Two Distinct Allosteric Binding Sites at α4β2 Nicotinic Acetylcholine Receptors Revealed by NS206 and NS9283 Give Unique Insights to Binding Activity-associated Linkage at Cys-loop Receptors*

Background: Positive allosteric modulators (PAMs) of α4β2 nicotinic acetylcholine receptors have significant therapeutic potential. Results: Two PAMs, NS206 and NS9283, were observed to have differential and additive pharmacological actions due to binding at distinct receptor sites. Conclusion: Modulator binding activity is linked to the specific binding position in Cys-loop receptors. Significance: Diverse PAM profiles increase possibilities for rational drug design and understanding of receptor function. Positive allosteric modulators (PAMs) of α4β2 nicotinic acetylcholine receptors have the potential to improve cognitive function and alleviate pain. However, only a few selective PAMs of α4β2 receptors have been described limiting both pharmacological understanding and drug-discovery efforts. Here, we describe a novel selective PAM of α4β2 receptors, NS206, and compare with a previously reported PAM, NS9283. Using two-electrode voltage-clamp electrophysiology in Xenopus laevis oocytes, NS206 was observed to positively modulate acetylcholine (ACh)-evoked currents at both known α4β2 stoichiometries (2α:3β and 3α:2β). In the presence of NS206, peak current amplitudes surpassed those of maximal efficacious ACh stimulations (Emax(ACh)) with no or limited effects at potencies and current waveforms (as inspected visually). This pharmacological action contrasted with that of NS9283, which only modulated the 3α:2β receptor and acted by left shifting the ACh concentration-response relationship. Interestingly, the two modulators can act simultaneously in an additive manner at 3α:2β receptors, which results in current levels exceeding Emax(ACh) and a left-shifted ACh concentration-response relationship. Through use of chimeric and point-mutated receptors, the binding site of NS206 was linked to the α4-subunit transmembrane domain, whereas binding of NS9283 was shown to be associated with the αα-interface in 3α:2β receptors. Collectively, these data demonstrate the existence of two distinct modulatory sites in α4β2 receptors with unique pharmacological attributes that can act additively. Several allosteric sites have been identified within the family of Cys-loop receptors and with the present data, a detailed picture of allosteric modulatory mechanisms of these important receptors is emerging.

the present data, a detailed picture of allosteric modulatory mechanisms of these important receptors is emerging.
In the search for new drugs targeting nicotinic acetylcholine receptors (nAChRs), 2 positive allosteric modulators (PAMs) have recently attracted interest due to their proposed advantages over receptor agonists, such as less tonic activation, desensitization, and up-regulation of receptors (1)(2)(3). Due to the important role of ␣4␤2 nAChRs in normal brain function and several disorders of the central nervous system (4 -7), ongoing drug discovery efforts seek to identify novel PAMs acting at these receptors. Indeed recent preclinical studies with NS9283 have shown early promise in the treatment of pain (8) and cognitive disorders (9). However, as only few ␣4␤2 nAChR PAMs have been described in the literature to date (1), overall understanding of allosteric modulation at this receptor subtype is sparse relative to other prominent Cys-loop receptors such as the ␣1␤2␥2 ␥-aminobutyric acid type A receptor (GABA A R) or the homomeric ␣7 nAChR.
When considering Cys-loop receptors broadly, it is clear that the variously described allosteric modulators have distinct pharmacological actions. For the ␣1␤2␥2 GABA A R, benzodiazepines increase GABA potency without affecting maximal GABA-evoked peak current levels (E max (GABA)) (10). By contrast, barbiturates can modulate maximal efficacy to levels above E max (GABA) in addition to modulating GABA potency (11). For ␣7 nAChRs, two broad classes of PAMs have been identified exemplified by NS1738 and PNU-120596 (12,13). Pharmacologically, NS1738 increase ACh-evoked peak current levels above E max (ACh) without otherwise affecting current waveforms, whereas PNU-120596 additionally delay desensitization resulting in large current waveform changes. Thus, it appears that at least three different modulatory phenotypes can be observed in the Cys-loop receptor family: (i) left-shift of the agonist CRR, (ii) increased peak current levels above E max of the endogenous agonist, and (iii) delay of desensitization. Depending on the compound and receptor, these actions can be observed independently or in combination.
The binding sites for Cys-loop receptor allosteric modulators have been identified in a number of cases. In GABA A Rs, the ␣␥-subunit extracellular domain (ECD) interface contains the binding site for benzodiazepines in a position homologous to the ␤␣-subunit agonist binding sites for GABA (10). Similar non-orthosteric binding sites in the nAChR ECD have also been proposed for galanthamine (14) and the ␣3␤4 nAChR modulator morantel (15). On the other hand etomidate has been shown to bind in the transmembrane domain (TMD) of GABA A Rs (16). In fact a cavity corresponding to the etomidate binding pocket has been suggested to be conserved across Cys-loop receptors, constituting the binding site for PU02 in 5-hydroxytryptamine type 3A receptors and ivermectin in the invertebrate GluCl (17,18). Similarly, the ␣7 nAChR PAMs NS1738 and PNU-120596 have been proposed to bind in an ␣7 TMDassociated binding pocket (19,20).
