Covalent Trapping of Methyllycaconitine at the α4-α4 Interface of the α4β2 Nicotinic Acetylcholine Receptor

Background: Methyllycaconitine is an antagonist at subtypes of the nicotinic acetylcholine receptor. Results: A reactive methyllycaconitine probe was covalently trapped by a cysteine introduced on the complementary face of the α4 subunit and only in the (α4)3(β2)2 nAChR stoichiometry. Conclusion: The α4-α4 interface on the α4β2 nAChR contains a methyllycaconitine binding site. Significance: Defining the molecular interactions of nAChR ligands at the α4-α4 interface may lead to superior therapeutics. The α4β2 nicotinic acetylcholine receptors (nAChRs) are widely expressed in the brain and are implicated in a variety of physiological processes. There are two stoichiometries of the α4β2 nAChR, (α4)2(β2)3 and (α4)3(β2)2, with different sensitivities to acetylcholine (ACh), but their pharmacological profiles are not fully understood. Methyllycaconitine (MLA) is known to be an antagonist of nAChRs. Using the two-electrode voltage clamp technique and α4β2 nAChRs in the Xenopus oocyte expression system, we demonstrate that inhibition by MLA occurs via two different mechanisms; that is, a direct competitive antagonism and an apparently insurmountable mechanism that only occurs after preincubation with MLA. We hypothesized an additional MLA binding site in the α4-α4 interface that is unique to this stoichiometry. To prove this, we covalently trapped a cysteine-reactive MLA analog at an α4β2 receptor containing an α4(D204C) mutation predicted by homology modeling to be within reach of the reactive probe. We demonstrate that covalent trapping results in irreversible reduction of ACh-elicited currents in the (α4)3(β2)2 stoichiometry, indicating that MLA binds to the α4-α4 interface of the (α4)3(β2)2 and providing direct evidence of ligand binding to the α4-α4 interface. Consistent with other studies, we propose that the α4-α4 interface is a structural target for potential therapeutics that modulate (α4)3(β2)2 nAChRs.

The ␣4␤2 nicotinic acetylcholine receptors (nAChRs) are widely expressed in the brain and are implicated in a variety of physiological processes. There are two stoichiometries of the ␣4␤2 nAChR, (␣4) 2 (␤2) 3 and (␣4) 3 (␤2) 2, with different sensitivities to acetylcholine (ACh), but their pharmacological profiles are not fully understood. Methyllycaconitine (MLA) is known to be an antagonist of nAChRs. Using the two-electrode voltage clamp technique and ␣4␤2 nAChRs in the Xenopus oocyte expression system, we demonstrate that inhibition by MLA occurs via two different mechanisms; that is, a direct competitive antagonism and an apparently insurmountable mechanism that only occurs after preincubation with MLA. We hypothesized an additional MLA binding site in the ␣4-␣4 interface that is unique to this stoichiometry. To prove this, we covalently trapped a cysteine-reactive MLA analog at an ␣4␤2 receptor containing an ␣4(D204C) mutation predicted by homology modeling to be within reach of the reactive probe. We demonstrate that covalent trapping results in irreversible reduction of ACh-elicited currents in the (␣4) 3 (␤2) 2 stoichiometry, indicating that MLA binds to the ␣4-␣4 interface of the (␣4) 3 (␤2) 2 and providing direct evidence of ligand binding to the ␣4-␣4 interface. Consistent with other studies, we propose that the ␣4-␣4 interface is a structural target for potential therapeutics that modulate (␣4) 3

(␤2) 2 nAChRs.
The neuronal nicotinic acetylcholine receptors (nAChRs) 4 are a family of ligand-gated ion channels that mediate fast synaptic transmission between cells both in the central and peripheral nervous system (1). These receptors are a pentameric protein assembly that can be formed by up to 12 different subunits in mammals (␣2-10 and ␤2-4) in different combinations, leading to a wide variety of individual subtypes, each with its own pharmacology (2). Of these receptors, the heteromeric ␣4␤2 nAChRs are the most widely expressed in the brain and are implicated in processes as diverse as cognition, consciousness, mood, nicotine addiction, and nociception among others (1). The homomeric ␣7 nAChR is not limited to expression in the central nervous system and is implicated in processes that include the cholinergic anti-inflammatory pathway and schizophrenia (3,4).
Each subunit within the pentameric nAChR complex contains a large extracellular domain, four transmembrane segments (M1-4) of which the second (M2) lines the channel pore, two short loops M1-M2 and M2-M3 that mediate signal transduction, and a large intracellular loop M3-M4 that contributes to channel conductance (1,2). Acetylcholine (ACh) binds at the interface of two subunits in the extracellular domain, which results in a series of conformational changes, culminating in the rotation of the physically distant M2 region that forms the channel pore. Thus, a conserved central hydrophobic leucine residue that is thought to physically occlude the channel pore in the closed state is moved out of the way to allow passive ion flow across the membrane (5)(6)(7).
