C57BL/6 mice (ages 3–6 months) used in this study were bred and maintained at the California Institute of Technology (Caltech) or the University of Colorado Boulder. After weaning at 21 days, same sex littermates were housed no more than 4 (Caltech) or 5 (Colorado) to a cage. Mice of the α4 subunit null mutant line (Marubio et al., 1999), the α6 subunit null mutant line (Champtiaux et al., 2002), and the hypersensitive α6L9’S transgenic mice (Drenan et al., 2008) were bred and maintained as above and genotyped as previously described (Marubio et al., 1999; Champtiaux et al., 2003; Drenan and Lester, 2012). Mice had free access to food and water and were maintained on a 12/12-h light/dark cycle at 22°C. All experiments were conducted with the approval of the California Institute of Technology Animal Care and Use Committee. C57BL/6J strain mice, as well as α4, α5, and β2 subunit null mutant mice on this background, were bred and maintained at the Institute for Behavioral Genetics, University of Colorado Boulder, CO, United States. Animal care and procedures were approved by the Animal Care and Utilization Committee of the University of Colorado, Boulder.

Rat experiments were conducted in adherence with the National Institutes of Health guidelines and approved by the Institutional Animal Care and Use Committee of Mt. Sinai, New York, NY, United States. Male Wistar rats (n = 8; Charles River Laboratories, Raleigh, NC) weighing 300 g at the start of experiments were housed in groups of 1–2 per cage in an environmentally controlled vivarium on a 12 h reverse light-dark cycle. Prior to the commencement of behavioral testing, all rats had ad libitum access to food and water.

TC299423, (E)-5-(pyrimidin-5-yl)-1,2,3,4,7,8-hexahydroazocine, was synthesized at Targacept. A full report on the synthesis of TC299423 is in preparation. Varenicline tartrate was also synthesized by Targacept. [¹²⁵I]-Epibatidine (2200 Ci/mmol), [³H]dopamine (3,4-[ring-2,5,6-3H], 30–60 Ci/mmol), [³H]choline (methyl-3H, 60–90 Ci/mmol), and carrier-free ⁸⁶RbCl were purchased from Perkin Elmer Life Sciences, Boston, MA, United States. α-Conotoxin MII (α-CtxMII) was obtained from J. Michael McIntosh, University of Utah, Salt Lake City, UT, United States. The following chemicals as well as all buffer components (Reagent Grade) were products of Sigma–Aldrich (St. Louis, MO, United States): (L)-nicotine hydrogen tartrate, mecamylamine, atropine, bovine serum albumin (BSA), (±)-epibatidine, HEPES, nomifensine, pargyline, and tetrodotoxin. All compounds were dissolved in physiological saline (0.9% NaCl). Concentrations refer to the free base.

For the mouse behavioral assays all drugs were dissolved in sterile 0.9% saline and were administered in an injection volume of 0.1 ml/g. Drug doses are expressed as free base.

Mice were habituated to the experimental room for 2 h prior to the experiment. Horizontal locomotor activity was recorded using an infrared photobeam activity cage system (San Diego Instruments; San Diego, CA, United States). After habituating to the experimental room, mice were placed singly in a novel cage. Ambulations were recorded when two contiguous photobeams were broken in succession, preventing activity from being recorded by most sedentary animals. Ambulation events were measured for 45 min with four 15 s intervals per min. Mice were given a total of two intraperitoneal (ip) injections. First, mice were pre-injected with saline or 1 mg/kg mecamylamine before being placed in the activity cages. After 8 min in the cage they were removed, injected with the second drug (saline, nicotine, or TC299423), and returned again to the cage for the 45 min session.

In preparation for the present study, a small-molecule discovery program using a library of ∼7000 compounds of modest drugability was conducted at Targacept. Initial screenings of the library were conducted with few-concentration assays of Ca²⁺ flux and nAChR binding in transfected cell lines. Molecules with potential nanomolar to micromolar affinity for α6β2^∗ nAChRs were sought. TC299423 (Figure 1), one of these compounds, was identified after our previous structure-activity relationship (SAR) studies on other nAChR subtypes (Breining et al., 2012; Zhang et al., 2012; Strachan et al., 2014) and on α6β2^∗ nAChRs (Breining et al., 2009; Grady et al., 2010; Strachan et al., 2014). TC299423 was also analyzed in computational studies on homology models of α6β2 binding sites analogous to those described previously (Strachan et al., 2014). Assays that assessed metabolism were also conducted (discussed below, see Supplementary Table S2). These initial experiments motivated the more detailed studies in the present paper, and in recently reported structure-function experiments (Post et al., 2015; Post et al., 2017).

