An Intersubunit Hydrogen Bond in the Nicotinic Acetylcholine Receptor That Contributes to Channel Gating*

The muscle nicotinic acetylcholine receptor is a large, allosteric, ligand-gated ion channel with the subunit composition α2βγδ. Although much is now known about the structure of the binding site, relatively little is understood about how the binding event is communicated to the channel gate, causing the pore to open. Here we identify a key hydrogen bond near the binding site that is involved in the gating pathway. Using mutant cycle analysis with the novel unnatural residue α-hydroxyserine, we find that the backbone N-H of αSer-191 in loop C makes a hydrogen bond to an anionic side chain of the complementary subunit upon agonist binding. However, the anionic partner is not the glutamate predicted by the crystal structures of the homologous acetylcholine-binding protein. Instead, the hydrogen-bonding partner is the extensively researched aspartate γAsp-174/δAsp-180, which had originally been identified as a key binding residue for cationic agonists.

agonist molecules induces a conformational change that leads to the opening of the ion channel.
It is now widely appreciated that a tryptophan residue (␣Trp-149) plays a key role in neurotransmitter binding by forming a cation-interaction with the quaternary ammonium group of acetylcholine (4), a result supported by structural data. However, many other important residues in the immediate vicinity of the binding site have been identified. In a classic experiment on the Torpedo nAChR (a close homologue of the muscle subtype), Czajkowski and Karlin (5) concluded that a key aspartate (␥Asp-174/␦Asp-180) from the complementary binding subunit could come within 9 Å of the agonist binding site. Mutation of this residue severely impacted receptor function, leading to a proposal that the negative charge of this aspartate interacted with the positive charge of the agonist (6,7). Subsequently, however, both the crystal structure of acetylcholinebinding protein (AChBP, a soluble protein homologous to the extracellular domain of the nAChR) (8) and the 4 Å cryo-electron microscopy structure of Torpedo nAChR (9,10) showed that this residue is positioned quite far from the agonist binding site (Fig. 1). Single-channel studies suggest that this residue is primarily important for ligand-induced channel gating rather than agonist binding (11,12).
␥Asp-174/␦Asp-180 is part of loop F, the most remote of the six loops originally proposed by Changeux and co-workers (1) to contribute to the agonist binding site. In the carbamylcholine-bound AChBP structure (13), a different F loop anionic residue, ␥Glu-176/␦Glu-182 (AChBP Glu-163), is positioned near loop C of the agonist binding site (Fig. 1). Specifically, ␥Glu-176/␦Glu-182 is within hydrogen-bonding distance of the backbone N-H at ␣Ser-191, which is located on the C loop between the aromatic binding box residue ␣Tyr-190 and the vicinal disulfide formed by ␣Cys-192 and ␣Cys-193. Loop F is generally disordered in the AChBP and nAChR structures, and hydrogen-deuterium exchange mass spectrometry on AChBP reveals that loop F and loop C are the most conformationally dynamic segments of the protein (14). It is generally accepted that agonist binding draws loop C inward, capping the aromatic binding pocket (13,15). In contrast, crystal structures of AChBP with antagonists bound reveal that loop C pulled away from the agonist binding site (15). Distinctions between antagonist-and agonist-induced motion have also been observed in loop F (16). As such, many investigators favor a gating model involving a contraction of the agonist binding site around an agonist molecule that is largely mediated by movements in loops C and F, although little is known about the nature of the specific interactions involved in this systolic motion.
In the present work, we evaluate several anionic residues in loop F and potential interactions with loop C. Using a combination of natural and unnatural mutagenesis, we find that, indeed, the backbone N-H of ␣Ser-191 does make a hydrogen bond to a loop F residue. However, the partner is not the glutamate seen in the AChBP crystal structure. Instead, the aspartate (␥Asp-174/␦Asp-180) originally identified by Czajkowski and Karlin (5) is the hydrogen-bonding partner.

EXPERIMENTAL PROCEDURES
Preparation of Unnatural Amino and Hydroxy Acids-All chemical reactions were performed under argon using solventcolumn dry solvents (17). Flasks and vials were oven-dried at 122°C and cooled in a desiccator box containing anhydrous calcium sulfate. Silica chromatography was carried out in accordance with the methods of Still et al. (18). The preparations of nitrohomoalanine (Nha), 2-amino-4-ketopentanoic acid (Akp), and ␣-hydroxyalanine (Aah) have been described previously (19 -21).
