Site-specific Charge Interactions of α-Conotoxin MI with the Nicotinic Acetylcholine Receptor*

We have tested the importance of charge interactions for α-conotoxin MI binding to the nicotinic acetylcholine receptor (AChR). Ionic residues on α-conotoxin MI were altered by site-directed mutagenesis or by chemical modification. In physiological buffer, removal of charges at the N terminus, His-5, and Lys-10 had small (2–4-fold) effects on binding affinity to the mouse muscle AChR and the Torpedo AChR. It was also demonstrated that conotoxin had no effect on the conformational equilibrium of either receptor, as assessed by the effects of the noncompetitive antagonist proadifen on conotoxin binding and, conversely, the effect of conotoxin on the affinity of phencyclidine, proadifen, and ethidium. Conotoxin displayed higher binding affinity in low ionic strength buffer; neutralization of Lys-10 and the N terminus by acetylation blocked this affinity shift at the αδ site but not at the αγ site. It is concluded that Ctx residues Lys-10 and the N terminal interact with oppositely charged receptor residues only at the αδ site, and the two sites have distinct arrangements of charged residues. Ethidium fluorescence experiments demonstrated that conotoxin is formally competitive with a small cholinergic ligand, tetramethylammonium. Thus, α-conotoxin MI appears to interact with the portion of the binding site responsible for stabilizing agonist cations but does not do so with a cationic residue and is, consequently, incapable of inducing a conformational change.

The muscle type nicotinic acetylcholine receptor (AChR) 1 is a pentameric ligand-gated cation channel with a subunit stoichiometry of ␣ 2 ␤␥␦ (see Ref. 1 for review). The homologous subunits surround a central pore that forms the ion conductive pathway. The AChR comprises two extracellular ligandbinding sites for ACh that lie at the interfaces between the ␣-␥and the ␣-␦-subunits (2,3). Competitive antagonists and natural toxins often have different affinities for the two sites that arise from the distinct contributions of the ␥and ␦-subunits to each site (4). For example, d-tubocurarine binds the ␣␥ sites of the Torpedo AChR with an affinity more than 100-fold greater than for the ␣␦ site (5).
A similar characteristic was observed for the marine snail ␣-conotoxins, some of which bind the mouse muscle AChR sites with affinities that differ by more than 10,000-fold, where the ␣␦ site is bound with higher affinity (see Ref. 6 for review). The differences in affinity have been attributed to three specific residues in mouse muscle AChR ␥-subunit (␥Lys-34, ␥Tyr-111, and ␥Phe-172) and the corresponding homologous residues in the ␦-subunit (7). Recent work (8) has also identified ␣-subunit residues that affect binding of ␣-conotoxin MI (Ctx) in a siteselective manner, which suggests that Ctx may bind with distinct configurations at the two sites. Ctx residues Gly-9 and Pro-6 (see Table I) and residues on the ␦-subunit and between Ala-7 and Pro-6 and binding site residues in the ␣-subunit appear to stabilize binding through hydrophobic interactions (9).
Significant affinity differences have also been observed for binding to the Torpedo AChR (10 -12); however, in this case the ␣␥ site is bound with higher affinity. Chiara et al. (12) showed that one AChR residue, ␥Tyr-111, and its ␦-subunit homolog, ␦Arg-113, accounted for much of the affinity differences in Torpedo AChR. They further suggested that the affinity change arose from specific charge repulsion between ␦Arg-113 and Ctx residue Lys-10. This Ctx residue is highly conserved among ␣-conotoxins; a cationic amino acid is present in this position (see Table I), with the exception of ␣-conotoxin SI, where it is in the adjacent position, or in SII where it is absent ( Table I). The homologous residue in ␣-conotoxin GI (Arg-9) was identified as being responsible for the high affinity binding of conotoxins to the ␣␥ site; when it was substituted for a proline, affinity decreased dramatically for the ␣␥ site (13). This observation was consistent with the low affinities of ␣-conotoxins SI and SII for the ␣␥ site, which have a Pro at this position.
Several studies suggest that ionic interactions may not be the critical aspect of the role of this residue in binding. Analysis of pairwise interactions in the mouse muscle AChR failed to reveal strong interactions between ␦Tyr-113 and Ctx Lys-10 (9). However, that analysis did not examine a charge-change at ␦Tyr-113 to Arg, the amino acid present in the Torpedo AChR. Studies on analogs of ␣-conotoxin GI that changed the corresponding Arg-9 to alanine displayed affinity changes that appeared site-independent (14), suggesting that the charge per se was not important to binding.
Conotoxin binding appears to block competitively the bind-ing of ␣-bungarotoxin and to block biological activity of AChRs. Yet, ␣-conotoxins do not appear to influence the conformation of the mouse AChR (15). The structural model of Unwin and co-workers (16,17) suggests the presence of transverse tunnels leading to a buried site for ACh binding. But such tunnels appear unlikely to permit access of a ligand the size of Ctx to the inner recesses of the binding site. An alternative model is that Ctx binds at the ACh-binding site entrance to block access of smaller ligands but may not interact closely with residues critical for stabilizing agonist binding. However, Bren and Sine (9) measured interactions of Ctx with ␣-subunit residues Tyr-93, , residues thought to be critical in stabilizing agonist binding. The evidence that Ctx interacts with AChR-binding site residues appears incompatible with the structural model of the AChR and with a role for cationic residues in binding.