Considerable structural information has been gained from the x-ray crystal structures of related bacterial (8 -10) and invertebrate (11) ligand-gated ion channels, soluble ACh-binding proteins (AChBPs) (12,13), and the electron microscopy structure of the muscle nAChR from Torpedo marmorata (6,14). Furthermore, the similarities in the structure at the ACh binding site between the ␣7, ␣4␤2 nAChRs and AChBPs have allowed researchers to use the AChBPs as a structural template to create homology models of receptor subtypes. This has been combined with protein-ligand docking studies and molecular dynamics simulations to understand and explain differences in ligand selectivity (15)(16)(17)(18). Recently the co-crystallization of a range of partial agonists and antagonists of ␣7 and ␣4␤2 nAChRs with the Lymnaea stagnalis and Aplysia californica AChBP has revealed ligand-receptor interactions in detail (19 -22).
Although the ␣7 and ␣4␤2 nAChRs share similar gene sequence, there are functional differences including ion selectivity, gating kinetics, and pharmacological profile (1,2). Furthermore, the ␣4␤2 nAChR is known to exist in two stoichiometries, (␣4) 2 (␤2) 3 and (␣4) 3 (␤2) 2 , which respond to ACh with different sensitivities thought to be mediated through binding to the ␣4-␣4 interface that is present in the (␣4) 3 (␤2) 2 stoichiometry, in addition to the consensus ␣4-␤2 agonist binding site (23)(24)(25)(26)(27). The binding site located in the ␣4-␣4 subunit interface has not been thoroughly investigated in terms of its binding to nAChR pharmacological agents and the functional effect these agents have on the receptor. Recently, it has been suggested that the selective binding of ligands to the ␣4-␣4 site could serve to allosterically modulate the receptor, positively or negatively influencing the response to ACh, similar to the benzodiazepine binding site in the GABA A receptors (26,28). These studies join a body of emerging evidence that pharmacological agents can exert their effects by binding to alternative subunit interfaces, distinct from the ␣-␤ interface containing the consensus agonist binding site (28 -30).
In this study we sought to determine the binding interactions of the known nAChR antagonist methyllycaconitine (MLA) at the ␣4␤2 nAChR. By measuring the inhibition of ACh responses by MLA on ␣4␤2 nAChRs, we demonstrate that binding of MLA reduces ACh activation in an apparently insurmountable manner. Using an MLA analog (MLA-maleimide) that is reactive to cysteine and by changing the ratio of subunits injected into the oocyte cell expression system, we demonstrate that MLA-maleimide is covalently trapped at an introduced cysteine (␣4(D204C)) on the complementary side of the ␣4-␣4 interface. We propose that MLA binds to the ␣4-␣4 site to inhibit ACh-mediated activation of the receptor. This interface and amino acid (␣4[D204]) has not been previously identified to interact with MLA.

Mutagenesis and Transcription of Recombinant ␣7 and ␣4␤2
Receptors-DNA encoding for the rat ␣4 subunit previously subcloned into the plasmid pSP64, rat ␤2 subunit subcloned into the plasmid pSP65, and the rat ␣7 subunit subcloned into the pBS SK(ϩ) vector were generous gifts from Professor Jim Boulter, University of California, Los Angeles, CA. DNA encoding for the hRIC-3 chaperone protein was a generous gift from Professor Millet Treinin, Hebrew University of Jerusalem. Sitedirected mutagenesis was performed to create cysteine substitutions using the QuikChange II site-directed mutagenesis kit (Stratagene, La Jolla, CA). Single colonies of Escherichia coli containing plasmid DNA were isolated, grown, and then the DNA-purified and sequenced to identify that the mutations were successfully incorporated.
␣7 DNA was linearized using SmaI, ␣4 was linearized with EcoRI, and ␤2 was linearized with HindIII before transcription. The ␣7 and RIC-3 mRNAs were transcribed in vitro using T7 mMessage mMachine TM transcription kit, and ␣4 and ␤2 mRNAs were transcribed using the SP6 mMessage mMachine TM transcription kit (Ambion Inc., Austin, TX). ␣7 mRNA was polyadenylated using the poly-A-tailing kit (Ambion). RNA was treated with DNase before purification, and RNA concentrations were measured by spectrophotometry using the Nanodrop (Thermo Fisher Scientific). RNA was mixed to the ratios stated in the text for ␣4␤2 nAChR expression.
Xenopus laevis Surgery, Oocyte Extraction, and Injection-The experiments were performed with animal ethics approvals from The University of Sydney. Female X. laevis was anesthetized with tricaine (850 mg/500 ml). Several ovarian lobes were surgically removed by a small incision on the abdomen. The X. laevis were allowed to recover from the surgery, and the time interval between surgical procedures on each frog was 12 months. Three recoverable surgeries were performed on each X. laevis, and a final terminal surgery was performed with the frog exposed to a lethal dose of anesthetic. The lobes were cut into small pieces and were rinsed thoroughly with oocyte-releasing buffer 2 (OR2 (Roche Applied Science); 82.5 mM NaCl, 2 mM KCl, 1 mM MgCl 2 , 5 mM HEPES (hemi-Na ϩ )). The lobes were digested with collagenase A (2 mg/ml in OR2) at room temperature. The oocytes were further washed with OR2 and stored in ND96 wash solution (96 mM NaCl, 2 mM KCl, 1 mM MgCl 2 , 1.8 mM CaCl 2 , 5 mM HEPES (hemi-sodium salt) supplemented with 2.5 mM sodium pyruvate and 0.5 mM theophylline) until ready for injection. Stage V-VI oocytes were selected and microinjected with 2 ng of mRNA in 50.6 nl of nuclease-free water. After injection, the oocytes were maintained at 18°C in the presence of ND96 wash solution augmented with 50 g/ml kanamycin.
