GABA_A receptors are ligand-gated chloride ionophores that mediate the fast synaptic as well as tonic extrasynaptic inhibitory effects of GABA (Mody and Pearce, 2004). They are pentameric assemblies of proteins (Cromer et al., 2002) derived from a family of 16 genes (α1–6, β1–3, γ1–3, δ, ϵ, θ, and π; Simon et al., 2004). Despite the theoretically large number of possible pentameric permutations that can arise from these 16 genes, only relatively few combinations are found within the central nervous system (McKernan and Whiting, 1996). Most of these native receptors contain α, β, and γ subunits in a 2:2:1 stoichiometry with an alternating αβαβγ arrangement as viewed from the synapse (Minier and Sigel, 2004).

In addition to binding sites for their endogenous ligand, GABA_A receptors also possess recognition sites for a number of pharmacologically active classes of compounds, including ethanol, barbiturates, neurosteroids, convulsants, and benzodiazepines (BZs; Sieghart, 1995). It is the clinical utility of compounds acting at this latter binding site that has attracted most attention. Thus, the so-called “classical” BZs such as diazepam, chlordiazepoxide, midazolam, alprazolam, and lorazepam all exert their pharmacological effects by potentiation of the inhibitory effects of GABA on certain populations of GABA_A receptors. Hence, GABA_A receptors comprising β, γ2, and either an α1, α2, α3, or α5 subunit contain a BZ binding site and collectively represent around 75% of the total brain GABA_A population (McKernan and Whiting, 1996; Sieghart and Sperk, 2002). It is therefore these GABA_A receptor subtypes that mediate the variety of pharmacological properties demonstrated by BZs, which include anxiolysis, anticonvulsant activity, sedation, myorelaxation, and cognition impairment. In addition, BZs interact with ethanol, have abuse potential, and are associated with adverse events such as rebound anxiety upon their sudden withdrawal (Woods and Winger, 1995).

Although sedation and myorelaxation are desirable characteristics of a medication, such as midazolam, which is given prior to certain surgical procedures, or in a BZ used as a hypnotic, these properties are liabilities in an anxiolytic drug (Argyropoulos and Nutt, 1999). Accordingly, a drug which retains the anxiolytic efficacy of existing BZs but is devoid of the sedation liability would offer distinct advantages over existing BZ anxiolytics (Atack, 2003). Initially, the search for such nonsedating anxiolytics focused on nonselective partial agonists, typified by bretazenil, which have equivalent affinity for the α1-, α2-, α3-, and α5-containing GABA_A receptor subtypes but, relative to diazepam, have reduced intrinsic efficacy (Haefely et al., 1990). In preclinical species, nonselective partial agonists are nonsedating anxiolytics, but these preclinical advantages failed, for a variety of reasons, to translate into clinical utility (Atack, 2003). Most notably, in humans bretazenil showed no evidence for a dissociation between anxiolytic and sedative effects (van Stevenick et al., 1996).

An alternative approach to develop nonsedating anxiolytics is to develop compounds that selectively target the GABA_A receptor subtype(s) involved in anxiolysis while having no effects at the subtype(s) involved in the sedative effectives produced by BZ site agonists. In this regard, molecular genetic (i.e., transgenic mouse) data have highlighted the involvement of α1- and α2-containing GABA_A receptors in the sedative and anxiolytic effects of diazepam, respectively (Rudolph et al., 1999; Löw et al., 2000; McKernan et al., 2000). In addition to the α2 subtype, pharmacological evidence has also implicated the α3 subtype in GABA_A-mediated anxiety since an α3-selective inverse agonist has been shown to be anxiogenic, whereas an α3-selective agonist is anxiolytic (Atack et al., 2005; Dias et al., 2005).

Subtype selectivity at the BZ binding site may be achieved either by selective affinity or selective efficacy (Atack, 2003). The prototypic efficacy selective compound is L-838417, a compound that binds with equivalent affinity to the four GABA_A receptor subtypes possessing a BZ binding site. It is a partial agonist at the α2, α3, and α5 subtypes but an antagonist (i.e., has no effects) at the α1 subtype (McKernan et al., 2000). This compound was shown to be a nonsedating anxiolytic in rodents (McKernan et al., 2000), but marked interspecies variations in the pharmacokinetics of this compound as well as metabolic issues limited its further development (Scott-Stevens et al., 2005). In the present study, we describe the properties of a compound, TPA023, which is essentially an antagonist at α1 and α5 subtypes and a low-efficacy partial agonist at the α2 and α3 subtypes. In rodents, this compound not only had anxiolytic-like and anticonvulsant activities but also had much reduced sedation, withdrawal, and ethanol interaction liabilities. Moreover, these nonsedating anxiolytic properties were also observed in primates.

Materials and Methods

Animal procedures were all performed as stipulated in the UK Animals (Scientific Procedures) Act of 1986.

Drugs

Diazepam, chlordiazepoxide, triazolam, and FG 7142 were purchased from Sigma-Aldrich (Gillingham, UK). [³H]Ro 15-1788 (70–87 Ci/mmol) and [³H]Ro 15-4513 (20–40 Ci/mmol) were purchased from PerkinElmer Life and Analytical Sciences (Boston, MA). Bretazenil was a gift from F. Hoffman-La Roche (Basel, Switzerland). TPA023 (Fig. 1) was synthesized as described elsewhere (Carling et al., 2005). [¹⁴C]TPA023 was produced by incorporation of [¹⁴C] from [¹⁴C]CO₂ by exchange into [¹⁴C]-2-fluorobenzoic acid. This precursor was reduced to the aldehyde and reacted with 3-chloro-4-(1,1-dimethyleth-1-yl)-6-hydrazinopyridazine to produce the imine N-[¹⁴C]-2-(fluoro)benzylidene-N′-(3-chloro-4-(1,1-dimethyleth-1-yl)-pyridazin-6yl. Subsequent formation of the chloroimidate was followed by reaction with the alcohol 2-ethyl-2H-[1,2,4]-triazol-3-yl)methanol to yield [¹⁴C]TPA023, which was purified to >99% radiochemical purity and a specific activity of 55 mCi/mmol using high-performance liquid chromatography.

