Unraveling amino acid residues critical for allosteric potentiation of (α4)3(β2)2-type nicotinic acetylcholine receptor responses

Neuronal nicotinic acetylcholine receptors (nAChRs) are promising drug targets to manage several neurological disorders and nicotine addiction. Growing evidence indicates that positive allosteric modulators of nAChRs improve pharmacological specificity by binding to unique sites present only in a subpopulation of nAChRs. Furthermore, nAChR positive allosteric modulators such as NS9283 and CMPI have been shown to potentiate responses of (α4)3(β2)2 but not (α4)2(β2)3 nAChR isoforms. This selective potentiation underlines that the α4:α4 interface, which is present only in the (α4)3(β2)2 nAChR, is an important and promising drug target. In this report we used site-directed mutagenesis to substitute specific amino acid residues and computational analyses to elucidate CMPI's binding mode at the α4:α4 subunit extracellular interface and identified a unique set of amino acid residues that determined its affinity. We found that amino acid residues α4Gly-41, α4Lys-64, and α4Thr-66 were critical for (α4)3(β2)2 nAChR potentiation by CMPI, but not by NS9283, whereas amino acid substitution at α4His-116, a known determinant of NS9283 and of agonist binding at the α4:α4 subunit interface, did not reduce CMPI potentiation. In contrast, substitutions at α4Gln-124 and α4Thr-126 reduced potentiation by CMPI and NS9283, indicating that their binding sites partially overlap. These results delineate the role of amino acid residues contributing to the α4:α4 subunit extracellular interface in nAChR potentiation. These findings also provide structural information that will facilitate the structure-based design of novel therapeutics that target selectively the (α4)3(β2)2 nAChR.

Drugs that enhance the function of brain nicotinic acetylcholine receptors (nAChRs) 3 are sought for their potential clinical applications in the management of nicotine dependence (tobacco smoking) and for improvement of cognitive deficits associated with neurological and psychiatric disorders (1). However, it has proven challenging to develop nAChR therapeutics selective for nAChR subtypes, and few are available clinically (2). Neuronal nAChRs are transmitter (ACh)-gated ion channels made up of five homologous or identical subunits (␣2-␣9 and ␤2-␤4), each one consisting of a large N-terminal extracellular domain and a 4-helix bundle transmembrane domain. The recently published X-ray structure of (␣4)2(␤2)3 nAChR confirmed this general subunit topology and provided further insight into the 3D structure of neuronal nAChRs (3). Whereas there are multiple neuronal nAChR subtypes with unique pathophysiological and pharmacological profiles, the ␣4and ␤2-containing nAChR subtypes are of particular interest (4,5). They constitute the majority of brain nAChRs (6) and are the primary nAChR subtypes that mediate the addictive effects of nicotine (7,8). In addition, they are important for neuronal survival and maintenance of cognitive performance and learning during aging (9), and there is a decrease in their number in the brain of Alzheimer's patients (10).
Drugs that enhance signaling through nAChRs include agonists and partial agonists that bind to the orthosteric (ACh) binding sites and directly activate the nAChR and positive allosteric modulators (PAM) that enhance ACh potency and/or efficacy by binding at sites distinct from the ACh-binding sites (11). Many years of effort have defined the ACh-binding sites at subunit interfaces within the extracellular domain and determined structural requirements for nAChR agonists (12). In contrast, compounds with diverse structural features potentiate nAChRs (1), and multiple PAM recognition sites have been proposed. These include pockets at the extracellular canonical (agonist-binding) and non-canonical subunit interfaces (13)(14)(15)(16)(17), within the transmembrane domain in the helix bundle of a single subunit (18), at the interface between adjacent subunits (19), and at the subunit's extracellular C-tail (20 -21). There is an increasing interest in the development of novel nAChR PAMs, as they provide potentially selective modulation of a subpopulation of nAChRs while avoiding the sustained receptor activation and non-physiological alteration in cholinergic transmission seen with agonists (11,22).
The effects of these amino acid substitutions were assessed by recording ACh-induced current responses (Ϯ CMPI, NS9283, or dFBr) of Xenopus oocytes injected with a mix of RNA encoding ␣4 containing the point mutation and WT ␤2 subunit RNA at a ratio of 8:1 to favor the expression of the 3␣4:2␤2 stoichiometry (Fig. 4A). The potentiation for each amino acid substitution was determined as R, the ratio of the peak current amplitude in the presence of 1 M PAM and 10 M ACh (EC 10 ) relative to the peak current amplitude elicited by 10 M ACh alone (Table 1). Consistent with previous reports (17,32), valine substitution at ␣4His-116 (␣4H116V) abolished the enhancement by 1 M NS9283 (R H116V ϭ 1. . Similar effects were seen for CMPI in receptors harboring ␣4H116L or ␣4H116A substitutions (Table 1), which indicates that the position ␣4His-116 is not essential for CMPI or dFBr modulation of (␣4)3(␤2)2 nAChR.