Electrophysiological Recording of Recombinant Receptors-Whole-cell currents were measured using a two-electrode voltage clamp with a Geneclamp 500B amplifier together with a Powerlab/200 (AD Instruments, Sydney, Australia) and Chart Version 3.5 for PC as previously described (31). The recording microelectrodes were filled with 3 M KCl and had resistances between 0.2 and 1 megaohms. 1-5 days post-injection oocytes held at Ϫ60 mV (unless otherwise stated) were used for recording. To prevent the activation of Ca 2ϩ -activated chloride currents, oocytes were superfused with calciumfree buffer (115 mM NaCl, 2.5 mM KCl, 1.8 mM BaCl 2 , 10 mM HEPES, pH 7.4) while recording until a stable base current was reached.
Concentration-response curves for agonist were constructed by measuring the peak current responses elicited from a range of agonist concentrations (10 nM-10 mM). Responses were normalized to the maximum current (I/I max ) to compare data from different oocytes. Oocytes were washed for 5 min between agonist applications for the ␣7(L9ЈT), 10 min for the ␣4␤2, and 20 min for the ␣7 and ␣4(D204C)␤2 receptors, to ensure receptors were not desensitized from the previous agonist application. To evaluate the mode of inhibition, ACh concentration-response curves were determined in the presence of MLA at the concentrations given under "Results." Where the curve was determined with preincubation, MLA was applied for 3 min before the solution was switched to contain the same concentration of MLA with ACh. Inhibitory concentration-response curves to MLA were determined with co-application of a range of MLA concentrations and the EC 50 concentration of ACh for the appropriate nAChR. Peak responses were normalized to the EC 50 current (I/I 50 ).
Rate of Reaction-MLA-maleimide was synthesized as previously described (32). The rate of covalent trapping of the reactive MLA-maleimide by the cysteine residue was measured by application of an EC 50 concentration of ACh to the receptor at regular 20-min intervals. Before the application of 1 M reactive MLA compound, MLA-maleimide, the peak currents (I 50 ) were required to be within 10% of each other. Once this was achieved, MLA-maleimide was applied for 30 s at 2.5 min post ACh application followed by ACh (EC 50 ) after a 17-min wash. This sequence was repeated until there was no longer a reduction in the peak currents to ACh.
Concentration-Response Curves-The amplitude of each current response to ACh (I) was normalized to the amplitude of the maximum current response to ACh (I max ). Normalized concentration-response curves were constructed and analyzed using GraphPad "Prism" version 5.0 according to Equation 1, where [A] is the ligand concentration, and n H is the Hill slope. Mean parameters of each curve were derived from at least three oocytes and are stated under "Results." Unless otherwise stated, the 95% confidence interval derived from the curve fitting is given in brackets.
Rate of MLA-maleimide Reaction to Introduced Cysteine-The change in response to an EC 50 concentration of ACh after cumulative time of MLA-maleimide addition (t) was expressed relative to the response to an EC 50 concentration of ACh before reagent addition at t ϭ 0. The data expressed in this way were fitted to a single exponential decay to obtain an estimate of the first order rate constant (k) and final current (I ∞ ) according to Equation 2, These were compared with the wild type using a Student's t test.
␣4-␣4 nAChR Homology Model-Homology models of the rat extracellular domain ␣4-␣4 subunit interface was prepared with MODELLER Version 9.10 (33) essentially as previously described (26) with the following changes: (i) the human sequence for ␣4 was substituted by the rat sequence (entry P09483 at the Protein Knowledgebase (34) as aligned by T-COFFEE Version 9.02 (35), (ii) the AChBP crystal structure with MLA in the binding site (22) (PDB code 2BYR) was substituted for the previously used structure with epibatidine, (iii) loop C was modeled only on the AChBP structure with MLA, and (iv) MLA was included in the homology models instead of nicotine. The best of 100 models according to the DOPE score (36) included in MODELLER were selected. The mutagenesis wizard in PyMOL (The PyMOL Molecular Graphics system, Version 1.5.0.1 Schrödinger LLC, New York) was used to change Asp-204 to a cysteine and select the rotamer placing the thiol closest to the maleimide reactive site. Next, a bond was added between the sulfur of the cysteine and the C4 (CH) of MLA-maleimide. The structure with the covalently bound MLA-maleimide was pre-processed with default settings and subjected to a restrained minimization with an root mean square deviation cutoff of 0.5 Å using the Protein Preparation Wizard of the Schrödinger Suite 2012 (Schrödinger Suite 2012 Epik Version 2.2, Impact Version 5.7, and Prime version 2.3, Schrödinger LLC, New York) to relax the strain caused by the added covalent bond.

