Introduction

Cognitive impairments associated with schizophrenia are among the core symptoms of the disease (Sharma and Antonova 2003); they correlate highly with functional outcome (Fett et al. 2011) and are not well treated with current antipsychotic therapies (Harvey and Keefe 2001; Keefe et al. 2007). Further understanding of the pathological processes leading to cognitive disruption in schizophrenia is a necessity if better treatments are to be made available to treat this important unmet clinical need.

γ-Aminobutyric acid (GABA)-ergic inhibitory neurotransmission in the brain controls the temporal fidelity of excitatory transmission and engenders oscillatory activities, certain types of which have been shown to be disturbed in schizophrenia (Uhlhaas et al. 2008). Specifically, disruption of gamma-band oscillations is suggested to be responsible, to some extent, for the cognitive difficulties experienced by schizophrenic patients (Cho et al. 2006). GABA exerts its modulatory effects, in part, via an action at GABAA receptors, which are heteropentameric membrane proteins that form a GABA-gated chloride channel. These receptors contain an α-subunit (α1, α2, α3 or α5), a β-subunit (mainly β2 or β3) and, in nearly all cases, a γ2 subunit in a 2:2:1 stoichiometry (McKernan and Whiting 1996). GABAA receptors are found synaptically as well as extrasynaptically. Synaptic receptors are believed to mediate fast phasic inhibition (Sieghart and Sperk 2002), whereas receptors located extrasynaptically are thought to mediate background tonic inhibition (Belelli et al. 2009).

Both genetic and medicinal chemistry approaches have been used to identify the potential physiological relevance of the various GABAA receptor subtypes (Nutt 2006). Interestingly, there is a growing body of evidence suggesting that the GABAA α5 receptor subtype may be a promising target for the modulation of learning and memory processing. GABAA α5-containing receptors are located at the base of dendritic spines of pyramidal cells in the hippocampus and are therefore strategically positioned to modulate excitatory glutamatergic input that arrives at the spines of these cells (Nutt 2006). Deletion of the receptor in mice has been shown to increase cognitive performance (Collinson et al. 2002) and to provoke spontaneous hippocampal gamma oscillations (Glykys et al. 2008), an effect mimicked in wild-type mice by dampening α5 receptor function with a selective negative modulator. In addition, functional disruption of the receptor via point mutation increased both temporal and spatial memory abilities in mice (Möhler 2009), although such a manipulation adversely affected the expression of murine latent inhibition (Gerdjikov et al. 2008) and pre-pulse inhibition (Hauser et al. 2005). Moreover, negative modulation of GABAA α5 receptors with selective inverse agonists has been shown to enhance learning and memory processing across several cognitive domains affected in schizophrenia using mice (Dawson et al. 2006), rats (Ballard et al. 2009), non-human primates (Ballard et al. 2009) and humans (Nutt et al. 2007). However, as all published preclinical studies, to our knowledge, have evaluated the potential pro-cognitive effects of GABAA α5-ergic drugs in normal animals or in acute deficit paradigms, e.g. following scopolamine disruption, the question remains as to whether possible pro-cognitive effects tested under such conditions may be extrapolated to schizophrenic patients. Preliminary clinical studies with MK-0777, a functionally selective GABAA α2 partial agonist, showed promising pro-cognitive activity in a small schizophrenic patient population (Lewis et al. 2008), although this initial optimism was not reflected in the data generated in a larger randomised trial where the compound was without effect (Buchanan et al. 2011). To date, as far as the authors are aware, the potential benefit of GABAA α5 negative modulation on cognitive impairments associated with schizophrenia has not been investigated in the clinic.