TC299423 can be considered an acrylic metanicotine analog (1 or 20a, Figure 1). In TC299423, the pyridine ring of nicotine was replaced by a pyrimidine ring, a substitution previously shown to increase α6^∗ selectivity (Breining et al., 2009). We also incorporated the N-methyl moiety of the metanicotine into an additional, 8-membered ring (a 1,2,3,4,7,8-hexahydroazocine) in an effort to test a hypothesis that conformational constraints imposed by a cyclic structure would confer additional selectivity by reducing the conformational flexibility of the linear metanicotine structure.

Data obtained with an oocyte expression system showed that TC299423 exhibits a binding mode similar to that of nicotine at α6β2 nAChRs, forming a relatively weak cation-π interaction with a conserved tryptophan residue termed TrpB (Post et al., 2015). At α4β2 nAChRs, this cation-π interaction also occurs. At α4β2 nAChRs, TC299423, like several other secondary ammonium agonists, makes an additional cation-π interaction with a conserved tyrosine termed TyrC2 (Post et al., 2017).

The bioavailability of TC299423 was studied by measuring its plasma and brain concentration in mice after ip or oral administration (Figure 2). TC299423 reached a maximum plasma concentration of 49 ± 18 ng/ml (0.26 μM) 0.08 h after a 0.3 mg/kg ip injection and had a half-life of 0.17 h in the plasma (Figure 2A). It reached a maximum brain-tissue concentration of 22 ± 0.43 ng/g (0.12 μM) 0.5 h after ip injection (Figure 2B). Oral administration at 1 mg/kg resulted in a maximum plasma concentration of 38 ± 10 ng/ml (0.20 μM) after 0.25 h with a half-life of 1.12 h in the plasma (Figure 2A). A maximum brain-tissue concentration of 42 ± 17 ng/g (0.22 μM) was attained at 0.25 h (Figure 2B). Thus, TC299423 enters the plasma, and passes into the brain, before being completely metabolized. We used a previously described radioligand displacement assay (Hussmann and Kellar, 2012) to obtain additional measurements of the TC299423 concentration in brain tissue ex vivo (Supplementary Table S1). Similar to the previous pharmacokinetic assay, these data confirm the penetration of TC299423 into the brain and its availability to nAChRs, for ≥20 min.

Some aspects of the microsomal metabolism of TC299423, in comparison to varenicline, were studied in preliminary experiments. In both human and rat microsomes, TC299423 exhibited a ∼7-fold longer half-life compared to varenicline (Supplementary Table S2). Using recombinantly expressed cytochrome P450 enzyme preparations, we also observed that TC299423 exhibits a more diverse degradation pathway than varenicline as it is metabolized by more isoforms of cytochrome P450 enzyme (Supplementary Table S2). These data suggest that several metabolic pathways contribute to TC299423 degradation. Thus, it is a candidate for more detailed human metabolism studies.

We used a variety of assays (whole-cell patch clamp, synaptosomal [³H]-dopamine release, synaptosomal [³H]-ACh release, [¹²⁵I]-epibatidine binding, and ⁸⁶Rb⁺ efflux) to measure TC299423 concentration-response relations for α4β2^∗, α6β2^∗, and α3β4^∗ nAChRs. Figure 3, Tables 1–3, and Supplementary Figure S1, summarize the data and show that TC299423 potently activates α6β2^∗ nAChRs. The TC299423 concentration-response relations for α4β2 and α6β2β3 nAChRs were measured by transiently transfecting fluorescently tagged, mutant α4-GFPβ2_DM or α6-GFPβ2_DMβ3 nAChRs in neuro-2a cells, voltage clamping the cells (whole-cell mode), and applying TC299423 by rapid microperfusion (see Materials and Methods, Figure 3A). GFP-labeled nAChR subunits were used to facilitate the identification of neuro-2a cells expressing the transfected nAChRs and the DM mutation was used to increase the surface receptor density (Henderson et al., 2014). EC₅₀ values were determined by fitting the concentration-response data to the Hill equation. The EC₅₀ of TC299423 on α4-GFPβ2_DM nAChRs was 0.1 ± 0.02 μM and on α6-GFPβ2_DMβ3 nAChRs was 0.04 ± 0.01 μM (Figure 3A and Table 1). The EC₅₀ for the α6-GFPβ2_DMβ3 nAChRs was significantly less than that for α4-GFPβ2_DM nAChRs (p < 0.05, two-tailed t-test). Thus TC299423 activated α6-GFPβ2_DMβ3 nAChRs significantly more potently than α4-GFPβ2_DM nAChRs.

