Subunit Interface Selectivity of the α-Neurotoxins for the Nicotinic Acetylcholine Receptor*

Peptide toxins selective for particular subunit interfaces of the nicotinic acetylcholine receptor have proven invaluable in assigning candidate residues located in the two binding sites and for determining probable orientations of the bound peptide. We report here on a short α-neurotoxin from Naja mossambica mossambica (NmmI) that, similar to other α-neurotoxins, binds with high affinity to αγ and αδ subunit interfaces (K D ∼100 pm) but binds with markedly reduced affinity to the αε interface (K D ∼100 nm). By constructing chimeras composed of portions of the γ and ε subunits and coexpressing them with wild type α, β, and δ subunits in HEK 293 cells, we identify a region of the subunit sequence responsible for the difference in affinity. Within this region, γPro-175 and γGlu-176 confer high affinity, whereas Thr and Ala, found at homologous positions in ε, confer low affinity. To identify an interaction between γGlu-176 and residues in NmmI, we have examined cationic residues in the central loop of the toxin and measured binding of mutant toxin-receptor combinations. The data show strong pairwise interactions or coupling between γGlu-176 and Lys-27 of NmmI and progressively weaker interactions with Arg-33 and Arg-36 in loop II of this three-loop toxin. Thus, loop II of NmmI, and in particular the face of this loop closest to loop III, appears to come into close apposition with Glu-176 of the γ subunit surface of the binding site interface.

The nicotinic acetylcholine receptor (nAChR) 1 found in muscle is a pentamer composed of four homologous subunits present in the stoichiometry ␣ 2 ␤␥␦ (fetal subtype) or ␣ 2 ␤⑀␦ (adult subtype). The subunits are arranged in a circular manner to surround a central channel in the order, ␣␥␣␦␤ or ␣⑀␣␦␤ (1-3). The two binding sites for agonists, competitive antagonists, and the slowly dissociating ␣-neurotoxins are formed at interfaces between the ␣␦ and ␣␥(⑀) subunit pairs. The extracellular domain in each subunit is formed principally from the aminoterminal 210 amino acids, which is followed by four membranespanning domains. Residues within the amino-terminal 210 amino acids have been shown to be the major contributors to the ligand binding sites and for dictating the order of assembly of subunits.
Three segments of the ␣ subunit, well separated along the linear sequence, harbor major determinants for ligand binding; these segments contain the key residues around Tyr-93, between Trp-149 and Asp-152, and spanning the region from Val-188 through Asp-200 (see Refs. 3 and 4 for reviews). Similarly, four discontinuous segments of the non-␣ subunits, appearing on the opposite face of the subunit, contain major determinants for ligand selectivity; in the ␥ subunit these segments contain the key residues Lys-34, between Trp-55 and Gln-59, between Ser-111 and Tyr-117, and between Phe-172 and Asp-174.
Since the early demonstration of irreversible neuromuscular blockade by the peptide from snake venom, ␣-bungarotoxin (5), and the use of labeled ␣-neurotoxins to identify the nAChR (6), these toxins have been the primary ligands employed for the identification and characterization of the muscle nAChR. Amino acid sequences are available for nearly 100 members of the ␣-neurotoxin family, which show a common basic structure consisting of three polypeptide loops emerging from a small globular core (7). ␣-Neurotoxins can be divided into the short (4 disulfide bonds and 60 -62 residues) and long neurotoxins (5 disulfide bonds and 66 -74 residues). Crystal and solution structure determinations reveal similar tertiary structures. Although these structurally well defined toxins are known to bind at the subunit interfaces (␣⑀ or ␣␥ and ␣␦), typically with a K D Յ 100 pM, little is known about their precise orientation with respect to the subunits that form the interfaces.
