Preclinical studies using genetically modified mice suggest that GABA_A-α₁ (Rudolph et al., 1999; McKernan et al., 2000), α₂ (Low et al., 2000), and α₅ (Collinson et al., 2002; Crestani et al., 2002)-containing receptors mediate the sedative/motor-impairing, anxiolytic, and memory impairing effects of benzodiazepines, respectively. Pharmacologically, various nonbenzodiazepine compounds have been described that show either selective affinity (e.g., zolpidem and indiplon) or selective efficacy (e.g., L-838,417, SL651498, TP003, TPA023) for subtypes of GABA_A receptors (Sanger et al., 1987; McKernan et al., 2000; Griebel et al., 2001; Dias et al., 2005; Atack et al., 2006). Such selectivity profiles have been argued to be the basis for selective behavioral profiles of these compounds in rodents and in some cases man. For example, zolpidem is selective for GABA_A-α₁-containing receptors, has a preferential sedative-hypnotic profile in animals, and is marketed as a sleep aid zolpidem tartrate (Ambien). Although additional compounds with various selectivity profiles have emerged, limited clinical data in patient populations exist to date (but see Basile et al., 2006; de Haas et al., 2007; Nutt et al., 2007).

One major area of interest has been to develop compounds with selectivity for GABA_A-α₂ and/or -α₃ over GABA_A-α₁ receptors, with the prospect of bringing a nonsedative anxiolytic to the market for the treatment of anxiety disorders (Atack, 2003). In this field, the only relevant clinical data, on a well characterized preclinical molecule, that have been forthcoming are on the α_2/3-selective molecule TPA023, which was compared with lorazepam in human volunteers in a double-blind, double-dummy crossover study (de Haas et al., 2007). TPA023 had a reduced liability to induce sedation and impair cognition in human volunteers compared with lorazepam. These clinical observations are consistent with lack of efficacy of this compound at human GABA_A-α₁ and -α₅ receptors as determined by in vitro electrophysiology (Atack et al., 2006). However, no data confirming that TPA023 has anxiolytic efficacy in man at doses not engendering side effects in human volunteers are available; to date, there is no translation of this preclinical profile to a meaningful therapeutic outcome. Thus, at this point, the increased preclinical understanding of the roles of some of the major types of GABA_A receptors has not been tested clinically. For this reason, we would emphasize the need for new molecules with novel selectivity profiles, which are amenable for testing in man, to aid in driving preclinical as well as clinical research.

In the current study, we introduce the novel compound NS11394 (Fig. 1), which binds with subnanomolar potency to human GABA_A receptors but which shows differential efficacy at these receptors, as determined by in vitro electrophysiology. NS11394 has a unique profile compared with other recently described compounds. Specifically, NS11394 shows a selectivity profile in the order of GABA_A-α₅ > α₃ > α₂ > α₁-containing receptors, with notably higher GABA_A-α₃ receptor efficacy compared with the majority of subtype-selective molecules recently described.

Given this unique profile, we have made efforts to determine the efficacy of NS11394 in animal models covering various disease areas, including anxiety and pain, making comparison with relevant reference compounds. In this, the first of two articles, we introduce NS11394 as a novel subtype-selective GABA_A receptor positive allosteric modulator and highlight: i) its in vitro binding and electrophysiology profile at human GABA_A receptors, as well as its selectivity over other targets; ii) its excellent pharmacokinetic profile in rodents; iii) the close correlation observed between pharmacokinetic and pharmacodynamic (receptor occupancy) properties; iv) its in vivo profile in animal models of anxiety-like behavior, in addition to its in vivo side effect profile in comparison to various benzodiazepine site-positive modulators. We chose to compare NS11394 to alprazolam, diazepam, or chlordiazepoxide in the efficacy and side effect models, because it is clear that benzodiazepines differ considerably on various parameters, including potency, half-life, selectivity, and metabolite formation to name a few (Chouinard, 2004; Mirza and Nielsen, 2006). Given these differences, we were keen not to restrict ourselves to comparing NS11394 to a single representative for the entire benzodiazepine class, albeit at the expense of not being exhaustive in profiling all three benzodiazepines in all models. Finally, in the accompanying article (Munro et al., 2008), we focus exclusively on the pain profile of NS11394, with comparative data to a range of GABA_A receptor modulators in rat models of acute, inflammatory, and neuropathic pain, supplemented by supportive in vitro spinal cord electrophysiological data.

Materials and Methods

Animals

Male rats (PVG; Harlan Netherlands B.V., Horst, The Netherlands; and Sprague-Dawley or Wistar; Taconic M&B, Ry, Denmark) and female mice (NMRI; Taconic M&B) were housed and habituated for at least 7 days before experimental procedures in Macrolon III cages (20 × 40 × 18 cm; 2 rats/cage or 8 mice/cage). Food (Altromin GmbH & Co., Lage, Germany) and water were available ad libitum, with the exception of restricted water in the conditioned emotional response procedure described below. Animals were allowed a minimum of 7 days of acclimatization in the animal facility before testing. Animals were in a temperature-controlled environment with a 13-h light/11-h dark cycle (lights on at 6:00 AM and off at 7:00 PM). All of the testing procedures were in accordance with Methods and Welfare Considerations in Behavioral Research with Animals (NIH Publication 02-5083, 2002) and licensed by the Animal Experiments Inspectorate, The Danish Ministry of Justice.

Compounds

NS11394 and zolpidem were synthesized at NeuroSearch A/S, Medicinal Chemistry Department, whereas paroxetine, chlordiazepoxide, Ro 15-4513 (Sigma-Aldrich, Vallensbæk Strand, Denmark), diazepam (Nomeco A/S, Copenhagen, Denmark), alprazolam (Cambrex Corporation, Charles City, IA), and clonazepam (Roche Diagnostics, Basel, Switzerland) were purchased from commercial sources. For all in vivo studies in mice and rats, NS11394 was dissolved in 5% Tween 80/Milli-Q water (Millipore Corporation, Billerica, MA) and administered orally in a dosing volume of 5 ml/kg for rat and 10 ml/kg for mouse. Diazepam, alprazolam, and chlordiazepoxide were dissolved in 5% Cremophor EL (polyethoxylated castor oil; BASF, Ludwigshaden, Germany) and administered intraperitoneally, whereas paroxetine was dissolved in saline and also administered intraperitoneally. All compounds were administered 30 min before behavioral procedures, unless otherwise stated. Doses are expressed as the milligram of salt weight per kilogram of body weight.

