which may provide flexibility for RAC3 to differentially regulate the function of different nuclear receptors.

Steroid and nuclear receptor coactivators (NCoAs) have been implicated in the regulation of nuclear receptor function by enhancing ligand-dependent transcriptional activation of target gene expression. We have previously isolated receptor-associated coactivator 3 (RAC3), which belongs to the steroid receptor coactivator family. In this study, we investigated the differential mechanisms by which RAC3 interacts with and modulates the transcriptional activity of different nuclear receptors. We found that the vitamin D receptor (VDR) and estrogen receptor ␤ interact with different ␣-helical LXXLL motifs of RAC3. Peptides corresponding to these motifs have diverse affinities for the VDR and estrogen receptor ␤, and mutation of specific motifs differentially impairs the ability of RAC3 to interact with these receptors in vitro. Consequently, these mutations inhibit the enhancement of transcriptional activation by these receptors in vivo. Furthermore, we found that the activation function-2 (AF-2) domain of the retinoid X receptor interferes with RAC3 binding to a DNA-bound VDR/retinoid X receptor (RXR) heterodimer, whereas the VDR AF-2 domain is required for this interaction. These results suggest a receptor-specific binding preference for the different LXXLL motifs of RAC3, which may provide flexibility for RAC3 to differentially regulate the function of different nuclear receptors.
The vitamin D receptor (VDR) 1 and estrogen receptor ␤ (ER␤) belong to the steroid/thyroid hormone receptor superfamily, which is a large class of ligand-dependent transcription factors that plays critical roles in regulating genes involved in a wide array of biological processes, including development and homeostasis (1). This superfamily can be divided into three subgroups. The ER␤ is a Type I receptor, which also includes receptors for steroids such as progestins, androgens, glucocorticoids, and mineralcorticoids. These receptors are coupled to heat shock proteins and sequestered to the cytoplasm in the absence of ligand. Upon hormone binding, they dissociate from the heat shock proteins, homodimerize, and translocate to the nucleus where they bind to cognate response elements consisting of palindromic repeats. The VDR is a Type II receptor like those for thyroid hormone (TR) and all-trans retinoic acid (RAR). These receptors are strictly nuclear and form heterodimers with the receptor for 9-cis retinoic acid (RXR). They also bind constitutively to response elements consisting of direct repeats (DRs). A third class of nuclear receptors is the orphan receptors, so-called because endogenous ligands for these proteins are currently unknown.
Most members of the nuclear receptor superfamily share a common domain structure. The N terminus contains the variable A/B region, which also includes the ligand-independent AF-1 activation domain. The highly conserved DNA binding domain and the C-terminal ligand binding domain (LBD) follow this region. The LBD contains the ligand-dependent AF-2 activation domain and also mediates dimerization of nuclear receptors. In the absence of ligand, nuclear receptors are able to repress basal transcription via functional interactions with the nuclear receptor corepressors SMRT and NCoR (2,3). SMRT and NCoR are found in complexes with the corepressor mSin3 and the histone deacetylase HDAC1, suggesting that transcriptional repression by nuclear receptors may involve histone deacetylation (4 -6). Ligand binding triggers the release of these corepressors and subsequent recruitment of coactivators through a drastic conformational change in the AF-2 domain of the receptor. Structural studies have demonstrated that helix 12, which contains the AF-2 domain, projects away from the LBD in the unbound RXR structure, but rotates nearly 180°to pack tightly against the LBD upon hormone binding in the RAR, TR, and ER (7)(8)(9)(10). This conformational change, together with induced changes in helices 3-5, is believed to facilitate interactions of the receptor with coactivators (11)(12)(13)(14)(15)(16).
