Identification of amino acids critical for the DNA binding and dimerization properties of the human retinoic acid receptor alpha. Importance of lysine 360, lysine 365, and valine 361.

Retinoic acid receptors (RARs) and retinoid X receptors (RXRs) activate target genes by binding to retinoic acid response elements (RAREs) as heterodimeric, asymmetrical complexes, and display a high degree of cooperativity in binding to RAREs. We have examined here the effect of lysine, cysteine, arginine, histidine, and tyrosine side chain chemical modification on the DNA binding, homo- and heterodimerization properties of the full-length human retinoic acid receptor α (hRARα). Lysines are the only residues to be engaged in the dimerization with human retinoid X receptor α (hRXRα) in the absence of DNA, whereas histidines are selectively involved in the homodimerization of hRARα in the presence of a RARE. Arginine modification affected the DNA binding activity of each type of dimer, whereas cysteines and tyrosines were primarily involved in the homo- or heterodimerization process in the presence of the same RARE. Modified lysines, interfering with the dimerization with hRXRα, were identified by receptor labeling and peptide mapping. They are located in the hormone binding domain eighth heptad repeat, at positions 360 and 365. In keeping with these results, mutation of Lys360, Val361, and Lys365 diminished strongly the DNA binding activity of hRARα as a homodimer or a heterodimer. Our results thus provide direct evidence for the differential involvement of basic, polar, or aromatic amino acids in the DNA binding, homodimerization, and heterodimerization properties of hRARα. Furthermore, they demonstrate the use of distinct dimerization interfaces and identify the type of amino acids involved in these protein-protein interactions.

display strong sequence homology in their DNA binding domain (DBD) with other members of the nuclear receptor superfamily, whereas the ligand binding domain (HBD) appeared to be poorly conserved (for review, see Refs. 1 and 2 and references therein). Nuclear proteins, initially described as factors able to potentiate the DNA binding activity of RAR␣ by establishing protein/protein contacts in vitro (3), were found to be homologous to RXR␣ and RXR␤ (4,5). The observed cooperativity in binding to DNA and transactivation is not limited to the RXR-RAR pair, but extend to VDR, T 3 R (4, 6 -9), proxisome proliferator-activator receptor (9), and orphan receptors (10). RAR/RXR heterodimers bind to retinoic acid response elements (RAREs) that consist, in most cases, of a direct repeat (DR) of the sequence PuGGTCA spaced by five nucleotides (9,11). However, RXR/RAR heterodimers can bind to RAREs with a spacing ranging from two to five nucleotides (12,13), and to core recognition sites arranged into palindromes, inverted palindromes (14) and inverted repeats.
DBDs of RAR␤ (15) and RXR␣ (16) contain, like the glucocorticoid receptor DBD (17), two perpendicular ␣-helical structures located at the C-terminal end of each zinc finger (noted CI and CII). The first ␣-helix maps to the so-called P box, and establishes direct contacts with base pairs in the major groove of the core recognition motif, whereas the second ␣-helix lies perpendicularly to the first one. The D box, located at the N terminus of the second zinc finger, is involved in the heterodimerization of RAR and T 3 R with RXR when bound to DR5 and DR4, respectively (18). Spacing restrictions are imposed by the interaction of the D box of RXR and DR boxes of T 3 R and RAR. The location of DR boxes is different according to the spacing of the two half-sites of the response element (18 -21). Although protein-protein interactions between DBDs of each dimerization partner are necessary and sufficient to impose the binding repertoire of the dimer (22), hormone binding domains (HBDs) are also taking a notable part in the dimerization process. Nine heptad repeats were identified in RAR, VDR, and T 3 R HBDs, and were predicted to be organized in a leucine zipper-like structure (23). The ninth heptad repeat is required for heterodimerization and different interfaces are involved in the absence or the presence of the cognate ligand (24). Amino acids located in this region of RAR, T 3 R, and RXR have a critical role in regulating the ligand-dependent homo-and heterodimerization properties of these receptors, and may even be determining the binding of RXR/RAR dimers to a DR1 element (22). The first and fifth heptad repeats have also been shown to be required for heterodimerization of hRAR␣ with hRXR␣ (25,26).
All of the studies mentioned above demonstrate that multiple interfaces are available to enable homodimerization or heterodimerization of RAR, T 3 R, VDR, and RXR. Chemical modification of proteins by specific reagents is an useful tool to identify critical residues for a given function of a protein (for review, see Ref. 27 and references therein). We and others have used cysteine-specific reagents to assess the function of these residues in the ligand binding activity (28 -31) or in the activation process of the GR (31,32). We also used a similar approach to study the ligand binding properties of hRAR␣ (33). In this report, we have assessed the contribution of cysteines, arginines, lysines, histidines, and tyrosines to the DNA binding, homodimerization, and heterodimerization of hRAR␣. We found that these residues are involved differentially in the different activities of the receptor. Functional, exposed NH 2 groups from lysyl residues were mapped to the hormone binding domain of the receptor and more precisely to the eighth heptad repeat motif. Our method therefore defines a general strategy to identify amino acids critical for nuclear receptor functions.

Plasmids and Bacterial Strains
The plasmid pHK1, containing the cDNA of hRAR␣ (34), was obtained from V. Giguere and R. M. Evans (Salk Institute, H.H.M.I., La Jolla, CA.). The pQE-9 vector was obtained from Diagen Gmbh (Dusseldorf, Germany). The hRAR␣ cDNA was obtained by polymerase chain reaction amplification and inserted into the pQE-9 vector as a BamHI-HindIII fragment, in order to generate an in-frame fusion protein made of a histidine tag followed by the sequence coding for the receptor. The identity of the amplified cDNA was confirmed by restriction mapping and sequencing. Furthermore, microsequencing of BrCNgenerated fragments of the purified protein yielded the expected amino acid sequences. Histidine residues do not to interfere with the ligand binding properties of hRAR␣ (35). F1-hRXR␣ was created by inserting in frame the hRXR␣ cDNA (36) as a HindIII-BglII fragment into the expression vector pF1 (IBI-Kodak, Rochester, NY) containing the Flag epitope.
DH5␣ (Life Technologies, Inc.) cells were used for routine subcloning procedures, and M15 or SG 13009 (Diagen) bacterial strains containing the Rep4 plasmid coding for the lac repressor were the host cells for overexpression of His-tagged hRAR␣. JM109 cells were used to overexpress F1-hRXR␣.