Cloning of cDNA and cRNA Preparation

cDNAs for the GABA_A receptor subunits α_1–6, β₂, β₃, and γ_2S were cloned from human hippocampus poly(A⁺) mRNA (Clontech, Mountain View, CA) using PCR. In brief, first-strand cDNA was obtained using oligo(dT) primer and Moloney murine leukemia virus reverse transcriptase (GE Healthcare Life Sciences, Buckinghamshire, UK). Full-length cDNA sequences were amplified by PCR reactions using 100 ng of first-strand cDNA per reaction, Expand HF polymerase (Roche Diagnostics), and gene-specific primer sets (MWG Biotech, High Point, NC). PCR conditions were: 1) 94°C 60 s; 2) 15 times (94°C 60 s, 55°C 60 s, 72°C 120 s); 3) 20 times (94°C 60 s, 55°C 60 s, 72°C 180 s); and 4) 72°C 10 min using a Robocycler (Stratagene, La Jolla, CA). Amplified products were polished with Pfu polymerase (Stratagene), purified on a QIAquick column (QIAGEN, Valencia, CA) and cloned into pCRScript (Stratagene) or pSwas. Several positive clones were sequenced bidirectionally, and verified clones were subcloned into either a pNS1 or pNS3 vector. pSws and pNS1/pNS3 are custom-designed vectors derived from the vectors pZErO-1 or pcDNA3 (Invitrogen, Carlsbad, CA), respectively. For cRNA production plasmids were linearized using a unique downstream polylinker enzyme (NotI, XhoI, or XbaI). cRNA was prepared and capped from the linearized cDNAs using the mMESSAGE mMACHINE T7 Transcription kit (Ambion, Austin, TX). RNA was purified using the RNeasy mini kit (QIAGEN), adjusted to a concentration of 0.5 μg/μl, and stored at -80°C until use. For establishment of stable cells lines, β subunits mutated to have a cation-permeable selectivity segment were used (Jensen et al., 2002).

Cell Culture and Stable Transfections

HEK-293 (ATCC1573) cell lines were propagated in culture flasks (Nalge Nunc International, Rochester, NY) at 37°C in a humidified atmosphere containing 5% CO₂ and 95% air. The growth medium consisted of Dulbecco's modified Eagle's medium (BE12-604/U1; Lonza Walkersville, Inc., Walkersville, MD) supplemented with 10% fetal bovine serum (Invitrogen). HEK-293 cells, seeded in a T12.5 culture flask (Nalge Nunc International) and cultured to 50 to 70% confluence, were transfected with a total of 1 μg of expression plasmids pNS3n(α_1–6), , and pNS3h(γ_2s) (n, z, and h denote NeoR, ZeoR, and HygR genes in the plasmids, respectively) using Lipofectamine Plus (Invitrogen) according to the manufacturer's protocol. Twenty-four hours after transfection, cells were detached using trypsin/EDTA (Invitrogen) and seeded in a T75 culture flask (Nalge Nunc International) with a tear-off lid. For stable expression, cells were selected in medium supplemented with 0.5 mg/ml G-418 (Sigma-Aldrich) and 0.125 mg/ml zeocin (Invitrogen) and 0.15 mg/ml hygromycin (Roche Diagnostics). Single clones were picked and propagated in selection media until sufficient cells for freezing were available; thereafter, the cells were cultured in regular culture media with one or more selection agents.

Preparation of Rat Cerebral Cortical Membranes

Wistar rat cerebral cortices were removed rapidly after decapitation, homogenized for 5 to 10 s in 10 volumes of Tris-HCl buffer (30 mM, pH 7.4), and pelleted by centrifugation at 27,000g for 15 min. All procedures were performed at 0 to 4°C unless otherwise indicated. After washing three times by resuspension in 10 volumes of ice-cold buffer and centrifugation at 27,000g for 10 min, the pellet was resuspended and homogenized in Tris-HCl buffer, incubated on a water bath at 37°C for 30 min, and then pelleted at 27,000g for 10 min. After one more wash, the pellet was resuspended in 10 volumes buffer and stored at -20°C until use. On the day of the experiment, the membrane preparation was thawed, pelleted at 27,000g for 10 min, and washed twice by resuspension in Tris citrate buffer (50 mM, pH 7.1) and centrifugation at 27,000g for 10 min. The final pellet was resuspended in Tris citrate buffer (500 ml/g buffer of original tissue).

Preparation of Cell Line Membranes

Cell culture medium was removed from confluent cell cultures; cells were rinsed once with Dulbecco's phosphate-buffered saline, harvested into a small volume of Dulbecco's phosphate-buffered saline by gently scraping the cells off of the bottom of the culture flask, and pelleted by centrifugation at 1550g for 10 min. All procedures were performed at 0 to 4°C unless otherwise indicated. The cell pellet was gently washed once in 15 ml of Tris-HCl or Tris citrate buffer (50 mM, pH 7.1), resuspended in the same buffer, homogenized using an Ultra-Turrax homogenizer (Rose Scientific Ltd., Edmonton, AB, Canada), and centrifuged at 27,000g for 10 min. The pellet was resuspended in 15 ml of Tris-HCl or Tris citrate buffer and stored at -80°C. On the day of the experiment, the membranes were thawed, pelleted at 27,000g for 10 min, and resuspended in Tris citrate buffer (50 mM, pH 7.1, α₁-, α₂-, α₃-, and α₅-containing cell lines) or KH₂PO₄ buffer (20 mM, pH 7.4, α₄- and α₆-containing cell lines) containing 100 mM KCl.

In Vitro [³H]Flunitrazepam, [³H]Ro 15-1788, and [³H]Ro 15-4513 Binding

Aliquots of 500 μl of rat cortical membranes or cell suspension (30–150 μg of protein per assay) were added to 25 μl of test compound, and 25 μl (1–5 nM, final concentration) of [³H]flunitrazepam (88 Ci/mmol; GE Healthcare), [³H]Ro 15-1788 (87 Ci/mmol; PerkinElmer Life and Analytical Sciences, Boston, MA), or [³H]Ro 15-4513 (28 Ci/mmol; PerkinElmer Life and Analytical Sciences) were mixed and incubated for 40 or 90 min (α₄- and α₆-containing cell lines) at 2°C. Nonspecific binding was determined in the presence of 1 μM clonazepam or 10 μM Ro 15-4513 (α₄ and α₆). Compounds were tested at 5 to 10 concentrations ranging from 0.01 nM to 30 μM. Binding was terminated by rapid filtration over Whatman GF/C glass fiber filters (Whatman, Clifton, NJ), and the amount of radioactivity on the filters was determined by conventional liquid scintillation counting using a Tri-Carb counter (PerkinElmer Life and Analytical Sciences).

In Vivo [³H]Flunitrazepam Binding in Mouse or Rat Brain

Groups of three female NMRI mice (25–28 g) or three male Wistar rats (180 g) were administered either orally with NS11394 or intraperitoneally with alprazolam, chlordiazepoxide, or diazepam. Twenty minutes before decapitation, the animals were injected intravenously (tail vein) with 5.0 μCi of [³H]flunitrazepam in 0.2 ml of saline (mice) or 20 μCi of [³H]flunitrazepam in 0.4 ml of saline (rats). After decapitation, mice forebrains were rapidly excised and homogenized in 12 ml of ice-cold Tris citrate buffer (50 mM, pH 7.1) using the Ultra-Turrax homogenizer. For rats, only half of the forebrain was homogenized in 15 ml of ice-cold buffer. Three aliquots of 1 ml (mice)/six aliquots of 1.5 ml (rats) were filtered through Whatman GF/C glass fiber filters and washed with 2 × 5 ml of ice-cold buffer. Groups of vehicle-treated animals served as control for estimation of total binding. Nonspecific binding was determined by animal groups that were administered 3.0 mg/kg i.p. clonazepam (mice) or 10 mg/kg i.p. clonazepam (rats) 10 min before [³H]flunitrazepam injection. Radioactivity on the filters was determined by conventional liquid scintillation counting using the Tri-Carb counter (PerkinElmer Life and Analytical Sciences).