Coactivators recruited by ligand-bound nuclear receptors include members of the SRC family of coactivators such as SRC1 (also known as NCoA-1), TIF2/GRIP1 (also known as SRC2 or NCoA-2), and RAC3/ACTR/pCIP/AIB1 (also known as SRC3 or NCoA-3) (reviewed in Refs. 17 and 18). SRC family members share an N-terminal basic helix-loop-helix/PAS-A/PAS-B domain of unknown function, centrally located receptor interaction domain, and C-terminal transcriptional activation domain. These cofactors interact with receptors in a hormone-and AF-2-dependent manner and enhance transcriptional activation by nuclear receptors. Both coactivators and receptors also have been demonstrated to interact with the general transcriptional activators CBP/p300 and PCAF (19 -26), suggesting that a large multi-protein complex is assembled at the target gene promoter to activate transcription. Furthermore, several coactivators, including SRC1, ACTR, PCAF, and CBP/p300, possess intrinsic histone acetylation activity, which disrupts nucleosomes (21,(27)(28)(29)(30). Therefore, the mechanism by which nuclear receptors activate transcription may entail the recruitment of a coactivator complex via the AF-2 domain that can modify chromatin structure, thereby facilitating access to the promoter by the general transcription machinery.
Intriguingly, members of the SRC family of coactivators have been found to contain several conserved motifs, termed NR boxes, with the consensus sequence LXXLL, where X is any amino acid (31). Motifs within the receptor-interacting domain and transcriptional activation domains of SRC1 and TIF2 have been demonstrated to mediate interactions with liganded nuclear receptors and CBP/p300, respectively (23,32). Crystallographic and protein structure prediction analyses have indicated that these motifs form amphipathic ␣-helices with the leucine residues comprising a hydrophobic surface on one face of the helix (11,12,14,24). The helical motif is able to interact with the AF-2 domain of the liganded receptor via a hydrophobic groove made up of residues from receptor helices 3, 4, 5, and 12 that is the result of the conformational change induced by hormone binding (11,14,16). Mutational analyses of the NR boxes of SRC1 and TIF2/GRIP1 have also uncovered a receptorspecific code of interaction, where different nuclear receptors require different NR boxes to interact with the coactivator (32)(33)(34). These studies indicate that flanking residues outside the NR box may also be important to nuclear receptor-coactivator interactions.
In this study, we investigate the mechanisms by which RAC3 regulates the function of the VDR and ER␤, for little is known concerning the regulation of these receptors by SRC coactivators, particularly RAC3. These analyses reveal receptor-specific interactions in which the VDR and ER␤ interact with different surfaces of RAC3. We demonstrate different preferences of these receptors for specific NR boxes of RAC3 and that single mutations in these LXXLL motifs are able to severely impair the ability of RAC3 to interact with and, thus, coactivate the VDR and ER␤. In analyzing the requirement of nuclear receptor AF-2 domains, we observe that the AF-2 domain of RXR can inhibit RAC3-RID binding to the DNA-bound VDR/ RXR heterodimer, whereas the AF-2 domain of VDR is required for this interaction. These data add a new level of complexity to the regulation of nuclear receptor activity by SRC coactivators and suggest that different classes of nuclear receptors may be regulated by RAC3 via different mechanisms.

EXPERIMENTAL PROCEDURES
Far Western Analysis-Far Western assays were carried out as described (20). Briefly, GST fusion proteins were expressed in DH5␣ cells and purified with glutathione-agarose beads (Amersham Pharmacia Biotech). Purified proteins were then separated by SDS-polyacrylamide gel electrophoresis and electroblotted onto a nitrocellulose membrane. Proteins were denatured with 6 M guanidine hydrochloride and renatured by the stepwise dilution of guanidine hydrochloride. Membranes were then blocked and hybridized overnight with 35 S-labeled protein.
The membrane was washed, and bound probe was detected by autoradiography. 35 S-Labeled probes were generated by Quick-coupled in vitro transcription/translation (Promega). For peptide competition experiments, the given concentration of peptide was added to the probe 10 min before hybridization with the membrane. Peptide sequences were as follows: NR box i peptide (LESKGHKKLLQLLTLSSDDRGHSSL), NR box ii peptide (LQEKHRILHKLLQNGNSP), NR box iii peptide (KKKENNALLRYLLDRDD), control peptide (GSGSATATLYENKPRP-PYIL). Radioactive bands were quantified by PhosphorImager using the ImageQuant software (Molecular Dynamics).