With respect to PAMs of ␣4␤2 heteromeric receptors, further complicating issues relate to existence of variant stoichiometries (21)(22)(23)(24). One stoichiometry, with a subunit ratio of 2␣:3␤, has relatively high sensitivity to the endogenous ligand ACh (EC 50 ϳ 1 M), whereas the other, with a subunit ratio of 3␣:2␤, shows a biphasic response to ACh (EC 50 ϳ1 and 70 M). The low sensitivity component of the biphasic response was recently shown to arise from ACh binding to a third binding site in the ␣␣-interface, which is only present in 3␣:2␤ receptors (24). Interestingly, such receptor stoichiometry differences can greatly affect PAM actions as illustrated by findings that two ␣4␤2 PAMs, Zn 2ϩ and NS9283, are both completely specific for 3␣:2␤ receptors (9,25). No PAMs have yet been shown to modulate the 2␣:3␤ stoichiometry ␣4␤2 receptors, an unfortunate fact given their potential importance in mediating dopamine release in the striatum (26).
In the present study, we describe a novel ␣4␤2 nAChR PAM, NS206, which modulates ACh-evoked responses at both common ␣4␤2 receptor stoichiometries (i.e. 2␣:3␤ and 3␣:2␤ receptors). Pharmacological actions of NS206 and the previously described PAM, NS9283, are compared and their binding sites mapped to specific receptor regions. The two compounds had differing modulatory actions where NS9283 gave a left shift of the agonist CRR, whereas NS206 predominantly increased agonist peak current levels and, interestingly, these modulatory actions proved additive when the compounds were co-applied. Aligning pharmacological actions for a range of PAMs (literature and current study) with their proposed binding sites, it appears that binding in ECD interfaces mostly result in agonist CRR left shifts, whereas TMD binding can further result in increased agonist-evoked peak currents, delayed desensitization characteristics, and direct agonist independent receptor activation. -6,7,8,9-tetrahydro-1H-benzo[g]indole-5-sulfonamide (NS206) was synthesized at NeuroSearch A/S according to methods described in patent WO 01/55110. 3-(3-(Pyridine-3-yl)-1,2,4-oxadiazol-5-yl)benzonitrile (NS9283) was likewise synthesized at Neuro-Search A/S as described by Timmermann et al. (9). The structures of NS206 and NS9283 were confirmed using MS and NMR and both compounds are of Ͼ98% purity. ACh (A9101) and all other chemicals were of analytical grade and purchased from Sigma unless otherwise specified.

Materials-3-N-Benzyloxy-3-hydroxyimino-2-oxo
Molecular Biology-Human cDNA of the ␣3, ␣4, ␤2, and ␤4 nAChR subunits were cloned and inserted into in-house plasmid expression vectors as previously described (12). Creation of a point-mutated ␣4-subunit with mutations H142V, Q150F, and T152L (numbering corresponds to the uniprot canonical sequence of the nAChR ␣4-subunit) have been described previously (24). Chimeric subunits, where the ECD of e.g. ␣4 was substituted with that of ␣3 or vice versa, were described previously (27). Briefly, these subunits were constructed by introducing a SalI site immediately upstream of the first transmembrane region in all relevant subunits. Chimeras were then constructed by restriction digestion and re-ligations (Fig. 5B). A point-mutated ␣4-subunit with mutations V77K, I80M, K160S, Y165F, and Y170Q (numbering of mutations corresponds to uniprot canonical sequence of the nAChR ␣3-subunit) were created using site-directed mutagenesis and restriction cutting and ligation using methods described previously (27). Custom designed oligos were ordered from Eurofins MWG Operon with the following sequences in which lowercase letters indicate positions introducing nucleotide mutations: 5Ј-GGT GTC  CAT GTC TCA GCT GGT GAA GGT cGA cGA Aaa AAA  CCA GAT gAT GGA GAC CAA CCT G-3Ј giving mutations  V77K and I80M and further 5Ј-CCA TCT TTA AGA GCT  CCT GTt cAA TCG AtG TGA CCT tCT TCC CGT TTG ATC  agC AAA ACT GTA CCA TGA AGT TCG-3Ј giving mutations  K160S, Y165F, and Y170Q. Mutations were confirmed by sequencing (MWG Operon) and cRNA was produced using the mMessage mMachine T7 Transcription kit according to the manufacturer's description (Ambion).