In Vitro Radioligand Binding

Human recombinant GABA_A receptors containing β3 and γ2 plus either α1, α2, α3, α4, α5, or α6 subunits were stably expressed in mouse fibroblast L(tk^-) cells, which were harvested, and membranes were prepared as described previously (Hadingham et al., 1993, 1996; Wafford et al., 1996). Male Sprague-Dawley rat (B&K Universal, Hull, UK) cerebellar and spinal cord P₂ membranes were prepared according to standard procedures.

A range of concentrations of TPA023 were incubated under equilibrium conditions (1-h incubation at room temperature) using the appropriate membrane preparations and 1.8 nM [³H]Ro 15-1788 as radioligand for α1-, α2-, α3-, or α5-containing human receptors and native rat receptors (with 10 μM flunitrazepam to define nonspecific binding) and 8.0 nM [³H]Ro 15-4513 for α4- or α6-containing receptors (with 10 μM Ro 15-4513 for nonspecific binding). The concentration required to reduce radioligand binding by 50% (IC₅₀) was calculated using XLfit (IDBS, Guildford, UK). These IC₅₀ values were converted to affinities (K_I values) using the Cheng-Prusoff equation (Cheng and Prusoff, 1973), for which the K_D values used were 0.92, 1.05, 0.58 and 0.45, 2.0, and 3.0 nM for binding of [³H]Ro 15-1788 to α1-, α2-, α3-, or α5-containing human receptors and rat cerebellum and spinal cord, respectively, and 5.0 and 6.5 nM for [³H]Ro 15-4513 binding to α4- and α6-containing receptors.

In Vitro Autoradiography with [¹⁴C]TPA023

Brains from two rats (male Sprague-Dawley; 260–300 g) were removed following decapitation, frozen in isopentane at -40°C on dry ice, and then stored at -70°C. Sections (12 μm) were subsequently cut with one brain in the sagittal plane and the other in the coronal plane and thaw-mounted onto glass microscope slides. Microscope slides were then incubated for 90 min in 50 mM phosphate buffer (pH 7.4) using 2.3 nM [¹⁴C]TPA023 (55 mCi/mmol) with alternate sections incubated in the absence and presence of 10 μM flunitrazepam, which was used to define nonspecific binding. Sections were then washed twice for 30 s each, dipped in distilled water, rapidly dried, and apposed to Kodak BioMax film (Eastman Kodak, Rochester, NY) for 7 weeks.

In Vitro Efficacy Measurements

The in vitro efficacy of TPA023 at human recombinant GABA_A receptors was assessed in the stably transfected mouse L(tk^-) cell lines used for the in vitro binding assays described above using whole-cell patch clamping (Brown et al., 2002). In brief, cells were grown as a monolayer on glass microscope coverslips, and membrane patches, with resistances in the region of 5 to 10 MΩ, were formed and then burst by suction. Cells were voltage clamped at -20 mV using an Axopatch 200B amplifier (Axon Instruments, Foster City, CA) with drug or wash solution application being made via a triplebarreled pipette system. Test compounds (chlordiazepoxide or TPA023) were preincubated for 30 s, and then a concentration of GABA equivalent to an EC₂₀ was applied for 5 s. For each cell, the nonselective agonist chlordiazepoxide (3 μM) was used as an internal standard to confirm the presence and extent of BZ site-mediated modulatory responses. Chlordiazepoxide, which typically increased the GABA EC₂₀-equivalent currents by between 80 and 200%, was used since its relatively low affinity and water solubility meant that it could be easily washed out prior to subsequent incubations with increasing concentrations of TPA023.

Current amplitudes were measured as percent modulation of the peak currents seen with GABA alone, and the data were transformed by expressing the potentiation of TPA023 relative to that seen with chlordiazepoxide. Concentration-effect curves for individual cells were constructed using GraphPad Prism (GraphPad Software, Inc., San Diego, CA), with curve-fitting being performed using a nonlinear least-squares method. From these data, maximum efficacy and EC₅₀ were determined for each cell (Brown et al., 2002).

Inhibition of in Vivo Binding of [³H]Ro 15-1788

The occupancy of rat and mouse brain BZ binding sites by TPA023 was measured by its ability to inhibit the in vivo binding of [³H]Ro 15-1788, essentially as described previously (Atack et al., 1999). Thus, animals pretreated with TPA023 received a tail-vein injection of [³H]Ro 15-1788 (70–87 Ci/mmol diluted 1:150 with saline; 5 μl/g for mice, 1 μl/g for rats; PerkinElmer Life and Analytical Sciences) 3 min prior to killing by stunning and decapitation. Brains were rapidly removed and homogenized in 10 volumes of ice-cold 50 mM potassium phosphate buffer (pH 7.4) and 300-μl aliquots filtered over Whatman GF/B filters (Whatman, Clifton, NJ). Filters were washed with 10 ml of ice-cold buffer, placed in scintillation vials, and after the addition of scintillation fluid, radioactivity (= membrane bound [³H]Ro 15-1788) was counted on a Beckman LS6500 scintillation counter (Beckman Coulter, Fullerton, CA). For each study, a separate group of rats or mice were dosed with bretazenil (5 mg/kg i.p. in 100% polyethylene glycol 300) to define the level of nonspecific in vivo binding of [³H]Ro 15-1788. In each experiment, the value of nonspecific binding defined using bretazenil was subtracted from all groups to give values of specific binding. The percent occupancy was defined as the percentage by which specific binding in the vehicle-treated group was inhibited by drug treatment. For example, if specific binding in the vehicle group was 2500 cpm and in the drug group was 250 cpm, then the occupancy would be 90%.

For rat studies, male Sprague-Dawley rats were used (250–310 g) with a vehicle of 0.5% methylcellulose and a dose volume of 1 ml/kg. To establish a dose-occupancy relationship, rats (n = 3–6/group) were dosed with either vehicle, bretazenil, or various doses of TPA023 (0.1–10 mg/kg), injected with [³H]Ro 15-1788 after 27 min, and killed at 30 min postdosing. Alternatively, for the time course study, rats (n = 5–7/group) were dosed with vehicle, bretazenil, or either 1, 3, or 10 mg/kg TPA023 and killed 0.5, 1, 3, 6, or 18 h later, having each received [³H]Ro 15-1788 3 min prior to killing.