Enormous efforts have been made, not least by employing various pharmacological interventions, to mimic some of the clinically observed symptomatology of schizophrenia using rodents (Neill et al. 2010; Wilson and Terry 2010; O’Tuathaigh and Waddington 2010). Two such strategies, the sub-chronic or early neonatal administration of phencyclidine (PCP), a N-methyl-d-aspartate receptor antagonist, induce a behavioural and neurobiological syndrome in rodents (upon cessation of treatment or in adulthood, respectively) that bears a remarkable similarity to some of the core symptoms observed in schizophrenic patients, including cognitive disruption (Jentsch and Roth 1999). Specifically, such treatment regimes in rodents have been repeatedly shown to disrupt preference for exploration of a novel object in the rat novel object recognition (NOR) task, a paradigm thought to be a measure of episodic recognition memory (McLean et al. 2010; Snigdha et al. 2010; Idris et al. 2010; Damgaard et al. 2010), as well as to disrupt performance on the extradimensional discrimination phases of an attentional set-shifting (ID/ED) task in rats, an assay thought to be a measure of fronto-cortical executive function (Goetghebeur and Dias 2009; Goetghebeur et al. 2010; Rodefer et al. 2005, 2008).

Therefore, given the proposed role of GABAA α5 receptors in cognitive functioning, together with the fact that modulators of these receptors have not yet been tested in “schizophrenia-relevant” cognitive assay/model systems, we decided to investigate the effects of RO4938581, a negative modulator of GABAA α5 receptor function (Ballard et al. 2009; Knust et al. 2009), in the NOR task following sub-chronic PCP (sub-PCP) administration and the ID/ED attentional set-shifting task following early postnatal PCP (neo-PCP) administration. RO4938581 has been reported to display a K i of 4.6 nM for GABAA α5β3γ2 receptors in [3H]flunitrazepam displacement binding and a pIC50 of ~8 with maximum efficacy of ~46% in whole-cell patch recordings (Ballard et al. 2009; Knust et al. 2009). RO4938581 displayed between 17- and 40-fold selectivity towards the α5-containing GABAA receptor compared with the α1-, α2- and α3-containing receptors in a [3H]flunitazepam displacement binding study (Knust et al. 2009), and no inhibition of the GABA current has been observed for α1-, α2- and α3-containing receptors at 3 μM (Ballard et al. 2009). Complementary immunohistochemical studies were performed to quantify parvalbumin-positive interneuron expression in the prefrontal cortex of sub-PCP- and neo-PCP-treated rats as a surrogate measure of animal model validity. A GABAA α5 receptor positive modulator (QH-II-066; Huang et al. 1996, 2000; Skolnick et al. 1997), was also tested in non-PCP-treated animals to assess the cognitive disrupting potential of increasing α5 receptor activity in the NOR task. Additional studies were performed with QH-II-066 and RO4938581, in an amphetamine-induced hyperactivity paradigm, to assess the effect of positive or negative modulation of GABAA α5 receptors on dopaminergic neurotransmission. As an internal control, top-line functional in vitro studies, using a fluorescent-based membrane potential assay and electrophysiological recordings in oocytes, were performed to confirm the published negative modulatory properties and selectivity of RO4938581 in our hands.

Materials and methods

In vitro experiments

FLIPR® membrane potential assay

Chinese hamster ovary cells (purchased from ChanTest, USA) expressing the α5β2γ2 (used as a measure of target engagement) and α2β2γ2 (used as a top-line surrogate measure of the previously published selectivity profile of RO4938581; Ballard et al. 2009; Knust et al. 2009) were used to detect changes in voltage across the cell membrane induced by GABA without or with the addition of allosteric modulators to the GABAA receptor. Previous studies have suggested that the pharmacology of the benzodiazepine site of the GABAA receptor is more dependent on the type of α-subunit (for a review, see Nutt 2006) or γ-subunit (McKernan et al. 1995) present and is not significantly influenced by the type of β-subunit (Hadingham et al. 1993) present. The assay was run in accordance with instructions from the manufacturer (Molecular Devices Corporation, Product R8126) using a FDSS7000 fluorescence reader (furo3, Ex480:Em540 and 96-well setting). Briefly, frozen stock of cells were plated 48 h prior to experiment at a density of 30,000 cells/well. On the day of experiment, the medium was replaced with 80 μl loading buffer (HBBS (+CaCl2 + MgCl2) buffer containing 20 mM HEPES and membrane potential RED dye) and incubated for 30 min at 37°C followed by 30-min incubation at room temperature. FDSS recordings were initiated with 20-s background measurements followed by the addition of 20 μl modulator. At the 3-min time point, 100 μl of GABA in chloride-free buffer [HEPES (20 mM, pH 7.4), Na-gluconate (140 mM), K-gluconate (2.5 mM), Ca-gluconate (1 mM), MgSO4⋅7H2O (2 mM), D(+)glucose (5 mM)] was added and the stimulation by GABA calculated as relative light units by subtracting basal signal (prior to GABA application) from the peak activation of GABA (Max − Min determined within the time interval 170–290 min).

