β2 Subunit Contribution to 4/7 α-Conotoxin Binding to the Nicotinic Acetylcholine Receptor*

The structures of acetylcholine-binding protein (AChBP) and nicotinic acetylcholine receptor (nAChR) homology models have been used to interpret data from mutagenesis experiments at the nAChR. However, little is known about AChBP-derived structures as predictive tools. Molecular surface analysis of nAChR models has revealed a conserved cleft as the likely binding site for the 4/7 α-conotoxins. Here, we used an α3β2 model to identify β2 subunit residues in this cleft and investigated their influence on the binding of α-conotoxins MII, PnIA, and GID to the α3β2 nAChR by two-electrode voltage clamp analysis. Although a β2-L119Q mutation strongly reduced the affinity of all three α-conotoxins, β2-F117A, β2-V109A, and β2-V109G mutations selectively enhanced the binding of MII and GID. An increased activity of α-conotoxins GID and MII was also observed when the β2-F117A mutant was combined with the α4 instead of the α3 subunit. Investigation of A10L-PnIA indicated that high affinity binding to β2-F117A, β2-V109A, and β2-V109G mutants was conferred by amino acids with a long side chain in position 10 (PnIA numbering). Docking simulations of 4/7 α-conotoxin binding to the α3β2 model supported a direct interaction between mutated nAChR residues and α-conotoxin residues 6, 7, and 10. Taken together, these data provide evidence that the β subunit contributes to α-conotoxin binding and selectivity and demonstrate that a small cleft leading to the agonist binding site is targeted by α-conotoxins to block the nAChR.

Neuronal nicotinic acetylcholine receptors (nAChRs) 1 comprise a large family of ion channels formed by the heteropentameric assembly of homologous subunits. ␣-Conotoxins, small disulfide-rich peptides isolated from the venom of predatory cone snails, potently and selectively block nAChRs (1). They act as competitive antagonists at the ACh binding site, which is formed at the interface between neighboring subunits. These ␣-conotoxins can distinguish between muscle and neuronal nAChRs subtypes and even between different binding sites within the same receptor. As a result, they are widely used as pharmacological tools for the subtype differentiation of nAChRs in native tissues (2).
The 4/7 (4 residues in loop 1 and 7 in loop 2) ␣-conotoxins are of particular interest given their ability to discriminate between diverse neuronal ␣-␤ nAChR subunit combinations. For example, MII was found to be selective for the ␣3 and especially the ␣6 subunit containing nAChRs (3,4) and aided the identification of the nAChR subunit composition in monkey striatum implicated in Parkinson disease (5,6). The 4/7 ␣-conotoxin PnIA preferentially blocks ␣3␤2 but also ␣7 rat nAChRs (7,8), whereas ␣-GID potently blocks ␣7 and ␣3␤2 nAChRs but also acts at ␣4␤2 nAChR at higher concentrations (9). MII, PnIA, and GID share an identical backbone structure common to all 4/7 ␣-conotoxins investigated (9 -11). Therefore, their specific selectivities are thought to arise from their different amino acid side chains. As a consequence, understanding the molecular determinants of their interaction with the receptor may help in the design of optimized pharmacological tools for the nAChRs with novel or improved selectivities. However, the lack of detailed information on the three-dimensional structure of the ␣-conotoxin binding site has hampered such projects.
