Chemical and Biochemical Issues Related to X-ray ... - ACS Publications

406

Bioconjugate Chem. 2001, 12, 406−413

Chemical and Biochemical Issues Related to X-ray Crystallography of the Ligand-Binding Domain of Estrogen Receptor r Steven W. Goldstein, Jon Bordner, Lise R. Hoth, and Kieran F. Geoghegan* Pfizer Global Research and Development, Eastern Point Road, Groton, Connecticut 06340. Received October 19, 2000; Revised Manuscript Received February 2, 2001

Careful attention to technical issues preceded successful crystallography of the ligand-binding domain of estrogen receptor R (ERR) complexed with CP-336156, a nonsteroidal estrogen agonist/antagonist. An affinity column based on immobilized estradiol was prepared according to the scheme of Greene et al. (Greene, G. L., Nolan, C., Engler, J. P., and Jensen, E. V. (1980) Proc. Natl. Acad. Sci. U.S.A. 77, 5115-5119). It was shown by X-ray crystallography that the major and less polar isomer of the affinity column precursor was 17R-((S)-2′,3′-epoxyprop-1′-yl)estra-1,3,5(10)-triene-3,17β-diol. This diastereomer was coupled to Thiopropyl Sepharose, with coupling monitored by observing loss of the phenolic absorption band of estradiol from the reaction supernatant, and gave an affinity matrix containing about 9 µmol of estradiol per milliliter of wet gel. Recombinant ERR ligand binding domain was selectively removed from E. coli cell lysate by binding to the column and was partly Scarboxymethylated by treatment with iodoacetic acid while bound to the column as described by previous workers. After being eluted from the column as a complex with drug, the receptor fragment was shown by mass spectrometry to be a mixture of differently modified forms. It was further S-carboxymethylated in solution, after which anion-exchange chromatography was used to isolate protein in which two of the four cysteine residues were S-carboxymethylated. This material, which afforded diffraction-quality crystals, was subjected to digestion with trypsin and peptide mapping analysis by HPLC coupled with mass spectrometry. For this experiment, the two previously unmodified cysteines were alkylated with 4-vinylpyridine to allow definitive identification. It was shown that Cys-417 and Cys-530 were S-carboxymethylated in the crystallized protein, while Cys-381 and Cys447 remained unmodified. Close attention to such technical issues may be important in structural studies of other nuclear receptors, a very important class of potential drug targets.

INTRODUCTION 1

ER belong to the nuclear receptor superfamily and are composed of discrete ligand-binding and DNA-binding domains. There are two principal isoforms, ERR and ERβ. In the presence of 17β-estradiol or another agonist, the receptors associate with accessory proteins known as coactivators to mediate the effects on gene expression that are the end-point of estrogen action in many cells or tissues. Details of these interactions have only recently begun to be elucidated (1-3). Crystallographic structural biology is one of the approaches being used to scrutinize this complex machinery. Structures have been reported for the ligand-binding domains of ERR (4-6) and ERβ (7) complexed separately with agonist ligands (17β-estradiol and diethylstilbestrol) and drugs of the class called selective estrogen receptor modulators (SERMs) (tamoxifen and raloxifene; for reviews of this drug class, see refs 8-10)). The results illustrate brilliantly how internal details of a proteinligand interaction can modulate the protein’s external conformation, as the packing site of the C-terminal structural element helix 12 is different in the agonist and SERM complexes. They complement other demonstra* To whom correspondence should be addressed. Phone (860)441-3601. Fax (860)441-3783. E-mail: kieran_f_geoghegan@ groton.pfizer.com. 1 Abbreviations: ER, estrogen receptor(s); ERR, estrogen receptor R; mCPBA, m-chloroperoxybenzoic acid; DTT, dithiothreitol, reduced form; RP-HPLC, reversed-phase HPLC.

tions that the nature of the ligand governs the types of interactions with coactivators available to the receptor and that this underlies the differences in action between agonists and SERMs. The results with ER form part of a more general picture that is emerging of how ligandprotein interactions modulate the function of nuclear receptors (11). Isolation of ER from natural or recombinant sources is necessary in many studies and an essential precursor to crystal growth. Affinity chromatography has been an important technical asset, as the high affinity of ER for immobilized derivatives of estradiol allows single-step isolations (12). The same high affinity, however, has also been a detriment, as conditions that are probably at least partly denaturing have been needed to elute bound receptor from columns. In all four publications to date describing crystallography of ER-ligand complexes (47), a single specific affinity matrix (13) and generally similar procedures have been used. In most of these published procedures, the receptor fragment was partly S-carboxymethylated while it was immobilized on the affinity column. When this was not done, the resulting structure contained an artifactual intermolecular disulfide (5). The exact outcome of these treatments has not been described in every case, and it has largely been left unstated but assumed that the modifications have no bearing on the substance of the results. We present peptide mapping analysis of a homogeneous protein preparation from which crystals were grown, data that clearly show the extent and

10.1021/bc000127d CCC: $20.00 © 2001 American Chemical Society Published on Web 04/14/2001