Electrophysiology-Experiments were performed as described previously (28). In brief, two-electrode voltage-clamp (TEVC) electrophysiology recordings were carried out using Xenopus laevis oocytes injected with ϳ25 ng of cRNA. Following injection, oocytes were incubated for 2-7 days at 18°C in modified Barth's solution containing: 90 mM NaCl, 1.0 mM KCl, 0.66 mM NaNO 3 , 2.4 mM NaHCO 3 , 10 mM HEPES, 2.5 mM sodium pyruvate, 0.74 mM CaCl 2 , 0.82 mM MgCl 2 , 100 g/ml of gentamycin, and pH adjusted to 7.5. For measurements, an oocyte was placed in a custom designed recording chamber where compound solutions could be added directly to the oocyte via a glass capillary. Pipette resistances were 0.6 -2.0 megaohm when submerged in OR2 containing: 90 mM NaCl, 2.5 mM KCl, 2.5 mM CaCl 2 , 1.0 mM MgCl 2 , 5.0 mM HEPES, and pH adjusted to 7.5. Following impalement the oocyte membrane potential was voltage clamped at a holding potential ranging from Ϫ40 to Ϫ80 mV using a Geneclamp 500B amplifier (Axon) and oocytes with leak currents exceeding 100 nA when clamped were discarded. Fresh solutions of NS9283, NS206, and ACh were prepared on the day of measurement in OR2 and applied to oocytes at a flow rate of 2.0 ml/min. Signals were low-pass filtered at 20 Hz and digitized at 200 Hz by a Digidata 1322A (Axon) and traces were recorded in Clampex 9.2 and subsequently analyzed in Clampfit. Except for datasets reproducing previously reported data, all datasets were constructed from at least five oocytes and conducted at minimally two individual experimental days.
Data Analysis-For data analysis, traces were baseline subtracted and responses to individual applications were read as peak current amplitudes. CRRs were fitted in GraphPad Prism 4 to a monophasic Hill equation constrained with a Hill slope of 1 and starting at 0, unless otherwise specified. Comparison of best fitting equation (monophasic versus biphasic) was carried out using the F-test in GraphPad Prism 4. When a biphasic fit was the statistically better model, data were fitted to the following equation, Y is the response, T is the maximum value of the fit, F is the fraction of the curve derived from the more potent component, ϫ is the concentration expressed on a logarithmic scale, log(EC 50 1) is the logarithm of the concentration giving halfmaximum response of the more potent component, and log(EC 50 2) is the logarithm of the concentration giving halfmaximum response of the less potent component.

RESULTS
NS206 (Fig. 1A) was initially identified as a PAM of ␣4␤2 nAChRs in a screening campaign specifically aimed at identifying allosteric modulators from compound libraries. As previously described for the PAM NS9283 ( Fig. 1B) (9), the screening was carried out using a HEK293 cell line stably expressing ␣4␤2 receptors in a Ca 2ϩ -flux assay and the Fluorescent Imaging Plate Reader. Interestingly, NS206 appeared to have a modulatory mechanism of action quite distinct from that of NS9283 and in the studies described below, we investigate these mechanisms using the two-electrode voltage-clamp electrophysiological technique on X. laevis oocytes. When considering studies conducted at ␣4␤2 receptors, it is important to note that these can express in two stoichiometries (2␣:3␤ or 3␣:2␤) with different ACh sensitivities as described under the Introduction. Fortunately, uniform populations of these stoichiometries are readily available in the X. oocyte system making this particularly suited for studies at ␣4␤2. Due to the different ACh sensitivities, maximal efficacious ACh-evoked currents (E max -(ACh)) in oocytes are typically obtained using 100 M ACh at the 2␣:3␤ stoichiometry and 1000 M ACh at the 3␣:2␤ stoichiometry (24). A commonly used methodology for testing the modulatory action of a PAM is co-applying the PAM with a submaximal concentration of an agonist, i.e. ACh in the case for ␣4␤2 receptors. Typically, the ACh concentration chosen (ACh-control) should evoke EC 10 -EC 30 currents when applied alone as this leaves a current "window" for observing positive as well as negative allosteric modulators.
NS206 Is a Novel PAM of Both ␣4␤2nAChR Stoichiometries-To evaluate the ability of NS206 to act as a modulator at ␣4␤2 nAChRs, the compound was tested on receptors expressed in oocytes injected with cRNA in ratios previously shown appropriate to give uniform populations of either a 2␣:3␤ or a 3␣:2␤ receptor stoichiometry (24). From representative traces at 2␣:3␤ receptors, NS206 were seen to modulate ACh-control currents in a concentration-dependent manner ( Fig. 2A). In fact, at the highest tested concentration of NS206 (10 M), the modulated response surpassed that of the E max (ACh) response. Experiments carried out with the 3␣:2␤ receptor revealed qualitatively similar observations (Fig. 2B). Although oocytes are not well suited for studying details of receptor desensitization kinetics, it was noted that current waveforms were not visually observed to differ in these experiments.
The concentration dependence of NS206 modulation at the two receptor stoichiometries was well approximated by the empirical Hill equation (Fig. 2C), suggesting the presence of a specific and saturable NS206 binding site. From the fitted curves, however, it is clear that full saturation is not observed at the highest tested concentration but precipitation within the NS206 solutions prohibited tests with higher concentrations. Whereas it could appear that NS206 gives higher modulatory efficacy at the 3␣:2␤ receptor, this cannot be firmly concluded based on these data as percentage modulation, in experiments such as these, is influenced by the specific ACh-control concentrations chosen. From the fitted graphs very similar EC 50 values of ϳ2-4 M are observed at the two receptor stoichiometries ( Table 1), suggesting that NS206 binding is largely unaffected by receptor stoichiometry. NS206 was unable to evoke any responses when applied by itself, i.e. without co-application of ACh at both receptor stoichiometries (Fig. 2, D and E) and engendered no displacement of [ 3 H]cytisine nor [ 3 H]epibatidine binding in native mouse and rat brain tissue (data not shown). Together, these data provide strong evidence of an allosteric mechanism of action and NS206 is therefore considered a PAM.
