Co-repressor Release but Not Ligand Binding Is a Prerequisite for Transcription Activation by Human Retinoid Acid Receptor α Ligand-binding Domain*

Nuclear hormone receptors coordinately regulate the activity of genetic networks through the recruitment of transcriptional co-regulators, including co-repressors and co-activators. Allosteric modulation of the ligand-binding domain by hormonal activators shifts the co-factor binding preference by defined structural changes in overlapping docking sites. We report here that mutations at conserved residues within the docking motif of the retinoic acid receptor α cause defects in dimerization, co-regulator association, and transcriptional regulation. Furthermore, although a minimal co-repressor receptor interaction domain is sufficient for receptor binding, flanking sequences appear to stabilize this interaction without interfering with ligand sensitivity. However, ligand sensitivity is changed by the K262A mutation, which requires much higher concentrations of all-trans-retinoic acid to promote co-repressor dissociation. Consequently, K262A functions as a dominant-negative mutant at low concentrations of all-trans-retinoic acid. As a result, transcriptional activation is mechanistically linked to co-repressor release.

receptors (RAR) function as potent transcriptional repressors becoming activators upon hormone binding. The repression activity by unliganded RAR and TR is mediated through the recruitment of either the silencing mediator for retinoid and thyroid hormone receptors (SMRT) or nuclear receptor co-repressor complexes that include mSin3A and a variety of histone deacetylases (3)(4)(5)(6)(7). In contrast, transcriptional activation by nuclear hormone receptors involves the recruitment of histone acetyltransferase complexes that include CREB-binding protein/p300, p300/CBP-associated factor, and members of the p160 family (SRC-1, GRIP1/TIF2, and ACTR/RAC3/P/CIP) (8 -14).
Nuclear receptors contain two conserved modules, the DNAbinding domain and the carboxyl-terminal ligand-binding domain (LBD). DNA-binding domains bind hormone response elements and thus direct receptors to appropriate target genes. LBDs are required for nuclear localization, homodimerization and/or heterodimerization, co-regulator association (including co-repressors and co-activators), and most importantly, ligand binding. The LBDs are composed of 12 helices in which helices 3-5 are the most conserved among receptors and define the nuclear receptor LBD signature motif for co-regulator recruitment. The transcriptional switch of the receptors from repressors to activators involves a ligand-induced conformational change, resulting in an exchange of co-repressors and co-activators. Using site-directed mutagenesis, we and others have previously shown that helices 3-5 of TR␤ and peroxisome proliferator-activated receptor ␥ play a pivotal role in binding co-regulators (15)(16)(17)(18)(19)(20). Reciprocally, SMRT and nuclear receptor co-repressor contain two short peptide motifs that are both necessary and sufficient for mediating co-repressor binding to unliganded RARs (15)(16)(17). Both motifs contain a hydrophobic core (I/L)XX(I/V)I. These two motifs, termed coreID I and coreID II (17 and 19 amino acids, respectively), are conserved in both position and in sequence between nuclear receptor co-repressor and SMRT. A third TR␤ binding motif aminoterminal to those previously mapped in nuclear receptor corepressor may also exist (21). Helices 3-5 also interact with co-activators through a LXXLL-containing signature motif. The p160 co-activator proteins contain three putative LXXLL motifs, but the affinity and specificity of the individual LXXLL motifs within p160 proteins toward different receptors are not well characterized.
RARs and RXRs form heterodimers in solution and on re-sponse elements containing direct repeat spaced by one (DR1) or five base pairs (DR5) with the half-site sequence AGGTCA (1). Although the LBD three-dimensional structures of RXRs and RARs are very similar (22,23), they exhibit distinct ligand binding properties toward stereoisomers. RXRs bind only to 9-cis-retinoic acid, whereas RARs can bind both 9-cis-retinoic acid and all-trans-RA (AT-RA), albeit with different affinities.
RXRs are thought to function as a silent partner whose ligands alone do not activate transcription activity of the RXR/RAR heterodimer. In contrast, RARs ligands alone are able to activate transcription.