Preparing ER Ligand-Binding Domain for Crystallography

location of the modifications in the protein that we crystallized. Our experience with these various technical issues came as we prepared recombinant ERR ligand-binding domain complexed with CP-336156 (lasofoxifene) (14, 15) for crystallography. The structural results (J. Pandit et al., unpublished) will appear elsewhere. Some technical aspects of the task of obtaining crystallographic-quality material were unusually subtle and challenging, and included elements not fully described in existing literature. Among these were crystallographic determination of the absolute conformation of the estradiol derivative needed to make the affinity column, details of its immobilization to Thiopropyl Sepharose, and the isolation of incompletely but uniformly S-carboxymethylated ERR protein. This note describes our experience with these key steps preceding successful crystal growth and structure analysis. EXPERIMENTAL PROCEDURES

Synthesis of 17R-Epoxypropylestradiol. 17R-(2′Propen-1′-yl)estra-1,3,5(10)-triene-3,17β-diol (1). To 100 mL of a 1 M allylmagnesium chloride solution in ether (100 mmol) was added a solution of estrone (4.05 g, 15.0 mmol) and THF (100 mL) over 20 min. The reaction was heated to reflux for 3 h and then allowed to cool to room temperature overnight. After pouring into an aqueous solution of ammonium chloride (250 mL) and extracting with ether (2 × 150 mL), the combined organic fraction was washed sequentially with water (100 mL), 5% NaHCO3 (50 mL), and brine (100 mL) and then dried over MgSO4 and concentrated to give the title compound (4.48 g, 96%). Recrystallization from ethyl acetate/cyclohexane afforded a single diastereomer, mp 106-107 °C (dec). 1H NMR (CDCl3) δ 0.90 (s, 3H), 1.2-1.7 (multiplets, 9H), 1.86 (ddt, J ) 2.7, 2.7, 12.5 Hz, 1H), 1.97 (m, 1H), 2.11 (m, 1H), 2.24 (dd, J ) 7.1, 13.5 Hz, 1H), 2.29 (m, 1H), 2.35 (dd, J ) 7.1, 13.5 Hz, 1H), 2.79 (d, J ) 5.0 Hz, 2H), 5.18 (m, 2H), 5.98 (ddt, J ) 7.3, 10, 17 Hz, 1H), 6.53 (d, J ) 2.7 Hz, 1H), 6.60 (dd, J ) 2.7, 8.4 Hz, 1H), 7.12 (d, J ) 8.4 Hz, 1H). MS (APCI) 295 (M-17, 100), 253 (18). 13C NMR (CDCl ) δ 14.6, 23.7, 26.5, 27.7, 29.9, 32.0, 35.1, 3 39.8, 42.0, 44.0, 46.7, 49.8, 82.8, 112.9, 115.5, 119.5, 126.7, 132.8, 135.0, 138.5, 153.6. Anal. Calcd for C21H28O2‚ 1/ EtOAc: C, 79.00; H, 9.04. Found: C, 78.78; H, 8.87. 3 3-(tert-Butyldimethylsilanyloxy)-17R-(2′Propen-1′-yl)estra-1,3,5(10)-triene-3,17β-ol (2). To a solution of tertbutyldimethylsilyl chloride (1.68 g, 11.1 mmol), imidazole (1.58 g, 23.2 mmol), and DMF (7 mL) was added 1 (9.01 mmol), and the reaction was heated to 35 °C for 3 h. An additional portion of silyl chloride (0.80 g, 11.6 mmol) was added, and the reaction was heated an additional 2 h at 35 °C. After cooling to room temperature, the reaction was poured into 2% aqueous NaHCO3 (200 mL) and extracted with ethyl acetate (2 × 100 mL). The combined organic layer was washed with water (100 mL) and brine (100 mL), dried (MgSO4), and concentrated to give an oil. This was chromatographed on silica gel with 5% ethyl acetate/hexane to give the title compound (3.61 g, 94%) isolated as an oil. 1H NMR (CDCl3) δ 0.16 (s, 6H), 0.89 (s, 3H), 0.95 (s, 9H), 1.2-1.7 (multiplets, 9H), 1.86 (ddt, J ) 2.7, 2.7, 12.5 Hz, 1H), 1.97 (m, 1H), 2.11 (m, 1H), 2.24 (dd, J ) 7.1, 13.5 Hz, 1H), 2.29 (m, 1H), 2.35 (dd, J ) 7.1, 13.5 Hz, 1H), 2.79 (d, J ) 5.0 Hz, 2H), 5.18 (m, 2H), 5.98 (ddt, J ) 7.3, 10, 17 Hz, 1H), 6.53 (d, J ) 2.5 Hz, 1H), 6.57 (dd, J ) 2.5, 8.3 Hz, 1H), 7.10 (d, J ) 8.3 Hz, 1H). 17R-((R,S)-2′,3′-Epoxyprop-1′-yl)estra-1,3,5(10)-triene3,17β-diol (3). To a solution of 2 (3.61 g, 8.47 mmol) and