Effect of amino acid substitutions in the transmembrane domain on modulation by CMPI
Within the transmembrane domain, the ␣4:␣4 subunit interface is formed by amino acid residues from the M3 helix of the ␣4 subunit providing the (ϩ) face and residues from the M1 helix of the ␣4 subunit providing the (Ϫ) face as well as amino acid residues from M2 of both subunits. The ␣4 and ␣3 subunits share high amino acid sequence identity within M1-M2-M3, Amino acid numbering is for mature protein according to the X-ray structure of human (␣4)2(␤2)3 nAChRs (PDB accession number 5KXI). Add 26, 25, and 31 to ␣4, ␤2, and ␣3 sequence numbers, respectively, to obtain amino acid numbers starting from the translational N terminus. ␤-Strands in A are color-coded (␤1, cyan; ␤2, red; ␤5, blue; ␤6, green) to match their sequence alignments in B.  Table 1.
nAChRs containing amino acid substitutions at positions ␣4Gly-41, ␣4Lys-64, and ␣4Glu-66, CMPI did not significantly alter the ACh concentration-response curve when co-applied with ACh, and ACh EC 50 in the presence of 1 M CMPI and was not significantly different from that in the absence of CMPI. For (␣4T126L)3(␤2)2 and (␣4T126I)3(␤2)2 nAChRs, co-application of 1 M CMPI slightly potentiated current responses at low ACh concentrations (Ͻ10 M) and slightly decreased current responses at higher ACh concentrations, producing a less steep ACh concentration-response curve but with no significant decrease in the ACh EC 50 . For (␣4Q124F)3(␤2)2 and (␣4Q124T)3(␤2)2 nAChRs, the poorly defined maximum response resulted in a large degree of

modulation of WT and mutants (␣4)3(␤2)2 nAChRs
Current responses to 10 M ACh alone or in the presence of increasing concentrations of CMPI were recorded from oocytes expressing WT and mutants (␣4)3(␤2)2 nAChRs. For each application peak current amplitude was quantified and normalized to peak current amplitude elicited by 10 M ACh alone within the same recording run. Replicas from the same oocyte were averaged and for each CMPI concentration (average (Avg) Ϯ S.E.) of data from several oocytes (N) were plotted (Fig. 5) and fit to Equation 1. The probability (P) that an I max differs from no potentiation (I max ϭ 100) was analyzed using a one-way analysis of variance with the Holm-Sidak test. Curve-fitting, parameters calculation, and statistics were performed in SigmaPlot 11 (Systat Software Inc.).   Table 3.