RESULTS
MLA Inhibition Is Surmountable at ␣7 nAChRs but Is Apparently Insurmountable at ␣4␤2 nAChRs-MLA is reported to be highly selective for the ␣7 nAChR compared with other receptor subtypes (37,38). To demonstrate this with the rat nAChR proteins, we performed a concentration-response curve to ACh for ␣7 and ␣4␤2 and then performed an inhibition concentration-response curve of MLA to the approximate EC 50 concentrations of ACh (100 and 300 M, respectively; Fig. 1, Table 1). As MLA and ACh would be expected to have different binding kinetics, we preincubated with MLA for 3 min to ensure that the MLA was at equilibrium before co-application with ACh ( Fig. 1). After a 17-min washout of the MLA, the current elicited by ACh was unchanged, demonstrating that the MLA inhibition was reversible for both subtypes. The IC 50 was ϳ1000-fold higher for ␣4␤2 (200 nM) than for ␣7 (140 pM). The rapid desensitization of the ␣7 nAChR can lead to erroneous measurement of peak currents without a fast-application system, as receptors may be in a desensitized state before the ACh solution, reaching equilibrium in the bath. To account for this, we introduced a pore mutation ␣7(L9ЈT) that abolishes rapid desensitization and increases the sensitivity to ACh. The EC 50 of ACh was decreased (␣7 ϭ 92 M, ␣7(L9ЈT) ϭ 1.3 M), but the IC 50 of MLA was unchanged (␣7 ϭ 140 pM, ␣7[L9ЈT] ϭ 150 pM) ( Fig. 1, Table 1), demonstrating that this mutation alters the intrinsic gating properties of the receptor but not the binding of MLA.
To determine the mode of inhibition, increasing concentrations of ACh were applied in the presence of the IC 50 concentration of MLA. In the ␣7(L9ЈT) receptor, the maximum current elicited by ACh in the absence of MLA could be reached in the presence of MLA if high enough concentrations of ACh were applied such that the inhibition was surmountable (Fig. 2, Table 2).
When increasing concentrations of ACh were co-applied with an IC 50 concentration of MLA to the ␣4␤2 nAChR, the maximum current that was elicited in the presence of MLA was significantly lower than that with ACh alone, but the EC 50 was not significantly different ( Fig. 2, Table 2). We repeated this experiment with a range of MLA concentrations, demonstrating that even at low MLA concentrations (Ͼ30 nM) the maximum ACh response was significantly different with the EC 50 unchanged, consistent with insurmountable inhibition (Fig. 2, Table 2). At the highest concentration of MLA applied, 1 M, the maximum current was significantly reduced by 78% and the EC 50 was significantly reduced, albeit by only 1.8-fold (p ϭ 0.041, Student's t test).
Mode of Antagonism Is Altered by MLA Preincubation-To further investigate the mode of inhibition of the ␣4␤2 nAChR, we compared the inhibition with and without a 3-min preincubation of MLA. Without preincubation, the IC 50 of MLA was significantly increased (IC 50 ϭ 1.5 M no incubation, cf. 200 nM with incubation, p Ͻ 0.05, Fig. 3, Table 2). When increasing concentrations of ACh were co-applied with MLA, the maximum current elicited was not significantly changed compared with the maximum current in the absence of MLA, such that the inhibition was surmountable (Fig. 3, Table 2). This result suggests that the insurmountable mode of inhibition depends on the binding of MLA to the receptor having reached equilibrium.
As MLA contains a tertiary amine, it would be expected to be charged at physiological pH and may exhibit voltage-dependent pore-block. To determine whether MLA binds to the channel lumen or to an alternative site leading to insurmountable inhibition, we determined a current-voltage (I-V) curve   with preincubation of MLA and demonstrated that there was no change in the relative inhibition of ACh-induced currents by MLA at various holding potentials (Fig. 4). This observation suggests that MLA does not block the pore of the channel in the transmembrane region, in contrast to other simple analogues of MLA that inhibit ␣4␤2 nAChRs (31). We instead focused on the subunit interfaces of the nAChR that do not contain the classical ␣4-␤2 ACh binding site. Due to the heteromeric nature of the ␣4␤2 nAChR, the alternative interfacial binding sites depend on subunit stoichiometry (Fig.  5). In addition to the ␣4-␤2 interfaces, the (␣4) 3 (␤2) 2 nAChR contains ␣4-␣4 and ␤2-␣4 interfaces, whereas the (␣4) 2 (␤2) 3 nAChR contains ␤2-␣4 and ␤2-␤2 interfaces. Given that irreversible inhibition was observed for the (␣4) 3 (␤2) 2 nAChR stoichiometry, the unique ␣4-␣4 interface suggested itself as a possible site of MLA binding. The high affinity for MLA binding to the ␣7-␣7 interface of the ␣7 nAChR also prompted investigation of the structurally related ␣4-␣4 interface present in the (␣4) 3 (␤2) 2 nAChR. In addition, recent investigations (26,27) have demonstrated that occupation of the ␣4-␣4 interface modulates the agonist response of the (␣4) 3 (␤2) 2 nAChR stoichiometry.