GST Pull-down Assay-Approximately 5 g of purified GST fusion protein was incubated with 5 l of 35 S-labeled protein with moderate shaking at 4°C overnight in binding buffer (20 mM HEPES, pH 7.7, 75 mM KCl, 0.1 mM EDTA, 2.5 mM MgCl 2 , 0.05% Nonidet P-40, 1 mM dithiothreitol, 1 mg/ml BSA). The bound protein was washed three times with binding buffer, and beads were collected by centrifugation. The bound protein was eluted in SDS sample buffer, subjected to SDS-polyacrylamide gel electrophoresis, and detected by autoradiography.
Site-directed Mutagenesis-NR box mutants were generated with the Quick-change site-directed mutagenesis system (Stratagene). The sequences of all mutant constructs were confirmed by dideoxynucleotide chain termination reactions using the T7 Sequenase protocol (U. S. Biochemical Corp.).
Gel Electrophoresis Mobility Shift Assay-The sequence of the DR3 element used for VDR/RXR gel-shift assays is AGCTTAAGAGGTCA-GAAAGGTCACTCGCAT. The double-stranded DR3 was end-labeled with [ 32 P]dCTP by standard Klenow fill-in reaction. The purified probe was incubated with 35 S-labeled receptors in binding buffer containing 7.5% glycerol, 20 mM HEPES, pH 7.5, 2 mM dithiothreitol, 0.1% Nonidet P-40, 1 g of poly(dI-dC) and 100 mM KCl. Wild-type or mutant GST-RAC3-RID was eluted from glutathione-agarose beads with 10 mM reduced glutathione and added to the binding reaction. The DNAprotein complex was formed on ice for 1 h and resolved on a 5% native polyacrylamide gel, which was subsequently dried and subjected to autoradiography.
Cell Culture and Transient Transfection-HEK293 and CV-1 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 5 g/l gentamycin at 37°C, 5% CO 2 . Cells were plated for transfection in Dulbecco's modified Eagle's medium supplemented with 10% resin charcoal-stripped fetal bovine serum in 12-or 6-well plates 1 day before transfection. HEK293 cells were transfected using the standard calcium phosphate method, whereas CV-1 cells were transfected using LipofectAMINE according to the manufacturer's protocol (Life Technologies, Inc.) Twelve hours after transfection, cells were washed with phosphate-buffered saline and refed fresh medium containing the indicated concentration of ligand. After 24 h, cells were harvested for ␤-galactosidase and luciferase activities as described (35). Luciferase activity was determined with a MLX plate luminometer (Dynex) and normalized relative to ␤-galactosidase activity.

VDR and ER␤ Interact with Multiple Surfaces of RAC3-We
have previously defined the minimal receptor interacting domain (RID) of RAC3 to be amino acids 613-752, which contains the first three LXXLL motifs ( Fig. 1A) (20). We wished to further determine if different receptors were capable of binding to the same regions of RAC3. To accomplish this, we purified a panel of GST-RAC3 fusion proteins, in total comprising the full-length RAC3 (Fig. 1A), and probed these fusions with [ 35 S]methionine-labeled VDR and ER␤ in a Far Western assay. The VDR, as expected, interacted in a ligand-dependent manner with GST-RAC3 613-752 in this assay (Fig. 1B). It also bound GST-RAC3 723-1034, which only contained NR box iii, in a ligand-dependent manner. The VDR did not interact with any other GST-RAC3 fragment, including GST-RAC3 342-646, which contained NR box i. It also appeared that the VDR interacted more strongly with GST-RAC3 723-1034 than with GST-RAC3 613-752, suggesting a more important role for NR box iii in the RAC3-VDR interaction. However, a different pattern was evident upon repeating this assay with 35 S-ER␤, for in addition to ligand-dependent interactions with GST-RAC3 613-752 and 723-1034, ER␤ also bound the 342-646 fragment, which contained only NR box i (Fig. 1C). These interactions were of approximately equal intensity. There was also a weak, ligand-independent interaction with GST-RAC3 1-407. Identical results were obtained for ER␣ (data not shown). A Coomassie Blue-stained polyacrylamide gel of the GST-RAC3 fusion protein confirmed the identity of each GST-RAC3 fusion protein and approximately equal protein concentrations in each lane (data not shown). Thus, the VDR and ER␤ display different binding patterns for RAC3 fragments, with the VDR interacting preferentially with regions containing NR box iii and the ER␤ interacting equally well with regions containing any of the three NR boxes and the N-terminal basic helix-loop-helix-PAS domain.