Oocyte electrophysiology

Xenopus laevis oocytes were isolated and treated with collagenase for 3–4 h before incubation for 24 h in Modified Barth's saline buffer [88 mM NaCl, 1 mM KCl, 15 mM HEPES, 2.4 mM NaHCO3, 0.41 mM CaCl2, 0.82 mM MgSO4, 0.3 mM Ca(NO3)2], supplemented with 2 mM sodium pyruvate, 0.1 U/l penicillin and 0.1 μg/l streptomycin. Stage IV oocytes were identified and injected with 4–42 nl of nuclease-free water containing 2–50 ng of cRNA encoding a mixture (1:1:1) of human GABA-α1, β3, γ2s (used as a top-line surrogate measure of the previously published selectivity profile of RO4938581; Ballard et al. 2009; Knust et al. 2009) or GABA-α5, β3, γ2s (used as a measure of target engagement). As outlined above, previous studies have suggested that the pharmacology of the benzodiazepine site of the GABAA receptor is more dependent on the type of the α-subunit (for a review, see Nutt 2006) or γ-subunit (McKernan et al. 1995) present and is not significantly influenced by the type of β-subunit (Hadingham et al. 1993) present. The oocytes were incubated at 18°C and used for electrophysiological recordings following 1–10 days after injection. Electrophysiological recordings were performed by two-electrode voltage clamp techniques at room temperature in a 2-ml bath chamber perfused with ringer solution (115 mM NaCl, 2.5 mM KCl, 10 mM HEPES, 1.8 mM CaCl2, 0.1 mM MgCl2, pH 7.5) at a flow rate of 4–6 ml/min. Oocytes were impaled with agar-plugged 0.5–1 MΩ electrodes, containing 3 M KCl, and voltage-clamped at −90 mV with an Oocyte Clamp Amplifier (0C-725-C) from Warner Instruments. GABA channels expressed in oocytes were activated by applying 3 μM GABA for 1 min at 5-min intervals. Tests of positive and negative modulators were performed by a pre-application of the test drugs for at least 4 min, followed by the co-application of test drug and GABA. GABA was dissolved in ringer or water and drugs were dissolved in DMSO at a stock solution of 10 mM. Further dilutions were performed in ringer to obtain the test concentration.

Off-target profiling

The interaction of RO4938581 with 100 binding sites, including all major classes of neurotransmitter receptor, uptake systems, ion channels and enzymes, was examined (CEREP, Celle L'Evescault, France). RO4938581 was tested in all assays at 10 μM. Each determination was made in duplicate.

Plasma and brain free fraction determinations

Free fractions in rat plasma (fu,plasma) and brain (fu,brain) were determined for RO4938581 and QH-II-066 using equilibrium dialysis in 96-well formats. The methodology was modified from (Kalvass and Maurer 2002). Freshly isolated rat brains were homogenized twice in two volumes of phosphate buffer (pH 7.4). Dialysis membranes (cutoff, 12–14 kDa; HTDialysis) were soaked in 80:20 PBS/ethanol and rinsed in deionized water before use. Brain homogenate or plasma was added to the donor side and spiked with a test compound to a final concentration of 1 μM. The receiver side contained phosphate buffer (pH 7.4). Equilibrium dialysis was performed by incubating at 37°C for 5 h. Subsequently, the compound concentration was measured in the buffer phase using liquid chromatography–mass spectrometry.