Bioavailability

Wistar rats were dosed with 3 mg/kg NS11394 either intravenously (n = 3, 1.5 mg/ml in Tween 80, clear solution) or orally (n = 3, 1.5 mg/ml in Tween 80, clear solution). At specified time points after intravenous (0.08, 0.5, 1, 2, 4, 6, and 24 h) or oral (0.5, 1, 2, 3, 4, 6, and 24 h) administration, blood samples were collected in EDTA-K⁺ tubes (Milian USA, Gahanna, OH), mixed, and kept on ice. Plasma was prepared by centrifugation at 1000g for 20 min and stored at -18°C before analysis. Plasma proteins were precipitated by adding 3 volumes of acetonitrile containing 100 ng/ml of an internal standard and centrifugation at 16,000g for 25 min at 5°C, after which the supernatant was transferred to a fresh tube and diluted with 1 volume of water. Samples were subsequently analyzed by liquid chromatography (Alliance 2795; Waters, Milford, MA) in combination with a triple quadrupole mass spectrometer (Micromass, Quattro Ultima; Waters). Detection of NS11394 and the internal standard was performed by selected reaction monitoring in electrospray positive ion mode, fragmenting protonated parent ion to a specific prominent product ion. Quantification of NS11394 was performed using quadratic regression (weighted 1/x). The calibration range was 10 to 5000 ng/ml in plasma, and samples above the upper limit were diluted 10 times with water and reanalyzed.

Pharmacokinetic-Pharmacodynamic Relationship

Animals were administered NS11394, [³H]flunitrazepam, and clonazepam as detailed in the in vivo binding experiments above and decapitated at fixed time points. Blood samples for plasma isolated from NS11394-dosed animals were treated as described in the bioavailability section. Furthermore, from each animal, half of the forebrain was transferred to a small plastic bag, frozen on dry ice, and stored at -20°C for later analysis, whereas the other forebrain half was used for in vivo binding. For analysis of brain concentrations of NS11394, brain tissue (0.1 g) was homogenized with 1-mm zirconia beads in 1000 μl of acetonitrile:water (80:20), containing 100 ng/ml internal standard, using a bead beater (Mini-BeadBeater-96+; Bio-Spec Products, Inc., Bartlesville, OK). The tissue homogenates were centrifuged at 16,000g for 25 min at 5°C, after which the supernatant was transferred to a fresh tube and diluted with 1 volume of water. Samples were analyzed as described in the rat bioavailability section. Quantification of NS11394 was performed using linear regression (unweighted fit).

Microsomal Stability

Microsomes were supplied by BD Biosciences (Franklin Lakes, NJ). In each of four tubes, 2.5-μl microsomes (20 mg/ml protein), 87.5 μl of COMIX (NADPH generating system; BD Biosciences), and 10 μl of test solution (10 μM compound in 20/80 MeCN/water) were mixed. The reaction in two of the tubes was immediately stopped with 20 μl of ice-cold MeCN (5% acetic acid). The supernatant was taken for later analyses by liquid chromatography-mass spectrometry. In the two remaining tubes, the reaction was allowed to continue for 60 min and, at this time point, stopped with 20 μl of ice-cold MeCN (5% acetic acid). Microsomal stability was calculated as the percentage of compound in the t = 60 sample relative to the t = 0 sample according to the formula:

Isolation and Injection of Xenopus laevis Oocytes

Adult female X. laevis (Nasco, Fort Atkinson, WI) were anesthetized with 0.28% tricaine (Sigma-Aldrich) in ice-cold water, and lobes of ovaries were removed surgically, after which the frogs were killed. After removal, the lobes were placed in a glass Petri dish in modified Barth's solution and cut into small pieces with surgical knifes. Lobular pieces were transferred to a 50-ml tube (Nalge Nunc International), and oocytes were dispersed as well as defolicated using 0.2% collagenase (Sigma-Aldrich) in Low-Ca Barth's (modified Barth's without CaCl₂) using gentle agitation. After dispersal for 1 to 2 h, oocytes were washed five times in Low-Ca Barth's followed by five times in modified Barth's and then placed in a glass Petri dish. Stage V and VI oocytes were selected and transferred to a clean glass Petri dish and maintained at 18°C in modified Barth's solution. For injection, the oocytes were placed in a custom-designed chamber in modified Barth's solution and injected with 25 to 50 nl of cRNA mixture using a Pico Pump (WPI, Sarasota, FL). The cRNA mixture contained GABA_A receptor subunits α_x, β₂, and γ_2S in the ratio of 1:1:3 and in a total concentration of 0.5 μg/μl. After injection, oocytes were maintained at 18°C in modified Barth's solution for 1 to 5 days.

Two-Electrode Voltage Clamp

Electrophysiological responses from X. laevis oocytes were measured using the two-electrode voltage-clamp technique. Single oocytes were placed in custom-designed recording chambers that were continuously perfused with >2 ml/min OR2. Recording electrodes were fabricated from borosilicate glass tubings with filament (BF150-110-10; Sutter Instrument Company, Novato, CA) using a DMZ-Universal puller (Zeitz, Augsburg, Germany), backfilled with 2 M KCl, and when submerged into OR2 solution, the electrode resistances were in the range of 0.5 to 1 MΩ. The oocyte was impaled using manual micromanipulators and allowed to equilibrate at a holding potential of -50 to -80 mV for at least 1 min to ensure a maximal leak current of 100 nA before the experiment was initiated. Currents were amplified by a Geneclamp 500B amplifier (Molecular Devices, Sunnyvale, CA), low-pass filtered at 20 Hz, digitized at 200 Hz by a Digidata 1322A (Molecular Devices), and then recorded as well as analyzed by a PC (Compaq Evo; Hewlett Packard Company, Palo Alto, CA) using the pClamp9 suite (Molecular Devices).

Drug solutions were applied through a capillary tube, with an inner diameter of 1.5 mm (Modulohm A/S, Herlev, Denmark), placed approximately 2 mm from the oocyte, and connected through Teflon tubing to a Gilson 233XL autosampler (Gilson, Middleton, WI). Gilson 735 software suite was used to control all of the Gilson equipment (233XL autosampler, 402 diluter, and Minipuls 3 pumps) and to trigger recording by pClamp9. A flow rate of 2.5 ml/min through the capillary tube during applications ensured a rapid exchange of liquid surrounding the oocyte (in the order of a few seconds). The application length was set to last 60 s, which was sufficient to obtain peak currents. The time interval between recordings was 5 min, during which the oocyte was perfused with OR2 through the capillary tube as well.

For each experimental set, GABA was freshly dissolved in OR2 in a concentration known to give rise to EC_5–25 elicited currents for a given GABA_A receptor subtype combination (0.5–5 μM), and this solution was then used for controls as well as a stock solution for dissolving the compounds to test in the experiment. A complete experimental set contained four control traces of GABA, a reference 0.5 μM diazepam trace, 10 GABA control traces, and finally eight test traces of a compound in increasing concentrations. The oocyte was discarded after one experimental set. Modulatory effects of diazepam were calculated by comparing the diazepam trace to the control trace immediately before. Likewise, modulatory effects of the compound in the test traces were obtained by comparing to the control immediately before the test traces. Oocytes with receptor expression levels, giving rise to a control GABA-evoked whole-cell currents either below or above the range of 250 to 2000 nA, were discarded. In addition, if the diazepam potentiation was below 70% or above 400%, the experiment was discarded. To enable comparison of effects of a compound between individual oocytes, all compound potentiations were normalized to the control diazepam potentiation on the same oocyte.