In an attempt to examine the molecular basis of co-regulator association with RAR␣, we have generated mutations within both conserved and diverged residues within RAR␣ helices 3 and 4 and systematically analyzed the properties of these mutants. We have identified critical residues for RAR␣ functions including dimerization, co-regulator association, and transcriptional activity. Our results suggest that the transcription activity of RAR␣ depends primarily on its ability to dissociate from co-repressors.
Transient Transfection-CV-1 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 50 units/ml of penicillin G, and 50 g/ml of streptomycin sulfate at 37°C in 7% CO 2 . For Gal4-RAR␣, CV-1 cells (60 -70% confluence, 48-well plate) were co-transfected with 16.6 -66.6 ng of pCMX-Gal4 and pCMX-VP16 (for mammalian two-hybrid assays) fusion constructs, 100 ng of pMH100-TK-Luc, and 100 ng of pCMX-LacZ in 200 l of Dulbecco's modified Eagle's medium containing 10% fetal bovine serum by the N- [1-(2,3-dioleoyloxy)propyl]-N,N,N-trimethylammonium methylsulfate-mediated procedure (7). For DR5-TK-Luc (see Fig. 5), the cells were transfected with 0.33 ng of pCMX-RAR␣, 100 ng of ␤-RARE-TK-Luc, and pCMX-LacZ. The amount of DNA in each transfection was kept constant by the addition of pCMX. The CV-1 cells were transferred to charcoal-stripped serum after adding DNA. After 24 h, the medium was replaced with or without all-trans-retinoic acid. The cells were harvested and assayed for luciferase activity 36 -48 h after transfection. The luciferase activity was normalized to the level of ␤-galactosidase activity. Each transfection was performed in triplicate and repeated at least three times.
Electrophoresis Mobility Shift Assays-GST fusion proteins were expressed in Escherichia coli DH5␣ strain and affinity purified by glutathione-Sepharose 4B beads. Immobilized GST fusion proteins were eluted by 20 M glutathione and dialyzed against 1ϫ phosphatebuffered saline. RAR␣ and RXR␣ synthesized in vitro (Promega) were incubated with a 32 P-labeled probe containing a DR5 element derived from the ␤RARE promoter containing the sense strand sequence 5Ј-GGT-AGG-GTT-CAC-CGA-AAG-TTC-ACT-C-3Ј with or without 1 M AT-RA. 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 of incubation at 25°C, the purified GST-SMRT or GST-ACTR fusion proteins were added followed by an additional 30 min of incubation. The final reaction mixtures were loaded onto a 5% polyacrylamide (29.2:0.8) nondenaturing gel followed by electrophoresis in 0.5ϫ TBE buffer. After electrophoresis, the gel was dried and subjected to autoradiography.

Identification of RAR␣ Residues Critical for Dimerization
with RXR␣-To determine the residues that are critical for RAR␣ function, we generated mutations on conserved and diverged residues within helices 3 and 4. Fig. 1A shows the nuclear receptor signature motif and the targets for site-spe-cific mutations in this study. Primarily, alanine substitutions were generated and tested for their ability to dimerize with RXR␣ (Fig. 1B). The ability of RAR␣ mutants to dimerize with wild-type RXR␣ was evaluated by mammalian two hybrid (M2H) assays and electrophoresis mobility shift assays (EMSA) on a template containing a DR5. M2H assays were conducted using Gal4-RXR␣ and VP-RAR␣ (LBD) along with a reporter construct (pMH100) containing multiple copies of Gal4-binding sites upstream of the thymidine kinase (TK) promoter fused with luciferase gene. Induction of the luciferase activity is an indicative of interaction of the partners. As a control, Gal4-RXR␣ with VP16 alone only gave a basal activity (Fig. 1B). In the presence of VP16-RAR␣, the reporter activity is induced, suggesting an association between RXR␣ and RAR␣. In the absence of AT-RA, all of the mutants except F249A and Q257A retain at least 75% of wild-type activity. As expected, the addition of 1 M AT-RA increased the reporter activity of the wild-type RAR␣, probably because of the recruitment of coactivators, which further increase transcriptional activation by Gal4-RXR␣/RAR␣. Interestingly, most of the RAR␣ mutants gave lower reporter activity than that of the wild type, indicating that these mutants have lower activation activity. This result suggests that these mutants might be defective in coactivator binding. Although unliganded F249A binds RXR␣ much less efficiently than wild type, the reporter activity of G4-RXR␣/RAR␣ (F249A) increased dramatically upon addition of 1 M AT-RA. To test the heterodimerization ability of wildtype or mutant RAR␣ with RXR␣ on DNA, in vitro synthesized RAR␣ and RXR␣ were mixed with a radiolabeled DNA duplex containing a DR5 element derived from the promoter of ␤RARE gene. Consistent with M2H assays, F249A and Q257A failed to heterodimerize with RXR␣ productively on DR5 (data not shown; see below). A likely possibility is that these two mutants failed to form heterodimer efficiently.