Side Chain and Backbone Mutagenesis-Conventional mutagenesis and unnatural mutagenesis, with the site of interest mutated to either an amber stop codon or a four-base frameshift codon (at ␣Ser-191), were achieved by a standard Stratagene QuikChange protocol. Sequencing through both strands confirmed the presence of the desired mutation. Mouse muscle embryonic nAChR in the pAMV vector was used. All the mutations were made in the presence of a background transmembrane mutation (␤L9ЈS) that lowers whole-cell EC 50 (24,25). In addition, the ␣-subunits contain a hemagglutinin epitope in the M3-M4 cytoplasmic loop for Western blot studies. Control experiments show that this epitope does not detectably alter EC 50 . mRNA was prepared by in vitro runoff transcription using the Ambion (Austin, TX) T7 mMessage mMachine kit. For conventional mutants, a total of 2.0 -4.0 ng of mRNA was injected in a ratio of 2:1:1:1 of ␣:␤:␥:␦. For suppression with unnatural amino and hydroxy acids, a total of 4.0 ng of mRNA was injected in an ␣:␤:␥:␦ subunit ratio of 10:1:1:1. Typically, 25 ng of tRNA was injected per oocyte along with mRNA in a ratio of 1:1 with a total volume of 50 nl/cell. As a negative control for suppression, truncated 74-nucleotide tRNA or truncated tRNA ligated to deoxycytosine adenosine was co-injected with mRNA in the same manner as fully charged tRNA. Data from experiments where currents from these negative controls were greater than 10% of the experimental were excluded. Frameshift suppression at ␣Ser-191 was used for simultaneous incorporation of two unnatural residues (26).
Electrophysiology-The function of mutant receptors was evaluated using two-electrode voltage clamp (see supplemental materials). Stage V-VI oocytes of Xenopus laevis were employed. Oocyte recordings were made 12-48 h after injec- tion in two-electrode voltage clamp mode using the Opus-Xpress 6000A instrument (Axon Instruments, Union City, CA). Oocytes were superfused with a Ca 2ϩ -free ND96 solution at flow rates of 1 ml/min before application, 4 ml/min during drug application, and 3 ml/min during wash. Holding potential was Ϫ60 mV. Data were sampled at 125 Hz and filtered at 50 Hz. Drug applications were 15 s in duration. Acetylcholine chloride was purchased from Sigma-Aldrich/RBI. Solutions ranging from 0.01 to 5000 M were prepared in Ca 2ϩ -free ND96 from a 1 M stock solution. Dose-response data were obtained for a minimum of eight concentrations of agonists and for a minimum of five cells. Dose-response relations were fitted to the Hill equation to determine EC 50 and Hill coefficient values. The doseresponse relations of individual oocytes were examined and used to determine outliers. The reported EC 50 values are from the curve fit of the averaged data.

A Backbone Mutation at ␣Ser-191 Has a Large Impact on
Receptor Function-In the process of probing the backbone flexibility surrounding the nAChR binding box, we mutated ␣Ser-191 to Sah. This produces an ester backbone linkage at this position, while preserving the side chain. Along with increasing backbone flexibility, this mutation removes the hydrogen bond-donating N-H group and replaces it with a nonhydrogen bond-donating O (Figs. 2 and 3). We have performed similar backbone mutations at several positions throughout the nAChR, and typically, the consequences are not dramatic (19,27). However, at ␣Ser-191, this subtle mutation leads to a 40-fold increase in EC 50 (Table 1 and Fig. 2B). We also made alanine and Aah mutations at this site. The side chain mutation alone had minimal impact on receptor function, producing no shift in EC 50 . However, receptor function of the ␣S191Aah mutant was dramatically impaired relative to ␣S191Ala, con-firming that the backbone, and not the side chain, at ␣Ser-191 is important for receptor function.
Evaluating Anionic Residues on Loop F-To evaluate the potential hydrogen bond from ␣Ser-191 to ␥Glu-176/␦Glu-182, we made several mutations at this glutamate. Surprisingly, all mutations have minimal impact, suggesting no critical role for this residue. Another nearby loop F glutamate residue,  ␥Glu-163169/␦Glu-163175, was also evaluated. Again, no significant impact was seen.
Fundamentally, the results of these conventional mutations strongly implicate the side chain of ␥Asp-174/␦Asp-180 in an electrostatic interaction, such as an ion pair or hydrogen bond. However, changing the side chain functionality from a carboxylate to an amide not only neutralizes the charge on the side chain, it also desymmetrizes it and introduces a potential hydrogen bond donor. To better understand the role of ␥Asp-174/␦Asp-180, we incorporated two unnatural amino acids.