Rat-Conditioned Emotional Response Test

The conditioned emotional response (CER) procedure was the same as that recently described by Mathiasen et al. (2007). In brief, eight standard operant chambers equipped with a 3-house light (center of ceiling), an operant lever (2 cm above the grid floor), and a valve-operated water spout on one wall were used (ENV-008: 32 × 25 × 25 cm; MED Associates, St. Albans, VT). Water-deprived male PVG rats were trained to associate-lever pressing with water reward. During 30-min sessions, rats were initially trained to make 10 lever presses to receive 1 reward (fixed ratio 10). Thereafter, animals were trained to maintain a high level of lever pressing by increasing the requirements culminating in a variable interval (VI) 60-s reward schedule. The house light remained on throughout all of the training sessions. Thereafter, CER training commenced. In the sequence of L1, D1, L2, and D2, each session consisted of four alternating periods, during which the house light was either on (L, 5 min) or off (D, 2.5 min). After the second D period, there was a third 5-min L period, although response rate data in this final period was not considered further and was a means to ensure that animals maintained baseline responding before the next day's session. During the D periods, a scrambled 0.4-mA, 0.1-s scrambled foot shock was applied according to a VI20-s schedule. The VI60-s reward schedule was maintained throughout both L and D periods. Response rates during the L and D periods were used to calculate a suppression ratio (SR) according to the following formula:

A suppression ratio of 0 indicates that the D has evoked conditioned fear and has completely suppressed lever pressing relative to the L. By contrast, an SR of 0.5 would indicate that the response rate during the L and D periods were equivalent, i.e., no fear to the D cue. Drug testing commenced only when a stable SR below 0.05 had been established. Of the eight animals that commenced training, seven reached this level of proficiency and were used in subsequent drug tests. A within-subject design was used in these studies, such that each individual animal was tested with all doses of NS11394 or alprazolam in a pseudo-randomized order. Test sessions were identical to training sessions with the exception that no foot shocks were delivered. Drug testing was conducted on Tuesdays and Fridays, with other days serving as baseline days to determine whether animals maintained normal performance levels between drug test days.

Four-Plate Testing

Four-plate testing (FPT), based on the method of Aron et al. (1971), consisted of a nontransparent Plexiglas box (25 × 18 × 16 cm) with a transparent lid constructed at NeuroSearch A/S (Bellerup, Denmark). The floor of the chamber consisted of four identical rectangular metal plates (11 × 8 cm), which were separated by a 4-mm gap and through which an electric foot shock (0.6 mA; 0.5 s) could be delivered manually using a foot pedal (BioSeb, Vitrolles, France). Female NMRI mice were placed individually in the chamber, and after a 15-s habituation period, the animal was given an electrical foot shock whenever crossing from one plate to the other. The definition of crossing between two metal plates was that the mouse had two paws on one plate and two on an adjacent plate. After the delivery of a foot shock, there was a 3-s pause before another foot shock was delivered manually. The number of punished crossings made by an animal was determined by the experimenter, blind to the treatment, over a 1-min period. An anxiolytic-like effect of drug treatment was inferred from an increase in the number of punished crossings made under drug treatment compared with the number of punished crossings made by control animals.

Marble Burying

Female NMRI mice were placed for 1 h in novel cages (1 mouse per cage, 20 × 30 cm) in which there were 20 glass marbles (15 mm in diameter) situated in four rows of five on top of 5 cm of sawdust. The mean number of glass marbles buried between 10 and 60 min was taken as an index of “anxiety”, i.e., the more marbles buried the more anxious the mouse (Broekkamp et al., 1986). A marble was classified as buried by an experimenter blind to treatment when at least two-thirds was covered by sawdust.

Locomotor Activity in Nonhabituated Mice or Rats

Mice or rats (Sprague-Dawley) were placed individually in transparent cages (30 × 20 × 25 cm) within activity frames (TSE Systems GmbH, Bad Homburg, Germany), and activity levels were determined over 30 min. The activity frames were equipped with 12 (6 × 2) infrared sensors. Locomotor activity was monitored automatically in the chambers and measured as the interruption of two consecutive infrared sensors. Interruptions were detected by a control unit and registered by a computer running ActiMot software (TSE Systems GmbH). All data are presented as mean distance (in meters) traveled ± S.E.M.

Sensorimotor Function in the Rat Assessed Using the Rotarod: Influence of Ethanol

The relative level of motor incoordination/ataxia in rats engendered by either NS11394 or diazepam (0.3–3 mg/kg i.p) in the presence/absence of ethanol (0.8 g/kg) was determined by assessing the ability of uninjured rats (Sprague-Dawley) to maintain balance on an accelerating rotarod (Ugo Basile, Comerio, Italy). The rotarod speed was increased from 3 to 30 rpm over a period of 180 s, so that the maximal time animals could maintain balance on the rotarod was 180 s and the minimal time was 0 s. Initially, rats received two training trials (separated by 3–4 h) on two separate days to acclimatize to the task. The dose of ethanol (0.8 g/kg) chosen for the interaction studies was selected based on prior reference dose response curves indicating that this dose was subthreshold for impairing rat rotarod performance.

Memory Impairment in Mice and Rats

Passive Avoidance. Mice were tested in a step-through passive avoidance task using four two-compartment chambers (MED Associates Inc): one compartment with light (10 × 12 cm) and the other compartment dark (16 × 20 cm) separated by a shutter with a small doorway. The floor of the box consisted of metal grid steel bars. The mice were handled on the day before training. On the training day, mice were introduced individually to the lit compartment and allowed to habituate for 60 s before access to the dark compartment was granted when the central door was raised. When a mouse completely entered the dark compartment (all 4 paws), the latency to enter was recorded, and two mild foot shocks (0.6 mA, 0.5 s, 5 s between foot shocks) were delivered through the grid floor. The animals remained in the dark compartment for 60 s, after which they were removed and returned to their home cage. Twenty-four hours after this training session, the mouse was again introduced to the lit compartment for 60 s, and thereafter, the latency to enter the dark chamber was taken as a measure of memory retention. If a mouse had not crossed to the dark compartment within 180 s, it was assigned this value and removed from the apparatus.