Helices 3 and 4 of RAR␣ Are Critical for Co-repressor Association-The ability of the unliganded RAR␣ mutants to bind SMRT was tested by M2H assays. M2H assays were conducted using Gal4-SMRT ID I ϩ II (amino acid 2064 -2307 of mouse SMRT, a fragment containing both ID I and II) and VP-RAR␣ ( Fig. 2A). Our data indicate that unliganded mutants V240A, F249A, Q257A, I258A, and L261A dramatically lost their SMRT binding activity (lanes 3, 6, and 8 -10). In the presence of 100 nM AT-RA, RAR␣ and SMRT interaction is completely abolished in all cases except mutant K262A (lane 11). We also tested whether these RAR␣ mutants bind SMRT in EMSAs. Using the same SMRT ID I ϩ II fragment used in M2H assays, we constructed pGEX-4T-1-SMRT ID I ϩ II for bacterial expression and subsequently affinity-purified the GST-SMRT ID I ϩ II fusion proteins. In vitro synthesized RAR␣ and RXR␣ were mixed with a radiolabeled DNA duplex containing a DR5 element derived from the promoter of the human ␤RARE gene. EMSAs were performed with or without purified GST-SMRT ID I ϩ II (Fig. 2B). Consistent with M2H assays, RXR␣/RAR␣ heterodimers with RAR␣ mutants V240A, F249A, Q257A, I258A, and L261A, significantly lost the ability to bind SMRT. The SMRT binding activity of mutants S232A, K244A, G248T, I254A, and K262A is moderately reduced. As with M2H assays, the addition of 100 nM AT-RA completely blocked SMRT ID association with RAR␣ except for mutant K262A (data not shown). These data indicate that the LBD signature motif is critical for co-repressor binding. We also noted that F249A and Q257A failed to heterodimerize with RXR␣ productively on DNA. This result is consistent with M2H data. A likely possibility is that these two mutants failed to bind DNA because they do not form heterodimers efficiently. We next measured the association between RAR␣ LBD mutants with coreID II (Fig. 2C) and found that SMRT coreID II domain alone is highly sensitive to mutations within RAR␣ LBD. Three patterns of association were identified. Most of the mutants, with the exception of I254A and K262A, failed to bind SMRT coreID II. In another group, S232A, K244A, and G248A bound ID I ϩ II but not the coreID II alone. Finally, association of coreID II and ID I ϩ II with RAR␣ LBD was abrogated in a third group consisting of V240A, F249A, Q257A, I258A, and L261A. Based on these results, we conclude that these residues are critical for co-repressor association and that sequences flanking coreID II may play a role in the association with RAR␣ LBD.
