Mutational Analysis Reveals That All-trans-retinoic acid, 9-cis-Retinoic acid, and Antagonist Interact with Distinct Binding Determinants of RARα*

Retinoids exert their pleiotropic effects on cell differentiation and proliferation through specific nuclear receptors, the retinoic acid receptors (RARs) and retinoid X receptors (RXRs). Two biologically highly active natural retinoids have been identified, all-trans-retinoic acid (t-RA) and 9-cis-retinoic acid (9-cis-RA). The RXRs exclusively bind 9-cis-RA, whereas the RARs bind both isomers of RA with comparable affinity. Recently published results suggest that RARs have the same binding site for t-RA and 9-cis-RA but with different determinants (1–3). Antagonist binding on RARα has been suggested to induce distinct conformational changes in comparison with agonist binding. To elucidate the region minimally required for efficient binding of agonist (t-RA and 9-cis-RA) and antagonist Ro 41-5253 to the RARα, we generated N- and C-terminally truncated mutants of the receptor. Characterization of these deletion mutant proteins using protease mapping and ligand binding experiments revealed that different parts of the ligand-binding domain are necessary for t-RA, 9-cis-RA, and antagonist binding. Three distinct regions of the ligand-binding domain of the human retinoic acid receptor-α are required for binding of t-RA (RARα187–402), 9-cis-RA (RARα188–409), and the antagonist Ro 41-5253 (RARα226–414).

Retinoic acid (RA) 1 has a broad spectrum of biological activities in vertebrate development and homeostasis (4 -6). Due to their fundamental role in the control of cell differentiation and proliferation, RA and synthetic retinoids are clinically very useful, predominantly in the treatment of leukemia and nonmalignant hyperproliferative disorders of the skin. However, retinoids have undesirable side effects such as hypervitaminosis A syndrome and teratogenicity (4,(7)(8)(9)(10). Two biologically active stereoisomers of RA have been identified, all-trans-retinoic acid (t-RA), and 9-cis-retinoic acid (9-cis-RA) (11)(12)(13). The direct biological effects of the retinoids are mediated by the activation of two subfamilies of nuclear receptors, the retinoic acid receptors (RARs) and the retinoid X receptors (RXRs) (13)(14)(15)(16). Both subfamilies consist of three receptor subtypes referred to as RAR␣, -␤, and -␥ (14,15,(17)(18)(19)(20) and RXR␣, -␤, and -␥ (13,16,21). RARs and RXRs belong to the nuclear hormone receptor superfamily, whose members act as ligandinducible transcription enhancer factors (22)(23)(24). On the basis of homology to other nuclear hormone receptors, the sequences of RARs are divided into six distinct regions designated A through F. The C domain constitutes a highly conserved DNAbinding domain, while the E domain confers the ligand binding properties of each receptor (25). RARs seem to operate effectively only as heterodimeric RAR⅐RXR complexes (26 -31), but RXR-independent transactivation by RARs has also been observed (32). The RXRs are able to activate genes via homodimers (33,34), but act predominantly as coregulators, enhancing the binding of RA, vitamin D3, thyroid hormone, and peroxisome proliferator-activated receptors to their response elements via heterodimerization (26 -31, 35). t-RA and 9-cis-RA are the natural ligands for RARs (14,36), whereas 9-cis-RA and phytanic acid are the natural ligands for RXRs (11,12,37). Both RA isomers compete with each other for RAR binding (38). A synthetic retinoid Ro 41-5253 has been identified as a selective RAR␣ antagonist (39), which, when bound to the receptor, induces a different conformational change as detected by limited proteolysis (40). Recently, the structure of the cocrystal hRAR␥-t-RA has been determined (1), and modeling of 9-cis-RA in the RAR␥ binding pocket suggests that the two isomers also compete for the same binding site in this receptor. In addition, two residues, Met-406 and Ile-410, have been shown to be critical, specifically for the binding of 9-cis-RA to hRAR␣ (2). Binding experiments using RA and different synthetic ligands have shown different requirements in the C-terminal part of the hRAR␣ but no distinction between t-RA and 9-cis-RA binding specificity (41). It has also been demonstrated recently that Cys-235, Arg-217, and Arg-294 of hRAR␣ play a role in antagonist binding, showing specific requirements for efficient antagonist binding to hRAR␣ (3). As shown by competitive binding experiments, t-RA competes with the antagonist for RAR␣ binding (40). This strongly suggests that the antagonist shares a common binding site with the RA isomers. To analyze the minimal structural requirements for an efficient binding of either t-RA, 9-cis-RA, or Ro 41-5253 to the hRAR␣, we generated N-and C-terminally truncated RAR␣ mutants. Characterization of these deletion-mutant proteins using protease mapping and ligand binding experiments revealed that different parts of the ligand-binding domain of RAR␣ are necessary for t-RA, 9-cis-RA, and antagonist binding. * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 1 The abbreviations used are: RA, retinoic acid; RAR, retinoic acid receptor; RXR, retinoid X receptor; t-RA, all-trans-retinoic acid; 9-cis-RA, 9-cis-retinoic acid; hRAR␣, human retinoic acid receptor ␣; hRAR␣-LBD, human retinoic acid receptor ␣-ligand-binding domain; RAR␣, retinoic acid receptor ␣; aa, amino acid; wt, wild type.  14 C-methylated protein molecular weight markers, and Amplify were obtained from Amersham. Restriction enzymes were obtained from Boehringer Mannheim. Trypsin (type I: from bovine pancreas) was purchased from Sigma.