To examine the role of charged residues in stabilizing binding of Ctx to the AChR and to examine whether Ctx interacts intimately with residues critical for agonist binding, we determined the binding properties of charge variants of Ctx, as well as several other Ctx mutants. We present data that the charge of Ctx Lys-10 and that of His-5 and the N terminus are relatively unimportant to the net binding affinity toward either the mouse muscle type AChR or the Torpedo electric organ AChR. We further demonstrate that binding of Ctx is conformationally insensitive. Nonetheless, the toxin is formally competitive with the ammonium moiety of ACh. This cationic toxin does not appear to take advantage of the innate charge stabilization the receptor confers on agonists and many competitive antagonists but nonetheless occupies the same steric space. Studies in low ionic strength, however, did reveal distinct contributions of charge interactions from the ␥and ␦-subunits, but these likely arise from interactions with negatively charged residues on the AChR.

EXPERIMENTAL PROCEDURES
Materials-All the enzymes for molecular biology, the pMALp2 plasmid vector, and amylose resin were supplied by New England Biolabs (Beverly, MA). Oligonucleotides were synthesized by Genosys (The Woodlands, TX). Culture media, competent cells, and isopropylthio-␤galactoside were from Life Technologies, Inc. DNA was sequenced using a kit, version 2.0, from U. S. Biochemical Corp. Peptidylglycine ␣-amidating enzyme was from Takara (Beverley, CA), and synthetic Ctx was purchased from Sigma. AChR-rich membranes were prepared from fresh or frozen Torpedo californica electric organ as described previously (18).
Synthesis and Modification of Conotoxins-␣-Conotoxin MI (Ctx) was prepared by peptide synthesis at the Baylor College of Medicine protein core facility. The reduced peptide was purified by preparative reversed phase HPLC (Vydac 250 ϫ 22 mm, C18); elution was by a linear gradient of CH 3 CN containing 0.09% trifluoroacetic acid. The peptide was renatured by oxidation of cysteines to form disulfides by stirring in air according to the procedures described by Zafaralla et al. (19) and by Myers (20). The renatured peptide was purified by reversed phase HPLC; it displayed the skewed peak profile characteristic of the conformational transition between two states (21). The elution profile was identical to that of a commercial preparation of Ctx (Sigma) and gave identical binding constants as assessed by inhibition of [ 3 H]ACh binding on AChR-rich membranes from Torpedo (22).
Two variants of Ctx were also prepared by peptide synthesis, Ctx Ala-7 3 Ser and Ctx Tyr-12 3 Trp. These were purified as the linear peptide. Oxidation of the sulfhydryls to form the correct disulfide bonding was carried out by oxidation in 1 mM GSSG. This redox buffer required only 30 min to 1 h for complete oxidation of the peptide, rather than the several days required for air oxidation. The peptides were then purified by preparative reversed phase HPLC as described above. These variants also displayed the skewed peak characteristic of the conformational transition observed for native Ctx. In all cases, the composition was confirmed by amino acid analysis. Chemical modifications of Ctx at the N terminus and Lys-10 were carried out as follows. Reductive methylation of Ctx was carried out by the method of Jentoft and Dearborn (23). Briefly, 2 mg of Ctx was reacted in 3 ml of 50 mM sodium phosphate, pH 7.0, 8.3 mM NaC-NBH 4 , and 8.3 mM H 2 CO overnight at 4°C. The reaction was terminated by addition of trifluoroacetic acid, and the product was purified by preparative reversed phase HPLC. The reaction resulted in addition of two methyl groups each at the N terminus and at Lys-10 (␣-N-dimethylglycyl, ⑀-N-dimethyllysyl-Ctx; Met-Ctx). Acetylation was carried out by reaction with acetic anhydride according to Means and Feeney (24). 2 mg of Ctx was dissolved in cold, 50% saturated sodium acetate (200 l) and kept on ice during the reaction. A total of 10 l of acetic anhydride was added in small aliquots at regular intervals over 1 h (acetyl-Ctx). The product was purified by reversed phase HPLC. For trinitrobenzoylation (TNB), 2 mg of Ctx was dissolved in 1 ml of sodium borate buffer, pH 9.5, with 20 mM trinitrobenzoyl sulfonate and incubated for 2 h at ambient temperature (ϳ21°C). The product was purified by reversed phase HPLC. The reaction yielded two peaks. Amino acid analysis and mass spectroscopy revealed that the larger second peak contained the product with reaction of TNB at both the N terminus and Lys-10.
Amino acid analysis of the products of chemical reactions revealed loss of detectable Lys and one of the two Gly residues after methylation and after reaction with trinitrobenzoyl sulfonate. The analysis of the acetylated Ctx was similar to that of unmodified toxin, as expected for an acid-labile adduct. Mass spectroscopy revealed parent ion masses of of 1493, 1549, 1577, and 1915 for Ctx, Met-Ctx, Ac-Ctx, and TNB-Ctx respectively; these masses corresponded to those expected for each modification. The loss of primary amines in the modified toxins was verified by loss of reactivity to fluorescamine. Fluorescence measurements showed that there were no amines available for reaction for each of the modified toxins, whereas Ctx showed reactivity that corresponded to the presence of two reactive sites.