In Vitro Transcription and Translation
The chicken progesterone receptor (form B), the human glucocorticoid receptor, the human RXR␣ and the human RAR␣ were synthesized by coupled transcription/translation using the Promega TnT lysate system and [ 35 S]methionine (Dupont NEN, 800-1000 Ci/mmol). Protein synthesis efficiency was monitored by trichloroacetic acid precipitation of an aliquot of the translation reaction.

Receptors Overexpression and Purification
Overexpression of His 6 -hRAR␣-Transformed M15 or SG13009 bacteria were grown in LB broth supplemented with 100 g/ml ampicillin and 25 g/ml kanamycin to an A 600 ϭ 0.7-0.9. Derepression by 1 mM IPTG proceeded for 3 h after which cells were pelleted and washed with 50 ml of ice-cold 1 ϫ PBS. Cells were resuspended in 10 ml of ice-cold buffer PNI 0 ␤ (2 ϫ PBS, pH 7.4, 0.4 M NaCl, 5 mM ␤-mercaptoethanol). The solution was brought to 150 g/ml of lysozyme and incubated 30 min on ice, then adjusted to 0.05% deoxycholate and incubated for 30 min at 4°C. Cells were lysed by two freeze-thawing cycles when required, and the lysate was brought to 20 mM MgCl 2 . The lysate was submitted to DNase I digestion (100 units/ml of extract) until the viscosity had significantly decreased. The homogenate was then centrifuged for 1 h at 100,000 ϫ g at 4°C, and the supernatant was adjusted to 10% glycerol and stored at Ϫ80°C.
Overexpression of F1-hRXR␣-A procedure similar to that described above was followed, except that LB broth did not contain kanamycin and derepression was initiated by 2 mM IPTG for 5 h. RXR was partially purified from bacterial extracts (40-fold) by ammonium sulfate precipitation and Mono-Q ion exchange chromatography.
Affinity Purification of His 6 -hRAR␣-500 l of agarose coupled to nitrilotriacetic acid (NiTA resin, Diagen Gmbh, Dusseldorf, Germany) were poured in a column (1 cm ϫ 10 cm) and packed by gravity flow. The resin was washed by 10 ml of PNGI 0 buffer (2 ϫ PBS, 0.4 M NaCl, 20% glycerol) and 2-5 ml of the bacterial extract were layered on the resin by 500 l of the fraction. The column was then washed successively with 80 ml of PNGI 0 buffer and 200 ml of PNGI 30 buffer (buffer PNGI 0 supplemented with 30 mM imidazole). All steps were performed at 4°C. At this stage, the receptor was eluted from the column by 2 ml of PNGI 200 buffer (buffer PNGI 0 supplemented with 200 mM imidazole) at room temperature. When polypeptides were submitted to microsequencing procedures, an additional purification step was used to obtain the modified receptor purified to homogeneity (see Fig. 2B). After performing the labeling step, samples were separated on an 8% SDSpolyacrylamide gel, and proteins were visualized by reversible KCl staining. The hRAR␣ polypeptide was then electroeluted using a Bio-Trap apparatus (Schleicher & Schuell, Dassel, Germany) in SDS-PAGE running buffer at 160 V overnight.

Electrophoretic Mobility Shift Assay (EMSA)
Oligonucleotides containing the DR5 response elements (see below) and the consensus half-site AGTTCA were end-labeled with T4 polynucleotide kinase and 10 Ci of ␥-[ 32 P]dATP (3000 Ci/mmol). The ability of hRAR␣ to heterodimerize with hRXR␣, homodimerize, or bind to DNA was assayed as follows.
Heterodimerization Assay-3 pmol of purified, native, or modified hRAR␣ were incubated with 10 -20 pmol of partially purified F1-hRXR␣, 20 fmol of the labeled DR5, 1 g of salmon sperm DNA in a binding buffer giving a final concentration of 20 mM HEPES, pH 7.4, 1 mM EDTA, 80 mM NaCl, 1 mM dithiothreitol, and 10% glycerol. Electrophoresis were run in 0.5 ϫ TBE on a 5% nondenaturing polyacrylamide gel run at 150 V for 3 h at room temperature. These conditions were modified when hRAR␣ binding to the DR5 DNA or to the half-site were assayed.
Homodimerization Assay-9 pmol of native or modified receptor were incubated as above with the labeled DR5 probe, except that 0.1 g of nonspecific DNA was used.
DNA Binding Assay-12 pmol of native or modified receptor were incubated with the labeled half-site (20 fmol) in the absence of nonspecific DNA. DNA binding reactions were for 30 min on ice, in a final volume of 20 l. Protein-DNA complexes were then resolved as above except that the stringency of the electrophoresis was reduced (0.25 ϫ TBE at 4°C). Gels were dried and autoradiographed at Ϫ70°C.

RXR Binding Assay
The protocol used is essentially derived from that described in Kurokawa et al. (12) and allowed for the quantitative analysis of His 6 -hRAR␣/ 35 S-hRXR␣ interaction in the absence of a specific DNA response element. Purified His 6 -hRAR␣, modified or not by treatment with a specific reagent, was diluted to a final concentration of 20 g/ml in 1 ϫ PBS and adsorbed (ϳ0.5 g/well) to a Corning 96-well tissue culture-treated plate for 16 h at 4°C. Wells were then washed three times with 1 ϫ PBS and coated for 3 h with a 5% fetal calf serum solution in 1 ϫ PBS to saturate nonspecific binding sites. Wells were washed three times with 1 ϫ PBS, and 10 6 cpm of 35 S-labeled hRXR␣ were loaded in each well and incubated for 90 min at 4°C in 30 l of binding buffer (20 mM HEPES, pH 7.8, 130 mM KCl, 1 mM EDTA, 1 mM ␤-mercaptoethanol, 0.05% Nonidet P-40, and 20% glycerol). Unbound hRXR␣ was removed by four washes with ice-cold binding buffer, and the specifically adsorbed 35 S-labeled hRXR␣ was released by incubation with 50 l of 0.1% SDS, 0.4 M HCl. Radioactivity was quantified by scintillation counting.