Materials-[
Plasmids-The human RAR␣ cDNA insert of pT7-RAR␣ (42) was subcloned as a MscI-BamHI fragment into pSG5-MscI giving pSG5-RAR␣. pSG5-MscI was constructed by replacing the EcoRI-BamHI fragment of pSG5 (43) with a linker containing an MscI site. RAR␣ deletion mutants were produced by PCR. The correctness of the sequence was confirmed by sequencing. For N-terminal deletions: removal of the EcoRI site of pET3a (44) was carried out by restriction with EcoRI, fill-in reaction and religation, yielding pET3a-⌬E. The linker (5Ј-TAT-GGAATTCTAGCGCTTA-3Ј), including a start codon and a subsequent EcoRI site, was ligated into the NdeI site of pET3a-⌬E giving pET3a-E. pSG5-M was constructed by ligation of the linker (5Ј-AATTGCTAC-CACCATGGAATT-3Ј) in the EcoRI site of pSG5. The mutants were generated by PCR using pT7-RAR␣ as template. The PCR products were restricted with EcoRI-BamHI (located in the 5Ј primer and 3Ј primer, respectively) and cloned into pET3a-E (expression in Escherichia coli for binding studies) or pSG5-M (in vitro translation for limited proteolytic digestion). For C-terminal deletions, the mutants were generated by PCR using pT7-RAR␣ as template, 5Ј primers (positions 466 -486), and 3Ј primer containing a BamHI site. The PCR products were restricted with SacI-BamHI and cloned into either the full-length pSG5-RAR␣ restricted with SacI-BamHI (in vitro translation for limited proteolytic digestion) or the truncated pET3a-RAR␣-⌬155 restricted with SacI-BamHI (expression in E. coli for binding studies).
Binding Assays, K d and IC 50 Determinations-RAR␣ deletion mutants were expressed in E. coli BL21(DE3)pLys (44). Expression and solubility problems occurred with the N-terminal deletion mutants hRAR␣-⌬188, -⌬189, -⌬225, -⌬226, and -⌬227 expressed in E. coli. These truncated proteins were synthesized with the TNT® coupled reticulocyte lysate system (Promega) for the determination of the IC 50 values. For these last mutant receptors, a 40 -60% specific binding was sufficient to allow the IC 50 determination. Since K d values, for either t-RA or 9-cis-RA, have been established for the C-terminal point mutants of hRAR␣ (aa 405-419) (2), we determined the binding activity of the C-terminally truncated receptors by Scatchard analysis for the same ligands. From the N-terminally truncated hRAR␣ (aa 187-227), no binding activity was determined previously, and we verified the binding activity by IC 50 determination. The binding assays (K d and IC 50 determinations) were performed as described elsewhere (3).
In Vitro Transcription, Translation, and Limited Proteolytic Digestion-Human RAR␣ (wt or truncated) in pSG5 was in vitro transcribed and translated in the presence of [ 35 S]methionine by using rabbit reticulocyte lysates as specified by Promega. The limited proteolytic digestions were performed as described elsewhere (40).