Points of attachment of ␣-neurotoxin within the nAChR binding sites have been examined by cross-linking chemically modified (8,9) or photoactivatable derivatives of ␣-neurotoxin (10 -13) and by simple ultraviolet irradiation without chemical modification (14). These labeling studies have suggested contacts with both ␣ and non-␣ subunits at the binding sites (see Refs. 2, 3, and 15 for reviews). Mutagenesis studies have also identified candidate residues in the principal loops of the ␣ (16) and non-␣ subunits (17) that contribute to ␣-toxin binding. Although most ␣-toxins do not distinguish between the two sites on the receptor, an ␣-toxin from the venom of Naja mossambica mossambica (NmmI) distinguishes between the two sites of the Torpedo receptor (18). Thus NmmI emerges as a potentially valuable ligand for determining regions of close approach between ␣-toxins and the non-␣ subunits at the binding site. Previous work showed that the ␣␥ and ␣␦ binding sites of the fetal mouse receptor exhibit similar affinities for NmmI (19). However, certain mutations in the NmmI toxin structure, and surprisingly also in the nAChR ␣ subunit common to both sites, resulted in nonequivalent reductions in affinity at the ␣␥ and ␣␦ binding sites (19). Here we examine binding of recombinant NmmI ␣-toxin to fetal and adult mouse AChRs and find that the affinity of NmmI for the ␣⑀ interface is 3 orders of magnitude lower than for the ␣␥ and ␣␦ interfaces. Using subunit chimeras and site-directed mutations in ␥ and ⑀ subunits, we show that the enhanced affinity conferred by the ␥ over the ⑀ subunit arises from Pro-175 and Glu-176 in the ␥ subunit. Mutant cycle analysis shows that Glu-176 interacts with cationic residues in loop II of the NmmI ␣-toxin.

EXPERIMENTAL PROCEDURES
Materials-␣-Conotoxin MI was purchased from American Peptide Company. 125 I-labeled ␣-bungarotoxin (␣-BgTx) (specific activity ϳ16 Ci/g) was a product of NEN Life Science Products.
NmmI Expression and Purification-A double-stranded synthetic NmmI cDNA containing the ␣-erabutoxin signal sequence and strategically placed restriction sites was subcloned into pEZZ vector encoding two IgG binding proteins from Staphylococcal protein A. The Staphylococcal protein A-NmmI fusion protein was expressed using Escherichia coli HB 101, cleaved, and purified as described in Ackermann and Taylor (19).
Construction of Mutant nAChR-cDNAs encoding mouse nAChR subunits were subcloned into a cytomegalovirus-based expression vector, pRBG4. All mutations were introduced using Quick Change TM Site-Directed Mutagenesis Kit (Stratagene) or by bridging two introduced or natural restriction sites with double-stranded oligonucleotides. Chimeras were also constructed by bridging natural or introduced restriction sites. After ligation of the fragments containing the mutated site or synthesized oligonucleotide into the original pRBG4 vector, the subcloned cassette was sequenced by the dideoxy method.
Ligand Binding Measurements-Cells were harvested in phosphatebuffered saline, pH 7.4, containing 5 mM EDTA, 2-3 days after transfection. They were briefly centrifuged, resuspended in potassium-Ringers buffer, and divided into aliquots for binding assays. Specified concentrations of NmmI and conotoxin MI were added to the samples 20 min prior to initiating the association rate assay with 125 I-labeled ␣-bungarotoxin. Dissociation constants of the ligands were determined from their fractional reduction of the initial rate of 125 I-labeled ␣-bungarotoxin association (20,21). Appropriate concentrations of ␣-conotoxin MI were used to block the ␣␦ interface but not that of ␣␥, when the distinctions in affinity for the ␣␥ and ␣␦ interfaces were not obvious (19).
Rate Measurements-The association rate for 125 I-labeled ␣-bungarotoxin was measured using a ␣-bungarotoxin concentration of 20 nM. At the specified time, cells were washed with 30 mM carbamylcholine in K ϩ -Ringers solutions and then washed two times with K ϩ -Ringers solution alone and counted. For the measurement of dissociation rates, we equilibrated the surface receptor with 40 nM of 125 I-labeled ␣-bungarotoxin for 2 h, then washed the cells two times with K ϩ -Ringers to remove the unbound ligand and resuspended the cells in K ϩ -Ringers with a 10-fold dilution. Unbound toxin was removed by washing at specific times, and the cells were counted (16).