Association of Co-activators with RAR␣ Mutants-Because helices 3 and 4 of TR␤ and peroxisome proliferator-activated receptor ␥ have also been shown to be critical for co-activator association (18 -20), we tested the interaction between RAR␣ helices 3 and 4 mutants and p160 co-activator proteins. M2H assays were carried out to determine in vivo association of p160 co-activator proteins with RAR␣ using Gal4-ACTR (SRC-1 or GRIP1) ID (receptor interaction domain) and VP16-RAR␣ (LBD). Among the three known p160 family members, RAR␣ bound ACTR/RAC3/PCIP the best, with SRC-1 less well and GRIP1 the least (Fig. 3A), whereas TR␤ bound SRC-1 better than ACTR and GRIP1. M2H assays were used to determine in vivo association of ACTR and wild-type or mutant RAR␣. The data in Fig. 3B show that the association between ACTR and RAR␣ is severely compromised for mutants V240A, K244A, F249A, Q257A, I258A, and L261A. Furthermore, we found that the interactions between ACTR and RAR␣ are ligand concentration-dependent (data not shown; see below). We then tested the effect of RAR␣ mutations on ACTR binding by EMSA. EMSA was conducted as described in Fig. 2 with or without GST-ATCR in the presence of 100 nM AT-RA. We found that GST-ACTR RID was able to supershift DNA-bound wild-type FIG. 2. SMRT binding activity of mutants in RAR␣ helices 3 and 4. A, mammalian two-hybrid assays of SMRT ID with RAR␣ helices 3 and 4 mutants. Association of SMRT ID I ϩ II with RAR␣ mutants was determined with or without the presence of 100 nM AT-RA. For A and C, the reporter activity of Gal4-SMRT with VP16 alone was defined as 1-fold. Fold activation is calculated as the ratio of the reporter activity of Gal4-SMRT ID I ϩ II with VP16-RAR␣ divided by the reporter activity in the presence of Gal4-SMRT ID I ϩ II with VP16 alone. B, SMRT binding of RAR␣ helices 3 and 4 mutants on DR5 element. GST-SMRT ID I ϩ II fusion protein was employed in EMSA assays to determine RAR␣ mutant SMRT binding activity. C, the sequence flanking coreID II stabilizes the association of SMRT and RAR␣ LBD. The activities of the wild-type RAR␣ to associate with SMRT ID I ϩ II and SMRT coreID II are defined as 100% binding activity. The percentage of SMRT binding was determined by luciferase activity (mutant)/luciferase activity (wild type).
Co-regulator Association with RAR␣ Correlates with Transcriptional Activity-The ability of these RAR␣ mutants to regulate transcription was examined by transient transfection assays. Wild-type and RAR␣ mutants were fused to the yeast Gal4 DNA-binding domain to generate Gal4-RAR␣ expression plasmids. Transient transfection assays were conducted in the presence or absence of 100 nM of AT-RA. In the absence of AT-RA, the degree of transcriptional repression activity of RAR␣ varies among these mutants. Notably, V240A, F249A, Q257A, I258A, and L261A lose more than 80% of their repression activity (Fig. 4A). The repression activity correlates with the ability of these mutants to interact with SMRT. Furthermore, the activation activity of mutants K244A, F249A, Q257A, L261A, and K262A was dramatically impaired at 100 nM AT-RA (Fig. 4B). We noted that liganded V240A, I258A, and L261A bind ACTR less efficiently than the wild-type RAR␣, but their ligand-dependent transcriptional activation is only partially inhibited. We propose that this observation may reflect the loss of SMRT binding activity by V240A, I258A, and L261A. In addition, although K262A and wild type bound ACTR equally well in the presence of 100 nM, K262A only moderately activated transcription in transient transfection assays. We interpret this to mean that K262A is still associated with co-repressor complexes in the presence of 100 nM AT-RA, as shown in Fig. 2 (see "Discussion").
Distinct Activities of RAR␣ Helices 3 and 4 Mutants F249A and K262A-During our analyses for ligand concentration-dependent association/dissociation of RAR␣ mutants and co-factors, we found that the association/dissociation profile of K262A with co-factors was distinct from those of other helices 3 and 4 mutations. Fig. 5A is an EMSA with wild-type RAR␣ and mutant K262A in the presence of GST-SMRT with or without increasing concentration of AT-RA. Similar to other mutants, K262A acquires reduced SMRT association activity (Fig. 5A, compared the ratio of SMRT-bound/unbound RAR␣). We noted that at 1 M, residue SMRT still remains associated with DNA-bound RXR␣/RAR␣ heterodimer. M2H assays were also carried out to determine ligand-dependent dissociation between SMRT and wild-type RAR␣ or K262A (Fig. 5B). Consistently, M2H assays demonstrated that dissociation between SMRT and K262A requires at least 1 M AT-RA as opposed to 100 nM for SMRT and wild type RAR␣. In contrast, heterodimer RXR␣/RAR␣ (K262A) binds ACTR in the absence of ligand in vitro (EMSA), indicating a better ACTR binding activity than that of the wild-type RAR␣ (Fig. 5C). However, in vivo association of K262A with ACTR requires the presence of AT-RA.