Alternatively, cross-linking experiments were performed to control for the effect of reagents on the dimerization of His 6 -hRAR␣ with hRXR␣ in solution. In these experiments, similar amounts of labeled hRXR␣ and purified His 6 -hRAR␣ were mixed and incubated in conditions similar to that used for EMSA, except that Tris-HCl was substituted for HEPES. Dimethyl suberimidate (Pierce) was added to a final concentration of 1 mM for 1 h at 4°C, and the cross-linking reaction was quenched by bringing the mix to 400 mM Tris-HCl, pH 7.4. Products were then resolved on a 8% SDS-PAGE.

Chemical Modification Procedures
A strategy was devised to assess the functionality of hRAR␣ after modification by amino acid-specific reagents and is detailed in Fig. 1. 150 pmol of purified hRAR␣ (5.8 M) were incubated in the presence of an increasing molar excess of the reagent, such that the ratio of the concentration of the reagent to the amino acid content of the receptor varied from 2 to 40. Note that the His 6 -hRAR␣ fusion protein contains 18 Cys, 25 Arg, 27 Lys, 13 His, and 10 Tyr residues. All reactions were carried out in a total volume of 26 l at 12°C, yielding reagents concentrations ranging from 0.012 mM to 13 mM. In all cases, control reactions were performed by omitting only the modifying reagent.
Arginine Modification-The procedure was essentially that described by Yamasaki et al. (37). Briefly, purified hRAR␣ was incubated in the dark for 120 min in the presence of p-hydroxyphenylglyoxal (HPG) in 10 mM Tris-HCl, pH 8.2. The reaction was quenched by bringing the reaction mix to 20 mM arginine.
Lysine Modification-Pyridoxal 5Ј-phosphate (PyrP) was used essentially as described earlier (27). PyrP was allowed to react with the receptor for 50 min in 20 mM HEPES, pH 6.8, in the dark, and the reaction was quenched by bringing the mix to 20 mM glycine. The resulting Schiff bases were then reduced by 5 mM sodium borohydride (NaBH 4 ) for 5 min at 4°C.
Cysteine Modification-N-Ethylmaleimide (NEM) was used in conditions adapted from that previously established in our laboratory (33). Briefly, purified hRAR␣ was incubated in 10 mM Tris-HCl, pH 7.4, for 1 h with varying concentrations of NEM, and the reaction was stopped by the addition of 20 mM ␤-mercaptoethanol.
Histidine Modification-diethyl pyrocarbonate (DEPC) was used in 50 mM sodium phosphate buffer, pH 6.9, as described in Cheng and Nowak (38). In this case, hRAR␣ was, prior to treatment with DEPC, desalted through a Chromaspin column (2 ϫ 0.5 cm, Clontech) to avoid quenching of the reaction by imidazole, which is used to elute hRAR␣ from the NiTA resin. The reaction was for 120 min and was stopped by bringing the solution to 20 mM imidazole.
Control reactions were treated similarly in the presence of vehicle (final concentration 0.2% in ethanol (HPG, NEM, DEPC) or H 2 O/ NaBH 4 (PyrP)). Upon completion of the chemical modification, the receptor was immediately submitted to the various assays as described above.

Receptor Labeling and Peptide Analysis
To further determine the localization of modified lysyl residues within the receptor, purified His 6 -hRAR␣ was labeled with [ 3 H]formaldehyde (DuPont NEN, 25-100 Ci/mmol). Labeled formaldehyde was diluted twice with unlabeled formaldehyde, and increasing concentrations of reagent were used to react with the receptor (0.3-12 mM final concentration). Variations in reagent concentration did not affect the qualitative outcome of the labeling (data not shown). Cyanogen borohydride (NaBH 3 CN) was used as a reducing agent and was added in equimolar concentration just prior to the addition of formaldehyde. The reaction was stopped by addition of NaBH 3 CN (0.5-2.0 mM, according to the concentration of formaldehyde used) and glycine (20 mM final concentration) after a 1-h incubation at 12°C. The receptor was then purified to homogeneity by electroelution as described above (see also Fig. 2B). Samples were lyophilized and dissolved in 70% formic acid. CNBr (Sigma) was added at a final concentration of 50 -100 mg/ml, and incubation proceeded overnight at room temperature. The reaction mix was lyophilized twice to remove formic acid and the excess of CNBr, and peptides were fractionated by reverse-phase high performance chromatography using a C18 Delta-Pack 15-m column (300 ϫ 7.5 mm; Waters, Milford, MA). Peptides were eluted by a linear gradient of H 2 O, 0.1% trifluoroacetic acid to 80% acetonitrile, 0.085% trifluoroacetic acid at a flow rate of 2 ml/min. Fractions were collected every 30 s, assayed for radioactivity by scintillation counting, and subjected to microsequencing analysis when appropriate.

Other Techniques
Western Blotting Procedure-Proteins were resolved on a 10% SDS-PAGE and transferred onto a nitrocellulose membrane. Immunodetection of RAR␣ was performed as described previously using the IBI Enzygraphic Web system (40). The anti-hRAR␣ monoclonal antibody R␣10 was raised against a synthetic peptide from the F domain of hRAR␣ (41) and purchased from Affinity Bioreagents (Neshanic Station, NJ).
Protein Assay-The protein content of receptor preparation was assayed by the Bradford assay (42) using bovine serum albumin as a standard. The percentage of receptor in the preparation was estimated by densitometry of silver-stained gels, using a Hoefer GS300 scanning densitometer.
Oligonucleotides-The following oligonucleotides and their complements, flanked by BamHI and HindIII sites, were synthesized: (i) a DR5 retinoic acid response element from the promoter P2 of the RAR-␤ gene (43) and (ii) a consensus half-site (44). The sequences of these oligonucleotides are as follows: (i) DR5: gatcGGGTAGGGTTCACCGAAAGT-TCAT, (ii) half-site: agctAGGAGGTCAAATGC. Core recognition sequences are indicated in bold.