Design and Construction of Synthetic Gene for Conotoxin M1-We have constructed a recombinant expression plasmid encoding a maltose-binding protein-Ctx fusion protein, using pMAL-p2 system (25). The system was chosen to direct secretion of the MBP fusion protein into the periplasmic space for proper disulfide bond formation. Codons for Ctx residues were chosen according to Escherichia coli codon usage (26). Glycine was appended to the Ctx sequence to allow cleavage by peptidylglycine ␣-amidation enzyme. Two stop codons were added at the end of the gene followed by a PstI site for cloning the synthetic gene into pMALp2 at the XmnI and PstI site. The XmnI site included sequence encoding the protease factor Xa cleavage site, located just 5Ј to the toxin gene sequence (25,27) such that cleavage of the fusion protein with factor Xa liberates Ctx with no additional amino acids. Complementary oligonucleotides were purified by gel electrophoresis in 20% polyacrylamide containing 7 M urea. They were then 5Ј-phosphorylated using polynucleotide kinase and annealed by combining equimolar amounts of each oligomer (300 pmol) in 100 l of 50 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , and 50 mM NaCl, followed by heating to 95°C for 5 min and slowly cooling to room temperature. The duplex DNA was gel-purified, digested with PstI, and inserted into the expression plasmid pMALp2 at XmnI and PstI sites. Transformants containing the conotoxin M1 gene (pMAL-Cono) were verified by DNA sequencing (28).
Expression and Purification of Recombinant Ctx-Cells containing pMAL-Cono clone were grown at 37°C in LB broth containing 0.2% glucose and 100 g/ml ampicillin to an absorbance of 0.5 at 600 nm, and then induced with 0.5 mM isopropylthio-␤-galactoside for 3 h. A periplasmic fraction was prepared by osmotic shock (29), and the fusion protein was purified by affinity chromatography over an amylose resin (30) column (10 ml) that had been equilibrated previously with 10 mM Tris-HCl, pH 7.4, 200 mM NaCl, 1 mM EDTA. The column was washed with 10 volumes of this buffer, and the bound MBP-conotoxin fusion protein was then eluted with buffer containing 5 mM maltose. Fractions were analyzed by SDS-PAGE.
The purified fusion protein (MBP-conotoxin) was cleaved with factor Xa at a w/w ratio of 1% for 48 h at room temperature followed by  The recombinant Ctx was then purified by HPLC on a semi-preparative Vydac C8 column. Elution was by a linear gradient of 0 -60% acetonitrile containing 0.09% trifluoroacetic acid at a flow rate of 3 ml/min. HPLC fractions were analyzed by Tricine/SDS-PAGE (31). To amidate Ctx, the pooled fractions were treated with peptidylglycine ␣-amidating enzyme according to a standard procedure (32) for 4 h at 37°C followed by repurification on HPLC as described above. Ctx mutants Lys-10 3 His, Lys-10 3 Pro, and His-5 3 Glu were produced as described above for Ctx with appropriate codon replacements in the synthetic oligonucleotides. The yields of the Ctx analogs were lower than that of native Ctx because the extent of factor Xa digestion was lower. Because of the limited amount of material, duplicate experiments on Torpedo AChR were carried out just once.
Binding Assays-Binding of the various conotoxin derivatives was measured by competitive inhibition of the initial rate of bungarotoxin binding, as described previously (33). Binding to Torpedo AChR was measured using the assay described by Schmidt and Raftery (34) or by filtration over GF/F filters. For filtration the following procedure was followed. Samples of AChR-rich membranes and competitive ligand were preincubated in HTPS (250 mM NaCl, 5 mM KCl, 3 mM CaCl 2 , 2 mM MgCl 2 , 0.002% NaN 3 , 20 mM Hepes, pH 7.0) containing 0.1% BSA for 30 min to reach equilibrium. The binding was initiated by addition of 125 I-␣-BgTx (0.5-3 nM), and the samples were allowed to incubate at ambient temperature for 45 min. Reactions were quenched by addition of 300 nM non-radioactive ␣-BgTx in HTPS. The samples were then filtered over Whatman type GF/F filters that had been presoaked in 1% polyethyleneimine. The filters were then washed with 10 ml of HTPS and counted in an ␥-radiation counter (Beckman Instruments). For low ionic strength experiments, binding was carried out in 10 mM Hepes, pH 7.0, with 0.1% BSA. Binding reactions were quenched at 30 s to remain in the linear range of the initial association rate.
Binding to mouse muscle type AChR expressed on the surface of tissue culture cells (BC3H1) was carried out as described earlier (35) using a 24-well plate assay or, alternatively, by a filtration assay (4). The latter assay was carried out by suspending the cells in potassium-Ringer buffer (140 mM KCl, 5.4 mM NaCl, 12.8 mM CaCl 2 , 1.7 mM MgCl 2 , 25 mM Hepes, 30 g/ml BSA, pH 7.4). Assays were incubated as described above and then filtered over Whatman type GF/C filters that had been presoaked overnight in 4% Carnation instant nonfat dry milk. The filters were then washed with 12 ml of phosphatebuffered saline and counted in an ␥-radiation counter (Beckman Instruments).
