Retinoid X Receptor α (RXRα) Helix 12 Plays an Inhibitory Role in the Recruitment of the p160 Co-activators by Unliganded RXRα/Retinoic Acid Receptor α Heterodimers*

Retinoid X receptor (RXR)/retinoic acid receptor (RAR) heterodimers control gene expression through recruitment of co-repressors or co-activators, depending on their hormone binding status. We show that the helix 12 of RXRα and RARα is critical for recruitment of the co-regulators and transcriptional regulation by RXRα, RARα, and RXRα/RARα. LG268, an RXR-specific agonist, was able to promote co-activator association with the heterodimers, but was unable to dissociate co-repressors. Reconstitution experiments in yeast demonstrated that LG268 was capable of activating transcription by RXRα/RARα through recruitment of the co-activator. We hypothesize that the inability to release co-repressors from RXRα/RARα is responsible for the inability of LG268 to activate RXRα/RARα heterodimers in mammalian cells. Deletion of RARα helix 12 (RXRα/RARα Δ403) abolished both hormone-dependent dissociation from co-repressors and hormone-dependent association with co-activators. Deletion of RXRα helix 12 (RXRα Δ443/RARα) resulted in a higher binding affinity for co-repressors. Unexpectedly, RXRα Δ443/RARα also gained hormone-independent co-activator binding activity. Moreover, LG268 became an antagonist to RXRα Δ443/RARα heterodimers. These data suggest that the helix 12 of RXRα plays an inhibitory role in the recruitment of co-activators by unliganded RXRα/RARα.

Whereas the DBD recognizes specific DNA sequences, the LBD dictates transcriptional activity by nuclear receptors through recruitment of co-repressors such as silencing mediator for retinoic acid and thyroid hormone receptor (SMRT) and nuclear receptor co-repressor or co-activators, including p160 family proteins (3,4).
Nuclear hormone receptors have been further subdivided into four classes based on their mode of action and DNA binding status (1). Retinoic acid receptors (RARs) and Retinoid X receptors (RXRs) are class II receptors that mediate retinoid signaling by regulating a complex gene network. RARs primarily form heterodimers with RXRs on DR5 (direct repeat spaced by 5 base pairs) or DR1-containing sequences. Conversely, RXRs are capable of forming homodimers on DR1 or heterodimers with other class II nuclear hormone receptors (1). The ability of RXRs to form homodimers or heterodimers with other receptors generates a complex mechanism by which the ligands of RXR may or may not exert effects on target genes, depending on the dimerization partner of the receptor (5). Receptors such as RARs and TRs (thyroid hormone receptors) dimerize with RXRs, generating a non-permissive complex in which RXR ligands are not capable of activating heterodimers. In such heterodimeric complexes, RXRs are referred to as silent partners. The mechanisms underlying this observation are currently unclear, but it has been proposed that RXRs within these heterodimers cannot bind their agonists (6,7). In contrast, permissive complexes between peroxisome proliferating activator receptors (PPARs) or vitamin D receptors and RXRs can be activated solely by RXR ligands.
Despite divergence in the primary sequences of LBDs of nuclear receptors, the structures of several LBDs are remarkably conserved (8,9). Structural comparison between RXR␣ apo-LBD (hormone-free) (10) and RXR␥ holo-LBD (hormonebound) (11) indicates that ligand binding triggers a conformational change within the receptors in which helix 12 of the RXR LBD makes a lid to a cavity containing helices 3, 4, and 5. Recent crystal and molecular studies have suggested that such a conformational change by helix 12 creates an interface for co-activator binding (12)(13)(14). Although the complex structure of the antagonist-bound PPAR␣ and co-repressor peptide has been solved (15), co-crystal structures between co-repressors and apo-receptors are unavailable. Nonetheless, biochemical and molecular studies indicate that the binding pockets for co-repressors and co-activators largely overlap (16 -19).
It is widely believed that, physiologically, class II receptors form high affinity DNA binding transcription factors as heterodimers, rather than as monomers (1). The high resolution structure of the liganded RXR␣/RAR␣ heterodimeric complex is now available (20) and provides insight into the molecular mechanism of dimerization; however, the crystal structure of the apo-heterodimer is currently unavailable, likely due to the unstructured helix 12. Therefore, understanding the roles of the helix 12 of dimerization partners in co-regulator recruitment is key to elucidating the molecular basis of transcriptional regulation by class II receptors.
In this report, we map the critical residues in the helix 12 of RAR␣ and RXR␣ for co-regulator association and transcriptional regulation by both receptors. Interestingly, although several mutants exhibit ligand-independent association with the co-activator, these mutants also acquire ligand-insensitive dissociation from the co-repressor. Our data indicate that LG268, a known RXR-specific ligand, promotes co-activator binding, but not co-repressor dissociation from heterodimers. We also found that RXR␣ helix 12 mutants displayed mild defects in all-trans-retinoic acid (AT-RA)-induced co-activator recruitment. Unexpectedly, deletion of RXR␣ helix 12 created a better co-activator binding surface within RXR␣ ⌬443/RAR␣ heterodimers. Furthermore, AT-RA, but not LG268, is capable of promoting recruitment of co-activators by RXR␣ ⌬443/RAR␣ heterodimers. These results indicate the unique requirement of helix 12 of RXR␣ and RAR␣ for recruitment of co-activators by the RXR␣/RAR␣ heterodimer.