Fear Conditioning in Rats. Fear conditioning was assessed in eight ventilated, sound-attenuated chambers (length × width × height, inner dimensions: 320 × 320 × 320 mm; D-61350, TSE Startle Response; TSE Systems GmbH), each consisting of a nonrestrictive Plexiglas cage (length × width × height: 100 × 60 × 70 mm) with a 9-rod grid floor (rod = 4 mm; distance between center of adjacent rods = 8.9 mm), inside which rats were placed. Two high-linearity speakers situated 6 cm on either side of the Plexiglas cage delivered background noise and tone cues. Sound intensities were measured using a sound level meter (model 2238; Brüel & Kjaer, Nærum, Denmark) in conjunction with a microphone placed within the chamber (model 4188). The movement of animals in the cages was recorded using a strain gauge load cell mounted directly below the animal. The output signal from the load cell was converted to arbitrary units using an analog-to-digital converter using appropriate software (TSE Systems GmbH). A calibration system (TSE Systems GmbH) was used to ensure comparable sensitivity between the eight systems. On the day before conditioning, animals were habituated to the conditioning chambers for 10 min, after which they were returned to their home cages. During this habituation session, rats were exposed to a constant background noise (65 dB, white noise). On the next day, the animals were conditioned to associate a brief tone cue (10 s, 5 kHz, 80 dB) with a single foot shock (0.6 mA, 0.5 s) delivered through the grid floor and which coterminated with the tone. The 10-s tone cue and associated shock were presented ∼3 min after the animals had been placed in the cages. After this single tone-shock pairing animals remained in the chambers for another 2 min. The entire conditioning session lasted ∼5 min with background noise (65 dB, white noise) throughout the session. Twenty-four hours after the conditioning session, rats were re-exposed to the chambers, and all of the details were identical, with the exception that no foot shock was delivered through the grid floor. The movement of the animals in the period before tone presentation (i.e., context fear conditioning) was recorded in 4-ms bins and averaged over ∼3 min. Thereafter, movement was recorded in 4-ms bins during the tone and averaged over the entire 10-s tone presentation (i.e., cue fear conditioning). Data are presented as average movement (arbitrary units) ± S.E.M. in each of these periods. When assessing the effects of vehicle, NS11394, and alprazolam on fear conditioning, an additional vehicle control group was run that was exposed to the chambers and tone cue but did not receive a foot shock. This latter “no-shock” group was a control to ensure that we had robust context and cue fear conditioning on the test day in the “vehicle-shock” group.

Data Analysis

For in vitro binding studies, IC₅₀ values were determined based on the equation where B is the binding in percentage of total specific binding; C is the concentration of test compound; and n is the Hill coefficient. Estimates of binding parameters were calculated with the nonlinear curve-fitting program GraphPad Prism (version 4.03; GraphPad Software, Inc., San Diego, CA). K_i values were calculated from IC₅₀ values using the equation by Cheng and Prusoff (1973). All results are given as mean ± S.E.M. For two-electrode voltage-clamp experiments, compound concentration-response curves were fitted to a sigmoidal dose-response curve using GraphPad Prism. For all behavioral studies, analysis of variance was used to analyze overall effects of treatments. When the F value was significant, this was followed by Dunnett's post hoc text with a significance level of p < 0.05 (SigmaStat 2.03; SPSS Inc., Chicago, IL).

Results

In Vitro Binding

The affinity of NS11394 for the benzodiazepine site of hGABA_A receptors containing the α₁, α₂, α₃, or α₅ subunits together with β₃ and γ_2S stably expressed in HEK-293 cells ranged from 0.1 to 0.8 nM. Diazepam was approximately 10-fold less potent with a range of 5 to 20 nM. Neither NS11394 nor diazepam showed affinity selectivity for any of these four receptor combinations. The affinity of NS11394 and diazepam at a mixed population of rGABA_A receptors in rat cortical membranes and these four cloned hGABA_A receptors expressed in cell lines was approximately the same, indicating no obvious species difference with respect to binding affinity (Table 1). Contrary to the potent binding at α₁-, α₂-, α₃-, and α₅-containing receptors, NS11394 had between 600 and 2000 times lower affinity for α₄- and α₆-containing hGABA_A receptors. Likewise, diazepam was considerably less potent at hGABA_A receptors containing either α₄ or α₆ subunits (Table 1).

NS11394 was also tested at 10 μM in a LeadProfiling-Screen (MDS Pharma Services, Bothell, WA) for binding to a broad range of over 60 different receptors, transporters, and ion channels. In addition to confirming high affinity for rat GABA_A benzodiazepine binding sites labeled by [³H]flunitrazepam, the screen also indicated a high selectivity relative to other targets. The only targets for which greater than 50% inhibition was noted were the human opiate κ (62% inhibition, [³H]diprenorphine), human adenosine A₃ (61% inhibition, [¹²⁵I]4-aminobenzyl-5′-N-methylcarboxamidoadenosine), and the rat sodium channel site 2 (84%, [³H]batrachotoxin). In follow-up studies, it was shown that NS11394 had a K_i of 9.6 and 2.7 μM at adenosine A₃ and opiate κ receptors, respectively. In-house patch-clamp studies showed that NS11394 inhibited Na⁺ channels in rat embryonic dorsal root ganglion cells, with a IC₅₀ of 16 μM (n = 2). No other significant off-target activities were seen (data not shown).

In Vitro Efficacy

NS11394-induced potentiation of GABA-evoked currents at human α₁-, α₂-, α₃-, or α₅-containing receptors expressed in combination with β₂ and γ_2S subunits in X. laevis oocytes can be seen in Fig. 2. A GABA concentration giving rise to approximately 5 to 25% of maximal GABA current (EC_5–25) was chosen as testing concentration for each receptor combination, and all of the datasets are indexed relative to the potentiating effects of 0.5 μM diazepam on the same oocytes. As seen, NS11394 modulated GABA responses at all four receptor combinations, with greatest efficacy at α₃- and α₅-containing receptors at which NS11394 engendered a maximal 52 and 78% potentiation relative to diazepam, respectively. NS11394 had less effect at α₂-containing receptors, with a maximal potentiation of 26% relative to diazepam. By contrast, NS11394 only weakly potentiated the GABA responses at α₁-containing receptors, with a maximal potentiation of 8% relative to diazepam. All data points were obtained from 5 to 25 individual oocytes (74 oocytes used in total). As expected from the binding experiments, no apparent selectivity could be observed with respect to functional potencies (see table in Fig. 2).

In Vivo Binding

NS11394 dose-dependently displaced [³H]flunitrazepam binding to mouse brain benzodiazepine receptors when administered 30 min before culling. The dose of NS11394 estimated to inhibit 50% [³H]flunitrazepam binding to mouse forebrain was 0.38 mg/kg. This study was repeated with mice culled at 120 min after NS11394 administration, with NS11394 inhibiting 50% [³H]flunitrazepam binding at 0.49 mg/kg (Fig. 3A). In studies in rats, NS11394 inhibited 50% [³H]flunitrazepam binding to rat brain benzodiazepine receptors at 1.3 and 0.69 mg/kg p.o., when administered 30 and 120 min before culling, respectively (Fig. 3B). In the rat, an additional time course study showed that NS11394 (3 mg/kg p.o.) inhibition of [³H]flunitrazepam binding to brain benzodiazepine receptors was long-lasting with occupancies of 70, 100, and 97% at 30, 180, and 360 min, respectively (data not shown). To support the in vivo efficacy and side effect studies described below, we also determined the in vivo receptor occupancy in mice treated intraperitoneally with various doses of alprazolam, chlordiazepoxide, and diazepam, the three benzodiazepines used as comparators. From Fig. 3C, it is clear that, at the doses tested, the three benzodiazepines attained receptor exposure on par with that seen with NS11394 in the mouse. Alprazolam, chlordiazepoxide, and diazepam inhibited 50% [³H]flunitrazepam binding to mouse forebrain at 0.17, 10, and 2.2 mg/kg, respectively.

