Interaction of d-tubocurarine analogs with the Torpedo nicotinic acetylcholine receptor. Methylation and stereoisomerization affect site-selective competitive binding and binding to the noncompetitive site.

Analogs of d-tubocurarine were used to determine the individual effects of methylation, stereoisomerization, and halogenation of d-tubocurarine on the affinity for each of the two acetylcholine (ACh) binding sites of the Torpedo nicotinic acetylcholine receptor (AChR) and for the noncompetitive antagonist site. Eight analogs were synthesized, including three new compounds: 7'-O-methyl-chondocurarine, 12'-O-methyl-chondocurarine, and 13'-bromo-d-tubocurarine. The two ACh sites differ in their affinities for d-tubocurarine by 400-fold, as shown by inhibition of [3H]ACh binding, whereas the affinity ratio for metocurine, the trimethylated derivative of d-tubocurarine, is reduced to 30 due to a decreased affinity for the high affinity site. Binding analysis of five d-tubocurarine analogs demonstrates that methylation of the phenols alone is responsible for the observed changes in affinity. Substitution with bromine or iodine at the 13'-position affected affinity at both sites with a net increase in site selectivity. Stereoisomers of d-tubocurare had decreased affinity for only the high affinity ACh site. Thus, the ring systems, including the 12'- and 13'-positions and the 1-position stereocenter, appear to be important in discriminating between the two ACh binding sites. Desensitization of the AChR was measured by increased affinity for [3H]phencyclidine. Binding to only the single, high affinity acetylcholine binding site, comprised by the alpha gamma-subunits, was required for partial desensitization of the AChR by d-tubocurarine and its analogs. Stronger desensitization, to the same extent observed in the presence of the agonist carbamylcholine, occurred upon binding by iodonated or brominated d-tubocurarine. Interaction of the analogs at the noncompetitive antagonist site of the AChR was also measured by [3H]phencyclidine binding. The bis-tertiary ammonium analogs of either the d- or l-stereoisomers bound to the noncompetitive antagonist binding site of the AChR with 100-fold higher affinity than the corresponding quaternary ammonium analogs.

The nicotinic acetylcholine receptor (AChR) 1 from Torpedo californica electric organ is a ligand gated cation channel com-posed of homologous subunits with a stoichiometry of ␣ 2 ␤␥␦ (Raftery et al., 1980;Noda et al., 1983). The five subunits each traverse the lipid bilayer and form a pseudo-symmetric pentameric rosette with the channel located at the central axis (Unwin, 1993). Channel opening is regulated by the binding of two molecules of acetylcholine (ACh) to sites on the extracellular surface of the protein (see Devillers-Thiery et al. (1993) for review). The ACh binding sites are nonidentical and can be distinguished by the differential binding of the plant alkaloid d-tubocurarine, a competitive antagonist (Neubig and Cohen, 1979). One ␣-subunit and the ␥-subunit of the AChR comprise the acetylcholine binding site with higher affinity for d-tubocurarine, while the second ␣-subunit and ␦-subunit comprise the site with lower affinity (Pedersen and Cohen, 1990;Blount and Merlie, 1989). Because the two ␣-subunits are identical, the distinct affinities of the two sites are likely to arise from interactions with the sites formed by differing aspects of the ␥and ␦-subunits.
Affinity labeling of ACh binding sites using sulfhydryl reactive compounds initially identified the ␣-subunit as the site of acetylcholine binding (Damle and Karlin, 1978;Wolosin et al., 1980). The snake venom toxin ␣-bungarotoxin could also be shown to bind with moderate affinity to ␣-subunit, even when denatured and proteolyzed (Wilson et al., 1984). Further studies using affinity labeling followed by mapping of the labeled sites to the amino acid sequence has identified residues in the ␣-subunit involved in acetylcholine binding: Cys-␣192 and Cys-␣193 (Kao et al., 1984) and the nearby residues, Tyr-␣190 (Abramson et al., 1989;Dennis et al., 1988), Tyr-␣198 (Middleton and Cohen, 1991), as well as residues more distant in the sequence, Trp-␣149 (Dennis et al., 1988), and Tyr-␣93 . The region ␣186 to ␣211 also binds ␣-bungarotoxin with low affinity and the residues identified by affinity labeling contribute substantially to this binding as judged by mutagenesis of these residues (Chaturvedi et al., 1993).
Several residues of the AChR that interact particularly with d-tubocurarine have been identified by affinity labeling or by site directed mutagenesis. The homologous residues Trp-␥56 and Trp-␦57 were identified by labeling and proteolytic mapping with [ 3 H]d-tubocurarine (Chiara and Cohen, 1992). Sitedirected mutagenesis of three residues of the mouse muscle AChR ␥-subunit (Ile-␥116, Tyr-␥117, and Ser-␥161) to the corresponding residues of the ␦-subunit could completely change the affinity to that characteristic of the ␣␦-site (Sine, 1993). The contribution of the subunits to the binding sites by particular amino acids is well characterized, but form an incomplete picture. Some of the amino acids are predicted to stabilize the quaternary ammonium of acetylcholine (e.g. Tyr-␣190, Tyr-␣93, Tyr-␣198, and Tyr-␥117: Cohen et al., 1991;Sine et al., 1994;Fu and Sine, 1994).
An alternative approach to examining the structure of the acetylcholine binding sites is to compare ligand structural analogs and determine quantitatively the influence of various substitutions upon binding energy (see Gund and Spivak (1991) for review). Many analogs of d-tubocurarine have been analyzed in search of a better muscle relaxant, but such studies utilized in vivo assays of potency that cannot be readily converted to binding affinity and do not take into account pharmacokinetics and other mechanisms such as open channel blockade. Further complications of d-tubocurarine structurefunction analysis was the correction of the structure by Everett et al. (1970) from a bis-quaternary ammonium to a monoquaternary, mono-tertiary ammonium as well as the appreciation by Soine (1978a, 1978b) that some previous synthetic procedures did not yield the expected derivatives (e.g. Marshall et al. (1967)). Interpretations of such experiments often emphasized the importance of bis-onium structure for antagonism (Sobell et al., 1972).
Experiments that examined the binding of d-tubocurarine and metocurine, its trimethylated derivative, revealed a difference in their site selectivity for binding to the two distinct acetylcholine binding sites on the AChR (Neubig and Cohen, 1979). Because the difference in structure lay in the methylation of the tertiary ammonium to a quaternary ammonium and methylation of the two phenols (see Fig. 1), we examined whether the difference could be ascribed to a particular site of methylation or resulted from smaller, additive effects on binding. In this report we describe the analysis of the binding properties of d-tubocurarine and 10 analogs. Binding is characterized for each of the acetylcholine binding sites and the noncompetitive antagonist site of the AChR. To determine how methylation at individual sites affects affinity, five analogs of d-tubocurarine were prepared, including two new compounds: 7Ј-O-methylchondocurarine and 12Ј-O-methylchondocurarine. In addition, two halogenated derivatives were prepared: 13Ј-iodo-d-tubocurarine (Menez et al., 1973) and 13Ј-bromo-d-tubocurarine. The results demonstrate that the charged nitrogens in d-tubocurarine need not be quaternary ammoniums for binding to the acetylcholine binding sites and that the changes in affinity are accounted for solely by methylation of the phenols. The effect of 13Ј-halogenation and stereoisomerization are also characterized. To further understand the binding of cholinergic ligands at the acetylcholine binding sites, it is desirable to determine a complete set of amino acids that interact with the ligands and to correlate the receptor sites that are important for binding with their points of contact on the ligand. The series of analogs presented here should facilitate such a characterization.