Yeast Methods-Yeast two-hybrid (Y2H) assays were carried out according to the manufacturer's protocol (Clontech). Liquid ␤-galactosidase activity was performed, and the values were derived from duplicate with two independent clones. Yeast three-hybrid (Y3H) assays were performed in strain Y190.
Electrophoresis Mobility Shift Assays-EMSAs were carried out according to our published protocol (17). Glutathione S-transferase (GST) fusion proteins were expressed in Escherichia coli DH5␣ and affinitypurified with glutathione-Sepharose 4B beads. Immobilized GST fusion proteins were eluted by 20 M glutathione and dialyzed against 1ϫ phosphate-buffered saline. RAR␣, RXR␣, and TR␤ synthesized in vitro (Promega) were incubated with a 32 P-labeled probe containing a response element with or without 1 M AT-RA. The probes include: DR5, 5Ј-GGT AGG GTT CAC CGA AAG TTC ACT C-3Ј; DR1: 5Ј-GCT TCA GGT CAG AGG TCA GAG AG-3Ј; DR4: 5Ј-GGT GGG TCA TCT GAG GTC AG-3Ј. DNA-protein binding was conducted in a reaction mixture containing 20 mM Hepes, pH 7.4, 50 mM KCl, 1 mM ␤-mercaptoethanol, and 10% glycerol. After 30 min at 25°C, the purified GST-SMRT or GST-ACTR fusion proteins were added, followed by an additional 30 min of incubation. The reaction mixtures were loaded onto a 5% polyacrylamide (29.2:0.8) non-denaturing gel followed by electrophoresis in 0.5ϫ TBE buffer (50 mM Tris, 50 mM boric acid, 25 mM EDTA pH 8.0). After electrophoresis, the gel was dried and subjected to autoradiography.

Association of RAR␣ Helix 12 Mutants with Co-repressors-
To investigate the roles of helix 12 in nuclear receptor function, we focused on helix 12 of RAR␣ and RXR␣. Fig. 1A shows an alignment of helix 12 sequences from several class II nuclear receptors. In this study, mutants were generated in the conserved residues of RAR␣ helix 12. We first tested whether these mutants bound SMRT by EMSAs. GST-SMRT (receptor interaction domain (RID)) fusion protein was expressed in bacteria and subsequently affinity-purified (17). EMSAs were conducted using a 32 P-labeled DNA duplex containing a DR5 element of the retinoic acid response element (␤RARE) of from human RAR␤2 gene. Both RXR␣ and wild-type or mutant RAR␣ were translated in vitro and incubated with the 32 Plabeled probe and purified GST-SMRT in the absence or presence of increasing concentrations of AT-RA. Wild-type RAR␣ displayed a dose-dependent hormone-induced dissociation from SMRT (Fig. 1B, lanes 1-4), whereas P408L (lane 9) and E412A (lane 17) bound SMRT weakly, and this binding was dramatically decreased in the presence of 10 nM AT-RA (lanes 10 and 18). Furthermore, RAR␣ ⌬403 (lanes 5-8), L409A (lanes [13][14][15][16], and M413A (lanes 21-24) did not fully dissociate SMRT, even at 1 M AT-RA. These data indicate that Pro-408 and Glu-412 play an important role in co-repressor binding and that Leu-409 and Met-413 are critical for hormone-dependent dissociation of co-repressors.
We next analyzed the association of co-repressor with helix 12 deletion (⌬403) and point mutants by mammalian twohybrid (M2H) assays using Gal4-SMRT (RID) and VP-RAR␣ (LBD) along with a reporter construct (pMH100) containing multiple Gal4 binding sites upstream of a TK basal promoter and the firefly luciferase gene. As shown in Fig. 1C, all helix 12 mutants showed high reporter activity, suggesting that these mutants bound SMRT in the absence of all-trans-retinoic acid (AT-RA), although the degree of association varied among the mutants. In the presence of 10 nM AT-RA, wild-type RAR␣ and several of the mutants displayed a dramatic reduction of the reporter activity. In contrast, significant amounts of the reporter activity were detected for RAR␣ ⌬403, L409A, and M413A in the presence of 10 nM AT-RA, and 50% of RAR␣ ⌬403 was still associated with SMRT at 1 M AT-RA. These results suggested that RAR␣ ⌬403 acquired aberrant ligand-insensitive co-repressor binding. Similarly, the reporter activity was diminished for point mutants L409A and M413A only at 1 M AT-RA, suggesting that complete dissociation of SMRT required more than 1 M AT-RA. In contrast, P408L and E412A displayed less reporter activity, indicating that they bound SMRT more weakly than that of wild-type RAR␣, but this binding was readily disrupted in the presence of 10 nM AT-RA. These results are consistent with those of EMSAs.
