Non-equivalent Ligand Selectivity of Agonist Sites in (α4β2)2α4 Nicotinic Acetylcholine Receptors

Background: The α4/β2 and α4/α4 interfaces of the (α4β2)2α4 nicotinic acetylcholine receptor house structurally different agonist sites. Results: Agonists of a certain size cannot bind the α4/α4 interface, which decreases efficacy. Conclusion: The ability to bind all agonist sites in (α4β2)2α4 receptors critically influences agonist efficacy. Significance: The finding adds a new level of complexity to structural mechanisms governing agonist efficacy. The α4β2 nicotinic acetylcholine receptor (nAChR) is the most abundant nAChR type in the brain, and this receptor type exists in alternate (α4β2)2α4 and (α4β2)2β2 forms, which are activated by agonists with strikingly differing efficacies. Recent breakthroughs have identified an additional operational agonist binding site in the (α4β2)2α4 nAChR that is responsible for the signature sensitivity of this receptor to activation by agonists, yet the structural mechanisms determining agonist efficacy at this receptor type are not yet fully understood. In this study, we characterized the ligand selectivity of the individual agonist sites of the (α4β2)2α4 nAChR to determine whether differences in agonist selectivity influence agonist efficacy. Applying the substituted cysteine accessibility method to individual agonist sites in concatenated (α4β2)2α4 receptors, we determined the agonist selectivity of the agonist sites of the (α4β2)2α4 receptor. We show that (a) accessibility of substituted cysteines to covalent modification by methanesulfonate reagent depends on the agonist site at which the modification occurs and (b) that agonists such as sazetidine-A and TC-2559 are excluded from the site at the α4/α4 interface. Given that additional binding to the agonist site in the α4/α4 interface increases acetylcholine efficacy and that agonists excluded from the agonist site at the α4/α4 interface behave as partial agonists, we conclude that the ability to engage all agonist sites in (α4β2)2α4 nAChRs is a key determinant of agonist efficacy. The findings add another level of complexity to the structural mechanisms that govern agonist efficacy in heteromeric nAChRs and related ligand-gated ion channels.

The ␣4␤2 nicotinic acetylcholine receptor (nAChR) is the most abundant nAChR type in the brain, and this receptor type exists in alternate (␣4␤2) 2 ␣4 and (␣4␤2) 2 ␤2 forms, which are activated by agonists with strikingly differing efficacies. Recent breakthroughs have identified an additional operational agonist binding site in the (␣4␤2) 2 ␣4 nAChR that is responsible for the signature sensitivity of this receptor to activation by agonists, yet the structural mechanisms determining agonist efficacy at this receptor type are not yet fully understood. In this study, we characterized the ligand selectivity of the individual agonist sites of the (␣4␤2) 2 ␣4 nAChR to determine whether differences in agonist selectivity influence agonist efficacy. Applying the substituted cysteine accessibility method to individual agonist sites in concatenated (␣4␤2) 2 ␣4 receptors, we determined the agonist selectivity of the agonist sites of the (␣4␤2) 2 ␣4 receptor. We show that (a) accessibility of substituted cysteines to covalent modification by methanesulfonate reagent depends on the agonist site at which the modification occurs and (b) that agonists such as sazetidine-A and TC-2559 are excluded from the site at the ␣4/␣4 interface. Given that additional binding to the agonist site in the ␣4/␣4 interface increases acetylcholine efficacy and that agonists excluded from the agonist site at the ␣4/␣4 interface behave as partial agonists, we conclude that the ability to engage all agonist sites in (␣4␤2) 2 ␣4 nAChRs is a key determinant of agonist efficacy. The findings add another level of complexity to the structural mechanisms that govern agonist efficacy in heteromeric nAChRs and related ligand-gated ion channels.
A central aim in nicotinic acetylcholine receptor (nAChR) 2 physiology and pharmacology is to understand the interactions between agonists and the agonist sites that influence gating efficacy. Functional studies of nAChRs have shown the importance of interactions between agonists and residues on the principal subunit for agonist efficacy (1,2), and crystal structures of the homolog acetylcholine-binding protein in complex with partial agonists of nAChRs have suggested that partial capping of loop C upon the binding of partial agonists in comparison with the complete capping induced by full agonists is a key determinant of agonist efficacy (3)(4)(5)(6). However, acetylcholinebinding protein crystal structures in complex with agonists of nAChRs also show that partial agonists establish water-mediated hydrogen bonds between their hydrogen bond acceptor moiety and the backbone atoms of hydrophobic residues on the complementary face (6,7), which could affect gating efficacy. Studies of the interactions between agonists and ␣4␤2 nAChRs by unnatural amino acid mutagenesis have confirmed hydrogen bonds between the complementary face and the hydrogen bond acceptor moieties of partial agonists, although not necessarily mediated by a water molecule (8,9). Further support for a role of complementary residues on agonist efficacy comes from structural and mutational studies that show that several residues contributed by the complementary subunit affect the efficacy of channel gating (7,10). Given that agonist sites in heteromeric nAChRs may have different complementary subunits and, hence, different structural elements, the impact of the complementary face on agonist efficacy raises the following fundamental question. Does non-equivalency of agonist sites influence agonist efficacy?