Pharmacokinetic-Pharmacodynamic Relationship in Mouse and Rat

The relationship between plasma concentration, brain concentration, and CNS in vivo receptor occupancy after oral administration of NS11394 was determined in mice and rats. NS11394 was dosed to separate animal groups (0.1–3 mg/kg p.o.) and plasma and brain samples taken at a single time point 30 min after dosing. What is evident from Fig. 4 is that, in both the mouse and rat, there is a strong relationship between plasma and brain concentrations and that both of these parameters correlate with CNS in vivo receptor occupancy, a pharmacodynamic marker. However, at equivalent doses in both species, brain and plasma concentrations are higher in mice compared with rat. From Fig. 4, it is clear that, in both species, the brain/plasma is ∼1 at all doses, indicating rapid equilibration between plasma and brain at the 30-min time point.

Profile of NS11394 in Rodent Models of Anxiety-Like Behavior

To evaluate the potential anxiolytic properties of NS11394, the compound was tested in three in vivo models using either alprazolam, chlordiazepoxide, or diazepam as reference comparators.

Rat CER.Figure 5A shows that NS11394 significantly affected the response rate of rats in the dark anxiety-provoking periods (F_4,24 = 11.5, P < 0.001) of a CER session, with all doses of NS11394 significantly increasing response rate compared with vehicle treatment (P < 0.01). Interestingly, NS11394 also affected response rate in the light periods (F_4,24 = 3.1, P < 0.03); however, this effect is not clearly dose-related, with a significant increase (P < 0.05) compared with control animals at 0.3, 1, and 10 mg/kg NS11394 but with no effect at 3 mg/kg. Figure 5B shows the main effect of NS11394 treatment on the suppression ratio parameter (F_4,24 = 15.5, P < 0.001), with all four doses engendering an increase in suppression ratio compared with vehicle treatment (P < 0.01). For comparative purposes, the effect of the benzodiazepine alprazolam at doses of 0.3 to 3 mg/kg are shown in Fig. 5, C and D. It is clear that alprazolam is also anxiolytic, as expected, with a significant effect at 1 to 3 mg/kg on both dark period response rate (F_3,15 = 6.5, P < 0.01) and suppression ratio (F_3,15 = 7.7, P < 0.01) parameters. However, contrary to NS11394, alprazolam at the higher dose also significantly reduces baseline response rate during the light period (F_3,15 = 5.0, P < 0.03), an effect clearly due to sedation/motor impairment based on visual observation.

Mouse FPT. Treatment with NS11394 significantly and dose-dependently increased the number of punished crossings made by mice (F_4,49 = 10.6, P < 0.001; Fig. 6A). All doses of NS11394 significantly increased the number of punished crossings made by mice compared with vehicle-treated mice (P < 0.01). For comparative purposes, we tested the benzodiazepine chlordiazepoxide, which also significantly increased punished crossings (F_3,33 = 7.3, P < 0.003; Fig. 6B), with mice dosed with 10 to 20 mg/kg differing significantly from vehicle-treated mice (P < 0.01).

Mouse Marble Burying Test. In the marble burying test (MBT), NS11394 treatment dose-dependently reduced the number of marbles that mice buried (F_4,34 = 13.0, P < 0.001; Fig. 6C). All doses significantly affected the burying behavior of mice compared with vehicle-treated animals (P < 0.01). The efficacy of NS11394 at 1 mg/kg p.o. was similar to the efficacy seen with paroxetine 10 mg/kg i.p. run in the same study (Fig. 6C, right-hand bar). In addition to using paroxetine as a positive control, we also assessed the effect of the benzodiazepine diazepam in the mouse marble burying test. Similar to NS11394, diazepam significantly reduced burying behavior (F_4,35 = 9.2, P < 0.001). All doses of diazepam reduced the burying behavior of mice compared with vehicle-treated control mice (P < 0.01; Fig. 6D), despite animals dosed with 3 mg/kg diazepam clearly showing signs of sedation.

Side Effect Profile of NS11394 Relative to Benzodiazepines in Rodents

Given that classical benzodiazepines are known to have a range of side effects coupled to their mechanism of action, NS11394 was evaluated in several in vivo models known to reveal different side effects and compared with either diazepam, chlordiazepoxide, or alprazolam.

Mouse and Rat Motor Activity. NS11394 had no significant effect on mouse motility (F_3,26 = 1.5, not significant) over a 30-min period up to a dose of 100 mg/kg p.o., although at the lowest dose tested, there was a tendency toward an increase in motility relative to the control group (Fig. 7A). By contrast, diazepam dose-dependently and significantly reduced motility (F_3,26 = 27.4, P < 0.001). Post hoc Dunnett's test showed that diazepam at all doses significantly (P < 0.01) reduced motility in mice compared with vehicle treatment (Fig. 7B).

In an equivalent study in rats, NS11394 was initially tested at doses of 3 to 30 mg/kg p.o., with no significant effect on motility in novel cages (data not shown). Therefore, NS11394 was retested at doses of 30 to 120 mg/kg p.o. In this case there was a significant effect of NS11394 (F_3,27 = 3.5, P < 0.03), with both 60 and 120 mg/kg p.o. engendering a significant (P < 0.01) reduction in motility compared with control animals (Fig. 7C). As seen in the mouse study, diazepam significantly and dose-dependently reduced motility in rats (F_3,27 = 25, P < 0.001), with all of the tested doses reaching significance compared to vehicle-treated animals (P < 0.01; Fig. 7D).

Rat Rotarod—Ataxia and Ethanol Interaction. Separate groups of rats were pretreated with either vehicle or a subthreshold dose of ethanol (i.e., a dose that does not in itself induce ataxia), in combination with various doses of NS11394 30 min before the rotarod test (Fig. 8A). Analysis of the data showed a significant main overall effect of NS11394 treatment (F_5,84 = 6.1, P < 0.01). Specifically, 120 mg/kg NS11394 significantly reduced time on the rotarod compared with control animals (P < 0.05). However, analysis showed no significant NS11394 × ethanol interaction (F_5,84 = 1.0, not significant), indicating no propensity for synergy between NS11394 and ethanol, with respect to rotarod performance, although from Fig. 8A, it would appear that some interaction was apparent at the 120 mg/kg dose.

Diazepam dose-dependently impaired rotarod performance in the absence of ethanol (F_3,56 = 48, P < 0.001), with 3 mg/kg diazepam engendering a significant impairment compared with the control group (P < 0.001; Fig. 8B). Furthermore, there was a significant diazepam × ethanol interaction (F_3,56 = 6.1, P < 0.001), indicating that the effects of diazepam were altered in the presence of ethanol. Further analysis indicated that there was a significant interaction between diazepam and ethanol at all three doses of diazepam tested (P < 0.001; Fig. 8B).

Mouse Passive Avoidance Test. Mice were trained to associate an environmental context with an aversive foot shock. Memory for the strength of this association was demonstrated in a test 24 h later where animals typically show an increased latency to enter the same context in which they previously received the foot shock. In separate groups of mice administered different doses of NS11394 before the training session, there was a significant dose-dependent impairment of 24-h recall performance (F_5,71 = 7.3, P < 0.001), indicated by the reduction in latency to cross to the dark compartment compared with control animals (Fig. 9A). At doses from 10 mg/kg, NS11394 significantly impaired memory compared with vehicle-treated animals (P < 0.01). There was also a significant impairment of 24-h recall in mice treated with chlordiazepoxide (F_3,44 = 10, P < 0.001), with post hoc tests indicating a significant reduction in crossover latency in mice treated with 20 mg/kg chlordiazepoxide compared with vehicle-treated mice (Fig. 9B).

