Molecular actions of smoking cessation drugs at α4β2 nicotinic receptors defined in crystal structures of a homologous binding protein

Abstract

Partial agonists of the α4β2 nicotinic acetylcholine receptor (nAChR), such as varenicline, are therapeutically used in smoking cessation treatment. These drugs derive their therapeutic effect from fundamental molecular actions, which are to desensitize α4β2 nAChRs and induce channel opening with higher affinity, but lower efficacy than a full agonist at equal receptor occupancy. Here, we report X-ray crystal structures of a unique acetylcholine binding protein (AChBP) from the annelid Capitella teleta, Ct-AChBP, in complex with varenicline or lobeline, which are both partial agonists. These structures highlight the architecture for molecular recognition of these ligands, indicating the contact residues that potentially mediate their molecular actions in α4β2 nAChRs. We then used structure-guided mutagenesis and electrophysiological recordings to pinpoint crucial interactions of varenicline with residues on the complementary face of the binding site in α4β2 nAChRs. We observe that residues in loops D and E are molecular determinants of desensitization and channel opening with limited efficacy by the partial agonist varenicline. Together, this study analyzes molecular recognition of smoking cessation drugs by nAChRs in a structural context.

Tobacco smoking is a major cause of premature death worldwide. To a large extent, the reinforcing effects of nicotine result from the direct activation of neuronal α4β2 nicotinic acetylcholine receptors (nAChRs), which triggers downstream events such as increased dopamine release in the mesolimbic system (1–4). Therefore, the α4β2 nAChR subtype has become a key target for development of therapeutic agents for smoking cessation (5, 6). In particular, high-affinity partial agonists for α4β2 nAChRs, such as lobeline (7), cytisine (8), and varenicline (8), are molecules of interest because they exhibit the unique property of acting as a mixed agonist/antagonist. High-affinity varenicline binding competes with nicotine at the α4β2 nAChR and, thereby, antagonizes the reward sensation of smoking (9, 10). However, varenicline activates α4β2 nAChRs with low efficacy and desensitizes, leaving channels unopened even at high binding occupancy, thereby minimizing withdrawal symptoms and increasing the success rate of smoking cessation attempts (9, 10). Understanding the molecular mechanism of partial agonism at the nAChR has great potential for designing novel smoking cessation compounds but, as yet, is hampered by the absence of the high-resolution structure of a eukaryote nicotinic receptor.

Current structural insight is derived from medium-resolution electron microscopy images of the nAChR from Torpedo marmorata (11), X-ray structures from prokaryotic nAChR homologs GLIC (12, 13) and ELIC (14), the Caenorhabditis elegans glutamate-activated chloride channel GluCl (15) and an X-ray structure from the monomeric mouse nAChR α-subunit extracellular domain in complex with α-bungarotoxin (16). However, significant progress has been made since the discovery of water-soluble pentameric acetylcholine binding proteins (AChBPs) from snails and the subsequent elucidation of their high-resolution ligand-bound crystal structures. AChBPs are homologs of the extracellular domain of nAChRs and are the best-studied structural models of ligand recognition by the extracellular domain of nAChRs (17–19).

In this study, we report the high-resolution X-ray crystal structures of an AChBP from the marine worm Capitella teleta (20), a unique nonmolluscan AChBP, in complex with two partial agonists, lobeline and the smoking cessation aid varenicline. These structures offer detailed insight into high-affinity interactions of lobeline and varenicline in the binding pocket of Ct-AChBP. We then used structure-guided mutagenesis and electrophysiological recordings of α4β2 nAChRs to reveal the key interactions of varenicline with the complementary face of the binding pocket that account for receptor desensitization and channel opening with limited efficacy. By defining structural determinants of these molecular actions at the α4β2 nAChR, we open avenues for the rational design of smoking cessation aids.

Results

Pharmacological Characterization of Ct-AChBP.

