Identification of Pairwise Interactions in the α-Neurotoxin-Nicotinic Acetylcholine Receptor Complex through Double Mutant Cycles*

α-Neurotoxins are potent inhibitors of the nicotinic acetylcholine receptor (nAChR), binding with high affinity to the two agonist sites located on the extracellular domain. Previous site-directed mutagenesis had identified three residues on the α-neurotoxin from Naja mossambica mossambica(Lys27, Arg33, and Lys47) and four residues on the mouse muscle nAChR α-subunit (Val188, Tyr190, Pro197, and Asp200) as contributing to binding. In this study, thermodynamic mutant cycle analysis was applied to these sets of residues to identify specific pairwise interactions. Amino acid variants of α-neurotoxin fromN. mossambica mossambica at position 33 and of the nAChR at position 188 showed strong energetic couplings of 2–3 kcal/mol at both binding sites. Consistently smaller yet significant linkages of 1.6–2.1 kcal/mol were also observed between variants at position 27 on the toxin and position 188 on the receptor. Additionally, toxin residue 27 coupled to the receptor residues 190, 197, and 200 at the αδ binding site with observed coupling energies of 1.5–1.9 kcal/mol. No linkages were found between toxin residue Lys47 and the receptor residues studied here. These results provide direct evidence that the two conserved cationic residues Arg33 and Lys27, located on loop II of the toxin structure, are binding in close proximity to the α-subunit region between residues 188–200. The toxin residue Arg33 is closer to Val188, where it is likely stabilized by adjacent negative or aromatic residues on the receptor structure. Lys27 is positioned closer to Tyr190, Pro197, and Asp200, where it is likely stabilized through electrostatic interaction with Asp200 and/or cation/π interactions with Tyr190.

The critical tool utilized in the initial identification of the receptor and in subsequent structural analyses is the family of three-fingered snake ␣-neurotoxins (for review, see Refs. 28 and 29), which form high affinity complexes with the receptor; for example, the ␣-bungarotoxin-receptor complex has a K d of ϳ10 Ϫ12 . The structure of the ␣-neurotoxins has been solved through nuclear magnetic resonance (30 -32) and x-ray crystallographic studies (33)(34)(35). These polypeptides (ϳ7 kDa) are characterized by three large loops which extend from a rigid globular domain held together by 4 or 5 conserved disulfide bonds. Even though ␣-neurotoxin structures have been solved, little is known about the structure of the toxin-receptor complex, and interacting residues have not been identified.
In previous work, we delineated residues involved in the binding interaction on both the ␣-neurotoxin, Naja mossambica mossambica (NmmI), and the mouse muscle nAChR interfaces (36). Four residues located on the receptor ␣-subunit and three residues located on the toxin structure were found to contribute significantly to high affinity binding. Even though wild-type NmmI displays an equivalent affinity for the two binding sites on nAChR, we showed through mutational analysis that the energetic contribution of selected residues differed at the two sites. Several of the toxin and receptor mutations studied differentially affected binding at the ␣␦ and ␣␥ sites, resulting in two distinct binding affinities (36). In this study, we have utilized double mutant cycles to explore whether any of these residues are involved in pairwise interactions at each binding site. Such information may not only lead to a better understanding of the structure of the toxin-receptor complex, but combined with the known structure of the toxin may eventually establish spatial constraints within the receptor architecture (37)(38)(39).

EXPERIMENTAL PROCEDURES
Materials-125 I-␣-Bungarotoxin (specific activity ϳ16 Ci/g) was obtained from NEN Life Science Products. ␣-Conotoxin M1 was purchased from American Peptide Company.
NmmI Expression and Purification-Recombinant wild-type and mutant NmmI were expressed as fusion proteins in Escherichia coli, and the free toxins were purified as described (36). Because several of the mutant toxins produced rather low affinity complexes with mutant receptors, relatively large amounts of NmmI ␣-neurotoxins were required for this study. Typically, 0.5-1.5 mg of toxin could be purified to homogeneity from 1 liter of cells.
NmmI Binding Measurements-Binding assays were carried out on assembled pentameric nAChRs expressed on the surface of intact cells. The cells were harvested by gentle agitation in phosphate-buffered saline plus 5 mM EDTA, centrifuged briefly, and resuspended in high potassium Ringer's solution. The cells were divided into aliquots for binding measurements (assay volume 200 l). Specified concentrations of NmmI were added to each tube containing receptor and allowed to bind for 5 h. NmmI dissociation constants were measured by competition against initial rates of 125 I-␣-bungarotoxin binding using 10 -20 nM concentrations (41).