Assessment of competitive binding between TMA and Ctx was carried out by monitoring their effects on the conformational state of the AChR. The conformation, in turn, was measured by the binding of ethidium. Ethidium preferentially binds the desensitized state of the AChR, and the extent of binding reflects the extent of desensitization of the AChR when the AChR and ethidium concentrations are low relative to the K D values for ethidium (ϳ500 nM). Ethidium binding was measured by fluorescence enhancement, essentially as described by Lurtz and Pedersen (36), except that fluorescence was monitored through a filter (550 nm cut on, Oriel number 59502) to obtain a higher signal to noise ratio. Binding to the NCA site was also measured by [ 3 H]phencyclidine binding using a centrifugation assay as described previously (22). Binding was carried out in low concentrations of [ 3 H]phencyclidine (ϳ1 nM) and is reported as bound over free, which compensates for the change in free [ 3 H]phencyclidine upon binding of competing ligand.
Binding was analyzed by nonlinear least squares fitting to Equation 1 for inhibition by competitive binding to two independent sites present in equal concentrations, where K I1 and K I2 represent the binding constants for the inhibitor, B max the maximum amount of binding at each site, and I the concentration of inhibitor. In some cases, binding was fit to Equation 2 for the binding of two sites present in differing concentrations.
This was necessary because 125 I-␣-BgTx binds with unequal rates to the two sites under some circumstances. Data for inhibition of Ctx binding to the mouse muscle AChR were generally fit with Equation 2, whereas data for inhibition of binding to Torpedo AChR could be fit with Equa-tion 1. For the analysis of ethidium fluorescence, the enhancement of binding by agonists was fit to the Hill Equation 3, or to the binding isotherm (the Hill equation with n ϭ 1). Inhibition of TMA binding by Ctx was fit to Equation 4 for inhibition at a single binding site.

RESULTS
Like other ␣-toxins that interact with the nicotinic AChR, ␣-conotoxins are cationic. Ctx carries a net charge of ϩ3.5, counting His-5 as 1 ⁄2 charge and including the N-terminal charge. To test whether these charged moieties play a significant role in the interaction of ␣-conotoxins with the AChR, we examined the binding energies of toxins that had been modified or mutated at the charged loci. One receptor residue that had been identified as constituting an electrostatically repulsive interaction was ␦Arg-113 in the Torpedo AChR. The homologous residue in the mouse muscle AChR is a Tyr. Therefore, we compared affinities of toxins in both these species as a means of assessing the importance of charge repulsion between this residue and Ctx. We utilized two approaches to generate Ctx structural analogs, expression of recombinant Ctx in E. coli, and chemical modification of synthetic Ctx.
Expression and Purification of Recombinant Ctx in E. coli-Ctx could be expressed and refolded in E. coli by attaching a synthetic gene to the secreted maltose-binding protein. The fusion protein was isolated by amylose affinity chromatography and Ctx cleaved by factor Xa. Purification by HPLC, amidation, and a final HPLC purification yielded a product that was indistinguishable from commercial Ctx and from synthetic Ctx, as judged by migration SDS-polyacrylamide gel electrophoresis and by HPLC (Fig. 1, insets). The net yield of recombinant toxin was about 50 g/liter of cell. Recombinant Ctx was compared with synthetic Ctx and with a commercial product (Sigma) for binding activity on mouse muscle AChR and on T. californica AChR by inhibition of 125 I-␣-BgTx binding as described under "Experimental Procedures." The affinities of each material were indistinguishable (data not shown). Proper refolding and disulfide bond formation was apparent from comigration of recombinant and synthetic Ctx on HPLC and because it is required for biological activity (37,38).
Effects of Ctx Modifications on Binding Affinity-Ctx was chemically modified by reductive methylation, by acetylation, or by trinitrobenzoylation, using standard reaction methods (see "Experimental Procedures"). Each modified toxin was purified by reversed phase HPLC and shown to be quantitatively modified at Lys-10 and at the N terminus by mass spectroscopy, by lack of reactivity to fluorescamine, and by amino acid analysis. Methylation will increase the bulk of the reacted amines but will not alter the charge; both the N-terminal amine and Lys-10 were found to be dimethylated. The effect of amine dimethylation on affinity was negligible ( Fig. 2 and Table II), demonstrating that the simple size of these amines was relatively unimportant for binding. Acetylation of the amines resulted in formation of a neutral acetamide with more steric bulk than the methylation. Nonetheless, there was less than a 2-fold effect on affinity for the mouse AChR and a 3-fold decrease for the Torpedo AChR ␣␥ site ( Fig. 2 and Table II). The small affinity changes suggest that the charges of Lys-10 and the N terminus do not contribute substantially to the binding energy. The modification by addition of a picric acid moiety, (TNB-Ctx) does cause a significant shift of binding to the ␣␥ site of Torpedo AChR, and the ␣␦ site of the mouse AChR of 13-15-fold. The effects on mouse ␣␥ and Torpedo ␣␦ sites were smaller (Table II).
These results indicated that the Lys-10 side chain charge was relatively unimportant for binding, although added steric bulk could affect binding at some sites. We further tested the ability of histidine, an aromatic cationic residue, and proline to affect the overall affinity at this position. Proline occurs as a natural substitution in ␣-conotoxin SI and SII at this position; both these toxins have substantially lower affinity for the AChR (13).