In this study, we characterized the pharmacological properties of an AChBP from the marine worm Capitella teleta, which is a unique nonmolluskan AChBP (20). To investigate the validity of Ct-AChBP as a potential model for the nAChR, we compared the binding properties of varenicline and various other ligands with two well-described molluskan AChBPs, namely AChBP from Aplysia californica (Ac-AChBP) and Lymnaea stagnalis (Ls-AChBP). Using competitive binding assays with ³H-epibatidine, we calculated K_i values for these ligands (Table 1). When Ct-AChBP is compared with the other AChBPs, we conclude that it has a distinct pharmacological profile. Ct-AChBP shows a low affinity for acetylcholine and a high affinity for nicotine. The affinity of varenicline for Ct-AChBP is more than 10-fold higher than for Ac-AChBP and is close to the high affinity for Ls-AChBP. The high affinity of lobeline and varenicline for Ct-AChBP resembles the high-affinity binding to α4β2 nAChR (7 ± 1 nM for lobeline and 0.18 ± 0.01 nM for varenicline; Table 1). Lobeline and varenicline both display partial agonist action at α4β2 nAChR (7, 9). The pharmacology of α4β2 nAChRs is complex because these receptors can occur in alternate stoichiometries and contain ligand binding sites with high and low sensitivity (21). In addition, these receptors can occur in different allosteric conformations and display cooperativity upon ligand binding, which are properties lacking in AChBPs. Despite these limitations, our binding data demonstrate high-affinity binding of lobeline and varenicline and suggest that Ct-AChBP is a useful addition to the family of AChBPs as a suitable model for structural studies.

Table 1.

Binding properties of Ct-AChBP in comparison with Ac-AChBP, Ls-AChBP, and α4β2 nAChRs

Ligand	Ct-AChBP	Ac-AChBP	Ls-AChBP	α4β2 nAChR	Ligand action
Lobeline	0.29 ± 0.06	1.2 ± 0.4	1,000 ± 64	7 ± 1	Partial agonist
Epibatidine	4.6 ± 0.3	11.4 ± 0.7	2.5 ± 0.2	0.05 ± 0.01	Agonist
Varenicline	5 ± 1	72 ± 8	3 ± 1	0.18 ± 0.01	Partial agonist
Strychnine	6 ± 1	38 ± 3	223 ± 26	819 ± 90	Antagonist
d-tubocurarine	140 ± 6	509 ± 38	171 ± 18	2 ± 1	Antagonist
Cytisine	247 ± 31	206 ± 52	219 ± 66	0.9 ± 0.1	Partial agonist
Methyllycaconitine	358 ± 66	15 ± 4	15 ± 6	ND	Antagonist
Nicotine	423 ± 59	583 ± 84	1,110 ± 231	6 ± 2	Partial agonist
α-conotoxin ImI	> 1 mM	4 ± 2	> 1 mM	ND	Antagonist
Levamisole	4.6 ± 0.8 μM	6.0 ± 0.4 μM	27 ± 3 μM	ND	Agonist
Acetylcholine	57 ± 6 μM	35 ± 5 μM	29 ± 3 μM	28 ± 4 μM	Agonist
Serotonine	171 ± 76 μM	319 ± 143 μM	893 ± 364 μM	8 ± 3 μM	Antagonist
GABA	> 1 mM	> 1 mM	> 1 mM	—	—

All measurements are in nanomolars unless otherwise noted. Binding constants (K_i values) were determined by using a competitive binding assay with ³H-epibatidine (n ≥ 3). ND, not determined.

X-Ray Crystal Structures of Ct-AChBP in Complex with Lobeline and Varenicline.