The acetylcholine-binding protein (AChBP) crystal structure solved by Sixma and colleagues (12) recently provided a template to create homology models of the ligand binding domain of receptor members of the Cys-loop ligand-gated ion channels family. Homology models of the ligand binding domain of ␣7, ␣3␤2, ␣3␤4, and ␣4␤2 nAChR subtypes are already providing useful insights into ligand-receptor interactions at the molecular level (13)(14)(15)(16). Our previous docking simulations of ␣-conotoxins ImI and PnIB on an ␣7 nAChR homology model revealed that these two peptides target a small cleft forming one of two possible entrances to the ACh binding site (17). We could also identify an equivalent cleft on an ␣3␤2 model as the likely region of the ␣3␤2 nAChR targeted by these peptides. Upon examination of this model, it was apparent that the ␤2 subunit contributed more than the ␣ subunit to the formation of this cleft, suggesting an important role of ␤-residues for binding and/or selectivity. In this study, we mutated three residues in the ␤2 subunit residues located in this cleft to characterize the contribution of the ␤ subunit to ␣-conotoxin binding. Each of these mutants influenced the binding of ␣-conotoxins MII, GID, and PnIA. Additionally, molecular surface analysis revealed striking shape complementarity between the highly conserved N-terminal half of the 4/7 ␣-conotoxins structure and the cleft. Finally, docking experiments showed that both MII and PnIA bind deep into the ␣3␤2 nAChR cleft, explaining at the molecular level the experimental results obtained from mutagenesis studies. Based on these results, we propose that the conserved cleft above the ␤9␤10 hairpin is the binding site for 4/7 ␣-conotoxins active at ␣7, ␣3␤2, and ␣4␤2 nAChRs and that the differences in the cleft residues provide the determinants directing ␣-conotoxin selectivity.

EXPERIMENTAL PROCEDURES
Materials-cDNAs encoding neuronal nAChRs were provided by J. Patrick (Baylor College of Medicine, Houston, TX) and subcloned into the oocyte expression vector pNKS2 (18). MII, PnIA, and GID were gifts from J. T. Blanchfield (School of Pharmacy, The University of Queensland, Australia), G. Hopping and M. Loughnan (Institute for Molecular Bioscience, The University of Queensland, Australia), respectively, and were synthesized as described previously (9,19,20).
Homology Modeling-The FASTA format of the ␣3 and ␤2 rat sequences were retrieved from the ligand-gated ion channel data base (pasteur.fr/recherche/banques/LGIC/). Their extracellular ligand-binding domains were aligned with the AChBP sequence as described previously (21).
A homology model of the ␣3␤2 nAChR subtype was built on a Silicon Graphics Octane R12000 work station using the MODELLER program (22). The AChBP structure (1I9B) was loaded in the INSIGHT II (Accelrys, San Diego, CA) environment and used as a template. Three models were built with a high "optimize level." The two options "optimize loop models" and "loop optimize level" were set to 3 and high, respectively. User disulfide was selected to assign the disulfide bonds as in AChBP except for the missing vicinal disulfide in the ␤2 subunit. The model with the lowest root mean square deviation compared with AChBP was refined further and used for these studies. Steepest descent energy minimizations were applied using the AMBER force field and DISCOVER program implemented in Insight II.
Mutagenesis-Mutagenesis of the ␤2 nAChR subunit cDNA was achieved using the QuikChange TM site-directed mutagenesis kit (Stratagene) following the manufacturer's instructions. Primers used to generate the mutants were from Proligo (Lismore, Australia). All mutations were confirmed by cDNA sequencing.