Hormone-dependent Co-activator Association with RAR␣ Helix 12 Mutants-The ability of RAR␣ helix 12 mutants to recruit co-activators such as ACTR was examined by EMSAs and M2H assays. EMSAs were carried out as described in Fig. 1B except that GST-ACTR (RID) fusion protein was used ( Fig. 2A). Wild-type RAR␣ exhibited a hormone-dependent association with ACTR, whereas mutants ⌬403 (lanes 5-8), L409A (lanes [13][14][15][16], and E412A (lanes 17-20) completely lost ACTR binding activity. Unexpectedly, we found that unliganded P408L bound weakly to ACTR (lane 9) and no further ACTR binding was induced by high concentrations of AT-RA (lanes 10 -12). Similarly, M413A also bound ACTR in the absence of ligand (lane 21), although this association was increased by AT-RA in a dose-dependent manner (lanes [22][23][24]. These data suggest that the conserved residues in the helix 12 of RAR␣ are critical for co-activator association. M2H assays were conducted as described in Fig. 1C using pCMX-ACTR (RID) and VP-RAR␣ (LBD). The association of RAR␣ with ACTR was measured in the presence of increasing concentrations of AT-RA. In the absence of AT-RA, no reporter activity was detected, suggesting no association with ACTR (RID) was detected for wild-type or any of the mutant RAR␣s (LBD) (Fig. 2B). In the presence of AT-RA, the reporter activity was increased, indicating that wild-type RAR␣ bound ACTR in a dose-dependent manner. However, we found that all mutants exhibited much lower reporter activity than that of the wildtype even in the presence of 1 M AT-RA. These data suggested that RAR␣ ⌬403, L409A, and E412A possess defects in coactivator binding. Similarly, P408L and M413A dramatically lost their ability to bind ACTR.
The aberrant association of these mutants with SMRT and ACTR suggested that RAR␣ helix 12 mutants have a defect in hormone-dependent transcriptional regulation. Transient transfection assays were employed to examine the ability of helix 12 mutants to regulate transcription by RAR␣. Wild-type and mutant RAR␣ were fused to the yeast Gal4 DNA-binding FIG. 1. Association of SMRT and mutants of RAR␣ helix 12. Transient transfection assays were carried out in CV-1 cells. A, alignment of helix 12 of selected nuclear receptors. The conserved residues are boxed. The stars are sites of the mutation. B, association of RAR␣ helix 12 mutants with SMRT on DR5 elements. SMRT binding was assayed by EMSAs on DR5 in the presence of increasing concentrations of AT-RA. RXR␣ and RAR␣ were in vitro translated. GST-SMRT was purified from E. coli. EMSAs were performed as described previously (17). C, ligand-dependent dissociation of RAR␣ from SMRT. Association of SMRT ID I ϩ II with RAR␣ mutants was determined with or without increasing concentration of AT-RA by mammalian two-hybrid (M2H) assays. The reporter activity of Gal4-SMRT with VP16 alone was defined as one. Activation was calculated as the ratio of the reporter activity of Gal4-SMRT ID I ϩ II with VP16-RAR␣/the reporter activity in the presence of Gal4-SMRT ID I ϩ II with VP16 alone. domain to generate Gal4-RAR␣ (LBD) expression plasmids. The Gal4-RAR␣ (LBD) expression vectors were co-transfected with MH100, and reporter activity was measured in the absence or presence of AT-RA. Our data indicate that all the helix 12 mutants repressed transcription in the absence of hormone ( Fig. 3A), although P408L, E412A, and M413A exhibited less repression activity. We also found that mutants ⌬403, L409A, and E412A completely lost activation activity in the presence of AT-RA, that wild-type RAR␣ was able to activate transcription in a hormone dose-dependent fashion, and that P408L and M413A activated transcription to a lesser degree. These data indicated that the conserved residues of RAR␣ helix 12 play a pivotal role in co-regulator recruitment and subsequent transcription regulation.
Roles of RAR␣ Helix 12 in Transcriptional Repression by RXR␣/RAR␣-Because RXR/RAR heterodimers are the functional unit for AT-RA signaling and because they bind the DR5 element, the effects of these RAR␣ mutations on transcription was further tested using a DR5-TK-Luc reporter construct (Fig.  3B). Without exogenous RAR␣, AT-RA activated reporter expression in a dose-dependent manner, suggesting the presence of endogenous RXRs and RARs in CV1 cells. Ectopic expression of wild-type RAR␣ enhanced AT-RA-induced activation. Introduction of mutants ⌬403 or Leu-409, moderately potentiated repression activity, consistent with their higher SMRT binding activity. Furthermore, at 10 -100 nM AT-RA, all helix 12 mutants displayed reduced AT-RA-induced activation activity relative to transfected wild-type RAR␣. In particular, expression of RAR␣ ⌬403 abolished basal activity (lanes 9 -12), indicating a dominant-negative effect. These data provide evidence that helix 12 of RAR␣ plays a critical role in AT-RA-induced activation by RXR␣/RAR␣ heterodimers.

Association of RXR Helix 12 Mutants with Co-regulators-
Having determined the effects of mutations of RAR␣ helix 12 on co-regulator binding and transcriptional activity, we next examined the properties of RXR␣ helix 12 mutants. We first determined whether RXR␣ helix 12 is critical for co-regulator recruitment in yeast two-hybrid (Y2H) assays. Because RXRs are capable of dimerizing with a set of nuclear receptors in mammalian cells and because yeast does not express nuclear receptors, Y2H assays provide a system to examine co-regulator association that excludes possible bridging effects by other receptors. Y2H expression vectors, pACTII-RXR␣ (LBD) and pGBT9 (Gal4 DBD)-SMRT (RID) or pGBT9-ACTR (RID), were co-transformed into yeast. The ability of these mutants to interact with co-regulators was measured in the absence or presence of hormone. In the absence of RXR agonists, ␤-galactosidase was expressed in cells co-transformed with pGBT9-SMRT (RID) and pACTII-RXR␣ (LBD) (Fig. 4A), whereas little ␤-galactosidase activity was detected in experiments with only one of the proteins. Although addition of LG268 did not have an effect on the interaction, 9-cis-RA moderately induced interaction between RXR␣ and SMRT. These data suggest that, in contrast to RAR ligands, which induce dissociation of co-repressors from RAR, RXR ligands do not disrupt RXR␣-co-repressor association.
