Key Residues in the Nicotinic Acetylcholine Receptor β2 Subunit Contribute to α-Conotoxin LvIA Binding*

Background: Toxins such as LvIA can help elucidate the physiological roles of nAChR subtypes. Results: Three residues in the β2 subunit were identified as critical to LvIA binding. Conclusion: The β complementary subunit plays a crucial role in the subtype selectivity of α-conotoxin LvIA. Significance: This study provides new insights into the unique selectivity of LvIA and more broadly into toxin-receptor interactions. α-Conotoxin LvIA (α-CTx LvIA) is a small peptide from the venom of the carnivorous marine gastropod Conus lividus and is the most selective inhibitor of α3β2 nicotinic acetylcholine receptors (nAChRs) known to date. It can distinguish the α3β2 nAChR subtype from the α6β2* (* indicates the other subunit) and α3β4 nAChR subtypes. In this study, we performed mutational studies to assess the influence of residues of the β2 subunit versus those of the β4 subunit on the binding of α-CTx LvIA. Although two β2 mutations, α3β2[F119Q] and α3β2[T59K], strongly enhanced the affinity of LvIA, the β2 mutation α3β2[V111I] substantially reduced the binding of LvIA. Increased activity of LvIA was also observed when the β2-T59L mutant was combined with the α3 subunit. There were no significant difference in inhibition of α3β2[T59I], α3β2[Q34A], and α3β2[K79A] nAChRs when compared with wild-type α3β2 nAChR. α-CTx LvIA displayed slower off-rate kinetics at α3β2[F119Q] and α3β2[T59K] than at the wild-type receptor, with the latter mutant having the most pronounced effect. Taken together, these data provide evidence that the β2 subunit contributes to α-CTx LvIA binding and selectivity. The results demonstrate that Val111 is critical and facilitates LvIA binding; this position has not previously been identified as important to binding of other 4/7 framework α-conotoxins. Thr59 and Phe119 of the β2 subunit appear to interfere with LvIA binding, and their replacement by the corresponding residues of the β4 subunit leads to increased affinity.

Nicotinic acetylcholine receptors (nAChRs), 3 which comprise many different molecular subtypes, are ligand-gated ion channels that are activated by the endogenous neurotransmitter acetylcholine (ACh) or exogenous nicotine (1,2). nAChRs are found in the neuromuscular junction, and in peripheral and central nervous systems throughout the animal kingdom, and play important roles in regulating synaptic transmission (3)(4)(5)(6)(7)(8)(9). Neuronal nAChRs are pentameric membrane-bound proteins, which are made up of ␣ (␣2-␣10) and ␤ (␤2-␤4) subunits (10). Pharmacological properties of the heteromeric nAChRs are influenced by the presence of ␤2 and/or ␤4 subunits (11). This study is part of an ongoing effort to elucidate the physiological role of each subtype of nAChR and the key binding residue determinants for selective ligands (12).
␣-CTx LvIA from Conus lividus was recently characterized and has high affinity for ␣3␤2 nAChRs, with an IC 50 of 8.7 nM (25). LvIA is notable for its ability to selectively block ␣3␤2 versus ␣6/␣3␤2␤3 or ␣3␤4 nAChRs. The residues in the ␤2 subunit that contribute to ␣-CTx LvIA binding to the ␣3␤2 nAChR remain unknown. We therefore performed a mutational study of the ␣3␤2 nAChR in which we assessed the influence of residues that line the ␤2 subunit on the binding of ␣-CTx LvIA.