Ex vivo experiments

Immunohistochemical quantification of parvalbumin-positive cells in the prefrontal cortex

Lister hooded rats from the sub-PCP and neo-PCP behavioural studies were killed 1 day after the behavioural studies were finalised. Animals were perfusion-fixed and brains were removed, immersion-fixed for 4 h in 4% PFA and then left in 30% sucrose for 48 h. Free-floating coronal sections (40 μm) were cut from vehicle- and PCP-treated rats. Every section was collected until 1.6 mm rostral to bregma (Paxinos and Watson, 4th edition). The sections were then immunostained with a parvalbumin antibody overnight (1:2,000, Sigma #P3088). Secondary antibody (biotinylated-IgG, Dako #0464) was applied for 1 h followed by ABC reaction. Finally, sections were visualized with diaminobenzidine.

Manual counting (2D) of parvalbumin-positive cells was performed blinded by a trained observer. For sampling at 3.7 mm rostral to bregma (Paxinos and Watson, 4th edition), sections were spaced by 240 μm to cover a total area of 880 μm in the caudal direction. The anatomical areas counted were limited to the prelimbic cortex and cingulate cortex, area 1. A ×10 objective and fixed lighting were used on all sections.

For each animal, both the left and right sides in four sections were analysed (eight areas total). In the sub-chronic vehicle-treated (n = 10) and early postnatal vehicle-treated (n = 10) groups, an average of 1,106 and 1,210 cells were counted per animal, respectively. However, of the total 80 areas counted in ten animals, 16 areas in the sub-chronic vehicle-treated and two areas in the early postnatal vehicle-treated groups were of poor quality and had to be omitted from the analysis. For this reason, the analysis was based on the average number of parvalbumin-positive cells in each hemisphere per section [173, SD = 19.3 (64 areas); 155, SD = 36 (78 areas)] in sub-chronic vehicle-treated and early postnatal vehicle-treated rats, respectively. These numbers were converted to 100% for the respective groups.

In vivo experiments

Animals and accommodation

For NOR experiments, 96 male Lister hooded rats (48 per experiment, 220–240 g at the time of testing; Charles River, Germany) were housed four per cage, with food and water available ad libitum (except during habituation, acquisition and retention trials).

For ID/ED experiments, 32 male Lister hooded rats (31 used in experiment, 1 rat used as cage mate, 250–300 g at the time of testing; dams obtained from Charles River) were housed two per cage. During training and test, animals were given access to water ad libitum, but were food restricted (11 g food per animal per day). During weekends, animals had access to food ad libitum until 12 h before the next testing day.

For amphetamine-induced hyperactivity, 80 male Wistar rats (40 per experiment, 180–240 g at the time of testing; Charles River) were housed four per cage. Wistar rats were chosen on the basis of historical in-house data showing an optimal window for the potential attenuation of amphetamine-induced hyperactivity using this strain.

For drug exposure evaluations, satellite animals (18 male Lister hooded rats, 220–240 g; Charles River) were housed in pairs.

All animals were housed under controlled conditions (12 h of light starting at 0600 hours, 20 ± 2°C, 30–70% humidity) in Macrolon (type III) cages, with standard sawdust bedding and environmental enrichment (plastic house and wooden chew blocks), and food and water available ad libitum (except when otherwise stated). All animal procedures were carried out in compliance with the European Commission Directive 86/609/EEC and with Danish law regulating experiments on animals. Every effort was made to incorporate the three R’s guideline principles (i.e. replace, reduce and refine) for the use of animals in research throughout the experiments described herein.