The lobeline-bound structure of Ct-AChBP (1.9 Å, crystallographic details in Table S1) is shown along the fivefold symmetry axis with the C terminus of the protein pointing toward the viewer in Fig. 1A and away from the viewer in Fig. 1B. The Ct-AChBP monomer has an architectural fold that closely resembles Ac-AChBP and can be superimposed with an r.m.s. deviation of 1.6 Å for 190 Cα atoms. Difference electron density at the subunit interfaces unambiguously revealed the occupancy of lobeline at all five binding sites (Fig. 1A, Inset). In addition, we observed electron density for an N-linked glycosyl chain at N122 (side view in Fig. 1C and Inset), which likely corresponds to paucimannose commonly found in insect cells (22) and partially wraps around the water-accessible part of the binding pocket at the subunit interface. Treatment with PNGase had no significant effect on the K_i value of nicotine, suggesting that the glycosyl chain attached to N122 does not affect the pharmacological properties of Ct-AChBP (Fig. S1). The varenicline-bound structure of Ct-AChBP (2.0 Å) is shown in Fig. 1D. The Ct-AChBP structures in complex with varenicline and lobeline are almost identical, and pentamers can be superimposed with an r.m.s. deviation of 0.4 Å for 1,053 Cα atoms.

Fig. 1.

X-ray crystal structures of AChBP from Capitella teleta (Ct-AChBP). Cartoon representation of Ct-AChBP in complex with lobeline, seen along the fivefold symmetry axis with the N terminus pointing away from the viewer (A) and pointing toward the viewer (B). Each of the five subunits is displayed in a different color. Lobeline molecules are shown as spheres. Oxygen, red; nitrogen, blue; sulfur, green. Insets show a detail of 2F_o-F_c electron density contoured around lobeline (ball-and-stick representation) at a σ level of 1.5. (C) Side view of Ct-AChBP in complex with lobeline with a subunit interface positioned toward the viewer. The principal face is shown in yellow; the complementary face is shown in blue. Lobeline molecules are shown as spheres, and the glycosyl chain N-linked to N122 is shown in magenta ball-and-stick representation. Inset shows a detail of 2F_o-F_c electron density contoured around the glycosyl chain at a σ level of 1.0. The paucimannose chain is composed of N-acetylglucosamine (NAG), α-d-mannose (α-MAN), and β-d-mannose (β-MAN). (D) Cartoon representation of Ct-AChBP in complex with varenicline, in the same orientation as Ct-AChBP in A. Varenicline is shown as spheres, and Inset shows a detail of 2F_o-F_c electron density contoured around varenicline (ball-and-stick representation) at a σ level of 2.0.

The architectural fold of the annelid Ct-AChBP is very similar to molluskan AChBPs, but there are notable structural differences in certain loops and β-strands. A structure-based sequence alignment of Ct-AChBP and Ac-AChBP is shown in Fig. S2A. Cartoon representations with magnified insets of superimposed monomers are shown for Ct-AChBP (magenta) and Ac-AChBP (yellow) in Fig. S2B. Notable differences between the two structures (highlighted in blue boxes) include a longer turn between the N-terminal α-helix and the β1-strand, F27-L28-D29, and the Cys-loop and β1-β2 loop, which form part of the interface between the transmembrane domain in integral Cys-loop receptors. The Cys loop in Ct-AChBP comprises only 8 amino acids (C137–C146), compared with 12 amino acids in molluskan AChBPs and 13 in membrane integral Cys-loop receptors. The β1-β2 loop in Ct-AChBP is 4 amino acids longer than in Ac-AChBP (M53–N56). Finally, Ct-AChBP has different conformations in the β5 strand, which lines the pentamer vestibule, and loop F, which forms part of the ligand binding site. Although subtle differences exist, these data show that the overall protein architecture is well conserved despite the relatively low sequence identity between Ct-AChBP and Ac-AChBP. Next, we investigated whether some of the structural differences at the ligand binding site could explain the observed changes in pharmacological properties in Ct-AChBP compared with other AChBPs (Table 1).

Lobeline and Varenicline Recognition in the Ct-AChBP Binding Pocket.