Peptide Synthesis-A two-step oxidation protocol was used to synthesize ␣-CTx LvIA as described previously (25). Because this protocol worked well, we did not attempt a simpler onestep oxidation approach. In this protocol, linear (see Fig. 1A) and folded (see Fig. 1B Table 1) were created using PCR and the QuikChange sitedirected mutagenesis kit (Stratagene) according to the manufacturer's instructions. Primers that contained the desired point mutation as well as at least 15 bases on either side of the mutation were synthesized. The mutagenic primers were extended and incorporated by PCR. DpnI was then used to digest the methylated, non-mutated parental cDNA. The point mutated DNA was inserted in the pSP65 or pGEMHE vector, which was transformed into DH5␣-competent cells. All the PCR mutations were sequenced to confirm incorporation of the desired mutation (19). The nomenclature for the point mutants lists the naturally occurring residue position followed by the change made, e.g. ␣3␤2[V111I] is a ␤2 subunit with the valine residue at position 111 position replaced by an isoleucine residue.
cRNA Preparation and Injection into Xenopus laevis Oocytes-Capped cRNA was synthesized in vitro following linearization of the plasmid containing template DNA encoding the rat ␣3, ␤2, and ␤4 subunits, as well as the various mutant subunits using the mMESSAGE mMACHINE in vitro transcription kit (Applied Biosystems/Ambion, Austin, TX), as described previously (20). The cRNA was purified using the Qiagen RNeasy kit (Qiagen). Their concentration was determined by absorbance at 260 nm. Mature X. laevis frogs were anesthetized by submersion in 0.1% 3-aminobenzoic acid ethyl ester, and oocytes were surgically removed. Two collagenase treatments lasting 1 h were performed at room temperature to remove follicle cells. RNA transcripts of wild-type ␣3 subunit with either wild-type ␤2 or mutant ␤2 subunit were mixed at a molar ratio of 1:1. Fifty nl of this mixture with ϳ10 ng of each cRNA was injected into each Xenopus oocyte and incubated at 17°C. Oocytes were injected within 1 day of harvesting.
Voltage-clamp Recording-Voltage-clamp recordings were performed 1-4 days after cRNA injection. All recordings were done at ϳ22°C room temperature. Briefly, oocytes were voltage-clamped at Ϫ70 mV and exposed to ACh and peptide in a 30-l cylindrical oocyte recording chamber, which was gravityperfused at a rate of 2 ml/min with ND-96 buffer. The ND-96 buffer consisted of 96.0 mM NaCl, 2.0 mM KCl, 1.8 mM CaCl 2 , 1.0 mM MgCl 2 , 1 M atropine, 5 mM HEPES, 0.1 mg/ml bovine serum albumin, pH 7.1-7.5. The oocyte was subjected once a minute to a 1-s pulse of 100 M ACh. Once a stable baseline was achieved, either ND-96 alone or ND-96 containing varying concentrations of the ␣-CTx LvIA was manually pre-applied in a static bath for 5 min prior to the addition of the agonist ACh pulse.
Data Analysis-Three to five ACh responses were averaged for the baseline responses of ND-96 after a 5-min incubation just preceding a test response, which was used to normalize evoked responses as a percentage of control response. The percentage of response of the toxin was divided by the pre-toxin baseline value to yield a percentage of response. The dose-response data were fitted to the equation: % of response ϭ 100/ (1 ϩ ([toxin]/IC 50 ) nH ), where nH is the Hill coefficient, using GraphPad Prism (GraphPad Software, San Diego, CA). Each data point of a dose-response curve is the average Ϯ S.E. from at least three oocytes. IC 50 values were determined by nonlinear regression analysis using Graph-Pad Prism.
Molecular Modeling-A molecular model of the interaction between LvIA and the ligand-binding domain of ␣3␤2 nAChR was built by homology using the NMR solution structure of LvIA (Protein Data Bank (PDB) identifier 2mdq) and the crystal structure of the complex between acetylcholine-binding protein (AChBP) and conotoxin PnIA variant (PDB identifier 2br8) as templates, as described previously (25). The molecular model was refined by a 30-ns explicit water molecular dynamics simulation carried out with the GROMACS 4.6.5 (26) package and the ff03 force field (27), using a procedure described previously (28,29). All the models of complexes involving ␤2 subunit mutants were generated by substituting residue side chains using Modeler 9v14 (30). This procedure refines the positions of the substituted side chain atoms as well as of those of the neighbor residues using a conjugate gradient minimization followed by a short molecular dynamics simulation. The molecular models were refined by a 2-ns explicit water molecular dynamics simulation, and the simulations of the T59K, V111I, and F119Q mutants were extended to 10 ns.