␣-Conotoxin Lys-10 3 Pro and Lys-10 3 His were made by expression in E. coli as MBP fusions. Substitution of His for Lys had a 5-fold effect on the affinity for the Torpedo AChR, and decreased the affinity for the mouse ␣␦ site more than 20-fold ( Fig. 3 and Table II). Pro-substitution decreased affinity on the Torpedo AChR 7-23-fold, whereas the change in affinity for the mouse ␣␦ site was more than 7000-fold. These results indicate that the mouse AChR is substantially more sensitive FIG. 2. Chemical modification of Ctx lysine 10 affects affinity weakly. Met-Ctx (Ⅺ), acetyl-Ctx (‚), and TNB-Ctx (ƒ) were compared with Ctx (E) for binding to the mouse AChR (A) or the Torpedo AChR (B). Binding was measured by inhibition of the initial rate of 125 I-␣-BgTx binding using the 24-well plate assay (A) and to the Torpedo AChR using the DE81 filter binding assay (B) as described under "Experimental Procedures." Each symbol represents the average of duplicate determinations that generally varied less than 5%. The data are normalized to the initial rate of binding observed in the absence of competitor. The solid lines represent the best fits to a model for inhibition at two binding sites (Equations 1 and 2). to changes in this position than the Torpedo AChR. The results with His indicate that change in side chain structure at this position can affect affinity substantially. The effects of Pro substitution must be interpreted more cautiously, as it will also constrain the backbone conformation, but the results do confirm that a Pro in this position likely contributes to the weaker affinity of conotoxins SI and SII. We examined the role of His-5 as a possible contributor to electrostatic stabilization. This residue was mutated to Glu, providing 2 units charge change.
Remarkably, Ctx His-5 3 Glu showed only a 3-6-fold change in affinity, indicating a modest effect of electrostatic attraction ( Fig. 4 and Table II).
We examined the importance of two additional conserved side chains by making the toxins using peptide synthesis. Tyr-12 of Ctx represents a residue conserved as either a Tyr or Phe residue among all ␣-conotoxins (Table I). We synthesized the analog Ctx Tyr-12 3 Trp and examined its binding properties ( Fig. 5 and Table II). The binding for the mouse AChR was unchanged from that of conotoxin MI but had higher affinity for the Torpedo ␣␥ site. Thus, this position is not acutely sensitive to a significant change in the size of the side chain. Alanine 7 of Ctx MI is highly conserved among the ␣-conotoxins. A change to serine affected affinity to both mouse and Torpedo AChR but affected Torpedo more strongly. This observation confirms the importance of this residue but also shows that the addition of a hydroxyl group can be tolerated without dramatic loss of affinity.
Conformational Changes in the AChR Do Not Affect Ctx Binding-The cation associated with agonists and many competitive antagonists affects the conformational equilibrium of the AChR, generally to increase the proportion of desensitized AChR upon binding. Even the smaller agonists such as TMA can desensitize the AChR. If a cationic moiety of Ctx bound the ACh-binding site in the same manner as the ammonium of ACh, it would be expected to affect the conformation as well.
We tested the ability of Ctx to affect the conformation by measuring the effect of noncompetitive antagonists on the affinity of conotoxin and, conversely, the effect of conotoxin on the affinities of proadifen and phencyclidine. As shown in Fig. 6, A and B, proadifen does not appreciably change the affinity of Ctx for either the mouse or Torpedo AChR (Table II), nor does tetracaine affect binding, an NCA that stabilizes the resting state of the Torpedo AChR (Table II). The presence of Ctx in nearly saturating conditions does not affect PCP binding to the Torpedo AChR and decreases the affinity of proadifen binding less than 2-fold, as measured by inhibition of [ 3 H]PCP binding (Fig. 6C). As a control, the agonist carbamylcholine increased the affinity of proadifen more than 10-fold (Fig. 6C). Ctx also does not affect the apparent affinity of ethidium, another strongly desensitizing noncompetitive antagonist (see below).
Electrostatic Dependence of Ctx Binding-The charge changes in Ctx affected by mutation or modification did not reveal any discrete charge-charge interactions critical for stabilizing binding; however, it was possible that such interactions might be revealed at low ionic strength conditions that would enhance electrostatic attraction or repulsion. We examined the ability of ionic strength to influence the affinity of Ctx and acetyl-Ctx for Torpedo AChR by inhibition of 125 I-␣-BgTx binding in 10 mM Hepes. As can be seen in Fig. 7, lowered ionic strength dramatically increased the affinity of Ctx (compare dotted lines in Fig. 7A). The change was more pronounced at the lower affinity ␣␦ site, suggesting that there are stronger charge interactions at this site. In low ionic strength buffer, inhibition by Ctx was characterized by a single site inhibition curve rather than a two-site inhibition curve as observed in physiological ionic strength. To ensure that this reflected similar binding of Ctx to each of the two sites, rather than an artifact of 125 I-␣-BgTx binding kinetics, we examined the abil-

FIG. 4. Ctx His-5 mutation alters affinity similarly at all sites.
Ctx His-5 3 Glu (‚) was produced by expression in E. coli, and its affinities for the mouse AChR (A) and the Torpedo AChR (B) were compared with Ctx (E). Binding was performed as described in Fig. 2. Each symbol represents the average of two determinations, and the data are normalized to the initial rate of binding observed in the absence of competitor. The solid lines represent the best fits to a model for inhibition at two binding sites.