Site-directed Mutagenesis-Point mutations were introduced in the hRAR␣ cDNA using the ExSite polymerase chain reaction-based site directed mutagenesis kit and all reactions were carried out as suggested by the manufacturer (Stratagene). Oligonucleotides were designed to convert Lys 360 , Val 361 , and Lys 365 into Thr, Gly, and Thr, respectively (mutations are indicated in bold, underlined characters). In addition, a silent mutation was introduced at position 356, converting the CTG codon (Leu) into CTC, thereby introducing a XhoI site used for the initial screening of mutants (indicated in bold). The mutagenic primers sequences were: K360T, 5Ј-CCGCTGCTCGAGGCGCTAACGGTCTA-CGTGCG-3Ј; V361G, 5Ј-CCGCTGCTCGAGGCGCTAAAGGGCTACGT-GCG-3Ј; K365T, 5Ј-CCGCTGCTCGAGGCGCTAAAGGTCTACGTGC-GGACGCGGAGG-3Ј. The second oligonucleotide (OLIG), used to synthesize the noncoding strand, had the following sequence: OLIG, 5Ј-CTCCTGCAGCATGTCCACCCGGTCCGGCTGCTCC-3Ј. Mutations were confirmed by sequencing, and fragments containing the desired mutations were subcloned into pQE9-hRAR␣. In addition, the identity of the overexpressed proteins was established using the anti-hRAR␣ antibody R␣10 (data not shown).
Peptide Sequence Analysis-Microsequencing of peptides was performed on a gas-phase sequencer (Applied Biosystems 470A) using the O3RPTH program. Phenylthiohydantoin derivatives of amino acids were identified on-line as described previously (45).
DNA Sequencing-All reactions were carried out using the Sequanase 2.0 sequencing kit as indicated by the manufacturer (U. S. Biochemical Corp.).

RESULTS
Experimental Strategy-Several experimental approaches can be considered to identify and localize functionally important amino acids of hRAR␣, necessary for either DNA binding or dimerization in the presence or the absence of DNA. Chemical modification of amino acids has some advantages over mutational studies, i.e. studying the protein under its native conformation, and it may also turn out to be a prerequisite for obtaining information about the nature and the localization of functionally important amino acids. Therefore, we used side chain-specific reagents to modify accessible groups in the native, purified hRAR␣, keeping in mind limitations imposed by optimal reaction conditions for each reagent and the stability of the receptor. These conditions were determined in preliminary experiments using crude extracts, and the reversibility of chemical modifications was assessed whenever possible. Indeed, the effects of thiol groups alkylation was reversible upon dithiothreitol treatment when methylmethanesulfonate or dithionitrobenzoate were used in place of NEM (see also Ref. 33).
PyrP of formaldehyde (lysine modification) and of cyclohexanedione (arginine modification) are known to generate shortlived derivatized products. Omitting stabilizing agents, such as NaBH 4 or NaBH 3 CN (lysine) or boric acid (arginine), yielded fully functional receptors in our assays, attesting that such modifications did not cause merely a denaturation of the protein. 2 A strategy was designed to assess the effect of amino acid modification on homodimerization, heterodimerization, and DNA binding properties of hRAR␣ and is described in Fig. 1. His 6 -tagged hRAR␣ was partially purified over a NiTA resin to produce soluble and functional receptor (see below and Fig. 2). The receptor was then treated in the presence or the absence of a side chain-specific reagent as described under "Experimental Procedures." Native or modified receptors were assayed by EMSA for their ability to bind to DNA as a monomer, a homodimer, or a heterodimer with hRXR␣, and the "RXR binding assay" was used to test the capacity of hRAR␣ to heterodimerize, in the absence of DNA, with hRXR␣ ( Fig. 3).
hRAR␣ was purified by a single step procedure using immobilized chelate affinity chromatography, which allowed for the recovery of milligram amounts of native receptor. We obtained routinely receptor preparations judged to be 50% pure with a major contaminant at 27 kDa, which accounted for more than 60% of contaminant proteins ( Fig. 2A, left panel). A proteolytic product of 48 -50 kDa resulted from the truncation of part of the A/B domain and was not retained on the NiTA resin ( Fig.  2A, lanes Wash). The N-terminal sequence of the 27-kDa protein was determined and is MYVASDLVMSNAYQXRT. Therefore this protein is not related to hRAR␣, but displays a strong homology with a bacterial, histidine-rich prolyl-isomerase. 3 The identity of the 52-kDa protein as hRAR␣ was confirmed by Western blot analysis using a monoclonal anti-hRAR␣ anti-body ( Fig. 2A, right panel).
Confirmation of chemical modifications was obtained by SDS-PAGE analysis of hRAR␣ polypeptides purified to homogeneity. Side chain modification could be monitored by an altered electrophoretic mobility of polypeptides, displaying molecular masses in the 50 -56-kDa range (Fig. 2B). This electrophoretic mobility shift parameter was thus used to assess the efficiency of side chain modification for each reagent (which was observed reproducibly for all reagents) and confirmed by radiolabeling with tritiated formaldehyde (see Fig. 6) and NEM, 4 as well as to verify receptor integrity.
Electrophoretic mobility shift assays were used to assess the dimerization properties and the intrinsic DNA binding activity 2 C. Rachez, unpublished results. 3  The purified His 6 -hRAR␣ was submitted to various treatments as described under "Experimental Procedures" and assayed for its dimerization and DNA binding properties. A complete description of the procedure is given in the text.

FIG. 2.
Purification of His 6 -hRAR␣. A, the vector used (pQE-9) contains a synthetic IPTG-inducible promoter driving the histidinetagged hRAR␣ cDNA expression. The receptor, following Escherichia coli extract preparation and purification under native conditions over a NiTA column, was characterized as follows. Crude extracts from IPTGtreated bacteria (Extract), aliquots from the last wash (Wash), from the eluate from the NiTA column (Eluate), as well as the resin itself were resolved on an 8% SDS-PAGE that was either silver-stained (left) or transferred on a nitrocellulose membrane and immunoprobed using an anti-RAR␣ monoclonal antiserum (right). Molecular masses are indicated on the left. The intact receptor migrated as a 52-kDa polypeptide, whereas a proteolytic product, not retained on the affinity matrix and therefore lacking part of the N terminus of the fusion protein, migrated as a 48 -50-kDa species. B, SDS-PAGE analysis of His 6 -hRAR␣ before and after chemical modification by NEM, PyrP/NaBH 4 , HPG, DEPC, and TNM. The native receptor was submitted to the various treatments and further purified as described under "Experimental Procedures." About 10 g of protein were resolved on a 12% SDS-PAGE. Proteins were detected by silver staining of the gel. Molecular masses are indicated on the left.