RESULTS
Chemical Synthesis of ␣-CTx LvIA-␣-CTx LvIA linear peptide ( Fig. 1A) was successfully synthesized with Fmoc (N-(9fluorenyl)methoxycarbonyl) chemistry, in which Cys residues were orthogonally protected using acid-labile S-trityl and acidstable S-acetamidomethyl groups. The acid-labile groups (tri-␤2 Subunit Contribution to ␣-CTx LvIA Binding of ␣3␤2 nAChR tyl) were first removed after cleavage of the assembled peptide chain from the resin. The linear peptide after the initial cleavage was purified by HPLC with a retention time of 27.7 min (Fig.  1C). Ferricyanide was used to close the first disulfide bridge, and iodine oxidation was used to subsequently close the second disulfide bridge. The fully folded peptide of ␣-CTx LvIA with Cys 1 -Cys 3 and Cys 2 -Cys 4 disulfide bonds (Fig. 1B) was purified again by HPLC, with a retention time of 27.9 min (Fig. 1D). The mass of the ␣-CTx LvIA matched that of the amidated sequence (calculated average mass, 1679.9 Da; observed, 1679.7). This synthesized fully folded peptide was utilized in all subsequent experiments.
Effect of Mutations of the ␤2 Subunit on Block by ␣-CTx LvIA-Previous studies using molecular modeling of related toxins suggested the nAChR positions that form the ligandbinding pocket of ␣-conotoxins (23,31,32). The residue positions of the ␤2 and ␤4 subunits that were suggested to form the LvIA-binding pocket in a previous modeling study (25) are highlighted in Fig. 2. We created point mutations of the ␤2 subunit where residues in this pocket were replaced with those found in the homologous position of the ␤4 subunit. These mutant receptors were then tested to determine toxin potency differences (Table 1). Seven nAChR ␤2 mutants were created, including Q34A, T59I, T59K, T59L, K79A, V111I, and F119Q. The concentration-response block by ␣-CTx LvIA on ␣3␤4 nAChR and wild-type and mutant ␣3␤2 nAChRs was investigated (Table 1 and Fig. 3). The potency at wild-type ␣3␤2   3). The subunit positions that were shown to contact LvIA according to a previous molecular modeling study (Luo et al. (25)) are underlined.  Fig. 3B).
Molecular Modeling-A molecular model of the interactions between LvIA and the wild-type ␣3␤2 nAChR showed that the ␤2 subunit positions considered for mutations are all potentially in contact with the conotoxin with the exception of position 34, as shown in Fig. 5A. LvIA had similar activity at wildtype and ␣3␤2 Q34A nAChR, in agreement with the absence of interaction of this position. The wild-type ␤2 residue Lys 79 can form a surface salt bridge with LvIA Asp 11 , and this interaction was found to be stable over a 30-ns molecular dynamics simulation (Fig. 5B). The three other substituted positions, i.e. positions 59, 111, and 119, are at least partly buried at the interface with LvIA. The change of activity of LvIA correlated with a change of buried surface solvent-accessible surface area at the interface for mutants of positions 59 and 119 (Fig. 5C). The molecular model did not provide a simple explanation for the decreased activity of LvIA at the V111I mutant, and we propose that this mutation could potentially result in conformational changes that cannot be modeled using short molecular dynamics simulations. The other mutated positions, i.e. 59 and 119, are located in ␤-strands, and the residues chosen for substitu-  Table 1.
a Time to 95% recovery after toxin washout. b Ͻ 3% recovery after 20 min; concentration of ␣-CTx LvIA was 10 M.
␤2 Subunit Contribution to ␣-CTx LvIA Binding of ␣3␤2 nAChR tions are not likely to disrupt this ␤-strand secondary structure. The 10-ns simulations of complexes incorporating the mutations T59K, V111I, or F1119Q resulted in similar binding modes with no change of binding site conformation (Fig. 6).