For mice studies, male Swiss-Webster mice were used (23–31 g; B&K Universal) with a vehicle of 0.5% methylcellulose and a dose volume of 10 ml/kg. For the Rotarod studies (see below), as soon as mice (n = 8–10/group) had completed their 2-min trial, they were taken, injected with [³H]Ro 15-1788, killed 3 min later, and processed as described above (in this experiment, a separate group of untested animals was used to define nonspecific binding using bretazenil). In the case of the anticonvulsant assay, mice that had received injections of pentylenetetrazole could not be used for occupancy measurements, and therefore satellite groups of mice (n = 5–6/group) were dosed with vehicle or TPA023, injected with [³H]Ro 15-1788 at t = 27 min, and then killed, and occupancy was measured 3 min later.

Rat Anxiolysis Assays

Elevated Plus Maze. Male Sprague-Dawley rats (approximately 250–300 g; n = 17–18/group) received p.o. either vehicle (0.5% methylcellulose; dose volume, 1 ml/kg), TPA023 (0.3, 1, or 3 mg/kg), or as a positive control and for comparison purposes the BZ full agonist chlordiazepoxide (5 mg/kg i.p. made up in isotonic saline). Thirty minutes after dosing, rats were placed in the center of an elevated plus maze for a 5-min trial, during which time their movements were monitored via a video tracking system (Dawson and Tricklebank, 1995). Their movements were analyzed using “Flexible Maze Software”, which calculated the time the rats spent on the open arms of the maze as well as the total distance traveled.

Fear-Potentiated Startle. Fear-potentiated startle was performed generally as described elsewhere (Kehne et al., 1988). Male Sprague-Dawley rats (250–290 g) were initially exposed to 10 100-dB tones and divided into four groups (n = 15/group) matched for their mean baseline startle amplitudes as measured using SR-LAB stabilimeter chambers (San Diego Instruments, San Diego, CA). Next, rats received 2 consecutive days of conditioning, during which they were trained to associate the presentation of a light with the delivery of a mild foot-shock (10 presentations of light for 3.7 s each, the final 0.5 s of which was accompanied by a mild 0.6-mA foot-shock). On the following day (day 4) rats were tested for fear-potentiated startle. Animals were dosed (p.o.) with either vehicle (0.5% methylcellulose; 1 ml/kg) or TPA023 (0.3, 1, or 3 mg/kg) and placed in the startle apparatus 30 min later. Following a 5-min acclimatization period, rats received 10 100-dB tones (50-ms duration, 30-s intertrial interval) to partially habituate them to the startle stimulus. This was followed by the test session itself, which was comprised of random noise stimuli (10 95 dB tones), half given 3200 ms after the presentation of the conditioning stimulus (light) and half given in darkness. All responses were measured over a 100-ms window beginning immediately on presentation of the tone, and the mean startle amplitude for each stimulus type was calculated for each rat. For every animal, the startle amplitude in the light minus the startle amplitude in the dark (the “difference score”) was also calculated.

Conditioned Suppression of Drinking. Male Hooded Lister rats (200–250 g; B&K Universal) were water-deprived for 22.5 h in each 24-h period for 2 days before and during the course of the experiment. On the 1st day of the experiment, rats were placed in the operant chamber for 30 min, and licking of the metal spout was reinforced with 0.1 ml of water according to a random interval (RI) 10-s schedule. This procedure rapidly established licking, and on the 2nd day, the schedule was increased to RI 60 s. On the 3rd day, the conditioning day, the session length was increased from 30 to 60 min. At 15, 32, and 48 min after the beginning of the session, the house light was illuminated for 60 s, and in the last second, a 0.4-mA foot-shock was delivered. Thus, the rats received three light-shock pairings, and as a consequence, when the light was presented for the third time, it suppressed their licking rates. The lick rate prior to and during the light presentation was recorded, and the ratio of the latter to the former defined the suppression ratio. Rats with mean suppression ratios ≥0.15 were excluded from compound testing; i.e., only rats that had learnt the light-shock relationship, as indicated by a mean suppression ratio <0.15, were included. On the 4th and 5th days, the rats were given a further 30-min RI 60-s session. On the 6th day, the test day, the procedure was identical to that on the conditioning day with the exception that electric shocks were not delivered during light presentations. The rats were given either vehicle (0.5% methylcellulose; 1 ml/kg) or one of three doses of TPA023 (1, 3, or 10 mg/kg) or diazepam (3 mg/kg p.o.) 30 min prior to the test session (n = 10–12/group).

Rat Sedation Assay

The response sensitivity test is a rodent model of sedation and/or behavioral disruption in which food-deprived rats are trained to pull a chain for access to food pellets according to a RI 30-s schedule. Full BZ agonists such as diazepam dose-proportionally reduce chain-pulling rates as a consequence of their sedative and muscle relaxant properties (Bayley et al., 1996).

Male PVG rats (270–350 g; B&K Universal), which were maintained at 85% of their free feeding weight by postsessional feeding, were trained to pull a chain suspended from the center of a standard operant conditioning chamber to receive a food pellet reward. Initially, one pellet was dispensed for a single chain pull, but the regulated probability interval schedule progressively increased until the average pellet-pellet interval was 30 s. On the testing day, rats (n = 8/group) were given (p.o.) either vehicle (0.5% of methylcellulose; 1 ml/kg) or one of three doses of TPA023 (3, 10, or 30 mg/kg) or 10 mg/kg diazepam immediately before the 60-min test session began.

Mouse Rotarod Assay

In the morning of the experiment, male BTKO mice (24–30 g; B&K Universal) were trained to walk on a Rotarod (Ugo Basile, Comeno, Italy) revolving at 16 times per minute until they could successfully complete three consecutive 120-s sessions without falling off. In the afternoon, mice (n = 8/group) were given either vehicle (0.5% methylcellulose; 10 ml/kg), TPA023 (1, 3, 10, or 30 mg/kg i.p.), or diazepam (3 mg/kg p.o.) 30 min before being placed on the Rotarod. The latency (seconds) to fall from the Rotarod was recorded. If the mouse did not fall from the Rotarod during the 120-s trial, the latency was recorded as 120 s. Immediately after the Rotarod trial, mice were given an i.v. injection of [³H]Ro 15-1788, and receptor occupancy was determined as described above.