of the native receptor. The ability of purified hRAR␣ to bind cooperatively with a tagged hRXR␣ on the ␤-RARE probe is shown in Fig. 3 (left panel, lanes2-4). Binding is specific (lanes 4 -6), and shifted complexes contain both RAR and RXR, as shown by the decreased mobility of these complexes observed in the presence of an anti-hRAR␣ antibody (lane 7) or of an anti-Flag antibody (lane 8), respectively. At higher hRAR␣ concentrations, binding of RAR homodimers could be observed on the same response element (Fig. 3, middle panel). Again, binding was specific (lanes 10 -12), and complexes were supershifted in the presence of an anti-hRAR␣ antibody (lane 13). Finally, we were able to refine experimental conditions (low stringency and high receptor concentration (0.6 M)), so that binding of hRAR␣ to a consensus half-site could be monitored. Complexes migrated faster than homodimers and heterodimers (compare lane 15 to lanes 4 and 10), bound specifically to the half-site (lanes [15][16][17] and were supershifted by the anti-hRAR␣ antibody. hRAR␣ is thus able to bind in these conditions to a consensus half-site as a monomer. Quantification of hRAR␣ binding to a DR5 response element as a monomer or a dimeric complex is therefore possible using the assays described above. To estimate the effect of specific amino acid modification of hRAR␣ on its heterodimerization properties in the absence of DNA, we set up a solid phase assay similar to that described by Kurokawa et al. (12), a system more reproducible and more suitable to statistical analysis, when compared to cross-linking experiments that yielded similar results (see Fig. 4D). The ELISA plate was loaded with purified His 6 -hRAR␣ and washed in order to produce a hRAR␣-coated plate. 25 fmol of [ 35 S]methionine-labeled cPR-B, rGR, hRXR␣, or hRAR␣ were then incubated with this matrix. As shown in the Fig. 3, panel B, only hRXR␣ bound strongly to the hRAR␣ matrix, whereas, in similar conditions, only a low amount of hRAR␣ remained associated to the matrix, showing that the stability of RAR/ RXR heterodimers is higher than that of hRAR␣ homodimers. Neither cPR-B, hVDR, nor rGR were able to bind to the matrix, and the interaction of hRXR␣ with the matrix was dependent on the presence of hRAR␣ (Fig. 3C). Thus this solid phase assay demonstrates the specificity and the saturability (Fig. 3C) of the interaction between hRAR␣ and hRXR␣, in the absence of DNA, and mirrors heterodimer formation in solution (Fig. 4D). Differential Involvement of Thiol Groups in hRAR␣ DNA Binding and Dimerization Properties-The effects of increasing concentrations of NEM on the hRAR␣ ability to bind to DNA as a monomer, homodimer, or heterodimer, as well as to interact  7, 8, 13, and 18). B, heterodimerization assay. ELISA plates were coated with purified His 6 -hRAR␣ and incubated with 35 S-labeled RAR, GR, PR, VDR, and RXR as described under "Experimental Procedures." In each case, receptor input was ϳ50 fmol (10 6 cpm), and results are expressed as the radioactivity stably bound to the matrix after four successive washes. Results are expressed as the mean of at least three independent assays for each receptor. C, saturability of the solid-phase assay. Nonspecific binding of labeled hRXR␣ was assayed by adding increasing amount of hRXR␣ to wells coated with bovine serum albumin (triangles). Total binding was measured similarly except that purified His 6 -hRAR␣ was used in place of bovine serum albumin (circles). The specific binding was plotted as the difference between total and nonspecific binding values for each hRXR␣ concentration (diamonds).
with hRXR␣ in the absence of DNA, are presented in Fig. 4. Purified hRAR␣ was treated with a 2-, 4-, 20-, or 40-fold molar excess over the actual Cys content of the receptor, a ratio that yielded NEM concentrations ranging from 0.2 to 4.0 mM. Fig.  4A shows that NEM impeded the formation of RXR/RAR heterodimers on the ␤-RARE in a dose-dependent manner. A complete loss of DNA binding was observed at a NEM/hRAR␣ ratio of 40, and a 50% decrease of this activity occurred at a ratio of 2.5. Similarly, NEM promoted a decrease in homodimer binding to the same response element (Fig. 4, panel B) which was also abolished at a ratio of 40, with a half-maximal effect noted at a ratio of 2.5. The intrinsic DNA binding activity of hRAR␣ was then measured by assaying its capability to bind to a half-site. Again, the receptor displayed a sensitivity to NEM treatment, although in a much less marked fashion, since approximately 50% of DNA binding activity was still detected at a NEM:Cys ratio of 16 (Fig. 4, panel C). Finally, protein-protein interactions with RXR were quantified by adsorption of labeled hRXR␣ on the hRAR␣ matrix after treatment with NEM. A dose-dependent decrease of the interaction of hRAR␣ with hRXR␣ was observed, but this effect appeared much less marked, with a 50% decrease of this activity occurring at a NEM:Cys ratio of 30 (Fig. 4D, upper panel). Chemical crosslinking experiments were run in parallel to validate the solid phase assay (Fig. 4D, lower panel). Similar results were obtained, showing a complete inhibition of heterodimer formation at a NEM:Cys ratio of 40.
Alkylation of cysteines had therefore a differential effect on each activity: homodimerization and heterodimerization in the presence of DNA were equally and highly sensitive to thiol groups alkylation, whereas the DNA-independent hRAR␣/ hRXR␣ dimerization appeared more resistant to this treatment. This is suggestive of a preferential involvement of accessible thiol groups in the stabilization of RXR/RAR heterodimers bound to DNA, and to a lesser extent in the non-DNA-dependent formation of heterodimers.