Mutations at the conserved residues within RXR␣ helix 12 displayed varying effects on SMRT binding (Fig. 4B). Although mutants F450A and M454A/L455A completely abolished association with SMRT, mutant L451A increased SMRT binding activity, and this interaction was partially inhibited in the presence of 9-cis-RA or LG268. Deletion of helix 12 (⌬443) strongly enhanced hormone-independent co-repressor interaction, whereas 9-cis-RA or LG268 reduced the interaction by   FIG. 2. Association of ACTR and mutants of RAR␣ helix 12. A, association of RAR␣ helix 12 mutants with ACTR on DR5 elements. ACTR binding was assayed by EMSA assays on DR5 in the presence of increasing concentrations of AT-RA similar to that of Fig. 1B. B, liganddependent association of RAR␣ with ACTR. Association of ACTR (RID) with RAR␣ mutants was determined with or without increasing concentrations of AT-RA by mammalian two-hybrid (M2H) assays. The reporter activity of Gal4-SMRT with VP16 alone was defined as one. The -fold activation was calculated as the ratio of the reporter activity of Gal4-ACTR (RID) with VP16-RAR␣/the reporter activity in the presence of Gal4-ACTR (RID) with VP16 alone. M2H assays were carried out similar to that of Fig. 1C. 50%. Our results suggest that helix 12 plays an important role in co-repressor recruitment by RXR␣.
Similar assays were also carried out to examine association between ACTR and wild-type or mutant RXR␣. Expression of pGBT9-ACTR or pACTII-RXR␣ alone did not activate ␤-galactosidase expression with or without ligands (data not shown). In the absence of 9-cis-RA or LG268, a high level of ␤-galactosidase activity was measured in cells expressing pGBT9-ACTR and pACTII-RXR␣, indicating that unliganded RXR␣ interacts with ACTR (Fig. 4C) in yeast. Moreover, 9-cis-RA and LG268 induced ␤-galactosidase expression an additional 3-to 4-fold, suggesting that both 9-cis-RA and LG268 were capable of promoting further association between RXR␣ and ACTR. Mutations or deletion of helix 12 completely abolished ligand-independent and ligand-induced RXR␣-ACTR interactions. These results indicate that helix 12 is essential for both hormoneindependent and -dependent recruitment of co-activators by RXR␣ in yeast.
The effect of helix 12 mutations on transcriptional regulation of RXR␣ was examined by transient transfection assays as described in Fig. 3A. Reporter activity was measured in the absence or presence of an RXR-specific agonist, LG268. Fig. 4D shows that, in the absence of ligand, Gal4-RXR␣ weakly acti-vated transcription. Mutations at the conserved residues had little effect on unliganded RXR␣ activity, except ⌬443 and L451A, which possessed moderate repression activity, consistent with their strong association with SMRT in yeast. In the presence of LG268, Gal4-RXR␣ was able to activate transcription, but mutation or deletion of helix 12 dramatically reduced this activation. These data are consistent with the proteinprotein interaction data from Y2H assays and suggest that helix 12 is critical for hormone-mediated transcriptional activation by Gal4-RXR␣.
RXR␣ Helix 12 Is Required for AT-RA-mediated Activation by RXR␣/RAR␣-The primary sequence of RXR␣ helix 12 is similar to that of other class II receptors, yet the role of RXR␣ helix 12 in transcriptional regulation by class II nuclear receptors is unclear. AT-RA is an RAR␣-specific ligand and is thought to be permissive in AT-RA-mediated transactivation by RXR␣/RAR␣ heterodimers. In contrast, RXR-specific ligands such as LG268 do not activate transcription by RXR␣/ RAR␣ and therefore RXR␣ is a non-permissive partner (1,5). These observations argue that RXR␣ plays a minor role in AT-RA-mediated transcriptional activation. To examine the role of RXR␣ helix 12 in transcriptional regulation by RXR␣/ RAR␣, RXR␣ helix 12 mutants were used to test the transcrip- tion activity of RXR␣/RAR␣ on a ␤RARE-derived (DR5) reporter construct with or without hormone. Fig. 5A shows that AT-RA increased reporter activity in the absence of transfected RXR␣ (ϳ9-fold). Ectopic expression of wild-type RXR␣ stimulated AT-RA-induced activation by an additional 10-fold. Mutations in RXR␣ helix 12 reduced the AT-RA-induced activation observed with RXR␣/RAR␣ by 50 -70%. In particular, exogenous expression of RXR␣ ⌬443 completely abolished AT-RAinduced activation above the endogenous background. These data indicate that helix 12 of RXR␣ is essential for optimal AT-RA-induced transcriptional activation by RXR␣/RAR␣ heterodimers.