Two-electrode voltage clamp recordings were performed in oocytes 1-10 days after cRNA injection at a holding potential of Ϫ70 mV. Pipettes were pulled from borosilicate glass and filled with 3 M KCl. Resistances were below 1 megohm. Membrane currents were recorded using a Turbo Tec-10CX or a Turbo Tec 05X Amplifier (npi electronic, Tamm, Germany) filtered at 200 Hz and digitized at 400 Hz. Version   Table I. Each point represents the average of at least three measurements. Error bars represent S.E. 8.53 Pulse software (HEKA Elektronik, Lambrecht, Germany) or Cell-Works software were used. The perfusion medium was automatically switched between ND96 with or without agonist (100 M ACh) using a custom-made magnetic valve system. A fast and reproducible solution exchange (Ͻ300 ms) for agonist application was achieved using a 50-l funnel-shaped oocyte chamber combined with a fast solution flow (ϳ150 l/s) fed through a custom-made manifold mounted immediately above the oocyte. ACh pulses were applied for 2 s at 4-min intervals. After each application, the cell was superfused for 1 min with agonist-free solution, and the flow was then stopped for 3 min. Peptide was mixed from a 10-fold stock into the static bath when responses to three consecutive agonist applications differed by less than 10%. Addition of toxin directly to the recording chamber conserved material and avoided adhesion of the toxin to tubing surfaces. To obtain estimates of potency, dose-response curves were fit to the data by the equation % response ϭ 100/{1 ϩ ([toxin]/IC 50 ) nH } using Prism software (GraphPad version 3.0 for Macintosh, San Diego, CA). To obtain estimates of toxin dissociation rates, agonist responses were measured at 2-min intervals under constant superfusion, and after stabilization of responses, oocytes were incubated with toxin in a static bath for 1 min. To obtain estimates of the toxin association rate, oocytes were continuously superfused with solution containing the indicated toxin concentration and ACh pulses were applied in 1-or 2-min intervals. Association and dissociation curves were fit to the data by the equations % response ϭ (100 Ϫ plateau) ϫ e ϪK ϫ time ϩ plateau and % response ϭ span (1 Ϫ e ϪK ϫ time ) ϩ plateau, respectively.
Docking Experiments-Docking simulations of ␣-conotoxins were carried out as described in Dutertre et al. (17). Briefly, MII and PnIA structures were retrieved from the protein data bank (Protein Data Bank codes 1MII and 1PEN, respectively) and docked onto an ␣3␤2 homology model using the program GOLD, version 1.2 (Genetic Optimization for Ligand Docking, Cambridge Crystallographic Data Centre, Cambridge, UK). One NMR/crystal structure was chosen for each ␣-conotoxin, as GOLD treats ligands as flexible molecules with side chain orientations optimized during calculations. As conotoxins are competitive antagonists, the active site radius was set at 20 Å from   Trp-147 to ensure the analysis focused on residues around the ACh binding site. From the 100 docked structures obtained for each conotoxin, the selection of the final docked structure was based on a low scoring function determined in GOLD (23).

RESULTS AND DISCUSSION
Homology Modeling Identifies Val-109, Phe-117, and Leu-119 in the ␤2 nAChR Subunit as Likely Residues Interacting with ␣-Conotoxins MII and PnIA-We previously identified two cavities at opposite sides of the ␤9/10 hairpin (C-loop of an ␣ subunit) on the surface of the nAChR, one large and easily accessible and one small and narrow, from which the ACh binding site could be reached (21). To differentiate between these two cavities and avoid confusion, we propose the use of "cavity" to refer to the larger one (below the ␤9␤10 hairpin, close to the cell membrane), and "cleft" to refer to the smaller one (above the ␤9␤10 hairpin). The cavity has recently been confirmed to represent the binding site used by the large snake neurotoxins to block the nAChR (24), whereas the cleft represents the likely region targeted by ␣-conotoxins (17,21). Docking simulations with ImI and PnIB structures on an ␣7 nAChR model revealed overlapping but different binding sites for ␣-conotoxins and snake toxins (17). Both ␣-conotoxins block the receptor by targeting a small cleft above the ␤9/10 hairpin that is also blocked by the small peptide antagonist toxin waglerin (25) and the non-peptidic antagonist d-tubocurarine (26). Concerning the ␣3␤2 nAChR, only one residue (␤2-Thr-57) weakly affecting 4/7 conotoxin binding has been identified on the ␤2 subunit compared with four on the ␣3 subunit (27,28). Interestingly, ␤2-Thr-57 is located above the ␤9␤10 hairpin, forming part of the cleft in ␣3␤2 together with three other residues, Val-109, Phe-117, and Leu-119. A potentially important role for additional residues in the ␤ subunit is suggested from binding kinetic studies where exchange of the ␤2 subunit by the ␤4 subunit (in combination with ␣3 or ␣6 subunits) strongly reduced the off-rate constants of PIA (29). Likewise, BuIA showed much slower dissociation if ␣2, ␣3, ␣4, or ␣6 subunits are combined with ␤4 instead of ␤2 (30). In an attempt to identify additional determinants in the ␤ subunit that contribute to the binding and selectivity of 4/7 ␣-conotoxins, we mu- tated Leu-119, which appears deep in the cleft, as well as Phe-117 and Val-109, which form part of the wall of the cleft (Fig. 1).