NS206 Is Selective for ␣4 Containing Receptors-To investigate the selectivity of NS206, the compound was tested in oocytes on additional common nAChRs.
At the homomeric ␣7 receptor, NS206 displayed no modulation of the ACh-control response (Table 1). When tested at various heteromeric receptors, these were expressed by inject- Additive Allosteric Modulation of ␣4␤2 nAChRs DECEMBER 13, 2013 • VOLUME 288 • NUMBER 50 ing cRNA in ratios to give the 3␣:2␤ stoichiometry. NS206 showed no modulation of the ganglionic ␣3␤4 nAChRs, however, the ␣4␤4 receptor was positively modulated with a potency of ϳ2 M (Table 1). This is similar to the potency observed at ␣4␤2 receptors and the maximal efficacy likewise appeared to be within the same range. Collectively, these data indicate that the presence of the ␣4-subunit is mandatory for PAM actions of NS206, whereas there is no dependence on the ␤-subunit.
Pharmacologically NS206 Increases Agonist Efficacy-Recently, the mechanism of NS9283 PAM action was described as a left-shift of the ACh CRR such that ACh appeared more potent in the presence of NS9283 at 3␣:2␤ receptors. Importantly, no noticeable differences were observed for E max (ACh) FIGURE 2. Allosteric modulation of ␣4␤2 nAChR currents by NS206 in X. laevis oocyte TEVC electrophysiology. A and B, representative traces of ACh-, ACh ϩ NS206-, and NS206-evoked ␣4␤2 currents. Oocytes were injected with cRNA for ␣4and ␤2-subunits in 1:4 (A) or 10:1 (B) ratios to yield (␣4) 2 (␤2) 3 and (␣4) 3 (␤2) 2 receptors, respectively, as indicated by insets. Modulatory efficacy of NS206 was evaluated by co-applications with a submaximal concentration of ACh typically giving rise to EC 10 -EC 30 current levels (ACh-control, Table 1). Applications of ACh or ACh ϩ NS206 (co-application) are indicated by bars above each trace. C, NS206 modulates both ␣4␤2 stoichiometries with comparable potencies. Background subtracted peak current amplitudes from experiments described above were normalized to the respective ACh-control current (I ACh-control ) and depicted Ϯ S.E. as a function of increasing NS206 concentrations of n ϭ 8 or 4 -10 experiments, respectively. Plotted data points were fitted to the empirical Hill equation using non-linear regression. D and E, NS206 does not activate (␣4) 2 (␤2) 3 (D) or (␣4) 3 (␤2) 2 (E) receptors in the absence of ACh. Oocytes were injected as described above and stimulated with ACh (left) or NS206 (right). { {denotes that applications are separated by washing steps or intermediate applications.

TABLE 1 NS206 or NS9283 modulation of ACh-evoked responses at wild-type, chimeric, and mutant nAChRs using X. laevis oocyte TEVC electrophysiology
Oocytes were injected with nAChR subunits in the indicated cRNA ratios to yield receptors in (␣*) 2 (␤*) 3 or (␣*) 3 (␤*) 2 stoichiometries. CRRs for NS206 and NS9283 were obtained by co-application with an EC 10 -30 concentration of ACh as shown in Fig. 2. ACh concentrations used to evoke EC 10 -30 currents (ACh-control) were adjusted for each receptor as indicated. Data were fitted to the empirical Hill equation using non-linear regression with a fixed bottom of 0 and a Hill slope of 1 and are presented as EC 50 in M and E max in % with 95% confidence intervals from the indicated number of experiments. At 2␣:3␤ receptors, ACh on its own was found to have an EC 50 value of ϳ1 M, whereas at 3␣:2␤ receptors the response was, as expected, biphasic with EC 50 values of ϳ1 and ϳ100 M, respectively (Fig. 3, A and B, and Table 2). The biphasic response in a uniform population of 3␣:2␤ receptors reflects the existence of two high sensitivity sites at the ␣␤-interfaces and a low sensitivity site at the ␣␣-interface and the observed potencies are, with the resolution taken into account, comparable with previous studies (21,23,24). Although not clearly evident in the graph (Fig. 3B), previous experiments demonstrate that 1 mM ACh is sufficient to give E max (ACh) responses (24). Interestingly, with NS206 present, the ACh CRRs were affected at both receptor stoichiometries. For the 2␣:3␤ stoichiometry, the peak current amplitudes were potentiated ϳ3-4fold at the highest ACh concentration accompanied with only a minor change in EC 50 value ( Fig. 3A and Table 2). Qualitatively similar potentiating effects of ϳ1.5 times were noted at the highest ACh concentrations for the 3␣:2␤ stoichiometry. However, the ACh CRR for this combination still appeared biphasic, albeit with an altered fraction between the two phases, and a 2-3-fold potency increase was noted for both EC 50 values ( Fig.  3B and Table 2).