Bioconjugate Chem., Vol. 12, No. 3, 2001 407

methylene chloride (85 mL) was added mCPBA (2.93 g, 17.0 mmol). After stirring overnight, the reaction was diluted with an additional 100 mL of methylene chloride and then washed sequentially with 2% KI (aqueous, 100 mL) and 10% aqueous sodium bisulfite (100 mL). The remaining solution was dried (MgSO4) and concentrated to give an oil which was chromatographed on silica gel (40% ethyl/hexane) to give two compounds (only clean fractions were collected): less polar diastereomer, 1.38 g; more polar diastereomer, 0.47 g. These compounds were not further characterized, but rather a solution of the less polar isomer (1.38 g, 3.12 mmol) and THF (10 mL) was treated with tetrabutylammonium fluoride (4.6 mL, 1 M in THF) at 0 °C. After 5 min the ice bath was removed and the reaction allowed to warm to room temperature. It was then diluted with ethyl acetate and washed with 5% NaHCO3 (2 × 50 mL) and brine (50 mL). This was dried over MgSO4 and concentrated to give a solid. Purification on silica gel with 60% ether/hexane gave the title compound as a white solid; mp 202-203 °C. 1H NMR (CDCl3) δ 0.90 (s, 3H), 1.2-1.7 (m, 9H), 1.85 (m, 1H), 2.10 (m, 2H), 2.27 (m, 1H), 2.36 (br s, 1H), 2.53 (dd, J ) 2.8,4.9 Hz, 1H), 2.80 (m, 2H), 2.85 (dd, J ) 4.2, 4.9 Hz, 1H), 3.26 (m, 1H), 4.72 (br s, 1H), 6.53 (d, J ) 2.7 Hz, 1H), 6.59 (dd, J ) 2.7, 8.3 Hz, 1H), 7.11 (d, J ) 8.3 Hz, 1H). 13C NMR (CDCl3) δ 14.0, 23.6, 26.4, 27.6, 29.8, 31.7, 35.0, 39.3, 39.7, 44.0, 46.8, 47.7, 49.7, 50.3, 83.7, 112.9, 115.4, 126.7, 132.8, 138.5, 153.5. Anal. Calcd for C21H28O3: C, 76.79; H, 8.59. Found: C, 76.48; H, 8.65. Crystallography of the Resolved Less Polar Isomer of 17R-Epoxypropylestradiol. Light yellow crystals of the purified less polar isomer were grown by slow evaporation from acetonitrile. A representative crystal (0.24 × 0.26 × 0.44 mm) was surveyed and a 1 Å data set collected on a Siemens R4/sealed tube diffractometer using copper radiation (λ ) 1.54178 Å) at room temperature. All crystallographic calculations were facilitated by the SHELXTL system. The crystal belonged to the monoclinic space group P21 with the following cell dimensions: a ) 6.824(1), b ) 26.662(2), c ) 9.839 Å, β ) 100.67(1)°, V ) 1759.2(2) Å3. There were two molecules in the asymmetric unit (Z ) 4). The structure was phased using direct methods and refined routinely to a R-index of 4.96%. Hydrogen positions were calculated wherever possible. The methyl hydrogens and hydrogens on oxygen were located by difference Fourier techniques. A final difference map revealed no missing or misplaced electron density (0.22, -0.28 eÅ-3). Preparation of the Affinity Matrix. With minor variations, the coupling procedure was based on the method used by Salman et al. (16) to immobilize 17R(6′-iodohex-1′-ynyl)estr-4-en-17β-ol-3-one on Thiopropyl Sepharose. Solutions and buffers were not bubbled with N2 before use in the present work, and coupling was monitored by observing the phenolic 280 nm absorption band of the intended ligand rather than by chemical assay of the beads. To generate the free-thiol form of Thiopropyl Sepharose 6B (Amersham Pharmacia Biotech), 9 g of the dry resin was slurried in 50 mL of 0.3 M NaHCO3, 1.1 mM EDTA, pH 8.4 containing 1.1 g of DTT, and the resulting mixture in a screw-capped polypropylene tube was inverted slowly and continuously for 1 h at 22 °C. After this, the resin was washed on a coarse glass sinter with (i) an aqueous solution made to 3 L containing 18 mL of acetic acid, 87.7 g of sodium chloride, and 1.1 g of disodium EDTA, and (ii) 300 mL of 0.3 M NaHCO3, 1.1 mM EDTA, pH 9.2. The reduced, filtered, and washed Thiopropyl Sepharose