RESULTS
Insensitivity of the ␣⑀ Site to NmmI-Certain competitive antagonists distinguish between the two binding sites of the nAChR because of species or subtype differences in the non-␣ subunits that form the ␣␥, ␣␦, and ␣⑀ binding sites. We therefore compared binding of NmmI to mouse fetal (␥ containing) and adult (⑀ containing) nAChRs expressed in HEK 293 cells. As described previously, NmmI does not distinguish between ␣␥ and ␣␦ sites of fetal receptors, binding to a single class of sites with a K D of 0.14 nM (Ref. 19; Fig. 1). Surprisingly, however, NmmI selects strongly between ␣␦ and ␣⑀ sites of the adult receptor, binding to the ␣⑀ site with three orders of magnitude lower affinity (K D ϭ 130 nM; Fig. 1). Thus the ⑀ subunit of the adult receptor contains residues that confer insensitivity to NmmI.
Measurements of NmmI binding also show apparently unequal fractions of high and low affinity sites in the adult receptor (fraction of high affinity sites ϭ 0.63 Ϯ 0.03; n ϭ 3), rather than a value of 0.5 expected for an equal abundance of ␣␦ and ␣⑀ sites. This apparent inequality could arise from distinct rates of association of the reporter ligand 125 I-labeled ␣-bungarotoxin with ␣␦ and ␣⑀ sites. To examine this possibility, we compared time courses of association of 125 I-labeled ␣-bungarotoxin for fetal and adult nAChRs (Fig. 2). For the ␣ 2 ␤␥␦ fetal receptor, association is well described by a single exponential component with a k on of 3.6 Ϯ 0.7 ϫ 10 6 M Ϫ1 min Ϫ1 (n ϭ 5), indicating indistinguishable ␣␦ and ␣␥ sites. On the other hand, for the ␣ 2 ␤⑀␦ adult receptor, association can be described by two exponential components with equal amplitudes, with a faster component of k on ϭ 3.9 Ϯ 0.5 ϫ 10 6 M Ϫ1 min Ϫ1 and slower component of k on ϭ 1.0 Ϯ 0.2 ϫ 10 6 M Ϫ1 min Ϫ1 (n ϭ 2). The different association rates could account for the apparently unequal fractions of sites detected by NmmI when it competes with the initial rate of 125 I-labeled ␣-bungarotoxin binding (17). The apparent fraction of the two sites will be biased toward the site with the more rapid rate of ␣-bungarotoxin binding.
We also compared time courses of dissociation of 125 I-labeled ␣-bungarotoxin from fetal and adult receptors. The two types of receptors did not show significant differences in dissociation time courses as estimated from the initial portion of the dissociation profiles with k off ϭ 2.5 Ϯ 0.5 ϫ 10 Ϫ4 min Ϫ1 for ␣ 2 ␤␥␦ and k off ϭ 2.6 Ϯ 0.6 ϫ 10 Ϫ4 min Ϫ1 for ␣ 2 ␤⑀␦ (n ϭ 3). The slow rates of dissociation preclude measurements over the entire time course because of progressive autolysis of the preparation. The ratios of dissociation to association rate constants yield equilibrium dissociation constants for ␣-bungarotoxin of 69 pM (␣␦ and ␣␥) for ␣ 2 ␤␥␦ and 67 pM (␣␦), 260 pM (␣⑀). These kinetic experiments demonstrate that ␣-bungarotoxin, has far less capacity than NmmI to distinguish between the ␣␥ and ␣⑀ sites.
Molecular Basis of Insensitivity of the ␣⑀ Site for NmmI Toxin-The ␥ and ⑀ subunits show high sequence identity in the extracellular domains (54% in mouse), and homologous residues should have virtually identical locations for their ␣-carbon backbone positions. Yet, NmmI binds 1000-fold more tightly to the ␣␥ than to the ␣⑀ site. To determine the structural basis of NmmI selectivity, we constructed subunit chimeras containing portions of the ␥ subunit substituted into the ⑀ subunit. Each chimera was coexpressed with complementary ␣, ␤, and ␦ subunits, followed by measurements of NmmI binding.