Residue Leu-119 Is Generally Important for ␣-Conotoxin Binding-A leucine in position ␤2-119 lies at the bottom of the cleft and is conserved in neuronal ␤4 and ␣7 nAChR subunits as well as in the muscle ␦, ␥, and ⑀ nAChR subunits. Thus, a leucine is present at this location in all non-␣ subunits (and the (Ϫ) face of the ␣7 nAChR) for which highly active ␣-conotoxins have been identified. Indeed, no potent ␣-conotoxin has been found for the ␣9 and ␣10 subunits, which have a negatively charged aspartic acid at this position. In combination with the docking studies, this suggests that the hydrophobic environment created by ␤2-Leu-119 and the corresponding leucine residues in other subunits plays an important role in the high affinity interaction with ␣-conotoxins. To test the hypothesis that Leu-119 is generally important for ␣-conotoxin binding, the activity of three different ␣-conotoxins that bind with high affinity to the ␣3␤2 nAChRs was tested on the hydrophilic ␣3␤2-L119Q mutant. The ␣3␤2/␣6␤2-selective ␣-conotoxin MII as well as the ␣3␤2/␣7-selective ␣-conotoxins PnIA and GID had a strongly reduced activity (Fig. 2, A-C), with 40-fold (PnIA), 165-fold (GID), and 300-fold (MII) higher IC 50 values at this mutant ␣3␤2 receptor (Table I). This is the strongest effect described thus far for a non-␣ subunit mutant affecting ␣-conotoxin binding. Based on its position at the bottom of the cleft and the fact that the activity of all three conotoxins was reduced by at least 40-fold upon introduction of a similar length hydrophilic residue (Gln), Leu-119 may play an important stabilizing role for the interaction with ␣-conotoxins, allowing them to bind deep into the nAChR cleft.
In addition to the four cysteines, which are involved in two disulfide bonds and are mostly buried within the peptide core, only serine and proline residues are common to all three ␣-conotoxins. Because mutation of Ser-4 in MII did not affect potency (31), we suspect that Leu-119 might interact with the ␣-conotoxin backbone or the conserved proline found in 4/7 ␣-conotoxins (third residue of loop 1). However, this possibility is difficult to test by double mutant cycle analysis because alteration of the conserved cysteine or proline residues in the conotoxin alters the peptide backbone conformation.