To evaluate the interaction between TPA023 and ethanol, procedures were carried out as described above, except that mice (n = 10/group) were pretreated with a subthreshold dose of ethanol (1.5 g/kg i.p.) that was coadministered with the same compounds and doses as described above. In addition, a vehicle-treated group that did not receive ethanol was also included. Again, upon completion of the Rotarod trial, mice were taken for occupancy measurements.

Primate Anxiolysis and Sedation Assays

Conditioned Emotional Response. Ten adult male squirrel monkeys (Saimiri sciureus, 0.7–1.2 kg) were placed in a standard Plexiglas primate chair placed with the distal portion of their tails clamped in a stock. Monkeys were trained to press a lever to obtain a fruit juice reward on a fixed reinforcement schedule that maintained their response rate >10/min. An experimental session was initiated by the extension of the lever into the operant chamber and the illumination of the house light. At an interval between 5 and 35 min after commencement of the session, a red cue light (conditioned stimulus) was illuminated for 60 s, during the last 0.5 s of which the monkeys occasionally (p = 0.1) received a mild electric tail-shock (1–7 mA). The end of the session was signaled by turning off the house light and retracting the response lever. On days when drugs were administered, the schedule differed only in that tail-shock was not delivered.

Once lever-pressing rates were stable and reliably suppressed during presentation of the red cue light (i.e., the monkey had made the association between the red light and possible tail shock), animals were dosed with vehicle (0.5% methylcellulose; 2 ml/kg) or TPA023 (0.3, 1, or 3 mg/kg p.o.) 30 min prior to testing. The experiment used a pseudo-Latin square design in which each animal received vehicle and each dose of TPA023. Lever pressing during the illumination of the conditioned stimulus, expressed as a percentage of the lever presses recorded in the 60 s prior to illumination, was subjected to one-way analysis of variance with repeated measures followed, where appropriate, by paired t tests.

Squirrel Monkey Lever Pressing. In this assay, behavioral disruption, for example, sedation or myorelaxation, is assessed by the ability of a compound to reduce the rate of responding to obtain a fruit juice reward and is analogous to the rat response sensitivity (chain-pulling) assay. Seven adult male squirrel monkeys (Saimiri sciureus, 0.7–1.2 kg) were placed in a standard Plexiglas primate chair. Monkeys were trained to press a lever to obtain a fruit juice reward on an fixed reinforcement schedule prior to being transferred to a regulated probability interval schedule under which they received a reinforcer, on average, every 30 s. An experimental session was initiated by the extension of the lever into the operant chamber and the illumination of the house light. After 30 min, the end of the session was signaled by turning off the house light and retracting the response lever.

Animals were dosed with vehicle (0.5% methylcellulose; 2 ml/kg) or TPA023 (1, 3, or 10 mg/kg p.o.) 30 min prior to testing. The experiment used a pseudo-Latin square design in which each animal received vehicle and each dose of TPA023. The lever-pressing rates of vehicle- or TPA023-treated monkeys were expressed as a percentage of the lever press rates recorded on the preceding day.

Pentylenetetrazole Anticonvulsant Assay

Male Swiss-Webster mice (24–30 g, n = 6/group; B&K Universal) were dosed i.p. with either vehicle (0.5% methylcellulose; 10 ml/kg) or doses of TPA023 (0.1, 0.3, 1, 3, or 10 mg/kg). Mice were injected again 30 min later with pentylenetetrazole (120 mg/kg s.c.), placed in a Plexiglas box, and observed for 30 min for full tonic convulsions. Mice were either classified as “unprotected” if they convulsed during the 30-min observation period or “protected” if they had not convulsed during that time. The percentage of mice protected at each dose was calculated, and the ED₅₀ dose was estimated using probit analysis.

Withdrawal Studies

To assess whether withdrawal symptoms can be precipitated upon cessation of chronic (7-day) treatment with TPA023, mice were dosed i.p. for 7 days with either vehicle (0.5% methylcellulose; 10 ml/kg), TPA023 (10 mg/kg), or triazolam (10 mg/kg). On the 8th day, mice (n = 8–10/group) were given the nonselective inverse agonist FG 7142 (40 mg/kg i.p.) in an attempt to precipitate withdrawal. Mice were then placed in an observation box for 1 h and scored according to whether or not they developed clonic seizures. FG 7142 was used to precipitate withdrawal since partial inverse agonists appear to be more sensitive in detecting an underlying dependence than the antagonist flumazenil (Martin et al., 1998).

A related experiment was performed in which the ability of an acute dose of TPA023 to substitute for FG 7142 and precipitate withdrawal in mice treated for 7 days with triazolam was evaluated. In this experiment, mice (n = 8–10/group) were dosed i.p. for 7 days with either vehicle (0.5% methylcellulose; 10 ml/kg) or triazolam (10 mg/kg). On day 8, these two groups were split, and animals were given either FG 7142 (40 mg/kg i.p.) or TPA023 (10 mg/kg i.p.). Scoring for clonic seizures was performed as described in the preceding paragraph.

Statistical Analyses

The effects of treatment were compared across groups using either parametric or nonparametric one-way ANOVA with post hoc comparisons of drug treatment being made relative to vehicle-treated groups using the appropriate parametric (Dunnett's or paired t test) or nonparametric (Dunn's multiple comparison) tests. Values shown are mean ± S.E.M.

Results

In Vitro Binding

TPA023 shows high affinity for α1-, α2-, α3-, and α5-containing GABA_A receptor subtypes (respective K_i values = 0.27, 0.31, 0.19, and 0.41 nM) but much lower (50- to 2000-fold less) affinities for the α4 and α6 subtypes (K_i values = 60 and 418 nM, respectively; Table 1). The affinity of TPA023 for recombinant human GABA_A receptor was comparable with that of native rat brain receptors. Thus, in rat cerebellum (primarily α1 subtype) the K_i of 0.33 nM correlated well with the K_i of 0.27 nM observed in recombinant human α1-containing receptors (Table 1). Similarly, in spinal cord (α2/α3 subtypes) a K_i of 0.32 nM compared well with the K_i values observed in recombinant human receptors of 0.31 and 0.19 nM for α2 and α3 subtypes, respectively.