Upon visual inspection of representative traces from the maximal ACh concentrations (1 mM) in the above experiments, it is seen that NS206 efficacy is not due to large changes in current waveform characteristics (Fig. 3, A and B, insets). The total application time for the ACh pulse is ϳ30 s and during this time period only limited current decay is observed for 2␣:3␤ receptors, whereas some decay is observed for 3␣:2␤ receptors. In the presence of NS206, the increases in maximal currents are readily evident; however, the overall current waveforms appear similar. Based on this it appears that NS206 is unlikely to have  Oocytes were injected with nAChR subunits in 1:4 or 4:1 cRNA ratios to yield receptors in (␣4) 2 (␤2) 3 or (␣4) 3 (␤2) 2 stoichiometries.

Receptor
Modulator  major effects at receptor desensitization characteristics at either of the two receptor stoichiometries.
To further investigate the fitted curvatures of ACh in the presence of NS206, CRRs with a higher number of data points were obtained using normalization to an E max (ACh) trace (100 M or 1 mM for 2␣:3␤ and 3␣:2␤, respectively). In these experiments, ACh CRR in the presence of NS206 clearly behaved in a monophasic nature at 2␣:3␤ receptors (Fig. 4A) but in a biphasic nature at 3␣:2␤ receptors (Fig. 4B). Because the non-linear regression routine creates a fit using the least squares method, the analysis of variance F-test was used to evaluate whether a simple monophasic model was preferred over the more complex biphasic model. For the 2␣:3␤ receptor, the low F-value of 2.2 (p ϭ 0.12) means that the simpler model is preferred, whereas a high F-value of 56 (p Ͻ 0.0001) show that the more complex biphasic fit is preferred for the 3␣:2␤ receptor. Modulatory increase of high concentration ACh-evoked currents was similar to the observations with the paired experiments (factor of ϳ3.6 and ϳ1.6 for 2␣:3␤ and 3␣:2␤ receptors, respectively) and whereas the EC 50 value for the 2␣:3␤ receptor was largely unaltered, EC 50 values at 3␣:2␤ receptors again appeared left-shifted by a factor of ϳ3-5 relative to published values for ACh alone (24). Although this is most likely a real effect it should be noted that fewer data points make up each phase in a biphasic fit, hence, the uncertainty in estimating EC 50 values increases. The fraction of high versus low sensitivity components at the 3␣:2␤ stoichiometry was ϳ0.6, which is higher than the ϳ0.2 reported for ACh alone. This likely reflects that maximal currents of the high sensitivity component are proportionally more affected by NS206 (factor of ϳ3.6) compared with the low sensitivity component (factor of ϳ1.6).
NS206 and NS9283 Have Additive Modulatory Effects-With different modes of action observed for NS206 and NS9283, the question arises as to whether an ␣4␤2 receptor can be modulated through both mechanisms simultaneously. To study this possibility, an ACh CRR was obtained at 3␣:2␤ receptors in the presence of NS9283 (31.6 M) and another ACh CRR was obtained in presence of NS206 ϩ NS9283 (3.16 ϩ 31.6 M).
The modulatory actions of NS9283 match previously published results, i.e. a monophasic left-shifted CRR with no significant increase in maximum peak currents at the highest ACh concentrations (Fig. 5). However, simultaneous application of both NS206 and NS9283 gives a unique mixture of modulatory effects not observed for either compound alone ( Fig. 5 and Table 2). (i) The ACh CRR is monophasic, whereas the CRR with NS206 alone is biphasic (Fig. 4B), (ii) at high ACh concentrations, the modulated response show efficacy levels that significantly surpasses that of NS9283 alone by a factor of ϳ1.7 (F (1,160) ϭ 633, p Ͻ 0.0001; maximal efficacy NS9283 ϩ NS206 versus NS9283 alone). Compared with the NS9283 alone addition, NS206 further left-shifted the CRR of ACh albeit only by a factor of ϳ2. Interestingly, in these experiments preincubation of the two modulators simultaneously gave rise to small currents of ϳ10% of E max (ACh) in the absence of ACh. This could indicate that the cumulative effects of the two modulatory compounds are sufficient to gate the ␣4␤2 receptor to a low degree.