DISCUSSION
Neuronal nAChRs are widely expressed in the CNS and peripheral nervous system in adults and during development, but the identification of which subtype is expressed in which nervous cell is challenging (33)(34)(35)(36)(37)(38). LvIA is the first ligand to be highly specific for ␣3␤2 nAChR, and it could potentially be used in physiological studies of this receptor (25). We sought here to gain further insights into the binding interactions of LvIA at this receptor through mutations of positions that have been shown to be important for the binding of other ␣-CTxs and ligands (24,39). These studies have shown that competitive nicotinic ligands of The nAChRs display different reversibility kinetics after block. C indicates control responses to ACh. Oocytes were exposed to 10 nM peptide for 5 min followed by repetitive application of ACh. ␣-CTx LvIA at 10 nM blocked ϳ55% current of wild-type ␣3␤2 nAChR with rapid reversibility (A), but did not block ␣3␤2[V111I] nAChR (B). LvIA at 10 nM blocked ϳ100% current of mutant receptors ␣3␤2[F119Q] nAChR with slow reversibility (C) and ␣3␤2[T59K] nAChR with slowest reversibility (D). FIGURE 5. Molecular modeling of the interaction between LvIA and ␣3␤2 wild-type and mutant nAChR. A, binding of LvIA (white) into the rat wild-type ␣3␤2-binding pocket, which comprises the ␣3 principal subunit (green) and the ␤2 complementary subunit (blue). The conformation of the side chains of the ␤2 positions that were mutated are displayed overlaid with those of the wild-type receptor. The side chains of the mutants are shown in different colors from those used for the wild-type structure. B, distance between the NZ atom of ␤2 Lys 79 and the CG atom of LvIA Asp 11 over a 2-ns molecular dynamics simulation. C, correlation between the differences of buried solvent-accessible surface area between wild-type and mutant complexes (⌬ BASA) and the IC 50 for the ␣3␤2 mutants.
nAChRs generally bind to both the ␣ subunits and the ␤ subunits that form a ligand-binding interface (18,32).
We investigated the influence of seven ␣3␤2 nAChR mutants on the binding of ␣-CTx LvIA. These residues were chosen based on previous findings with the related ␣-CTx LtIA (23 (Tables 1 and 2; Figs. 3 and  4). The three other mutations, Q34A, K79A and T59I, had little or no detectable effect on ␣-CTx LvIA activity (Tables 1 and 2; Figs. 3 and 4). ␣-CTx LvIA at 10 nM blocked ϳ55% of the current of wild-type ␣3␤2 with rapid reversibility but blocked Ͼ95% of the current of ␣3␤2[F119Q] and ␣3␤2[T59K]; the block of the latter two nAChRs had much slower reversibility after toxin washout when compared with that observed for the wild-type ␣3␤2 nAChR (Fig. 4). The substitution of Phe 119 of the ␤2 subunit by Gln, which is present in the homologous position of the ␤4 subunit (␣3␤2[F119Q]), resulted in a 15-fold increase in ␣-CTx LvIA potency. The mutation T59K caused an 11-fold increase sensitivity for ␣-CTx LvIA, partly due to a decrease in off-rate (Fig. 4D). A similar finding has been reported for the 4/4 ␣-CTx BuIA (19). The potency of BuIA at ␣3␤2[F119Q] and ␣3␤2[T59K] increased 8-and 20-fold, respectively, when compared with wild-type ␣3␤2, with very slow off-rates. However, BuIA had a faster off-rate but similar IC 50 at ␣3␤2[V111I] versus wild-type ␣3␤2, in contrast to LvIA, which has a 15-fold decrease in potency at ␣3␤2[V111I] when compared with wild-type ␣3␤2 (Table 3) (19). Thus, we suggest that BuIA and LvIA have overlapping, yet distinct binding interactions with the receptor. Overall our data suggest that the three positions on the receptor, 59, 111, and 119, are key to LvIA binding. Of course, because we examined only a finite number of mutations, we cannot exclude the possibility that other positions might also be important.