An ideal nAChR to investigate the role of functional nonequivalency of agonist sites on agonist efficacy is the (␣4␤2) 2 ␣4 nAChR because this protein possesses two types of agonist sites, which are functionally and structurally non-equivalent (11,12), and because a fully concatenated version of the receptor that aids the study of individual subunit interfaces is available (13). The (␣4␤2) 2 ␣4 nAChR is one of two alternate forms of the ␣4␤2 nAChR (14), the most abundant nAChR type in the brain and a key mediator of the rewarding and reinforcing effects of nicotine (15,16). Despite having structurally identical agonist binding sites at their ␣4/␤2 interfaces, the alternate receptors display strikingly different sensitivities to activation by agonists and to high-affinity desensitization (14,17). These differences are accounted for partly by an additional operational agonist binding site at the signature ␣4/␣4 subunit interface of the (␣4␤2) 2 ␣4 nAChR (11,12,17). From studies of the ACh sensitivity of (␣4␤2) 2 ␣4 nAChRs with an impaired or ablated ␣4/␣4 agonist site (11,17), it appears that binding of ACh (submaximal concentrations) to the agonist sites at the ␣4/␤2 interfaces produces efficacious gating but that occupation of all three agonist sites (at high concentrations) is necessary for maximal receptor activation. Hence, a likely functional consequence of agonist exclusion from the site at the ␣4/␣4 interface is partial agonism.
Here, we have applied the substituted cysteine accessibility method to concatenated (␣4␤2) 2 ␣4 receptors to address for the first time the agonist selectivity of individual agonist sites in this receptor type. We show that the site at the ␣4/␣4 interface does not accommodate agonists larger than the smoking cessation drug varenicline and that agonists excluded from the site behave as partial agonists. Thus, the ability to engage all agonist sites influences agonist efficacy in (␣4␤2) 2 ␣4 nAChRs and possibly other related neurotransmitter gated ion channels with a (␣␤) 2 ␣ subunit stoichiometry (e.g. the (␣3␤4) 4 ␤4 nAChR (18) and heteromeric glycine receptor (19)).

EXPERIMENTAL PROCEDURES
Mutagenesis and Expression in Oocytes-Xenopus laevis oocytes were prepared from whole ovary tissue obtained from the European Xenopus Resource Center (Portsmouth, UK). Human cRNAs encoding wild type or mutant concatenated (␣4␤2) 2 ␣4 nAChRs were injected into freshly isolated Xenopus oocytes as described previously (11,13). The fully concatenated form of the (␣4␤2) 2 ␣4 nAChR was engineered as described previously (11,13). Briefly, the signal peptide and start codon were removed from all of the subunits but the first (a ␤2 subunit), and the subunits were bridged by AGS linkers. Only the last subunit in the construct (a ␣4 subunit) contained a stop codon. The subunits were subcloned into a modified pCI plasmid vector (Promega) using unique restriction enzyme sites flanking the N and C termini of each subunit. To introduce a mutation into a specific subunit of the concatemeric (␣4␤2) 2 ␣4 nAChR, the mutation was first introduced into the subunit subcloned into the modified pCI plasmid using the QuikChange site-directed mutagenesis kit (Stratagene). The presence of the mutation and the absence of unwanted mutations were confirmed by sequencing the entire cDNA insert (SourceBioscience). The mutated subunit was then ligated into the concate-mer using unique restriction enzyme sites. To confirm that the mutated subunit was incorporated into the concatemer, the subunit was cut from the concatemer using unique restriction enzyme sites, and then its nucleotide sequence was verified by DNA sequencing (SourceBioscience). For clarity, mutations in the concatemeric receptors are shown as superscript positioned in the (ϩ)-or (Ϫ)-face of the mutated subunit. Thus, in ␤2 L146C _␣4_␤2_␣4_␣4 receptors, the mutation L146C is located in the (Ϫ)-face of the ␤2 subunit occupying the first position of the linear sequence of the concatemer, whereas in ␤2_␣4_␤2_␣4 T152C _␣4 receptors, ␣4T152C is in the (Ϫ)-face of the ␣4 subunit in the fourth position of the linear sequence of the concatemer. All concatemeric constructs were assayed for integrity using restriction enzyme digestion and the LT reporter mutation (L9ЈT in the second transmembrane domain) as described previously (11).