Alanine Exchange of Val-109 and Phe-17 Specifically Enhances Binding of MII and GID-In our model of the ␤2 subunit, residues Val-109 and Phe-117 lie adjacent to each other on two neighboring antiparallel ␤-strands and above the Leu-119 residue, potentially forming the walls of the ␣-conotoxin binding cleft (Fig. 1). By mutating the residues into the smaller alanine, we expected to weaken the interaction with ␣-conotoxins. Unexpectedly, both the F117A and V109A mutations produced IC 50 values for MII, GID, and PnIA little different from those at the wild type receptor, suggesting that Phe-117 and Val-109 do not have close interactions with the ␣-conotoxins  (Fig. 2, A-C, Table I). The same effect was observed for V109G (except for PnIA, where a slight decrease in affinity was observed). However, the off-rates of MII were affected by all three mutations (Table II), resulting in a very slowly reversible block by MII (Fig. 3A). A similar effect on the binding was observed with GID (Fig. 3B). In contrast, the off-rates of PnIA appeared to be hardly affected by the mutations (Fig. 3C). The finding that alanine exchange of Val-109 or Phe-117 decreased the off-rate constants of MII and GID without an obvious change in the IC 50 value implied that a similar reduction in on-rate had occurred to balance the off-rate change or that the IC 50 value had not been determined accurately because of the short 3-min preincubation times. To obtain a more accurate estimate of the potency, on-rates of MII binding to the wild type and F117A mutant were compared. As expected, the F117A mutant displayed a much slower on-rate for MII compared with its on-rate at the wild type receptor (Fig. 4A). Calculated on-and off-rates for MII at the ␣3␤2F117A mutant produced a K i value of 0.69 nM (Fig. 4B), ϳ 10-fold lower than the experimentally determined IC 50 value ( Fig. 2A, Table I). Unfortunately, the long incubation times necessary to determine the real IC 50 values at the ␣3␤2F117A and Val-109 mutants were not practical because of run-up/down effects on nAChR current seen in oocytes and the large amounts of toxin needed for superfusion. Therefore, to complement the IC 50 value determinations, we compared off-rates as a measure of toxin affinity that is independent of toxin concentration or preincubation time. Because not only the off-rate constants but also the on-rate constant of GID and MII on the Val-109 and Phe-117 mutants were decreased, we propose that these residues comprise part of a short access path or an intermediate binding state that ␣-conotoxins transition past before reaching (or leaving) their high affinity bound state.
The Effect on Val-109 and Phe-117 Exchange Also Occurs in Combination with the ␣4 nAChR Subunit-The results presented above show that a small modification of the ␤2 subunit can have a strong and specific effect on conotoxin binding, suggesting that the ␤ subunit contributes to ␣-conotoxin binding through direct and specific interactions. However, it cannot be excluded that this is a secondary effect because of an alteration in the conformation of the ␣ subunit induced by the mutation in the ␤ subunit. To address this aspect, we investigated whether these ␤2 effects were preserved in combination with a different ␣ subunit. Both MII and GID showed nanomolar activity at the ␣4␤2 nAChR subtype, whereas no activity was observed for PnIA at concentration up to 1 M. When the ␣4 subunit was combined with the ␤2-F117A mutation, both GID and MII showed an ϳ10-fold increase in activity (Fig. 5A) that was likewise paralleled by a clearly decreased off-rate (results not shown). This further indicates that the binding modes of MII and GID within the binding pocket are conserved across the ␣4␤2 and ␣3␤2 receptor interfaces and that the ␤ subunit contributes directly to ␣-conotoxin potency. Whereas the overall higher IC 50 values at the ␣4␤2 combination suggest that the contribution of the ␣4 subunit to conotoxin binding is smaller than that of the ␣3 subunit, the inactivity of PnIA could also be explained by a residue in PnIA that prevents an interaction with the ␣4 subunit.
The Length of the Side Chains in Position 10 of the ␣-Conotoxins Determines the Strength of the Interaction with the Val-109 and Phe-117 Mutants-Sequence comparison of PnIA with MII and GID (Fig. 6) shows that there are 7 residues in PnIA (Leu-5, Pro-7, Ala-7, Ala-10, Pro-13, Asp-14, and Tyr-15), which are different from the corresponding residues in GID or MII and thus are likely candidates to account for the poor activity of PnIA at the ␣4␤2 interface and the low affinity of PnIA (compared with GID and MII) at the ␣3␤2 mutants. Because MII has a leucine and PnIA an alanine at position 10, it appears that a long or bulky aliphatic side chain enhances ␣-conotoxin interactions with the Phe-117 and Val-109 mutants. To further test this hypothesis, we investigated the activity of the A10L-PnIA analogue on both ␣3␤2 and ␣4␤2 mutants. As predicted, replacing Ala-10 by a longer leucine residue enhanced PnIA affinity for ␤2 receptors containing the Phe-117 or Val-109 mutations (Figs. 2D and 3D; Tables I and II). Indeed, a markedly reduced off-rate constant and an up to 23-fold lower IC 50 value were observed for A10L-PnIA at the ␤2 mutants (Table I: 3.99 nM on ␤2-F117A; 3 nM on ␤2-V109A; 2.4 nM on ␤2-V109G) compared with the wild type ␣3␤2 receptor (55 nM). Moreover, this A10L modification converted PnIA into a potent antagonist at the ␣4␤2-V109G receptor (Fig. 5B).