The decreased reporter activity noted in Fig. 5A with the helix 12 mutants could be due to the inability of mutant RXR␣/ RAR␣ heterodimers to recruit co-activators or the inability of RXR␣/RAR␣ to dissociate from co-repressors. To address this issue, we examined the association profiles of RXR␣/RAR␣ with co-repressors or co-activators by EMSAs. We first tested disso-

FIG. 4. Co-regulator recruitment and transcriptional regulation by helix 12 mutants of RXR␣.
A, association of RXR␣ with SMRT in yeast. Yeast expression plasmids, pGBT9-SMRT (RID) and pACTII-RXR␣ (LBD), were co-transformed into yeast strain Y190. Transformed cells were grown in the absence of presence of RXR ligands. Liquid ␤-galactosidase assays were determined according to manufacturer's protocol. ␤-Galactosidase activity was normalized by total cell numbers and time. B, association of SMRT with RAR␣ helix 12 mutants in yeast. Y2H assays were carried out similar to A. C, association of ACTR with RAR␣ helix 12 mutants in yeast. Y2H assays were carried out similar to A. D, transcription activity by Gal4-RXR␣ in CV-1 cells. Assays were carried out as described in Fig. 3A. ciation of SMRT from wild-type or mutant RXR␣/RAR␣ heterodimers with increasing concentrations of AT-RA (Fig. 5B). We found that mutant heterodimers with L451A (lane 9) or M454A/L455A (lane 13) display a slightly higher SMRT binding activity than that of wild-type RXR␣/RAR␣. Furthermore, AT-RA-induced dissociation of SMRT was partially impaired in these mutants (lane 14). The ability of mutant heterodimers to recruit ACTR was also examined in the presence or absence of hormone (Fig. 5C). We found that the AT-RA-induced ACTR association with RXR␣/RAR␣ was slightly affected by the RXR␣ helix 12 point mutations.

FIG. 5. Recruitment of co-regulators and transcriptional regulation by RXR␣/RAR␣ on DR5 by EMSAs.
A, effects of RXR␣ helix 12 mutants on transcriptional regulation by RXR␣/RAR␣. Wild-type and mutant RXR␣ were cotransfected with a reporter construct and internal control plasmid, CMX-␤-galactosidase, into CV-1 cells. B, co-repressor recruitment by RXR␣ helix 12 in EMSAs. Wild-type or mutant RXRs were mixed with RAR␣ followed by the addition of 32 P-labeled probe, ligand, and purified GST-SMRT (RID). C, co-activator recruitment by RXR␣ helix 12. EMSAs were performed as in panel above. than that of wild-type RXR␣/RAR␣ (lanes 13-16), and SMRT dissociation was less sensitive to AT-RA than that of wild-type RXR␣/RAR␣ (lanes 5-8).
Because co-repressor release is a pre-requisite for co-activator binding and subsequent transcriptional activation (17), we hypothesized that, although RXR agonists promote co-activator association, they may not be able to induce dissociation of co-repressors. To address this question, we used EMSAs to test whether LG268 is capable of promoting recruitment of coactivators and release of co-repressors. As shown in Fig. 6B, addition of AT-RA (lanes 2 versus 3) or LG268 (lanes 2 versus 4) strongly induced association with ACTR with the wild-type heterodimer. However, whereas AT-RA completely disrupted SMRT binding (lanes 5 versus 6) with the wild-type heterodimers, LG268 did not (lanes 5 versus 7). In the context of the RXR␣ ⌬443/RAR␣ heterodimer, AT-RA was still capable of promoting ACTR association (lanes 9 and 10). Interestingly, addition of LG268 weakly inhibited ACTR association (lanes 9 versus 11) and did not have an effect on SMRT binding (lane 14). We also examined whether these results can be recapitulated using a DR1 probe, which is also a known response element for RXR␣/RAR␣. Experiments were carried out in the same manner as Fig. 6B, except that the probe was a DR1containing element derived from the human CRABPII gene. Fig. 6C demonstrates that, whereas AT-RA strongly induced ACTR association and SMRT dissociation on DR1 with the wild-type heterodimer (lanes 2 versus 3 and 6), LG268 was unable to promote dissociation of SMRT (lanes 5 versus 7). Similar to the DR5 probe, LG268 inhibited recruitment of ACTR by the RXR␣ ⌬443/RAR␣ heterodimers on DR1 probe (lanes 9 versus 11).
We further tested whether this observation was unique to RXR␣/RAR␣ heterodimers or common to other non-permissive heterodimers such as RXR␣/TR␤. The effect of LG268 on recruitment of co-regulators by RXR␣/TR␤ was further examined using the thyroid hormone response element, DR4, derived from the human DI gene (Fig. 6D). While thyroid hormone (T3) weakly promoted ACTR binding (lanes 2 and 3), LG268 significantly enhanced ACTR association (lanes 2 and 4). We also found that T3 induced significant dissociation of SMRT from RXR␣/TR␤, but LG268 did not (lanes 5-7). In the context of RXR␣ ⌬443/TR␤ heterodimers, T3 had a more pronounced effect on promoting ACTR binding compared with wild-type heterodimers (lanes 2 and 3 versus 9 and 10). Similar to that of RXR␣ ⌬443/RAR␣ heterodimers, LG268 weakly inhibited ACTR binding (lanes 9 versus 11) to RXR␣/TR␤. Our results indicate that helix 12 of RXR␣ has an inhibitory effect on co-activator recruitment by unliganded heterodimers but is critical for the agonistic effect of LG268. Moreover, the failure of LG268 to induce dissociation of co-repressors from receptors is common to RXR␣/RAR␣ and RXR␣/TR␤ heterodimers.