EXPERIMENTAL PROCEDURES
Materials-AChR-rich membranes were isolated from Torpedo californica electric organ (Marinus Inc., Long Beach, CA) as described in Pedersen et al. (1986) with the addition of calpain inhibitors I and II (10 mg/kg organ). Purified membranes typically contained 1-2 nmol of acetylcholine (ACh) binding sites/mg of protein measured by binding of [ 3 H]ACh as described below. Membranes were stored in 37% sucrose, 0.02% NaN 3 at Ϫ80°C under argon. Lower specific activity fractions (0.2-0.5 nmol of acetylcholine binding sites/mg of protein) were used for [ 3 H]ACh binding assays. Lauryl sulfate, ␣-bungarotoxin (␣-BgTx), carbamylcholine, l-bebeerine and Tris were from Sigma. Diisopropyl fluorophosphate and thiophenol were from Aldrich. Ci/mmol for binding assays and was shown to be 80% radiochemically pure by assay for ability to bind the AChR. l-Bebeerine was crystallized from 4:1 methanol:methylene chloride prior to using in assays. Meproadifen was synthesized according to Krodel et al. (1979) as described by Pedersen (1995).
Synthesis of d-Tubocurarine Analogs-The structures of the d-tubocurarine analogs used in this study are shown in Fig. 1. Three compounds were obtained commercially: d-tubocurarine (7Ј,12Ј-dihydroxy-6,6Ј-dimethoxy-2,2Ј,2Ј-trimethyltubocuraranium chloride), metocurine (6,6Ј,7Ј,12Ј-tetramethoxy-2,2,2Ј,2Ј-tetramethyltubocuraranium diiodide), and l-bebeerine ((1␤)-7Ј,12Ј-dihydroxy-6,6Ј-dimethoxy-2,2Ј-dimethyltubocuraran). The remaining compounds were synthesized as described below. The synthesis and purification of each compound was conveniently monitored using reversed phase high pressure liquid chromatography (HPLC). The elution of the compounds is shown in Fig. 2. HPLC was also used to establish purity of newly synthesized compounds, particularly the lack of contamination by related compounds. This was particularly important for proper interpretation of differences in binding affinity for the various compounds. Electrospray mass spectroscopy was routinely used to confirm the structure and assess possible FIG. 2. Reversed phase HPLC separation of curare compounds. A mixture of d-tubocurarine and analogs (2.5 nmol each) were injected onto a C18 reversed phase column and eluted with a gradient of solvent B (---). Elution was monitored by absorbance at 280 nm (--). Peak numbers correspond to the assignments in Fig. 1. Compounds IX and X were not included as they coelute with compound VI. Solvent A was H 2 O, 0.1% trifluoroacetic acid, solvent B was CH 3 CN, 0.09% trifluoroacetic acid. contamination by other curare compounds. The (M ϩ n14)/z peaks observed in mass spectra of I by Naghaway and Soine (1979) were not seen. Electrospray mass spectroscopy is likely a gentler method that is less prone to breakdown of the specimen; nonetheless we frequently observed a minor (M Ϫ 58)/z peak, even with commercial compounds after recrystallization.
Tubocurine (7Ј,12Ј-Dihydroxy-6,6Ј-dimethoxy-2,2Ј-dimethyltubocuraran)-Tubocurine (II) was obtained by demethylation of d-tubocurarine using the sodium thiophenoxide method of Shamma et al. (1966). This procedure was also used by Naghaway and Soine (1978b) to obtain II in good yield. Sodium thiophenoxide was prepared by treatment of thiophenol with NaOH in ethanol followed by addition of toluene and distillation to remove ethanol and water. Crystals of sodium thiophenoxide were collected by filtration and stored dessicated. d-Tubocurarine (5 g, 6.5 mmol) was dissolved in 150 ml of ethanol with mild heating. Sodium thiophenoxide (2.57 g, 19.4 mmol) was added in 10 ml of ethanol and allowed to stir for 3 h at room temperature. Precipitated NaCl was then removed by filtration and the filtrate concentrated to dryness by rotary vacuum evaporation. The residue was suspended in 500 ml of freshly distilled 2-butanone and refluxed for 4 h when HPLC indicated that the reaction had neared completion. The material was reconcentrated to dryness, suspended in water, and then dissolved by adding HCl. Any remaining insoluble material was removed by filtration. The filtrate was brought to pH 8 with saturated NaHCO 3 and extracted three times with an equal volume of ether. The ether extract was washed with 1 mM NaHCO 3 and concentrated to dryness to leave crude II. Tubocurine was crystallized from methanol. The crystallization was inefficient and required several repetitions to obtain a reasonable yield. The net yield was 2.091 g of white crystalline powder (3.52 mmol; 54%). Another 9% of the product was recovered by conversion to III with crystallization (see below). The tubocurine was pure as judged by HPLC. Electrospray mass spectroscopy revealed the m/z ϭ 596.3 and m/z ϭ 298.2, the expected values for mono-and diprotonated II, respectively.