NR Box Peptides Differentially Compete with Nuclear Receptors for RAC3 Binding-We then wanted to investigate the relative importance of individual NR boxes within the RAC3-RID in mediating the interactions between RAC3 and the VDR or ER␤. Peptides were synthesized corresponding to NR boxes i, ii, and iii, which were incubated with 35 S-labeled receptor and 1 M ligand before probing the GST-RAC3 613-752 fragment in the Far Western assay. With the VDR, peptides corresponding to the second and third LXXLL motif were able to compete away the RAC3-RID interaction with VDR in a dosedependent manner ( Fig. 2A). Upon quantifying the data, it was evident that peptide iii was a more potent inhibitor than peptide ii, whereas the peptide comprising NR box i had little, if any, effect on the VDR-RID interaction (Fig. 2B). A control experiment demonstrated that the effects of these peptides were specific, for a random peptide did not alter the interaction between the 35 S-VDR and GST-RAC3-RID (Fig. 2C). We again identified a different pattern when the same experiment was done with the ER␤ (Fig. 2D). Here, all three peptides were able to compete efficiently for ER␤ binding with the RAC3-RID, with peptide ii being the most potent (Fig. 2E). Thus, these data reveal receptor-specific preferences for interactions between RAC3 and different nuclear receptors, for the VDR and ER␤ have different affinities for the NR boxes of RAC3.
NR Box Mutations Can Impair RAC3 Interactions with Nuclear Receptors in Vitro-To assess the integrity of the LXXLL motif in mediating the interaction between the RAC3-RID and nuclear receptors, we used site-directed mutagenesis to switch the leucine residues of each motif to alanines (Fig. 3A). The mutants were made using the GST-RAC3-RID fusion as the template and tested for their ability to interact with the VDR or ER␤ in GST pull-down assays. The wild-type RAC3-RID was able to pull down a significant amount of 35 S-VDR in the presence of 1 M vitamin D (Fig. 3B). This interaction was specific, for GST alone pulled down much less 35 S-VDR. Mutations in NR boxes i or ii displayed wild-type binding. However, when NR box iii was mutated, the RID-VDR interaction was greatly reduced to a level only slightly higher than background binding to GST alone. GST-RAC3 342-646, which contained only NR box i, also had minimal binding to the VDR, consistent with the Far Western assay (Fig. 1B). Equal protein concentrations of each GST fusion confirmed the specificity of these findings. Thus, these data support the above observations in implicating NR box iii as being most critical to RAC3 interaction with the VDR.
The wild-type RID was also able to pull down significant amounts of 35 S-ER␤ in the GST pull-down assay (Fig. 3C). In contrast to the VDR, an alanine substitution for leucine in any of the three NR boxes weakened the interaction of the RAC3-RID with ER␤, with the mutation of NR box ii being the most deleterious, again supporting the results of LXXLL peptide competition experiments (Fig. 2B). However, with each mutation, significant binding above background between the RAC3-RID and 35 S-ER␤ was still observed. Furthermore, the GST-RAC3 342-646 fragment, with only NR box i, was able to interact efficiently with ER␤ (Fig. 3C) as in the Far Western assay (Fig. 1C). These data suggest that although all three motifs are capable of interacting with ER␤ separately, none of them is absolutely required for the interaction. In contrast, NR box iii of RAC3 appears to be essential for the interaction with the VDR.
RAC3-RID Interactions with DNA-bound Nuclear Receptors-The above data provide compelling evidence that the VDR interacts specifically with RAC3 in solution via the LXXLL motifs of the RAC3 RID, particularly NR box iii. To gain further insight into the function of NR boxes in coactivator-VDR interactions on a heterodimeric complex bound to DNA, we performed gel-shift assays with VDR/RXR heterodimers on a DR3 element in the presence of wild-type or mutant RAC3-RID (Fig. 4). The VDR/RXR heterodimer bound strongly to the 32 P-labeled DR3 probe and was unaffected by the addition of GST alone (Fig. 4A, lanes 1 and 2). The addition of the RAC3-RID resulted in a ligand-dependent shift of the

FIG. 2. Peptides corresponding to the NR boxes of the RAC3-RID can compete for VDR and ER␤ binding with the RAC3-RID.