Hormone-independent Activation by RXR␣ ⌬443/RAR␣ Heterodimers in Yeast-We next tested whether our in vitro observations with EMSAs can be recapitulated in vivo using yeast three-hybrid (Y3H) assays. Plasmids expressing Gal4-DBD-ACTR (RID), pGAD-RAR␣ (LBD), and RXR␣ ⌬443 were prepared and transformed into yeast strain Y190. Based on our EMSA data (Fig. 6), we anticipated that expression of the ␤-galactosidase activity induced by the formation of ACTR (RID)/RXR␣ ⌬443/RAR␣ trimeric complex will be higher than that of ACTR (RID)/RXR␣/RAR␣. As shown in Fig. 7A, expression of Gal4-DBD-ACTR (RID)-RXR␣ and RAR␣ moderately induced expression of ␤-galactosidase activity. Replacement of wild-type RXR␣ with RXR␣ ⌬443 further increased the reporter activity 2-fold, suggesting that mutant RXR␣ ⌬443/ RAR␣ heterodimers bind ACTR better than wild-type RXR␣/ FIG. 6. Effects of RXR␣ helix 12 deletion on co-regulator recruitment. EMSAs were carried out according to "Materials and Methods." A, effects of helix 12 deletion on the recruitment of co-activators by RXR␣/RAR␣ heterodimers on DR5 probe. B, effects of LG268 on coregulator association with RXR␣/RAR␣ heterodimers on DR5 probe. C, effects of helix 12 deletion on co-regulator recruitment by RXR␣/RAR␣ heterodimers on DR1. D, effects of LG268 on co-regulator association with RXR␣/TR␤ heterodimers on DR4. RXR homodimers do not bind these elements under these assayed conditions (data not shown).
RAR␣. Addition of AT-RA promoted association of ACTR with both RXR␣/RAR␣ and RXR␣ ⌬443/RAR␣ and activated reporter activity, suggesting that deletion of RXR␣ helix 12 does not prevent AT-RA-induced recruitment of ACTR. Taken together, these data suggest that helix 12 of RXR␣ is inhibitory to the recruitment of co-activator by unliganded RXR␣/RAR␣.
LG268 Is Able to Activate Transcription by RXR/RAR in Yeast-Our EMSA data also suggested that LG268 is capable of promoting co-activator recruitment by RXR␣/RAR␣ in the absence of co-repressors. We reasoned that LG268 is incapable of activating RXR␣/RAR␣ in mammalian cells, because LG268 does not promote co-repressor dissociation from the heterodimers. Thus, we predicted that LG268 would promote coactivator association and subsequent transcriptional activation by RXR␣/RAR␣ in cells that lack SMRT and nuclear receptor co-repressor, such as yeast. To test this hypothesis, the transformed yeast cells were grown in the absence or presence LG268, and the reporter activity was measured. As shown in Fig. 7B, LG268 was able to promote association between ACTR and RXR␣/RAR␣ heterodimers and activate reporter activity (lanes 1 and 2). However, the LG268-induced reporter activity relies on the integrity of RXR␣ helix 12, because LG268 did not promote association between ACTR and RXR␣ ⌬443/RAR␣ (lanes 3 and 4). These data suggest that RXR␣/RAR␣ is able to respond to LG268 and activate transcription in the absence of co-repressors. DISCUSSION In this study, we have examined the role of helix 12 of RAR␣ and RXR␣ on co-regulator recruitment and transcriptional regulation by RXR␣/RAR␣. Our results indicate that mutations at conserved residues within helix 12 of both RXR␣ and RAR␣ effect co-regulator association and subsequent transcriptional regulation. Deletion of helix 12 not only promoted co-repressor binding but also promoted recruitment of the co-activator in a hormone-independent manner. We also found that, although LG268, a RXR agonist, was able to promote co-activator association, it was unable to dissociate co-repressors from heterodimers. Our results provide evidence for an unexpected role of RXR␣ helix 12 in recruitment of co-regulators and further suggest a mechanistic link between activation and repression.
RAR␣ Helix 12 Mutants Have Two Defects in Both Co-repressor and Co-activator Association-Sequence comparison of class II nuclear receptors suggested an important role of helix 12. Our data indicate that the conserved residues of RAR␣ helix 12 are critical for the recruitment of both co-repressors and co-activators. At physiological concentrations of AT-RA (10 -100 nM), all RAR␣ helix 12 mutants except M413A acted as dominant-negative mutants (Fig. 3B). Based on their biochemical properties, these mutants can be grouped into three categories. First, Pro-408 and Glu-412 mutations displayed reduced co-repressor association and subsequent repression activity. Intriguingly, both mutants also lost hormone-dependent co-activator binding activity. The second group is represented by M413A. Unliganded M413A bound co-repressor as well as wild-type, but AT-RA barely caused co-repressor dissociation. Indeed, at 1 M AT-RA, residual amounts of SMRT still bound RXR␣/RAR␣ (M413A) on DR5. Additionally, M413A acquired hormone-independent association with ACTR and AT-RA promoted further ACTR association in a dose-dependent manner. This mutant is reminiscent of the K262A mutant we described previously (17), suggesting that these two residues serve common functions, preventing hormone-independent co-activator association and promoting hormone-induced dissociation of co-repressors. The third group includes ⌬403 and L409A, which possessed two defects, hormone-dependent dissociation of co-repressors and hormone-dependent recruitment of co-activators. Although mutant M413A was capable of interacting with ACTR in the absence of AT-RA in EMSAs ( Fig.  2A), it did not interact with ACTR in M2H assays (Fig. 2B). Presumably, M413A remains associated with co-repressors in mammalian cells in the absence of AT-RA. Indeed, Gal4-RAR␣ (M413A) only weakly repressed transcription (Fig. 3A). These data suggest that transcriptional regulation by RAR␣ depends primarily on dissociation of co-repressors and release of corepressors is a pre-requisite for activation.