Modulation by NS9283, but Not NS206, Depends on the ␣␣-Interface-Both NS206 and NS9283 act as positive modulators and are therefore expected to bind at one or more site(s) that are distinct from the orthosteric agonist (ACh) binding sites. As stated previously, the 3␣:2␤ stoichiometry selectivity of NS9283 could be reconciled with a hypothesis that only this stoichiometry presents the binding site for NS9283 (9). A major difference between the two stoichiometries is the presence of an ␣␣-interface with an additional ACh binding site in the 3␣:2␤ receptor (24). To investigate whether this additional ACh binding site influences modulatory actions, both compounds were tested on a 3␣:2␤ receptor with three point mutations H142V, Q150F, and T152L in the ␣-subunit (␣4 3M ) (Fig.  6A, top inset). These mutations convert key ␣-subunit complementary side residues to the corresponding ␤-subunit residues, thereby making the ␣␣-interface more ␣␤-like. When expressed with wild-type ␤2 in oocytes, the resulting 3␣ 3M :2␤ receptors contain three sites that display comparable affinities for agonists and give monophasic ACh CRRs (24), i.e. the receptor behaves like a 2␣:3␤ receptor albeit with higher current levels.  At wild-type 3␣:2␤ receptors, NS9283 is a PAM with ϳ700% efficacy (relative to the EC 10 -30 ACh-control) and an EC 50 value of ϳ3 M (Fig. 7A and Table 1), which is similar to previously reported data of E max ϭ 638% and EC 50 ϭ 3.3 M (9). However, NS9283 loses all its modulatory actions at the 3␣ 3M :2␤ receptor (Fig. 7A and Table 1). Modulatory actions by NS206, on the other hand, were unaffected by the mutations (Fig. 7B and Table 1), as both efficacy and EC 50 values at the mutated receptor are similar to the values obtained at wild-type receptors. Hence, consistent with the lack of subunit stoichiometry selectivity, NS206 binds to a site entirely distinct from the NS9283 binding site.
Determinants of NS206 Modulatory Action-In another attempt to identify the receptor region(s) important for NS206 modulatory actions, the compound was tested on chimeric receptors where the ECD of ␣4 was substituted with that of ␣3 or vice versa (Fig. 6B). These chimeras have previously been shown to assemble into fully functional nAChRs with characteristics dominated by the ECDs (27). Of particular importance, the ACh sensitivities are very similar to the respective wild-type receptors enabling regular selection of ACh-control concentrations (Table 1). For the experiments described below, chimeric receptors were expressed in a 3␣:2␤ stoichiometry.
At the ␣4/␣3ϩ␤2 chimera, i.e. an ␣4␤2 ECD coupled with an ␣3␤2 TMD, efficacy and potency of NS9283 were observed similar to that at wild-type receptors (Fig. 8A and Table 1). No activity was seen with the reverse ␣3/␣4 ϩ ␤2 chimera in agreement with the described lack of efficacy at ␣3-containing receptors (data not shown) (9). Hence, modulatory actions of NS9283 appear entirely dependent on the ECD of ␣4 fully consistent with the above mapping of its binding site to the ␣␣-interface of ␣4␤2 receptors. Furthermore, the normal modulatory behavior of NS9283 at the ␣4/␣3 ϩ ␤2 chimera underscore the usefulness of chimeras in this context. NS206 positively modulated the ␣3/␣4 ϩ ␤2 chimera with an EC 50 of ϳ2 M albeit with very low efficacy ( Fig. 8B and Table  1). In contrast, no modulation was observed with the reverse chimeric receptor ␣4/␣3 ϩ ␤2 (Fig. 8B and Table 1). Because   . Amino acids in ␣4and ␣3-nAChR subunits affecting modulation by NS206 and NS9283. A, efficacy determinants for NS206 and NS9283 modulation are located in different regions of the receptor as illustrated by a ribbon diagram model of the three-dimensional structure of the T. marmorata ␣␤-subunit interface (Protein Data Bank 2BG9) (29). In the top inset, a homology model for an ␣4␣4-interface (24) shows that three amino acids on the complementary side of the ␣4-subunit differ from those of an ␣4␤2-interface. Upon mutations of these (H142V, Q150F, and T152L), which convert the ␣␣-site to resemble an ␣␤-site, modulation by NS9283 but not NS206 is abolished. In the bottom inset, a distance threshold of 12 Å to the TMD reveals that only five amino acids in this part of the ECD region differ between ␣3and ␣4-subunits. Mutation of these to the corresponding ␣4 residues in the chimeric ␣3/␣4-subunit (V77K, I80M, K160S, Y165F, and Y170Q) introduces full NS206 modulatory efficacy at the ␣3 5M /␣4␤2 receptor. B, schematic showing the design of an ␣4/␣3 chimera. A SalI site was silently introduced to a few amino acids prior to the beginning of the TMD in both ␣3and ␣4-subunits. The illustrated chimera, where the ␣4-ECD (dark shaded) is fused to the ␣3-TMD (white), was then constructed by restriction digestion and re-ligation (27). C, alignment of regions of the ECDs of ␣4 and ␣3 showing that the five differing amino acids described in A are located in loop ␤1-␤2 or loop ␤6-␤7 . the potency at the ␣3/␣4 ϩ ␤2 chimera was similar to that at wild-type receptors it appears reasonable to assume that NS206 bind in the TMD region. However, for this argument to hold, the lack of efficacy must be explained by suboptimal propagation of structural changes upon compound binding. Given that the chimera contains entire domains from two different subunits, this hints to determinants in the interface between the two domains. ECD regions believed to be important for this interface include loop ␤1-␤2 , the Cys-loop (loop ␤6-␤7 ), and loop ␤8-␤9 (Fig. 6A, bottom