The Ligand Binding Domain of the Human Retinoic Acid Receptor (cid:103) Is Predominantly (cid:97) -Helical with a Trp Residue in the Ligand Binding Site*

Retinoic acid exerts its many biological effects by in- teraction with a nuclear protein, the retinoic acid receptor (RAR). The details of this interaction are unknown due mainly to the lack of sufficient quantities of pure functional receptor protein for biochemical and struc- tural studies. We have recently subcloned the D and E domains of human RAR (cid:103) for expression in Escherichia coli . Using nickel-chelation affinity chromatography with a polyhistidine amino-terminal tail, purification of the DE peptide with a pI of 5.18 was accomplished to greater than 98% purity. Scatchard analysis and fluorescence quenching techniques using the purified protein indicate a very high percentage of functional molecules ( (cid:62) 95%) with a K d for retinoic acid ( t -RA) of 0.6 (cid:54) 0.1 n M . Circular dichroism spectra of the purified domains predict a predominantly (cid:97) -helical structure ( (cid:59) 56%) with little (cid:98) sheet present. No significant changes in these structural characteristics were observed upon binding of t -RA. Inspection of the amino acid sequence within these domains identified a single tryptophan residue at position 227. Modeling the amino acid sequence in this region as an (cid:97) -helical structure indicates that this tryptophan is adjacent to alanine 234, which corresponds to alanine 225 in RAR (cid:98) that has previously been linked to the ligand binding site. Fluorescence

Retinoic acid exerts its many biological effects by interaction with a nuclear protein, the retinoic acid receptor (RAR). The details of this interaction are unknown due mainly to the lack of sufficient quantities of pure functional receptor protein for biochemical and structural studies. We have recently subcloned the D and E domains of human RAR␥ for expression in Escherichia coli. Using nickel-chelation affinity chromatography with a polyhistidine amino-terminal tail, purification of the DE peptide with a pI of 5.18 was accomplished to greater than 98% purity. Scatchard analysis and fluorescence quenching techniques using the purified protein indicate a very high percentage of functional molecules (>95%) with a K d for retinoic acid (t-RA) of 0.6 ؎ 0.1 nM. Circular dichroism spectra of the purified domains predict a predominantly ␣-helical structure (ϳ56%) with little ␤ sheet present. No significant changes in these structural characteristics were observed upon binding of t-RA. Inspection of the amino acid sequence within these domains identified a single tryptophan residue at position 227. Modeling the amino acid sequence in this region as an ␣-helical structure indicates that this tryptophan is adjacent to alanine 234, which corresponds to alanine 225 in RAR␤ that has previously been linked to the ligand binding site. Fluorescence of this tryptophan was quenched in a dose-dependent manner on the addition of t-RA, confirming that Trp-227 is within the ligand binding site. Tryptophan flourescence quenching analysis also demonstrates that a single retinoic acid molecule is bound per receptor and suggests that receptorligand interactions occur within the amino-terminal portion of the predominantly ␣-helical ligand binding domain.
The RARs, like other members of this nuclear receptor superfamily, have a modular structure consisting of six domains denoted A through F (11)(12)(13). The E region, or ligand binding domain, is 85-90% conserved among the RARs and has liganddependent transactivation and dimerization functions (14 -16).
Information pertaining to the interaction between the RARs and their ligand requires large quantities of purified receptor protein in a functional three-dimensional conformation. Several attempts have been made at purification of recombinant full-length RAR␣, -␤, and -␥ expressed in eukaryotic cells (6,17,19), Sf9 insect cells using the baculovirus expression system (18,19), and a variety of expression systems in Escherichia coli (20,21). Expression of subfragments corresponding to the RAR ligand binding domain (i.e. domains DEF, EF, etc.) in E. coli improved yields as compared to expression of full-length receptors (17,22,23). However, in every case, regardless of yield, the purified receptor preparations were insufficiently active with respect to ligand binding and were therefore unsuitable for structural or biochemical analysis. Improved expression and purification techniques described here using nickel chelation chromatography and a modified thrombin cleavage produced milligram quantities of hRAR␥ receptor protein containing the D and E domains. This DE␥ protein was found to have binding kinetics similar to that of the native full-length receptor. Circular dichroism analysis of the expressed protein suggests that the secondary structure of the ligand binding domain is predominantly ␣-helical and contains very little ␤ sheet. This observation is in good agreement with the reported crystal structure of the ligand binding domain of RXR␣, a closely related member of the steroid/thyroid hormone receptor superfamily (24), and that determined from structure prediction modeling. Interestingly, the relative pattern of secondary structural components for this peptide did not change significantly upon binding to ligand.