Direct Evidence for the Distinct Contribution of Arg, Lys, Tyr, and His Residues to the Dimerization and DNA Binding Properties of hRAR␣-The specific effect of the thiol alkylating reagent NEM described above prompted us to investigate the role of other side chain groups using a similar approach. Compounds known to be selective and active in mild conditions in order to preserve the receptor were selected and tested for their effect(s) on hRAR␣ functions. HPG, PyrP/NaBH 4 , TNM, and DEPC were chosen to modify select Arg, Lys, Tyr, and His residues, respectively. Results for each reagent are summarized in Fig. 5A. They are further expressed as the ratio of the concentration of the reagent to that of the target residues Procedures" and the ability of hRAR␣ to form homodimers on the DR5 RARE was analyzed by EMSA. C, effect of NEM treatment on the DNA binding activity of hRAR␣. 12 pmol of native or modified His 6 -hRAR␣ were allowed to bind to a labeled half-site, and complexes were resolved as above. Results were quantified by excision of the radioactive bands and scintillation counting, and are shown in panel A of Fig. 5. D, upper panel, effect of NEM treatment on the heterodimerization of hRAR␣ with hRXR␣ in the absence of RARE. ELISA plates were coated using ϳ1 nmol of native or NEM-modified hRAR␣ and washed to generate a RAR-coated matrix. [ 35 S]Methionine-labeled RXR was then incubated with this matrix, and its specific adsorption onto the RAR matrix was quantified by assaying the radioactivity remaining stably bound to microtiter wells (see "Experimental Procedures"). Lower panel, identical amounts of hRXR␣ and purified His 6 -hRAR␣ were used in conditions similar to that used for EMSA assays, and cross-linked by 1 mM dimethyl suberimidate sulfate. Products were resolved on a 8% SDS-PAGE and identified by autoradiography. Labeled hRXR␣ migrated as a 54-kDa species, whereas the cross-linked product migrated as a 100 -110-kDa polypeptide. A, purified His 6 -hRAR␣; X, labeled hRXR␣; V, vehicle. present in the hRAR␣ preparation required to observe a 50% decrease of the property tested (noted as R 50 , Fig. 5B).
Treatment of peptides containing lysine residues by vitamin B 6 (PyrP) leads to the formation of Schiff bases that are readily hydrolyzed in aqueous solution. A reduction of these Schiff bases by sodium borohydride is necessary to obtain an irreversible derivatization of properly exposed ⑀-amino groups (27,46), and indeed omitting this step, whether PyrP or formaldehyde were used, yielded fully functional receptors. Additionally, treatment by NaBH 4 or NaBH 3 CN alone had no effect on receptor activities, showing that the observed effect is the result of Schiff base reduction. The derivatization of hRAR␣ by PyrP/NaBH 4 yielded a polypeptide that bound poorly to DNA as a monomer and as a homodimer, with observed R 50 of 2.0 for both assays. More surprisingly, PyrP/NaBH 4 was found less efficient at inhibiting heterodimer formation on the ␤-RARE since the measured R 50 was around 10. The DNA-independent heterodimerization of hRAR with hRXR␣ was also strongly affected by PyrP/NaBH 4 treatment, since the R 50 was around 4.0. These results indicate that heterodimerization of His 6 -hRAR␣ with hRXR␣ is able to overcome partially the effect of PyrP/NaBH 4 on the DNA binding activity of RXR/RAR heterodimers and consequently suggest that ⑀-amino groups of lysine residues are less critical for this activity.
The guanidyl group of arginine reacts specifically with HPG between pH 7.00 and 9.00 at 20°C and modify less than 5% of other residues in these conditions (37). Treatment of hRAR␣ with this reagent affected strongly its ability to bind to a DR5 element as a homodimer, whereas binding to DNA as a heterodimer was affected to a lesser extent, with R 50 of 7.5 and 12.5, respectively. The DNA binding activity of the monomeric receptor was found to be equally affected by arginine modification, with a R 50 of 13. Very interestingly, heterodimerization without DNA was much less sensitive under these conditions, since the observed R 50 was in the 40 -45 range. Thus, it is reasonable to conclude that arginine residues are very important for the DNA binding activity of RAR and for RXR/RAR heterodimer formation in the presence of DNA, whereas they do not appear to be implicated in this phenomenon in the absence of DNA. It is interesting to note that, in opposition to PyrP/NaBH 4 -modified RAR, HPG-modified receptor does not see its loss in DNA binding activity partially alleviated upon heterodimerization.
Nitration of tyrosine by TNM is highly specific and selective, although some mild reactivity with cysteinyl and methionyl residues has been reported for some proteins (see Ref. 27 and references therein). This compound was poorly active on the DNA binding activity of RAR␣, with a R 50 above 40, but displayed a better efficiency at inhibiting homo-and heterodimer formation in the presence of a ␤-RARE, with comparable R 50 values of 15 and 17 respectively. TNM treatment affected RXR/ RAR dimerization in the absence of DNA with a 2-fold lower efficiency (R 50 ϭ 30), showing that tyrosine residues have a predominant contribution to the RAR dimerization activity in the presence of DNA.
O-Carbethoxylation by DEPC of histidyl residues located in the active site of various enzymes has been widely used to inhibit their activity. DEPC showed a very strong specificity on the homodimerization of purified His 6 -hRAR␣ (R 50 ϭ 15) but was totally inactive at inhibiting heterodimer formation in the presence of DNA, and showed a mild effect on the DNA binding activity of the receptor to the half-site (R 50 ϭ 34). Heterodimerization in the absence of DNA was also barely affected by DEPC treatment, with a R 50 above 40. This result thus clearly demonstrate that one or several histidyl residues are involved in the homodimerization process, whereas they appeared to be dispensable for heterodimerization with hRXR␣.