408 Bioconjugate Chem., Vol. 12, No. 3, 2001

beads (now containing 0.48-0.84 mmol of reactive thiol, according to the manufacturer’s specifications) were then added to a 50 mL polypropylene screw-cap tube containing 81 mg (0.25 mmol) of 3 dissolved in 18 mL of DMF. Some spontaneous warming of the mixture was noticed. This addition was followed by 30 mL of 0.3 M NaHCO3, 1.1 mm EDTA, pH 9.2, after which a small portion of the mixture (about 0.05 mL) was quickly withdrawn, and the full tube was capped. The sample withdrawn was centrifuged to pellet the suspended beads, and the absorption spectrum of the supernatant (0.02 mL diluted by addition to 1 mL of aqueous sodium phosphate buffer, pH 7.4) was recorded. The local maximum near 280 nm due to the phenol absorption band had an absorbance of 0.12 in a 1 cm cuvette. The reaction mixture was inverted slowly and continuously overnight at 22 °C, after which a check of the absorption spectrum of the supernatant from a small sample (done as before) showed no trace of the phenol band. This indicated complete attachment of the estradiol derivative to the resin. The beads were filtered on a coarse sinter and washed with 300 mL of 0.3 M NaHCO3, 1.1 mM EDTA, pH 9.2, after which they were added to a solution of 0.16 g of iodoacetamide in 45 mL of 0.3 M NaHCO3, 1.1 mM EDTA, pH 8.4, in a screw-capped polypropylene tube. The resulting slurry was inverted continuously and slowly for 1 h at 22 °C to alkylate remaining free thiol groups on the Thiopropyl Sepharose and then washed in turn with 0.3 L of each of the following: (i) 0.3 M NaHCO3, 1.1 mM EDTA, pH 8.4, also containing 1.5 g of DTT; (ii) the same solution without DTT; (iii) 0.05 M Tris HCl, pH 7.4; (iv) water; (v) 25% methanol in water; (vi) 50% methanol in water; (vii) 70% methanol in water. The beads were then stored at 4 °C as a slurry in 70% methanol in water. The absorption spectrum of a slurry of a small portion of the derivatized beads in sodium phosphate buffer, pH 7.4, clearly showed the presence of the phenol absorption band of the immobilized estradiol derivative. Based on complete coupling of the epoxide to Thiopropyl Sepharose, the gel was calculated to contain 9 µmol of estradiol per milliliter. Estradiol Affinity Chromatography. This step was performed at 22 °C using the method of Hegy et al. (17) for the most part, but with important modifications after ERR was eluted from the affinity column. A 11.4 g amount of Escherichia coli cells harboring the expression construct for ERR was resuspended in 50 mL of 0.1 M Tris HCl, 0.1 M KCl, 1 mM EDTA, 4 mM DTT, pH 8.5, containing leupeptin (25 µg/mL) and aprotinin (5 µg/mL), and lysed by sonication on ice (4 × 45 s treatment at power setting 4 with a 40% duty cycle using a Branson Sonifier). The lysate was centrifuged at 18 000g for 30 min, the pellet was discarded, and the supernatant was recovered and centrifuged again in the same way. The resulting clarified 50 mL sample was loaded (0.5 mL/min) onto a 5 mL column of the estradiol-Sepharose affinity medium previously equilibrated with lysis buffer. After loading, the column was washed sequentially at 1 mL/ min with 25 mL of lysis buffer, 25 mL of 0.05 M Tris HCl, 0.7 M KCl, 1 mM EDTA, 1 mM DTT, pH 8.5, and 25 mL of 0.01 M Tris HCl, pH 8.1. After this, the column was treated (0.5 mL/min) with 18 mL of 5 mM iodoacetic acid in 0.01 M Tris HCl, pH 8.1, made by 1:200 dilution of a 1 M solution of iodoacetic acid in 1 M Tris base with 0.01 M Tris HCl, pH 8.1. Flow was then stopped, and the column (protected from light) was allowed to stand at 22 °C for 18 h. It was then washed (1 mL/min in each case) with 25 mL of 0.01 M Tris HCl, pH 8.1, and with

Goldstein et al.

25 mL of 0.05 M Tris HCl, 0.25 M NaSCN, 1 mM EDTA, 1 mM DTT, 10% DMF, pH 8.5. Finally, bound ERR was eluted (0.5 mL/min) as a broad peak using 50 µM CP336156 in the same buffer, and a 60 mL product pool was recovered. Mass spectrometry (see Results) showed that the eluted protein was unevenly carboxymethylated. Following dialysis against 0.02 M Tris HCl, 5 mM DTT, pH 8.2, the product (32 mg of ERR in 42 mL) was further treated with 10 mM iodoacetic acid for 2 h at 22 °C under argon in the dark. The reaction was followed by mass spectrometry with the objective of maximizing the yield of twice-carboxymethylated ERR and was stopped by adding DTT to 20 mM. The product was then dialyzed against 0.02 M Tris HCl, 5 mM DTT, pH 8.2, and fractionated by anion-exchange chromatography on a MonoQ HR 10/ 10 column (Amersham Pharmacia Biotech) using a NaCl gradient in 0.02 M Tris HCl, pH 8.2. The main peak consisting of twice-carboxymethylated ERR was isolated and used for crystal growth trials. Digestion and peptide mapping to locate the sites of S-carboxymethylation were performed as described below. Mass Spectrometry. Mass spectra of intact proteins were obtained using an open-access system (Dr. J. G. Stroh) consisting of a PE Sciex API 100 single-quadrupole spectrometer with electrospray sample interface. Multistage MS of tryptic digests was obtained using a Finnigan LCQ ion-trap spectrometer interfaced with a HewlettPackard Model 1090 HPLC (see below). Tryptic Digestion. Purified complex of ERR with CP336156 (0.6 mg in 1.2 mL) was desalted by RP-HPLC on a Poros R2/H 2.1/30 column as described (18), and the protein peak (now lacking drug) in two equal aliquots was dried in a centrifugal concentrator (SpeedVac, Savant). To the dried residue of one portion in a microfuge tube was added 0.05 mL of 0.25 M Tris HCl, 6 M GnHCl, 1 mM EDTA, 5 mM DTT, pH 8.8. The sample was flushed with argon, closed, incubated at 37 °C for 2 h, and then allowed to come to 22 °C and treated with 1 µL of a 10% solution of 4-vinylpyridine in ethanol, giving a final concentration of 4-vinylpyridine of about 0.02 M. The tube was purged with argon and stored in the dark at 22 °C for 30 min, after which the sample was again desalted by RP-HPLC and dried in the SpeedVac. The dried sample was redissolved in 0.05 mL of 0.4 M NH4HCO3, 8 M urea, pH 8.5 and incubated for 1 h at 37 °C. It was then diluted to 0.2 mL with water, giving final concentrations of 0.1 M NH4HCO3 and 2 M urea. Trypsin (Promega chemically modified) was added to give a 1:50 ratio of trypsin to ERR, the sample was incubated for 2 h at 37 °C, and a further addition of trypsin was made to give a 1:25 ratio of trypsin to ERR. Digestion was then allowed to continue for 18 h at 37 °C. Digests were then analyzed by LC-MS using a Vydac C18 narrowbore column (type 218TP52) and the Finnigan LCQ. Solvents were A: 0.1% TFA, and B: 0.085% TFA in acetonitrile. RESULTS