To compare the efficacy and potency of TC299423 for native α4β2^∗ and α6β2^∗ nAChRs, we measured TC299423-induced [³H]dopamine release from striatal synaptosomes (Figure 3B and Table 1). Two classes of nAChR subtypes are responsible for nicotine-induced [³H]dopamine release from striatal synaptosomes: α-conotoxin MII (α-CtxMII)-sensitive and α-CtxMII-insensitive nAChRs. These classes are composed of α6β2^∗ and α4(non-α6)β2^∗ nAChRs, respectively. Accordingly, we measured concentration-response relations for the α-CtxMII-sensitive, and -insensitive, TC299423-induced [³H]dopamine release. As above, the EC₅₀ values were obtained from fits to the Hill equation. The EC₅₀ of TC299423 for α-CtxMII-insensitive [³H]dopamine release was 0.13 ± 0.03 μM and that for α-CtxMII-sensitive release was 0.03 ± 0.01 μM (Figure 3B and Table 1). Thus, consistent with the patch-clamp data above, the TC299423–induced [³H]dopamine release assay also suggests that it activates α6β2^∗ nAChRs more potently than α4β2^∗ nAChRs. Both data sets suggest that TC299423 activates α6β2^∗ nAChRs ∼3-fold more potently than α4β2^∗ nAChRs. We also studied the effects of TC299423 on [³H]dopamine release using striatal synaptosomes from α5-KO mice. Genetic deletion of the α5 subunit did not significantly affect the potency of TC299423-induced α-CtxMII-sensitive, or -insensitive, [³H]dopamine release (Table 1).

To study the potency of TC299423 for native α3β4^∗ nAChRs, we prepared synaptosomes from the mouse interpeduncular nucleus, a region with dense α3β4 nAChR expression (Shih et al., 2014), and measured TC299423-induced [³H]ACh release. The EC₅₀ for TC299423-induced [³H]ACh release was 8.0 ± 0.4 μM (Table 1). Thus, TC299423 was much less potent at eliciting [³H]ACh release from interpeduncular nucleus synaptosomes than it was at eliciting [³H]dopamine release from striatal synaptosomes, suggesting that the drug activates α6β2^∗ and α4β2^∗ nAChRs more potently than α3β4^∗ nAChRs.

The α4 and β2 nAChR subunits can form functional nAChRs with different subunit stoichiometries [α4₍₃₎β2₍₂₎, α4₍₂₎β2₍₃₎] and agonist sensitivities (low-, high-sensitivity, respectively). Receptors with the α4₍₃₎β2₍₂₎ stoichiometry are much less ACh-sensitive than those with the α4₍₂₎β2₍₃₎ stoichiometry (Nelson et al., 2003; Kuryatov et al., 2005). To compare the efficacy and potency of TC299423 for these two receptor stoichiometries, we measured TC299423-elicited ⁸⁶Rb⁺ efflux from mouse thalamic and cortical synaptosomes (Table 2). The α4(non-α6)β2^∗ nAChRs mediate nicotinic agonist-induced ⁸⁶Rb⁺ efflux from these synaptosomes (Marks et al., 2009). The nicotinic antagonist dihydro-β-erythroidine (DhβE) blocks high-sensitivity α4β2^∗ nAChRs more potently than low-sensitivity α4β2^∗ nAChRs (Marks et al., 1999) and can be used to distinguish between the two receptor stoichiometries. TC299423 had an EC₅₀ of 0.6–2.0 μM for the DhβE-sensitive α4β2^∗ nAChRs and ≥14 μM for the DhβE-insensitive α4β2^∗ nAChRs (Table 2). Thus, TC299423 more potently activates high-sensitivity α4₍₂₎β2₍₃₎, than low-sensitivity α4₍₃₎β2₍₂₎, nAChRs.