Inspection of the amino acid sequence within the D and E domains revealed a single tryptophan residue at position 227, which was subsequently used for fluorimetric titration. At a concentration of protein determined by amino acid analysis, fluorescence was quenched in a dose-dependent manner upon addition of t-RA with a transition point equal to the concentration of protein in the reaction. Cogan analysis (33) of this curve indicates a single ligand binding site within DE␥. Further, the position of the transition point indicates that Ͼ95% of the receptor protein is present in an active conformation and that * 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 U.S.C. Section 1734 solely to indicate this fact.
Trp-227 is located in close proximity to the ligand binding site. The assignment of Trp-227 within the ligand binding site is further supported by a previous report that Ala-234 (Ala-225 in RAR␤) is a contact amino acid within the ligand binding site (25). In an ␣-helical structure, this alanine would be expected to be immediately adjacent to Trp-227 supporting a structural model for the ligand binding pocket.

EXPERIMENTAL PROCEDURES
Materials-Expression vector pET15b, host strain E. coli BL21(DE3), restriction grade thrombin and His⅐bind TM resin were purchased from Novagen, Inc. pSG5/hRAR␥ was kindly provided by Dr. P. Chambon (26). Taq polymerase, PCR buffers, and deoxynucleoside triphosphates were obtained from Perkin-Elmer Corp. Amplification was performed in an Ericomp Easycycler TM (San Diego). Ligand binding assays were carried out with t-RA, retinol, and retinal purchased from Sigma and [11, H]t-RA (47.5 Ci/mmol) from Dupont NEN. 9-cis RA was synthesized by Bristol-Myers Squibb Central Chemistry (Wallingford, CT). Retinoids were used under yellow fluorescent light to minimize photodegradation. Oligonucleotides used in the PCR reactions were synthesized by Genosys Biotechnologies, Inc. (Woodlands, TX). Restriction enzymes and DNA ligase were purchased from New England BioLabs, Inc. All other chemicals used were of reagent grade and purchased from Sigma.
Construction of pET15b/DE␥ Expression Vector-The truncated DE␥ receptor cDNA was constructed and amplified from the full-length hRAR␥ cDNA (26) using PCR with the upstream forward primer 1 (36-mer) 5Ј-CTCGCATAGACCCATATGTCCAAGGAAGCTGTG-CGA-3Ј and the downstream reverse primer 2 (35-mer) 5Ј-GCGCGCG-GATCC(TTA)CATTTCAGGGTTCTCCAGCA-3Ј. The underlined nucleotides in each primer represent the hybridizing portion, and the nucleotides in boldface type specify the NdeI and BamHI restriction sites, respectively. The codon specifying the translation stop is in parentheses in primer 2. After 5 min of preheating at 95°C, each of 20 PCR cycles was composed as follows: denaturation, 1 min at 94°C; annealing, 1 min at 55°C; and synthesis, 2 min at 72°C. The 50-l PCR mixture contained 50 pmol of each primer, 3 ng of pSG5/hRAR␥ vector, 0.2 mM deoxynucleotide triphosphates, 50 mM KCl, 10 mM Tris-HCl, pH 8.3, 1.5 mM MgCl 2 , 0.001% gelatin, and 1.0 unit of Taq polymerase. Following amplification, the reaction was extracted with an equal volume of chloroform:isoamyl alcohol (24:1) and then precipitated and washed with 100% ethanol. The resultant cDNA was digested to completion with NdeI and BamHI and gel purified and ligated into the similarly digested recipient pET15b vector. The ligation mixture was used to transform competent E. coli strain DH5␣ cells. The sequence of the ligated insert from a transformed colony was confirmed using a modification of the Sanger dideoxy method (U. S. Biochemical Corp.) and was designated pET15b/DE␥.