Binding of t-RA to the N-and C-terminally Truncated RAR␣
Mutants-An important region for the t-RA binding (aa 186 -198) was defined previously in hRAR␣ by a N-terminally truncated receptor (41). A C-terminal deletion to position 403 resulted in a moderate decrease in affinity for the same ligand (41). On the other hand, a C-terminal deletion to position 404 of hRAR␣ exhibited a K d of 0.3 nM for t-RA (2). To determine the precise region of the hRAR␣-LBD required for an efficient t-RA binding, we used binding assays and protease mapping to probe t-RA binding to N-and C-terminally truncated hRAR␣-LBD ( Figs. 1 and 2).
A K d of 1.1 nM for t-RA was obtained with the hRAR␣-LBD (aa 155-462) ( Table I). This value is in the range of the K d published for the full-length hRAR␣ expressed in E. coli or COS cells (0.67 nM (42), 1.7 nM (45)), and also for the hRAR␣-LBD expressed in E. coli (6.2 nM (46); 0.6 nM (3)).
Among the N-terminally truncated receptors, the hRAR␣-⌬187 exhibited good protection against trypsin digestion when bound to t-RA. A strongly reduced signal was observed for the digestion of the hRAR␣-⌬188 truncated receptor under the same conditions. At the C-terminal end of the receptor, the same tendency was observed with the two mutants hRAR␣-⌬402 and hRAR␣-⌬401. In regard to the binding activity of these mutant receptors, the N-terminally truncated hRAR␣-⌬187 exhibited a K d value 4.5-fold higher than the wild type hRAR␣-LBD, whereas a deletion of a single further amino acid abolished the binding activity in such a way that neither K d nor IC 50 could be determined (Table II). Similar experiments were carried out using N-terminally truncated mutants of hRAR␥. We found that hRAR␥-⌬189 (hRAR␣-⌬187) was able to bind t-RA with an affinity comparable with the full-length hRAR␥, whereas hRAR␥-⌬190 (hRAR␣-⌬188) showed no detectable binding activity to this ligand (data not shown). Therefore, it appears that, while the C-terminal region of the D domain of RAR␣ (aa 188 -199) is required for efficient binding of the t-RA, the C-terminal part of the E domain (aa 403-419) and the entire F domain can be deleted without affecting the ability of the RAR␣ to bind its natural t-RA ligand (see Fig. 3).
Binding of 9-cis-RA to the N-and C-terminally Truncated RAR␣ Mutants-Two contradictory results have been published concerning the binding of 9-cis-RA to C-terminal deletion mutants of hRAR␣. Lefebvre et al. (41) observed that a deletion up to position 403 of the hRAR␣ expressed in E. coli resulted in a moderate decrease in affinity for 9-cis-RA (41). On the other hand, a small region (aa 405-419) within the ligand binding domain of a truncated hRAR␣ was demonstrated to be required for the 9-cis-RA binding (2). In our study, a K d value of 1 nM was determined for the binding of 9-cis-RA to the hRAR␣-LBD. This value is in agreement with the range of the reported data for the truncated receptors expressed in E. coli (0.3 nM for the hRAR␣-⌬419 (2) and 1.2 nM for the hRAR␣-LBD (3)).
After the limited trypsin digestion of the N-terminally truncated hRAR␣ liganded to 9-cis-RA, the corresponding protec-tion was almost completely abolished for the mutant hRAR␣-⌬189, and a reduction of the signal was already observed for the mutant hRAR␣-⌬188 (Fig. 2). At the C-terminal end of the receptor, the digestion of the hRAR␣-⌬409 liganded to 9-cis-RA by trypsin yielded a band of the appropriate size, whereas the truncated hRAR␣ at the next residue (hRAR␣-⌬408) exhibited only a faint band (Fig. 1). IC 50 and K d determinations for the binding of the 9-cis-RA to the above mentioned truncated receptors confirmed the N-and C-terminal deletion borders of the hRAR␣ concerning the RA isomer. No 9-cis-RA binding was detectable to the hRAR␣-⌬189, whereas a 5-fold higher IC 50 value was obtained for the binding of the t-RA isomer to the hRAR␣-⌬188 in comparison with the hRAR␣-LBD (Table II). At the C-terminal end, a 35-fold higher K d value, was obtained for the hRAR␣-⌬409, while the hRAR␣-⌬408 bound the 9-cis-RA more than 150-fold less efficiently than the hRAR␣-LBD (Table I). From these results, it is clear that the C-terminal region of the D domain of RAR␣ (aa 189 -199) is required for an efficient binding of 9-cis-RA. Furthermore, the C-terminal part of the E domain (up to aa 409) is also crucial for the binding efficiency of this ligand to the hRAR␣ (see Fig. 3).