Identification of Critical Lysine Residues in the hRAR␣ Polypeptide-Results detailed above establish clearly a major role for lysyl residues in the dimerization of hRAR␣ with hRXR␣, and most notably in the absence of DNA. Formaldehyde also reacts specifically and irreversibly with lysine residues in the presence of NaBH 4 or NaBH 3 CN and can therefore be used in place of PyrP. Indeed, this compound displayed an activity similar to that of PyrP/NaBH 4 on the DNA binding and FIG. 5. Differential effect of amino acid side chains on the DNA binding and dimerization activities of hRAR␣. A, binding properties of hRAR␣ after treatment with increasing concentrations of side chain modifiers (expressed as the ratio of each reagent concentration to that of the targeted amino acid present in the receptor preparation). 1) quantitative analysis of the ability of hRAR␣ to form heterodimers with hRXR␣ in the presence of a DR5 RARE. EMSA were performed as described in Fig. 4. The retarded and free labeled DNA were quantified and results expressed relative to the level of retarded material before chemical treatment (in percent of a control, untreated sample, see lanes V in Fig. 4). Results are averaged from three to five independent assays, with a standard deviation not exceeding 15%. 2) quantitative analysis of the ability of hRAR␣ to homodimerize on a DR5 RARE. EMSAs were run as described above, as well as quantitation and plotting of the experimental data. 3) quantitative analysis of the DNA-binding activity of hRAR␣. Purified, modified hRAR␣ was allowed to bind to a single half-site and complexes were resolved by EMSA (see also Fig. 4). 4) quantitative analysis of the strength of the RXR/RAR interaction after selective chemical modification. hRAR␣ was purified on a NiTA affinity column and modified as described above. Native polypeptides were adsorbed to microtiter wells to generate a RAR-affinity matrix which was subsequently incubated in the presence of 10 6 cpm of 35 S-hRXR␣. Values are expressed as the mean of triplicate points from two independent experiments. B, relative efficiency of amino acid modifiers to inhibit the dimerization and DNA binding activities of hRAR␣. The fold excess of reagent necessary to observe a 50% inhibition of the activity tested was drawn from data points shown in panel A and is called R 50 . When a 40-fold excess of reagent was not sufficient to inhibit the response by 50%, the R 50 is indicated as Ͼ40. NA indicates that no significant inhibition was observed at a 40-fold molar excess. dimerization properties of hRAR␣, although R 50 values were found to be 8 -10 times higher than those obtained with PyrP. 5 This lower efficiency may be related to the lower steric hindrance provided by this reagent, when compared to the bulky aromatic PyrP molecule. Tritiated formaldehyde was therefore used to modify properly exposed ⑀-amino groups of the receptor, which was then cleaved at methionine residues using cyanogen bromide in acidic conditions. Fourteen fragments are predicted to be generated by such a cleavage, and peptides were resolved using reverse phase chromatography (Fig. 6). As shown in Fig.  6A (upper panel), the A 220 profile identified 12-14 peaks which were observed reproducibly with different receptor preparations. Most of the detected radioactivity (Ͼ90%) coeluted with peptides at ϳ40% acetonitrile (Fig. 6A, lower panel), although they were poorly separated. Three labeled peaks (9, 10, and 11) were identified reproducibly in this region. These fractions were isolated, repurified by reverse phase HPLC, and cleaved by endoproteinase Arg-C. Digested samples were fractionated, yielding profiles shown in Fig. 6B. This second chromatography step showed that peptides present in peaks 9, 10, and 11 could be resolved into two species with similar retention times, but different relative labeling rates. Thus this result shows that peptides 9, 10, and 11 are identical but labeled differentially by tritiated formaldehyde, thereby affecting the hydrophobicity of the molecule. This hypothesis was further confirmed by microsequencing analysis that identified the Leu-Gln-Glu-Pro-Leu sequence as the N terminus of these peptides. This sequence corresponds to a fragment of hRAR␣ mapping from Leu 351 to Met 373 including the eighth heptad repeat (23) and two lysines separated by a Arg-C cleavage site (Fig. 8).
Mutation of Lysine 360, Valine 361, and Lysine 365 of hRAR␣ Diminishes Its DNA Binding Affinity-The contribution of Lys 360 and Lys 365 to the dimerization activities of hRAR␣ was further tested by site-directed mutagenesis. These residues were converted into threonine, generating two receptor mutants referred to as K360T and K365T. In addition, structure predictions (see Fig. 8B) showed that Val 361 could also form part of the dimerization interface of the receptor, and this was similarly tested by mutating Val 361 into a glycine (V361G). As shown in Fig. 7, all mutated receptors displayed, at similar receptor concentrations (Fig. 7A), a lower affinity for the DR5 RARE, whether they were allowed to homodimerize (Fig. 7B) or to form heterodimers with hRXR␣ (Fig. 7C). Quantification of these results showed that mutant K360T had an affinity decreased by 75% for the DR5 RARE when binding as a homodimer, when compared to the wild type molecule. Similarly, V361G and K365T bound this probe with a 2-fold decreased affinity. The impact of mutations was different when hRAR␣ mutants bound to the same probe as heterodimers with hRXR␣: Lys 360 appeared to be involved to a similar extent in heterodimer formation than in homodimer formation, since its mutation into a threonine reduced its binding efficiency by 75%. In contrast, mutation of Val 361 and Lys 365 induced a less pronounced effect on heterodimer binding to the DR5 probe, this property being especially marked for Val 361 (Fig. 7C). These data are thus in perfect agreement with chemical modification data and show that Lys 360 and Lys 365 are equally engaged into homodimer and heterodimer formation on a DR5 response element, whereas Val 361 is more prominently engaged in homodimerization. All three mutants bound all-trans retinoic acid with an affinity similar to that of the wild-type receptor (K D Ϸ 1-4 nM) (35), demonstrating that the three-dimensional structure of the HBDs of these mutated receptors is not drastically modified. A full characterization of the biological properties of these mutants is in progress.

DISCUSSION
Asymmetrical dimerization interfaces located in the DBD of RAR␣ impose a binding polarity to RXR/RAR heterodimers, such as RXR always binds to the 5Ј half-site of DR5, DR4, and DR2 response elements (12,19). On the contrary, this polarity is inverted on a DR1 response element (18,20,22,44). Other sequences located in the extended C-terminal region of the DNA binding domain (H box) (47) and in the hormone binding domain (24 -26) are involved in the dimerization process. However, the contribution of distinct amino acids to this process has been established for only two hydrophobic residues of hRAR␣ (Met 377 and Leu 384 ), located at the N and C terminus of the ninth heptad repeat (24) (see also Fig. 7). No data are at present available with respect to the contribution of polar (Cys), charged (Arg, Lys, His), and aromatic (Tyr, Trp) amino 5 C. Rachez, unpublished data.
FIG. 6. Localization of exposed lysyl residues in the native hRAR␣ polypeptide. After labeling with tritiated formaldehyde, His 6 -hRAR␣ was purified and cleaved at methionine residues using cyanogen bromide. A, BrCN-generated peptides from hRAR␣ were separated by reverse phase HPLC and detected by following on-line the A 220 (upper panel). Peaks are identified by numbers according to their elution time. The radioactive content of fractions from the RP-HPLC column was assayed by scintillation counting of one-tenth of each 1-ml fraction (lower panel). B, A 220 profile and radioactivity content of fractions 9, 10, and 11 after cleavage by Arg-C. The A 220 of the eluate is shown for each peptide and the radioactive content of each fraction is indicated by a black star for the most radioactive peak, and a empty star for the less radioactive peak. Values for this typical experiment were: peak 9, 350 and 180 cpm; peak 10, 164 and 117 cpm; peak 11, 699 and 75 cpm.
acids to the dimerization activity of this receptor. As described in this study, we have devised a strategy that allows for a precise quantitation of the role of such amino acids in the DNA binding and dimerization activities of hRAR␣ using specific amino acid modifiers.