Synthesis and Crystallography. 17R-Epoxypropylestradiol was prepared as described by el Garrouj et al. (19) (Scheme 1). Briefly, estrone was treated with excess allyl Grignard to give a single diastereomeric tertiary alcohol (1) after recrystallization. The phenolic hydroxy was then protected as the tert-butyldimethyl silyl ether (2), the terminal olefin was oxidized with mCPBA, and the intermediate silyloxy epoxides were separated chromatographically into pure fractions. These products were not fully characterized, but the major diastereomer was subjected to desilylation with fluoride ion to give 3,



Figure 1. Crystallographically defined structure of 17R-((S)-2′,3′-epoxyprop-1′-yl)estra-1,3,5(10)-triene-3,17β-diol. Two views of the molecule are shown. In the upper molecule, the A ring with the 3-hydroxyl group is to the right and the epoxide is to the extreme left. In the lower molecule, the A ring is at the lower left and the epoxide is to the extreme right. Scheme 1

crystallized, and identified as S at the epoxide (Figure 1) by crystallography (see Experimental Procedures). This compound was then coupled to the solid support. Coupling. Complete disappearance of the absorption band corresponding to estradiol from the supernatant of the coupling reaction indicated quantitative attachment of the steroid epoxide to Thiopropyl Sepharose (see Experimental Section). The product was stored as a slurry in 70% methanol at 4 °C and performed well after several months of storage. Affinity Chromatography and Modification of Cysteines. A 29 kDa protein fragment consisting of methionine followed by residues 301-553 of ERR (SWISSPROT accession P03372; see Figure 2 for the amino acid sequence) was expressed as a soluble intracellular protein in E. coli and purified by affinity chromatography on immobilized estradiol (Figure 3). ERR was treated with iodoacetic acid while bound to the column, as described by other groups reporting successful crystallization of an ER-ligand complex (4, 6, 7). After elution of the ERR with CP-336156, anion-exchange chromatography on MonoQ resolved the complex into three main peaks (Figure 4A), and ES-MS (Figure 4B) showed that about 5% of the polypeptide was unmodified, 60% was singly

Figure 2. The amino acid sequence of recombinant ERR ligandbinding domain numbered according to the full-length sequence of ERR. Theoretical masses: unmodified protein 28 852 Da; once-carboxymethylated 28 910 Da; twice-carboxymethylated 28 968 Da; three times carboxymethylated 29 026 Da. An initiator methionine preceded this sequence, but was absent from the purified product.

carboxymethylated, and 35% was doubly carboxymethylated. Size-exclusion chromatography on a Superdex 75 HR 10/30 column showed that the protein was a dimer of the 29 kDa subunits (data not shown), and ES-MS of the MonoQ peaks (not shown) showed that these were (in order of elution) homodimer of singly carboxymethylated ERR, heterodimer of once- and twice-modified ERR, and homodimer of twice-carboxymethylated ERR. In an effort to make the protein more homogeneous, it was treated in solution with iodoacetic acid (see Experimental Section). ES-MS was used to follow pilot reactions, and it was found that double carboxymethylation was readily accessible, while a third S-carboxymethylation would require more forcing conditions to be completed. A 33 mg sample in 43 mL was therefore treated with 0.02 M iodoacetic acid (2 h, 22 °C) until twicecarboxymethylated product was predominant (mass of 28 970 in Figure 4C). The reaction was quenched with DTT, and anion-exchange chromatography on MonoQ


Goldstein et al.

Figure 3. SDS-polyacrylamide gel electrophoresis showing affinity chromatography of ERR ligand binding domain on a column packed with the Thiopropyl Sepharose conjugate of 3. The procedure used was essentially that described in reference (17).

(Figure 4D) resolved the product into a minor peak (heterodimer of once- and twice-carboxymethylated ERR) and a major peak which was the homodimer of twicecarboxymethylated ERR (mass observed 28 971 Da) (Figure 4D, inset). The purified twice-alkylated protein was then used successfully for crystal growth. Identification of S-Carboxymethylation Sites. To allow definitive assignment of the modification sites, the two remaining cysteines were S-pyridylethylated using 4-vinylpyridine before the protein was digested with trypsin and subjected to LC-MS peptide mapping. Having each of the four cysteines in one of two stable, alkylated forms created the possibility of showing conclusively which were carboxymethylated and which were unmodified in the crystallized ERR. LC-coupled mass analysis was conducted using a Finnigan LCQ ion-trap spectrometer. The instrument was programmed to enable data-dependent multistage analysis of any mass peak exceeding a threshold, including a narrow-range scan that allowed confident assignment of its mass and subsequent analysis of a sequencerelated pattern of fragment ions created by collisioninduced dissociation in the ion trap (20). Postrun analysis of the data using the SEQUEST algorithm (21) frequently allows a peptide corresponding to the experimental ion to be identified even in a large database, with the best success tending to come for peptides of about 8-20 amino acids. For known protein sequences, if used with due care, this analysis is as reliable as chemical peptide sequencing. Using a combination of automated and manual interpretation, peptides accounting for all but a few small fragments of ERR were accounted for in the peptide map (Figure 5). The SEQUEST program was run more than once to search for Cys-containing peptides modified by S-carboxymethylation (+58 Da) and for Cys-containing peptides modified by S-pyridylethylation (+103 Da), and UV and mass peaks unaccounted for after this process were interpreted manually by reference to the sequence.