To experimentally investigate MLA binding at the ␣4-␣4 interface, we elected to employ covalent trapping, a methodology that uses cysteine mutagenesis in combination with synthetically derived reactive probes and computational models to interrogate the location and nature of ligand binding at functional nAChRs expressed in Xenopus oocytes (31,32). To this end we manipulated the MLA molecule such that it contained a reactive maleimide group suitable for covalent trapping with cysteine mutants (Fig. 6). Maleimides react rapidly, selectively, and irreversibly with thiolate-bearing residues such as the amino acid cysteine even in the presence of other nucleophiles such as lysine (39). We previously demonstrated that MLAmaleimide reacts covalently to ␣7 nAChR cysteine mutants in a time-dependent manner (32). In the ␣7 nAChR, the cysteine mutation that reacted with MLA-maleimide was located in loop F. Because this loop is highly flexible and in the ␣4 subunit is different in sequence and length compared with the ␣7 subunit, the corresponding residue in ␣4 could not be reliably identified from a sequence alignment alone. Thus, to determine the optimal position for the cysteine to react with MLA-maleimide in the ␣4-␣4 interface, we constructed a dimeric ␣4-␣4 homology model (see "Experimental Procedures") with MLA in the binding site. The residue located closest to the MLA-succinimide moiety was aspartate 204 on the complementary face of the binding site. To test the feasibility of covalent bond formation, we introduced the cysteine in this position computationally, created the bond to the MLA-maleimide moiety, and relaxed the structure with a root mean square deviation cutoff of 0.5 Å on all heavy atoms. The small relaxation required of the complex suggested that the covalent bond could form between the introduced cysteine and MLA-maleimide without marked changes in the inhibitor binding mode (Fig. 5).  The generation of the MLA-maleimide-reactive probe involves the formal removal of two hydrogen atoms to afford the unsaturated maleimide unit. This small structural change was not expected to significantly change the potency of the ligand. To demonstrate this, we performed an experiment on oocytes expressing (␣4) 3 (␤2) 2 nAChRs, where the ACh-elicited current was measured. After a 10-min washout period, 100 nM MLA-maleimide was preincubated for 3 min then co-applied with 100 M ACh, then the same procedure was repeated with 100 nM MLA (Fig. 6). There was no significant difference in the inhibition of ACh-elicited currents with either 100 nM MLA or 100 nM MLA-maleimide (I/I ACh ϭ 0.57 Ϯ 0.07 100 nM MLA, 0.57 Ϯ 0.08 100 nM MLA-maleimide; p Ͼ 0.05, paired Student's t test). This experiment was repeated on oocytes expressing (␣4) 2 (␤2) 3 nAChRs. There was no significant difference in currents elicited by 1 M ACh when inhibited by 100 nM MLA compared with 100 nM MLA maleimide (I/I ACh ϭ 0.44 Ϯ 0.03, 100 nM MLA; 0.45 Ϯ 0.02, 100 nM MLA-maleimide; p Ͼ 0.05, paired Student's t test). This demonstrates that MLA-maleimide inhibits either receptor stoichiometry with the same potency as MLA (Fig. 6).
Effect of Cysteine Mutations to Receptor Function-The target ␣4(D204C) mutation was introduced using site-directed mutagenesis. To ensure that the ␣4(D204C) mutation did not cause any gross changes to receptor function, we performed concentration-response curves to ACh and inhibition concentration-response curves to MLA for a 1:1 and 1:10 injection ratio. There was a small shift (ϳ5-fold reduction in EC 50 compared with wild type) in the concentration-response curve to ACh for the ␣4(D204C)␤2 nAChR injected in a 1:1 ratio, with an EC 50 of 26 M. When a 1:10 injection ratio was evaluated, the EC 50 of ACh (EC 50 ϭ 1.9 M) was not significantly different from wild type (p Ͼ 0.05, Fig. 3, Table 1). When the inhibition curve to MLA was performed for the 1:1 ratio, the IC 50 was 27 nM, ϳ10-fold more potent than the wild-type ␣4␤2 nAChR. When a 1:10 injection ratio was performed, the IC 50 of MLA (IC 50 ϭ 120 nM) was not significantly different from wild type (p Ͼ 0.05; Fig. 3, Table 1). These results suggest that the introduction of the cysteine mutation did not make gross changes to the structure that might interfere with receptor function and that the potency of MLA to the mutants was similar to the wild type for either stoichiometry. 2 nAChR-Based on the ␣4-␣4 homology model with MLA in the binding site, we hypothesized that MLA-maleimide would react with the introduced cysteine to form a covalent bond, irreversibly reducing the current elicited by ACh (Fig. 5). We tested this hypothesis by applying 10 M ACh to oocytes injected with ␣4(D204C) and wild-type ␤2 subunits in a 1:1 ratio to predominantly express the (␣4[D204C]) 3 (␤2) 2 stoichiometry and then re-applied ACh to ensure the current was reproducible. We then applied 1 M MLA-maleimide for 30 s and washed with buffer for 17 min to remove any MLA within the perfusion system. We then applied 10 M ACh again and measured the reduction in the ACh-elicited current before MLA-maleimide addition (Fig. 7). This MLA-maleimide concentration was ϳ10-fold greater than the IC 50 for MLA. A similar protocol for MLA IC 50 concentration has previously been employed to covalently trap MLA-maleimide in the binding site of ␣7 receptor cysteine mutants (32). After 30 s of incubation with the MLA-maleimide, the ACh-elicited current was reduced by 51 Ϯ 0.08% (mean Ϯ S.E., n ϭ 5) (Fig. 7). As MLAmaleimide was not present during the ACh application, we conclude that the inhibition in current was due to an irreversible reaction of the maleimide group with the introduced cysteine at D204C.