Expression of DE␥ in E. coli and Preparation of Crude Soluble Extracts-For the preparative purification of the HIS 6 -DE␥ fusion, 12 ϫ 10 6 BL21(DE3)/pET15b-DE␥ cells were introduced into 1 liter of minimal media (MM/C) (42 mM Na 2 HPO 4 -7H 2 O, 22 mM KH 2 PO 4 , 8.6 mM NaCl, 19 mM NH 4 Cl, 5 g/ml FeCl 3 -6H 2 O, 1 mM MgSO 4 -7H 2 O, 100 M CaCl 2 , 0.0001% thiamine, 0.4% glucose, 50 g/ml carbenicillin) prewarmed to 37°C. The inoculum was freshly prepared in MM/C from a frozen stock of the expression host BL21(DE3)/pET15b-DE␥. The cells were grown at 37°C in a New Brunswick model G25 shaker at 350 rpm. At 1.0 A 600 (approximately 18 h), the temperature of the culture was rapidly shifted to 26°C by swirling on ice water. Expression of the fusion protein was induced with 1 mM isopropyl-␤-D-thiogalactopyranoside (Boehringer Mannheim) for 2 h at 26°C with continuous shaking at 350 rpm. Cells were harvested by centrifugation at 4,200 ϫ g for 5 min at 4°C, resuspended in 50 ml of ice-cold 50 mM Tris-HCl, pH 8.0. After addition of lysozyme and Triton X-100 to final concentrations of 200 g/ml and 0.1%, respectively, the cells were quickly frozen on dry ice, thawed at 30°C, and then sonicated for 30 s until no longer viscous. After the addition of NaCl to 0.5 M, the lysate was centrifuged at 15,000 ϫ g for 20 min at 4°C. The protein concentration in the supernatant (crude soluble cytoplasmic extract) was determined by the Bradford method using bovine serum albumin as the standard (27).
Purification of DE␥ by Nickel Chelation Chromatography-The DE␥ protein was purified from 150 -200 ml of crude soluble cytoplasmic extract in a batchwise fashion by mixing with a 5-ml bed volume of Ni 2ϩ -charged His⅐Bind resin equilibrated in binding buffer (5 mM imidazole, 500 mM NaCl, 20 mM Tris-HCl, pH 7.9). The resin was purchased from Novagen and prepared according to the manufacturer's instructions. After gently rotating 45-60 min at room temperature, the resin was packed into a 2.5 ϫ 100-cm Econocolumn (Bio-Rad) and washed six times by resuspension in 50 ml of binding buffer, followed by four washes in 50 ml of wash buffer (60 mM imidazole, 500 mM NaCl, 20 mM Tris-HCl, pH 7.9). When the concentration of protein in the final wash was less than 100 g/ml, the resin was equilibrated with cleavage buffer (50 mM Tris-HCl, pH 8.0, 100 mM NaCl, 2.5 mM CaCl 2 ). The amount of protein bound to the resin was determined by the Bradford method (27). Thrombin (Novagen) was added to a final concentration of 0.5 units/mg protein in a volume of 20 -30 ml of cleavage buffer and allowed to cleave for 12-14 h at 25°C while rotating slowly in a vessel with no discernible airspace. The column was drained and then rinsed with cleavage buffer without CaCl 2 . Phenylmethylsulfonyl fluoride was added to the eluate from a 100 mM stock to a final concentration of 0.1 mM to inhibit thrombin activity. The purity of the DE␥ protein was determined by laser densitometry of silver-stained 12% SDS-polyacrylamide gel electrophoresis gels (28) and by amino acid analysis. 2 Isoelectric Focusing-The isoelectric point of the purified DE␥ protein was determined using the Pharmacia Phast System. Pharmacia broad pI calibration standards were run in parallel lanes with DE␥ on a Phast Gel IEF 3-9. The gel was prefocused at 2000 volts, 2.5 mA, 3.5 watts for 75 V-h at 15°C. After loading and prerunning the gel at 200 V, 2.5 mA, 3.5 watts for 15 V-h at 15°C, the samples were focused for 410 V-h under the same conditions as in the prefocusing step. The gel was silver stained according to the manufacturer's instructions and dried under hot air for 10 min. Migration distances for each of the standards were measured and plotted versus isoelectric points. The pI for the DE␥ protein was determined by comparison with this standard curve.
