Ligand Binding at the α4-α4 Agonist-Binding Site of the α4β2 nAChR Triggers Receptor Activation through a Pre-Activated Conformational State

The α4β2 nicotinic acetylcholine receptor (nAChR) is the most abundant subtype in the brain and exists in two functional stoichiometries: (α4)3(β2)2 and (α4)2(β2)3. A distinct feature of the (α4)3(β2)2 receptor is the biphasic activation response to the endogenous agonist acetylcholine, where it is activated with high potency and low efficacy when two α4-β2 binding sites are occupied and with low potency/high efficacy when a third α4-α4 binding site is occupied. Further, exogenous ligands can bind to the third α4-α4 binding site and potentiate the activation of the receptor by ACh that is bound at the two α4-β2 sites. We propose that perturbations of the recently described pre-activation step when a third binding site is occupied are a key driver of these distinct activation properties. To investigate this, we used a combination of simple linear kinetic models and voltage clamp electrophysiology to determine whether transitions into the pre-activated state were increased when three binding sites were occupied. We separated the binding at the two different sites with ligands selective for the α4-β2 site (Sazetidine-A and TC-2559) and the α4-α4 site (NS9283) and identified that when a third binding site was occupied, changes in the concentration-response curves were best explained by an increase in transitions into a pre-activated state. We propose that perturbations of transitions into a pre-activated state are essential to explain the activation properties of the (α4)3(β2)2 receptor by acetylcholine and other ligands. Considering the widespread clinical use of benzodiazepines, this discovery of a conserved mechanism that benzodiazepines and ACh potentiate receptor activation via a third binding site can be exploited to develop therapeutics with similar properties at other cys-loop receptors.


Introduction
Nicotinic acetylcholine receptors (nAChRs) are members of the cys-loop receptor superfamily of ligand-gated ion channels (LGICs) and form a pentameric subunit arrangement around a defining the efficacy of an agonist, giving a framework to explain how partial agonists activate receptors with low efficacy.
Modulators of cys-loop receptors include many classes of compounds that are used either clinically or in research, including barbiturates and benzodiazepines that modulate GABA A Rs, and compounds such as estradiol and PNU-120596 that modulate nAChRs. These modulators bind to the receptor complex and, in combination with an agonist, alter the conformational transitions between states to ultimately favour an open state, potentiating the response to the agonist. To achieve this, different steps in the activation pathway can be altered by the modulator depending on the location of their binding site and the conformational change induced to the receptor. For instance, PNU-120596 is a positive modulator of α7 nAChRs that primarily decreases transitions into the desensitized state [29], while diazepam modulates GABA A Rs by increasing transitions into a pre-activated state [30]. These different mechanisms of modulation lead to differences in the pharmacological properties that underlie the physiological response to a modulator.
Typically, activation of Cys-loop receptors has been studied by developing kinetic models of activation and fitting data from high-resolution single-channel recordings to these models. Acquiring single-channel recordings from α4β2 nAChRs has been difficult, but an alternative method using simpler linear models and whole-cell electrophysiological recordings has recently been described to explain how benzodiazepines potentiate GABA activation of GABA A Rs through binding at a third site homologous to the GABA-binding site [30]. Similar to benzodiazepines, both ACh and NS9283, a ligand that augments cognitive function in rodents, bind at the third α4-α4 binding site of 3α4:2β2 nAChRs to potentiate receptor response, and may increase receptor activation via a similar mechanism.
Here we delineate currents elicited from α4-β2 (Sazetidine-A or TC-2559) and α4-α4 (NS9283) activation. By developing two linear kinetic models that can differentiate the transitions between receptor states, we show that the increased efficacy of activation when a third agonist (NS9283) is bound is due to a shift in the equilibrium from the ligand-bound closed state to a pre-activated or "flip" state. We propose that ligands that bind at a third homologous binding site will potentiate receptor activation by the same mechanism.

Molecular Biology
cDNA encoding human nAChR α4 and β2 subunits were cloned, and cRNA was prepared as previously described [31]. A triple-point-mutated β2 m subunit with the mutations V136H, F144Q, and L146T was constructed as previously described [17].