MLA-maleimide Is Covalently Trapped at the (␣4[D204C]) 3 (␤2)
To illustrate that the maleimide group was required for the irreversible inhibition, we evaluated whether MLA bound to the receptor could inhibit the reaction between the cysteine mutant and MLA-maleimide. 10 M ACh was applied to oocytes injected with a 1:1 ratio of ␣4(D204C) and ␤2 mRNA at least twice to ensure a consistent response. 10 M MLA was applied for 3 min to equilibrate MLA binding followed by coapplication of 10 M MLA and 1 M MLA-maleimide for 30 s. After a 17-min washout to remove MLA and MLA-maleimide, 10 M ACh was re-applied, and the response was compared with 10 M ACh before MLA and MLA-maleimide application. After a 3-min preincubation of MLA (10 M) followed by a co-application of MLA (10 M) and MLA-maleimide (1 M), the inhibition by MLA-maleimide was significantly attenuated, and the response to ACh (10 M) after a 17-min washout was unchanged relative to the initial response to ACh (I/I ACh ϭ 1.08 Ϯ 0.08, p Ͻ 0.01; paired Student's t test). This competition experiment demonstrates that MLA binds to the (␣4[D204C]) 3 (␤2) 2 nAChR and blocks the reaction of MLA-maleimide at the cysteine residue, presumably by preventing MLA-maleimide binding to the nAChR.
To measure the full extent of the reaction of MLA-maleimide to the (␣4[D204C]) 3 (␤2) 2 receptor, we performed a rate of reaction experiment. We applied 10 M ACh to oocytes injected with ␣4(D204C) and wild-type ␤2 subunits in a 1:1 ratio to predominantly express the (␣4[D204C]) 3 (␤2) 2 stoichiometry and then re-applied ACh to ensure the current was reproducible. We then applied 1 M MLA-maleimide for 30 s and washed with buffer for 17 min to remove any MLA within the perfusion system. We then applied 10 M ACh again and measured the reduction in the ACh-elicited current before MLA-maleimide addition (Fig. 7). This process was repeated until the current reached a plateau. The decrease in the maximum current was plotted against the cumulative time and fitted with an exponential decay. The maximum current inhibition was 82% (I ∞ ϭ 0.18 Ϯ 0.04, n ϭ 5), and the first-order decay constant was k ϭ 0.03 s Ϫ1 (Fig. 7, Table 3). This was significantly different from the wild-type ␣4␤2 nAChR, where MLA-maleimide applica-tion and wash inhibited the ACh-elicited current by 38% (I ∞ ϭ 0.62 Ϯ 0.12, n ϭ 4, p Ͻ 0.01) after 120 s incubation, and experiments from individual oocytes did not fit a single exponential decay curve (Fig. 7, Table 3). Incubation of MLA (1 M for 30 s) alone followed by a 17-min wash to the (␣4[D204C]) 3 (␤2) 2 receptors under identical conditions to MLA-maleimide resulted in a 17% (I ∞ ϭ 1.05 Ϯ 0.1, n ϭ 3, p Ͻ 0.01) reduction of ACh-elicited currents after 120 s of cumulative incubation and did not fit a single exponential decay curve (Fig. 7, Table 3). The reduction in current observed when MLA-maleimide was applied to the wild type is most likely a nonspecific effect of the moderately high concentrations of ligand applied, as incubation of 1 M MLA for a cumulative 120 s followed by a 17-min wash to the wild-type receptor led to a similar reduction in current (I ∞ ϭ 0.61 Ϯ 0.06, n ϭ 4, p Ͼ 0.05, Table 3).