FIG. 1. Sensitivity of nAChR expressed as ␣ 2 ␤␥␦ and ␣ 2 ␤⑀␦ in HEK cells to ␣-neurotoxin from Naja mossambica mossambica (NmmI). HEK cells were transfected with cDNAs encoding wild type subunits to form cell surface receptors with the composition ␣ 2 ␤␥␦ (q) and ␣ 2 ␤⑀␦(E). ␣-Neurotoxin (NmmI) binding was measured by competition with an initial rate of ␣-bungarotoxin binding. kobs/kmax is the ratio of initial rates for 125 I-labeled ␣-bungarotoxin binding in the presence and absence of NmmI. Data are plotted according to the equation: , where k T,␣␦ and k T,␣⑀ are the ␣-bungarotoxin association rates for the ␣␦ and ␣⑀ sites, K ␣␦ and K ␣⑀ are the equilibrium dissociation constant for NmmI of the respective sites, and k max ϭ 0.5 k T,␣␦ ϩ 0.5 k T,␣⑀ . For ␣ 2 ␤␥␦, the data are fit to a single class of sites. The fraction of sites with an affinity corresponding to ␣␦ and ␣⑀ is reflected by the inflection. The apparent fraction of sites of ␣␦ (0.62) to ␣⑀ (0.38) in ␣ 2 ␤⑀␦ that is greater than the predicted equal population of sites since k T,␣␦ Ͼ k T,␣⑀ (cf. Fig. 2).
We first screened with chimeras containing ␥ sequence from the amino terminus to junctions ranging from positions 74 to 173 of the ⑀ subunit. Each of these chimeras confers low affinity for NmmI, characteristic of the wild type ␣⑀ site (Fig. 3), indicating that NmmI selectivity arises from residues carboxylterminal to position 173. By contrast, moving the chimera junction just four residues to position 177 increases NmmI affinity to that of the native ␣␥ site (K D ϭ ϳ80 pM; Fig. 3). Thus residue differences between positions 174 and 177 confer NmmI selectivity for the ␣␥ over the ␣⑀ site.
Sequence comparison of the ␥ and ⑀ subunits reveals only two mismatched residues between positions 174 and 177 (Fig. 4A). We therefore constructed point mutations at these two positions of the ␥ subunit and measured NmmI binding to the resulting mutant receptors. The point mutations ␥P175T and ␥E176A reduce affinity to values intermediate to those of the wild type ␣␥ and ␣⑀ sites (P175T, K D,mt /K D,wt ϭ 36; E176A, K D,mt /K D,wt ϭ 16; Fig. 4, Table I). Combining the two mutations into a single ␥ subunit reduces affinity to approach that of the wild type ␣⑀ site (Fig. 4, Table I). The converse double mutation in the ⑀ subunit, ⑀T176P and ⑀A177E, increases NmmI affinity to equal that of the wild type ␣␥ site (Fig. 4, Table I). Thus, the residue pairs at homologous positions ␥Pro-175/⑀Thr-176 and ␥Glu-176/⑀Ala-177 fully account for the 1000-fold selectivity of NmmI for the ␣␥ over the ␣⑀ site.
Sequence alignment shows that the ␦ subunit contains the same residues in this region as the ␥ subunit, consistent with high affinity of the ␣␦ site. We therefore attempted to produce a low affinity ␣␦ site by introducing the residue determinants in ⑀ that reduce ␥ affinity into equivalent positions of the ␦ subunit. The mutations ␦P181T and ␦E182A, singly or combined, do not affect appreciably NmmI affinity (Table I), indicating that substitutions of other residues unique to the ⑀ subunit into the ␦ subunit are required to decrease affinity of NmmI for the ␣␦ site.
Residues in NmmI That Interact with Selectivity Determinants in the ␥ and ⑀ Subunits-Because Pro-175 of the ␥ subunit likely orients Glu-176 to come into close apposition with a cationic residue on the NmmI toxin, we asked whether an anionic residue at the homologous position to 176 in the ⑀ subunit stabilizes a cationic residue in the central loop of NmmI. Because Coulombic interactions can be effective over relatively long distances, and both attractive (opposite charges) and repulsive (like charges) forces can be generated, we measured binding of the mutant toxins, K27E, R33E, and R36E to receptors containing the mutation ␥E176K. Each pair of receptor-toxin mutations is equivalent to a charge reversal between receptor and toxin, and should the distance relationships be appropriate, charge reversal could preserve a stabilizing interaction, if Coulombic forces prevail.