A concentration of ␣-bungarotoxin considerably above its K d (60 pM) was used. The assay relies on the fractional reduction of the initial rate of ␣-bungarotoxin binding in the presence and absence of NmmI. ␣-Bungarotoxin rates are less affected by the receptor mutations, and these reductions will not influence the determined NmmI K d as long as ␣-bungarotoxin dissociation is slow relative to the time frame of the initial rate assay.
Data were analyzed using least squares fits to the Hill equation or to two sites of equal population but different affinities. Nonspecific binding was determined in the presence of 10 mM carbamylcholine, 300 M dimethyl d-tubocurarine, or 4 M cobra ␣-toxin (Naja naja siamensus), depending on the particular receptor mutant. Binding assays conducted in the presence of ␣-conotoxin M1 were carried out in an identical manner. Concentrations of ␣-conotoxin M1 utilized for assays with the receptor mutations were as follows: wild type (300 nM), V188D (300 nM), V188K (300 nM), Y190T (1 M), Y190F (300 nM), P197I (1 M), and D200Q (1 M) and were based on previously determined K d values for ␣-conotoxin M1 (40).
Homology Modeling-Using erabutoxin b as a template (an ␣-neurotoxin of known crystal structure (33) and possessing 60% residue identity with NmmI), segments of NmmI were modeled on the basis of conserved regions and the common disulfide linkages. The final conformation of the ␣-carbon backbone was adjusted to account for the two unique prolines in erabutoxin b and the single proline in NmmI. The modeled structure was then relaxed by unrestrained steepest descent minimization of 5000 iterations with the program Discover (MSI, 1997).

RESULTS
nAChR Binding Site-Of the three segments of linear sequence in the ␣-subunit, the region encompassing residues 180 -200 appears to be the most critical for NmmI recognition. Substitutions at residues 188, 190, 197, and 200 resulted in substantial decreases in toxin binding affinity (36), indicating major roles of these residues in ␣-neurotoxin recognition. Determinants in this region include a conserved aromatic residue (Tyr 190 ), two positions where neuronal and muscle receptors differ (Val 188 and Pro 197 ), and a negatively charged residue (Asp 200 ). The mutations Y190F and Y190T both resulted in a loss of affinity of ϳ2 kcal/mol at the ␣␦ site and almost 4 kcal/mol at the ␣␥ site (Table I), suggesting the importance of the aromatic hydroxyl group. Introducing a positive charge at position 188 (V188K) was also unfavorable and destabilized the toxin-receptor complex by 1.8 kcal/mol (at the ␣␦ site) and 3.5 kcal/mol (at the ␣␥ site), whereas introduction of a negative charge (V188D) at this position had either no effect or far less of an influence (Ϫ0.09 kcal/mol at the ␣␦ site and 1.8 kcal/mol at the ␣␥ site). The selective loss in binding affinity observed with the introduction of a positive charge but not a negative charge suggested destabilization resulted from coulombic repulsion of the highly cationic ␣-neurotoxin. Elimination of the negative charge at position 200 (D200Q) decreased binding by 66-fold (2.5 kcal/mol) selectively at the ␣␥ site. ␣-Neurotoxin Structure and Binding Site-Sequence comparisons between members of the ␣-neurotoxin family have identified ϳ12 residues that are highly conserved, three of which are positively charged (Lys 27 , Arg 33 , Lys 47 ; NmmI numbering). Extensive mutagenesis studies carried out by Menez and co-workers (42-44) using a homologous ␣-neurotoxin with 60% residue identity to NmmI (erabutoxin a) defined the toxin binding surface to encompass ϳ680 Å 2 and about 10 amino acids located on loop II and the tips of loops I and III. Five of these residues appear to play the predominant role in binding (Ser 8 , Gln 10 , Lys 27 , Arg 33 , Lys 47 ) (Fig. 1). The binding site utilized for the NmmI toxin appears to be similar yet not identical to that defined for erabutoxin a (36). Whereas Gln 10 in erabutoxin a and Glu 10 in NmmI affect binding differently, similar to erabutoxin a, the three conserved positive residues are critical for the NmmI-mouse muscle nAChR high affinity interaction.