Although our data indicate that replacement of the conserved residues of RAR␣ helix 12 effects co-activator binding, we cannot exclude the possibility that the defect in some of the mutations was due to an inability of these mutants to bind hormone. Structural studies have indicated that the only residue in the helix 12 of RAR␥ in close proximity to AT-RA is Ile-410 (ϳ4.1 Å) (10), whereas the closest residue in the helix 12 of RXR␣ to 9-cis-RA is Leu-451, but a direct interaction is unlikely given the substantial distance between Leu-451 and 9-cis-RA (ϳ7.1 Å) (11). Indeed, it has been shown that deletion of helix 12 caused only a moderate decrease on AT-RA binding, suggesting RAR␣ helix 12 plays a minor role in AT-RA binding (23). Furthermore, none of the residues within helices 12 of ligand-bound RXR␣/RAR␣ heterodimers are within the range for an interaction with the synthetic ligands (20). Our data are FIG. 7. Association of ACTR and RXR␣/RAR␣ in yeast. A, association of RXR␣ ⌬443/RAR␣ heterodimers with ACTR in yeast. Yeast three-hybrid assays were used to examine the interaction between ACTR and RXR␣ ⌬443/RAR␣. Y3H expression plasmid, pBridge, harboring Gal4 DBD-ACTR RID and RXR␣ or RXR␣ ⌬443 was co-expressed with pACTII-RAR␣ (LBD) in yeast strain, Y190. Yeast cells were grown with or without AT-RA. B, effects of RXR agonist, LG268, on the recruitment of ACTR by RXR␣/RAR␣ in yeast. Transformed yeast cells were same as in A. RXR-specific agonist, LG268, was added during growth and ␤-galactosidase activity was measured. also consistent with the structure of T3-bound TR␤⅐GRIP1 complex (12) in which Leu-454, Glu-457, and Val-458 of TR␤ (corresponding to residues Leu-409, Glu-412, and Met-413 of RAR␣) directly contact the LXXLL-containing ␣-helix from GRIP1 and are not located near the T3 binding site. There is no structure data available for apo-RAR␣ complexed with co-repressor peptides, but studies from antagonist-bound PPAR␣⅐SMRT complex suggested that the helix 12 of PPAR␣ lies parallel to the SMRT peptide at a distance too great for an interaction between helix 12 and the co-repressor peptide (Ͼ10 Å) (15). The SMRT peptide is held by helices 3-5 in the structure. However, it is possible that helix 12 of antagonist-bound PPAR␣ adopts a distinct conformation that is not present in apo-RAR␣ or apo-RXR␣. Further structural studies will help to address this issue.
The Role of RXR␣ Helix 12 in Retinoid Signaling-Using Y2H assays, we detected interactions between ACTR and unliganded RAR␣ as well as RXR␣ (Fig. 4). This result is unexpected, because we did not detect interactions between RAR␣ and ACTR in M2H assays (17) or in GST pull-down assays (data not shown). It is likely that there are molecules, either produced by yeast or in yeast medium, mimicking receptor ligands. Alternatively, receptors may interact with co-regulators through bridging factors in yeast. Our Y2H data indicate that, in the absence of RAR␣, RXR␣ ⌬443 not only gains corepressor binding activity but also loses ligand-mediated coactivator recruitment, indicating that helix 12 is critical for recruitment of co-regulators by RXR␣. Consistent with this observation, helix 12 mutations had severe effects on ligandinduced transcriptional activation by Gal4-RXR␣. We show that RXR ⌬443 does not potentiate AT-RA-mediated activation of the DR5 reporter construct (Fig. 5A). However, our EMSA data indicated that at 100 nM AT-RA, RXR␣ ⌬443/RAR␣ completely released SMRT and was able to recruit co-activator (Fig. 6A). This discrepancy is likely due to the fact that the concentration of GST-ACTR is much higher than that of GST-SMRT in our EMSA assays. Furthermore, the ratio of the [co-repressors]/[co-activators] in the cells may also determine whether RXR␣ ⌬443/RAR␣ represses or activates transcription.