Figure 2 shows autoradiographic images of rat brain sections labeled with [¹⁴C]TPA023 alone (total binding) or [¹⁴C]TPA023 in the presence of flunitrazepam (nonspecific binding). In the absence of flunitrazepam (Fig. 2, left panels), binding sites for [¹⁴C]TPA023 showed a relatively wide-spread distribution in cerebellum as well as cortical and subcortical regions, typical of the ubiquitous distribution of BZ binding sites in the brain. However, in the presence of flunitrazepam (Fig. 2, right panels), no appreciable radioligand binding could be detected. The complete blockade of [¹⁴C]TPA023 binding by flunitrazepam suggests that TPA023 is highly specific for the BZ site of GABA_A receptors and does not bind with high affinity to other recognition sites in the brain.

In Vitro Efficacy

TPA023 potentiated the currents produced by an EC₂₀-equivalent concentration of GABA in human recombinant GABA_A receptors containing β3, γ2 plus either α1, α2, α3, or α5 subunits as shown in Fig. 3, with the quantitation of these concentration-effect curves being presented in Table 2. GABA_A receptors containing either an α1 or α5 subunit show essentially no modulation of the GABA EC₂₀ current by TPA023 (maximum modulation = 1 and 6%, respectively), and therefore TPA023 is essentially an antagonist at these subtypes. On the other hand, TPA023 modulated the GABA EC₂₀ responses by 12 and 33% at α2- and α3-containing receptors, respectively (Table 2). When expressed relative to the modulation produced by the nonselective full agonist chlordiazepoxide, the efficacy of which was measured in the same cells used to characterize the TPA023 responses, TPA023 had a relative efficacy of 0.11 and 0.21 at the α2 and α3 subtypes, respectively (Table 2 and Fig. 3, inset).

In Vivo Binding

For doses of TPA023 in the range 0.1 to 10 mg/kg (p.o.), the occupancy of rat brain BZ binding sites (i.e., the inhibition of in vivo [³H]Ro 15-1788 binding) was dose-dependent (Fig. 4A), with the dose of TPA023 estimated to inhibit by 50% the in vivo binding of [³H]Ro 15-1788 (ID₅₀) being 0.42 mg/kg. Based upon an ID₅₀ of 0.42 mg/kg, the percent occupancy at doses ranging between 0.03 and 30 mg/kg was calculated (Table 3) and forms the basis for the interpretation of the rat behavioral data.

In addition to being dose-dependent, occupancy of rat brain BZ binding sites by TPA023 was also time-dependent (Fig. 4B). Thus, at 1, 3, and 10 mg/kg (p.o.), occupancy was greatest 0.5 h after dosing, with respective maximum occupancies of 76, 89, and 98% falling to 29, 34, and 73% 6 h after dosing with only ≤12% remaining at 18 h.

TPA023 Has Anxiolytic-like Activity in Rats

Elevated Plus Maze.Figure 5A shows the time spent on the open arms of the maze expressed as a percentage of the total trial time (5 min; left panel) as well as the total distance traveled (right panel). A one-way ANOVA showed that there was a significant effect of treatment on the percentage of time on the open arms (F_4,83 = 8.30, p < 0.001). More specifically, vehicle-treated animals spent 17 ± 3% of their time on the open arms, whereas the time for rats treated with the full agonist chlordiazepoxide (30 ± 3%) was significantly (p < 0.01; Dunnett's t test) longer. TPA023 produced a dose-proportional increase in the time spent on the open arms, which achieved statistical significance (p < 0.05 and 0.01) at doses of 1 and 3 mg/kg (28 ± 3 and 34 ± 2% time spent on open arms, respectively). Thus, TPA023 had anxiolytic-like activity in the elevated plus maze assay at doses (1 and 3 mg/kg p.o.) corresponding to 70 and 88% occupancy (see Table 3), whereas chlordiazepoxide was anxiolytic at a dose (5 mg/kg i.p.) corresponding to 21 ± 5% occupancy (data not shown).

Treatment also had an effect on the total distance traveled (F_4,83 = 4.91, p < 0.01; one-way ANOVA). Hence, the distance traveled was significantly greater (p < 0.01 and 0.05; Dunnett's t test) for chlordiazepoxide and 3 mg/kg TPA023 (30 ± 2 and 31 ± 3 m, respectively) relative to vehicle-treated rats (20 ± 2 m), indicating that at these doses neither compound caused sedation, which would be reflected in a reduced total distance traveled. Although the locomotor-stimulating effect of TPA023 suggests little evidence of sedation, it also raises the possibility that the anxiolytic effect is a false positive caused by this increased activity (Dawson and Tricklebank, 1995). Consequently, the effects of TPA023 were assessed in two nonlocomotor-based anxiolytic assays, fear-potentiated startle, and conditioned suppression of drinking.

Fear-Potentiated Startle.Figure 5B shows the effects of TPA023 on the magnitude of fear-potentiated startle produced by a 95-dB auditory stimulus in the absence and presence of a conditioning stimulus (light). The panel on the left shows the startle amplitude under both the dark (safe) and light (threatened) conditions, whereas that on the right shows the mean within-subject difference between the startle amplitudes in the dark and light conditions (i.e., the fear-potentiated startle response). Across groups, there was no significant effect of treatment on the baseline (dark) response, suggesting that TPA023 did not produce overt sedation or myorelaxation effects. Within-group analyses showed that the startle responses in the light were significantly greater than in the dark for the vehicle and 0.3 mg/kg (p < 0.001) but not the 1 or 3 mg/kg groups. These differences represent a clear fear-potentiated startle response in rats treated with either vehicle or 0.3 mg/kg TPA023.

For each animal the difference score (light minus dark) was calculated (Fig. 5B, right panel), and an analysis of these difference scores showed that there was a significant effect of treatment (F_3,56 = 5.68, p < 0.01; one-way ANOVA) on the extent of this fear-potentiated startle. More specifically, the fear-potentiated startle responses were significantly attenuated at doses of 1 and 3 mg/kg TPA023 (p < 0.01; Dunnett's t test) such that responses in the dark and the light were comparable.