Chondocurarine (7Ј,12Ј-Dihydroxy-6,6Ј-dimethoxy-2,2,2Ј,2Ј-tetramethyltubocuraranium Diiodide)-Chondocurarine (III) was synthesized from II by methylation with methyl iodide as described by Dutcher (1952). Tubocurine (1 g, 1.7 mmol) was dissolved in 18 ml of methanol and then reacted with 3.8 ml of methyl iodide overnight. Crystals of III formed spontaneously, and further crystallization was induced by addition of CHCl 3 to 25 ml. The crystals were collected and a second batch obtained from the dried filtrate: after dissolving the dried material in warm methanol an equal volume of CHCl 3 was added and crystals formed within several hours. A total of 1.32 g (1.5 mmol, 88%) was obtained after dessication. The product was pure as judged by HPLC, but III coelutes with I in our HPLC system (see Fig. 2). Since I was a potential product of the methylation reaction, the sample was analyzed by electrospray mass spectroscopy. This revealed a peak of m/z ϭ 312.2, corresponding to the expected double ion of III. There was no contamination by I which would produce a peak of m/z ϭ 609.3. A minor mass peak (ϳ7%) was also observed at 566.1 and corresponds to (M Ϫ 58)/z. O,O-Dimethyltubocurine (6,6Ј,7Ј,12Ј-Tetramethoxy-2,2Ј-dimethyltubocuraran)-O,O-Dimethyl tubocurine (IV) was synthesized by N-demethylation of VII as described by Naghaway and Soine (1978b). This method is analogous to the synthesis of II from I that is described above. Metocurine (0.502 g, 0.55 mmol) was converted to the chloride salt by stirring with freshly prepared AgCl in methanol. The dried metocurine chloride was dissolved in 10 ml of ethanol and stirred with 0.33 g of sodium thiophenoxide for 3 h. Precipitated NaCl was then removed by filtration. The dried filtrate was refluxed in 32 ml of freshly distilled 2-butanone for 3 h, reconcentrated to an oil and resuspended in 25 ml CHCl 3 . An equal volume water was added and the aqueous phase extracted four more times with CHCl 3 . The combined CHCl 3 extracts were concentrated to a yellow oil, dissolved in 50 ml of 10% HCl and extracted with ether (7 ϫ 50 ml). The aqueous phase was neutralized with NaOH and extracted with with CHCl 3 (4 ϫ 40 ml). The combined extracts were concentrated to dryness to yield the crude product (0.3 g, ϳ88%) which contained several unidentified contaminants, as detected by HPLC. These contaminants were removed by cation exchange chromatography over CM-Sephadex 25 (Pharmacia Biotech Inc.). A 1.5 ϫ 30-cm column was equilibrated with 10 mM HEPES, pH 7.0. After applying the product, the column was washed with 0.2 M NaCl, 10 mM HEPES, pH 7.0, and then eluted with an exponential gradient from 0.2 to 1.5 M NaCl. Fractions were assayed by absorbance at 280 nm and by HPLC. Fractions containing pure product were pooled, made basic with NaOH, and extracted with CHCl 3 . This yielded 0.18 g (0.29 mmol, 54%) pale yellow powder that was pure as judged by HPLC and by UV absorbance spectroscopy. Mass spectroscopy gave the expected m/z ϭ 623.3.
Fractions containing pure VI were neutralized with acetic acid and lyophilized yielding a yellow powder that contained significant salt contamination. This was rechromatographed over the same column and eluted with a step change to 0.4 M NH 4 HCO 3 , pH 9.0. The product was dried by rotary vacuum evaporation, redissolved in water, and brought to pH 2.0 with trifluoroacetic acid. This material was applied to a preparative C18 reversed phase HPLC column (22 ϫ 150 mm, Beckman Ultraspherogel) and eluted with a step change to 30% acetonitrile. The eluate was dried to a yellow powder that was essentially salt free (0.127 g, 0.18 mmol, 15%).
Fractions from the first cation exchange column that contained V also contained significant quantities of VI. To completely purify V, the fractions were neutralized with acetic acid, lyophilized, and redissolved in 10 ml of 100 mM NH 4 HCO 3 , pH 9.0. This material was reapplied to the same cation exchange column and eluted with a gradient from 100 to 400 mM NH 4 HCO 3 , pH 9.0. This acheived sufficient separation of VI from V. The fractions containing pure V were pooled and concentrated to dryness by rotary vacuum evaporation (41 mg, 54 mol, 5%).
Compounds V and VI were deemed pure by HPLC. HPLC traces of V contained no visible contaminants, while VI contains a small impurity visible as a shoulder on peak VI in Fig. 2. This minor contaminant was not a product of the reaction but appeared during purification. Absorption spectra of V and VI in 10 mM HCl (Fig. 3) revealed extinction coeffecients at 280 nm similar to those of chondocurarine (III) and metocurine (VII). Mass spectra of the two compounds were similar yielding the predicted double ion mass peak (m/z ϭ 319.2).
Since the two products are similar and have identical molecular weights, the location of the methoxy groups were assigned by 13 C NMR. The data for the region, including the methoxy groups, are as follows: Compound III: 56.09 and 56.13 ppm; compound V: 56.00 ppm, 56.10 ppm, and 59.80 ppm; compound VI: 55.76 and 56.03 ppm (2 ϫ); compound VII: 55.8 ppm, 56.2 ppm (2 ϫ), and 59.8 ppm. The peak at 59.8 ppm is characteristic of the middle aryl methoxy substituent (Breitmaier and Voelter, 1987) and thus identifies compound V as the 7Ј-Omethylchondocurarine and compound VI as 12Ј-O-methylchondocurarine. The NMR data corresponded well to those obtained by Koike et al. (1981) for the corresponding analogs of l-bebeerine.
Bromo-d-tubocurarine (13Ј-Bromo-7Ј,12Ј-dihydroxy-6,6Ј-dimethoxy-2,2Ј,2Ј-trimethyltubocuraranium chloride)-Bromo-d-tubocurarine (IX) was synthesized by reaction of bromine with I. The reaction is completely analogous to the iodination described in Menez et al. (1973) for the synthesis of VIII. d-Tubocurarine (1 g, 1.3 mmol) was dissolved in 100 ml of methanol, to which was added 2.6 ml of 1 M Br 2 in methanol. The reaction was stopped after 1 min by addition of 2.6 ml of 1 M sodium thiosulfate. The reaction was dried, redissolved in 100 ml of water, the pH adjusted to 10.5, and the material chromatographed over an AG11A8 (Bio-Rad) ion retardation resin. The pooled fractions were concentrated to dryness and redissolved in 6 ml of 12% acetonitrile. This was applied to a 22 ϫ 250-mm C18 reversed phase HPLC column (Vydac) in six batches and eluted with a gradient of acetonitrile from 10 to 50%. The pure fractions were pooled and concentrated to dryness (0.39 g, 0.46 mmol, 35%). The product was pure as judged by HPLC and by mass spectroscopy (m/z ϭ 689.2 and 687.2 and m/z ϭ 345.2 and 344.2 for the protonated double ion).
Ligand Binding Assays-Binding assays were carried out in HEPES-Torpedo physiological saline solution (HTPS: 250 mM NaCl, 5 mM KCl, 3 mM CaCl 2 , 2 mM MgCl 2 , 0.02% NaN 3 , 20 mM HEPES, pH 7.0). Ligand binding assays with [ 3 H]ACh, [ 3 H]PCP, or with [ 3 H]d-tubocurarine were performed by centrifugation as described by Pedersen (1995). For [ 3 H]ACh binding assays, the AChR-rich membranes, or a less pure side fraction of membranes from the discontinuous sucrose gradient fractionation, were first incubated with diisopropylfluorophosphonate to inactivate acetylcholinesterase. Membranes (100 g) were then incubated in HTPS at room temperature with the indicated concentrations of ligand for 30 min and then centrifuged at 19,000 ϫ g for 30 min in a TOMY MTX-150 microcentrifuge to separate bound from free ligand. The free ligand concentration was determined by counting an aliquot of the supernatant. Bound ligand was determined by counting the pellet after dissolving in 10% SDS. Nonspecific binding was determined by inclusion of a competitive inhibitor in high concentration.