The ligand binding site of Ct-AChBP is found at the interface between two subunits and ligands interact with amino acids from both the principal face (containing loops A, B, and C) and complementary face (containing loops D, E, and F). Lobeline contacts residues from the principal and complementary face of the binding pocket in Ct-AChBP resembling interactions in the complex with Ac-AChBP (Fig. 2 A and B and Table S2). In the complex of Ct-AChBP, the tertiary amine of the piperidine ring of lobeline is stabilized through cation–π interactions with conserved aromatic residues of the principal face, including W153 (loop B) and Y201 (loop C). Similar to the complex in Ac-AChBP (23), lobeline exposes a subpocket in Ct-AChBP by changing the rotameric state of F102 (loop A) from the g- to t-conformation (homologous to Y91 in Ac-AChBP; ref. 24). Lobeline forms four hydrogen bonds that converge upon residues of loop B (shown as dashed lines in Fig. 2A). The lobeline hydroxyl group interacts with the S152 and W153 backbone oxygens. The tertiary amine of the piperidine ring binds with the W153 backbone oxygen, and the lobeline carbonyl oxygen interacts with the W153 indole nitrogen. The interactions on the complementary face are predominantly hydrophobic and include contacts with I118, L126, and I128 (loop E). The lobeline carbonyl oxygen also interacts with the backbone nitrogen of I128 through a water molecule that occupies the same position as observed in the lobeline complex with Ac-AChBP (Fig. 2B).

Fig. 2.

Molecular recognition of lobeline. (A) Amino acid interactions of lobeline in Ct-AChBP. The principal face is shown in yellow; the complementary face is shown in blue. Side chains are shown in ball-and-stick representation (yellow for principal face; pink for the complementary face). Dashed lines indicate hydrogen bonds. A water molecule between lobeline and I128 is represented as a red sphere. (B) Amino acid interactions of lobeline in Ac-AChBP (PDB ID code 2BYS). (C) Lobeline adopts different configurations in the Ct-AChBP and Ac-AChBP pocket. The configuration in Ct-AChBP (yellow) corresponds to α-lobeline, whereas the configuration in Ac-AChBP (white) is the opposite stereochemical configuration.

A remarkable difference between Ct-AChBP and Ac-AChBP binding pockets is the lack of the highly conserved aromatic residue in loop D (Y53 in Ac-AChBP and I64 in Ct-AChBP). In addition, superposition of Ct-AChBP and Ac-AChBP binding pockets reveals that the lobeline molecule adopts different configurations in Ct-AChBP and Ac-AChBP (Fig. 2C). In Ct-AChBP, the lobeline piperidine ring adopts a chair conformation, versus a half-chair conformation in Ac-AChBP. In addition, the hydroxyl group is in the (S)-configuration in Ct-AChBP, versus the (R)-configuration in Ac-AChBP. As a result, the lobeline configuration observed in Ct-AChBP corresponds to the α-lobeline configuration, whereas the configuration observed in Ac-AChBP is the opposite stereochemical configuration. These results demonstrate that subtle differences exist in the binding pocket architecture of Ct-AChBP and other AChBPs, which might account for different ligand binding properties (Table 1). In addition, lobeline is sufficiently flexible to adapt to subtle differences in the AChBP binding pocket architecture and suggests it might also do so in different Cys-loop receptors.