Binding of  to the N-and C-terminally Truncated RAR␣ Mutants-The antagonist Ro 41-5253 has been reported to induce a different conformational change when bound to the hRAR␣ (40). It has also been shown recently that Ro 41-5253, as well as other antagonists, exhibit distinctly different requirements for efficient binding to hRAR␣ (3). In the present study, we determined an IC 50 value of 16 nM for the binding activity of the antagonist Ro 41-5253 to the hRAR␣-LBD (Table  I). Limited trypsin proteolysis of the N-and C-terminally truncated hRAR␣ bound to Ro 41-5253 led to the definition of the N-terminal deletion border between the residue D226 and the residue Lys-227 and the C-terminal deletion border between the residue L414 and M413 of the E region of the hRAR␣-LBD. Concerning the binding activity of Ro 41-5253 to the truncated hRAR␣s, 24-and 15-fold higher IC 50 values for hRAR␣-⌬227 and hRAR␣-⌬413 have been observed, respectively, in comparison with the hRAR␣-LBD (Tables I and II). Remarkably, the hRAR␣-⌬226 and the hRAR␣-⌬414 mutant exhibited only a 2and 3.5-fold reduction in IC 50 , respectively, in comparison with the hRAR␣-LBD. These results evidence distinctly different requirements of the N-and C-terminal regions of the ligandbinding domain of the hRAR␣ for binding Ro 41-5253 compared with the two other ligands t-RA and 9-cis-RA (see Fig. 3) .   FIG. 2. Effect of t-RA (lanes 11-16),  9-cis-RA (lanes 17-22), and Ro 41-5253  (lanes 23-27) on limited tryptic digestion of RAR␣ wild type (wt) and Nterminal deletion mutants. Controls for wt and truncated receptor expression are in lanes 1-10. The experimental conditions were the same as described in the legend to Fig. 1.

DISCUSSION
It has been shown that the binding of both t-RA and 9-cis-RA induces a different conformational change in hRAR␣ to that observed with the binding of the antagonist Ro 41-5253 (40). At present, several arguments allow one to think that, depending on the chemical structure of the ligands, distinctly different determinants are required for the efficient binding to the RARs. Recently published results also showed distinct determinant requirements for the binding of different ligands to the estrogen receptor (47). The recently described crystal structure of the RAR␥ ligand-binding domain bound to t-RA has clarified the ligand binding interactions. Modeling 9-cis-RA in the hRAR␥ binding pocket revealed that the binding site of this receptor was likely to be very similar to that of t-RA (1). In addition, three residues have been shown to play a significant role in ligand binding (agonist and antagonist) to hRAR␣, whereas Cys-235, Arg-217, and Arg-294 have been shown to be important residues of the hRAR␣-LBD, specifically for antagonist interactions (3). Furthermore, Arg-269 of the hRAR␤ (Arg-276 of hRAR␣ and Arg-278 of hRAR␥) has been shown to be a crucial residue for RA binding (48,49). These findings and the highly conserved amino acid homology of the E domain of the RARs (92%) (23) suggest that the binding conditions of t-RA to the three subtypes of receptors RAR␣, -␤, and -␥ are very similar. It also indicates that these ligands may compete for a unique binding site in the RARs.
In the present study, the 30-fold higher K d value for 9-cis-RA binding to the C-terminal deletion hRAR␣-⌬409 reinforces the role of the residue Ile-410 in the 9-cis-RA binding as shown by Tate and Grippo (2). We have provided evidence by this study that the complete F domain can be removed without disturbing the ability of the hRAR␣ to bind any ligand. The C-terminal part of the E domain of the hRAR␣ (to aa 414) is sufficient for the efficient binding of the antagonist Ro 41-5253, whereas further deletion of the E domain of the hRAR␣ to positions 409 or 402 for 9-cis-RA and t-RA, respectively, does not disturb the ability of the receptor to bind these two ligands. Interestingly, t-RA and 9-cis-RA require the C-terminal part of the D domain of the hRAR␣ (from aa 187 and 188, respectively), while the D domain and the N-terminal part of the E domain (aa 155 to 226) are not required for the efficient binding of the antagonist Ro 41-5253.