Among the cationic residues in NmmI, Lys-27 showed the strongest interaction with ␥Glu-176. The receptor mutation FIG. 2. Kinetics of 125 I-labeled ␣-bungarotoxin association with the nAChR expressed as ␣ 2 ␤␥␦ and ␣ 2 ␤⑀␦ in HEK cells. Top panel, association of 5 nM 125 I-labeled ␣-bungarotoxin with 200 pM wild type ␣ 2 ␤␥␦ (q) and ␣ 2 ␤⑀␦(E) receptors. The data for ␣ 2 ␤␥␦ are fit by a single exponential approach to equilibrium with a k on of 3.6 Ϯ 0.7 ϫ 10 6 M Ϫ1 min Ϫ1 , whereas for ␣ 2 ␤⑀␦ a two-exponential fit of equal amplitudes is used k on of 3.9 Ϯ 0.5 ϫ 10 6 M Ϫ1 min Ϫ1 and 1.0 Ϯ 0.2 ϫ 10 6 M Ϫ1 min Ϫ1 . Bottom panel, dissociation of 125 I-labeled ␣-bungarotoxin after equilibration of 40 nM toxin with 20 pM receptor, washing twice with K ϩ -Ringers solution, and resuspending with dilution in K ϩ -Ringers solution. Averaging three such dissociation experiments yields k off ϭ 2.5 Ϯ 0.5 ϫ 10 Ϫ4 min Ϫ1 for ␣ 2 ␤␥␦ and 2.6 Ϯ 0.6 ϫ 10 Ϫ4 min Ϫ1 for ␣ 2 ␤⑀␦. ␥E176K decreases affinity of the ␣␥ site for NmmI by nearly 3 orders of magnitude. Similarly, the NmmI mutation K27E decreases affinity of NmmI for the ␣␥ site by approximately 2.5 orders of magnitude. However, combining both ␥E176K and K27E results in a complex that is more stable by 4.5 orders of magnitude of NmmI concentration than expected for noninteracting pairs of residues (Fig. 6). Thus charge reversal of ␥E176K and K27E preserves a stabilizing interaction, suggesting that electrostatic force between ␥Glu-176 and Lys-27 is a primary factor in stabilizing the toxin-receptor complex.
We used thermodynamic mutant cycle analysis to determine free energy of interaction between charged pairs of residues in the receptor and NmmI (22) (Scheme I). In this mutant cycle, the asterisk indicates the presence of a mutation in either the receptor (R) or the NmmI toxin (T).
The loss of energy, ⌬⌬G, arising from substitution from wild type into mutant is calculated from the dissociation constant (K D ) as follows.
The coupling energy, ⌬⌬G INT , is defined in terms of the respective dissociation constants (K) of the complexes, where the ⌬G°values are the standard free energies for formation of the toxin/receptor complex. If the mutations do not interact, the two differences in standard free energies should be equal because the effect of mutating the receptor should be independent of whether or not toxin is mutated (Equation 3).
Similarly, mutating the toxin should be independent of the receptor mutations (Equation 4). On the other hand, if the two SCHEME 1.

TABLE I
The influence of ␣-toxin and receptor mutations at the ␥, ␦, and subunits on Naja mossambica mossambica (NmmI) ␣-toxin association with the nicotinic acetylcholine receptor Dissociation constants were calculated from competition with the initial rate of the 125 I-labeled ␣-bungarotoxin binding. Receptor was expressed as ␣ 2 ␤␥␦ or ␣ 2 ␤␦ by transfection of cDNAs encoding four respective sets of subunits. K D is dissociation constant for ␣␦, ␣␥, or ␣ sites by fitting a two-site analysis. The ratios of dissociation constants of mutant (mt) to wild type (wt) were calculated using an average or mean value of at least two measurements involving separate transfections. ⌬⌬G is free energy of binding calculated from Equation 1 in the text. ⌬⌬G INT was calculated using Equation 2 in the text. Values less than unity were inverted and indicated with a minus sign.