125 I-␣-BgTx Binding Assay-Binding of 125 I-␣-BgTx was measured using the DE-81 filter binding method of Schmidt and Raftery (1973). AChR-rich membranes were incubated in 60 l of HTPS containing 0.1% BSA with 125 I-␣-BgTx. After incubation for 45 min or for 1 h the reaction was diluted 5-fold into 10 mM Tris, pH 7.4, 0.1% Triton X-100, 0.1% BSA with 300 nM ␣-BgTx to stop any further binding. Aliquots (60 l) of the samples were then spotted onto DE-81 filters. The filters were batch-washed together twice with 100 ml of 10 mM Tris, pH 7.4, 50 mM NaCl, 0.1% Triton X-100, blotted, and then counted for bound 125 I-␣-BgTx. In HTPS, the incubation time with 125 I-␣-BgTx is within the linear portion of the binding reaction and this condition measures the initial rate of binding. When competing ligands were included they were preincubated with the AChR for 30 prior to addition of 125 I-␣-BgTx. Extended incubation in HTPS reveals binding to approximately half of the sites measured by [ 3 H]ACh binding, whereas incubation in 10 mM Tris, pH 7.4, 0.1% Triton X-100, 0.1% BSA shows stoichiometric binding. These results are consistent with those of Conti-Tronconi et al. (1990) that indicate negative cooperativity of ␣-BgTx binding to the AChR in physiological buffers.
Ligand Binding Data Analysis-Inhibition data were analyzed by nonlinear least squares fitting of the data to models for single site inhibition, B I ϭ A/(1 ϩ I/K app ) ϩ Bcg, and for inhibition at two equimolar sites, B I ϭ A{1/(1ϩ I/K1 app ) ϩ 1/(1 ϩ I/K2 app )} ϩ Bcg, where B I is the concentration of bound ligand, A the maximum concentration of bound ligand at each site, I the inhibitor concentration, Bcg the nonspecific or background level of binding, and K app the concentration of inhibitor required to produce a 50% effect. Inhibition of 125 I-␣-BgTx binding by halogenated analogs was not well fit by this equation but could be better described using the following equation: A 1 and A 2 represent variable site stoichiometry. Although such a description is inconsistent with the expected 1:1 ratio of ACh binding sites on the AChR, this equation nontheless consistently fit this particular data set better than other models. Nonlinear least squares fitting was performed using the program Sigmaplot (Jandel Scientific version 4.1 or Windows version 2.0). Equilibrium dissociation constants for inhibitors (K I ) were determined from the K app values: where L is the free radioactive ligand concentration, and K d is the equilibrium dissociation constant for the radioactive ligand. The K d for [ 3 H]ACh was determined in independent experiments to be 17 nM. For inhibition of the initial rate of 125 I-␣-BgTx binding, K I ϭ K app .
For [ 3 H]PCP binding experiments performed in the absence of carbamylcholine, binding was sometimes increased due to the allosteric effects of binding to the ACh sites. Such data were fit to an equation describing simple binding: B I ϭ AI/(I ϩ K) ϩ Bcg. For some ligands, this effect was followed by direct competitive inhibition at higher concentrations of ligand. These data were modeled using an equation to describe binding modulated by a direct inhibition function.
In this equation, B 0 represents the binding of [ 3 H]PCP in the absence of other ligands, A is the amplitude of binding induced by ligand I, K 1 is the corresponding dissociation constant, and K 2 is the binding constant for inhibition. Data from experiments performed in the presence of carbamylcholine were fit to the equations described above for single site competitive inhibition.
Ancillary Methods-Protein assays were performed using a bicinchoninic assay (BCA microassay, Pierce) with bovine serum albumin as a standard. HPLC was performed using a Beckman 125 pump; detection was by absorbance using Beckman 166 variable wavelength detector; gradient formation and data collection were computer-controlled using Beckman System Gold software. Mass spectroscopy was carried out at the Baylor Mass Spectroscopy Core facility. NMR spectroscopy was carried out at the Baylor NMR facility by Dr. Mohan Chari.
pH titrations were carried out with 100 M solutions of d-tubocurarine, chondocurarine, 7Ј-O-methylchondocurarine, 12Ј-O-methylchondocurarine, and metocurine in buffers of varying pH. The buffers used were sodium phosphate from pH 6 -7, Tris-HCl from pH 7.25 to 9, and sodium carbonate from pH 9.25 to 11. Each buffer was at 46.7 mM final concentration. Absorbance measurements were carried out on a Beckman DU-50 spectrophotometer.

RESULTS
Analogs of d-tubocurarine (I) were synthesized to analyze the effects of specific modifications on the binding affinity for the AChR. The procedures for the isolation and synthesis of many analogs and derivatives of d-tubocurarine have been described previously (e.g. Dutcher, 1946Dutcher, , 1952. Compounds II, III, IV, VIII, IX, and XI were, therefore, prepared essentially according to published procedures with the modifications indicated under "Experimental Procedures." In each case they were the predominant expected product, and the structure was corroborated by mass spectroscopy. To examine the particular effects of O-methylation on binding to the AChR, two new compounds were synthesized: 7Ј-Omethylchondocurarine (V) and 12Ј-O-methylchondocurarine (VI). The UV absorption spectra of the starting material, compound III, and compounds V, VI, and VII in acid and in base are shown in Fig. 3. The absorbance peak of III increases and shifts from 280 to 290 nm with increased pH. The change in absorbance is most pronounced in the region from 295 to 310 nm and is similar to the changes seen with d-tubocurarine (data not shown; Kalow, 1954). The change in the spectra are ascribed to titration of the 7Ј-and 12Ј-phenols to phenolate ions. Because the corresponding methoxy groups are untitrat- able, the fully methylated VII shows no appreciable spectral changes. Titration of V and VI resulted in spectral changes similar to each other, but that differed from III in having approximately half the increase in absorbance near 300 nm. This is consistent with titration of only one phenol each.
By measuring absorbance changes, Kalow (1954) calculated pK values of 8.1 and 9.1 for the two phenols of d-tubocurarine and assigned them to the 12Ј and 7Ј-positions, respectively. The pK values were determined for I, III, V, and VI by measuring the change in absorbance at 300 nm with pH ( Table I). The titrations data for I and III were well fit by two titratable groups. The pK values determined for I agree with the results of Kalow (1954). Titrations of compounds V and VI were well fit to a single titratable groups with pK values that differ by less than 0.3 pH units. However, the pK values are reversed in the order expected from the assignments of Kalow (1954). The changes in the pK values from the values of d-tubocurarine likely reflect the absence of the two other titratable groups (one ammonium and one phenol) in each of these analogs. The results nevertheless demonstrate the presence of one titratable phenol in each compound, consistent with the structures for V and VI given in Fig. 1.