Rat Fear Conditioning Test. In the rat fear conditioning (FC) test, animals were trained to associate a context and a short auditory tone cue with an aversive foot shock. Animals treated with vehicle and exposed to a context and a tone cue associated with foot shock (vehicle shock group) froze more when exposed to the same conditions 24 h later in the absence of foot shock, compared with vehicle-treated animals exposed to context and tone cue but never shock (vehicle no-shock group). In the studies described here, the differences between the vehicle no-shock and vehicle shock groups were as follows: i) context fear conditioning, 46.7 ± 3.1 and 34.8 ± 3.3, respectively (P < 0.01); and ii) cue fear conditioning, 40.7 ± 4.2 and 18.5 ± 1.8, respectively (P < 0.001). In the study of NS11394 and alprazolam described below, drug effects are shown relative to the vehicle shock group only (Figs. 9, C–F).

Separate groups of rats were administered various doses of NS11394 or alprazolam, in addition to a vehicle-treated control group, 30 min before training for context-shock and cueshock conditioning. When tested 24 h later, there was a near significant effect of treatment on context fear conditioning (F_6,57 = 2.1, P < 0.06; Fig. 9, C and D) for both compounds and a clear significant effect of treatment on cue fear conditioning (F_6,57 = 3.7, P < 0.003; Fig. 9, E and F). Further analysis indicated that the 3 mg/kg dose of both NS11394 and alprazolam significantly impaired cue fear conditioning compared with the vehicle control group. Nonetheless, from Fig. 9, C and D, it is evident that both NS11394 and alprazolam also showed a dose-dependent trend to impair context fear conditioning.

Pharmacokinetic Profile of NS11394

Microsomal Stability. When NS11394 was incubated with liver microsomes derived from mouse or rat at a concentration of 1 μM for 1 h, approximately 90% of parent compound was still present in the media after this time. These data indicate that NS11394 is relatively stable with respect to first pass metabolism and support the bioavailability data described below.

Bioavailability in the Rat. Intravenous administration of 3 mg/kg NS11394 indicated that it had a low-moderate clearance rate of 0.14 l/h/kg and an elimination half-life of 2.8 h. After oral administration, 3 mg/kg NS11394 was well absorbed with a bioavailability of 82% and a prolonged mean plasma concentration level of ∼1200 ng/ml between 30 min and 6 h (Fig. 10), giving an area under the curve_{0–24 h} of 17,160 h · ng/ml and a half-life of >6 h. Because T_max was reached at 30 min, it is unlikely that the prolonged exposure was due to slow absorption. The volume of distribution of 0.8 l/kg is similar to body water (∼0.71 l/kg), indicating fast equilibration between plasma and other body organs. A summary of these data is given in the table accompanying Fig. 10.

Summary

To give an overall visual snapshot of the therapeutic index for NS11394 and to aid in orientating the discussion, we show the separation between plasma concentrations at minimal effective doses of NS11394 engendering efficacy in rodent anxiety models relative to plasma concentrations attained at doses inducing side effects (Fig. 11). At high doses assessed in side effect models, for which measured plasma concentrations of NS11394 were not determined, we estimated the concentrations based on linear regression analysis of mouse and rat data in Fig. 4 (r² = 0.99, P < 0.0005). Figure 11 also depicts the in vitro K_i from binding studies (Table 1) and the functional potency (EC₅₀) of NS11394 at GABA_A receptors from in vitro electrophysiology studies (Fig. 2).

Discussion

NS11394 represents a novel subtype-selective GABA_A receptor positive allosteric modulator, with a unique profile compared with other GABA_A receptor molecules published to date. In particular, NS11394 has substantially greater hGABA_A-α₃ efficacy compared with several recently described subtype-selective molecules (McKernan et al., 2000; Dias et al., 2005; Atack et al., 2006) while maintaining low hGABA_A-α₁ efficacy (Griebel et al., 2001). Similar to recently described subtype-selective GABA_A modulators, NS11394 shows no affinity selectivity among hGABA_A-α-₁, α₂-, α₃-, α₅-containing receptors, although NS11394 has lower affinity for hGABA_A-α₄-or α₆-containing receptors. Furthermore, the affinity of NS11394 for rat (cortex) and hGABA_A-α₁-, α₂-, α₃-, α₅-containing receptors is equivalent, an important consideration in cross-species translation. Functionally, the selectivity profile of NS11394 is dependent upon the α subunit present in the hGABA_A receptor complex, with an efficacy order of α₅ >α₃ >α₂ >α₁. However, functional efficacy data of this sort are somewhat tricky to obtain because the modulatory effects are highly dependent on the GABA concentration used. Within a given receptor combination, the GABA EC₅₀ can vary significantly between individual cells, giving rise to large variation in modulator efficacy. Comparing data for different receptor combinations, this problem is exacerbated because it is difficult to ensure that exactly the same EC_x concentration of GABA is used for each combination. Therefore, we chose to index the modulatory actions of NS11394 to diazepam. Diazepam is described as being equally effective at the four GABA_A receptor subtypes at which we tested (Sieghart, 1995), although this may be an approximation as we consistently see modulation of GABA responses in the order of α₃ ≥ α₂ >α₁ >α₅. Thus, NS11394 has the order of efficacies seen in Fig. 2 relative to diazepam, but relative to GABA, the order would be somewhat different; in particular, the α₅ efficacy would in effect be lower.

The relationship between the pharmacokinetic (plasma and brain concentrations) and pharmacodynamic (CNS receptor occupancy) properties of NS11394 was an important consideration in driving our in vivo studies. There was a linear increase in plasma and brain concentrations, with increasing NS11394 dose and a brain/plasma ratio of ∼1, indicating that NS11394 readily crossed the blood-brain barrier. Once in the brain, NS11394 readily occupied benzodiazepine receptors, and there was a dose-linear increase in receptor occupancy paralleling the increases in plasma and brain concentrations. In rat, NS11394 had a bioavailability of ∼80% and long plasma exposure not dependent on absorption, consistent with long receptor exposure seen for up to 6 h. In vitro stability data were excellent because NS11394 showed >90% stability when incubated for 1 h with rat/mouse microsomes, indicating limited first-pass metabolism and supporting the bioavailability data. In summary, NS11394 has outstanding pharmacokinetic and pharmacodynamic properties, making it a highly useful compound for in vitro and in vivo studies since data generated can confidently be attributed to the parent compound.