Nha is an analogue of glutamate that contains a nitro (NO 2 ) group, which is isosteric and isoelectronic to a carboxylate, but it has no charge and is a much weaker hydrogen bond acceptor ( Fig. 2A). 6 Incorporation of Nha at ␥Asp-174/␦Asp-180 yields a slightly less dramatic effect than the ␥D174N/␦D180N and ␥D174Q/␦D180Q mutants, producing a 24-fold shift in EC 50 relative to ␥D174E/␦D180E (Fig. 2B). Thus, charge neutralization at this site significantly affects receptor function but cannot fully account for the EC 50 shift seen in the ␥D174N/ ␦D180N and ␥D174Q/␦D180Q mutants.
The second unnatural amino acid analogue incorporated at ␥Asp-174/␦Asp-180 was Akp. Akp is isoelectronic to Asp but possesses a methyl ketone functionality in place of the carboxylate ( Fig. 2A). As such, Akp is a desymmetrized analogue of Asp, similar to Asn, but it has different electrostatic properties (less polar, weaker hydrogen bond acceptor, and cannot donate a hydrogen bond). When Akp is incorporated at ␥Asp-174/ ␦Asp-180, its effect upon receptor function is roughly as deleterious as the Asn mutation (Fig. 2B). Mutations at individual binding sites (␣␤␥D174N␦ and ␣␤␥␦D180N) showed substan-tial (ϳ50-fold) and approximately equivalent increases in whole cell EC 50 (supplemental Table 1).
Mutant Cycle Analysis Reveals a Strong Interaction between the ␣Ser-191 Backbone and the Side Chain of ␥Asp-174/␦Asp-180-Mutant cycle analysis was performed between several of the side chain mutations at ␥Asp-174/␦Asp-180 and the ␣S191Sah mutation (Figs. 3 and 4). Briefly, mutant cycle analysis is used to determine the pairwise interaction energy between two residues in a protein using the equation given in Fig. 3. If the two residues do not interact, the change in free energy for the simultaneous mutation of both residues should simply be the sum of the free energy of each of the individual mutations. However, for residues that interact, the change in free energy for the double mutation will be non-additive. EC 50 -based mutant cycle analysis has been used to investigate interactions in Cys loop receptors by other researchers (30 -32). Further discussion can be found in the supplemental materials.

DISCUSSION
The AChBP crystal structures transformed the study of nAChRs by providing high resolution structural data about the ligand binding domain of these proteins. In addition to refining our existing structural knowledge of the receptor, obtained from decades of careful biochemical research, it served as a valuable starting point for new structure-function studies on the receptor. However, AChBP is not a ligand-gated ion channel and, in fact, shares less than 25% homology with its nearest Cys loop relative, the ␣7 nAChR (8). As such, some fundamental structural differences must exist between AChBP and actual Cys loop receptors, particularly pertaining to residues involved in mediating communication between the binding site and the ion channel pore.
Arguably the biggest discrepancy between the AChBP structures and prior biochemical studies of the nAChR concerns the loop F residue ␥Asp-174/␦Asp-180. A remarkable cross-linking study in the Torpedo nAChR indicated that this residue can come within 9 Å of the vicinal disulfide on the C loop, located at the heart of the agonist binding site (5). However, in AChBP, the residue that aligns with ␥Asp-174/␦Asp-180 is not at all near the agonist binding site. In addition, the cryo-electron microscopy structure of the Torpedo nAChR, which is believed to be in the desensitized or closed state, places this residue tucked deeply inside a ␤-sheet-lined hydrophobic pocket, over 15 Å from loop C (9, 10).
Although ␥Asp-174/␦Asp-180 is remote to the agonist binding site in AChBP, another loop F anionic residue, ␥Glu-176␦Glu-182, appears to make a hydrogen bond to a backbone N-H that is integral to the aromatic box of the agonist binding site when the agonist carbamylcholine is bound. Using nonsense suppression methodology, we have been able to specifically remove the backbone N-H in question by replacing ␣Ser-191 with its ␣-hydroxy analogue. Consistent with the AChBP images, this subtle structural change has a large effect on receptor function, suggesting that a hydrogen bond to this moiety is important. However, consistent with prior mutational analyses (6,7,28), we find that ␥Glu-176/␦Glu-182 does not play a large role in receptor function. This suggests that AChBP does not provide an accurate model of the muscle nAChR in this region. Given that the sequence alignment in this region shows a number of insertions in nAChR relative to AChBP, combined with the fact that the F loop is believed to be involved in gating the nAChR (AChBP does not gate), it is not surprising that AChBP would be an unreliable model here.