Drugs and treatment

PCP (1-(1-phenylcyclohexyl)-piperidine hydrochloride, synthesised in-house) was dissolved in 0.9% sodium chloride to obtain a dose of 5 mg/kg (sub-chronic administration paradigm; see “Sub-chronic PCP treatment” for dosing regimen) or 20 mg/kg (early postnatal administration paradigm; see “Early postnatal PCP treatment” for dosing regimen). d-Amphetamine ((+)-α-methylphenethylamine sulphate; Unikem A/S, Copenhagen, Denmark) was dissolved in 0.9% sodium chloride to obtain a dose of 0.5 mg/kg and was administered s.c. at time 0 min before testing in the hyperactivity paradigm. Modafinil (2-[(diphenylmethyl)sulfinyl]acetamide; Sequoia Research Products, UK) was suspended in 0.5% methylcellulose (Sigma, Denmark) to a dose of 64 mg/kg. This dose of modafinil was selected based on previous published findings (Goetghebeur and Dias 2009; Goetghebeur et al. 2010; Waters et al. 2005). Due to its short half-life in rats (Waters et al. 2005), modafinil was dosed twice (p.o.) for ID/ED testing: once 30 min prior to the first discrimination and again 30 min before the fifth (ID2R) discrimination stage. QH-II-066 (7-acetyleno-1,3-dihydro-1-methyl-5-phenyl-2H-1,4-benzodiazepin-2-one, synthesised in-house) was dissolved in 25% solutol (Sigma) to obtain doses of 0.3, 1 and 3 mg/kg and was administered s.c. 30 min before acquisition training in NOR or amphetamine administration in amphetamine-induced hyperactivity. RO4938581 (3-Bromo-10-difluoromethyl-9H-imidazo[1,5-a] [1,2,4]triazolo[1,5-d] [1,4]benzodiazepine, synthesised in-house) was dissolved in 100% PEG-400 (Sigma) to doses of 0.1, 0.3, 1 and 10 mg/kg and was administered p.o. 60 min before acquisition training in NOR (0.3 and 1 mg/kg) or amphetamine administration in amphetamine-induced hyperactivity (0.1, 1 and 10 mg/kg) and 30 min before testing in ID/ED (1 mg/kg). The doses of RO4938581 were selected based on previous published findings (Ballard et al. 2009; Knust et al. 2009). All doses are expressed as milligrams per kilogram free base and were injected in a volume of 5 ml/kg.

Sub-chronic PCP treatment

Rats were administered sub-chronic PCP (5 mg/kg, i.p., b.i.d. for 7 days) or saline (1 ml/kg i.p. b.i.d. for 7 days) at 7 am and 7 pm, followed by a 7-day washout period prior to behavioural testing (as described in “Early postnatal PCP treatment”). This treatment regime is in accordance with previous published methods (see Egerton et al. 2005; Rodefer et al. 2005, 2008).

Early postnatal PCP treatment

The early postnatal PCP paradigm as applied in the current study has been previously described (Broberg et al. 2008, 2009). Briefly, timed pregnant Lister hooded rats were obtained at gestational day 15 from Charles River and were housed individually until delivery. The day of parturition was counted as postnatal day (PND) 0. On PND 5, male pups were cross-fostered and randomly assigned to a lactating dam. Female members of the litters were removed at this time. Furthermore, on PND 7, 9 and 11, pups were treated subcutaneously (s.c.) with either vehicle (0.9% isotonic saline) or PCP (20 mg/kg in a 10-ml/kg dose volume). Treatment with PCP within the first 2 weeks of a rat’s postnatal life corresponds to the second late trimester in human pregnancy in terms of neurodevelopmental changes (Bayer et al. 1993; Clancy et al. 2001) and fits with the hypothesis that exposure to PCP during this trimester increases the probability for the progeny to develop schizophrenia (Deutsch et al. 1998; Green et al. 1994). Pups were weaned on PND 25 and housed two per cage until behavioural testing (as described in “Novel object recognition task: apparatus and behavioural testing ”).