Ligand Binding Assays-In binding assays and competition experiments, 1.0 -3.0 pmol of purified DE␥ receptor protein plus 3.5 g of uninduced crude extract, which served as a carrier and had negligible specific binding of its own, was routinely added to binding buffer (50 mM Tris-HCl, pH 8.0, 100 mM NaCl) giving a final volume of 1.0 ml. In the binding assays, the protein was incubated with 0.1-10.0 nM [ 3 H]t-RA. For competition experiments, the protein mixture was preincubated with 1.0 nM [ 3 H]t-RA at 4°C for 30 min, after which various concentrations of unlabeled retinoids solubilized in EtOH were added. The final concentration of EtOH did not exceed 1-2% (v/v). The reactions were carried out in the dark at 4°C for 16 h, after which 0.5-ml ice-cold equivalent particle size, lyophilized dextran-coated charcoal (EqDCC), prepared according to Dokoh et al. (29), was added. The samples were vortexed, placed on ice for 10 min, and then centrifuged 15 min at 14,000 ϫ g at 4°C. Total dpm of the supernatants was measured in a Beckman LS6000 IC scintillation counter. Nonspecific binding was determined in the presence of 100-fold molar excess unlabeled t-RA. Scatchard analysis (30) was performed for determination of K d . Apparent K d values for t-RA, 9-cis RA, retinol, and retinal were calculated using the IC 50 determined from competition curves and the Clark equation (31).
Circular Dichroism Measurements-CD spectra were recorded with a Jasco J720 spectropolarimeter (Jasco Inc., Tokyo). The instrument was calibrated at ambient temperature such that a 1 mg/ml solution of (ϩ)Ϫ10-camphor sulfonic acid had an ellipticities ratio of 2.0 when measured at 192.5 and 290 nm in a cell with a 0.1-mm optical path length. All experimental measurements were made at ambient room temperature (22.2 Ϯ 0.1°C) on samples of 50 M protein in buffer containing 50 mM Tris (pH 7.4), 100 mM NaCl, 0.1 mM EDTA, and 0.025 mM phenylmethylsulfonyl fluoride. In some samples, t-RA was added in equimolar amounts to protein. CD spectra were recorded using buffer alone as the base line. For each sample, four spectra were collected and averaged (two spectra for two separate preparations of each sample). All spectra are reported in terms of molar ellipticity, [], where [] ϭ MRW /(10Lc) and MRW is the measured mean residue ellipticity, L is the path length, and c is the protein concentration. CD spectra were deconvoluted by least squares analysis with the four-basis set of Yang and co-workers (32). Estimates of secondary structure were obtained by normalizing the fitted parameters to yield percentages of helix, sheet, turn, or random coil.
Fluorescence Measurements-Fluorescence measurements were made with a Perkin-Elmer model LS-5B (Perkin-Elmer Corp.) luminescence spectrometer at 25°C using a slit width of 5 nm. t-RA was added to a 5 M solution of purified DE␥ protein in 50 mM Tris-HCl, pH 8.0, 100 mM NaCl. The protein concentration was determined by amino acid analysis. 2  For denaturation studies, flourescence measurements were obtained by diluting the purified DE␥ protein in 6 M guanidine HCl, and emission spectra were obtained as described above in the absence of t-RA.