Regarding efficacy, previous data with an oocyte expression system showed that TC299423 is a partial agonist at α6β2 nAChRs, with an efficacy 59% that of ACh itself (Post et al., 2015). In the present experiments on neurotransmitter release, we found additional evidence that TC299423 is a partial agonist at α6β2^∗ receptors: the maximal release by TC299423 was 50–54% that of nicotine (Table 1).

We find that TC299423 is roughly as efficacious as nicotine for high-sensitivity α4β2^∗ nAChRs, but a partial agonist on low-sensitivity α4β2^∗ nAChRs (Tables 1, 2). Interestingly, it also appears to be at least as efficacious as nicotine for α4β2^∗ nAChRs lacking the α5 nAChR subunit (Figure 1, Tables 1, 2, and Supplementary Table S1). We also find that TC299423 is roughly as efficacious as nicotine for α3β4^∗ nAChRs.

To determine whether TC299423 displays a higher affinity for β2^∗, than β4^∗, nAChRs, we measured its ability to displace [¹²⁵I]epibatidine binding to membranes from three brain regions (see Materials and Methods). [¹²⁵I]Epibatidine binds predominantly to the α4β2^∗ subtype in cortical membranes, α3β4^∗ in IPN membranes from β2KO mice, and α6β2^∗ in striatal membranes from α4KO mice (Grady et al., 2010). Thus, the set of K_i values for TC299423 to displace [¹²⁵I]epibatidine binding in these three regions provides a way to measure the apparent affinity of these three subtypes. Based on these measurements, the K_i values for the α4β2^∗, α6β2^∗, and α3β4^∗ subtypes were 0.24 ± 0.04, 1.4 ± 0.6, and 18.0 ± 0.7 nM, respectively (Table 3). Thus, TC299423 binds with higher affinity to α4β2^∗ and α6β2^∗, than α3β4^∗ nAChRs. K_i values primarily reflect agonist binding to the high-affinity, desensitized state, whereas EC₅₀ values for functional assays reflect agonist binding to the free-receptor state. The relationship between the K_i for [¹²⁵I]epibatidine displacement (Table 3) and the EC₅₀ for [³H]dopamine release (Table 1) for the α4β2^∗ and α6β2^∗ nAChRs is consistent with previous results showing that α6^∗ nAChRs display a ∼10-fold lower K_i than EC₅₀, whereas α4^∗ nAChRs display a ∼100-fold difference. Here, we observed a ∼20-fold difference for α6β2^∗ nAChRs and a ∼500-fold difference for α4β2 nAChRs. This comparison suggests that agonist-induced, steady-state desensitization is less pronounced for α6β2^∗, than α4β2^∗, nAChRs (Grady et al., 2012).

TC299423 was assessed for binding to 70 different receptors and other targets at a concentration of 1 μM (Supplementary Table S3). Assays included the nicotinic drug [³H]epibatidine binding to α-bungarotoxin-insensitive nAChRs, the 5-HT3 inhibitor [N-methyl-³H]GR65630 to mouse receptors in N1E-115 cells, and human receptors expressed in a clonal cell line. Other assays included monoamine oxidase A, monoamine oxidase B, 5-HT transporter, norepinephrine transporter, and dopamine transporter, as well as other. G-protein-coupled receptors, ion channels, enzymes, and transporters. The results showed that, consistent with the results reported here, TC299423 bound to neuronal nAChRs (measured by [³H]epibatidine displacement). It also bound to ATP-sensitive potassium channels, but with a much lower affinity than to nAChRs. Only these two sites showed significant binding. Thus, off-target effects are unlikely to confound our experimental data at TC299423 concentrations ≤1 μM.