Subunits
uncertainty in ACh EC 50 calculation, and that precluded determination of any significant effect of CMPI on the ACh EC 50 .
In the course of writing this manuscript, the X-ray structure of the agonist-bound (␣4)2(␤2)3 nAChR was published (5). Table 3 The effect of CMPI on ACh dose-response curve of (␣4)3(␤2)2 nAChRs Current responses to increasing concentrations of ACh (alone and in the presence of 1 M CMPI) were recorded from oocytes expressing WT and mutants (␣4)3(␤2)2 nAChRs. For each application, peak current amplitude was quantified and normalized to peak current amplitude elicited by 1 mM ACh alone within the same recording run. Replicas from individual oocytes were averaged. For each drug application (average Ϯ S.E.), data from several oocytes (N) were plotted (Fig. 6) and fit to Equation 1. The probability (P) that the ACh EC 50 in the presence of CMPI differs from ACh EC 50 in the absence of CMPI was analyzed by t test. Curve-fitting and parameter calculation was performed in SigmaPlot 11 (Systat Software Inc.). One-way analysis of variance of ACh EC 50 values for WT and mutant (␣4)3(␤2)2 nAChRs in the absence of CMPI revealed no statistically significant difference (p ϭ 0.079). Pairwise t test of ACh EC 50 of (␣4K64T)3(␤2)2, (␣4T126L)3(␤2)2, or (␣4T126I)3(␤2)2 versus WT (␣4)3(␤2)2 nAChR revealed p values of 0.39, 0.06, and 0.01, respectively. This structure contains only two ␣4 subunits and lacks the ␣4:␣4 subunit extracellular interface, the structural domain harboring the CMPI and NS9283 binding sites. We generated an (␣4)3(␤2)2 nAChR homology model using protein superim-position, as described under "Experimental Procedures," then we performed computational docking analyses for CMPI and NS9283 in the absence of ACh (Fig. 8). CMPI docking at the ␣4:␣4 subunit extracellular interface revealed two favorable For each ligand, conserved amino acid residues that make up the agonist site aromatic box and amino acid residues that significantly reduced its effects when substituted to the corresponding residue from the ␤2 subunit are shown in Connolly surface representation colored by atom charge (positive, blue; negative, red). Residues are labeled using the single letter code with those from the ␣4 (ϩ) face underlined.
In view of the limitations associated with the prediction of binding modes by computational docking using homology models, it was unexpected that calculations based upon the AChBP and (␣4)2(␤2)3 nAChR structures would result in similar predictions concerning the location of bound CPMI and NS9283. However, both docking analyses predicted close proximity and interactions between CMPI or NS9283 and the amino acid residues that we identified using mutational analyses as molecular determinants for CMPI or NS9283 potentiation of the (␣4)3(␤2)2 nAChR. Furthermore, CMPI docking at the ␣4:␣4 subunit extracellular interface in the absence and presence of ACh predicted higher affinity in the absence than the presence of ACh, and the binding mode in the absence of ACh was more consistent with the effects of amino acid substitutions on CMPI potentiation of (␣4)3(␤2)2 nAChR.
A total of 15 amino acid positions, 10 within the extracellular domain and 5 within the transmembrane domain, were substituted individually to their corresponding amino acid residues in the ␣3 and/or ␤2 nAChR subunits. Within these substitutions, one position (His-116) selectively reduced potentiation by NS9283, whereas three positions (Gly-41, Lys-64, and Glu-66) selectively reduced potentiation by CMPI. Substitutions at two positions (Gln-124 and Thr-126) reduced potentiation by CMPI and NS9283. Amino acid substitution at the other nine positions did not significantly reduce potentiation by CMPI or NS9283, and none of the substitutions reported here had a significant effect on (␣4)3(␤2)2 nAChR potentiation by dFBr.
To gain further insight into the binding modes of CMPI/ NS9283, we performed computational analyses docking CMPI and NS9283 at the ␣4:␣4 subunit extracellular interface in the absence and presence of ACh bound at this interface. Our computational analyses (summarized in Fig. 7 and Fig. 8) predicted stable binding of CMPI and NS9283 at the ␣4:␣4 interface in the absence or presence of ACh but with different amino acid residue contacts. The functional consequences of amino acid substitutions and the predicted interactions of docked CMPI/ NS9283 with these amino acid residues establish that NS9283/ CMPI potentiate (␣4)3(␤2)2 nAChR by replacing ACh at the ␣4:␣4 subunit interface by binding to pockets that completely (NS9283) or partially (CMPI) overlap the ACh-binding site.