Residue 10 in PnIA has previously been shown to be an important determinant of selectivity between ␣3␤2 and ␣7 receptors (8). A comparison of a range of conotoxins with sequence similarity in the second loop showed that the length of the side chain in position 10 correlated with an ␣7 versus ␣3␤2 selectivity. This correlation was also found to be valid for GID and PnIA but not for MII, suggesting that MII has an altered binding mode. However, the similar influence of ␤2 mutants on GID, PnIA, and MII affinity indicate that these 4/7 ␣-conotoxins orientate similarly within the cleft formed by the ␣/␤ interface and that the critical position 10 side chain likely faces the ␤2 subunit in all three ␣-conotoxins.
Shape and Size Complementarity between 4/7 ␣-Conotoxins and the Cleft in nAChRs-A high affinity complex requires either a large contact area between receptor and ligand surfaces with multiple points of interaction or shape complemen- tarity that favors the formation of a smaller number of specific interactions (e.g. a ligand bound in a protein cleft). Given their relatively small size, the high affinity complex observed for ␣-conotoxins suggests that a tight fit between the ␣-conotoxin and a cleft on the nAChR surface is likely to occur. The ␣7 nAChR cleft appears to be ϳ7 Å deep, 13.5 Å long, and 9.5 Å wide (6.6 Å at the narrowest point). These dimensions are almost identical to those found for the ␣3␤2 nAChR (7.5 Å deep, 14.5 Å long, and 9.8 Å wide (6.8 Å at the narrowest)), and are mostly conserved in shape. For comparison, we measured the dimensions of the N-terminal hydrophobic patch of ␣-conotoxins. In MII it is 12 Å long (Gly-to Val-7) and 7 Å wide, and in PnIA it is 11.2 Å long (Gly-1 to Pro-7) and 6.5 Å wide, whereas GID has a flexible tail that complicates the measurements (a tail-truncated analogue has similar dimension to PnIA and MII, and the tail could be accommodated by the flexible loop F of nAChR). These comparisons reveal a striking size and shape complementarity between 4/7 ␣-conotoxins and the nAChR cleft, strongly suggesting a lock-and-key interaction (Fig. 7).
Docking Models of PnIA and MII Interactions with ␣3␤2 nAChR-To verify that 4/7 ␣-conotoxins can interact with the nAChR cleft via their conserved hydrophobic patch (residues 6, 7, and 10), docking of PnIA and MII structures onto an ␣3␤2 nAChR model were simulated using the program GOLD. Analysis of the results from docking simulations confirmed both hypotheses: the small cleft is the binding site for 4/7 ␣-conotoxins, and the ␣-conotoxins bind to the ␣3␤2 nAChR by presenting their conserved N-terminal shape (Fig. 7). Indeed, although an active site radius set to 20 Å would have allowed ligands to dock at other locations around the C-loop, all structures were found docked in the cleft above the ␤9/10 hairpin. The final orientation of PnIA and MII docked into the receptor cleft presented in Fig. 8 was chosen from the cluster of lowest energy structures produced by GOLD and on the basis of how well it fitted the available experimental data. Both ␣-conotoxins bind in a similar way to the docking mode found for PnIB at the ␣7 nAChR (17). It is interesting to note that although MII still binds to ␣3␤2 from its hydrophobic patch similarly to PnIA, its position in the cleft appears slightly shallower compared with PnIA, as expected given the greater width of its C-terminal half (see Fig. 7). Interestingly, MII is observed to make additional contacts with ␣3␤2 nAChR residues, giving an explanation for its very slow off-rate and associated higher affinity compared with