Oocyte Preparation and Electrophysiology
This study was carried out in strict accordance with the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes. The protocol for this specific study was approved by the Animal Ethics Committee of the University of Sydney (Protocol number: 2013/5915). All surgical procedures were carried out in under anesthesia induced by immersion in a tricaine solution to ameliorate suffering. All Xenopus were monitored during surgery and for the following 5-7 days to ensure both anesthesia and the surgery did not cause adverse effects. Mature female Xenopus laevis frogs were anaesthetised with 0.17% tricaine (buffered with 0.06% sodium bicarbonate) for 15 minutes, after which the loss of righting reflex was confirmed before transferring on to ice where surgeries were performed. A small (1-2 cm) abdominal incision was made through both the skin and muscle layer with surgical knives. Ovary lobes were removed with a pair of forceps, and kept in oocyte releasing 2 (OR2) buffer (82.5 mM NaCl, 2 mM KCl, 1 mM MgCl 2 , 5 mM HEPES hemisodium; pH 7.4). The skin and muscle layer were sutured separately, and frogs were allowed to recover for six months before they were reselected for surgeries. A total of five recoverable surgeries with 6-12 months recovery periods between surgeries were performed on each frog to reduce the adverse effects of Xenopus transportation into Australia, as approved by the Animal Ethics Committee of the University of Sydney (Protocol number: 2013/5915) before a terminal surgery was performed, in which a lethal dose of tricaine (0.5%) was used. Briefly, lobes from ovaries of female adult X. laevis were removed and defolliculated to obtain isolated oocytes. Oocytes were injected with a total of *25 ng of cRNA encoding human WT receptor as shown earlier to express 3α4:2β2 and 2α4:3β2 [17], and incubated for 2-5 days at 15-18°C.
Oocytes were subjected to two-electrode voltage-clamp electrophysiological testing using a custom-built system where solutions were applied directly to the oocytes via a glass capillary tube placed in the vicinity of the cell. This ensured a solution exchange rate faster than two seconds. Ligand was applied for 30 sec and peak current amplitudes were measured. To minimise the effects of long desensitized states, the number of applications of ligand at each oocyte was minimized rather than performing full concentration-response curves on individual oocytes.

Data Analysis
Concentration-response curves for Sazetidine-A and TC-2559 were fitted using GraphPad Prism 6 to a monophasic Hill equation with Hill Slope of 1: Where I max is the maximum current, EC 50 is the concentration that produces the half-maximal response and [A] is the concentration of ligand.
Concentration-response curves for ACh were fitted using GraphPad Prism 6 to a biphasic Hill equation with Hill slopes of 1: Where Frac1 and Frac2 are the fractions and EC 50(1) and EC 50 (2) are the half-maximal concentrations of high and low-sensitivity phases of the concentration-response curves, respectively.

Results
To understand how additional ligand binding at the α4-α4 site enhances receptor activation compared to when only two α4-β2 sites are occupied, we used simple linear kinetic models to interpret the data. This requires ligands that selectively bind at the α4-β2 and α4-α4 sites to activate the receptor.

Saz-A and TC-2559 Selectively Activate via the α4-β2 Interface
To determine if Saz-A and TC-2559 were indeed activating the receptor exclusively through binding at the α4-β2 sites, concentration-response relationships (CRR) of Saz-A and TC-2559 were constructed for wild-type (WT) 2α4:3β2 and 3α4:2β2 receptors. Saz-A was a full agonist and TC-2559 acted as a super agonist at 2α4:3β2 receptors with maximum responses of 110% and 360%, respectively, when normalised to the response of a saturating concentration of ACh (1 mM), although the maximum response to TC-2559 was difficult to estimate as the top of the concentration-response curve may not have been reached (Table 1, Fig 1). However, Saz-A and TC-2559 were both partial agonists at 3α4:2β2, eliciting a maximum response of 19% and 49% respectively, ( Table 1, Fig 1D), consistent with these molecules binding at only two α4-β2 sites to activate the receptor, in comparison to ACh that binds at three (two α4-β2 and an α4-α4). Furthermore, both ligands exhibited similar potencies at 3α4:2β2 and 2α4:3β2 receptors with EC 50 values of 23 and 10 nM for Saz-A and a 5-fold difference with EC 50 values of 3300 and 642 nM for TC-2559, respectively, (Table 1) suggesting that both receptors were activated via identical sites. Both Saz-A and TC-2559 do not seem to affect the recovery of the current after application (Fig 1B and 1C). Nevertheless, a contribution to Saz-A and TC-2559 activation from binding at the α4-α4 interface cannot be excluded from these experiments. Therefore, Saz-A and TC-2559 were applied to mutated 3α4:2β2 HQT receptors where the α4-β2 binding site is abolished. The β2 HQT construct was designed such that the complementary (-) side of the agonist binding pocket resembled the α4(-) face so that the receptors contain only α4-α4-like binding sites [11,17,32]. Neither Saz-A or TC-2559 activated 3α4:2β2 HQT mutant receptors (Fig 1D and 1E; Table 1), demonstrating that Saz-A and TC-2559, at the concentrations tested, bind only at α4-β2 interfaces to activate 3α4:2β2 nAChRs and can therefore be used to mimic ACh activation when two α4-β2 binding sites are occupied.