Oocyte Electrophysiology-Two-electrode voltage clamp recordings on oocytes were carried out as described previously (11,13). Concentration-response curves for agonists were obtained as described previously (11). Concentration response curves were plotted using Prism version 5.0 (GraphPad Software, San Diego, CA). The agonist activation concentration response curve data were first fit to the one-component Hill equation, I ϭ I max /(1 ϩ (EC 50 /x) nH ), where EC 50 represents the concentration of agonist inducing 50% of the maximal response (I max ), x is the agonist concentration, and nH is the Hill coefficient. When agonists induced biphasic receptor activation, the concentration-response curve data were fit to the sum of two Hill equations, as described previously (14).
Methanethiosulfonate (MTS) Modification of Substituted Cysteines-2-Trimethylammonium ethyl methanethiosulfonate (MTSET) (Toronto Research Chemicals, Toronto, Ontario, Canada) was used to covalently modify the introduced free cysteines (20). The accessibility of the introduced cysteines to MTSET was determined using the following protocol. Current responses elicited by 5-s pulses of ACh were measured every 5 min from oocytes expressing wild type or mutant receptors until the ACh responses varied by less than 5% for four consecutive pulses (stabilized responses). For ␤2_␣4_ ␤2_␣4 T152C _␣4 and ␤2_␣4_␤2_␣4 T152C,H142A _␣4 receptors, which exhibited monophasic sensitivity to ACh (Figs. 1C and 6, A and B), the concentrations of ACh pulses were 1 and 1.15 mM, respectively. These concentrations corresponded approximately to EC 50 ϫ 5 (see Table 1). For ␤2 L146C _␣4_␤2_␣4_␣4 receptors, which exhibited biphasic sensitivity to ACh (see Fig. 1D), the concentration of ACh pulses was EC 50-2 (1.6 mM) ( Table 1). We have previously reported that impairment of any of the agonist sites in ␤2_␣4_␤2_␣4_␣4 receptors may produce biphasic ACh-induced receptor activation due to the presence of mutated and wild type agonist sites in the concatamer (11). The biphasic curves comprise a high sensitivity component that is contributed mainly by the unaltered agonist sites and a low sensitivity component that is contributed by the mutated and non-mutated agonist site (11). Higher concentrations of ACh (i.e. EC 50-2 ϫ 5) were not used on ␤2 L146C _␣4_␤2_␣4_␣4 receptors to minimize possible ion channel blockade by ACh and/or chronic receptor desensitization. After stabilization, the MTSET reagent (1 mM) was bath-applied for 120 s, followed by a 90-s wash with Ringer's solution. Prelimi-nary experiments showed that this concentration and time of MTSET exposure were sufficient to reach maximal inhibition of the ACh-evoked responses in all mutant receptors tested. After washing, ACh was applied again every 5 min until the amplitude of the responses was constant. The average of the current amplitudes prior to application of MTSET was the control response current (I initial ), and the average of current amplitudes after rinsing was the average response after MTSET application (I after MTSET ). The effect of MTSET reagents was estimated using the equation, % Change ϭ ((I after MTSET /I initial ) Ϫ 1) ϫ 100.
Rate of MTSET Modification-The rate of covalent modification of substituted cysteines by MTSET was determined by measuring the effect of sequential applications of subsaturating concentrations of MTSET using a protocol based on one previously used on GABA A receptors (21). The concentration of MTSET causing subsaturating effects was determined separately for each mutant receptor. These concentrations were 1 M for ␤2_␣4_␤2_␣4 T152C _␣4 and ␤2_␣4_␤2_␣4 T152C,H142A _␣4 receptors and 10 M for ␤2 L146A _␣4_␤2_␣4_␣4 receptors. The responses to ACh prior to MTSET reagent application were first stabilized as follows. EC 50 ϫ 5 or EC 50-2, depending on the receptor under study, was applied for 5 s, followed by a recovery time of 70 s. Immediately after the recovery time, a pulse of a ligand to be tested later for protection (e.g. ACh, cytisine, varenicline, TC-2559, or sazetidine-A) was applied for 10 s, followed by a 3 min and 40-s wash with Ringer solution. This cycle was repeated until the ACh responses stabilized (Ͻ5% variance of peak current responses to ACh on four consecutive applications) (Fig. 3A). Ligands to be tested for their ability to protect the introduced cysteine residues from MTSET reactions (i.e. protectants) were applied during the stabilization of the ACh responses to correct for any process of desensitization and/or ion channel blockade that could develop during the protection assays described below. MTSET was then applied using the following sequence of reactions. At time 0, ACh was applied for 5 s, followed by a period of recovery of 70 s; MTSET was then applied for 10 s, followed by a recovery period of 20 s (Fig. 3A). Immediately after the recovery time, the protectant was applied for 10 s, after which time the cell was washed with Ringer's solution for 3 min and 40 s. This cycle was repeated until the peak current responses to ACh no longer changed, indicating completion of the MTSET reaction. After completion of the MTSET reaction, ACh and protectant were applied as described above to demonstrate that the observed changes in ACh responses were induced by MTSET (Fig. 3A).