Our results indicate that the hydrogen-bonding partner for the backbone N-H of ␣Ser-191 in the nAChR is instead the side chain of ␥Asp-174/␦Asp-180. Based on the available structural and functional data, we suggest that this hydrogen bond exists in the open state only (11,28,29). As others have seen, a number of mutations of this side chain profoundly affect receptor function. Here we employ several relatively subtle mutations. The fact that substantial functional consequences are seen suggests a precise structural role for this side chain in at least one state crucial for activating the channel. Furthermore, all mutations at ␥Asp-174/␦Asp-180 that significantly impact function also show strong coupling to the ␣S191Sah backbone mutation via mutant cycle analysis. The nature of the coupling is as one would expect from the hydrogen-bonding model. Mutation at either site has a strong effect; however, once the ␣S191Sah mutation is introduced, removing any possible hydrogenbonding interaction, mutations at ␥Asp-174/␦Asp-180 have a much smaller impact. Interestingly, no specific role for the side chain of ␣Ser-191 is found as the ␣S191A mutant gives essentially wild-type behavior. 7 However, when the alanine side chain is combined with the ␣-hydroxy backbone mutation, the same coupling to ␥Asp-174/␦Asp-180 is observed.
We propose that the movement of loop F of the complementary subunit from a position remote to the agonist binding site to one of close proximity to loop C of the principal subunit is a key early structural change associated with nAChR gating. Driving this structural reorganization is the formation of a hydrogen bond between the side chain of ␥Asp-174/␦Asp-180 and the backbone N-H of ␣Ser-191. In the closed state of the wild-type receptor, ␣Ser-191 and the C loop project out into solution, away from the bulk of the receptor, whereas ␥Asp-174/␦Asp-180 of the F loop projects deep within a hydrophobic cavity. Although the energetic desolvation penalty of burying a charged residue within a hydrophobic cavity is significant, 8 it alone is apparently not sufficient to overcome other structural elements that bias this conformation of the F loop. However, agonist binding induces a centripetal movement of the C loop, bringing the backbone N-H of ␣Ser-191 in closer proximity to the F loop. This structural change makes possible a hydrogen bond between ␥Asp-174/␦Asp-180 and the C loop backbone. We hypothesize that the formation of this hydrogen bond, along with the energetic solvation benefit of moving ␥Asp-174/ ␦Asp-180 into an aqueous environment, provides sufficient driving force to move ␥Asp-174/␦Asp-180 out of its pocket, inducing a movement of the F loop toward the C loop. 9 This structural rearrangement of loop F contributes to the gating pathway. Using rate-equilibrium free energy relationships, Auerbach and co-workers (34) have also concluded that ␥Asp-174/␦Asp-180 moves early in the gating process.
Our results provide an explanation for the cross-linking studies of Czajkowski and Karlin (5), and they are not in conflict with available structural information. The cryo-electron microscopy images of the Torpedo receptor show greater than 20 Å separation between ␣Ser-191 and ␥Asp-174/␦Asp-180 in the closed state. Still, the distance of less than 9 Å that is suggested by the cross-linking studies is plausible, provided that the residues are free to move closer, as occurs in our model of channel gating. A comparison of AChBP structures with and without agonist bound likewise shows that motions of loops C and F are the dominant structural rearrangements that occur when agonist binds. The fact that the hydrogen bond acceptor in loop F differs between AChBP and nAChR is not surprising, given that loop F is strongly implicated in nAChR gating and AChBP did not evolve to have a gating function.
This model is also consistent with the ␥174/␦180 mutants described here. Lengthening the side chain of this key F loop residue (␥D174E/␦D180E) effects a modest improvement in receptor function, either because the longer side chain can more easily reach its ␣Ser-191 hydrogen-bonding partner or because it fits more poorly in the hydrophobic pocket, destabilizing the closed state. Any mutation that eliminates side chain charge has a significant impact on function, which is expected, given that these mutant side chains are poorer hydrogen bond acceptors and that they experience a much lower energetic solvation benefit upon moving from the hydrophobic pocket into an aqueous environment.
In conclusion, mutant cycle analysis involving a novel backbone mutant has identified an important interaction between an F loop residue that has long been thought to contribute to receptor function and the peptide backbone of loop C. The hydrogen bond between the side chain of ␥Asp-174/␦Asp-180 and the backbone of ␣Ser-191 likely forms upon agonist binding and is part of the agonist-induced conformational changes that lead to channel opening. Along with contributing new insights into the gating pathway of the nAChR, our results reconcile a long-standing discrepancy between early biochemical studies of the receptor and structural models from the AChBP systems.