The reactive probe MLA-maleimide was applied to the (␣4[D204C]) 2 (␤2) 3 nAChR in oocytes derived from a 1:10 injection ratio (Fig. 6, Table 3). We applied 1 M ACh to oocytes injected with ␣4(D204C) and wild-type ␤2 subunits in a 1:10 ratio, to predominantly express the (␣4[D204C]) 2 (␤2) 3 stoichiometry, and then re-applied ACh to ensure the current was reproducible. We then applied 1 M MLA-maleimide for 30 s and washed with buffer for 17 min to remove any MLA within the perfusion system. We then applied 1 M ACh again and measured the reduction in the ACh-elicited current before MLA-maleimide addition (Fig. 7). After 30 s of incubation with the MLA-maleimide, the ACh-elicited current was reduced by 11 Ϯ 0.02% (mean Ϯ S.E., n ϭ 3) (Fig. 7). This was significantly different to the (␣4[D204C]) 3 (␤2) 2 stoichiometry (p Ͻ 0.01),   a Where data from individual oocytes did not fit to an exponential curve, the I ∞ is defined as the I 120 . b Individual oocytes could not be fitted to an exponential curve fit. demonstrating that MLA-maleimide was not reacting to the introduced cysteine in the (␣4[D204C]) 2 (␤2) 3 .

DISCUSSION
We have demonstrated that MLA-maleimide, a reactive analog of MLA, is covalently trapped at an introduced cysteine (D204C) on the complementary face of the ␣4 subunit only in the (␣4) 3 (␤2) 2 stoichiometry where an ␣4-␣4 interface is present, unequivocally demonstrating that MLA binds at the ␣4-␣4 interface. Covalent trapping of a reactive analog at an introduced cysteine has previously been used to integrate computational techniques such as homology modeling and molecular dynamics, with functional analysis of the trapped ligand-receptor complex (29,40). Here we have used this method for a different purpose, to reveal the interfacial binding site of a ligand in a heteropentameric nAChR.
There are four structurally distinct subunit interfaces within the two known stoichiometries of the ␣4␤2 nAChR, two of which, ␣4-␣4 and ␣4-␤2 contain ACh binding sites (26 -27). The reactive MLA-maleimide is only trapped by the (␣4[D204C]) 3 (␤2) 2 nAChR cysteine mutant where the receptor contains an ␣4-␣4 interface, demonstrating unambiguously that MLA-maleimide and by extension MLA also binds at this site. No covalent trapping was observed for the (␣4[D204C]) 2 (␤2) 3 nAChR that contains two ␤2-␣4 interfaces modified by cysteine but no ␣4-␣4 site. Preincubation of the (␣4) 3 (␤2) 2 nAChR with MLA results in insurmountable inhibition of Ach-induced currents. The covalent trapping experiments in this work may implicate the ␣4-␣4 binding site in this mode of receptor antagonism. Indeed, the binding of MLA to the ␣4-␣4 interface should not be surprising given the high affinity of the structurally related ␣7-␣7 interface of the ␣7 nAChR for MLA. Furthermore, the elegant work of Palma et al. (41) has demonstrated that binding of a single MLA molecule to the ␣7 nAChR is sufficient to observe receptor inhibition despite the fact that five ␣7-␣7 interfaces are available for MLA binding. In the current context, binding of a MLA molecule to the ␣4-␣4 binding site displayed in the (␣4) 3 (␤2) 2 may also be sufficient to observe insurmountable inhibition.
Recent investigations by Harpsøe et al. (26) and Mazzaferro et al. (27) have demonstrated that binding of ACh at the ␣4-␣4 interface modulates the response to ACh in the (␣4) 3 (␤2) 2 nAChR stoichiometry. The observation that MLA and ACh can both bind at the ␣4-␣4 interface appears contradictory to our pharmacological data; MLA and ACh would be expected to compete at the ␣4-␣4 binding site. Thus, the inhibition by MLA would be surmountable with high enough concentrations of ACh and parallel shifts of the ACh concentration-response curves in the presence of different concentrations for MLA would be expected (42). Under our experimental conditions, this was not observed (Fig. 2). Limitations in our experimental conditions, specifically ACh-induced desensitization of receptors and the osmotic effects of molar concentrations of ACh, prevent applying ACh in higher concentrations (Ͼ10 mM) to determine if the inhibitory effects of MLA are in fact surmountable. These complexities may contribute to the absence of a parallel shift in the concentration-response curves and/or the insurmountable inhibition.