Delineating 3α4:2β2 Receptor Activation from α4-β2 and α4-α4 Binding Interfaces NS9283 selectively binds to the α4-α4 binding site to increase receptor activation [6]. 3α4:2β2 receptor activation from α4-β2 and α4-α4 binding interfaces can then be obtained by co- Experimentally derived values are obtained from concentration-response relationship curve fitted to a hill co-efficient of one using prism 6. 1 Mean EC 50s values and 95% confidence interval derived from the curve-fitting are shown in the bracket. 2 Mean I max normalized 1 mM ACh with ± SEM derived from the curve-fitting is shown. 3 Estimated P o,max ± SEM is obtained by normalizing with 1 mM ACh + NS206 (10 μM).
applying NS9283 with Saz-A or TC-2559 (Fig 2). Co-application of Saz-A (1 μM) and TC-2559 (10 μM) with NS9283 (10 μM) had peak currents comparable to max ACh (1 mM) application, with no significant effect on current recovery after drug application that suggests no effect on desensitization ( Fig 2B and 2C). NS9283 potentiation curves were generated by co-applying 10 μM NS9283 with varying concentrations of Saz-A and TC-2559 to WT 3α4:2β2 nAChRs ( Fig 2D and 2E). For Saz-A, co-application of NS9283 (10 μM) increased the maximum response by 4.4-fold and decreased the EC 50 value by 2-fold (Table 1), while for TC-2559 the maximum response was increased by 2.9-fold and the EC 50 value was decreased 3.1-fold (Table 1). Thus, NS9283 potentiation of Saz-A or TC-2559 resulted primarily in an increased efficacy of receptor activation, that is comparable to the increase observed when a third molecule of ACh binds to the α4-α4 interface. Therefore, we used Saz-A or TC-2559 in combination with NS9283 to determine how binding at the third α4-α4 site increases activation levels of the receptor.

Estimating Maximum Open Probability
Ideally, single-channel recordings would be used to estimate the P o,max values; however this is technically highly challenging for α4β2 nAChRs due to channel rundown and desensitization [33]. To overcome this, an alternate method utilizing co-application of ACh and NS206, a positive allosteric modulator (PAM), was used to estimate the maximum open probability (Est. P o,max ) at 3α4:2β2 receptors. NS206 binds at an alternative site to NS9283, increasing the maximum response of the receptor above the maximum response to ACh [19,34]. We have assumed that when both the positive allosteric modulator and native ligand ACh are bound, the probability of the receptor being opened is approaching one, and was previously successfully used for other members of LGICs [35]. All other responses are then expressed relative to this maximum response (Fig 3).  The application of 10 μM NS206 increased the response of the receptors to either 10 μM or 1 mM ACh. When co-applied with high concentrations, NS206 typically elicited a second smaller peak when the solution was washed off (Fig 3B). This is likely due to a combination of desensitized states and the binding kinetics of both ACh and NS206. The mean peak current amplitude of ACh max (1 mM) was increased by a factor of 2.0 (n = 7) when co-applied with NS206 for 3α4:2β2 receptor, allowing for an estimate of maximum open channel probability, while 10 μM NS206 increased the response to 10 μM ACh by nearly 4-fold ( Fig 3C). Thus, the relative open probability for a saturating concentration of ACh is only half the Est. P o,max obtained with co-applied NS206. The Est. P o response curves for the partial agonists (Saz-A and TC-2559) alone or when co-applied with NS9283 were obtained by transforming the maximum responses normalized to 1 mM ACh to an Est. P o,max (Fig 3D and 3E; Table 1). The resulting P o,max for Saz-A and TC-2559 that were used to calculate the values of the equilibrium constants were 9.1% and 20%, respectively, and in the presence of NS9283 these values increased to 39% and 59% (n = 6), respectively (Fig 3D and 3E; Table 1).