Interactions of varenicline with residues of the principal and complementary binding site in Ct-AChBP are shown in Fig. 3A and compared with interactions of nicotine (25) [Protein Data Bank (PDB) ID code 1UW6; Fig. 3B] and carbamylcholine (25) (PDB ID code 1UV6; Fig. 3C) in complex with Ls-AChBP (Table S2). In the X-ray crystal structure of Ct-AChBP, varenicline forms interactions on the principal face of the binding site with aromatic amino acids, which are highly conserved among different nAChR subtypes. On the complementary face of the binding site, varenicline forms mostly hydrophobic interactions with residues that are weakly conserved. In Ct-AChBP, these residues on the complementary face of the binding site are I118, L126, and I128 in loop E and I64 in loop D (Fig. 3A). Interestingly, varenicline also exposes a subpocket in Ct-AChBP by changing the rotameric state of F102 (loop A) from the g- to t-conformation, which we also observed in the lobeline-bound structure (F102 is homologous to Y89 in Ls-AChBP; Fig. 3 B and C). Also similar to lobeline, varenicline forms cation–π interactions and multiple hydrogen bonds that converge on residues of loop B, one of which involves a water molecule that forms a bridge between the varenicline benzazepine nitrogen and the W153 carbonyl oxygen. A second water molecule forms hydrogen bonds between the varenicline pyrazino nitrogen and the backbone nitrogen and oxygen of I128 and Q116, respectively. Remarkably, this water molecule occupies a position that is almost identical to the water molecule observed between nicotine and residues L102 and M114 in the complex with Ls-AChBP, and the water molecule in the lobeline complexes with Ct-AChBP (Fig. 2A) and Ac-AChBP (Fig. 2B). This result indicates that lobeline, varenicline, and nicotine, which act as partial agonists at α4β2 nAChRs, share a common molecular recognition that involves water molecules acting as a bridge with residues of the binding pocket. In contrast, carbamylcholine (Fig. 3C), which acts as a full agonist at α4β2 nAChRs, forms interactions with the residues L102, L112, and M114 on the complementary face that are mainly hydrophobic and lack the contribution of water molecules to these interactions. These results confirm that the nature of ligand interactions with residues of the complementary face possibly contribute to discrimination between full and partial agonists in different Cys-loop receptors (26).

Fig. 3.

Molecular recognition of varenicline and comparison with nicotine and carbamylcholine. Amino acid interactions observed for varenicline bound to Ct-AChBP (A), nicotine bound to Ls-AChBP (B; PDB ID code 1UW6), and carbamylcholine bound to Ls-AChBP (C; PDB ID code 1UV6). (D) Superposition and different C loop conformations observed for AChBP complexes with varenicline (blue; Ct-AChBP), nicotine (green; Ls-AChBP) and α-conotoxin ImI (red; Ac-AChBP). Varenicline is shown in white sticks. (E) Residues of the complementary face of Ct-AChBP involved in ligand interactions with varenicline were mapped onto a structure-based sequence alignment of AChBPs from different species that was seeded with human nAChR sequences. Residues are colored in shades of blue by using an identity threshold of 50%. Residues involved in critical contacts with varenicline are indicated in black. Ct, C. teleta; Ac, Aplysia californica; Ls, Lymnaea stagnalis; Bt, Bulinus truncatus; Hs, Homo sapiens. Numbers above the sequence alignment correspond to amino acid numbering for Ct-AChBP; numbers below the alignment correspond to the human β2 nAChR.

We have analyzed conformational changes in loop C upon binding of agonists, partial agonist, and antagonist in more than 30 AChBP crystal structures solved to date (27). Consistent with this analysis, we find that varenicline produces a contraction of loop C (7.79 ± 0.08 Å; Fig. 3D), which is intermediate between full agonists (<7 Å distance) and the extended state observed for antagonists (>11 Å distance) (27). For example, the previously characterized partial agonist tropisetron in complex with Ac-AChBP (26) produces a contraction of loop C of 7.96 ± 0.30 Å (27), which is very similar to varenicline in complex with Ct-AChBP. This result indicates that, similar to other partial agonists cocrystallized with AChBPs, varenicline causes an incomplete contraction of loop C around the binding site, which may contribute to limited efficacy of channel opening in α4β2 nAChRs. Similar to the lobeline complex with Ac-AChBP, we find that lobeline produces an exceptionally strong contraction, usually associated with full agonists, of loop C around the ligand binding site in Ct-AChBP. The distance measured between the carbonyl oxygen of the conserved loop B aromatic residue (W153 in Ct-AChBP and W145 in Ac-AChBP) and the tip of loop C (γS atom of C196 in Ct-AChBP and C188 in Ac-AChBP) is 6.54 ± 0.06 Å in the lobeline complex with Ct-AChBP and 6.74 ± 0.34 Å in the Ac-AChBP complex. Together, these results indicate that for most ligands, the extent of C-loop contraction correlates to their mode of action at nAChRs, but exceptions exist, such as lobeline.