Protein Analysis-The structural predictions for the DE␥ protein were performed using the Gene-Works 2.4 DNA-protein analysis software (IntelliGenetics, Inc.).

Expression and Purification of hRAR␥ DE Domain from E.
coli-The host E. coli strain, BL21(DE3), transformed with the vector pET15b/DE␥, produces a protein of the predicted molecular mass for HIS 6 -DE␥ (approximately 32.4 kDa) upon induction with 1 mM isopropyl-␤-D-thiogalactopyranoside (Fig. 1,  lane 1). This protein was not detectable in uninduced cells, nor was it detected in cells transformed with the parent vector alone, pET15b (data not shown). About 90% of the HIS 6 -DE␥ expressed was present in the soluble fraction when induced as described. The remainder was recovered in the insoluble pellet formed by centrifugation of the lysate at 15,000 ϫ g. Induction in minimal media at 26°C for 2 h was critical to the expression of HIS 6 -DE␥ in the soluble fraction. Induction at 30°C for 2 h reduced the level of soluble fusion protein by 50%, while induction at 37°C for 2 h reduced the level beyond detection. The hydrophilic D domain appears to be important for the expression of soluble protein, since similar constructs of the E domain, which did not contain the D domain, were expressed entirely as insoluble inclusion bodies (data not shown).
The HIS 6 -DE␥ peptide was purified from the crude soluble cytoplasmic extract by nickel chelation chromatography. The yield was approximately 7 mg of DE␥ from 383 mg of crude protein (Table I) representing about 2% of the total soluble protein and is reported as the average of five experiments. The HIS 6 -DE␥ fusion is undetectable in the void volume (Fig. 1,  lane 2), indicating that the expressed fusion protein was efficiently bound to the affinity resin. Repeated washing of the affinity column with 60 volumes of a 5 mM imidazole buffer followed by 40 volumes of a 60 mM imidazole buffer was necessary to completely remove contaminating E. coli proteins from the resin.
The purified DE␥ protein was eluted by thrombin cleavage of the resin-bound HIS 6 -DE␥ fusion. Based on quantitative estimates using laser densitometry, greater than 85% of the HIS 6 -DE␥ peptide was cleaved by thrombin. Approximately 60% of the cleaved product was recovered in the column eluate, and the purity of this protein was greater than 98% (Fig. 1, lane 5). Results from amino acid analysis of the protein in this fraction indicate that the composition of residues in the purified DE␥ protein is consistent with that determined from the primary sequence 2 so that the protein preparation is unlikely to be contaminated. Subsequent treatment of the column with buffer containing 400 mM imidazole elutes the remaining DE␥ protein (Fig. 1, lane 6). Since the DE␥ in this fraction does not bind retinoic acid (data not shown) and spontaneously precipitates upon collection, it most likely contains molecules that are incorrectly folded.
The isoelectric point (pI) for the purified DE␥ protein was determined on a Pharmacia IEF 3-9 Phast Gel and was found to be 5.2. This experimentally obtained pI value is in good agreement with that calculated from the primary amino acid sequence (5.18) (Fig. 2).