inset) (29). An alignment of ␣3 and ␣4 with the ␣-subunit of the Torpedo marmorata AChR identifies a number of amino acid differences. Because a 4-Å resolution structure of the Torpedo AChR is available (Protein Data Bank 2BG9), this alignment further allows qualified guesses as to the proximity of these differing amino acids to the TMD (Fig.  6A, bottom inset). Within 12-Å distance of the TMD, ␣3 and ␣4 differ in only five amino acid positions, three in the Cys-loop and two in loop ␤1-␤2 (Fig. 6C). To investigate whether the identity of these five amino acids affect modulatory actions of NS206, an ␣3 5M /␣4 chimera was created, where the five residues in the ECD of ␣3 were point mutated to the corresponding ␣4 residues (V77K, I80M, K160S, Y165F, and Y170Q). Injecting cRNA for ␣3 5M /␣4 ϩ ␤2 in oocytes gave rise to receptors that behaved like the regular non-mutated ␣3 ϩ ␤2 receptor with respect to ACh sensitivity (EC 50 of 380 (320 -450) M versus EC 50 of 290 (240 -340) M, respectively; for simplicity these values are from monophasic Hill fits, although both receptors clearly show biphasic behavior). Furthermore, NS9283 showed no modulatory efficacy consistent with the lack of an ␣4␣4interface (Fig. 8A). Nevertheless, a large difference was noted when testing modulatory actions of NS206 at this chimera, as the efficacy level at this mutant receptor was restored to values comparable with that at wild-type ␣4␤2 (Fig. 8B, Table 1). These data indicate that NS206 indeed does bind in the TMD in a manner dependent on the ␣4-subunit. Furthermore, whereas NS9283 is fully capable of propagating its modulatory actions across a chimeric ECD/TMD interface, NS206 needs an interface that closely resembles that of the ␣4-subunit for full efficacy.

DISCUSSION
Drug discovery efforts aimed at identifying novel PAMs of ␣4␤2 nAChRs are progressing due to the important role of this receptor class in normal brain function. However, to date very few new chemical entities modulating these receptors have emerged in the literature limiting both the current understanding but also new investigative opportunities. To this end, the present study identifies the novel compound NS206 as a potent and efficacious PAM of ␣4␤2 nAChRs. Structurally, this compound does not resemble the previously described NS9283 and comparison of these two compounds indeed reveals important pharmacological differences.
NS206 displays all the hallmark features of a PAM at ␣4␤2 receptors. When applied alone to ␣4␤2 expressing oocytes in electrophysiological experiments, it evokes no detectable current; however, when co-applied with ACh, clear potentiation of current levels is observed with no apparent change in current waveforms. The level of potentiation is concentration-depen-dent and well approximated by the empiric Hill equation, indicating a saturable binding site for NS206. Moreover, no radioligand displacement was noted in binding studies. With respect to selectivity profile, NS206 was highly selective for ␣4* over ␣3␤4 and ␣7 nAChRs. No apparent differences in PAM activity were noted between ␣4␤2 and ␣4␤4, indicating that NS206 activity is linked to the presence of the ␣4-subunit. Heteromeric nAChRs such as ␣4␤2 are known to express in two stoichiometries with ratios between the subunits being either 2␣:3␤ or 3␣:2␤. Whereas NS9283 only modulates 3␣:2␤ receptors, NS206 displayed allosteric modulation at both receptor stoichiometries.
To study the pharmacological actions of NS206 further, ACh CRRs were run in the absence and presence of the compound. At 2␣:3␤ stoichiometry for ␣4␤2 receptors, NS206 increased ACh-evoked peak current levels and fitted curvatures of the two CRRs remained identical with no change in ACh potency. Thus, modulatory actions of NS206 caused a ACh concentration-independent fixed increase in evoked currents by a factor of 3-4. For 3␣:2␤ receptors, deciphering ACh CRR data are more complicated due to the existence of two ␣␤-sites and one ␣␣-site. This causes an ACh CRR to be biphasic with observed potencies at the two components of ϳ1 and ϳ70 M (␣␤-site activation and ␣␤-site plus ␣␣-site activation, respectively) (24). Interestingly, in the presence of NS206 the allosterically modulated ACh CRR remains biphasic albeit with an altered ratio between the components and small increases in potency. Current levels at EC 100 of ACh were increased by a factor of ϳ1.6, which is less than seen for the 2␣:3␤ stoichiometry but, nevertheless, similar type of effect.
Collectively these data show that the main effect of NS206 is modulation of ACh-evoked peak current levels, i.e. increased "gain." Incidentally, the 2␣:3␤ receptor, where NS206 induces a proportionally larger increase in maximal peak current amplitudes, is also the stoichiometry with the lowest unitary conductance (21). This can be reconciled with a notion that ACh-evoked channel activation via two agonist sites (2␣:3␤ receptor) gives lower gating efficiency compared with activation via three sites (3␣:2␤ receptor). Thus, the lower gating efficiency at 2␣:3␤ receptors allows for a greater potential for modulation by e.g. NS206. This again explains the altered ratio between the two ACh CRR components at 3␣:2␤ receptors. In support of this, recent nAChR ␣7 data show that activation via two binding sites results in intermediate mean channel open times, whereas activation via three binding sites arranged nonconsecutively (similar to the arrangement for 3␣:2␤ receptors) results in maximal mean channel open times (30).