FIG. 3. Association of RAR␣ with p160 family proteins.
A, RAR␣ binds ACTR with highest affinity. The strength of co-activator association was examined by M2H assays with or without corresponding ligands. B, association of RAR␣ with ACTR ID in M2H assays. Gal4-ACTR ID was co-transfected with VP-RAR␣ (LBD) for M2H assays. The reporter activity from Gal4-ACTR ID ϩ VP16 alone was set as 1-fold. C, association of liganded RAR␣ with ACTR on DR5. ACTR binding was assayed by EMSA assays on DR5 in the presence of 100 nM AT-RA. Note that F249A binds weakly to ACTR. Note that the ratio of GST-ACTRbound/unbound G248T is comparable with that of S232A. Although K262A binds to ACTR equally as well as that of wild type at 100 nM AT-RA, K262A binds less tightly than wild type at 1 M AT-RA in M2H assays (Fig. 5D). Under the same conditions, F249A can only associate with ACTR in the presence of 1 M AT-RA.
To understand how co-regulator association correlates with the transcriptional activity of K262A, transient transfection assays were employed. Unliganded wild-type Gal4-RAR␣ re-presses transcription efficiently (Fig. 5E, lane 5), whereas K262A moderately repressed basal transcription (lane 13). The reduced repression activity is consistent with the reduced SMRT binding activity shown in Fig. 5 (A and B).  16). Surprisingly, F249A strongly activates transcription at 1 M AT-RA. This result is consistent with the observation that F249A acquires co-activator association at 1 M AT-RA shown in Fig. 5D.
We further examine the activity of the mutants on a DR5containing reporter construct (Fig. 5F). In the presence of AT-RA, the expression of the reporter activity is induced in the absence of exogenous RAR␣. Presumably, this activity is derived from the endogenous RARs (lanes 2-4). Exogenous expression of the wild-type RAR␣ further induced the expression of the reporter activity (lanes 6 -8). Intriguingly, we found that although F249A binds ACTR weaker than the wild-type RAR␣ at 1 M AT-RA as shown in Fig. 5D, the transcription activity of F249A is comparable with that of wild-type RAR␣. Furthermore, at lower AT-RA concentration, expression of K262A resulted in lower reporter activity than that in the absence of AT-RA (lanes 14 and 15 compared with lanes 2 and 3), suggesting that K262A inhibits endogenous RAR activity and that K262A functions as a dominant-negative mutant of the wildtype RAR␣. DISCUSSION Site directed mutagenesis was used to dissect the mechanistic links between co-repression, co-activation, and ligand binding. Our results indicate that residues within the receptor signature motif contribute in specific ways to both co-repressor and co-activator binding (Table I and Fig. 6). One unusual mutation is K262A, which results in increased co-activator and reduced co-repressor binding but functionally acts as a dominant-negative mutation. This implies an altered co-repressor off rate leading to the suggestion that co-repressor release is dominant to co-activator binding. This in turn implies that the principle role of ligand is to induce co-repressor release, enabling the signature motif to attract the co-activator. Taken together, these data strongly suggest that transcriptional repression and activation are mechanistically linked.