Ligand Binding Characteristics of Purified DE␥-The ligand binding properties of DE␥ were first analyzed using a modified charcoal absorption binding assay. Fig. 3a depicts   hRAR␥ Domain E Contains an ␣-Helical Ligand Binding Site data gives a linear plot indicating a single class of binding sites (Fig. 3b) with a dissociation constant, K d ϭ 0.6 Ϯ 0.1 nM (n ϭ 3), in good agreement with the K d determined for full-length RAR␥ from nuclear extracts of transiently transfected COS cells (7). Results for other naturally occurring retinoids were obtained from competition experiments (Table II) using the purified DE␥ protein or crude bacterial extracts containing full-length hRAR␥. The order of binding was similar for both sources of receptor protein (i.e. t-RA Ͼ 9-cis RA Ͼ retinol Ͼ retinal). A similar order of binding has been previously reported for the full-length receptor protein (7,17,20), providing further evidence that the binding properties of the isolated DE domains are consistent with those for the full-length receptor. Secondary Structure Analysis-Analysis of the secondary structure of DE␥ was performed by circular dichroism. Fig. 4B shows the CD spectrum obtained for 50 M purified DE␥ protein. In the absence of t-RA (line b), ␣-helix content is indicated by the negative ellipticity with minima at 222 and 208 nm and a peak maximum near 192 nm. Deconvolution of this spectrum with the unconstrained four-basis set of Yang and co-workers (32) suggests that DE␥ contains a large portion of helical conformation (25-30%) with little ␤ sheet (Ͻ8%). Constraint of the basis set such that the sum of the secondary structures is equal to 100% increases the estimated helical content to 56% and decreases the ␤ sheet to 0%. This result agrees well with the Garnier prediction (34) for this protein depicted in Fig. 4A. The CD spectrum for DE␥ in the presence of 10 Ϫ6 M t-RA was also recorded (Fig. 4B, line a). The magnitude of the peaks at 222, 208, and 192 nm all increased, suggesting an increase in the helical content of the ligand-bound protein. However, upon unconstrained deconvolution, differences between the CD spectra in the presence or absence of t-RA were slight; the ␣-helical content increased to 29 -34% (from 25-30%) with no change in the amount of detectable ␤ sheet. These results suggest that a major change in the secondary structure of the DE␥ protein is not occurring upon ligand binding. However, changes in the tertiary structure of the protein not detected by CD analysis may still have profound consequences for protein function.
Fluorescence Analysis of Receptor-Ligand Interaction-Inspection of the amino acid sequence of the DE␥ peptide reveals a highly conserved tryptophan residue within the ligand binding domain (Trp-227) (Fig. 4A). This residue allows the use of fluorescence quenching to assess the integrity of the purified DE␥ protein preparation with respect to two parameters: threedimensional structure and ligand binding.
The tryptophan fluorescence emission spectrum was used to evaluate the integrity of the three-dimensional structure of the folded DE␥ protein. The fluorescence of the native protein exhibits two maxima at 320 and 340 nm (Fig. 5). Upon denaturation with 6 M guanidine HCl, Trp-227 fluorescence shifts to higher wavelengths (340 -350 nm), indicating increased exposure of that residue to the solvent. This shift is most likely the result of a repositioning of the tryptophan residue into a more polar, less shielded environment and is an indication that the purified DE␥ protein was properly folded prior to denaturation. At the same time, relative fluorescence maxima of both the native and denatured protein remained the same (68 versus 75%), consistent with the view that fluorescence quenching was not the result of changes in conformation.
Trp-227 fluorescence can also be used to monitor binding kinetics of the purified receptor protein. Titration of 5 M DE␥ protein with increasing concentrations of t-RA reveals significant quenching of the 340-nm fluorescence peak (Fig. 6a). Since the concentration of t-RA at the inflection point of the curve (5.1 M) is very close to the concentration of receptor protein in each reaction vessel (5.0 M determined by amino acid analysis), 2 greater than 95% of receptor molecules are folded into a functionally active conformation. Further, linearization of the fluorescence quench data by the method of Cogan (33) (Fig. 6b) was used to demonstrate the presence of one ligand binding site per receptor molecule (n ϭ 1.08) from the slope of the Cogan curve.