Structure-Guided Mutagenesis Analysis of Varenicline Interactions in Human α4β2 nAChRs.

To investigate the importance of varenicline interactions in α4β2 nAChRs, we used structure-based sequence alignments of AChBPs seeded with human β2 subunit nAChR sequences to identify homologous residues of varenicline contact sites (I118, L126, and I128 in loop E and I64 in loop D) in Ct-AChBP (Fig. 3E). These residues likely contribute to αβ and ββ interfaces in α4β2 nAChRs (21, 28, 29). In the human β2 subunit, these residues on the complementary face of the binding site correspond to V111, F119, and L121 (loop E) and W57 (loop D). Because the smoking cessation action through α4β2 nAChRs has been best described for varenicline (9), we compared relative affinity and efficacy for varenicline and desensitization by varenicline and the full agonist acetylcholine. Using two-electrode voltage-clamp recordings from oocytes, we measured ligand-activated currents of nAChRs composed of wild-type α4 subunits coexpressed with wild-type β2 subunits or four different β2 mutants, namely W57A, V111F, F119A, and L121F. To investigate changes in the volume of the binding pocket and importance of aromaticity, we mutated residues to phenylalanine if nonaromatic and to alanine if aromatic. Examples of current responses evoked by various concentrations of varenicline are shown in Fig. 4 A–D.

Fig. 4.

Mutation of homologous residues in α4β2 nAChRs affects desensitization and efficacy of channel opening. (A) Example traces from oocytes expressing wild-type α4β2 and activated with a fixed concentration of acetylcholine (30 μM) and desensitized by increasing concentrations of varenicline (0.01–300 nM). (B) When applied alone, concentrations of varenicline more than 300 nM cause channel opening, but with low efficacy compared with 1 mM acetylcholine. (C) Desensitization of mutant α4β2 nAChRs, obtained by coexpression of wild-type α4-subunits and W57A β2-subunits, is abolished. (D) The W57A mutation eliminates the relative difference in efficacy of channel opening between the partial agonist varenicline and the full agonist acetylcholine. (E) Bar graph summarizes the effect of mutations on the relative efficacy of channel opening by varenicline normalized against the saturating response with acetylcholine on each mutant α4β2 nAChRs (*P < 0.01). (F) Concentration-activation (solid lines) and concentration-inhibition (dashed lines) curves for varenicline on wild-type (red) and W57A receptors (green). The activation responses are normalized to the maximum response to 1 mM acetylcholine (=1). Inhibition responses are normalized to the half-maximal response to 30 μM acetylcholine (=0.5). (G) Models illustrate characteristic differences between antagonists (red), partial agonists (blue), and full agonists (green). Together with an intermediate contraction of loop C (magenta), the interactions with W57 and L121 (solid lines for varenicline), critically determine the rate of desensitization and limit the efficacy of channel opening by varenicline.

In agreement with previous studies, varenicline activates α4β2 nAChRs with an EC₅₀ value of 1.0 ± 0.3 μM (n = 3–11) and low efficacy (14.7 ± 0.6%, n = 3–11) compared with the full agonist acetylcholine (EC₅₀ = 56.4 ± 9.1 μM, efficacy = 100%, n = 4–10). Consistent with previous studies (9, 30), we observe that varenicline concentrations between 3 and 300 nM desensitize ACh-induced responses of wild-type α4β2 nAChRs (Fig. 4A), without causing significant channel opening when applied alone in this concentration range (9, 30). This effect on desensitization of α4β2 nAChRs is thought to occur at clinically effective concentrations of varenicline and, likely, contributes to smoking cessation relief by reduction of craving during nicotine abstinence (9). In contrast, at concentrations more than 300 nM, varenicline alone produces channel opening with high affinity, but low efficacy relative to the response obtained with a saturating concentration of the full agonist acetylcholine (Fig. 4B). This effect likely reduces nicotine reinforcement during smoking lapses, because varenicline binds with higher affinity to α4β2 nAChRs than nicotine (9).