In the rat CER test, NS11394 induced a robust anxiolytic-like effect, which is unlikely to be secondary to any analgesic effect, because i) no foot shock is delivered on drug test days, and further, ii) in the companion article, we demonstrate no effect of NS11394 in the tail-flick or hot-plate models of acute nociceptive pain (Munro et al., 2008). Although it could be argued that NS11394 had greater efficacy in CER compared with alprazolam, it is clear that efficacy of alprazolam was limited by motoric impairment at higher doses. However, NS11394 generally increased baseline response rate, a disinhibitory effect also described with low doses of benzodiazepines (Mathiasen et al., 2007). NS11394 was also effective in two mouse models of anxiety. Overall, the minimal significantly effective dose of NS11394 in rodent anxiety models was 0.1 to 0.3 mg/kg, equating to receptor occupancy in mouse/rat of ∼20% (Fig. 11). By contrast, NS11394 only reduced locomotor activity at 60 to 120 mg in rat, with no effect in mice up to 100 mg/kg, marginally affected rat rotarod performance at 120 mg/kg, and did not significantly interact with ethanol in rat up to 120 mg/kg (Fig. 11). Thus, NS11394 has a benign side effect profile, even at doses suprathreshold for inducing full receptor occupancy in forebrain. By contrast, diazepam significantly reduced locomotor activity in mouse and rat, impaired rat rotarod performance, and engendered considerable ethanol interaction at doses (0.3–3 mg/kg) that could readily be ascribed to occupancy at GABA_A receptors (Mirza and Nielsen, 2006). The doses of diazepam engendering side effects clearly overlap with doses effective in animal models of anxiety (e.g., Griebel et al., 1999a; Mathiasen et al., 2007). The basis for the effect of NS11394 on rat motility at doses suprathreshold for inducing full receptor occupancy are unclear but might feasibly be related to NS11394 acting at GABA_A-α₄-or α₆-containing receptors at high doses, because activation of these receptor subtypes induces motoric impairment (Ebert et al., 2006). However, we are cautious about this interpretation, as we tested NS11394 at GABA_A-α₄ or α₆ containing receptors combined with γ₂, whereas current literature suggests that α₄ and notably α₆ subunits combine preferentially with a δ subunit (Ebert et al., 2006). The effect of NS11394 on rat motility at doses suprathreshold for inducing full receptor might also be the result of off-target effects, although these were few, and follow-up on specific targets indicated weak potency compared with GABA_A receptors.

Although the therapeutic index for NS11394 is exceptional with respects to sedation, ataxia, and ethanol interaction, a different pattern emerges when cognition is considered. NS11394 impaired memory for aversive events in both the mouse (PA) and rat (FC). High efficacy of NS11394 at GABA_A-α5 receptors possibly explains these effects given that positive modulation of GABA_A-α5 receptor is an important determinant of benzodiazepine-induced memory impairment (Maubach, 2003). However, what is surprising is the apparent better therapeutic index for NS11394 compared with alprazolam and chlordiazepoxide. When comparing separation between anxiolytic potency and potency to impair cognition for chlordiazepoxide (FPT versus PA) and alprazolam (CER versus FC), this is ∼2 to 3-fold. For NS11394, this separation is 100-fold in mouse (MB versus PA) and 10-fold in rat (CER versus FC) (Fig. 11). Although, overall, the therapeutic index of NS11394 is less impressive with respect to memory impairment, it is still noteworthy that, whereas anxiolytic efficacy is seen at ∼20% brain occupancy in both mouse and rat, memory impairments are apparent at much higher brain receptor occupancy (Fig. 11).

Based on our data, what do we believe is responsible for the excellent in vivo anxiolytic efficacy and low side effect profile of NS11394? With respect to in vivo anxiolytic efficacy, there are several possibilities: i) substantial efficacy at GABA_A-α₃ receptors; ii) sufficient efficacy at GABA_A-α₂ receptors; and iii) a combination of efficacy at α₂ and α₃ receptors. Although compounds with efficacy at GABA_A-α₁ receptors are often very efficient in anxiety models such as CER (Mathiasen et al., 2007), knowledge from our in-house program indicates that compounds with no efficacy at GABA_A-α₁ receptors can still demonstrate efficacy in CER as long as they demonstrate sufficient efficacy at GABA_A-α₃ receptors (>30% using our electrophysiology methods). Furthermore, in our hands, nonselective weak partial agonists with efficacy of ∼15% at all receptor subtypes are rarely active in CER (data not shown). Thus, we believe it unlikely that the low efficacy of NS11394 at GABA_A-α₁ receptors contributes significantly to its anxiolytic effects. Regarding the low side effect profile of NS11394 compared with benzodiazepines, two factors probably work in synergy. First, NS11394 is a partial agonist at all subtypes, a profile known to impart an improved side effect profile (Haefely et al., 1990), and second, the efficacy at GABA_A-α₁ receptors is low.

Having discussed what we believe explains the excellent efficacy versus side effect profile of NS11394, it is fair to say that the in vitro electrophysiology profile for NS11394 is somewhat inconsistent with conclusions derived from work with gene knock-in mice (Rudolph et al., 1999), implicating GABA_A-α₂ receptors as most important in the anxiolytic effect of benzodiazepines. However, pharmacological studies with GABA_A-α₁ selective antagonists are also inconsistent with gene knock-in mouse data in implicating GABA_A-α₁ receptors in anxiolytic efficacy (Shannon et al., 1984; Griebel et al., 1999b; Huang et al., 1999; Paronis et al., 2001; Rowlett et al., 2005). Indeed, ocinaplon, an arguably GABA_A-α₁ selective molecule, demonstrates anxiolytic efficacy in GAD patients without engendering sedation (Basile et al., 2006). Furthermore, based on data with weak positive and negative GABA_A-α₃ subtype-selective modulators, others like us advocate that this receptor subtype alone is sufficient to engender an anxiolytic response in rodents (Atack et al., 2005; Dias et al., 2005). Thus these discrepancies and our data with NS11394 suggest that results generated with gene knock-in animals should perhaps not be considered pivotal but rather as parallel datasets.

When comparing NS11394 to other subtype-selective ligands described in the literature, it is important to appreciate that functional selectivity is a rather loose term. For example, as stated earlier in our hands, diazepam is a somewhat more effective modulator at GABA_A-α_2/3 over GABA_A-α_1/5 receptors. In addition, adipiplon/NG2-73, described as a α₃-preferring GABA_A partial agonist (www.neurogen.com) using our methods, is a nonselective partial agonist (∼40% at α₁- and α₃-containing receptors, unpublished data). Another point to consider is that many subtype-selective compounds described may not necessarily be good in vivo tools. For example, whereas anxiolytic efficacy for L-838,417 has been described in the mouse (McKernan et al., 2000; van Bogaert et al., 2006), its pharmacokinetic profile in this species is poor after peroral administration (Scott-Stevens et al., 2005), and we see only 10% of parent compound remaining in our rodent microsomal stability assays (data not shown), making it reasonable to query conclusions from in vivo studies with this molecule.

In conclusion, NS11394 represents a potent and novel subtype-selective GABA_A receptor modulator with a unique selectivity profile compared with other recently described molecules. We believe that this profile largely explains the potency of NS11394 in animal models of anxiety and excellent therapeutic index. NS11394 also shows excellent pharmacokinetic properties that correlate with pharmacodynamic endpoints, allowing for a powerful translational approach. Given the confidence that we have in attributing in vivo effects to NS11394 per se, this makes it an invaluable tool in posing questions regarding the relevant GABA_A subtype selectivity profile necessary for efficacy in emerging therapeutic areas such as pain (see Munro et al., 2008; Knabl et al., 2008).