FIG. 5. Modification of Ctx positions 7 and 12 affect binding to the AChR.
Ctx A7S (Ⅺ) and Ctx Y12W (‚) were produced by peptide synthesis, and their affinities as well as that for Ctx (E) for the mouse AChR (A) and Torpedo AChR (B) were determined by the glass fiber filtration assays described under "Experimental Procedures." Each symbol represents the average of three determinations, and the data are normalized to the initial rate of binding observed in the absence of competitor; standard deviations for each point were generally less than 5%. The solid lines represent the best fits to a model for inhibition at two binding sites.
ity of d-tubocurarine to inhibit the rate of 125 I-␣-BgTx binding under the same conditions (Fig. 7B). The data clearly reveal two-site inhibition, with affinities for d-tubocurarine similar to those observed at physiological ionic strength. Therefore, Ctx binding has similar affinities to the ␣␥ site and ␣␦ site in low ionic strength.
We also determined the affinity of acetyl-Ctx (Fig. 7A) to test whether neutralizing the N terminus and Lys-10 was important to the charge interactions observed in low ionic strength buffer. Acetyl-Ctx displayed clear two-site binding in low ionic strength, whereas binding was well fit by single site function in physiological buffer. Therefore, we could not, a priori, assign sites to the high and low affinity components. To determine unambiguously which site corresponded to the high affinity component, we carried out inhibition by acetyl-Ctx with the ␣␥ site blocked by including 1 M d-tubocurarine (Fig. 7B, ⌬). The inhibition was well fit by a single site binding function with a K value that corresponded to that of the low affinity component (Table III). This demonstrated that the high affinity component reflected acetyl-Ctx binding to the ␣␥ site and that the lower affinity reflected binding to the ␣␦ site.
In low versus high ionic strength, the shift in affinity of acetyl-Ctx was similar to the shift observed for Ctx itself at the high affinity, ␣␥ site, ϳ10 -30-fold. Acetylation, therefore, does not affect ionic interactions at the ␣␥ site. The ionic strength-induced change in affinity must arise from charges other than the N terminus or Lys-10. The ionic strength shift of acetyl-Ctx affinity at the low affinity ␣␦ site was much smaller (1-3-fold), particularly when compared with the shift of Ctx itself (ϳ300-fold). Thus, the ionic strength change affects interactions between negatively charged residues at the ␣␦ site and Ctx residues Lys-10 or the N terminus or both. The attractive interactions at the ␣␥ site must be mediated by other residues since the shift by ionic strength is unaffected by acetylation.
Ctx Competes with a Small Agonist-The data strongly suggested that charged moieties on Ctx do not participate in binding and that there was not a functional group that acted as the equivalent of the ammonium group present in most agonists and antagonists. In addition, Ctx did not affect the conformational state of the AChR as would be expected if it occupied the same space as that occupied by an agonist ammonium group. Nonetheless, Ctx appears to be competitive with other toxins and with ACh; we have shown that Ctx blocks [ 3 H]ACh binding and that it blocks [ 3 H]d-tubocurarine binding (data not shown).
However, it was possible that Ctx sterically blocked binding by interaction with the sites that stabilize moieties outside of the ammonium group, such as the acetate of ACh or the larger bulk of other toxins. To test whether Ctx directly interacts with the AChR in the most critical part of the ACh-binding sites, the portion that stabilizes ammonium binding, we tested whether Ctx was directly competitive with the small agonist TMA.
To assess whether Ctx and TMA were mutually competitive, we utilized the observation that Ctx does not affect the conformation of the AChR, whereas TMA desensitizes the AChR. To assess the conformational state, we utilized the fluorescent enhancement of ethidium upon binding the AChR. Ethidium preferentially binds the desensitized conformation (22,39). Conditions of relatively low ethidium and low AChR concentrations were chosen such that the fluorescence increase due to ethidium binding would reflect the increase in affinity due to desensitization. Fig. 8A shows the increases in fluorescence caused by titration of ethidium/AChR suspensions with carbamylcholine and with TMA. Both ligands caused concentrationdependent changes in fluorescence that correspond to the binding of ligand and concomitant desensitization of the AChR. Carbamylcholine increased fluorescence with a Hill coefficient of 1.5 and a K obs of 170 nM, which corresponds closely to the values for equilibrium binding (data not shown). The K obs for TMA was 3 M with a Hill coefficient of 1.
To determine whether conotoxin competes with TMA binding, Ctx was titrated into suspensions of ethidium/AChR that contained varying concentrations of TMA. Example titrations with 0, 10, 20, and 100 M TMA are shown in Fig. 8B. The curve with no TMA demonstrates that Ctx by itself has only a small effect on ethidium fluorescence. As titration with Ctx displaced TMA from the ACh-binding sites, the AChR displayed less ethidium binding and fluorescence as the desensitized conformation was not stabilized by Ctx. In each case the inhibition was well fit by a single site inhibition function with a background fluorescence approaching that found in the absence of TMA. The K obs was determined for each titration and replotted against the TMA concentration (Fig. 8C). There is a linear relationship between the TMA concentration and the K obs , indicative of direct competition.