The use of these reagents first evidenced a peculiar role of exposed cysteinyl residues, which are mostly involved in the binding of dimers to a DR5 RARE. These residues are thus distinct from that implicated in the coordination of zinc atoms in the two zinc-finger motifs of the receptor. Only a few cysteinyl residues can be found in regions potentially involved in the dimerization of retinoic acid receptors in the presence of a response element. Two of them (Cys 148 and Cys 174 ) are, however, flanking the newly described H box (47), which has been shown to be required for specific and cooperative binding of purified DNA binding domains of RAR and RXR to DR2 and DR5 RAREs.
Arginine alteration affected equally well the DNA binding and the DNA-dependent dimerization of hRAR␣. This result is in agreement with their scattered location along the receptor sequence. Indeed, arginine residues are found in the DNA binding domain at sites close to or in P, D, and H boxes and also at the fifth position in the first and sixth heptad repeat of the HBD. The weak activity of HPG on the dimerization in the absence of DNA would rather suggest that critical arginine residues are likely to be found in the DNA binding domain.
Tyrosine nitration appeared to impede preferentially the dimerization of hRAR␣ in the presence of a DR5 RARE, whereas the DNA binding and the dimerization without DNA were affected to a lesser extent. This aromatic amino acid is abundant in the HBD, but is more rarely found in the DBD. Again, Tyr 98 and Tyr 100 are located in a critical region for dimerization, the DR box mapping from amino acid 90 to 102-104. A third tyrosine has also a potentially critical location, at position 122 and therefore very close to the D box of hRAR␣. Tryptophan did not display any significant role in the various activities tested (data not shown), in agreement with the location of this unique tryptophanyl residue in the ligand binding pocket (48).
Carbethoxylation of histidines revealed that these residues are specifically involved in the homodimerization activity of hRAR␣, but not in heterodimerization with hRXR␣ in the presence of DNA. This finding is particularly interesting when considering the position of these residues. As DEPC was poorly active on other activities (DNA binding, heterodimerization), a probable location of the functionally important residue(s) is in region(s) involved in dimerization outside of the HBD. Likely candidates are located at position 99 in the DR box and 125 in the D box, and therefore in close vicinity with the already mentioned tyrosine residues at position 98, 100, and 122.
The use of PyrP/NaBH 4 and of formaldehyde/NaBH 3 CN evidenced a peculiar role of the ⑀-amino group of lysines. Indeed, these groups appeared to be involved strongly in the stabilization of the interaction of hRAR␣ with hRXR␣ in the absence of DNA, and were as such unique. Since the hormone binding domain is suspected to play a major role in this case, we scrutinized its sequence to locate lysine residues. Very interestingly, these residues can be divided into two groups: (i) conserved among members of the nuclear receptor family known to dimerize with RXR (VDR, T 3 R, RAR) and (ii) unique to RAR. Another salient feature of these amino acids is their location along the primary sequence of RAR, since they are clustered at the N and the C terminus of the E domain, and are, in the latter case, located in the eighth and ninth heptad repeats, which were proposed as putative structural subdomains of the HBD (27) (see Fig. 8). The two labeled lysines residues were indeed found in the eighth heptad repeat (Lys 360 ) and between the eighth and ninth heptad repeat (Lys 365 ). Very interestingly, the crystal structure of hRXR␣ HBD homodimers revealed a crucial role of this particular region. It contains helices 9 and 10, which encompass the eighth and ninth heptad repeat motifs, respectively, in which Lys 405 (helix 9) and Lys 417 (helix 10) establish salt bridges between the two RXR monomers (49) (see also Fig. 8A). These two RXR lysine residues are located in a position analogous to that of hRAR␣ Lys 360 and Lys 365 (Fig. 8A). Molecular modeling of this region, based on the hRXR␣ and hRAR␥ HBD structures, suggests that these two lysines are located in an ␣-helical structure and oriented in such a way that they could form a dimerization interface, together with Val 361 (Fig. 8B). Thus a role similar to that of hRXR␣ Lys 405 and Lys 417 can be envisioned for hRAR␣ Lys 360 and Lys 365 in the dimerization process, a hypothesis significantly strengthened by mutagenesis data. Indeed, Lys 360 ap- The wild type (WT) and mutant receptors were overexpressed in E. coli and partially purified by NiTA affinity chromatography. Chromatographic fractions were analyzed by Western blot analysis, and receptor content was estimated by densitometry scanning of the membrane. Results are calculated and expressed relative to the amount of the wild type receptor. B, homodimerization properties of K361T, V361G, and K365T. The ability of each receptor to bind as a homodimer to the DR5 ␤-RARE probe was tested by EMSA as described under "Experimental Procedures." The amount of receptor binding to the probe was estimated by excision of the bands and scintillation counting. Binding activities are calculated relative to the DNA binding activity of the wild type receptor and to receptor concentration in the preparation. C, heterodimerization properties of K361T, V361G, and K365T. The ability of each receptor to bind as a heterodimer to the ␤-RARE in the presence of hRXR␣ was tested by EMSA as described under "Experimental Procedures." Results are calculated and expressed as in B. Values shown in bar graphs are average values from three independent experiments. peared to be equally involved into homo-and heterodimerization processes, whereas Lys 365 , and more noticeably Val 361 , are contributing more prominently to the homodimerization activity of hRAR␣. We cannot, however, rule out allosteric effects due to an interaction of this part of the HBD with the DBD of hRAR␣.
Our method allows for the location of potentially important amino acids for the dimerization activities of hRAR␣, and these predictions can be confirmed by site-directed mutagenesis. In addition, this approach can now be extended to other response elements (DR2 and DR1) to distinguish and assay the relative contribution of hydrophobic and charge interactions to the dimerization activities of retinoid receptors, when different interfaces (and therefore different amino acids) are used. Finally, it provides direct evidence for the differential involvement of amino acid side chains in homo-and heterodimerization of hRAR␣.