The Effects of Methylation on Binding to the Agonist Sites of the AChR-The binding of d-tubocurarine analogs to the ACh binding sites was analyzed by competitive inhibition of [ 3 H]ACh binding. The data for the various N-and O-methylated analogs is shown in Fig. 4. Each set of data was fit to an equimolar two-site binding model by nonlinear regression; the fitted curve is shown by the corresponding solid lines. The K app values determined for inhibition at each site were used to calculate the corresponding dissociation constants, K I values, as described under "Experimental Procedures." Inhibition of [ 3 H]ACh binding by d-tubocurarine (I) shows biphasic inhibition (Fig. 4A) with a 450-fold site selectivity (the ratio of K I2 / K I1 , Table II). The dissociation constants (K I values, Table II) agree well with previously published binding constants determined by direct binding of [ 3 H]d-tubocurarine and by competitive binding assays (Neubig and Cohen, 1979;Pedersen and Cohen, 1990). The higher affinity site is comprised by the ␣and ␥-subunits of the AChR and will be referred to as the ␣␥-site in the following discussion; the low affinity site will be referred to as the ␣␦-site.
The site selectivity of metocurine, the trimethylated analog of d-tubocurarine, is reduced compared with that of d-tubocurarine. This is a result of a 10-fold decrease in affinity for the ␣␥-site with no significant change in the affinity for the ␣␦-site ( Fig. 4A; Table II). To determine the effect of methylation at the amines, the binding of the bis-tertiary ammonium analog, tubocurine (II), and the bis-quaternary ammonium analog, chondocurarine (III), were examined (Fig. 4B). The affinities at each site were comparable with those of d-tubocurarine, exhibiting less than 2-fold changes in the K I values, and demonstrate that tertiary and quaternary ammoniums interact equally well at the binding site (Table II). This result further suggested that methylation of the two phenols was responsible for the affinity change displayed by metocurine at the ␣␥-site. This was confirmed by the binding of the bis-tertiary ammonium analog, O,O-dimethyl tubocurine (IV), which had binding affinities indistinguishable from metocurine ( Fig. 4B and Table II).
To examine whether the affinity change due to methylation of the phenols could be accounted for by a single modification or resulted from effects exerted by both modifications, the binding of 7Ј-O-methylchondocurarine (V) and 12Ј-O-methylchondocurarine (VI) was determined (Fig. 4C). Comparison of the K I

FIG. 4. Effect of methylation of d-tubocurarine analogs on binding to the AChR as determined by inhibition of [ 3 H]ACh binding. AChR-rich membranes (100 g; 36 pmol of ACh binding sites) were incubated with 100 nM [ 3 H]ACh and the indicated concentrations of competing ligand in 1 ml of HTPS. Bound [ 3 H]ACh was then determined after removal of free [ 3 H]ACh by centrifugation as described under "Experimental
Procedures." Each panel shows a separate experiment that included controls with no added ligand (Ⅺ) and with 100 M carbamylcholine (E). A, d-tubocurarine (I, q), metocurine (VII, ç). B, chondocurarine (III, ࡗ); tubocurine (II, f); O, O-dimethyltubocurine (IV, å). C, 12Ј-O-methylchondocurarine (VI, q), 7Ј-O-methylchondocurarine (V, ç). Each set of data was fitted to a model for inhibition at two equimolar independent sites (190). Each data point is the average of duplicate determinations that generally varied less than 5%. values for V and VI with the parent compound III (Table II) indicates that 7Ј-O-methylation has no effect on the binding to the ␣␥-site and increased affinity 2-fold for the ␣␦-site. Methylation at the 12Ј-position decreases affinity at the ␣␥-site 7-fold and at the ␣␦-site 3-fold. Thus, 7Ј and 12Ј methylation both affect binding affinities, but the effect exerted by 12Ј modification applies to both sites, whereas the effect of 7Ј modification is weaker and applies only at the ␣␦-site.
The effect of methylation on binding affinity to the ␣␥and ␣␦-site were also examined by inhibition of the initial rate of binding of 125 I-␣-BgTx. The net site selectivity observed was lower using this assay (typical selectivity for d-tubocurarine was 50 -150-fold). Nonetheless, the changes in site selectivity were similar to those observed using inhibition of [ 3 H]ACh binding. The conclusions on the effects of methylation on the affinity and site-selectivity are similar (data not shown).
Effect of 13Ј-Halogenation on Binding Affinity-The more pronounced effect of 12Ј-O-methylation on binding affinity suggested that this portion of d-tubocurare interacted directly at the binding site. The structure of d-tubocurarine shows this portion of the molecule pointing into the surrounding solution. Iodination at the adjacent 13Ј-position had been demonstrated by Menez et al. (1973), and this compound was shown to be biologically active, but its binding properties have not been characterized in detail. Therefore, we examined the binding properties of the iodo-and bromo-derivatives of d-tubocurarine, compounds VIII and IX. Inhibition of [ 3 H]ACh binding to the ACh binding sites is shown in Fig. 5A, and the corresponding K I values are listed in Table II. Both VIII and IX displayed 2-fold increased affinity for the ␣␥-site and 4-fold decreased affinity for the ␣␦-site. This resulted in a significant increase in site selectivity to more than 2000-fold.
Examination of the affinities of d-tubocurarine and iodo-dtubocurarine by inhibition of the initial rate of binding of 125 I-␣-BgTx yielded a similar 5-fold increase in site selectivity for iodo-d-tubocurarine ( Fig. 5B; for I, K I1 ϭ 45 nM, K I2 ϭ 2.5 M, and selectivity is 57-fold; for VIII, K I1 ϭ 50 nM, K I2 ϭ 15 M, and selectivity is 300-fold). Bromo-d-tubocurarine (IX) displayed a similar inhibition pattern. However, the inhibition by the halogenated derivatives could not be fit using the equation for inhibition at two equimolar sites. These data were consistently better fit to an equation with variable site stoichiometry. A ratio of 2 to 1 of high affinity to low affinity sites was typically observed. The inhibition curves for all other derivatives tested could be fit well using the model with two equimolar sites. This difference may be related to the ability of the halogenated analogs to desensitize the AChR more strongly as shown below in Fig. 8, thereby causing noncompetitive effects on the binding of 125 I-␣-BgTx at the low affinity site. Similar noncompetitive effects on ␣-toxin binding have also been observed by a desensitizing noncompetitive antagonist (Krodel et al., 1979).
The High Affinity Site Is the ␣␥-Site for the d-Tubocurarine Derviatives-To ensure that the effects of methylation and halogenation on site selectivity were not the result of more radical affinity changes that resulted in inversion of site selectivity between the two sites, [ 3 H]d-tubocurarine was used as the radioligand at a concentration such that binding was primarily to the ␣␥, high affinity site. The ability of the methylated and halogenated analogs to compete for this binding at concentrations consistent with binding to the ␣␥-site demonstrated that no inversion of site selectivity had occurred (data not shown).