Novel object recognition task: apparatus and behavioural testing

Prior to the initiation of behavioural studies, a predefined inclusion/exclusion criterion was adopted for subsequent NOR data analysis. Animals that did not reach a total object exploration time of 15 s, either during a maximum allowed 3-min acquisition session or a full 3-min test trial period, were excluded from the data analysis. Furthermore, for the test trial period, animals that reached the 15-s object exploration criterion but did not explore both objects for a minimum of 2 s were excluded from the data analysis (e.g. a rat exploring object A for 15 s and object B for 0 s would be excluded. A rat exploring object A for 2 s and object B for 13 s would be included). This inclusion/exclusion criterion was adopted to ensure that each animal included in the experimental analysis had explored each object in acquisition and trial periods and to control for chance exploration. NOR testing was performed in a rectangular arena surrounded by non-transparent Plexiglass walls (dimensions, 95 × 45 × 50 cm). The arena was placed on a downflow air ventilation table in order to minimize the spread of allergens and was surrounded by a black curtain. Animals were habituated to the empty arena in groups of four in 2 × 10-min sessions on day 1. On day 2, individual animals were introduced to the arena for an acquisition session with two identical objects (two opaque Perspex pyramids, 10 × 10 × 6 cm, or two domed glass paperweights, 8 × 8 × 8 cm) for a maximum allowed period of 3 min. The acquisition session was terminated early once the rat reached a total exploration time of 15 s or continued for the full 3-min period until 15 s of active exploration had been achieved. This protocol, although introducing some variation in total test arena (non-object) exploration time between subjects, was an attempt to ensure equal exploration of objects across subjects during the acquisition session and thus encourages the encoding of a uniform object memory trace for use in the test trial. The animals were then transferred to their home cages for a 1-h inter-trial interval. Following this period, the animal was reintroduced into the arena for the test session. Here, a triplicate of the familiar and a novel object were placed in the arena and the animal was allowed to explore during a full 3-min test session. A camera mounted above the arena relayed the animal’s general behaviour (e.g. distance travelled) to a computer equipped with Ethovision tracking software (Noldus, The Netherlands). Object interaction, recorded on video, was scored manually by an experimenter blinded to the dose groups on completion of the experiment. Object exploration was defined as sniffing, licking or touching the object whilst facing it. All testing was performed on day 8 or 9 after the last PCP injection.

Attentional set-shifting task (ID/ED): apparatus and behavioural testing

The test apparatus consisted of a 44 × 64 × 30-cm black non-transparent Perspex test box divided into three equal-sized areas (a starting area separated from a choice area subdivided in two). In the choice area, two 11-cm diameter terracotta pots were placed recessed into the box floor (with a 2-cm lip remaining above the floor) and separated by a divider. Odour (oils from The Body Shop®, UK and Denmark) and media cues (not previously presented to the rats) were added to the pots. The test box was placed on a downflow air ventilation table in order to minimize the spread of allergens and mixing of the odours.