Protection Assays with Ligands-To determine whether the accessibility of the incorporated cysteines could be altered by the presence of agonists or antagonist, the following protocol was used. Peak current responses to 5-s pulses of ACh (EC 50 ϫ 5 or EC 50-2 ) were stabilized as described above, after which time MTSET was applied using the following sequence. At time 0, ACh was applied (5 s), followed by 70-s recovery; MTSET and the protectant for agonists or IC 50 x5 for the antagonist Dh␤E were then co-applied for 10 s, followed by a recovery period of 3 min and 40 s (Fig. 3B). This cycle was repeated nine times (90 s in total). At the end of this cycle, ACh and protectant were applied as described for the MTSET reaction rate protocol. The For cytisine protection assays on ␤2 L146A _␣4_␤2_␣4_␣4 receptors, the concentration of cytisine used was 0.3 M, which corresponds approximately to an EC 10 . Preliminary studies showed that higher concentrations of cytisine slowed the MTSET reaction to the extent that it was not possible to measure the rate over the time course of the experiments. At lower concentrations (1-10 nM), cytisine did not protect the substituted cysteine. At the end of each protection assay, the cells were exposed to maximal MTSET to ensure that the previously protected mutant cysteines were still accessible. To study the effects of MTSET on ␤2_␣4_␤2_␣4 T152C,H142A _␣4 receptors, the ACh pulses were 1.15 mM (EC 50 ϫ 5; see Table 1), whereas the protectants tested, TC-2559 and sazetidine-A, were used at 20 M (approximately EC 50 ϫ 5). Preliminary experiments indicated that at these concentrations, TC-2559 and sazetidine-A permitted accurate measurements of the rate of MTSET reaction.
For all rate experiments, the decrease in the peak current response to ACh was plotted versus cumulative time of MTSET exposure. The change in response to ACh after cumulative time of MTSET addition (t) was expressed relative to the response to ACh prior to the MTSET addition, at t ϭ 0. The data expressed in this way were fit with a single exponential decay curve to obtain an estimate of the first order rate constant (k 1 ) and final current response (I ∞ ) according to the equation, I EC ϫ 5 ϭ I ∞ ϩ (1 Ϫ I ∞ )e Ϫk1t . A second-order rate constant (k 2 ) was calculated by dividing k 1 by the concentration of MTSET used (21).
Structure Homology Modeling and Docking-Sequences of the human ␣4 and ␤2 nAChR subunits were obtained from the ExPASy proteomics server (22) with accession numbers P43681 (␣4) and P17787 (␤2). The homopentameric Lymnaea stagnalis acetylcholine-binding protein structure (Protein Data Bank code 1UW6) (23) was used to generate models of the extracellular domain of the (␣4␤2) 2 ␣4 nAChR, as described previously (24). Molecular docking of nicotinergic ligands at the agonist binding sites of the ␣4␤2 homology models was investigated using the Lamarckian genetic algorithm search method as implemented in AutoDock version 4.0, as described previously (25).
Statistical Analysis-Data analysis was carried out using nonlinear regression analysis included in the Prism software package (GraphPad Software). An F-test determined whether the one-site or biphasic model best fit the concentration response data; the simpler one-component model was preferred unless the extra sum-of-squares F-test had a value of p less than 0.05. Statistical analysis on agonist efficacies or on MTSET accessibility was conducted using a one-way analysis of variance, followed by a post hoc Dunnett's or Bonferroni test. Statistical differences between control and test log EC 50 values were determined using unpaired Student's t tests.

RESULTS
Effects of Cysteine Mutations-We characterized the agonist selectivity of the agonist binding sites in (␣4␤2) 2 ␣4 nAChRs by establishing if agonists could slow down the thiol-specific reagent MTSET reaction with free cysteines introduced into the complementary face of the agonist binding sites. We assumed that agonists that slowed down the reaction impeded MTSET accessing the cysteine residues (20,21). To characterize individual agonist sites, we took advantage of the availability of concatenated (␣4␤2) 2 ␣4 nAChRs (␤2_␣4_␤2_␣4_␣4 nAChRs) and substituted cysteines into individual subunit interfaces (Fig. 1, A and B). Agonist binding sites on ␣4/␤2 interfaces are located at the interface between the first (a ␤2 subunit) and second (an ␣4 subunit) positions of the concatemeric receptor (␣4/␤2-1 interface) and between the third (a ␤2 subunit) and fourth (an ␣4 subunit) positions (␣4/␤2-2 interface) of the concatemeric receptor (11). The third agonist site is located at the interface between the ␣4 subunits located in the fourth and fifth subunit positions of the concatenated receptor, and the complementary face for this site is contributed by the ␣4 subunit in the fourth position of the ␤2_␣4_␤2_␣4_␣4 receptor ( Fig. 1A) (11). To study the agonist site at the ␣4/␣4 interface, we introduced ␣4T152C into the fourth position of the concatemer to engineer the mutant ␤2_␣4_␤2_␣4 T152C _␣4 receptor (Fig. 1A). ␣4T152 occupies a position homologous to that of ␤2L146 on loop E on the complementary side of agonist sites on ␣4/␤2 interfaces (Fig. 1B), where it appears to act as a gatekeeper residue (26). To study an agonist site located at a ␣4/␤2 interface, we incorporated ␤2L146C into the first position of the ␤2_␣4_␤2_␣4_␣4 (␣4/␤2-1 interface) and produced a ␤2 L146C _␣4_␤2_␣4_␣4 mutant receptor (Fig. 1B). The agonist site at the ␣4/␤2 interface adjacent to the fifth subunit of the concatemer (␣4/␤2-2 interface) was not studied because we found previously that the agonist sites at ␣4/␤2-1 and ␣4/␤2-2 interfaces respond similarly to alanine or cysteine substitutions (11), suggesting functional equivalency of these sites.