Figure 4. Use of anion-exchange chromatography and mass spectrometry to analyze the extent of S-carboxymethylation in ERR and to isolate twice S-carboxymethylated ERR. (A) Analytical fractionation on MonoQ HR 5/5 of ERR after elution from estradiol-Sepharose by CP-336156. Peak labels indicate indications from mass spectrometry; e.g., 1CM-2CM indicates a heterodimer of once- and twice-carboxymethylated ERR. (B) ESMS of the protein sample applied to MonoQ in panel A. (C) ESMS of ERR subjected to further treatment with iodoacetic acid in solution, but before the anion-exchange step in the next panel. (D) Preparative anion-exchange chromatography on MonoQ HR 10/10 of the ERR after the secondary treatment with iodoacetic acid. Inset: ES-MS of the material purified in the major peak, showing that twice-carboxymethylated ERR (mass theor. 28 968 Da) is the major component.

Cys-417 and Cys-530 were detected only in the Scarboxymethylated form, showing that these were the two cysteines that were alkylated in the crystallized ERR. Cys-381 and Cys-447 were detected only in S-pyridylethylated form, showing that they were unmodified in the crystallized protein. DISCUSSION

The preparation of protein-drug complexes for crystallography is often quite straightforward. Drug is added to soluble protein, and a concentrated solution of the complex is combined with a range of solutions that often induce protein crystallization. Conditions that yield


Figure 5. LC-MS peptide mapping of a tryptic digest of ERR ligand binding domain, S-alkylated with iodoacetic acid first and subsequently with 4-vinylpyridene. Cysteine-containing peptides were detected with Cys in either S-carboxymethyl or S-pyridylethyl form, as indicated on the figure.

acceptable crystals are ultimately discovered by a combination of strategy and chance. This method does not appear to have worked so far with ER, as no published reports describe success with it, and we were also unsuccessful with it. Instead (4-7), ER ligand-binding domain has been allowed to bind to the affinity matrix introduced by Greene et al. (13), treated while bound to the affinity medium with iodoacetic acid to alkylate some of the four sulfhydryl groups in each polypeptide and then eluted from the column with soluble ligand in a buffer containing 0.25 M NaSCN, a chaotrope and potential denaturant, as well as (in one case) 10% DMF. Why this method works when the conventional one does not is open to speculation. An important possibility is that the ER binding site must be fully occupied with ligand, and that this is not achieved by adding ligand to protein in solution. It is possible, for example, that the column method monomerizes the ER fragment, forcing each polypeptide to be separately displaced from the column by soluble hormone or drug and so giving full occupancy. The Greene affinity column is not the first affinity matrix applied to ER isolation. Vonderhaar and Mueller (22) prepared 17R-n-propylthiol estradiol which was immobilized on polyvinyl-(N-phenylenemaleimide) at a level of 0.08 mmol/g of dried resin (presumably about 0.03


mmol/mL of drained wet gel, or 3% of the ligand density of material made according to Greene). The same authors made another affinity medium in which estradiol was immobilized by diazotization at the 2 or 4 position, although this was unstable on storage. Both their products removed estrogen receptor from a homogenate of rat uterus, but the high affinity exhibited by the receptor for these media caused difficulty. Estradiol bound to ERR is wholly engulfed by the protein (4), and it is difficult to see how estradiol immobilized through an A ring atom can be an effective affinity ligand if immobilized ligand is bound in the same orientation as soluble ligand. Sica et al. (23) tried other linkages of estradiol to agarose through A ring substituents, but the resulting media were ineffective. Among the many options that they examined, better success came from coupling 17β-estradiol hemisuccinate to amino groups on an immobilized amino acid copolymer. Bound receptor was eluted by warming the medium to 30 °C in the presence of soluble estradiol. Efforts to produce media based on immobilized estradiol have continued; for example, a detailed discussion of a medium prepared using 17β-estradiol 17-hemisuccinate linked to bovine serum albumin and immobilized on agarose appeared in reference (24). Greene et al. (13, 25) described the synthetic plan (Scheme 1) for preparing the affinity medium used by all workers so far reporting crystal structures and showed that bound receptor could be eluted by soluble ligand in a buffer containing 10% DMF and 0.25-0.50 M sodium thiocyanate. We followed this general track to obtain the crystal structure of ERR complexed with CP-336156 (not shown here), but made departures from and additions to previously described methods. 17R-Epoxypropylestradiol is the key reagent required to prepare the affinity column. Its diastereomers at the epoxide center have previously been used in attempts to prepare compounds that bind irreversibly to ER (19, 26) and were shown to differ not only in their chromatographic properties but also in their affinities for ER. To the best of our knowledge, the stereochemical identities of the two isomers have not previously been reported. The diastereomers were prepared and resolved by chromatography as described (19), and crystals suitable for X-ray diffraction were prepared by slowly evaporating a solution of 3 in acetonitrile. Analysis of the diffraction pattern identified 3 as the (S) epoxide (Figure 1), and this compound was used to prepare the affinity matrix by coupling it to Thiopropyl Sepharose. Proteins with free cysteines are problematic for crystal growth, because long incubations of droplets create the risk of disulfide formation that destroys molecular homogeneity. For example, when cysteines were left unmodified in ERR, crystals were still obtained but an artifactual intermolecular disulfide was present in the final structure (5). This problem was avoided by the two groups who treated the column-bound protein with iodoacetic acid (4, 6). In addressing these problems, we encountered important issues that have not been discussed in detail elsewhere. The first was that ERR eluted from the affinity column with CP-336156 was inhomogeneous with respect to charge (Figure 4A) and mass (Figure 4B). As homogeneity is so highly valued in efforts to grow protein crystals, we “topped up” the chemical modification of cysteines by treating the protein-drug complex in solution with iodoacetic acid until doubly S-carboxymethylated ERR was the major species (Figure