A, Far Western assay using 35 S-VDR to probe GST-RAC3-RID in the presence of 1 M vitamin D and given concentration of each peptide. B, the data from A was quantified by PhosphorImager and plotted as percent GST-RAC3-RID binding to 35 S-VDR versus peptide concentration. 100% RID binding represents the density of the band in the absence of peptide. ࡗ, peptide I; f, peptide ii; OE, peptide iii. C, Far Western assay demonstrating that inhibition of 35 S-VDR interaction with GST-RAC3-RID by peptide iii is specific. Peptide iii abolishes nearly all the interaction, whereas a control, random peptide has no effect. D, Far Western assay using 35 S-ER␤ to probe GST-RAC3-RID in the presence of 1 M estradiol and given concentration of each peptide. E, the data from D was quantified by PhosphorImager and plotted as the percent GST-RAC3-RID binding to 35 S-ER␤ versus peptide concentration. ࡗ, peptide I; f, peptide ii; OE, peptide iii. heterodimeric complex to a slower-migrating form (arrow, lanes 3 and 4). The mutation of NR box i had little effect on the ability of the RID to shift the complex (lane 5). However, mutating NR box ii diminished the formation of the RID-VDR/ RXR complex, whereas mutating NR box iii nearly abolished the formation completely (lanes 6 and 7). Consistently, GST-RAC3 342-646, with only NR box i, was unable to shift the VDR/RXR complex (lane 8). These results differ slightly from the GST pull down, in which only the NR box iii mutation inhibited interaction with VDR alone (Fig. 3B), suggesting that both motifs ii and iii may contribute to the interaction with DNA-bound VDR/RXR heterodimer.
We next analyzed the involvement of nuclear receptor AF-2 domains in regulating the interaction between the RAC3-RID and the VDR/RXR heterodimer (Fig. 4B). Using the gel-shift assay, we compared the ability of the RID to bind the wild-type heterodimer versus the VDR/RXR443 and VDR402/RXR heterodimers, in which the AF-2 domain of RXR or VDR had been deleted, respectively. As demonstrated above, the RID was able to bind the DNA-bound, wild-type VDR/RXR complex (Fig. 4B,  lane 2). Deletion of the VDR AF-2 domain did not affect the formation of the heterodimer-DNA complex but resulted in the loss of the RID-shifted complex (lane 4). This suggests that the VDR AF-2 domain is required for interaction of RAC3 with the heterodimer and that RXR AF-2 domain alone is not sufficient for the interaction. Interestingly, deletion of the RXR AF-2 domain resulted in a much stronger shift of the heterodimeric complex by the RAC3-RID (lane 6) without affecting heterodimer formation (lane 5), suggesting that the RXR AF-2 domain can inhibit the interaction between RAC3 and VDR/ RXR. The strong interaction was abolished upon deletion of the VDR AF-2 domain (lane 8), further supporting a requirement of VDR AF-2 helix 12 for RAC3 binding to the DNA-bound heterodimer.
Finally, we compared the RAC3 NR box preferences of VDR/ RXR versus VDR/RXR443 (Fig. 4C). Intriguingly, the VDR/ RXR443 heterodimer displayed different NR box preferences. Mutation of NR box i or iii greatly reduced the shift by the RAC3-RID (Fig. 4C, lanes 3 and 5), whereas mutation of NR box ii only slightly weakened the binding (lane 4). Thus deletion of the RXR AF-2 domain resulted in a switch in the NR box requirements, with NR boxes i and iii being most important for VDR/RXR443 compared with NR boxes ii and iii for VDR/RXR. An autoradiograph confirmed equal expression levels of each 35 S-labeled receptor (Fig. 4D). This finding supports the hypothesis that multiple LXXLL motifs provide RAC3 with the flexibility to adapt to different configurations of a nuclear receptor dimer.