An important caveat to this approach is that the accuracy of the Est. P o is dependent on response of the potentiator and ACh having an open probability of 1 [36]. Therefore, the Est.

Models of α4β2 Receptor Activation
To determine how binding at the α4-α4 site alters receptor transitions during activation, two simple linear kinetic models were used to describe receptor activation when two α4-β2 binding sites are occupied. For simplicity the binding of two ligands at α4-β2 sites is combined into a single step. The first model (Model 1) is a three-step model designed to determine whether the third ligand alters: (i) the binding affinity of ligands at the α4-β2 site (cooperativity); (ii) the transitions between closed and open states (gating) or (iii) the transitions between open and a distinct closed state in the presence of ligand (desensitization). The second model (Model 2) expanded the gating transitions to investigate whether a pre-activation (flipping) step is altered. We then determined which equilibrium constant was principally modulated in these models when NS9283 bound to the α4-α4 site potentiated the Saz-A or TC-2559 responses.
Model 1. The open probability for the first model is given by ( Fig 4A): Where E is the gating constant, [A] is the agonist concentration, K is the dissociation constant of Saz-A or TC-2559, and D is the desensitizing constant. By rearranging Eq 1 to isolate [A]: The open probability curve will fit to a Hill equation with a slope coefficient of one, since the ligand binding steps have been simplified by aggregating into a single transition. It follows from Eq 2 that the maximum open probability (when [A] approaches infinity) is given by: and the half-maximum effective concentration is: We first hypothesized that a third ligand (NS9283) bound at the α4-α4 site, NS9283, increases the gating transitions (E) in Model 1 by a constant multiplication factor C E (Fig 4; Scheme 1). The open probability in the presence of NS9283 (P NS o:max ; the superscript denotes when NS9283 is co-applied) and half-maximum effective concentration (EC NS 50 ) can be described by substituting E with C E .E in Eqs 3 and 4 respectively. The value of C E can then be calculated from the ratios of the EC 50 values and the maximum open probabilities in control and when potentiated by NS9283, as follows: Using Eq 7, the values of the constant 'C E ' calculated from the change in EC 50 and P o,max values when NS9283 was applied in addition to Saz-A and TC-2559 were 8.7 and 9.2 respectively ( Table 2). As NS9283 binds selectively to the α4-α4 site, we assumed that it has the same multiplier effect; regardless of which partial agonist is bound at the α4-β2 site. Hence, the values of C E for Saz-A and TC-2559 were averaged to 8.94. Equilibrium constants were then calculated using the following equations. Rearranging Eq 3: An equivalent expression in the presence of NS9283: By subtracting Eq 8 from Eq 9.
Rearranging to give D.
Rearranging Eq 3 to give E.