Receptor
Toxin  4. Assignment of the residues contributing to ⑀ subunit insensitivity to NmmI binding. A, top, the junctions of two chimeras, ␥173⑀ and ␥177⑀. The shaded bar represents ␥ subunit sequence followed the nonshaded ⑀ subunit. Junctional amino acids are shown by the bar. Amino acids from ⑀ are designated in italics. Numbering is for ␥ subunit. Only two differences at positions ␥175 and 176 (⑀176/177) exist in this region. A, bottom, mutations (*) were introduced at ␥ or ⑀ subunits to examine the contributions at these two positions. K D and the K D,mt /K D,wt are shown in a logarithmic scale as described in Fig. 2. B, sequence alignment around 175 and 176 in the ␥ subunit. The superscripts in italics show reported determinants for binding of various ligands. Superscripts are: waglerin, W (30); ␣-conotoxin MI, CM (29); anionic residues for agonists, A (25,26), and acetylcholine, ACh (36). mutations interact, the bracketed differences should not be equal. When applied to the ␥E176K/K27E pair, mutant cycle analysis reveals a substantial free energy of interaction of Ϫ5.9 kcal/mol (Table I). Similar analysis of the NmmI mutations, R33E and R36E, reveal modest interaction-free energies of Ϫ2.7 and Ϫ2.2 kcal/mol, respectively (Table I). The overall results indicate close approach of cationic residues in the central loop of the NmmI toxin and ␥Glu-176 of the binding site, with the most proximal charged residue being Lys-27 of the toxin.

DISCUSSION
The ␣-neurotoxins are a family of three-fingered peptide toxins found in venom of elapid snakes (7). They have proven to be invaluable tools for the isolation and study of the nAChR because of their high affinities and slow rates of dissociation from the receptor (5,6). Although the isolated ␣-subunit of the receptor retains the capacity to bind ␣-BgTx, whereas isolated ␤, ␥, or ␦ subunits do not, the ␣-toxins bind with far lower affinity to the ␣ subunit than to the intact receptor. Moreover, small agonists and antagonists do not compete with ␣-toxin binding to the isolated ␣ subunit at expected concentrations (23). These observations point to a predominant, but not sole, contribution to the ␣-neurotoxin binding coming from the ␣ subunit. Our previous work showed that , and Asp-200 of ␣ subunit contribute to NmmI binding (19). Also glycosylation at positions 189 and 187, yielding oligosaccharides uniquely found in cobra and mongoose nAChR, reduced ␣-BgTx binding substantially (16).
Although the ␣ subunit appears to be the predominant site of ␣-toxin binding, chemical cross-linking and mutagenesis studies show that non-␣ subunits are close to the site of ␣-neurotoxin binding (Refs. 8 -14, 16, 17, and see Refs. 3 and 15 for reviews). The results described here further illustrate the role of neighboring non-␣ subunits in contributing to high affinity ␣-toxin binding, as NmmI binds to ␣⑀ interfaces of the adult type of nAChR (␣ 2 ␤⑀␦) with 3 orders of magnitude lower affinity than to the ␣␥ and ␣␦ interfaces of the fetal receptor (␣ 2 ␤␥␦). Binding studies, initially using chimeras and subsequently point mutants, show that ⑀Thr-176/⑀Ala-177 (Pro/Glu in ␥/␦) contribute entirely to insensitivity of the ␣⑀ interface to NmmI. The observation that ␣-bungarotoxin association is only slightly affected by the ␥ and ⑀ sequence differences suggests that this region of the ␥, ⑀, and ␦ subunits is not used equivalently for stabilization of the entire family of bound ␣-neurotoxins. At the present time, it is unclear whether stabilization from this region is unique to some of the short ␣-neurotoxins, or the long ␣-neurotoxins, such as ␣-bungarotoxin, acquire the bulk of their stabilization energy from other portions of the structure. Distinct differences in specificity between the short and long neurotoxins have been noted for the ␣7 subtype of nAChR (24).
Residues at the ␥175/176 positions were previously unrecognized as determinants of ligand binding. However, they are immediately adjacent to ␥Asp-174, which was shown by crosslinking to be ϳ9 Å away from  in the ␣ subunit (25,FIG. 5. Structure of an ␣-neurotoxin from Naja mossambica mossambica (NmmI) and mutations studied. An energy minimization model of NmmI described in (19), is shown with the mutated side chains of loop II. The concave face of the toxin is facing the viewer.