Effects of NR Box Mutations on RAC3 Coactivation Function in Vivo-RAC3 has previously been shown to enhance the transcriptional activity of the RAR and progesterone receptor (37). However, its effect on VDR and ER␤ function in vivo has not been demonstrated. To address this, we performed transient transfection assays in HEK293 and CV-1 cells using luciferase reporters harboring either two copies of the vitamin D response element of the osteopontin gene for VDR studies or a consensus ERE element for ER␤ studies (Fig. 5). Transfection of the VDR into CV-1 cells minimally activated the vitamin D response element driven reporter (Fig. 5A). However, treating

FIG. 4. Mutation of RAC3 NR boxes inhibits RAC3-RID binding to DNA-bound VDR/RXR heterodimer.
A, gel-shift assay of the effects of NR box mutations (mut) on RAC3-RID binding to DR3-bound VDR/RXR. 1.5 l of each 35 S-labeled nuclear receptor was added to a binding reaction (see "Experimental Procedures") containing 1 M vitamin D, equal amounts of the indicated GST fusion protein, and the [ 32 P]dCTP-labeled DR3 probe. The arrow indicates the RID-receptor complex. wt, wild type. B, the RXR AF-2 domain can interfere with RID binding to VDR/RXR, whereas the VDR AF-2 domain is required for the interaction. The gel-shift assay was performed as in A, except the AF-2-truncated RXR443 or VDR402 was used where indicated. *, nonspecific band from lysate. C, the VDR/RXR443 heterodimer has different NR box preferences than the wild-type receptor heterodimer. The gel-shift assay was performed as in A. *, nonspecific band from lysate. D, autoradiograph confirming the equal expression of the 35 S-labeled receptors used in B and C. these cells with vitamin D strongly stimulated its activity. Cotransfection of RAC3 further enhanced VDR transcriptional activation by approximately 50%, consistent with the coactivation function of RAC3 (37).
We then analyzed the role of the NR boxes in mediating the ability of RAC3 to potentiate VDR activity. Mutations of each NR box of the RAC3-RID were made in the context of the full-length RAC3 protein and tested for their ability to coactivate the VDR in transient transfection assays (Fig. 5A). Mutation of NR box i did not inhibit RAC3 enhancement of VDR transactivation, consistent with its inability to block the interaction of RAC3 with VDR in vitro. However, the RAC3-NR box ii or iii mutations reduced the function of RAC3 in enhancing VDR activity (Fig. 5A), consistent with their requirement for RAC3 binding to the DNA-bound VDR/RXR heterodimer in gel-shift assays (Fig. 4A). Taken together, it is clear that NR boxes ii and iii are both involved in RAC3 regulation of VDR function, whereas the role of NR box i appears minimal.
We then repeated these experiments with the ER␤ (Fig. 5B). Estradiol treatment of CV-1 cells transfected with the ER␤ and the ERE-luciferase reporter activated reporter expression approximately 8-fold. Cotransfection of wild-type RAC3 resulted in a strong enhancement of ER␤ activity. It is evident from these data that RAC3 is a more potent coactivator for the ER␤ than for the VDR. Cotransfection of RAC3 expression plasmids containing mutations in NR boxes i, ii, or iii all suppressed the ability of RAC3 to coactivate the ER␤, with the NR box ii mutant having the greatest and NR box iii mutant having more modest effects on RAC3 function. Thus, these in vivo data correlate with the in vitro data in implicating all three NR boxes of the RAC3-RID as being important for RAC3 regulation of ER␤ function.
Finally, we decided to assess the functional consequences of the antagonism of RAC3 interaction with DNA-bound VDR/ RXR by the RXR AF-2 domain observed in gel-shift assays (Fig.  4B). To do this, we compared the ability of RAC3 to coactivate VDR/RXR and VDR/RXR443 activities by transient transfection assays (Fig. 5C). When wild-type VDR and RXR were expressed in HEK293 cells, RAC3 was able to enhance transcriptional activation by approximately 50% (Fig. 5C, left). However, when VDR was coexpressed with RXR443, RAC3 displayed a 2.5-fold enhancement of receptor activity (Fig. 5C,  center). As expected, the VDR402 mutant was transcriptionally inactive, and RAC3 could not modulate its activity (Fig. 5C,  right). Thus, these in vivo data are consistent with the gel-shift data in demonstrating that the RXR AF-2 domain can inhibit RAC3 modulation of the VDR/RXR heterodimer, whereas the VDR AF-2 domain is absolutely required for this regulation.