Our data demonstrate that deletion of helix 12 appears to create a better co-repressor interaction surface for RXR␣ (Fig.  4) as well as RXR␣/RAR␣ and RXR␣/TR␤, an observation similar to that of two previous reports (24,25). Interestingly, our FIG. 8. Models for ligand-mediated transcriptional regulation by RXRs/RARs. A, AT-RA-mediated activation occurs in a three-step process. We hypothesized that co-repressor release is a pre-requisite for transcriptional activation by RXR/RAR heterodimer (17). Because the co-activator binding pocket overlaps with the co-repressor binding pocket, co-repressor must dissociate from heterodimers prior to co-activator recruitment. Non-permissive heterodimers such as RXRs/RARs and RXRs/TRs bind co-repressor in the absence of ligand. AT-RA will dissociate co-repressors from the heterodimer followed by recruitment of co-activators. B, LG268 is unable to activate transcription in non-permissive heterodimers such as RXRs/RARs and RXRs/TRs in mammalian cells. Because LG268 does not promote dissociation of co-repressors from non-permissive heterodimers, it is unable to activate transcription. C, RXR␣/RAR␣ heterodimer responds to LG268 and recruits co-activators in yeast. RXR heterodimers with vitamin D receptor or PPARs can be activated by RXR agonists as well. data indicated that mutations in the helix 12 of RXR␣ have either positive (Leu-451 and M454A/L455A) or negative effect on the association of RXR␣ with co-repressor. This result cannot be explained by a model in which the length, rather than the specific sequence of the RXR␣ helix 12, masks the corepressor binding pocket (25).
Intriguingly, we note that RXR␣ ⌬443 or unliganded RAR␣ did not bind co-activator individually. However, DNA-bound RXR␣ ⌬443/RAR␣ heterodimers acquired a hormone-independent co-activator binding activity. Our Y2H assays also verified that RXR␣ ⌬443/RAR␣ gained a better binding affinity to ACTR than wild-type RXR␣/RAR␣ (Fig. 7). Furthermore, AT-RA is capable of inducing further interaction between RXR␣ ⌬443/RAR␣ and ACTR. These results suggest: 1) deletion of helix 12 of RXR␣ did not effect hormone binding activity of RAR␣; 2) helix 12 of RXR␣ blocks co-activator binding and therefore plays an inhibitory role in the recruitment of coactivator by unliganded RXR␣/RAR␣ heterodimers; 3) helix 12 of RXR␣ partially overlaps with the co-activator binding pocket within the heterodimers. The fact that deletion of helix 12 of RXR␣ creates a better co-repressor and co-activator binding surface for RXR␣/RAR␣ heterodimers further strengthens the notion that the co-repressor and co-activator binding pockets are partially overlapped within the heterodimers. Together, these data pose an additional complexity of the allosteric effects mediated by dimerization of the partners. Future studies on the three-dimensional structure of unliganded heterodimers will advance our understanding of the molecular data.
Our finding that LG268 is able to induce recruitment of co-activators to heterodimers raises an important question. Do co-activators bind the liganded RXR or unliganded RAR within the heterodimers? Further studies will be required to address this question. Our EMSAs showed that, although AT-RA was able to promote recruitment of the co-activator by RXR␣ ⌬443/ RAR␣ and RXR␣ ⌬443/TR␤, LG268 was not. Indeed, addition of LG268 moderately inhibited co-activator binding to RXR␣ ⌬443/RAR␣ and RXR␣ ⌬443/TR␤, implying that binding of LG268 causes a conformational change, either within RXR␣ or RAR␣, such that it generates a poor co-activator binding pocket, thus acting as an antagonist. However, it remains to be seen whether this observation is common to all RXR agonists.
Why Are RXRs Silent Partners in Non-permissive Heterodimers?-The reasons why RXR agonists are unable to activate transcription by RXR/RAR heterodimers are unclear. It was proposed that only RARs can bind ligand; the RXR partner being incapable of this activity (7). Interestingly, two reports suggested that RXR/TR heterodimers do respond to 9-cis-RA, which is a RXR ligand (26,27). Our EMSA and Y3H data demonstrated that LG268, an RXR-specific agonist, promotes association of heterodimers with co-activators on DNA, suggesting that LG268 does bind to RXR␣/RAR␣. Nonetheless, we found that LG268 did not promote dissociation of co-repressor from RXR␣/RAR␣ heterodimers on either DR1 or on DR5 binding sites. Indeed, a similar result was observed for the RXR␣/ TR␤ heterodimers. These data indicate that the inability of LG268 to dissociate co-repressor is independent of receptor DNA binding. Based on these data, we propose a model in which AT-RA and T3 not only promote dissociation of co-repressors from heterodimers but also promote association with co-activators (Fig. 8A). Because co-activator and co-repressor binding surfaces on the receptors have significant overlap and LG268 is incapable of promoting release of co-repressors from non-permissive heterodimers, LG268 is unable to activate transcription by non-permissive heterodimers, including RXR␣/ RAR␣ and RXR␣/TR␤ (Fig. 8B). Our model further predicts that heterodimers, which do not bind co-repressors on DNA, can be activated by RXR agonists (Fig. 8C). Indeed, unliganded PPAR␥/RXR and vitamin D receptor/RXR, which do not bind co-repressors on EMSA data (28,29), 2 can be activated by RXR agonists. Furthermore, our results raise the possibility that RXR agonists can activate RXR␣/RAR␣ heterodimers provided there is a high ratio of [co-activator]/[co-repressor] (30). Indeed, LG268 was able to promote the recruitment of ACTR by RXR␣/ RAR␣ and activate the expression of a reporter gene in yeast, which does not express co-repressors. Together, our findings provide an alternative mechanism by which RXRs function as silent partners for RXR/RAR and RXR/TR heterodimers.