FIG. 6. Linkage of the free energy of binding between charge modifications on the ligand and receptor. A and B, inhibition of the initial rate of 125 Ilabeled ␣-bungarotoxin binding to cell surface nAChR expressed as ␣ 2 ␤␥␦ by wild type NmmI (A) and K27E mutant NmmI (B). C and D, inhibition of the initial rate of 125 I-labeled ␣-bungarotoxin binding to cell surface nAChR expressed as ␣ 2 ␤(␥E176K)␦ by wild type (C) and K27E mutant (D) NmmI ␣-toxin. The dashed line in D is the predicted curve when the coupling energy (⌬⌬G INT ) between ␥E176K and K27E is 0. The deviation of observed affinity (D) from that predicted by no linkage between the residues at the ␣␥ site (dashed line) produces a large coupling energy of Ϫ5.9 kcal/mol. 26) and was shown to influence the affinity of quaternary agonists and antagonists (27,28). Moreover, the adjacent residues of ␥Phe-172 (␦Asp-178, ⑀Ile-173) are known to confer site-selectivity to the smaller competitive peptide inhibitors such as ␣-conotoxin MI (29) and waglerin (30). The equivalent region of the ␣7 subunit (Asp-163, Ile-164, and Ser-165), which presumably forms a homomeric pentamer of subunits, constitutes part of a putative Ca 2ϩ binding region that faces the ligand binding site (31). At the ␣ subunit interface of the binding site, both aromatic  and anionic (Asp-200) residues were mapped to the ␣-toxin binding surface (19). Here, we identify another anionic residue in the ␥ subunit, Glu-176, perhaps restricted in its position by a neighboring secondary amino acid Pro-175, on the ␥ subunit, as a crucial residue for binding.
The significant linkage between Glu-176 and cationic residues in loop II of the toxin suggests that an electrostatic interaction contributes to the tight binding of the NmmI/nAChR complex. Because the linkage obtained from charge reversal is greater for Lys-27 than for Arg-33 and Arg-36, one would predict that the portion of loop II proximal to loop III of the toxin, is closest to the ␥ subunit (cf. Fig. 5). In this analysis, the loss of free energy (⌬⌬G) associated with a single charge mutation results from all interactions between the charged residue and its multiple neighboring residues. The pairwise interactions (⌬⌬G INT ) resulting in charge reversal of specified residues in the interacting molecules should then highlight the strength of interaction coming from the paired charged residues. In the absence of significant changes in conformation or hydration, ⌬⌬G INT from Equation 2 should largely reflect the Coulombic interaction between the respective paired residues (32).
Extensive studies on a related short neurotoxin, erabutoxin a, involving mutations at 36 toxin positions clearly revealed the importance of the tips of loops situated on the concave face of the toxin (33,34). These investigations showed that the K27E mutation of erabutoxin a decreases its affinity more than 100fold for Torpedo nAChRs. Photo-activable p-azidobenzoyl and p-azidosalicyl groups attached to Lys-26 (analogous position at Lys-27 of NmmI) of neurotoxin II labeled ␥ and ␦ subunits of the receptor upon photolysis (11,12). Three different photoactivatable groups attached to the equivalent residue Lys-23 of a long neurotoxin, toxin 3, also labeled predominantly the ␥ and ␦ subunits in preference to the ␣ subunit (13). Thus mutagenesis and chemical labeling studies showed a crucial role of lysine at position 27 and its proximity to ␥ and ␦ subunits. Here, our mutant cycle studies delineate the interaction between Glu-176 of the ␥ subunit and Lys-27 of NmmI toxin. The largest linkage in loop II between K27E and ␥E176K (⌬⌬G ϭ Ϫ5.9kcal/mol) and smaller linkages (⌬⌬G INT ϭ Ϫ1.6 to Ϫ3.0 kcal/mol) found previously between K27E and ␣ subunit residues of Val-188,  correlate well with the labeling studies. A more complete elucidation of ␣-toxin-receptor interactions should enable us to orient a docked ␣-toxin with respect to the subunit interfaces, as well as refine existing models of the structure of the extracellular domain of the receptor (4).