When wild-type α4 subunits are coexpressed with W57A β2 subunits, we observe that the mutation abolishes varenicline-mediated desensitization of ACh responses of these mutant α4β2 nAChRs (Fig. 4 C and F). This result suggests that the loop D residue W57, which is homologous to I64 in Ct-AChBP, forms a critical interaction with varenicline to mediate transitions to a densensitized state of the α4β2 nAChRs. In addition, we observe that the W57A mutation eliminates the relative difference in efficacy of channel opening compared with the full agonist acetylcholine (Fig. 4 D and F). Three other mutations in α4β2 nAChRs at positions homologous to those involved in varenicline contacts in Ct-AChBP have similar, but less pronounced effects on varenicline efficacy compared with the W57A mutation. L121F produces channel opening by varenicline with an efficacy of 81.9 ± 4.4% of ACh (n = 7–15), for V111F the efficacy is 39.2 ± 0.5% (n = 4–13), and for F119A the efficacy is 39.2 ± 0.5% (n = 4–13) (summarized in Fig. 4E). Remarkably, these point mutations have little effect on the affinity of varenicline compared with the pronounced changes in efficacy (Table S3). W57A (EC₅₀ = 2.0 ± 0.3 μM, n = 4–11) and L121F (EC₅₀ = 2.6 ± 0.5 μM, n = 7–15) result in a small decrease in affinity for varenicline, whereas F119A exhibits a higher affinity (EC₅₀ = 0.6 ± 0.2 μM, n = 4–9). V111F does not significantly alter the affinity of varenicline (EC₅₀ = 1.3 ± 0.1 μM, n = 4–13). Together, these results illustrate the predictive value of the cocrystal structure of Ct-AChBP with varenicline. The pronounced effects of homologous mutations in the β2 nAChR subunit on receptor desensitization and efficacy of channel opening by varenicline indicate that the binding mode in α4β2 nAChRs is likely similar to the one observed in the crystal structure. In particular, the interactions of varenicline with W57 and L121 on the complementary binding site of the β2 nAChR subunit possibly have essential contributions to mediate transitions to a desensitized receptor state and to limit the efficacy of channel opening by the partial agonist varenicline.

Discussion

Partial agonism at pentameric ligand-gated ion channels arises from a fundamental molecular property, which is to trigger channel opening with lower probability than full agonists at equal levels of receptor occupancy. Partial agonists open these channels with low efficacy through a mechanism that is common to different members of the Cys-loop receptor family. Using single channel recordings from the muscle nAChR and GlyR, Lape et al. (31) demonstrated that partial and full agonists have similar kinetics for the closed to open transition of the channel pore, but that the response to partial agonists is limited by an early conformational change, termed flipping, that precedes channel opening. At the structural level, flipping possibly corresponds to an early conformational change after ligand recognition by the extracellular ligand-binding domain, whereas the ion channel domain is still closed, similar to domain closure observed for glutamate receptors (32, 33). Despite these insights, the precise interactions of partial agonists with the receptor binding site are not understood.

From new crystal structures of lobeline and varenicline in complex with Ct-AChBP, together with a comparative analysis of more than 30 existing AChBP structures (27), common structural and functional properties of partial agonists start to emerge. Typically, partial agonists stabilize loop C in a conformation that is intermediate between the fully contracted state observed for full agonists and the extended state observed for antagonists (26, 27). It is likely that low efficiency of flipping by partial agonists for the muscle nAChR and GlyR as described by Lape et al. (31) can be understood in terms of incomplete contraction of loop C as described in crystal structures of AChBP.