DISCUSSION
Information concerning the three-dimensional structure of members of the steroid/thyroid hormone receptor superfamily has been hindered by the inability to purify sufficient quantities of receptor protein for NMR or x-ray crystallographic anal-  hRAR␥ Domain E Contains an ␣-Helical Ligand Binding Site ysis. Several laboratories have reported on the use of expression systems in different species as a method for increasing yields of heterologous protein, but the yields of purified protein with functional activity have generally been low (6,(17)(18)(19)(20)(21)(22)(23). Another approach has been to express each domain separately under the assumption that a three-dimensional structure obtained in this way would be identical to that of the same domain in the context of the full-length receptor protein. The most successful use of this approach has been the expression and resulting NMR analysis of the DNA binding domain of the glucocorticoid (35,36), estrogen (37,38), retinoic acid (39), and retinoid X (40) receptors. The structure of the DNA binding domain has rendered it particularly amenable to functional expression in both prokaryotic and eukaryotic systems due mainly to the hydrophilicity of the amino acids in this zinc finger domain. Other domains, such as the ligand binding domain E, have proven more difficult to produce in high yields due to the large number of hydrophobic residues in this region. The method described here consistently produced 5-8 mg of purified DE␥ protein from 1 liter of bacterial culture, representing approximately 30% recovery of expressed protein (Table I). Purity and integrity of protein obtained in this way was assessed by a number of criteria. Silver-stained polyacrylamide gel electrophoresis gave a single band of the expected molecular mass, 31,000 daltons (Fig. 1, lane 5). The experimen-tally determined pI of the purified DE␥ protein (Fig. 2) agrees with that calculated from the amino acid sequence (5.2 versus 5.18). Binding and Scatchard analysis of the purified protein gave a K d of 0.6 nM (Fig. 3). This value is consistent with that obtained using crude protein extracts (7). In addition, competition binding assays using other naturally occurring retinoids (Table II) gave apparent K d values similar to those published for the full-length receptor (17,20) with the same order of affinity: t-RA Ͼ 9-cis RA Ͼ retinol Ͼ retinal.
Since the DE␥ protein appeared to be purified to homogeneity by the methods described using the above criteria, it was appropriate to evaluate the secondary structure characteristics using circular dichroism. Initial analysis revealed that light scattering of the sample hindered interpretation of the CD spectra at wavelengths below 190 nm. Previous studies have shown that truncating the CD spectrum at 190 nm results in poor estimates for the amount of ␤ sheet and turn conformations found in the protein (44,45). However, the same studies found that because the CD for an ␣-helix dominates the spectrum, the analysis for ␣-helix is reliable regardless of the wavelength range. Given these considerations, the purified DE␥ protein was found to contain a high degree of ␣-helical structure. Constrained deconvolution of the CD spectra to quantitate helix, sheet, or random coil conformations suggests an ␣-helical content as high as 56%, in good agreement with a value of 62% ␣-helical content determined using the Garnier algorithm (Fig. 4A). Further support for this observation comes from the recently published crystal structure of a closely related member of the steroid/thyroid superfamily, RXR␣ (24). This structure contains approximately 65% ␣-helix with only 4% ␤ sheet.
On addition of retinoic acid, only minor changes in helical content were observed (Fig. 4b), indicating that major alterations in the secondary structure of DE␥ were unlikely to occur upon ligand binding. Recently, CD analysis of the thyroid hormone receptor, another member of the steroid/thyroid hormone receptor superfamily, indicated a high proportion of ␣-helical structure (46). Interestingly, addition of thyroid hormone produced only minor changes in the CD spectra obtained, in agreement with the results reported here.
The finding of significant helical content for DE␥ is noteworthy in light of the fact that the structure of another protein known to bind retinoids, retinol-binding protein, is primarily ␤ sheet. Retinol binding protein is a member of a family of proteins whose architecture consists of a ␤ barrel formed by two orthogonal ␤ sheets and four turns of ␣-helix at its carboxyl terminus (47,48). Based on the results of this study, DE␥ is structurally distinct from the retinol binding protein family. In support of this observation, one-dimensional NMR spectra of DE␥ did not show any of the typical markers for ␤ sheet conformation. 3 Few ␣-proton resonances are observed down field of water, and there is little chemical shift dispersion among the amide proton resonances that are typical of proteins with significant ␤ sheet or turn conformations.