Recent structural studies have demonstrated that helices H7, H9, H10, and H11 of RAR␣ are involved in heterodimerization with RXR␣ (Ref. 26 and Fig. 6A). Our results are consistent with their findings because helices 3 and 4 mutants possessed at least 60% of the heterodimerization activity in M2H assays of the wild-type RAR␣. The exceptions were F249A and Q257A, which are absolutely conserved within class II nuclear receptor (Fig. 1A). We also note that both F249A and Q257A fail to bind SMRT. It is possible that the global structure of mutants F249A and Q257A is dramatically disturbed, so that they lose most of the LBD-associated activities. Gln 257 is located at the middle of helix 4, and both of its amide groups can donate hydrogen bonds, whereas the carbonyls of Leu 252 , Thr 250 , and Phe 249 can accept the hydrogen bond (not shown). Phe 249 is positioned in the middle of the loop connecting helices H3 and H4. Thus, mutations of either Gln 257 to Ala or Phe 249 to Ala will abolish these hydrogen bonds and very likely destabilize helices 3 and 4 and the whole structure of the molecule. Both F249A and Q257A mutations create large cavities (Fig. 6, C-E), and as a consequence the protein structure may tend to relax. This could result in significant structural rearrangements, indirectly altering the dimerization interface involving helices H7, H9, H10, and H11. In addition, mutation of Phe 249 to Ala would probably result in rotation of the Gln 257 side chain as modeled by Swiss-PdbViewer (Fig. 6E) and also very likely destabilize the loop connecting helices H3 and H4. One might speculate that Phe 249 and Gln 257 are part of the folding nucleus. Folding nuclei are known to include conserved amino acid residues (29 -31). Mutation of the residues critical for the formation of the nucleation core could also result in folding defects. However, the exact structural changes caused by Gln 257 to Ala as well as Phe 249 to Ala mutations have yet to be determined. Although M2H assays indicated that F249A did not interact with RXR␣, EMSAs showed that F249A was able to associate with RXR␣, albeit with much less affinity than that of wild type. This discrepancy could be due to the fact that RAR␣ LBD was used in M2H assays, whereas full-length RAR␣ was used for the EMSA assays. One possibility is that the RAR␣ DNA-binding domain may contribute to heterodimerization with RXR␣ on DNA.
We have previously shown that SMRT coreID association is highly sensitive to mutations within the LBD of TR␤ (16). All of the TR␤ mutants examined failed to bind SMRT ID. Our data show that most RAR␣ helices 3 and 4 mutants, with the exception of I254A and K262A, failed to bind SMRT coreID II (Fig.  2C). These results imply the existence of a common co-repressor-interacting surface between the LBD of RAR␣ and TR␤. Intriguingly, we found that some of these mutants, such as S232A, K244A, and G248T, failed to bind coreID II and could still associate with SMRT ID I ϩ II. These data suggest that the sequence flanking coreID II contacts these residues and stabilizes the interaction between SMRT and RAR␣. However, mutants including V240A, F249A, Q257A, I258A, and L261A failed to bind both SMRT coreID II or SMRT ID I ϩ II. One common characteristic of these residues is that they are highly conserved among nuclear receptors. These data suggested that The results were summarized from M2H assays (co-regulator association) and transient transfection assays (transcriptional activity). The activity of the wild-type RAR␣ is shown as ϩϩϩ. The activity of the mutant RAR␣ is denoted: ϩϩϩ, 80-100%; ϩϩϩ/ϩϩ, 50-80%; ϩϩ, 30-50%; ϩϩ/ϩ, 10-30%; ϩ, 5-10%; ϩ/Ϫ, 0-5%; Ϫ, 0%. The percentage is defined as mutant activity/wild-type activity. The assays were carried out in the absence of AT-RA (SMRT association and repression) or presence of 100 nM AT-RA (ACTR association and activation). Note that V240A, I258A, and L261A bind ACTR weakly (only 5%), but their activation activity is only partially impaired (50 -80%). In contrast, wild-type RAR␣ and mutant K262A bind ACTR equally well, but K262A only moderately activates transcription. We found that at 100 nM, K262A still associates with SMRT.