Conditioned Suppression of Drinking.Figure 5C shows the mean number of licks in the period prior to and during a conditioned stimulus (left panel), with the extent to which the lick rate was suppressed (the suppression ratio) being shown in the right panel. A one-way ANOVA showed that there was a significant effect of treatment on the suppression ratio (F_4,52 = 3.82, p < 0.001). Hence, presentation of the conditioned stimulus (light) produced a dramatic attenuation in the lick rate as the animals anticipated receiving a foot shock. However, this attenuation was significantly reduced (p < 0.05) by the nonselective agonist diazepam (3 mg/kg p.o.) as well as by 3 and 10 but not 1 mg/kg TPA023. In addition, at no dose did TPA023 reduce the rate of responding in the period prior to presentation of the conditioned stimulus, suggesting that at these doses TPA023 does not cause sedation. Thus, in this assay, TPA023 had efficacy at doses corresponding to 88 and 96% occupancy but not at a 1 mg/kg dose (70% occupancy) that was efficacious in the elevated plus maze and fear-potentiated startle assays.

TPA023 Is Nonsedating in Rats

In addition to implicit measures of sedation in the anxiolytic assays (i.e., total distance traveled in the elevated plus maze and responses prior to presentation of the conditioned stimulus in the fear-potentiated startle and conditioned suppression of drinking assays), the propensity for TPA023 to cause sedation was evaluated in the rat chain-pulling assay (Fig. 6). Data collapsed across the 1-h trial is presented in Fig. 6A, whereas Fig. 6B illustrates the time course of responding, with response rates being divided into 10-min intervals. In this assay, vehicle-treated animals responded at around 75% of their baseline response rates. This is in part due to a reduction in the response rate over the duration of the trial, presumably due to fatigue and/or reduction in thirst. Statistical analysis showed that there was a significant effect of treatment on performance averaged over the trial (Fig. 6B; F_4,35 = 9.1, p < 0.0001; one-way ANOVA) and, more specifically, that the nonselective agonist diazepam (10 mg/kg p.o.) produced a marked impairment in the rate of responding such that the response rate is less than one-third that of vehicle-treated animals (p < 0.01; Dunnett's t test). Even at a dose of 30 mg/kg (corresponding to an occupancy of 99%; Table 3), TPA023 did not produce a significant impairment in responding rates averaged over the 60-min period of the assay (Fig. 6A). More detailed analysis over the time course of the experiment (Fig. 6B) showed that at each time point, there was a significant effect of treatment (F_4,35 = >4.32, p < 0.01; one-way ANOVA) with diazepam being significantly different from vehicle animals at each time point (p < 0.01; Dunnett's t test). On the other hand, although there was an apparent effect of TPA023 over the first 20-min performance, this was not statistically significantly different from vehicle at any time point (p > 0.05). It is also worth noting that this tendency for reduced performance with TPA023 at the early time points may be a nonspecific effect (for example, compound palatability) since it does not reflect the pharmacokinetics of the compound. Thus, maximum plasma concentrations of TPA023 are achieved 0.5 to 2 h after oral dosing (data not shown), yet performance in the chain-pulling task at 0.5 to 1 h is comparable with vehicle-treated animals.

TPA023 Is a Nonsedating Anxiolytic in Primates

Figure 7 shows the effects of TPA023 in the squirrel monkey-conditioned emotional response and lever-pressing assays, which are, respectively, anxiolytic and sedation assays analogous to the rat-conditioned suppression of drinking and chain-pulling assays. In the conditioned emotional response assay, presentation of the conditioned stimulus (light) reduced the responding rate to less than 10% of baseline levels (i.e., responding rate in the preceding, drug-free day). There was a statistically significant effect of treatment on the responding rate upon presentation of the conditioning stimulus (F_3,26 = 9.46, p < 0.05). Hence, the attenuation in response could be reversed in a dose-dependent manner by TPA023 at all doses tested (0.3, 1, and 3 mg/kg; Fig. 7A), with responding rates rising from 43% (t₉ = -3.72, p < 0.05) to 76% (t₉ = -4.87, p < 0.05) of baseline between doses of 0.3 and 3 mg/kg. In the lever-pressing test of sedation, there was a tendency for the rates of responding to be lower in TPA023 compared with vehicle-treated animals, but none of these differences achieved statistical significance (Fig. 7B).

TPA023 Has a Reduced Ethanol Interaction Compared with Diazepam

Figure 8 shows Rotarod performance at different doses of TPA023 (1–30 mg/kg i.p.) compared with diazepam in the absence (Fig. 8A) and presence (Fig. 8B) of a subthreshold dose of ethanol (1.5 g/kg i.p.). Upon completion of the Rotarod test, mice were taken and occupancy measured using an in vivo [³H]Ro 15-1788 binding assay, which permits behavioral performance to be plotted as a function of BZ binding site occupancy (Fig. 8C).

There was a significant effect of treatment on Rotarod performance (p < 0.01; Kruskal-Wallis nonparametric one-way ANOVA). Hence, in the absence of ethanol, all vehicle-treated mice were able to complete the 120-s trial without impaired performance (Fig. 8A), and none of the TPA023 groups, including the highest dose tested (30 mg/kg i.p., corresponding to 100% occupancy; mean trial time = 112 ± 8 s), significantly impair Rotarod performance. On the other hand, 3 mg/kg p.o. diazepam (corresponding to 52% occupancy) produced a significant impairment in performance (p < 0.05; Dunn's multiple comparison test), with mice lasting, on average, 75 ± 17 s before falling off the Rotarod.