The results cannot be accounted for by inhibition through a ternary complex of AChR, TMA, and Ctx unless one ligand conformationally reduces the binding of the second ligand with a coupling energy of more than 3.4 kcal/mol. This value was determined from the maximum TMA concentration used (1 mM) and the K obs for TMA binding. The experiment instead suggests direct, steric competition between TMA and Ctx, where Ctx occupies the space normally taken by the agonist ammonium moiety. DISCUSSION We examined various charged Ctx residues to determine whether they were critical for high affinity binding to the AChR, in a manner similar to the cationic moiety of agonists. Elimination of charge at Lys-10 and the N terminus affected binding affinity 2-4-fold. A charge reversal at His-5 consistently changed affinity 6-fold at all sites. Such modest changes show that individual charge-charge interactions are not a highly significant component of stabilizing Ctx binding in physiological buffer, despite the generally conserved cationic nature of these toxins. Binding affinity to Torpedo AChR was, however, significantly modified upon neutralization of Lys-10 when examined at low ionic strength, and the effect was much stron- FIG. 6. Ctx binding to the AChR is independent of conformation. A, BC3H-1 cells were incubated in the indicated concentrations of Ctx in the absence (E) or presence (Ⅺ) of 26 M proadifen, and the initial rate of binding of 125 I-␣-BgTx was measured using the 24-well assay as described under "Experimental Procedures." The value for the presence of excess BgTx to define background binding is also shown (OE). The solid curves are the best fits to a model for inhibition at two binding sites (Equation 2). The K I1 values for this experiment are 11 and 5 nM in the absence and presence of proadifen, respectively. B, Torpedo AChR-rich membranes (3.4 nM ACh sites) were incubated with the indicated concentrations of Ctx in the presence of 26 M proadifen (Ⅺ) or no added ligand (E). The initial rate of 125 I-␣-BgTx (0.75 nM) binding was measured using the glass fiber filtration method described under "Experimental Procedures." The solid curves represent the best fits to a model for binding two independent sites. C, Ctx does not alter the affinity of proadifen.

TABLE III Charge effects on binding observed in low ionic strength buffer
The K values for binding of conotoxin M1 and acetyl-conotoxin to the AChR in low ionic strength buffer (10 mM Hepes) were determined as described in Fig. 7 and under "Experimental Procedures." Errors are the standard deviation of the number of independent determinations indicated by n.
ger at the ␣␦ site than the ␣␥ site. The results indicate an attractive ionic interaction of Lys-10, the N terminus, or both with the ␣␦ site that is not present at the ␣␥ site. The two sites must have distinct orientations of charged residues or the toxin must bind in distinctly different orientations at each site.
We conclude that Lys-10 or His-5 do not interact with the AChR in the same manner as cationic groups of agonists that are required for channel activation. This is shown by the modest affinity changes upon modification and by the general failure of Ctx to modulate the conformational state of the AChR. Ctx does not appear to interact with other residues that influence the conformational equilibrium between the resting and desensitized states of the AChR. However, Ctx does formally compete with the binding of TMA, suggesting that steric hindrance prevents the simultaneous occupancy of both Ctx and TMA. Consequently, Ctx does interact with the portion of the binding site responsible for stabilizing agonist cations but does not do so with a cationic residue and is therefore incapable of inducing a conformational change.
Our binding data reveal affinities of Ctx for the Torpedo ␣␥ site that differ from those observed by others (10,11). Some of the larger differences can be accounted for by the distinct buffer conditions used as follows: Hann et al. (13) and Groebe et al. (10) used low ionic strength buffers to measure binding in the presence of detergents. However, the data of Chiara et al. (12) was carried out in physiological buffer that revealed affinities 5-10-fold higher than we observed. Our binding data on the mouse muscle AChR is in complete agreement with that of others (7). We also obtained similar binding data for commercial, synthetic, and recombinant Ctx and observed consistent inhibition of [ 3 H]ACh binding and of [ 3 H]d-tubocurarine binding (22) (data not shown). The source of the difference remains unknown.
Ionic Interactions in Binding-It has long been argued that localized charges stabilize the binding of agonists and competitive antagonists. Cholinergic agonists universally carry a positively charged group (40), and for the AChR, a simple quaternary ammonium in the form of TMA is a completely competent agonist. Localization of several anionic residues (␥Asp-180 and ␣Asp-152) in the vicinity of the ACh-binding sites has supported the idea of charge stabilization of ligand binding (41,42), but it remains unclear whether these residues act as countercharges to the ligand ammonium. Experiments that examined the effect of ionic strength on ligand binding were more consistent with a diffuse charge distribution near the binding site than a single countercharge in close proximity (43). ␣-Toxins of snakes and marine snails are generally basic and have binding that is strongly dependent on ionic strength (34,44), suggesting that ionic interactions constitute a significant component of the net binding energy.