Inspection of the amino acid sequence within the DE␥ protein revealed a tryptophan residue at position 227. This residue was used to evaluate the native structure of DE␥ as well as the kinetics of ligand binding. Denaturation of the protein with 6 M guanidine causes a shift in the emission maxima from 340 to 350 nm (Fig. 5). This shift suggests a repositioning of Trp-227 into a more polar environment (41). Since the relative fluorescence does not change, this shift is most likely due to structural differences common upon denaturation but not a change in autofluorescence. Thus, any quenching observed upon the addition of ligand will be due to ligand binding only, further supporting the position of Trp-227 within the ligand binding site.
Upon addition of increasing t-RA concentration, the stoichiometry of ligand binding to purified DE␥ protein can be measured. This method has been used to evaluate the binding of retinoic acid and other naturally occurring retinoids to the cytoplasmic retinoic acid binding proteins (41,42). Fig. 6a shows that the fluorescence of Trp-227 is quenched in a dosedependent manner. An inflection in this binding curve can be expected at the concentration of ligand, which saturates all the binding sites present within the protein tested. As can be seen in Fig. 6a, fluorescence is progressively quenched upon addition of increasing amounts of t-RA. The inflection point of this titration curve occurs at a point slightly greater than 5 M t-RA. Since the concentration of DE␥ protein in each reaction vessel was adjusted to 5 M based on amino acid analysis, it is evident that stoichiometric binding of a single binding site occurs for greater than 95% of the purified protein molecules. Linearization of the fluorimetric titration data for DE␥ using the method of Cogan et al. (33) confirms only one t-RA binding site per receptor molecule (n ϭ 1.08) (Fig. 6b). This result appears to be a reliable estimate of the number of binding sites (n) since this method is most accurate when high affinity binding occurs, as is the case for DE␥ (43).
The results presented here predict that the ligand binding pocket of RAR␥ will be composed of at least one ␣-helical structure containing the tryptophan at position 227. A model of this region (Fig. 7) shows the location of Trp-227. Interestingly, this residue is immediately adjacent to an alanine residue at position 232. This residue is analogous to Ala-225 in RAR␤, which has been recently demonstrated to be an important recognition residue for receptor-selective retinoids (25). Maksymowych et al. (49) have predicted a similar helix-turnhelix motif in this region for all members of the steroid/thyroid hormone receptor superfamily. Taken together, these results 3 L. Mueller, personal communication. FIG. 6. Fluorimetric titration of purified DE␥. a, intrinsic fluorescence of DE␥ protein is quenched upon addition of increasing amounts of t-RA. Fluorescence was measured at 340 nm emission after excitation at 280 nm using a slit width of 5 nm. b, titration quench data are linearized to determine the number of binding sites per receptor molecule (n) after the method of Cogan et al. (33). Po ϭ protein concentration; Ro ϭ retinoid concentration; ␣ ϭ (F Ϫ F max /Fo Ϫ F max ) where Fo ϭ initial fluorescence, F ϭ fluorescence at each Ro, and F max ϭ fluorescence at maximum quench.
hRAR␥ Domain E Contains an ␣-Helical Ligand Binding Site suggest that the region in the amino-terminal portion of the ligand binding domain of the RARs contains the ligand recognition site and that all the members of this superfamily are likely to contain receptor-ligand contact points in this region. In addition, modeling the position of 9-cis RA within the ligand binding domain of another member of this superfamily, RXR␣, from the crystal structure (24) predicts receptor-ligand contacts within the hydrophobic portion of the homologous helix that is described here for RAR␥, further supporting this region as the ligand binding site.
In summary, this work describes an efficient, one-step method for the production and purification of milligram quantities of human RAR␥ DE domains expressed as heterologous protein in E. coli. The purified DE␥ receptor protein is of high purity and significant biological activity. The protein exhibits binding affinities consistent with the full-length receptor for t-RA and other naturally occurring ligands. Secondary structural analysis using circular dichroism is consistent with a predominantly ␣-helical polypeptide with only minor alterations in secondary structure evident upon ligand binding. Fluorescence quenching techniques were used to identify a single hydrophobic, ␣-helical site within the amino-terminal portion of the ligand binding domain, which contains a tryptophan residue in the ligand binding pocket. Future studies will assess the structural and biochemical properties of this binding site.