4C). Anion-exchange chromatography was then used to purify the doubly S-alkylated form (Figure 4D), and this was the protein used to grow diffraction-quality crystals. The main question is how many and which cysteines need to be carboxymethylated in order to allow crystal growth. Peptide mapping of a tryptic digest was used to determine which cysteines had been modified in our ERR cocrystallized with CP-336156. Cys-417 and Cys-530 were found to be S-carboxymethylated, and Cys-381 and Cys447 were found not to be (Figure 5; also, see Figure 2 for the amino acid sequence of the recombinant ERR). This result indicated a difference between our results and those of Brzozowski et al. (4). Their paper indicated only that Cys-381 was uniformly modified and Cys-447 was not modified. Their protein-ligand complexes were prepared as described in Hegy et al. (17), which states that triply carboxymethylated ERR with Cys-447 unmodified was the main product in their protein separately cocrystallized with estradiol and raloxifene. The crystal structures show that Cys-447 is sufficiently buried to be highly unlikely to react with iodoacetic acid in the folded protein. Thus it would appear that crystallizable forms of ERR described so far are those in which Cys-447 is unmodified, Cys-417 and Cys-530 are modified, and Cys381 either is or is not modified. In the protein as used by Shiau et al. (6) (Protein Data Bank accession 3ERD), carboxymethylation was modeled into the X-ray structure on the basis of electron density at Cys-417 and Cys-530, but the detailed results of protein chemical analyses were not reported in the paper (although ES-MS was stated to have been done on the modified protein). All of the above represents discussion in detail of material that some will consider purely technical and even marginal. To this it can only be replied that nuclear receptors represent a very large family of molecules of enormous biological significance, and that analysis of their structures is a process now in its infancy. The gains represented by work on a small number of nuclear receptors now need to be followed by success with more family members. The sharing of detailed experimental information can often be helpful and occasionally essential in accelerating such progress. ACKNOWLEDGMENT

We thank Dr. P. K. LeMotte for providing cells expressing the recombinant fragment of ER. LITERATURE CITED (1) McKenna, N. J., Xu, J., Nawaz, Z., Tsai, S. Y., Tsai, M. J., and O’Malley, B. W. (1999) Nuclear receptor coactivators: multiple enzymes, multiple complexes, multiple functions. J. Steroid Biochem. Mol. Biol. 69, 3-12. (2) Weatherman, R. V., Fletterick, R. J., and Scanlan, T. S. (1999) Nuclear-receptor ligands and ligand-binding domains. Annu. Rev. Biochem. 68, 559-581. (3) Kraichely, D. M., Sun, J., Katzenellenbogen, J. A., and Katzenellenbogen, B. S. (2000) Conformational changes and coactivator recruitment by novel ligands for estrogen receptor-R and estrogen receptor-β: Correlations with biological character and distinct differences among SRC coactivator family members. Endocrinology 141, 3534-3545. (4) Brzozowski, A. M., Pike, A. C. W., Dauter, Z., Hubbard, R. E., Bonn, T., Engstro¨m, O., O ¨ hman, L., Greene, G. L., Gustafsson, J. A., and Carlquist, M. (1997) Molecular basis of agonism and antagonism in the oestrogen receptor. Nature 389, 753-758.