Previous data show that mutant mice expressing hypersensitive α6L9′S^∗ nAChRs are more sensitive to nicotine-induced CPP than WT mice, suggesting that α6^∗ nAChR activation on its own can induce CPP (Drenan et al., 2012). Our functional data show that TC299423 exhibits a modest preference for activating α6^∗ nAChRs. Thus, CPP was used to test whether the activation of α6^∗ nAChRs by TC299423 is rewarding in mice (Figure 4). We tested the ability of TC299423 to establish CPP in both WT, and mutant α6L9′S, mice. Mice were injected with zero (saline-injected), 1, 4, 12, or 25 μg/kg TC299423 during CPP training (see Materials and Methods). TC299423 had a highly significant overall effect on the CPP difference score in the mutant α6L9′S mice [One-way ANOVA, F_(4,45) = 6.706, p = 0.0003], but not in WT mice [One-way ANOVA, F_(4,60) = 1.442, p = 0.2312] (Figures 4A,B). Post hoc comparisons (Tukey HSD) showed that the scores of α6L9′S mutant mice injected with 1 and 25 μg/kg TC299423 were significantly greater than the saline-injected controls (p < 0.05, Figure 4B). The ability of low doses of TC299423 to elicit CPP in the α6L9′S mice suggests that it is rewarding for these mice. Interestingly, the CPP difference scores for the saline-injected WT and mutant controls also suggest that WT mice habituate to the stress of repeated confinement in the least preferred chamber (i.e., their CPP difference scores are positive), while α6L9′S mutant mice become sensitized to it (i.e., their CPP difference scores are negative). For the WT mice, the CPP dose-response data were fitted to the Michaelis–Menten equation (see Materials and Methods) with an ED₅₀ value of 4 ± 3 ng/kg, whereas the lowest dose tested for the α6L9′S mice elicited a near maximal response, consistent with an ED₅₀ value < 1 ng/kg (Figures 4A,B). This confirms that mice with hypersensitive α6L9′S^∗ nAChRs were more sensitive to TC299423-induced CPP than WT mice. Thus, low doses of TC299423 are more rewarding for the hypersensitive α6L9’S mutant mice than the WT.

Even though TC299423 had no significant overall effect on CPP in the WT mice (one-way ANOVA, Tukey HSD, p = 0.23), a t-test revealed a marginally significant increase in CPP difference score for the 0.012 mg/kg TC299423-injected group, compared to the saline-injected control (p = 0.047) (Figure 4A). Using this dose, we examined the importance of individual nAChR subunits on TC299423 reward-related behavior. We compared WT mice to α6KO, α4KO, and β2KO mice, all injected with 0.012 mg/kg TC299423. Here, we observed a significant overall different response to TC299423 among the genotypes [one-way ANOVA, F_(3,47) = 4.825, p = 0.005] (Figure 4C). Genetic deletion of the α6 and β2 subunits dramatically reduced the CPP difference scores (p < 0.05), whereas deletion of the α4 subunit had no effect on the difference scores (Figure 4C). These data suggest that (1) a low dose of TC299423 (0.012 mg/kg) is weakly rewarding for WT mice, and (2) this rewarding effect is mediated by the activation of α6(non-α4)β2^∗ nAChRs.

The CPP data above suggest that low doses of TC299423 are weakly rewarding for WT mice but do not provide any information about the potential effects of TC299423 on the motivational properties of other substances, such as food and nicotine. Previous results show that nAChR partial agonists such as varenicline and (+)-PHT (Figure 1) block nicotine-induced IVSA and CPP (Rollema et al., 2007; Carroll et al., 2015). Thus, we tested the effects of TC299423 on male Wistar rats trained to press a lever for food pellets or intravenous nicotine infusions (Figure 5).

Rats were trained using 45 mg food pellets on a FR5TO20 schedule. Rats were then permitted to respond for food (n = 8) or nicotine infusions (0.03 mg/kg/inf) (n = 9) during 1 h daily sessions under the FR5TO20 schedule. TC299423 or vehicle was administered (ip, 1 mL/300 g weight, doses indicated in Figure 5) 20 min before placement in operant boxes. Compared to vehicle, rats pre-injected with doses of 0.01, 0.025, 0.05, or 0.08 mg/kg TC299423 showed no statistically significant changes in responding for food pellets or nicotine infusions [one-way ANOVA: Nicotine (p = 0.6331), Food (p = 0.9967)]. In neither case did TC299423 change the number of active lever presses. Thus, low-dose TC299423 did not inhibit the rewarding effects of food or nicotine.