By comparing modulatory actions of NS206 to those of NS9283 (9) it is clear that the compounds behave very differently. NS9283 only works on the 3␣:2␤ receptor stoichiometry and its modulatory actions left-shift the ACh CRR and give monophasic properties. The exact left-shift degree appears to be somewhat methodology dependent, because a ϳ600-fold shift is seen in oocytes but only a ϳ60-fold shift in patch clamp (31). The reasons for this discrepancy are not clear at the moment, but irrespective of which testing regime was used, peak current levels in the presence of NS9283 did not surpass that of EC 100 of ACh. As discussed above, this is in contrast to the pharmacological actions of NS206 and, interestingly, these two mechanisms can work additively. An ACh CRR at 3␣:2␤ receptors in the presence of both NS9283 and NS206 displays the core characteristics of both compounds, i.e. large left-shift in potency and monophasic fitted curvature (NS9283 actions) combined with current levels above that of EC 100 of ACh (NS206 actions). The differences in modulatory actions and additive effects make it likely that NS9283 and NS206 bind in distinct sites at the receptor. Previously, etomidate has been shown to modulate [ 3 H]flunitrazepam binding in the GABA A R benzodiazepine site, indicating that binding to separate sites in Cys-loop receptors can indeed have additive or even synergistic effects (32).
Given that NS9283 is dependent on the ␣␣-interface it was speculated that this interface presents the binding site (9). Indeed, three mutations associated with the binding of ACh in the ␣␣-site turned out to be of vital importance for NS9283 modulatory actions. This indicates a shared mechanism of action with other modulators of Cys-loop receptors that bind in non-canonical ECD interfaces. The most prominent example is benzodiazepine modulation at GABA A Rs (10), but interface binding sites have also been proposed for galanthamine and morantel (14,15). Whereas the case of NS9283 appears similar to that of benzodiazepines there is one important difference: the ␣␣-interface binds ACh and is hence canonical. This suggests two possibilities: either NS9283 directly substitutes for ACh in the orthosteric binding pocket or NS9283 binds in addition to ACh and through local interactions acts to increase ACh binding potency. Regardless of which of these possibilities is correct both would lead to a full contribution of the ␣␣-interface in gating and are hence mechanistically consistent with an ACh CRR left-shift as the only pharmacological observation.
Modulatory actions of NS206 on the other hand were unaffected by mutations in the ␣␣-interface, however, in studies with chimeric ␣-subunits (␣4/␣3 and ␣3/␣4), modulation was only seen with the ␣3/␣4 chimera. Efficacy levels were severely impaired but the potency remained similar to that at wild-type ␣4␤2 receptors. This indicates retention of key binding determinants in the ␣4 TMD and hints that efficacy determinants reside on both the ECD and TMD. To address this, an ␣3/␣4 chimera was constructed in which five ␣3 ECD amino acids in close proximity of the TMD were mutated to the respective amino acids in ␣4. These mutations turned out to give "gain of function," i.e. full modulatory efficacy of NS206 was restored in the ␣3 5M /␣4 ϩ ␤2 receptor. Given that only efficacies change in these experiments it appears reasonable to assume that binding is dominated by the ␣4 TMD and that amino acids at the interface between the ECD and TMD are involved in propagating modulatory actions to other parts of the receptor. This would be consistent with the proposed binding sites observed for etomidate at GABA A Rs, PU02 at 5-hydroxytryptamine type 3A receptors, ivermectin at an invertebrate GluCl channel, and NS1738 and PNU-120596 at ␣7 (16 -20). Future studies will show whether NS206 in fact bind to a binding site that is the ␣4␤2 nAChR homologue of e.g. the PU02 site.
Summing up, it is clear that ␣4␤2 nAChRs can be allosterically modulated by two distinct mechanisms. Of the three phenotypes of allosteric modulation mentioned under the Intro-duction, the two modulators tested here each give rise to only one of the phenotypes, neither of which dramatically changed desensitization characteristics. Allosteric modulators that exhibit several modulatory phenotypes in one molecule have previously been described (see Introduction) but the fact that the actions of two "single phenotype" modulators are additive indicates limited cross-talk between their binding sites, i.e. the actions of one is fully independent on the other. Aligning the data obtained here with that of modulators at other Cys-loop receptors, a picture begins to emerge. Notably, the pharmacological action of a modulator that binds in an ECD interface is likely one of agonist potency left shifting (benzodiazepines, galanthamine, and NS9283), whereas a modulator that binds in the ECD/TMD interface likely gives increased "gain" (etomidate, NS1738, PNU-120596, and NS206). This knowledge adds another layer of refinement in drug discovery that again can have therapeutic implications. For instance, galantamine is used for treatment of mild Alzheimers and a compound with actions similar to NS206 would based on the current data be expected to act additively, whereas a compound similar to NS9283 would instead compete for binding.