A dose of ethanol (1.5 g/kg i.p.) was selected such that it produced no impairment in performance by itself (Fig. 8B). A comparison of the performance between ethanol-treated animals showed a clear effect of drug treatment (p < 0.0001; Kruskal-Wallis nonparametric one-way ANOVA), with significant impairments being observed in the 30 mg/kg TPA023 and diazepam groups (p < 0.05 and 0.001, respectively; Dunn's multiple comparison test). Hence, when ethanol was given in combination with diazepam (3 mg/kg, receptor occupancy, 50%) it had a profound effect on the ability of mice to remain on the Rotarod (latency, 9 ± 2 s). A dose of TPA023 (3 mg/kg), which produced occupancy (58 ± 6%) comparable with that seen with 3 mg/kg diazepam (50%), showed no sign of impairment with ethanol, indicating that the ethanol interaction of TPA023 is much reduced relative to diazepam (Fig. 8C). Indeed, it is only at a dose of TPA023 (30 mg/kg i.p.) corresponding to an occupancy of 100% that any significant interaction between ethanol and TPA023 is observed, and even then it is less pronounced than that produced by diazepam (Fig. 8C).

TPA023 Is Anticonvulsant

Figure 9A shows that in vehicle-treated animals, pentylenetetrazole (120 mg/kg s.c.) caused tonic seizures in 6:6 mice (100%). TPA023 (i.p.) gave dose-proportional protection against pentylenetetrazole-induced convulsions with a dose of 10 mg/kg, giving full protection (6:6 animals), with the ED₅₀ for this anticonvulsant effect being 1.4 mg/kg. With respect to in vivo [³H]Ro 15-1788 binding measured in a satellite group of mice (Fig. 9B), occupancy was dose-dependent with an ID₅₀ of 1.7 mg/kg and 10 mg/kg, which gave full protection against pentylenetetrazole-induced seizures and produced an occupancy of 84 ± 2%. Thus, the correspondence of the anticonvulsant ED₅₀ (1.4 mg/kg) and the occupancy ID₅₀ (1.7 mg/mg) indicates that an occupancy of around 50% affords 50% protection from pentylenetetrazole-induced seizures.

TPA023 Is Neither Associated with Dependence nor Does It Precipitate Withdrawal

Figure 10A shows the effects of termination of 7-day treatment with either vehicle, TPA023, or the nonselective full agonist triazolam, with withdrawal effects being precipitated on day 8 by a single injection of the nonselective partial inverse agonist FG 7142. Clearly, the dose of FG 7142 used (40 mg/kg i.p.) does not induce seizure activity since the vehicle/FG 7142 group demonstrated no seizure activity. Moreover, 7-day treatment with TPA023 (10 mg/kg i.p., which corresponds to around 80% occupancy; see Figs. 8 and 9) does not appear to alter the sensitivity to FG 7142 on day 8 since again no seizure activity (including myoclonic seizures) was observed. On the other hand, treatment for a week with triazolam (10 mg/kg i.p.) rendered animals much more sensitive to the effects of FG 7142, with 50% of animals experiencing a clonic seizure. This is presumably a consequence of triazolam producing an alteration in GABA tone such that a dose of FG 7142 which otherwise would have no effect was able to cause convulsant activity.

A similar experiment was performed (Fig. 10B) where in addition to FG 7142 (40 mg/kg i.p.), TPA023 (10 mg/kg i.p.) was also used to try precipitate withdrawal following treatment for a week with either vehicle or triazolam (10 mg/kg i.p.). Like before, FG 7142 produced no seizure activity in vehicle-treated mice, but in triazolam-treated mice, there was a 50% incidence of clonic seizures. Dosing mice with TPA023 on day 8 produced no seizure activity (including myoclonic seizures) in mice treated for a week with either vehicle or triazolam.

Discussion

TPA023 is a triazolopyridazine possessing high (subnanomolar) affinity for GABA_A receptors containing an α1, α2, α3, or α5 subunit and much lower affinity for those receptors possessing an α4 or α6 subunit and, as shown autoradiographically using [¹⁴C]TPA023, has no appreciable affinity for other binding sites in the brain. It is a structural analog of L-838417, the latter of which like TPA023 lacks efficacy at the α1 subtype (McKernan et al., 2000; Rowlett et al., 2005). However, the two compounds differ in that whereas L-838417 has efficacy relative to the full agonist chlordiazepoxide of 0.34 and 0.39 at the α2 and α3 subtypes, respectively, the corresponding efficacies for TPA023 were 0.11 and 0.21. Moreover, whereas L-838417 has appreciable efficacy at the α5 subtype (relative efficacy, 0.36), TPA023 is essentially an antagonist at this subtype (relative efficacy, 0.05), which given that α5-selective inverse agonists can enhance aspects of cognition (Chambers et al., 2004), suggests that TPA023 may have a reduced cognition impairing liability compared with L-838417 (although this remains to be systematically evaluated). In addition, the variable interspecies difference in the pharmacokinetics of L-838417 (Scott-Stevens et al., 2005) precluded a detailed analysis of the in vivo properties of this compound. Thus, compared with L-838417, TPA023 has greater subtype selectivity (i.e., should produce essentially no in vivo effects via the α5 subtype) and has lower efficacy at the α2- and α3-containing subtypes. It should be emphasized that the decision to further develop TPA023 rather than L-838417 was not due to the lower α2 and α3 efficacies of TPA023 compared with L-838417 but rather was related to metabolic issues.

TPA023 readily penetrates the brain and gives time- and dose-dependent occupancy of GABA_A receptor BZ binding sites with a rat brain ID₅₀ of 0.42 mg/kg p.o. In mice, TPA023 was presumably well behaved in terms of formulation, absorption, and metabolism since the ID₅₀ values from three separate [³H]Ro 15-1788 in vivo binding experiments were relatively consistent (2.4, 2.1, and 1.7 mg/kg). Since the potency of BZ site occupancy by TPA023 appeared reproducible, doses producing behavioral responses in rats were interpreted in terms of BZ site occupancy derived from the rat dose-occupancy relationship (Fig. 4A, Table 3).

Given that TPA023 possesses a novel in vitro efficacy profile, it was evaluated in a number of assays of anxiety and sedation in rodents as well as in primates (squirrel monkeys; Table 4). TPA023 was not only anxiolytic in rats in an unconditioned model of anxiety (the elevated plus maze) but also in conditioned models (fear-potentiated startle and conditioned suppression of drinking). Furthermore, the more stringent conditioned suppression of drinking assay required higher levels of occupancy for efficacy (3 mg/kg, 88% occupancy) compared with the milder anxiety state produced in the elevated plus maze (1 mg/kg; 70% occupancy). In the elevated plus maze, the level of occupancy required for a significant anxiolytic effect (70%) is clearly much higher than the occupancy required for the positive control, nonselective full agonist chlordiazepoxide (21%). Since partial agonists require higher levels of occupancy to produce an anxiolytic response relative to full agonists (Facklam et al., 1992), these data clearly indicate that TPA023 behaves as a partial agonist in vivo.