Rearranging Eqs 5 and 6 to give K The values obtained for D, E and K ( Table 2) were substituted into Eq 1 to predict the P o at a given concentration of Saz-A or TC-2559. The predicted P o response curve for both Saz-A and TC-2559 co-applied with NS9283 was derived by substituting the same values into Eq 1, with the exception that E was substituted with C E .E. The predicted P o response curves describe the experimentally derived data points well (Fig 4C). Further, when the values for D, E and K were determined using an Est. P o for 1 mM ACh of 0.4 or 0.45 and substituted into Eq 1, the predicted P o response curves also describe the experimentally derived data points well, demonstrating that the exact value for P o is not necessary (S1 Fig). We next explored the hypothesis that NS9283 decreases the desensitization constant (D) by a constant factor C D (Fig 4B; Scheme 2). The maximum open probability and the EC 50 for this scheme are: The ratio of EC 50 and P o,max values gives: This relationship is not supported by the experimental data, where NS9283 causes a decrease in the EC 50 value and an increase in the P o,max . Thus, the ratio of EC NS 50 and P NS o;max values will be smaller than in control. With Saz-A the ratio in control is 253 nM, but only 29 nM when potentiated by NS9283, while with TC-2559 the ratio is 3.1 μM, but 336 nM when potentiated by NS9283. Hence, it is unlikely that NS9283 elicits its action by a decrease in desensitization. Finally, we investigated whether NS9283 binding to the α4-α4 site alters the binding equilibrium (K) of Saz-A or TC-2559 by a constant factor C K (Fig 4B; Scheme 3). However, the P o,max does not depend on the value of K (Eq 3) and is predicted to remain unchanged when K is modulated. This is inconsistent with the experimental data and was rejected as a possibility.
Taken together, this model predicts that a third ligand increases gating transitions from the closed to the open state to potentiate receptor function, primarily increasing the efficacy of the response. However, the gating step in this model is likely to be an oversimplification of the actual gating transitions. Transitions into a pre-activated intermediate state has been essential to explain the efficacy of partial agonists and the actions of modulators at other LGICs [27,28,30]; Hence we chose to test whether transitions into a pre-activated state were altered when NS9283 was bound at the α-α interface.
Model 2. This model expanded the gating step from Model 1 to include a pre-activated closed state between the ligand-bound and open states and hence serves as an extension of Model 1 (Fig 5).
Since desensitization was found to be unnecessary to describe the action of NS9283 in Model 1, it was not included in this model for simplicity. Therefore, Model 2 is a three-step model with a binding, pre-activated and gating step equivalent to that used by Gielen et al (2012) to describe the potentiation of GABA A Rs by diazepam. The open probability for this model is given by the following: Where K is the dissociation constant, F is the pre-activating equilibrium constant and E is the gating constant. It also follows that the maximum open probability is given by: Thus, both F and E will have an influence on the observed macroscopic efficacy. The halfmaximum effective concentration is: A Third Ligand Bound at the α4-α4 site Regulates Pre-Activating Transitions in Model 2 We investigated the possibility that NS9283 increases the pre-activation constant, F, by a factor, C F , in Model 2 ( Fig 5B; Scheme 2). By substituting F with C F .F in Eqs 18 and 19, the value of C F can be determined numerically from the equation: The equilibrium constants K, F and E can then be calculated using the following equations (Gielen et al., 2012). Rearranging Eq 18: An equivalent expression in the presence of NS9283: By subtracting Eq 21 from Eq 22 and re-arranging in terms of E: The value of E can then be calculated using Eq 23. F was calculated by rearranging Eq 18, similar to Eq 12 above, and K can then be calculated by substituting C F into Eq 13 and multiplying the right-hand side by F. From Eq 20, the value of C F was calculated for NS9283 potentiation of Saz-A and TC-2559 activation was calculated, averaged to 8.94 (as discussed in Model 1) and used to determine the values for E, F and K ( Table 2). These values were then used to predict the P o from Eq 17 for a given concentration of agonist for Saz-A and TC-2559 with and without NS9283. The predicted P o response curves correspond well with the experimental data (Fig 5C and 5D). Thus, the mechanism of NS9283 potentiation of Saz-A and TC-2559 could be refined by an increase in the equilibrium of the pre-activating step (Fig 4C and 4D).