Effect of Altered Stereochemistry on Binding-Various stereoisomers related to d-tubocurarine have been isolated from natural products. Most have been characterized as noncholinergic, suggesting that they bind the AChR poorly despite the the structural homology to d-tubocurarine. l-Bebeerine (X) is a stereoisomer of tubocurine (II) with an inverted configuration at carbon 1 (Fig. 1). Comparison of the binding properties should reveal the importance of the correct stereoconfiguration. The dimethiodide of l-bebeerine, a stereoisomer of chondocurarine, was also synthesized (compare XI versus III in Fig. 1). Fig. 6A and the corresponding K I values in Table II. Inhibition by X was incomplete at 300 M and K I2 is therefore poorly determined. Higher concentrations of X could not be used as they disrupted membrane pelleting in the assay. Therefore binding of X was also examined by inhibition of the initial rate 125 I-␣-BgTx binding as shown in Fig. 6B. These data are well fit by a single inhibition constant of 5.6 M, suggesting no site selectivity. This value is reasonably consistent with the K I1 determined by inhibition of [ 3 H]ACh binding, but differs substantially from K I2 . Inhibition of 125 I-␣-BgTx binding by d-tubocurarine is also shown (Fig. 6B, K I1 ϭ 25 nM and K I2 ϭ 4 M). Inhibition of [ 3 H]ACh binding by XI was also described by a two-site fit (Fig. 6A). The value for K I2 was similar to those of the d-isomers but K I1 was substantially higher. The K I1 value for X and XI was similar (Table II). Thus, the primary effect of stereoisomerization appears to be a 50-fold reduction in affinity at the ␣␥ site with only a small effect at the ␣␦-site. , and controls with no added ligand (E) and with 100 M carbamylcholine (Ⅺ). Each set of data was fitted to a model for inhibition at two equimolar independent sites (--). Each data point is the average of duplicate determinations that generally varied less than 5%. B, 125 I-␣-BgTx binding: AChR-rich membranes (2.5 nM in 60 l HTPS) were preincubated with the competing ligand for 30 min and then 125 I-␣-BgTx was added to 2 nM and bound 125 I-␣-BgTx determined after further 45-min incubation as described under "Experimental Procedures." Data are shown for d-tubocurarine (I, q) and iodo-d-tubocurarine (VIII, f). The data for I was fit to a model for inhibition at two equimolar independent sites, whereas the data for VIII was fit to a similar model that incorporated variable site stoichiometry (--). Controls are shown for no added competitor (E) and 300 nM ␣-BgTx (Ⅺ). Each data point in B is the average of duplicate determinations that generally varied less than 10%.

Interaction of Curare Analogs with the Noncompetitive Antagonist Binding Site-The data for inhibition of [ 3 H]ACh bind-
ing and 125 I-␣-BgTx binding by l-bebeerine and iodo-d-tubocurarine suggested the presence of allosteric effects in addition to strictly competitive binding at the ACh sites. One potential source of allosteric modulation is through the noncompetitive antagonist (NCA) site of the AChR. Therefore, the binding of the curare analogs to the NCA site was examined by inhibition of [ 3 H]PCP binding. PCP binds at the NCA site with ϳ5-fold higher affinity to the desensitized conformation than the resting conformation. The following experiments were carried out using a low concentration of [ 3 H]PCP (ϳ1 nM). Because this concentration is substantially lower than the dissociation constant (near 1 M), only a small fraction of the AChR are occupied. The amount bound, therefore, may reflect changes in affinity due to allosteric modulation (e.g. by agonist binding to the ACh binding sites) in addition to being inhibitable by direct competitive binding at the NCA site. The allosteric regulation is illustrated by the data performed in the presence of carbamylcholine, which induces desensitization, and therefore results in more observed binding than in its absence (Fig. 7,  compare filled with open squares). The inclusion of 1 mM carbamylcholine also serves to block binding to the ACh sites by the competing ligand. Tubocurine inhibited [ 3 H]PCP binding in the presence of carbamylcholine, whereas chondocurarine had little effect (Fig.  7A, filled symbols). In the absence of carbamylcholine, both compounds increased [ 3 H]PCP binding to a level near 60% of the binding observed in the presence of carbamylcholine (Fig.  7A, open symbols). For tubocurine, this was followed by inhibition of binding at higher concentrations (Fig. 7A, open triangles) yielding a bell-shaped curve. The enhanced binding in-duced at the lower concentrations was presumably due to desensitization of the AChR induced upon binding of the ligand at the high affinity ACh binding site (␣␥-site). The increase in binding was well fit to curves describing binding at a single site (Fig. 7A, solid lines), and the K app values (230 nM for chondocurarine) were consistent with titration of the high affinity binding sites. The K app was substantially lower than the K I2 for binding to the low affinity ␣␦-site.
The l-isomers of chondocurarine and tubocurarine, dimethyll-bebeerine and l-bebeerine, were likewise examined for their effects on [ 3 H]PCP binding (Fig. 7B). In the presence of carbamylcholine, l-bebeerine could fully inhibit binding, whereas dimethyl-l-bebeerine required 100-fold higher concentrations for inhibition (Fig. 7B, filled symbols). The l-isomers also increased binding in the absence of carbamylcholine, as did the d-isomers, but the effect was only 2-fold (Fig. 7B, open symbols), and only l-bebeerine inhibited binding at higher concentrations (open triangles). Thus, [ 3 H]PCP binding to the NCA site was inhibited ϳ100fold more potently by the bis-tertiary analogs, tubocurine, and l-bebeerine, than their bis-quaternary counterparts, chondocurarine and dimethyl-l-bebeerine. The K I values for inhibition are given in Table III. The K I values were decreased in the presence of carbamylcholine, suggesting that tubocurine and l-bebeerine bind with higher affinity to the NCA site in the desensitized conformation. Likewise, metocurine increased [ 3 H]PCP binding in a manner similar to that of chondocurarine whereas O,O-dimethyl tubocurine displayed a pattern similar to tubocurine (data not shown). Thus, only the bis-tertiary compounds bind the NCA site with appreciable affinity.