PAM recognition at the ␣4:␣4 extracellular interface
CMPI recognition at the ␣4:␣4 subunit extracellular interface differs from that for NS9283 and for agonists that bind at the ␣4:␣4 subunit extracellular interface, especially in regard to the role of amino acid residues in the complementary ␣4(Ϫ) face. Amino acid substitutions at positions ␣4Gly-41 and ␣4Lys-64 to the corresponding amino acid residues in the ␤2 subunit (methionine and threonine, respectively) selectively reduced potentiation by CMPI but not NS9283 or dFBr. In agreement with these findings, our computational analyses in the absence of ACh predicted CMPI binding in close proximity to ␣4Gly-41 and ␣4Lys-64 with the latter predicted to form hydrophobic interactions with the chlorobenzene and isoxazole of CMPI. This hydrophobic interaction between ␣4Lys-64 and the chlorobenzene ring appears essential for CMPI potentiation, as replacement of the chlorobenzene ring by non-substituted benzene or cyclohexane increased CMPI EC 50 by ϳ60and ϳ250-fold, respectively (31).
In addition to CMPI interactions with non-conserved amino acid residues of the ␣4(Ϫ) face, CMPI in its most favorable binding mode in the absence of ACh interacts with conserved core aromatics of the ␣4(-) and ␣4(ϩ) faces. The CMPI piperidine ring, which is a strong base and can be protonated, occupies the same position as the carbonyl group of ACh and is in close proximity to (and predicted to form H-bonds) with both the carbonyl group of ␣4Trp-154 and the hydroxyl group of ␣4Tyr-204. The methyl substitution at the pyrazole ring is predicted to form -alkyl bonds, two with the benzene and pyrrole rings of ␣4Trp-156 (␣4 (ϩ) face) and one with the benzene of ␣4Trp-62 (␣4(Ϫ) face). Indeed, the methyl substitution enhanced the potency of CMPI by ϳ7-fold over non-substituted pyrazole (H instead of CH 3 ), and replacement of the methyl by an ethyl group further enhanced CMPI potency as an ␣4␤2 nAChR potentiator (31). We predict that longer alkyl substituents would further enhance the hydrophobic interaction of CMPI with these core aromatic residues in the ␣4:␣4 subunit interface.
Despite binding with high affinities at the ␣4:␣4 interface, CMPI and NS9283 lack agonist activity in wild-type (␣4)3(␤2)2 nAChRs. Their binding at the ␣4:␣4 extracellular interface potentiates current responses triggered by ACh binding at the ␣4:␤2 agonist-binding sites. The higher affinity of CMPI and NS9283 for the ␣4:␣4 versus ␣4:␤2 interface could explain their lack of agonist activity, as binding of at least two agonist molecules is required for channel gating (25). Indeed, NS9283 acted as an agonist in (␣4)3(␤2)2 nAChRs mutated to contain two ␣4:␣4-like agonist-binding sites (17). Although more studies are needed to fully understand the pharmacology of the ␣4:␣4 drug-binding sites, the work presented here provides comprehensive analyses of the role of the ␣4:␣4 extracellular interface in the allosteric modulation of ␣4␤2 nAChRs. The results of this study will facilitate the design and development of novel ␣4␤2 nAChR-selective ligands with potential experimental and clinical applications.

cDNA plasmids and site-directed mutagenesis
pSP64 poly(A) plasmids with cDNA encoding for human ␣4 or ␤2 nAChR subunit were generously provided by Dr. Jon Lindstrom (University of Pennsylvania). Point mutations were introduced into plasmid expression vectors coding for the ␣4 nAChR subunit using the QuikChange II Site-Directed Mutagenesis Kit (Agilent Technologies) then confirmed by DNA sequencing (GENEWIZ, LLC., South Plainfield, NJ). For each point mutation within ␣4, two custom-designed complimentary oligos containing the desired mutation were ordered from Integrated DNA Technologies (Coralville, IA). Amino acid numbers are based on the mature ␣4 subunit; add 28 to convert to amino acid numbering starting from the transitional N terminus (methionine 1) of the ␣4 subunit.