Homology modeling of NmmI using the solved crystal structure of erabutoxin a as a template is shown in Fig. 1. Locations of each of the residues studied here are indicated. R33E is near the tip of loop II, whereas K27E is closer to the top and opposite side of loop II. K47A is located on an exposed surface of loop III. The mutations K27E and R33E resulted in large changes in binding affinities for the NmmI-nAChR interaction at both the ␣␦ and ␣␥ interfaces (Table I). A shift in binding affinity of over 4 orders of magnitude was observed with R33E at the ␣␥ binding interface and nearly 3 orders at the ␣␦ binding interface (Table I). The mutation K47A did not exhibit site selectiv-ity for ␣␥ and ␣␦ sites.
Double Mutant Cycles-Double mutant cycle analyses were applied to the three ␣-neurotoxin variants (K27E, R33E, and K47A) and six ␣-subunit nAChR variants (V188D/K, Y190T/F, P197I, and D200Q) described in order to delineate potential pairwise interactions. This method is based on simple additivity or non-additivity of mutations. If two residues are interacting, then the sum of the free energy change of the single mutations will usually not equal the free energy change measured with both mutations (45)(46)(47). This is shown by Equation 1: where ⌬⌬G (X) represents the change in free energy caused by a mutation at site X on one interacting species relative to its wild type, ⌬⌬G (Y) represents the change in free energy caused by a mutation at site Y on the other species relative to its wild type, ⌬⌬G (X, Y) represents the change in free energy caused by both mutations when present together, and ⌬⌬G INT (coupling energy) is the measure of the interaction of the two components that are mutated. If the two residues are not linked or interacting, ⌬⌬G INT will equal 0, and if the two residues are interacting then the value of ⌬⌬G INT may be either positive or negative depending on whether the interaction between the mutated residues reduces or enhances affinity (47). ⌬⌬G INT can also be described in terms of the equilibrium constants (39): Mutant Pairs and Site Selectivity-Dissociation constants and changes in free energy for each of the mutant pairs analyzed are shown in Table I. Fig. 2 (A and B), shows typical binding curves obtained with the ␣-subunit mutations Y190T and P197I, respectively, when assayed with each of the toxin mutations; the wild-type toxin/wild-type receptor curve is also shown for comparison. As seen in Table I and Fig. 2, many of the mutant pairs result in large reductions in the overall affinity of the toxin-receptor complex. The dissociation constant for the R33E/V188D pair at the ␣␥ binding site was too large to measure precisely but corresponded to a K d of more than 0.5 mM. The R33E/P197I mutant pair resulted in a loss of affinity of 8.0 kcal/mol (␣␥ site) or a loss in K d of 6 orders of magnitude compared with the wild-type/wild-type interaction. Mutant pairs involving the toxin mutation K27E also resulted in large destabilizations ranging from 3.5 to 7.5 kcal/mol.
As expected by the demonstration of site selectivity conferred by mutations on the ␣-neurotoxin or receptor when analyzed separately, all of the mutant pairs showed two distinct binding affinities presumably arising from the ␣␦ and ␣␥ sites. Analysis of the binding curves yielded Hill coefficients ranging from 0.3 to 0.8, indicating the presence of two classes of binding sites. When these curves were fit to a two-site analysis of equal population, differences in affinity between the two sites ranged from 9-to 85-fold.
To confirm that these binding curves reflected disparate affinities at the ␣␦ and ␣␥ subunit interfaces and to ascertain which site possessed the high affinity binding, assays were carried out in the presence of ␣-conotoxin M1. ␣-Conotoxin M1 has a more than 10,000-fold selectivity for the ␣␦ binding site over the ␣␥ site on the wild-type mouse receptor (K d , ␣␦ ϭ 0.45 nM, K d , ␣␥ ϭ 20 M) (22,27). ␣-Conotoxin M1 site selectivity is maintained with each of the receptor mutations studied here despite some changes in the absolute K d values. 2 Thus, when Loops I, II, and III of the toxin are labeled. This structure was obtained through an energy minimization with the solved structure of erabutoxin a, which shares a 60% amino acid sequence identity with NmmI. Sequence identities of the two 62-amino-acid peptides are shown.