Iodo-d-tubocurarine increased [ 3 H]PCP binding to the same extent as carbamylcholine whereas d-tubocurarine increased binding to only 60% of that level (Fig. 8). In each case, the K app   FIG. 6. l-Bebeerine (X) and N,N-dimethyl-l-bebeerine (XI) bind the agonist sites with low affinity. A, [ 3 H]ACh binding: AChR-rich membranes (100 g; 36 pmol of ACh binding sites) were incubated with 100 nM [ 3 H]ACh and the indicated concentrations of competing ligand in 1 ml of HTPS. Bound [ 3 H]ACh was then determined as described under "Experimental Procedures." Data are shown for l-bebeerine (X, q), N,NЈ-dimethyl-l-bebeerine (XI, ç, and controls with no added ligand (Ⅺ) and with 100 M carbamylcholine (E). Each set of data was fitted to a model for inhibition at two equimolar independent sites (OO). B, 125 I-␣-BgTx binding: AChR-rich membranes (1.2 nM ACh binding sites) were incubated with the indicated concentrations of competing ligand for 30 min. To measure the initial rate of binding, 125 I-␣-BgTx was added to 2 nM and further incubated 1 h as described under "Experimental Procedures." Data are shown for d-tubocurarine (I, q) and l-bebeerine (X, f) with corresponding fits of the data to models for inhibition at two sites or one site, respectively (--). Controls are shown for no added competitor (E) and 400 nM ␣-BgTx (Ⅺ). Each data point in B is the average of duplicate determinations that generally varied less than 10%.  ). B, N,NЈ-dimethyl-l-bebeerine (E, q); l-bebeerine (É, ç). For data obtained in the presence of carbamylcholine, the solid lines represent the best fit to a model for inhibition at a single site. For the data obtained in the absence of carbamylcholine (E), the solid lines represent the best fit to a model of a simple binding function. When inhibition was also observed at high concentrations (É), the fit was to a binding function modulated by an inhibition function, as described under "Experimental Procedures." for the binding increase was consistent with titration of only the high affinity, ␣␥-site. Bromo-d-tubocurarine increased binding to the same extent as iodo-d-tubocurarine (data not shown). Only the halogenated analogs increased binding to the level observed in the presence of carbamylcholine.

DISCUSSION
The work presented in this article correlates the structure of d-tubocurarine with its site selectivity for the nicotinic acetylcholine binding sites. The initial observation that metocurine has 10-fold lower affinity for the ␣␥-site and unchanged affinity for the ␣␦-site, as compared with d-tubocurarine, provided a starting point for determination of structural features that affect site selectivity. The data demonstrate that methylation of the phenolic groups alone was responsible for the affinity changes observed with metocurine. It was further shown that halogenation at the 13Ј-position and stereoisomerization at the 1-position also affect site selectivity. Together the data support the notion that the phenyl ring bearing the 12Ј and 13Ј substituents and the fused rings that bear the carbon 1 stereocenter interact with residues important for site selectivity.
Independent Effects of Methylation upon Binding Affinity-The insignificant changes in affinity observed upon changes in N-methylation from tertiary to quaternary ammoniums demonstrated that there is no requirement for a quaternary ammonium in d-tubocurarine for binding to the ACh sites. While the importance of the positive charge is undisputed, this observation is consistent with the general lack of correlation be-tween the potency of agonists and the successive state of methylation of the positive center associated with agonists (see Gund and Spivak (1991) for review). Recent studies of mutant AChR have indicated the importance of Tyr-␣93, Tyr-␣190, Tyr-␣198, and Tyr-␥117 Fu and Sine, 1994) to the binding of metocurine by stabilization of the quaternary ammonium. The results present here show that a quaternary ammonium interaction is not required. It may be that tertiary ammonium ligands are equally well accomodated by these particular residues or that other portions of metocurine actually interact with these residues.
As N-methylation did not account for the observed affinity changes in metocurine, the unmethylated II was compared with its O,O-dimethyl analog, IV, to reveal differences due to methylation at only the phenolic groups. Those differences fully account for the change in affinity between d-tubocurarine and metocurine. The individual contributions of each O-methyl modification were established using 7Ј-O-methylchondocurarine and 12Ј-O-methylchondocurarine and demonstrated that each methylation contributed to the total affinity change. The free energy of binding (⌬G) to each of the sites was calculated from the binding constants for compounds III, V, VI, and VII (Table IV). The change in free energy of binding (⌬⌬G) relative to III was then calculated for each compound at each site. From the values it can be seen that the free energy changes due to methylation at the individual sites when summed are nearly equal to the free energy change when metocurine is compared with chondocurarine. This demonstrates that the effects of individual methylations account for the observed affinity changes of metocurine compared with chondocurarine. The individual effects of methylation are, therefore, additive and unlikely to involve allosteric or synergistic effects on binding and the changes in binding due to each methylation can be interpreted separately.
The effect of 7Ј-O-methylation is weaker and increases binding affinity only to the ␣␦-site. Methylation at the 12Ј-phenol decreases binding affinity at both sites but to different extents, having a nearly 10-fold effect on binding to the ␣␥-site. The compensatory effects at the ␣␦-site result in unchanged affinity when both 7Ј-and 12Ј-O-methylations are present. Halogenation of 13Ј-position produced smaller affinity changes at each site, but in opposite directions and resulting in a substantial change (5-fold) in site selectivity. The adjacent 12Ј and 13Јpositions have the strongest effects observed with the modifications tested here and suggests that this part of the structure interacts directly with the binding site. It is tempting to speculate that this portion of d-tubocurarine interacts particularly with a portion of the site that affects site selectivity, particularly residues of the ␥ and ␦ subunits, but such a conclusion must be tempered by the observation that mutagenesis of an ␣-subunit residue can also affect site selectivity (Tyr-␣190; Sine et al., 1994).
Allosteric Interactions-To determine the binding constants for the d-tubocurarine analogs, we relied upon inhibition of [ 3 H]ACh binding with saturating concentrations of [ 3 H]ACh and calculated the K I values based on the affinity of [ 3 H]ACh for the AChR. However inhibition at the low affinity and high affinity sites reflect different situations. Competition for binding at the high affinity site occurs with [ 3 H]ACh presumably present on the low affinity site, whereas competition for the low affinity site presumably has the competing ligand present at the high affinity site. In the case of the low affinity site, it is possible that the equilibrium dissociation constant for [ 3 H]ACh differs from the experimentally determined value because of the presence of the competing ligand at the high affinity site. This may compromise the calculation used to determine the K I2 H]PCP was determined as described under "Experimental Procedures." The effects of d-tubocurarine (I, E) and iodo-dtubocurarine (VIII, Ⅺ) are shown. Data for the absence of added ligand (छ) and for the presence of 1 mM carbamylcholine (f) and 50 M proadifen (ࡗ) are also shown. The solid lines represent the best fits to a single site binding function. For I, K app ϭ 90 nM, and for VIII, K app ϭ 60 nM. Each point is the average of duplicate determinations which varied less than 10%.
at the low affinity site (see "Experimental Procedures"). The dominant conformational change is the equilibrium between the resting and desensitized conformations. The Torpedo AChR is 10 -20% desensitized in the absence of ligand (Cohen and Strnad, 1987 and references therein), defining an allosteric equilibrium constant of M ϭ 0.1-0.2 for this conformational change. d-Tubocurarine and its variously methylated analogs will desensitize the AChR to 60%, effectively changing M to near 1. This could potentially result in a 5-fold change in affinity for [ 3 H]ACh. More importantly, the change will be similar for each of the methylated analogs as they desensitize to the same extent (Figs. 7 and 8 and associated text). Thus, while the accuracy of the binding constant to the low affinity site may be subject to this systematic error, the comparison between the methylated analogs remains valid. The strong agreement between the competitive inhibition analysis and the results of direct binding experiments (Neubig and Cohen, 1979) suggests that the error is small.