Preincubation with MLA was performed to ensure the binding of MLA to the nAChRs was in equilibrium before ACh application, which along with a parallel shift in the concentration-response curves is a necessary precondition for the proper analysis of a Schild plot to characterize competitive inhibition (42). The preincubation was required to observe maximal insurmountable inhibition of the ␣4␤2 nAChR in this study, which suggests a slow onset of MLA binding to the ␣4-␣4 interface. Similar observations have been made for human ␣7 nAChRs where preincubation has been demonstrated to increase the potency of MLA inhibition of JN403-induced currents (43). In our covalent trapping study, the pseudo first order rate constant for binding and reaction of MLA-maleimide (1 M) at the (␣4[D204C]) 2 (␤2) 3 was k ϭ 0.038 Ϯ 0.002 s Ϫ1 , providing an estimated second-order rate constant of k ϭ 3.8 Ϯ 0.2 ϫ 10 4 M Ϫ1 s Ϫ1 . The pseudo first-order reaction rate constant for the binding and reaction of the same reactive probe MLA maleimide (10 nM) to the ␣7(S188C:L9ЈT) nAChR of k ϭ 0.0058 Ϯ 0.0014 s Ϫ1 provides an estimated second-order rate constant of k ϭ 5.8 Ϯ 1.4 ϫ 10 5 M Ϫ1 s Ϫ1 . This is comparable to the k on ϭ 2.04 ϫ 10 6 M Ϫ1 s Ϫ1 for binding of [ 3 H]MLA to the ␣7 nAChR from rat membranes, which serves as an upper limit for the rate of the covalent trapping (44). The lower estimated second-order rate constant for covalent trapping at the (␣4) 3 (␤2) 2 receptor subtype may in part reflect slower binding to the ␣4-␣4 interface. The surmountable mode of inhibition that we demonstrate for the ␣4␤2 nAChR in the absence of preincubation may also be a function of the relative time course of antagonist binding with respect to ACh binding. A slower time course of MLA binding would result in lower concentrations of ACh required to fully occupy the three agonist binding sites if the ligands were co-applied rather than the impractically high concentrations of ACh thought to be required if the binding of MLA was allowed to reach equilibrium before ACh application.
The covalent trapping experiment resulted in the ␣4(D204C)␤2 receptor eliciting ϳ14% of the current compared with the current before the reaction. It is tempting to speculate that this is a result of a mixed population of receptor stoichiometries, leading to some current elicited by (␣4[D204C]) 2 (␤2) 3 receptors that do not trap the inhibitor. However, we have previously demonstrated that when MLA-maleimide is trapped in the ␣7-␣7 interface by the analogous ␣7(S188C:L9ЈT) receptor, we also do not see full inhibition, suggesting that the reaction does not go to completion or that bound MLA does not always covalently trap in the binding site (32). It is, therefore, not reliable to use this cysteine-trapping method as a proxy measure of different stoichiometric populations.
There is emerging evidence that ligands can exert pharmacological effects by binding to alternative subunit interfaces of heteromeric nAChRs at locales distinct from the ␣-␤ interface containing the consensus agonist binding site (28 -30). These studies include an x-ray co-crystallographic study of the binding of non-competitive inhibitors and allosteric modulators to the AChBP coupled with radioligand binding studies of these inhibitors and modulators to AChBPs with mutations that resemble subunit interfaces that do not contain a principal ␣ subunit (29), the application of the substituted cysteine accessibility method to identify the binding site of the binding of the positive allosteric modulator morantel at the ␤3-␣2 interface of the ␣2␤3 nAChR (30), and a photo-affinity labeling study of the positive allosteric modulators physostigmine and galantamine at the Torpedo californica muscle-type nAChR that identified three binding sites, including an ␥-␤ subunit interfacial site (28). Furthermore, these non-agonist binding sites resemble the now well established modulatory benzodiazepine binding site on GABA A receptors.
In this study we have employed covalent trapping to directly reveal MLA binding to the ␣4-␣4 interface in functional (␣4) 3 (␤2) 2 nAChRs, the first time that ligand binding at the ␣4-␣4 interface has been directly demonstrated. The identification of this new ligand binding site and the delineation of the pharmacological profile in functional receptors open new avenues for development of therapeutics by targeting this site.
There has been a concerted effort to identify modulators of nAChRs with therapeutic benefits. Agonists selectively targeting ␣4␤2 or ␣7 nAChRs have been demonstrated to improve cognitive function in various learning and memory paradigms (45) as well as possess analgesic properties (46). Positive allosteric modulators only act in the areas of the brain where ACh was released, preserving the spatial and temporal activation by the endogenous neurotransmitter (46). The (␣4) 3 (␤2) 2 -selective positive allosteric modulator NS9283 increases the peak maximal current in response to ACh and augments AChevoked Dh␤E-sensitive currents in thalamocortical neurons and improves or alleviates pharmacological deficits in a wide variety of paradigms, including sensory gating, vigilance, attention, episodic, and reference memory (47). The selectivity exhibited by NS9283 for the (␣4) 3 (␤2) 2 nAChR stoichiometry implicates binding to the ␣4-␣4 interface. However, there is little understanding of the structural rearrangements after binding at the ␣4-␣4 site that lead to positive or negative modulation, an understanding of which has the potential to lead to the identification of superior therapeutics. Our results suggest that MLA analogues may assist in addressing this question and also serve as lead compounds for ␣4-␣4-targeted therapeutics, one that could be identified by a combination of molecular modeling, in silico docking, and functional electrophysiological techniques.
In conclusion, we have demonstrated for the first time that MLA binds to the ␣4-␣4 subunit interface of the ␣4␤2 nAChR through covalent-trapping of a modified MLA molecule. MLA binding at the ␣4-␣4 interface leads to an apparently insurmountable mode of inhibition. We also demonstrated that trapping molecules to an introduced cysteine in individual subunits can identify otherwise ambiguous binding sites in ligandgated ion channels. We propose that the ␣4-␣4 binding site can be targeted to produce more selective pharmacological agents that can then be used to better understand the role of the different stoichiometries of the ␣4␤2 nAChR.