Receptor expression in Xenopus oocytes
cDNA plasmids of wild-type (WT) and mutant nAChR subunits were linearized with AseI (h␣4) and PvuII (h␤2), then cRNA transcripts suitable for oocytes expression were prepared from linearized plasmid using SP6 mMESSAGE mMACHINE high yield capped RNA transcription kits (Ambion) following the manufacturer's protocol. RNA concentration was estimated from the A 260 (with 260/280 ratio Ͼ1.9) and mixed at ratios 8␣:1␤ or 1␣:8␤ to express nAChRs with subunit stoichiometries of 3␣:2␤ or 2␣:3␤, respectively. The final concentration of the subunits RNA mixture was adjusted to 100 -200 ng/l with nuclease-free water, and oocytes were injected with a volume containing 5-20 ng of cRNA.
Oocytes-positive female X. laevis were purchased from NASCO (Fort Atkinson, WI), and ovarian lobules were harvested surgically following animal use protocols approved by the Texas A&M Health Sciences Center Institutional Animals Care and Use Committee. The lobules were treated with 3 mg/ml collagenase type 2 (Worthington Biomedical) with gentle shaking for 3 h at room temperature in Ca ϩ2 -free OR2 buffer (85 mM NaCl, 2.5 mM KCl, 1 mM MgCl 2 , 5 mM HEPES, pH 7.6). Stage V and VI oocytes were selected, injected with cRNA, and incubated at 18°C in ND96-gentamicin buffer (96 mM NaCl, 2 mM KCl, 1.8 mM CaCl 2 , 1 mM MgCl 2 , 5 mM HEPES, 50 g/ml gentamicin, pH 7.6) for 24 -72 h to allow receptor expression.

Electrophysiological recording
A standard two-electrode voltage-clamp technique was used as described previously (30,33) to acquire whole cell current from oocytes expressing WT or a mutant nAChR in response to drug(s) application. Xenopus oocytes were voltage-clamped at Ϫ50 mV using Oocyte Clamp OC-725C (Warner Instruments) in a custom-made small recording chamber and under continuous perfusion with recording buffer (100 mM NaCl, 2 mM KCl, 1 mM CaCl 2 , 0.8 mM MgCl 2 , 1 mM EGTA, 10 mM HEPES, pH 7.5). An automated perfusion system made up of glass reservoirs and Teflon tubing (Warner Instruments) was used to control recording chamber perfusion and drug application. Each recording run included 3-6 drug applications (10 s each) separated by 90-s wash intervals unless otherwise specified in the figure legends, and oocytes were washed for 3-5 min between runs. Each drug concentration was tested at least two times per oocyte and repeated on a number of oocytes (N) as specified in Tables 1-3. Current traces were digitized at 50 Hz and analyzed, and peak current amplitudes in response to drug applications were quantified using Digidata 1550A interface/ pCLAMP 10.4 software (Molecular Devices). CMPI and NS9283 were prepared as 10 mM stock solution in DMSO, and dFBr was prepared as 5 mM stock solution in water. Stock solutions were stored at Ϫ20°C and diluted to final concentrations in recording buffer on the day of experiments. At the highest concentration tested, DMSO was present at 0.1%, which had no effect on ACh responses.

Data analysis
Data analyses, parameter calculations, and statistical analyses were performed using Excel (Microsoft Corp.) and Sigmaplot v11 (Systat Software Inc.). For concentration-dependent potentiation of ACh-induced currents by CMPI, peak current amplitudes in response to co-applications of CMPI with 10 M ACh were normalized to the peak current amplitude elicited by 10 M ACh alone within same recording run. For ACh concentration-response curves in the absence or presence of CMPI, peak current amplitudes in response to co-applications of increasing concentrations of ACh (Ϯ1 M CMPI) were normalized to the peak current amplitude elicited by 1 mM ACh alone within same recording run. Replicas from the same oocyte were combined (average Ϯ S.D.; the oocyte with S.D. of Ͼ30% of its average was rejected), and data (average Ϯ S.E.) from N oocytes were plotted and fit to a 3-parameter Hill equation, where Ix is the normalized ACh current in the presence of CMPI at concentration x, I max is the maximum potentiation of current, h is the Hill coefficient, and EC 50 is the CMPI concentration producing 50% of maximal potentiation. For CMPI enhancement, I 0 ϭ 100, and the probability (P) that the I max differs from no potentiation (I max ϭ 100) was analyzed using one-way analysis of variance with the Holm-Sidak test and is reported in Table 2. For ACh concentration response, I 0 ϭ 0, and the probability (P) that the ACh EC 50 in the presence of CMPI differs from ACh EC 50 in the absence of CMPI was analyzed by the t test and is reported in Table 3.