Simultaneous binding to the NCA site could also potentially alter the pattern of binding by the competing ligand through allosteric effects. However, the curare analogs that include a quaternary ammonium do not bind the NCA site with sufficient affinity to interfere in the assays (Figs. 7 and 8). The tertiary ligands tubocurine and O,O-dimethyltubocurine bind the NCA site with moderate affinity (Table III) and with a slight, 2-fold preference for the desensitized conformation. Thus, the conformational changes due to binding at the NCA are unlikely to produce more than a 2-fold error in the K I for the competing ligand, and such effects would only occur near 100 M concentrations. This would only affect the latter part of the curves for inhibition at the low affinity site.
The inhibition of the initial rate of 125 I-␣-BgTx binding does not require knowing the affinity for 125 I-␣-BgTx and will not be subject to the same constraint as for [ 3 H]ACh binding. For the methylated analogs, somewhat lower site selectivity was routinely observed using this assay. Nonetheless, the changes in site selectivity were fully consistent with those observed by [ 3 H]ACh binding, supporting the conclusions listed above. Interpretation of this kinetic assay may, nonetheless, have other caveats, as suggested by the unusual pattern observed when halogenated analogs were used.
Stereoisomerization Affects Binding to the ␣␥-Site-The effect of altered stereochemistry at the carbon 1, one of the two stereocenters in curare compounds, revealed dramatically decreased affinity (50-fold) to the ␣␥-site, when the l-isomers are compared directly with the corresponding d-isomers (Table II). The effects of binding at the ␣␦-site were initially difficult to assess because of the incomplete inhibition of [ 3 H]ACh binding by l-bebeerine (Fig. 6). Nonetheless, l-bebeerine fully inhibited the initial rate of 125 I-␣-BgTx binding with a K app of of 5.6 M, suggesting equivalent binding at each site. Comparison of the bis-quaternary isomers, chondocurarine versus N,NЈ-dimethyll-bebeerine, which do not bind the NCA site, indicated no change in binding affinity to the ␣␦-site. Thus, proper stereo-chemistry appears to be important primarily for binding to the ␣␥-site.
Energy minimization of l-bebeerine using the MM2 algorithm (Hyperchem version 2.0) was compared with the energy minimized structure of d-tubocurarine (data not shown). An exact alignment of the molecules in the vicinity of the 2Јammonium and the carbon 1Ј stereocenter was achieved with the major deviations in the fused ring structure associated with the carbon 1. While it is tempting to speculate that inversion at the carbon 1 decreases binding affinity due to local changes in that vicinity, and therefore that portion of tubocurare interacts most closely with the ACh sites, it is equally possible that perturbation of the structure as a whole affects the binding affinity. Nonetheless, the ring system that includes the 1-position must interact uniquely with the ␣␥-site and suggests that site-specific interactions are closely associated with this portion of the molecule. Testing the other stereoisomer of d-tubocurare, d-bebeerine (chondodendrine), which has inverted stereochemistry at the carbon 1Ј, should elucidate this issue. The potency of methylated d-bebeerine was 40-fold less than d-tubocurarine (King, 1935) suggesting little activity of this stereoisomer. If that result relates to binding to the ACh sites it would suggest that changes in both stereocenters vitiates binding and that both portions of the structure are important.
Interaction with the NCA Site-l-Bebeerine could potently inhibit binding of [ 3 H]PCP to the NCA site of the AChR in the presence of high concentrations of agonist. N,NЈ-dimethylation reduced the affinity 100-fold. A similar inhibition pattern was seen for the d-isomers. Although d-tubocurarine has been characterized as a voltage-dependent open channel blocker on the frog neuromuscular junction at M concentrations (Colquhoun et al., 1979), the affinity for the NCA site in Torpedo AChR has been shown to be low (K d ϳ 9 mM; Cohen et al., 1985). For both l-bebeerine and the d-isomer, tubocurine, binding at the NCA site is preferentially to the desensitized conformation as seen by the ability to inhibit binding in the presence and absence of agonist (Table III). The large loss of affinity upon quaternization of these compounds suggests that the ammonium binds in a sterically constrained portion of the NCA site, which is within the ion channel, or disrupts a particular interaction with the tertiary ammonium (such as a hydrogen bond).
Desensitization by Binding to the ␣␥-Site-Occupancy of the ␣␥-site by the tubocurare analogs increased the observed binding of [ 3 H]PCP. This increase in binding is interpreted as higher affinity for [ 3 H]PCP due to desensitization of the AChR. The l-isomers induced significantly less [ 3 H]PCP binding than the d-isomers. This likely reflects the inability of the altered structure to induce the desensitized conformation. d-Tubocurarine and its methylated analogs produced partial desensitization (ϳ60%), whereas the iodo-and bromo-d-tubocurarine analogs desensitized the AChR to the same extent as the agonist carbamylcholine. The K app for increased [ 3 H]PCP binding (Figs. 7 and 8) is correlated with binding to the high affinity ␣␥-site. There was no evidence that binding to the ␣␦-site  (Boyd and Cohen, 1984) and with the effects of d-tubocurarine on the binding affinity of [ 3 H]histrionicotoxin .
Desensitization by ACh appears to require binding to only a single site, as suggested by a Hill coefficient of 1 for the slow transition to the state that has high affinity for ACh (Neubig et al., 1982). Desensitization by proadifen of mouse muscle type AChRs when expressed in limited subunit combinations (␣ 2 ␤␥ 2 and ␣ 2 ␤␦ 2 ) suggests that desensitization more strongly affects agonist affinity at the ␣␥-site (Sine and Claudio, 1991). That result and the data presented here suggest that the binding energy that drives desensitization is derived predominantly, but not exlusively, from binding to the ␣␥-site. Nonetheless, the dramatic difference in the K app for channel opening (near 100 M; Neubig et al., 1982), which reflects binding to the resting conformation, and the equilibrium affinity (ϳ 20 nM) dictates significant conformational changes at both binding sites upon desensitization. The complete absence of effect on [ 3 H]PCP affinity by the d-tubocurare analogs upon binding the ␣␦-site is unlikely to be due only to a lesser contribution of this site to desensitization but probably also reflects the weaker preference of d-tubocurarine for the desensitized conformation, compared with the strong preference of agonists.
By determing features of the structure that affect site-selectivity, clues to the orientation of d-tubocurarine in the binding site relative to the surrounding ␣␥ and ␣␦ subunits are obtained. This initial set of data begins a description of the structural basis of binding affinity of tubocurare compounds. The series of analogs described should serve as a starting point for a point-to-point mapping of contact sites between amino acids on the AChR and specific functional groups on d-tubocurarine. This can be accomplished by studying changes in affinity of the analogs upon mutation of specific binding site residues. Such studies should lead to full description of the ACh binding sites and be able to elucidate the relative location of amino acids within the site.