Wild type and mutant receptors, expressed heterologously in Xenopus oocytes, were activated in a concentration-dependent manner in response to ACh (Fig. 1, C and D). For ␤2_ ␣4_␤2_␣4 T152C _␣4 receptors, the concentration-response curve for ACh was monophasic and shifted to the right by 2.4fold, compared with wild type (Fig. 1C and Table 1). For ␤2 L146C _␣4_␤2_␣4_␣4 receptors, the ACh concentration-response curve was biphasic (p Ͻ 0.001; n ϭ 6), comprising a high-and a low-sensitivity component ( Fig. 1D and Table 1). Introduction of single point mutants into individual agonist sites of concatenated (␣4␤2) 2 ␣4 nAChRs may produce biphasic ACh concentration-response curves, depending on the extent to which mutations decrease the sensitivity of the mutated agonist site to ACh (11). The high-sensitivity component of the curve is contributed mostly by intact agonist sites, whereas the low-sensitivity component is contributed by both intact and impaired agonist sites (11). Mutant receptors were also responsive to partial agonists cytisine, varenicline, TC-2559, and sazetidine-A (Table 1). For ␤2_␣4_␤2_␣4 T152C _␣4 receptors, the EC 50 for cytisine or TC-2559 was not different from wild type, but the EC 50 for varenicline was 4-fold lower (Table 1). In addition, the competitive antagonist dihydro-␤-erythroidine (Dh␤E) inhibited the responses activated by ACh with wild type potency (IC 50 ϭ 0.25 (0.07;0.43) M versus wild type IC 50 ϭ 0.28 (0.024;0.033) M; mean (minimum;maximum for 95% CI)). By comparison, ␤2 L146C _␣4_␤2_␣4_␣4 receptors retained wild type sensitivity for varenicline but displayed reduced sensitivity for cytisine and TC-2559 (Table 1). These results indicate that ␣4T152C and ␤2L146C on loop E of the complementary component of the agonist site at the ␣4/␣4 and ␣4/␤2 interfaces, respectively, affect agonist sensitivity. The findings are in agreement with a previous report that noted that L119C, L119C, and L121C, the corresponding cysteine substitutions in the vertebrate muscle nAChR subunits ␥, ⑀, and ␦, respectively, treated with an MTS reagent, display reduced affinity for dimethyl-Dtubocurarine and ␣-conotoxin M1 (27). The relative efficacies of agonists, compared with ACh, were not affected by the ␣4T152C or ␤2L146C substitutions (Table 1). Overall, the findings indicate that cysteine substitution of ␣4T152 in the agonist site at the ␣4/␣4 interface or ␤2L146 in the agonist site at the ␣4/␤2-1 interface is tolerated.
Rates of MTSET Reaction with Substituted Cysteines-We tested first whether the substituted cysteines were accessible to MTSET using the protocol depicted in Fig. 2A. Exposure to a saturating concentration (1 mM) of MTSET decreased the ACh   Table  2). In contrast, MTSET decreased the ACh (EC 50-2 , 1.6 mM) responses of ␤2 L146C _␣4_␤2_␣4_␣4 receptors by 71 (87;56)% (mean (95% CI)) ( Fig. 2, C and D, and Table 2). Longer periods of MTSET applications produced no further changes in the ACh responses. Differences in the levels of ACh response reductions induced by MTSET suggest that both types of agonist sites contribute differentially to receptor activation. The ACh responses of wild type receptors were not affected by exposure to MTSET ( Fig. 2D and Table 2), demonstrating that the changes in the amplitude of agonist responses of the mutant receptors were due to MTSET reaction with the substituted cysteines. Exposure to 1 mM MTSET also decreased the responses of ␤2_␣4_␤2_␣4 T152C _␣4 to cytisine (EC 50 ϫ 5 ϭ 70 M) or varenicline (EC 50 ϫ 5 ϭ 12 M) (Fig. 2, E and F), and the extent of the decrease was comparable with that observed for the responses of ACh in this mutant receptor (Table 2). Exposure to the reducing agent dithiothreitol (DTT) (1 mM; 120 s) fully reversed the effects of MTSET on the ACh responses of ␤2_␣4_␤2_␣4 T152C _␣4 receptors. This shows that any modification of endogenous disulfide bonds in ␤2_␣4_␤2_ ␣4 T152C _␣4 receptors by DTT had minimal effects on receptor function. In contrast, DTT (1 mM, 120 s) had no effects on the responses of ␤2 L146C _␣4_␤2_␣4_␣4 receptors (Fig. 2, B and C), further indicating differences between the agonist sites. Longer DTT applications had lethal effects on oocytes.