In combination with the conformational stabilization of loop C, the detailed molecular interactions of partial agonists with residues of the binding pocket appear equally important. Because most of the residues on the principal subunit are highly conserved among different nAChR subtypes, we focused our attention to the variable residues on the complementary subunit, which likely mediates partial agonist effects of varenicline in certain nAChR subtypes, including the important α4β2 subtype. In the crystal structure of Ct-AChBP in complex with varenicline, a water molecule that occupies a common position in nicotine and lobeline complexes with AChBP plays a key role in establishing hydrogen bonds with the backbone atoms of hydrophobic residues of loop E. In addition, a highly conserved aromatic residue of loop D, W57 in α4β2, also plays a key role in defining the architecture of the complementary subunit. The critical contribution of this residue has been demonstrated in other Cys-loop receptors, including α7 nicotinic receptors (34), GABA_A receptors (35), 5-HT₃ receptors (36–38), and glycine receptors (39). Consistent with earlier observations (26), we took the assumption that these specific contacts of ligand and binding site residues might be of key importance for discriminating agonists from partial agonists. We determined the relative contribution of these interactions to partial agonist effects of varenicline by mutating homologous positions of the complementary subunit at the high-affinity α4/β2 interface in α4β2 nAChRs and comparing the effect of mutations on the efficacy of channel opening relative to the full agonist acetylcholine. We found that mutations W57A (loop D) and L121F (loop E) have the most pronounced effects and almost completely eliminate the difference in efficacy between varenicline and acetylcholine. These results indicate that both interactions with loop D and E contribute to efficiency of channel opening. Less pronounced effects were observed for mutations V111F and F119A in loop E, indicating that these interactions are less critical for this.

The molecular property of varenicline to mediate channel opening of α4β2 nAChRs with high affinity but low efficacy is believed to contribute to the clinical effects during smoking cessation relief because it limits the effects of nicotine during moments of smoking relapse. In addition, at concentrations even lower than those required for channel opening, varenicline also causes pronounced receptor desensitization, an effect that minimizes reward sensation mediated through signaling pathways involving α4β2 nAChRs. We observed that the mutation W57A (loop D) abolishes receptor desensitization of α4β2 at low occupancy of varenicline, indicating that the interaction with W57 is critical in mediating a transition to a desensitized state. Based on our findings, we propose a model (Fig. 4G) in which the molecular action of the smoking cessation aid varenicline relies on (i) incomplete contraction of loop C around the ligand binding site, (ii) critical hydrophilic interactions with residues of loop E, which occur through a water molecule in the ligand binding site, and (iii) interactions with the highly conserved aromatic residue of loop D, which play a pivotal role in mediating transitions to channel states with low open probability.

In summary, we show high-resolution crystal structures of a unique AChBP with structural features and pharmacological properties distinct from previously identified AChBPs. The high-affinity binding of the smoking cessation aids lobeline and varenicline, which act as partial agonists at α4β2 nAChRs, suggests that Ct-AChBP is a useful model to study atomic interactions for this receptor subtype. Using structure-guided mutagenesis, we revealed the importance of key interactions in loop D and E of the α4/β2 subunit interface of α4β2 nAChRs. We identified single point mutations in the binding pocket that abolish receptor desensitization and eliminate the relative difference in efficacy of channel opening by varenicline and the full agonist acetylcholine. In combination with an intermediate contraction of loop C, we suggest that these interactions define the molecular recognition of partial agonists and function as key recognition events that precede channel opening, previously described as flipping. Together, these data provide detailed insight into the molecular mechanism underlying partial agonism at nAChRs. This study opens perspectives for the design of new partial agonists to treat disorders associated with specific Cys-loop receptor subtypes.

Methods

Ct-AChBP was expressed in Sf21 insect cells and purified on nickel Sepharose and Superdex 200 (GE Healthcare). Crystallization trials were carried out by vapor diffusion in the presence of 1 mM α-lobeline hydrochloride (Sigma-Aldrich) or 1 mM varenicline hydrochloride. Details on X-ray crystallography, radioligand binding assays, and electrophysiological recordings are described in SI Methods.