As far as ligand residues contributing to binding are concerned, the highly conserved Ser-Xaa-Pro motif in the first loop of ␣-CTxs contains a small ␣-helix important for nAChR binding. ␣-CTx LtIA is atypical because it lacks this Ser-Xaa-Pro motif and has been suggested to bind a novel microsite on the ␣3␤2 nAChR (19). ␣-CTx LtIA potentially interacts with ␤2 Phe 119 and ␤2 Lys 79 because the Phe 119 and Lys 79 mutants disrupted LtIA binding (19), but mutations of these positions were without effect for activity of ␣-CTxs MII, PnIA, and GID (18). By contrast, the mutation F119Q increased affinity of LvIA, and the mutation K79A did not affect LvIA activity. The mutation V111I in the ␤2 subunit was previously reported to have only a small effect on the activity of 4/7 ␣-CTxs MII, PnIA, GID (18),  and LtIA (23) ( Table 3). By contrast, this mutation decreased LvIA activity by 15-fold. LvIA displays the conserved Ser-Xaa-Pro motif in its first loop, suggesting that it adopts a similar binding mode to most 4/7 ␣-CTxs. The sequence in the second loop of ␣-CTx LvIA is therefore probably a determinant of its unique selectivity and distinct binding site. Molecular models indeed suggest that this second loop, especially residues Asn 9 , Val 10 , Asp 11 , and Pro 13 , interacts with the ␤2 subunit. Interestingly, the 4/7 ␣-CTx PeIA also potently blocks ␣3␤2 nAChRs and has similar residues in its second loop (31).
The mutant K79A does not show a significant difference in activity from the wild-type nAChR, but the molecular models suggest that the Lys 79 residue establishes a stable charge interaction with LvIA Asp 11 (Fig. 5B). It has been proposed that surface salt bridges can have little contribution to binding affinity because the favorable charge-charge interaction can be counterbalanced by the negative entropic effect of restraining the conformation of the side chain (40). This compensation of enthalpy for entropy components between apo and bound states is a potential explanation for the innocuous nature of the K79A substitution. Indeed, the Lys 79 side chain is highly exposed to the solvent and should therefore have considerable conformational freedom in the absence of the toxin. The side chain of Lys 79 was restrained during the molecular simulations of the bound toxin, suggesting a significant entropic cost to the immobilization of the side chain.
The molecular models suggest that substitutions at positions 119 and 59 increase the solvent-accessible surface area buried at the interface, and this increase correlates with higher affinities of LvIA observed experimentally (Fig. 5C). In particular, the three mutations, T59I, T59L, and T59K, incrementally introduce longer side chains at position 59, and they result in increasing inhibitory potency of LvIA. The introduction of a positively charged Lys at position 59 results in the burial of a positively charged group, which could be detrimental to binding, but this residue can potentially interact with the negatively charged ␤2 subunit Glu 61 , which is proximally located (Fig. 5A). The F119Q mutation resulted in better complementarity at the interface by creating further interactions, especially with LvIA residues Val 10 and Pro 13 , resulting in the largest buried surface area among all mutants in this study, in agreement with the highest inhibitory activity of LvIA among all mutants.
It is interesting to compare the trends in LvIA binding to ␣3␤2 versus ␣3␤4 relative to the individual residue substitutions at the three key positions of 59, 111, and 119. In principle, the decreases in IC 50 values associated with the T59K and F119Q substitutions should more than compensate for the increased IC 50 associated with the V111I substitution. Nevertheless LvIA is 17-fold more potent at ␣3␤2 than at ␣3␤4. The non-additivity of the single point mutant effects can probably be explained by the spatial organization of these positions because the side chain at position 119 is sandwiched by those of positions 111 and 59. The ␤4 subunit, which displays bulkier side chains at these positions than the ␤2 subunit, should present a different interface to LvIA than the ␤2 subunit single point mutants.
In conclusion, we have identified three residues in the nAChR ␤2 subunit that are key to the binding interaction of LvIA with the ␣3␤2 nAChR. Furthermore, molecular modeling indicates that the sequence of residues in the second loop of LvIA is particularly important for high affinity for the ␣3␤2 nAChR. These findings help provide insights into the unique selectivity profile of this toxin. Understanding interactions between different ␣-CTxs and ␣3␤2 nAChR should further help to elucidate the molecular pharmacology of this subtype.