Normalised Response
Next, we determined the rates of covalent modification of the introduced cysteines by measuring the effect of successive subsaturating applications of MTSET on ACh current responses using the protocol described under "Experimental Procedures." The decrease in ACh current responses was plotted against cumulative duration of MTSET exposure and fit with a single exponential decay curve, which yields a pseudo-first-order rate constant (k 1 ). To correct for the concentration dependence of the rate (21), a second order rate constant (k 2 ) ( Table 3) was calculated by dividing k 1 by the concentration of MTSET used. This correction was needed to compare the rate of MTSET reactions on both types of agonist sites. For both mutant receptors and for all agonists tested, except cytisine on ␤2 L146C _␣4_␤2_␣4_␣4 receptors, the maximal effects of MTSET observed in the rate experiments were consistent with those measured using maximal concentrations of MTSET, indicating that the MTSET reactions went to completion. When cytisine was used to stabilize the ACh responses of ␤2 L146C _␣4_␤2_␣4_␣4 receptors, the maximal effects of MTSET were ϳ2.3-fold smaller than the maximal effect observed when any of the other agonists were tested was used to stabilize the ACh current responses of ␤2 L146C _␣4_␤2_␣4_␣4 receptors (Fig. 3F). Application of 1 mM MTSET at the end of the cytisine protection assay did not increase the plateau (not shown). These findings suggest that prolonged exposure to cytisine produces long term desensitization in a subset of ␤2 L146C _␣4_␤2_␣4_␣4 receptors, which could make these receptors unavailable for MTSET reaction. This possibility is consistent with previous studies, which have shown that repeated exposure of ␣4␤2 nAChRs to submaximal concentrations of cytisine elicits high-affinity, long term desensitization of receptor function (28). For both mutant receptors, the control MTSET reaction rate was not affected by the application of any of the agonists tested, including cytisine on ␤2 L146C _ ␣4_␤2_␣4_␣4 receptors, during the stabilization phase of the rate protocol used (Fig. 3A and Table 3), which proves that any changes in the responses to ACh pulses observed during the protection assays are due to changes in MTSET reaction rates. Comparison of the calculated k 2 values indicated that the fastest reaction of MTSET occurred at ␣4T152C (Table 3), suggesting that this residue is more accessible than ␤2L146C to covalent modification by MTSET.
Agonist Sites Have Different Ligand Selectivity-To determine the ligand selectivity of the agonist site at the ␣4/␣4 and ␣4/␤2-1 interfaces, established (␣4␤2) 2 ␣4 agonists (i.e. ACh, cytisine, varenicline, TC-2559, and sazetidine-A) (29) were tested for the ability to protect the substituted cysteines from MTSET reactions. We reasoned that co-application with agonists should decrease the rate of MTSET reaction with substituted cysteines if the co-applied agonists competed with this reagent for access to the cysteine-substituted agonist sites. In  (11), ACh protected ␣4T152C and ␤2L146C from MTSET reactions (Fig. 3, C and  D). Cytisine and varenicline also protected the agonist sites at the ␣4/␤2 and ␣4/␣4 interfaces (Fig. 3, E-H). Protection could have also occurred if occupation of the agonist sites by ACh, cytisine, or varenicline induced structural rearrangements, resulting in a decreased accessibility of the substituted cysteines to MTSET. To test this possibility, we determined the effect of Dh␤E, a competitive antagonist that is thought not to cause agonist-like conformational changes in the agonist binding site of ␣4␤2 nAChRs (30) and which we have shown previously to bind the agonist sites at the ␣4/␤2 and ␣4/␣4 interfaces (11). In agreement with our previous studies (11), Dh␤E was effective at protecting ␣4T152C from covalent modification by MTSET  Table 1) and Var (EC 50 ϫ 5, 12 M; see Table 1) Table 2 for statistical analysis of data in D and F. The arrows in B and C indicate application of ACh, whereas in E and F, they represent the application of cytisine or varenicline, respectively. Error bars, S.E.  (Table 3). ACh, cytisine, and varenicline differed in their ability to protect both types of agonist sites from MTSET, suggesting that these agonist sites have different affinity for the agonist sites in (␣4␤2) 2 ␣4 nAChRs. This finding is consistent with studies showing that the binding poses of these agonists differ in ␣4␤2 nAChR (5,(7)(8)(9)(10).