Molecular modeling
Homology modeling and docking simulations (CDOCKER) were performed using the Discovery Studio 2017 molecular modeling package from Accelrys. Homology models based on two different X-ray crystallographic structures were constructed as follows.
A homology model of the extracellular domain of the human (␣4)3(␤2)2 nAChR was constructed from the structure of the Lymnaea stagnalis AChBP crystallized with carbamylcholine (PDB accession number 1UV6; Ref. 37) by replacing subunits A, C, and E with ␣4 residues and B and D with ␤2 residues. The computer-generated alignment was checked to ensure insertions/deletions occurred outside secondary structure motifs (␣-helices or ␤-sheets). The alignment between ␣4 and AChBP began with ␣4His-9 (amino acid numbering is for mature protein according to the X-ray structure of human (␣4)2(␤2)3 nAChRs (PDB accession number 5KXI; add 26 to obtain amino acid numbers starting from the translational N terminus) and with AChBP L1 and required ␣4 insertions of residues Asn-31, Asp-78, Tyr-79, Gly-105, and Phe-144 and an AChBP deletion at residue Tyr-168. The alignment between ␤2 and AChBP began with ␤2 Asp-2 (amino acid numbering is for mature protein; add 25 to obtain amino acid numbers starting from the translational N terminus) and with AChBP L1 and required ␤2 insertions of residues Lys-19, Lys-20, Gly-27, Phe-74, Asp-75, Asp-99, and Phe-13 and AChBP deletions of residues Ser-162, Ser-186, Cys-187, and Cys-188. Carbamylcholine molecules were placed into the two ␣4:␤2 agonist sites and the ␣4:␣4 interface site in orientations approximating that of carbamylcholine in the crystal structure, and the model was minimized to Ϫ2.30 ϫ 10 4 kcal/mol. The carbamylcholine molecule was removed from the ␣4-␣4 interface site before docking.
Additional docking on the 1UV6.PDB-based model was performed after first identifying (by docking) the preferred binding site for acetylcholine, placing an acetylcholine molecule in the ␣4:␣4 binding site, and minimizing (Ϫ22,937 kcal/mol final energy). A 12 Å binding-site sphere was used with its center between ␣4Gln-122 and ␣4Cys-198, and eight randomly positioned CMPI and NS9283 molecules were docked within this binding-site sphere using the same parameters described above.
For docking, eight structurally different molecules of CMPI and NS9283 were seeded together in random orientations and used to dock to the ␣4:␣4 interface "agonist" sites for each model using CDOCKER. CDOCKER is a CHARMm-based molecular dynamics simulated annealing program that treats the ligand as fully flexible while maintaining a rigid receptor (39,40). A binding-site sphere of radius 12 Å was used that was centered on the location of carbamylcholine or nicotine in the respective models. We configured the docking parameters so that each of the 8 seeded ligands was subjected to 40ϩ hightemperature molecular dynamic-induced structural alterations and 40ϩ random translation/rotation reorientations, producing a total of 1600ϩ attempts at docking. For each attempt, the ligand and the amino acids within the binding sphere were subjected to simulated annealing, and a full potential final minimization step from which the ligand-receptor interaction energies was calculated. The orientation of the ligand and the calculated interaction energies were collected for the 50 lowest energy solutions for each seeded ligand.
Author contributions-A. K. H. designed and supervised the research and wrote the paper. Z.-J. W., F. D., T. S. M., and K. R. performed the research and the analyzed data. D. C. C. performed the molecular modeling. All authors contributed to reading and editing the manuscript.