␣-conotoxin M1 is included in the ␣-neurotoxin-receptor binding assays, it protects the ␣␦ site nearly completely, with the residual ␣-neurotoxin binding observed only at the ␣␥ site. Fig.  2C shows binding curves of K27E/V188D assayed in the presence and absence of 300 nM ␣-conotoxin M1. In the absence of ␣-conotoxin M1 the curve gives a Hill coefficient of 0.6 with K d values of 57 nM and 1.8 M. In the presence of ␣-conotoxin M1, the number of sites decreased by ϳ50% and the resulting Hill coefficient increased to 0.9, consistent with residual ␣-neurotoxin binding to a single class of sites. Because ␣-conotoxin M1 will preferentially protect the ␣␦ site, the observed K d of 1.3 M reflects lower affinity binding of K27E at the ␣␥ site. Accordingly, identical experiments were carried out with each of the toxin-receptor pairs presented in Table I. In all cases, higher affinity ␣-neurotoxin binding was found to correspond to the ␣␦ site and lower affinity to the ␣␥ site, as indicated in Table I.
Coupling Analysis-The coupling coefficient ⍀ and coupling energy ⌬⌬G INT for each toxin-receptor pair (at both the ␣␦ and ␣␥ binding sites) were determined using Equations 1 and 2 (Table I and Fig. 3). Two of the mutant pairs studied gave strong coupling energies above 2.0 kcal/mol and seven pairs gave coupling energies of 1.5 kcal/mol or above. Despite very large shifts in the overall affinities of the mutant pairs relative to that of the wild-type toxin-wild-type receptor complex, the majority of the pairs showed simple additivity of free energy within experimental error (coupling energies approaching 0, 0.1-0.5 kcal/mol). Most strikingly, all of the mutant pairs involving K47A gave values very close to ⌬⌬G INT ϭ 0, indicating that the introduced receptor mutations and the toxin mutation K47A do not grossly alter the respective protein structures. A few mutant pairs gave values between 1.0 and 1.3 kcal/mol; intermediate energy values in this range are difficult to interpret because of cumulative errors of addition of free energies.
As can be seen in Table I, the strengths of the observed linkages are not identical at the ␣␦ and ␣␥ binding sites (Fig. 3,  A and B). Because the energy contributions of the individual residues alone were found to differ at the two sites, a difference in the coupling energy at the two sites might also be expected.

FIG. 2. Equilibrium binding of wild-type and mutant NmmI
␣-neurotoxin to wild-type and mutant nAChR. A, binding of wildtype NmmI to wild-type nAChR (q) and binding of the mutant toxin K47A (), K27E (OE), or R33E (f) to the nAChR Y190T. B, binding of wild-type NmmI to wild-type nAChR (q) and binding of the mutant toxin K47A (), K27E (OE), or R33E (f) to nAChR P197I. C, binding of wild-type NmmI ␣-neurotoxin to wild-type nAChR (q) and binding of the toxin-receptor mutant pair K27E/V188D in the absence (OE) or presence (f) of 300 nM ␣-conotoxin M1. Binding determinations for NmmI toxins were measured as the fractional reduction in the initial rates of 125 I-␣-bungarotoxin binding in the absence of NmmI (k max ) or in the presence of the indicated amounts of NmmI (k obs ). The curves for the wild-type NmmI-wild-type nAChR interaction and for K27E/V188D in the presence of conotoxin are least squares fits to the Hill equation with n H ϭ 1.0. The remaining curves are least squares fits to two binding sites present in equal populations.

Interactions of Receptor Residue Val 188 with Toxin Residues
Arg 33 and Lys 27 -The two strongest interactions were found between the mutant pairs R33E/V188D and K27E/V188D at the ␣␦ interface (coupling energies of 2.6 and 2.1 kcal/mol, respectively). The R33E/V188D pair also appeared to have a strong coupling at the ␣␥ interface (Ͼ1.5 kcal/mol), but a precise number could not be obtained. The very large overall loss in the binding affinity required an unachievable production level of mutant toxin. On the other hand, no significant coupling was observed at the ␣␥ interface with the K27E/V188D mutant pair containing the other charge substitution studied in loop 2.