PnIA. Indeed, although ␤2-L119 appears in close proximity to the conserved P6 residue, and position 10 residues are found between ␤2-V109 and ␤2-F117 in both PnIA and MII, only His-9 of MII makes an electrostatic interaction with ␤2-Glu-57 possible. However, the hydrophobic interaction between ␣-conotoxins (positions 6, 7, and 10) and the nAChR cleft (Leu-119, Val-109 and Phe-117) appears to play a significant role in the high affinity complex observed. Mutagenesis experiments carried out by others have identified 5 residues influencing MII and PnIA binding (27,28). Although ␤2-Thr-57 is located in the cleft, and ␣3-Ile-183 is in close proximity of both docked structures, residues Lys-183, Pro-180, and Gln-196 are far from the binding site. We propose that their effect on peptide binding is likely to be indirect, as they are all located on the structurally important ␤9/10 hairpin and include two proline mutants (P180T and Q196P), expected to have a significant structural effect. The orientation of 4/7 ␣-conotoxins in the cleft explains the functional data obtained from the ␤2 mutants.
Pharmacophore-We developed a minimal antagonist pharmacophore for 4/7 ␣-conotoxins based on our mutagenesis and docking results (Fig. 9). PnIA has a very hydrophobic N-terminal patch, comprising Leu-5, Pro-6, Pro-7, Ala-9, and Ala-10 ( Fig. 9A). MII also possesses a similar hydrophobic core (Pro-6, Val-7, and Leu-10) that we take to represent a minimum antagonist pharmacophore (Fig. 9B). When compared with the ␤2 subunit pharmacophore (Fig. 9C), the distances found in conotoxins are compatible with the direct interaction between the conserved proline in position 6 and ␤2-Leu-119 that we observed in our docking model. We propose that Pro-6 in 4/7 conotoxin acts as an anchor and stabilizes the conotoxin-receptor complex. Additional hydrophobic contacts such as Pro-7 (PnIA) or Val-7 (MII) and Ala-10 (PnIA) or Leu-10 (MII) with ␤2-Val-109 and ␤2-Phe-117 strengthen the interaction, whereas secondary interactions such as His-9 (MII) and ␤2-Glu-57 determine the selectivity. A comparison of residues in the equivalent position of the receptor pharmacophore in other ␤ subunits and (Ϫ) ␣7 reveals important differences likely to influence ␣-conotoxin selectivity (Fig. 9C). For example, the polar Thr-119 residue present in ␤3 may explain the absence of ␣-conotoxins acting at nAChRs containing this subunit.
Conclusions-We have demonstrated the predictiveness of nAChR homology models based on the AChBP structure and characterized a small cleft as the common binding site for 4/7 ␣-conotoxins. Point mutations in this cleft revealed a major effect of ␤2-Leu-119 on ␣-conotoxin affinity and important influences of ␤2-Phe-117 and ␤2-Val-109 on ␣-conotoxin binding kinetics. Furthermore, we could demonstrate a specific interaction of residue 10 in ␣-conotoxins with the ␤2 subunit. Docking simulations of PnIA and MII at ␣3␤2 confirmed these results by showing a direct interaction between ␤2 and ␣-conotoxin residues. These studies indicate that ␣-conotoxins block the nAChRs by a lock-and-key interaction, where a primary hydrophobic interaction between ␤2-Leu-119 and the conserved Pro of 4/7 ␣-conotoxins "locks" the ligand in its cleft, and secondary complementary interactions contribute to ligand selectivity. Based on homology models of nAChRs, the identification of ligand-accessible residues in the cleft (Leu-119, Val-109, and Phe-117) and the determination of a minimal antagonist pharmacophore, a rational approach to the design of subtypeselective nAChR modulators can now be pursued.