Locomotor responses to novel stimuli are an index of animal exploration and anxiety that can also provide a predictive factor for responses to rewarding drugs (Antoniou et al., 2008). Mice that exhibit greater motor activity (high responders) in a novel environment are more likely to be susceptible to rewarding properties of drugs, compared to mice that exhibit low activity (low responders). Thus, we used a locomotor assay to compare the in vivo effects of low doses of TC299423 and nicotine on WT and α6L9′S mice (see Materials and Methods, Figure 6A). Mice were injected ip with saline, nicotine (0.08 mg/kg), or TC299423 (0.09 mg/kg) and their ambulatory activity in a novel environment was recorded (Figure 6A). (Doses of 0.08 mg/kg nicotine and 0.09 mg/kg TC299423 are equimolar, 0.5 μmol/kg). To determine whether the effects of nicotine and TC299423 on ambulation were driven by nAChR activation, the mice were also pre-injected with either saline or mecamylamine (1 mg/kg) (Figure 6A). Both genotype and drug treatment significantly affected ambulation [two-way ANOVA: genotype, F_(1,72) = 15.98, p < 0.005; drug effect, F_(5,72) = 7.70, p < 0.005]. Nicotine and TC299423 at these doses dramatically increased the ambulatory activity of the mutant α6L9′S, but not WT, mice (Figure 6A). The increases were blocked by mecamylamine pre-injections. Neither TC299423 nor nicotine affected WT ambulation, and there were no significant differences between the ambulatory activity of WT and α6L9′S mice pre-injected mecamylamine. Interestingly, even the saline-injected α6L9′S mice showed significantly more ambulation than the WT (p < 0.01). This difference was also blocked by mecamylamine pre-injection (Figure 6A), suggesting that it is mediated by the endogenous activation of α6L9′S^∗ nAChRs. Thus, consistent with the CPP results above and previous data (Drenan et al., 2008), the locomotor response of the mutant α6L9′S mice was more sensitive to nicotine and TC299423 than that of the WT mice. The locomotor data also confirm the activation of α6L9′S^∗ nAChRs by a low dose of TC299423 (0.09 mg/kg in this case).

Previous data show that nicotine is antinociceptive and this effect is mediated by α4β2^∗ (Damaj et al., 2007) and α6^∗ nAChRs (Wieskopf et al., 2015). Because TC299423 is an agonist for α4β2^∗ nAChRs at higher doses, we tested its antinociceptive properties at a dose of 0.3 mg/kg in WT mice using a hot plate assay, and compared its antinociceptive properties to those of 0.3 mg/kg nicotine and 0.3 mg/kg varenicline (Figure 6B). The drug treatments significantly affected responses of WT mice measured with the hot plate test [One-way ANOVA, F_(3,27) = 6.826, p = 0.0014]. The antinociception elicited by TC299423 was similar to that elicited by 0.3 mg/kg nicotine. The antinociceptive effects of both nicotine and TC299423 were significantly greater than that of varenicline (Figure 6B, p < 0.05).

The anxiolytic effects of nicotine may contribute to its rewarding effects, and appear to be mediated by α4β2^∗ nAChRs (Turner et al., 2010; Anderson and Brunzell, 2012). Marble burying behavior is a widely used (Deacon, 2006; Yohn and Blendy, 2017) measure of anxiety-related and compulsive behavior in mice, though interpretation may be complex (Thomas et al., 2009; Wolmarans de et al., 2016). To compare the anxiolytic properties of TC299423, varenicline, and nicotine, WT mice were administered saline, nicotine (0.3 mg/kg), varenicline (0.3 mg/kg), or TC299423 (0.3 mg/kg) and evaluated with the marble burying test (Figure 6C). Overall, the drugs significantly affected the number of marbles buried [One-way ANOVA, F_(3,33) = 45.65, p < 0.001]. Mice buried significantly fewer marbles following nicotine, TC299423, or varenicline administration than following saline administration. Further, mice buried fewer marbles after nicotine or TC299423 injections than varenicline injections (p < 0.001, Figure 6C). Note that neither nicotine, nor TC299423, significantly reduced locomotion in WT mice (Figure 6A), rendering it unlikely that the decrease in marble burying is caused by physical sedation.