Analysis of the direction of the free energy changes observed in the double mutant cycles involving V188D or V188K and the toxin mutations R33E and K27E are consistent with the involvement of coulombic attractions and repulsions between the introduced charged mutations. For example, the single toxin mutation R33E and single receptor mutation V188D resulted in changes in free energy from the wild-type counterpart of 5.7 and 1.8 kcal/mol, respectively (at the ␣␥ interface). The summations of the individual changes in free energy is then 7.5 kcal/mol, which corresponds to a K d of ϳ42 M. However, the observed K d of Ͼ500 M with the R33E/V188D mutant pair is larger then predicted for non-interacting sites and corresponded to a coupling energy ⌬⌬G INT Ͼ 1.5 kcal/mol. A larger than predicted K d might be expected if these two residues are in close apposition in the complex and experience coulombic repulsion from the two introduced proximal negative charges. Conversely, for the R33E/V188K interaction (at the ␣␥ interface) the K d found in the binding experiment was lower than predicted for additivity, (predicted K d ϭ 860,000 nM, ⌬⌬G ϭ 9.3 kcal/mol, experimental K d ϭ 63,000 nM, ⌬⌬G ϭ 7.7 kcal/mol) corresponding to a ⌬⌬G INT ϭ Ϫ1.6 kcal/mol. Again, this finding is consistent with coulombic attraction of the introduced negative charge on the toxin (R33E) and positive charge on the receptor (V188K).
The linkages between these mutant pairs were further explored by examining a network of mutant cycles proceeding among positive, neutral, and negative substitutions shown in Fig. 4. Here additional cycles involving R33E with K188D and K27E with K188D are also analyzed. The mutant cycle for R33E/K188D results in a very strong coupling coefficient of more than 100 at both the ␣␦ and ␣␥ binding interfaces (2.8 and Ͼ3.0 kcal/mol, respectively). This result reveals a previously unobserved strong linkage at the ␣␥ interface. The same analysis carried out with the K27E/K188D mutant pairs also gives roughly equal linkages at the two sites but 10-fold smaller than those observed with R33E with the same receptor mutations.
Other Interactions-The toxin-receptor mutant pairs K27E/ Y190F, K27E/Y190T, K27E/P197I, and K27E/D200Q all showed coupling energies of 1.5 kcal/mol or higher at the ␣␦ binding site with no significant couplings observed at the ␣␥ binding site. Approximately equal coupling values were found with either Y190T or Y190F and the K27E substitution (Table  I). Apart from its coupling to position 188 on the receptor, the R33E substitution showed coupling to the Y190T mutation. This was only observed at the ␣␥ binding site.

DISCUSSION
Using site-directed mutagenesis, we previously identified residues involved in the high affinity interaction between the ␣-neurotoxin NmmI and the mouse muscle nAChR. The goal of this study was to identify specific pairs of interacting residues between the toxin and the receptor utilizing double mutant cycles (39,47). This method has been successfully developed to identify pairwise interactions between the scorpion toxins and the potassium channel (37)(38)(39). In those studies, the known structure of the toxin was utilized to establish spatial locations of residues on the potassium channel of unknown structure.
In theory, if two residues are coupled either directly or through another residue then the effect of the double mutation will not be equal to the sum of the effects of the two single mutations (47,48). Therefore, if ⌬⌬G INT deviates from 0 (or ⍀ deviates from unity) the two mutations under study exhibit an interaction. In practice, small ⌬⌬G INT values may exist between residues separated by great distances (49). Conversely, coupling energies greater than 1.5 kcal/mol are generally associated with short distances between the two residues under study (46). In our case, the errors associated with binding measurements of two nonequivalent sites located on one receptor molecule ranged from 10 to 20%. Because of the cumulative errors when summing single mutations, the errors associated with our linkage values ranged from 30 to 50%. Therefore, we have only considered coupling energies above 1.5 kcal/mol for our analysis.
Besides direct interactions or interactions mediated through a proximal residue, large deviations from additivity can also occur when gross structural changes in the individual molecules result from the introduced mutations. If a mutation results in a global conformational change then it would be expected to be linked to a large number of residues. Therefore, the lack of linkages with nearby residues becomes a good indication that structural integrity of the interacting molecules has been maintained. In addition, failure to detect an interaction does not exclude the close proximity of the two residues. This may be due to either weak interactions between the two residues or to interactions that are compensatory yielding a minimal net change. This is especially true in the case of mutant cycles in which the reference side chain, typically the naturally occurring amino acid, exhibits a dominant influence (50).