Sine et al. (7) identified several residues critical for the site selectivity of Ctx binding to mouse muscle receptor, including residues ␥Lys-34, ␥Tyr-111, and ␥His-172. The homologous residue to ␥Tyr-111 in Torpedo ␦-subunit is ␦Arg-113. Chiara et al. (12) demonstrated that ␦Arg-113 dramatically weakened binding of Ctx to the ␣␦ site, and they inferred an electrostatic repulsion with Lys-10. At the Torpedo AChR ␦-subunit, acetylation of Lys-10 does appear to cause an increase in affinity of 2-4-fold, consistent with lessened repulsion. However, this change is much smaller than the 1000-fold effect observed by Chiara et al. (12). Our data suggest that the ␦Arg-113 interaction is not through electrostatic repulsion at the N terminus, His-5, or Lys-10 of Ctx. We cannot rigorously exclude an interaction through Ctx Arg-2, which we have not modified; however, it appears unlikely that Arg-2 is responsible because conotoxins GI and SIA bind with similar selectivity to the Torpedo AChR (13) and have a Glu and Tyr in the homologous positions, respectively (see Table I). We conclude that Lys-10 of Ctx and ␦Arg-113 are not in close apposition because electrostatic repulsion is not alleviated by charge neutralization. This is in agreement with the conclusions of Groebe et al. (14) who used a Arg-9 3 Ala mutation of ␣-conotoxin GI to analyze binding to the Torpedo and mouse AChRs.
Ionic interactions become more apparent in low ionic strength, where we observed similar affinities of Ctx for both sites on the Torpedo AChR, indicating a larger change in affinity for the ␣␦ site. This shift is consistent with stronger attraction among opposite charges. If charge repulsion among cationic residues had been significant, lower ionic strength should have exacerbated the repulsion, resulting in relatively lower affinity. The observation of higher affinity, rather than lower, at the ␣␦ site, reinforces the conclusion that charge repulsion between Lys-10 and the ␦Arg-113 of the Torpedo AChR is not a factor that causes site selectivity in binding.
In low ionic strength, reducing the charge on Ctx by acetylation of Ctx and the N terminus results in a larger shift on the Torpedo ␣␦-subunit than the ␣␥-subunit. This suggests that attractive forces are lost upon neutralization and that they are more significant for the ␣␦-subunit than the ␣␥-subunit. The results further suggest the presence of distinct charge distributions at the ␣␥ and ␣␦ sites. For the ␣␦ site, binding of Ctx is enhanced 300 -600-fold upon lowering ionic strength, and this is reduced to a 1-3-fold effect for diacetyl-Ctx. This suggests that the primary stabilizing ionic interactions occur from the ␣␦ site to Lys-10 or the N terminus. In contrast for the ␣␥ site, the binding affinity of Ctx is enhanced about 30-fold upon lowering ionic strength, and this is similar for diacetyl-Ctx. Thus, ionic strength effects are similar even when charge is neutralized at Ctx Lys-10 and the N terminus. This suggests that the weaker stabilizing ionic interactions at the ␣␥ site occur through one or more of the remaining ionic residues on Ctx, Arg-2 or His-5. The His-5 3 Glu mutation shows similar effects at both sites of Torpedo AChR, suggesting a role for Arg-2; however, those experiments will need to be conducted in low ionic strength.
Examination of the ␥and ␦-subunit N-terminal domain sequences does not lead to immediate insights into the difference in amino acids that could account for the binding in low ionic strength. Since the ionic changes reflect attractive interactions, the difference in binding should reflect differences in the distribution of anionic residues near the Ctxbinding sites. There are nearly two dozen such possible sites, some near the binding site loops and many not. Unfortunately, it is also unclear whether such differences in charge distribution will be important for binding since charge changes have slight effects on binding in physiological ionic strength.
Conotoxin Binds Independently of Conformation-Our data shows that Ctx binding is affected less than 2-fold by noncompetitive antagonists on both Torpedo and mouse AChR and that Ctx affects the binding of NCAs on the Torpedo AChR less than 2-fold. Ctx, therefore, binds equivalently to the desensitized and resting forms of the AChR. The results are consistent with a similar finding of Prince and Sine (15) that Ctx binding to mouse muscle AChR was insensitive to the presence of proadifen. It can be inferred that Ctx does not interact strongly with residues that mediate conformational changes even though Ctx competes for ACh binding.
An alternative interpretation is that Ctx acts as a cap to block an entrance to the binding site but does not intimately contact the residues that normally stabilize binding. A capping mechanism should have permitted us to observe simultaneous binding of Ctx and a small ligand, such as TMA. We tested this notion explicitly by examining competition of Ctx with TMA. The mutual inhibition was linear over a 300-fold range of TMA concentrations, which represents an interac-tion energy of Ͼ3.4 kcal/mol. Allosteric effects were unlikely to account for this interaction since we had shown Ctx binding to be insensitive to the fundamental conformational change of desensitization. The interaction most likely represents steric repulsion. Such a conclusion suggests that Ctx inserts a moiety into the most critical part of the ACh-binding site and does not merely cap access to the site. The moiety, however, is clearly not one of the Ctx cationic residues because they affect binding only weakly.
Our results show that charge-charge interactions between Ctx and the ACh-binding sites are weak, at best, under physiological conditions but reveal distinct arrangements of charged residues at each site or, as suggested by Sugiyama et al. (8), that Ctx binds in distinct orientations at the two sites. They furthermore show intimate binding of Ctx with the AChbinding site that is independent of cationic residues on Ctx and that do not affect the conformation of the AChR. The lack of effect on conformation may be a consequence of not placing a cation in the critical part of the binding site but seems remarkable nonetheless.