In contrast, TC-2559 and sazetidine-A protected only ␤2L146C from MTSET reaction (Fig. 3, I-L, and Table 3), suggesting that these compounds do not bind the agonist site at the ␣4/␣4 interface. Overall, therefore, the protection assays show that the agonist site at the ␣4/␣4 interface excludes some agonists that are capable of engaging the site at the ␣4/␤2-1 interface.

DISCUSSION
We (11) and others (12) have shown previously that the (␣4␤2) 2 ␣4 nAChR contains three functional agonist binding sites for ACh. Binding of ACh to two sites produces effective gating, but engagement of the three agonist sites produces maximal activation. Two of the sites are located at ␣4/␤2 interfaces and are thus structurally identical. The third agonist site is at the ␣4/␣4 interface, where it occupies a position homologous to that of the agonist sites at ␣4/␤2 interfaces. We wanted to investigate whether the agonist sites at ␣4/␣4 and ␣4/␤2 interfaces have differing ligand selectivity and whether such difference could impact agonist efficacy at (␣4␤2) 2 ␣4 nAChRs. Our findings indicate that the agonist site at the ␣4/␣4 interface is a

Rates of covalent modification of cysteine-substituted ␤2_␣4_␤2_ ␣4_␣4 nAChRs
Rate constants (k 2 ) were calculated as described under "Experimental Procedures" and represent the mean value Ϯ S.E. of 3-6 individual oocytes. k 2C /k 2PA represents the ratio of the rates of MTSET reactions obtained in the control rate (k 2C ) and protection assays (k 2PA ). For ␤2_␣4_␤2_␣4 T152C _␣4, the concentrations of agonists used in the protection assays were EC 50 ϫ 5: ACh, 1 mM; Cyt, 55 M; Var, 12 M; TC-2559, 15 M. For ␤2_␣4_␤2_␣4 T152C _␣4 receptors, the concentrations of Var (45 M) and TC-2559 (65 M) were also EC 50 ϫ 5, but for Cyt and ACh, the concentrations were 0.3 M (EC 10 ) and 1.6 mM (EC 50-2 ), respectively. For both mutant receptors, the Saz-A concentration used was 1 M, a concentration that produced maximal effects (EC 100 ) (13). For ␤2_␣4_␤2_␣4_ T152C,H142A ␣4, the concentrations of ACh, TC-2559, and Saz-A pulses were 1.15 mM and 20 M, respectively. For all three mutant receptors shown, statistical differences were determined by Student's t tests or one-way analysis of variance. *, statistical differences between the rate constants determined under control and protected assay conditions (*, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. ϩ , statistical differences between the control rates calculated for the covalent modification of ␣4T152C and ␤2L146C by MTSET ( ϩ , p Ͻ 0.05; ϩϩ , p Ͻ 0.01; ϩϩϩ , p Ͻ 0.001). key determinant of agonist efficacy; occupancy of this site increases agonist efficacy, whereas exclusion from the site leads to partial agonism. This is likely to apply to other pentameric ligand gated ion channels assembled from ␣ and ␤ subunits into a (␣␤) 2 ␣ receptor stoichiometry (e.g. (␣3␤4) 2 ␣3 (18) and heteromeric (␣␤) 2 ␣ glycine receptors (19)). Given that the two stoichiometries of the ␣4␤2 nAChR may express in neurons (31)(32)(33)(34)(35), the findings may encourage a fresh outlook and new impetus in the design of ␣4␤2-targeted compounds. From the observations of 1) the effects of the cysteine substitutions on the sensitivity of the receptors to agonists, 2) the effects of saturating concentrations of MTSET on the amplitude of the current responses to ACh, 3) the extent of DTTdriven recovery from MTSET reactions, 4) the effects of cytisine on the maximal effects of MTSET, and 5) the rates of MTSET reaction and the extent of agonist-dependent protection from MTSET reactions, we confirm that the sites at ␣4/␣4 and ␣4/␤2 interfaces in (␣4␤2) 2 ␣4 nAChRs are functionally non-equivalent (11,17). Functional non-equivalence of the agonist sites may be attributed largely to differences in the complementary side of the agonist sites. Because the principal face is contributed by ␣4 subunits at both types of binding sites, binding interactions between agonists and the principal face of the agonist sites are possibly similar in the two sites. In contrast, the complementary face is contributed by the ␣4 subunit in the ␣4/␣4 interface and by a ␤2 subunit in the ␣4/␤2 interface, which probably provides interface-specific structural elements for agonist site-ligand interactions.