For example, when the naturally occurring side chain Val 188 on the common ␣-subunit of two binding sites is converted to both cationic and anionic side chains therein creating two parallel cycles, different coupling energies are achieved at the ␣␦ and ␣␥ interfaces (Fig. 4). However, if we coalesce the two cycles by considering a direct substitution from a cationic to an anionic side chain, then differences between the two sites virtually vanish. This suggests the naturally occurring valine common to both small cycles imparts the asymmetry, and the influence of an inserted charge on ␣-neurotoxin binding is similar at both sites. Interactions intrinsic to the reference residue or steric constraints could influence the ␣␦ and ␣␥ sites in a differential manner (50). A more appropriate frame of reference might come from a neutral side chain isosteric with substituted residues or a side chain with minimal steric perturbation (alanine).
Of the 36 residue pairs studied here (18 at each of the ␣␦ and ␣␥ binding sites), 25 gave ⌬⌬G values below 1.0 kcal/mol, indicating simple additivity. These results suggest that the gross structural changes do not occur in the interacting molecules with the introduced mutations. Rather, the relatively few but large coupling energies that were observed support the specific interactions between these residues.
The strongest linkages observed are with the R33/V188 and K27/V188 toxin-receptor pairs. The strength of the coupling observed between these pairs varied at the ␣␦ and ␣␥ binding sites and also with the different amino acid substitutions examined. The toxin residue 33 and receptor residue 188 appear to be interacting at both the ␣␦ and ␣␥ binding sites with coupling energies as high as 2.8 and Ͼ3.0 kcal/mol observed within the network of cycles. Toxin residue 27 and receptor residue 188 also appear to be interacting at both sites but to a lesser extent, as the coupling energies observed between this pair were generally lower than with the 33/188 pair. In contrast, the toxin residue Lys 47 did not show any interaction with the receptor residue Val 188 .
Coupling energies ranging from 1.5 to 1.9 kcal/mol were also found with the toxin residue Lys 27 and the three receptor residues (Tyr 190 , Pro 197 , and Asp 200 ), all at the ␣␦ interface. These results suggest that Tyr 190 , Pro 197 , and Asp 200 are close enough in the receptor structure each to be interacting with Lys 27 . Another possibility is that some or all of these observed linkages are mediated through a third residue. The lack of coupling observed between these paired residues at the ␣␥ binding site does not preclude their interaction or their close proximity. However, it does demonstrate that the energetic contributions of Tyr 190 , Pro 197 , and Asp 200 to toxin binding differ at the two sites. These data also suggest that the toxin is binding with different orientations to the two ligand sites, where interactions with Tyr 190 , Pro 197 , and Asp 200 are less critical for the ␣␥ site, but further experiments will be necessary to address this point.
An initial model of binding is proposed from these data. The two conserved toxin cationic residues Arg 33 and Lys 27 , located on loop II of the toxin structure, are complexing with key receptor residues located on the ␣-subunit region between 180 and 200. More specifically, we suggest that the toxin residue Arg 33 is adjacent to the receptor residue Val 188 and is probably stabilized by adjoining negative or aromatic residues located on the receptor structure. One such candidate may be Tyr 190 , which did show a linkage with Arg 33 at the ␣␥ binding site. Other possibilities include residues located on the ␦/␥ subunits (see below). Lys 27 also appears to be positioned in the vicinity of Val 188 but closer to the residues Tyr 190 , Pro 197 , and Asp 200 . In this case, the lysine cation on the toxin may be directly stabilized through electrostatic interactions with Asp 200 and cation/ interactions with Tyr 190 .
The involvement of cationic residues near the tip of loop 2 on the toxin and the receptor sequence between residues 180 and 200 has been implicated from single residue mutations, chemical labeling, and binding of toxin to receptor peptide fragments (50 -55), but previous studies have not pinpointed specific residue interactions nor have they distinguished differences in ␣-neurotoxin binding between the two binding sites. Homology modeling (26) and labeling experiments (12) have indicated that residues 180 -200 on the ␣-subunit are located at the interface formed at the ␣␦ and ␣␥ ligand binding sites. Therefore, besides the ␣-subunit residues studied here, it is likely that the toxin loop II residues are interacting with ␦/␥ subunit residues. On the other hand, K47A, which is located on loop III of the toxin structure, does not appear to be interacting with this area of the ␣-subunit. Further studies aimed at identifying linkages between toxin residues and receptor residues on the ␦ and ␥ subunits should provide the additional constraints necessary to describe the toxin orientation and positioning at the two receptor binding sites.