The final hypothesis to be tested was that NS9283 increases the gating constant E of Model 2 by the constant C E (Fig 5B; Scheme 3). Modulation of E is predicted to be affect both the maximum response, P NS o;max , and the half-maximum concentration, EC NS 50 . The value of C E can be calculated as before from Eq 17, and the average value of 8.94 was used. The potentiation of the EC 50 and P o,max are described by: which shows that the EC 50 and P o,max of the NS9283 curves are dependent only on the values of C E , the agonist potency (EC 50 ) and the P o,max and not on the specific values of E, F and K (Table 2). Notably, the same equations would be derived for a simple Castillo-Katz mechanism with a single binding step and gating transition. The predicted relative response curves describe a shift in the EC 50 and P o,max with NS9283 potentiation, with predicted EC 50 values of 13 nM and 140 nM, similar to the measured values of 11 nM and 195 nM for Saz-A and TC-2559, respectively. However, the maximum response overshoots the experimental data, predicting a P o,max of 47% and 69% in comparison to the measured values of 39% and 59% for Saz-A and TC-2559, respectively (Fig 4C and 4D). Further, when the values for K, F and E were determined using an Est. P o for 1 mM ACh of 0.4 or 0.45 and substituted into Eq 17, the predicted P o response curves also describe the experimentally derived data points well when the pre-activated constant F, rather than the gating constant E, was altered (S1 Fig). Taken together, this suggests that the increase in gating transitions observed in Model 1 is primarily due to an increase in transitions into the pre-activated state, with the gating transitions unchanged, when the third α4-α4 site is occupied to primarily increase the efficacy of the response.

Discussion
Here, we demonstrate that binding of an agonist to the α4-α4 binding site of the 3α4:2β2 receptor increases the equilibrium constant favouring transitions to the pre-activated, 'flip' state. We used site-selective ligands coupled with two distinct linear kinetic models to interrogate how the activation mechanism is altered when the third α4-α4 agonist-binding site is occupied in addition to the two α4-β2 sites.
When NS9283 bound to the α4-α4 site, it enhanced both the potency and efficacy of Saz-A or TC-2559 activation via the α4-β2 sites, and the concentration-response curves could be described by the alteration of just one transition in our kinetic models. Both these models indicate that when the third α4-α4 site is occupied by NS9283, the equilibrium constants defining the transitions from the closed to the open states are increased. Further, the second model incorporating a pre-activated step indicates more specifically that the transitions to the preactivated state are increased. Although it would be ideal to obtain P o,max from single channel recordings, the changes in the Est. P o,max did not make any differences to the conclusions that were reached and a similar method to estimate open probability with an allosteric modulator, has been used for other cys-loop receptors to compare the magnitude of an allosteric shift between different receptor subtypes [35,36].
The α4β2 nAChR is considered to be an important target for treating schizophrenia, attention deficit hyperactivity disorder (ADHD) and pain; and several drugs have been shown to increase pro-cognitive and analgesic effects in pre-clinical as well as clinical trials [3][4][5][6]. However, only one drug, varenicline, has come to the market in recent years that is used for smoking cessation [37]. A key aspect of drug development is the selectivity for specific receptor subtypes that is complicated by the fact that α4β2 exists in two different stoichiometries, the 2α4:3β2 and 3α4:2β2. We and others have shown that the presence of an α4-α4 agonist interface in the 3α4:2β2, in addition to the α4-β2 interface present in both stoichiometries, greatly change the pharmacological properties of the receptor [11,17,18]. Because of this additional agonist binding site, several agonists (e.g. ACh, epibatidine), partial-agonists (e.g. NS3573), antagonists (e.g. methyllycaconitine) and modulators (e.g. Zn 2+ ) act differently on the two stoichiometries [15,18,19,38]. Further, ligands that selectively act on the α4-β2 and the α4-α4 agonist binding sites like Saz-A [8] and NS9283 [19,34] respectively, provide us with a great tool to tease out signal transduction initiated from specific agonist binding sites, along with serving as lead molecules in structure activity relationship (SAR) studies that enable us to target the two stoichiometries more specifically.
While traditionally positive allosteric modulators have been defined as ligands that increase the response of the native agonist without activating the receptor themselves, NS9283 binds to the same site as the native agonist ACh. This has recently led NS9283 to be referred to as either an agonist [39], or a site-selective agonist [20] rather than as a positive allosteric modulator. Mechanistically NS9283 is behaving like an agonist, but cannot activate receptor when only one binding site is occupied. Targeting such a site can present advantages for therapeutics, as compounds that bind at these sites only increase the response to the native ligand, rather than activate all available receptors, potentially reducing the side-effects of the drug [40].