A key finding of our studies is that the agonist site at the ␣4/␣4 interface, but not the agonist sites at ␣4/␤2 interfaces, excludes agonists of a certain size. Comparisons of the MCL and ASA of agonists suggest that the site at the ␣4/␣4 interface is readily occupied by small ligands (e.g. ACh, cytisine, and varenicline). Larger agonists (e.g. TC-2559 or sazetidine-A) cannot enter the site, constraining these compounds to trigger ion channel gating through the agonist sites at ␣4/␤2 interfaces only. A single residue on the complementary face of the site at the ␣4/␣4 interface, ␣4H142, impedes entry to agonists with a critical size, thus acting as a molecular sieve in the agonist site at the ␣4/␣4 interface and, hence, as a key determinant of partial agonism at (␣4␤2) 2 ␣4 nAChR. Given that maximal activation of (␣4␤2) 2 ␣4 nAChRs increases by additional binding to the agonist site at the ␣4/␣4 interface (11,12), exclusion from the agonist site at the ␣4/␣4 interface inexorably leads to partial agonism.
How could structurally identical agonist sites be functionally non-equivalent? Functional non-equivalence could arise by different subunit environments surrounding the agonist binding sites, as suggested for the ␣1␤2␥ 2s GABA A receptor (36). In the case of the alternate ␣4␤2 nAChRs, it may be that functionally relevant interactions between the fifth subunit and neighboring subunits are receptor-specific, which could influence agonist effects. For example, in addition to contributing the primary side of an operational agonist site, the fifth subunit in the (␣4␤2) 2 ␣4 nAChR could influence binding of agonists to the sites at ␣4/␤2 interfaces such that agonist binding poses to those sites are different from those at the (␣4␤2) 2 ␤2 stoichiometry. Some of these interactions could be weakened or strengthened upon agonist binding, and this could influence the ability of occupied receptors to reach the shut states immediately preceding gating (flipped or priming states) (37,39). It has been suggested that full and partial agonists differ in that partial agonists are less effective than full agonists in inducing flipped receptor states (37). The suggestion that the fifth subunit may be directly involved in the coupling of agonist binding to ion channel gating is consistent with recent electron microscopy studies of Torpedo marmorata nAChRs that suggest that displacement of the accessory subunit (␤1), driven by motions of the inner and outer sheets of the ␤ barrel of neighboring agonist binding ␣1 (␣1 ␥ ) induced by agonist binding, plays a greater determining role in gating than hitherto thought (39). We show here that in the case of the (␣4␤2) 2 ␣4 nAChR, the fifth subunit critically shapes the functional properties of this receptor type.
We are unable at this time to propose an adequate kinetic model for the activation/desensitization cycle of the (␣4␤2) 2 ␣4 nAChR by agonists, because single channel data for this receptor are lacking. Nevertheless, on the basis that the (␣4␤2) 2 ␣4 receptor efficiently activates when only two agonist sites are engaged by ACh and that these partially occupied receptors activate and desensitize with affinities comparable with those determined for the (␣4␤2) 2 ␤2 receptor (17), we propose that the (␣4␤2) 2 ␣4 can display (␣4␤2) 2 ␤2-like or (␣4␤2) 2 ␣4 activity, depending on the concentration of ACh. This implies that there are two pathways for the activation of (␣4␤2) 2 ␣4 nAChRs (Fig. 7, A and B). For simplicity, we have omitted from the scheme shown in Fig. 7, A and B, open and desensitized recep-   Table 2 for statistical analysis of the data. D and E, rates of reaction of MTSET at ␤2_␣4_␤2_␣4 T152C,H142A _␣4 nAChRs in the presence and absence of TC-2559 or Saz-A. TC-2559 and Saz-A were applied during the stabilization of the responses to ACh to correct for any process of desensitization that may have occurred during the assay. Normalized ACh currents in the absence and presence of TC-2559 (D) or sazetidine-A (E) were plotted versus cumulative time of MTSET and fit with single exponential functions (see "Experimental Procedures"). Data points were normalized to ACh currents at time 0 and are the mean Ϯ S.E. of at least four experiments. The concentration of TC-2559 and Saz-A used during both the stabilization of the EC 50 ϫ 5 ACh (1.15 mM) responses and the protection assays was 20 M.
In summary, we show that non-equivalent agonist selectivity of the agonist sites of the (␣4␤2) 2 ␣4 nAChR constrains some agonists to trigger receptor activation through binding to agonist sites at ␣4/␤2 interfaces. Because full receptor occupation induces maximal receptor activation, exclusion from the agonist site at the ␣4/␣4 interfaces reduces agonist efficacy. This work gives further insights into how agonist efficacy is determined in pentameric ligand gated ion channels.