From the crystal structure of the hRAR␥-LBD bound to t-RA and the modeled 9-cis-RA in the hRAR␥ binding pocket, less distance was expected between the ligand 9-cis-RA and the C-terminal region of the hRAR␥ (helix-11 and helix-12) in comparison with the t-RA spatial disposition (1). It was also clear that the C-terminal part of helix-1 overlaps with the  Remarkably, all of helix-1 can be deleted without disturbing antagonist binding, whereas this completely abolishes agonist binding. An explanation for this could be that the C-terminal part of helix-1 may stabilize the spatial location of helix-3 required specifically for the binding of the agonists.
Taken together, these results indicate that three different fragments of hRAR␣ are minimally required for the binding of t-RA, 9-cis-RA, and Ro 41-5253. The data shown here suggest that distinct determinants of hRAR␣ are either directly involved in the binding of these three ligands or, alternatively, they are needed to maintain the binding pocket in an appropriate conformation as a prerequisite for the binding of each of the isomers of RA as well as Ro 41-5253.
Limited trypsin digestion of either wild type or truncated receptor yielded the same digested peptides corresponding to the potential cleavage sites (data not shown). The undigested mutant receptors showed positions in the gel that were in accordance with their calculated molecular weights (Fig. 1,  lanes 2-13; Fig. 2, lanes 2-10). Also, the molecular weights of the fragments as determined from the SDS gel were in good accordance with the magnitude of the truncation of the mutants (Figs. 1 and 2). This indicates that the binding of either the agonists t-RA and 9-cis-RA or the antagonist Ro 41-5253 to the truncated receptors induces the same conformational change in all the mutants as that observed with the full-length RAR␣. In this study, we demonstrate that with trypsin digestion of 9-cis-RA-bound full-length hRAR␣, the resulting 30-kDa fragment is less protected against further proteolysis than the one derived from the analogous experiment with t-RA. This was evidenced by the time course of tryptic digestion of RAR␣ bound to t-RA versus 9-cis RA-bound receptor (Fig. 4). These results were further confirmed by tryptic proteolysis with increasing protease concentrations (data not shown). These differences in the proteolytic resistance cannot be explained by distinct affinities of the RA isomers to the RAR␣, because both exhibit K d values in the range of 1 nM (Tables I). The decreased stability of the 30-kDa fragment obtained from RAR␣ bound to 9-cis-RA could reflect the higher off-rate of 9-cis-RA from RAR␣ in comparison with the t-RA. Displacement assays have demonstrated that 9-cis-RA exhibits about 2-fold higher off-rates from murine RAR␣ than t-RA (38). Time course experiment using Ro 41-5253 as ligand yielded also a decreased stability of the 25-KDa antagonist characteristic fragment compared with the stable 30-kDa fragment observed with t-RA (Fig. 4). Incubation of the receptor with Ro 41-5253 or 9-cis-RA instead of t-RA in the digestion assay could result in more unliganded receptors, or fragments, for a short period of time. During this time, they might change to a more relaxed conformation, more accessible to the protease.
In conclusion, we have evidenced that three distinct regions of the of the hRAR␣-LBD are required for the efficient binding of t-RA, 9-cis-RA, and the antagonist Ro 41-5253. t-RA binding requires the region of the hRAR␣-LBD from amino acid 187 to amino acid 402, 9-cis-RA requires the region of the hRAR␣-LBD from amino acid 188 to amino acid 409, while the antagonist requires the region of the hRAR␣-LBD from amino acid 226 to amino acid 414. In vitro synthesized [ 35 S]methionine-labeled RAR␣ was preincubated with Me2SO alone, with 100 nM of either t-RA or 9-cis-RA or with 10 M of the antagonist Ro 41-5253. Trypsin solution was added, giving a final concentration of 100 g/ml. Alternatively, an equal volume of water was added. Incubations were for the indicated time intervals at room temperature. Cleavage was immediately stopped by boiling with SDS sample buffer. Samples were electrophoresed through a SDS-polyacrylamide gel, and the dried gel was autoradiographed. The sizes of molecular weight markers are indicated. The resistant protein fragment occurring in the presence of RA isomers is marked by a diamond, and the resistant fragment characteristic of the antagonist is indicated by a star.