Goldstein et al. (5) Tanenbaum, D. M., Wang, Y., Williams, S. P., and Sigler, P. B. (1998) Crystallographic comparison of the estrogen and progesterone receptors ligand binding domains. Proc. Natl. Acad. Sci. U.S.A. 95, 5998-6003. (6) Shiau, A. K., Barstad, D., Loria, P. M., Cheng, L., Kushner, P. J., Agard, D. A., and Greene, G. L. (1998) The structural basis of estrogen receptor/coactivator recognition and the antagonism of this interaction by tamoxifen. Cell 95, 927937. (7) Pike, A. C. W., Brzozowski, A. M., Hubbard, R. E., Bonn, T., Thorsell, A. G., Engstro¨m, O., Ljunggren, J., Gustafsson, J. K., and Carlquist, M. (1999) Structure of the ligand-binding domain of oestrogen receptor beta in the presence of a partial agonist and a full antagonist. EMBO J. 18, 4608-4618. (8) Grese, T. A., and Dodge, J. A. (1998) Selective estrogen receptor modulators (SERMs). Curr. Pharm. Des. 4, 71-92. (9) Bryant, H. U., and Dere, W. H. (1998) Selective estrogen receptor modulators: an alternative to hormone replacement therapy. Proc. Soc. Exp. Biol. Med. 217, 45-52. (10) Dhingra, K. (1999) Antiestrogens - tamoxifen, SERMs and beyond. Investigat. New Drugs 17, 285-311. (11) Egea, P. F., Klaholz, B. P., and Moras, D. (2000) Ligandprotein interactions in nuclear receptors of hormones. FEBS Lett. 476, 62-67. (12) Seielstad, D. A., Carlson, K. E., Katzenellenbogen, J. A., Kushner, P. J., and Greene, G. L. (1995) Molecular characterization by mass spectrometry of the human estrogen receptor ligand-binding domain expressed in Escherichia coli. Mol. Endocrinol. 9, 647-658. (13) Greene, G. L., Nolan, C., Engler, J. P., and Jensen, E. V. (1980) Monoclonal antibodies to human estrogen receptor. Proc. Natl. Acad. Sci. U.S.A. 77, 5115-5119. (14) Rosati, R. L., Da Silva Jardine, P., Cameron, K. O., Thompson, D. D., Ke, H. Z., Toler, S. M., Brown, T. A., Pan, L. C., Ebbinghaus, C. F., Reinhold, A. R., Elliott, N. C., Newhouse, B. N., Tjoa, C. M., Sweetnam, P. M., Cole, M. J., Arriola, M. W., Gauthier, J. W., Crawford, D. T., Nickerson, D. F., Pirie, C. M., Qi, H., Simmons, H. A., and Tkalcevic, G. T. (1998) Discovery and preclinical pharmacology of a novel, potent, nonsteroidal estrogen receptor agonist/antagonist, CP336156, a diaryltetrahydronaphthalene. J. Med. Chem. 41, 2928-2931. (15) Ke, H. Z., Paralkar, V. M., Grasser, W. A., Crawford, D. T., Qi, H., Simmons, H. A., Pirie, C. M., Chidsey-Frink, K. L., Owen, T. A., Smock, S. L., Chen, H. K., Jee, W. S., Cameron, K. O., Rosati, R. L., Brown, T. A., DaSilva-Jardine, P., and Thompson, D. D. (1998) Effects of CP-336, 156, a new, nonsteroidal estrogen agonist/antagonist, on bone, serum cholesterol, uterus and body composition in rat models. Endocrinology 139, 2068-2076. (16) Salman, M., Ruiz, A. A., Stotter, P. L., and Chamness, G. C. (1987) A progesterone receptor affinity chromatography reagent: 17 R-hexynyl nortestosterone Sepharose. J. Steroid Biochem. 26, 383-391. (17) Hegy, G. B., Shackleton, C. H., Carlquist, M., Bonn, T., Engstro¨m, O., Sjo¨holm, P., and Witkowska, H. E. (1996) Carboxymethylation of the human estrogen receptor ligandbinding domain-estradiol complex: HPLC/ESMS peptide mapping shows that cysteine 447 does not react with iodoacetic acid. Steroids 61, 367-373. (18) Geoghegan, K. F. (1996) Modification of amino groups, in Current Protocols in Protein Science. (J. E. Coligan, B. M. Dunn, H. L. Ploegh, D. W. Speicher, and P. T. Wingfield, Eds.) pp 15.2.1-15.2.18, John Wiley, New York. (19) el Garrouj, D., Aumelas, A., and Borgna, J. L. (1993) Steroidal affinity labels of the estrogen receptor. 1. 17 R-(Bromoacetoxy)alkyl/alkynylestradiols. J. Med. Chem. 36, 2973-2983. (20) Jonscher, K. R., and Yates, J. R. (1997) The quadrupole ion trap mass spectrometer - a small solution to a big challenge. Anal. Biochem. 244, 1-15. (21) Eng, J. K., McCormack, A., L., and Yates, J. R., III. (1994) An approach to correlate tandem mass spectral data of peptides with amino acid sequences in a protein database. J. Am. Soc. Mass Spectrom. 5, 976-989.

Preparing ER Ligand-Binding Domain for Crystallography (22) Vonderhaar, B., and Mueller, G. C. (1969) Binding of estrogen receptor to estradiol immobilized on insoluble resins. Biochim. Biophys. Acta 176, 626-631. (23) Sica, V., Parikh, I., Nola, E., Puca, G. A., and Cuatrecasas, P. (1973) Affinity chromatography and the purification of estrogen receptors. J. Biol. Chem. 248, 6543-6458. (24) Feng, W., Graumann, K., Hahn, R., and Jungbauer, A. (1999) Affinity chromatography of human estrogen receptoralpha expressed in Saccharomyces cerevisiae. Combination of heparin- and 17β-estradiol-affinity chromatography. J. Chromatog. A 852, 161-173.

Bioconjugate Chem., Vol. 12, No. 3, 2001 413 (25) Greene, G. L., and Jensen, E. V. (1982) Monoclonal antibodies as probes for estrogen receptor detection and characterization. J. Steroid Biochem. 16, 353-359. (26) Poirier, D., Labrie, C., Merand, Y., and Labrie, F. (1990) Derivatives of ethynylestradiol with oxygenated 17 alphaalkyl side chain: synthesis and biological activity. J. Steroid Biochem. 36, 133-142.

BC000127D

Chemical and Biochemical Issues Related to X-ray ... - ACS Publications

Recommend Documents