DISCUSSION
In this study, we have investigated the role of the NR boxes of RAC3 in mediating the ability of this coactivator to bind and coactivate the VDR and ER␤. We found that NR box iii is most critical to VDR binding, whereas NR boxes i, ii, and iii are involved in ER␤ interaction. Peptides corresponding to these respective motifs were able to compete with the RAC3-RID for VDR and ER␤ binding. The integrity of the motifs themselves was also important, for mutations in specific NR boxes inhibited RAC3 interaction with these receptors in solution and when bound to DNA. The AF-2 domain of VDR is required for binding of the RAC3-RID to DNA-bound VDR/RXR, whereas the AF-2 domain of RXR was able to antagonize this interaction. Removal of this inhibitory AF-2 helix of RXR enhances ligand-dependent binding of RAC3 to the VDR/RXR443 heterodimer and alters the NR box requirements. Furthermore, the mutation of NR box ii or iii blocked the ability of RAC3 to enhance transcriptional activation by the VDR in vivo. In contrast, mutation of NR boxes i, ii, or iii reduced RAC3 coactivation of ER␤. Together, these in vitro and in vivo studies suggest a mechanistic difference in the manner by which RAC3 regulates VDR and ER␤ activities.
The NR boxes are highly conserved among the SRC family of coactivators (18). Our data and that of others clearly reveal that multiple motifs are necessary for high affinity interactions with nuclear receptors (23,32,33). With the DNA-bound VDR/ RXR heterodimer, we found that mutation of NR boxes ii or iii of RAC3 weakens the interaction with the RAC3-RID. In contrast, NR box iii of RAC3 is the only critical motif for interaction with VDR in solution. Therefore, it is likely that each motif binds to each monomer of the receptor heterodimer, consistent with the structure of a peroxisome proliferator-activated receptor ␥-LBD dimer co-crystallized with a fragment of SRC1 containing two NR boxes (12). Our data on ER␤ suggest that all three NR boxes of the RAC3-RID are involved for a wild-type interaction, whereas the presence of two motifs is sufficient for a strong interaction, motif ii being most important. In light of this finding, RAC3 may utilize motif ii in combination with motif i or iii for efficient interaction with the ER␤ homodimer. The integrity of the other motif may be critical to the overall conformation of the coactivator or potentially make an additional contact with another region of the receptor. Support for the latter possibility can be found in recent studies detailing the enhancement of the N-terminal AF-1 activation function of nuclear receptors by SRC coactivators (38,39). The presence of multiple NR boxes also likely provides coactivators the flexi-

FIG. 5. Mutation of RAC3 NR boxes blocks coactivation of VDR and ER␤ activity in vivo.
A, mutation (mut) of NR box ii or iii inhibits RAC3 enhancement of VDR activity. CV-1 cells were transfected with expression plasmids for VDR, wild-type (wt), or mutant RAC3 and the Sppx2-luciferase reporter and treated with 40 nM vitamin D for 24 h where indicated. Upon harvesting cells, luciferase activity was measured and normalized to ␤-galactosidase activity. B, mutation of NR box i, ii, or iii inhibits RAC3 enhancement of ER␤ activity. CV-1 cells were transfected with expression plasmids for ER␤, wild-type, or mutant RAC3 and the ERE-luciferase reporter and treated with 10 nM estradiol for 24 h where indicated. Upon harvesting cells, luciferase activity was measured and normalized to ␤-galactosidase activity. C, deletion of the RXR AF-2 domain results in enhanced coactivation of VDR/RXR activity by RAC3. HEK293 cells were transfected with the appropriate nuclear receptor expression plasmids along with expression plasmids for RAC3 and Sppx2-luciferase and treated with 10 nM vitamin D where indicated. Upon harvesting cells, luciferase activity was measured and normalized to ␤-galactosidase activity. bility to interact with a broad range of nuclear receptors, resulting in the different preferences that are observed between nuclear receptors and distinct motifs, depending on the precise structural nuances of each receptor-coactivator interface. This is evident upon comparing the NR box requirements of the VDR/RXR heterodimer versus those of VDR/RXR443. Deletion of the AF-2 helix of RXR not only enhances RAC3-RID binding to the heterodimer but also switches the NR box preferences from motifs ii/iii to motifs i/iii. Finally, amino acids flanking the NR boxes also likely contribute to the specificity of interaction (15,32), for, despite the high homology between the RAC3 NR boxes, peptides comprising each motif and surrounding residues displayed different affinities for VDR or ER␤ binding. Thus, it is clear that the multiple NR boxes do not serve merely redundant functions.