In the rat response sensitivity assay of sedation, there was only very mild and transient sedation at the maximum dose tested (30 mg/kg, 99% occupancy). This lack of overt sedation was confirmed in the mouse Rotarod assay, in which mice showed no impairment of performance at 30 mg/kg (>99% occupancy). Further indirect measures of locomotor activity also confirmed the lack of sedative effects of TPA023 since it did not impair total distance travelled on the elevated plus maze or response rates in the absence of the conditioned stimulus in either the fear-potentiated startle or conditioned suppression of drinking assays.

Unlike nonselective full agonists, TPA023 showed only a mild interaction with ethanol. This mild interaction with ethanol (i.e., significant impairment in Rotarod performance in mice treated with TPA023 plus ethanol but not TPA023 alone) only occurred at a dose (30 mg/kg) giving 100% occupancy of mouse brain BZ binding sites. Given that TPA023 has some, albeit very low, affinity for α4- and α6-containing GABA_A receptors, it is possible that the ethanol interaction could be due to a modest degree of occupancy at these subtypes. Alternatively, the reduced ethanol interaction relative to the full agonist diazepam might be α2- and/or α3-mediated and could merely reflect the much lower intrinsic efficacy of TPA023.

The in vivo properties of TPA023 are presumably mediated via α2- and/or α3-containing GABA_A receptors. As such, it is not possible to determine which of these subtypes mediates the anxiolytic effects of TPA023, although the greater relative efficacy at the α3 subtype (0.21) compared with the α2 subtype (0.11) might suggest that the α3 is the subtype responsible for the anxiolytic-like behavior. This would disagree with data from transgenic mice (Löw et al., 2000) in which the α2 or α3 subunit is rendered insensitive to BZ binding by point-mutating the histidine residue crucial for BZ binding to an arginine (Wiedland et al., 1992). Clearly, it could be argued that the relatively low α2 efficacy of TPA023 is sufficient to mediate the anxiolytic effects of this compound. However, the fact that an α3-selective inverse agonist and an α3-selective agonist are anxiogenic and anxiolytic, respectively, suggests that this receptor subtype plays a role in anxiety (Atack et al., 2005; Dias et al., 2005).

The lack of significant sedation in rats, mice, and rhesus monkeys is consistent with the lack of efficacy of TPA023 at α1-containing GABA_A receptors. Hence, α1-containing GABA_A receptors mediate the sedative effects of nonselective BZs as shown not only with transgenic mice (Rudolph et al., 1999; McKernan et al., 2000; Rudolph and Möhler, 2004) but also pharmacologically, in that compounds devoid of α1 efficacy are nonsedating (McKernan et al., 2000; Johnstone et al., 2004; Rowlett et al., 2005). Moreover, compounds with selectivity for the α1 subtype, such as zolpidem and indiplon, are hypnotics (Sanger and Depoortere, 1998; Foster et al., 2004). Nevertheless, the α1 subtype is not solely responsible for the sedating properties of diazepam, and the other GABA_A subtypes also play a role, even if to a much lesser extent. For example, in α1 knock-in mice diazepam still produces α2-, α3-, and/or α5-mediated effects on Rotarod performance (McKernan et al., 2000). Hence, although the very low intrinsic efficacy of TPA023 at the α2 and α3 subtypes did not produce overt sedation in preclinical species, it remains to be seen whether TPA023 produces sedation in humans, especially since a nonsedating anxiolytic profile in preclinical species does not necessarily translate into human (e.g., bretazenil; van Steveninck et al., 1996).

It could be argued that the preclinical profile of TPA023 is not dissimilar from that of nonselective partial agonists such as bretazenil, which in preclinical species is, like TPA023, also a nonsedating anxiolytic with a reduced ethanol interaction that shows little propensity to develop dependence and does not precipitate withdrawal from a full agonist (Haefely et al., 1990; Martin et al., 1995; Griebel et al., 1999). However, TPA023 differs in two important respects from nonselective full and partial BZ agonists such as diazepam and bretazenil: 1) it lacks efficacy at the α1 and α5 subtypes and therefore exerts its in vivo effects through α2- and α3-containing receptors rather than the α1, α2, α3, and α5 receptors modulated by the nonselective BZ agonists, and 2) it is a weak partial agonist at the α2 and α3 subtypes, with an efficacy at the α2 and α3 subtypes of only 0.11 and 0.21% relative to chlordiazepoxide that is therefore considerably lower than that of bretazenil (Atack, 2003). Consequently, any difference between the pharmacological properties of TPA023 and nonselective full agonists may be a result of either loss of α1 and α5 efficacy or partial rather than full agonist efficacy at the α2 and α3 subtypes or a mixture of these two factors.

Taken together, these data all suggest that TPA023 differs from nonselective full agonists in that it is has anxiolytic-like activity in the absence of sedative, ethanol interaction, and dependence liabilities. Indeed, in baboons TPA023 does not generalize to lorazepam and was not self-administered (Ator, 2005). On the other hand, TPA123, a compound which possesses a degree of α1 agonist efficacy, showed modest levels of self-administration and generalization to lorazepam, suggesting that efficacy at the α1 subtype may be associated with an addiction liability (Ator, 2005). Clearly, given the past disappointments with the nonselective BZ site partial agonists failing, for a variety of reasons, to translate into clinically useful drugs (Atack, 2003), the critical issue for subtype selective compounds like TPA023 is whether or not the novel preclinical pharmacological profile translates into a clinical benefit in human. Nevertheless, compounds such as TPA023 highlight the potential for selectively targeting particular GABA_A receptor populations using the strategy of subtype selective efficacy and exploit the increased understanding of BZ pharmacology afforded by molecular genetic and pharmacological approaches (Atack, 2003; Rudolph and Möhler, 2004).