ϩϩϩ ϩϩϩ/ϩϩ ϩϩϩ ϩϩ ϩϩ/Ϫ these conserved residues are absolutely essential for co-repressor interaction. Among all of the mutants analyzed, K262A displays characteristics that are distinct from others. First, K262A binds SMRT more poorly than wild-type RAR␣ both in EMSA and M2H assays. Consistently, Gal4-RAR␣ (K262A) repressed transcription to a lesser extent than the wild-type RAR␣. Second, K262A did not respond to ligand properly. Dissociation of K262A with SMRT required a high concentration of AT-RA (1 M). This aberrant ligand responsiveness could be derived from a differential ligand binding affinity of K262A. However, previous studies on RAR␥ and RXR␣ suggested that ligand binding, dimerization, and DNA binding of this mutant were not affected (23,27). Intriguingly, unliganded K262A binds ACTR better than that of wild-type RAR␣ in EMSA assays. Similar to the wild-type RAR␣, addition of ligand up to 100 nM enhanced the association of K262A with ACTR. However, at 1 M of AT-RA, the association of K262A with ACTR was only slightly enhanced. These results suggest that Lys 262 has a dual function, preventing ligand-independent association of the co-activators and promoting ligand-dependent dissociation of the corepressors. The distinct co-regulator association profiles of K262A are also reflected in its ability to regulate transcription. In fact, K262A functions as a dominant-negative mutant at low concentrations (Ͻ100 nM) of AT-RA (Fig. 5F, lanes 2, 3, 14, and   16). At 1 M AT-RA, K262A is able to activate transcription, although the level of activation is much lower that that of the wild-type RAR␣. It is apparent that loss of SMRT binding, as shown in M2H assays and EMSAs, correlates with loss of repression activity by F249A (Fig. 2). A similar observation has been reported for mutant F249R (28). Similar to our findings, this mutant did not activate transcription at 50 nM AT-RA. However, our results indicate that even though ACTR association of F249A is not as strong as that of the wild-type RAR␣ at 1 M AT-RA (Figs. 3, B and C, and 5D), F249A activated transcription equally well, if not better (Fig. 5, E and F). This is likely due to the fact that F249A does not associate with a co-repressor whose dissociation is a prerequisite for co-activator association and that the overall transcriptional activation activity depends on both the dissociation of co-repressors and association of co-activators. Two additional results suggest that the limiting step for transcriptional activation by RAR␣ is the dissociation of the co-repressor. First, although V240A, I258A, and L261A bind less efficiently to ACTR than the wild-type RAR␣, the ligand-dependent transcriptional activation by these mutants is only partially inhibited. We propose that this observation is largely due to the loss of SMRT binding activity. Second, unliganded K262A binds ACTR better than wild type in EMSAs (Fig. 5C), but unliganded K262A was unable to activate transcription in transient transfection assays (Fig.   FIG. 6. Ribbon diagrams of RXR␣ and RAR␣ based on the crystal structure of liganded RXR␣/RAR␣ heterodimer (Protein Data Bank code 1DKF) (26). A, the dimeric arrangement of RXR␣/RXR␣ heterodimer is viewed perpendicular to the dimer axis. Mutated residues of RAR␣ are represented in red and gray (Gly 248 and Phe 249 depicted in gray are behind Ile 254 in this view) and denoted with residue number (van der Waals' radii of the side chains are shown). RAR␣ helix H3 is depicted in blue, helix H4 is in pink, and helix H12 is in yellow. RAR␣ helices H7, H9, H10, and H11 forming in part the heterodimer interface are depicted in light green. RXR␣ helices H7, H9, H10, and H11 as well involved in heterodimer interface formation are depicted in banana green. Wire frame models of bound ligands are depicted in gray. Unmodeled in 1DKF region between H1 and H3 is not shown. B, structural model of RAR␣ rotated ϳ90°relative to A. Mutated residues are denoted as in A; helix H3 is depicted in blue, helix H4 is in pink, and helix H12 is in yellow. C, structural model of RAR␣. Orientation is the same as in B. Only Gln 257 and Phe 249 are depicted by their side chain van der Waals' radii and residue number. D, structural model of RAR␣ viewed as in C. Gln 257 is mutated to Ala. E, structural model of RAR␣ viewed as in C. Phe 249 is mutated to Ala. Molecular simulations were performed using GlaxoSmithKline R & D Deep View v3.7 Swiss-PdbViewer. 5E). We interpret this to mean that in M2H assays, K262A is strongly associated with co-repressor complexes in the absence of ligand. Indeed, at 100 nM, K262A still remained associated with SMRT (Fig. 5, A and B) and functioned as a dominantnegative mutant (Fig. 5F). Although the co-regulator association property is not identical to that of PML-RAR␣, the transcription activity of K262A is reminiscent of PML-RAR␣ at low concentrations of AT-RA. In summary, we conclude that helices 3 and 4 of the RAR␣ LBD play a critical role in RAR␣ functions and that co-repressor release is a prerequisite for co-activator association and consequently transcriptional activation.