Our finding that the AF-2 domain of RXR can interfere with RAC3-RID binding to a DNA-bound VDR/RXR heterodimer is consistent with studies suggesting allosteric inhibition of coactivator binding to RAR/RXR by the RXR AF-2 domain (36). This inhibition may be the result of competition between the AF-2 domain of RXR and the LXXLL motif for the coactivator binding site on the other receptor (11,12,36). In the antagonistbound ER␣-LBD crystal structure, the AF-2 domain occupies the coactivator binding groove, mimicking the hydrophobic interactions of the NR box peptide with this domain in the agonist-NR box peptide-receptor complex (11). Biochemical studies with RAR/RXR and SRC1 support these observations, for binding of RAR-and RXR-specific ligands enhance SRC1 interaction with the receptor dimer relative to the interaction in the presence of either ligand alone (36). Presumably, one ligand binding recruits a single NR box to the receptor dimer, which displaces the AF-2 domain from the coactivator binding site and relieves allosteric inhibition, allowing the second ligand to bind the other receptor monomer. This, in turn, enhances the interaction with coactivator by recruiting a second NR box (36). In the case of wild-type receptors, hormone does stimulate RAC3-RID binding to the heterodimer, but only weakly compared with the VDR/RXR443 dimer, where a very strong, vitamin D-dependent interaction is observed. These observations are confirmed by in vivo studies demonstrating that RAC3 can coactivate VDR/RXR443 activity to a greater extent than VDR/ RXR activity. Hormone binding and RID recruitment must not be able to displace every RXR AF-2 domain from the coactivator binding site of the partnering receptor; thus, fewer RID molecules are able to bind in the presence of the RXR AF-2 domain. This suggests that the AF-2 domain of RXR plays a critical role in regulating RAC3 modulation of receptor function. However, other possibilities may explain this finding, foremost being the hypothesis that deletion of the AF-2 domain of RXR results in a conformational change of the VDR/RXR443 dimer that enhances its affinity for the RAC3-RID.
Our data demonstrate for the first time that RAC3 can enhance the transcriptional activation function of the VDR and ER␤ and that this coactivation activity depends on different NR box requirements. Several other cofactors have been found to stimulate VDR activity, including SRC1, GRIP1/TIF2, NCoA-62, and the DRIP (VDR-interacting proteins) complex (13, 40 -42), whereas SRC1 can coactivate the ER␤ (43). The role of multiple coactivators in the function of the VDR in vivo is unknown, but several possibilities exist that suggest that the function of these coactivators is not completely redundant. First, the relative contribution of each coactivator may depend on cell or tissue type and/or coactivator levels in these cells. RAC3 is expressed at a high level in placenta, heart, and HeLa cells relative to TIF2 and SRC1 (20); thus, it may serve a more prominent role in receptor function in these cells. Second, dif-ferent coactivators may serve different functions that in total result in maximal transcriptional activation by the VDR. For example, RAC3 can interact with CBP; thus, RAC3 may recruit CBP to the VDR. SRC1 has intrinsic histone acetylation activity, and the DRIP VDR-interacting proteins complex may remodel nucleosomes (21,44), which also may contribute to the overall function of the VDR in stimulating target gene expression. Finally, we cannot rule out the existence of a complex containing multiple coactivators, which synergize to potentiate VDR activity.
In summary, our data establish RAC3 as a potent coactivator of the vitamin D receptor and estrogen receptor ␤. Interestingly, RAC3 modulates the function of these receptors differently via interactions that depend on specific LXXLL motifs in the RAC3 receptor-interacting domain. Although the biological role of RAC3 in nuclear receptor function remains to be explored, this study sheds light on the molecular mechanisms of RAC3 regulation of receptors that will hopefully lead to a better understanding of SRC coactivator function in vivo.