Rather than being a property of the closed to open transition as previously supposed, partial agonism in the Cys-loop receptor family can be ascribed to intermediate preactivation step that has been termed a 'flip' or a 'prime' state [27,28,41,42]. Single-channel recordings fitted with mechanisms that include a pre-activated state have demonstrated that the low efficacy activation of Cys-Loop receptors such as GlyRs by taurine, the muscle nAChR by tetramethylammonium [27] and 5-HT 3A Rs by tryptamine [42] is determined by the affinity of the ligand to the pre-activated states. These discoveries that the efficacy of an agonist is in large part determined by perturbations of a pre-activated state thus have implications for drug development. Ligands that are targeted to the interface of extracellular domains, such as diazepam and NS9283, are likely to potentiate the response of the native agonist through favouring transitions into the pre-activated state. This will manifest itself primarily through an increase in both potency and efficacy of the native agonist, but the magnitude of these increases will be dependent on the intrinsic efficacy of the agonist. The combined knowledge of the intrinsic efficacy of the native ligand, the binding sites that distinct stoichiometries make available and the likely mechanism of action of a ligand binding at putative binding site can then be used in the early stages of drug development to predict the pharmacological effects of a drug at the receptors of interest.
With this understanding of how the binding of a third agonist at the α4-α4 site, modulates receptor activity, we can also propose a plausible kinetic model of ACh activation for the 3α4:2β2 receptor (Fig 6). In this scheme, the binding of two molecules of ACh at the α4-β2 site, similar to Saz-A or TC-2559, would elicit a conformational change of the receptor into a preactivated state, whereby the receptor conformation can then transition into the open state. When the third molecule of ACh binds to the α4-α4 site, similar to NS9283, the conformational change of the receptor into the pre-activated state becomes much more likely, driving an increased probability of the receptor being in the open state. When measuring macroscopic responses, the combination of the higher binding affinity of ACh at the α4-β2 binding sites, and the increased efficacy when three binding sites are occupied, leads to biphasic concentration-response curve where ACh is essentially a partial agonist at low concentrations. After prolonged periods in the open state, the receptor transitions to a desensitized state that is resistant to activation by agonist binding. This pre-activated state described in the current study can nAChRs Are Activated via a Pre-Activated State well be a representation of multiple intermediate states that make the 'conformational wave' that is represented in a single state. This model is supported by activation mechanisms of other LGICs, including the glycine receptor (GlyR), where the pre-activated state has been detected in single channel recordings [27]. In these models, both the nature of the agonist and the number of agonist molecules bound alter transitions into the pre-activated states, while the transition into the final gating step is a fast, stereotyped reaction that has little dependence on the nature of the agonist [27,41]. We did not consider channel openings from un-liganded and mono-liganded receptors in our linear kinetic model, and without measurements from single channel recordings we can only speculate on their importance. Undoubtedly these states exist, albeit extremely rarely, and we cannot make any conclusions on their relevance from macroscopic recordings [43,44]. Ideally, we would be able to directly test this model of ACh-activation by fitting data generated from single-channel recordings to our scheme and estimating the equilibrium rate constants for kinetic schemes. However, measuring single-channel recordings from α4β2 nAChRs is problematic due to channel rundown and desensitization, which prevents reliable estimates of the open probability (P o ), and the values of equilibrium rate constants from being derived [45]. Despite this, simulations using macroscopic data combined with simple kinetic models can be a useful tool to understand allosteric receptor activation, even considering for the limitations of modelling macroscopic data where receptors are in different conformational states at any given time during agonist application [25,46]. This is best exemplified by the discovery that benzodiazepines modulate transitions into the pre-activated state at γ-aminobutyric acid (GABA A Rs) using macroscopic data from Xenopus oocyte electrophysiology [30].
In summary, we have proposed and tested plausible activation models for the 3α4:2β2 nAChR isoform. By systematically analysing macroscopic data with two models, we demonstrate that when a third agonist molecule is bound at the α-α interface, transitions into the preactivated state are favoured to increase the activation levels, and ultimately the efficacy of the response. Considering the widespread clinical use of benzodiazepines, this discovery of a conserved mechanism that benzodiazepines and ACh potentiate receptor activation via a third binding site can be exploited to develop therapeutics with similar properties at other cys-loop receptors.