The attentional set-shifting procedure employed in our laboratories has been described in detail previously (Broberg et al. 2008, 2009; Goetghebeur and Dias 2009; Goetghebeur et al. 2010; Pedersen et al. 2009). In brief, rats were required to locate a food reward (1/2 Honey Nut Cheerio, Nestlé, Denmark) on the basis of digging media or odour as the relevant perceptual dimension. Prior to the test, rats were trained to dig for a food reward in terracotta pots filled with standard bedding media. For 3 days, rats were habituated to the test pots filled with cage bedding and food rewards. On day 4, all rats were then presented with two different media and, thereafter, two different odours, and were required to distinguish which of the two media or two odours were associated with the food reward. Subsequently (on day 5), rats were subjected to the following discriminations: simple discrimination (SD), compound discrimination (CD), intradimensional shift 1 (ID1), intradimensional shift 2 (ID2), intradimensional shift 2 reversal (ID2R), extradimensional shift (ED) and extradimensional shift reversal (EDR). The SD test required the rat to discriminate between two different cues within the same dimension (media only, no odours). In the CD, the pots contained the same media as in the SD, but with odours now added to them. New sets of media and odours were employed in ID1, ID2 and ED. In the ID2R and EDR, the incorrect stimuli from ID2 and ED were now correct, and vice versa (see Table 1). Animals were required to make six consecutive correct trials in order to advance to the next discrimination. Rats were allowed one “discovery” trial at the start of each training or test discrimination problem in which they were allowed to self-correct a dig in the wrong pot. The discovery trial was counted towards the number of trials to performance criterion if the rat dug first in the correct pot, but not if the rat self-corrected an initial incorrect dig. The number of trials to reach the criterion level of performance at each test stage was recorded. Omissions were defined as an animal’s refusal to participate in the task (sniff or dig in either pot) for greater than 15 consecutive minutes. After an omission, animals were returned to the home cage to rest for approximately 30 min before testing was resumed, continuing from the trial the animal had reached prior to its omission break. A cutoff time of 6 h or 100 trials per discrimination (whichever came sooner) was imposed, after which the animal was not included in the study. Throughout testing, Honey Nut Cheerio dust was sprinkled over both pots (relevant/irrelevant) to reduce the risk that animals may have smelt the reward in the relevant pot. All testing was conducted between PND 56 and 96.

Table 1 Order of discriminations and examples of possible combinations of stimulus pairs in the ID/ED task
Full size table

Amphetamine-induced hyperactivity

The amphetamine-induced hyperactivity procedure employed was adapted from previously published protocols (Sotty et al. 2009a, b). Briefly, rats were placed in the test cage (macrolon type III) situated in a U-frame equipped with four infrared light sources and photocells. The light beams crossed the cage 4 cm above the bottom of the cage which was covered by a thin layer of standard bedding material. Recording of a motility count required the interruption of adjacent light beams, thus avoiding counts induced by stationary movements of the rat. Animals were allowed a 1-h habituation period to the test apparatus prior to amphetamine administration.

Exposure studies

In separate groups of Lister hooded rats, blood samples and brains were collected to determine plasma and brain levels of QH-II-066 and RO4938581 [sampling times, 30, 60, 120 and 180 min (for RO4938581) following drug administration]. QH-II-066 was dosed at 3 mg/kg, s.c. (5 ml/kg) in 25% solutol. RO4938581 was dosed at 1 mg/kg, p.o. (5 ml/kg) in 100% PEG-400. Three animals were used at each time point. Cardiac blood was obtained under isoflurane anaesthesia followed by decapitation. Total drug concentrations were determined in harvested EDTA plasma and brain homogenates using ultra performance LC chromatography followed by MS/MS detection. Free (unbound) plasma and brain levels were calculated at each time point by multiplying the measured total concentrations with the free fractions in plasma and brain obtained from equilibrium dialysis in vitro.

Statistical analysis

For immunohistochemical quantification of parvalbumin-positive cells in the prefrontal cortex of sub-PCP- and neo-PCP-treated rats, data are expressed as per cent parvalbumin-positive stained cells in the prefrontal cortex. Student’s t test was used to investigate statistical differences between test groups (P < 0.05).

For novel object recognition experiments, the data are presented as: (1) mean exploration time (in seconds) of novel or familiar object; (2) mean total exploration time of novel and familiar object (in seconds); (3) mean total distance travelled (in centimetres). Paired t tests (exploration time) and one-way analysis of variance (ANOVA), followed by appropriate multiple comparison post hoc analysis (Bonferroni’s t test, total exploration and distance travelled), were used to investigate statistical differences between test groups (P < 0.05).

For the attentional set-shifting experiment, data are presented as the mean number of trials required to achieve six consecutive correct responses in each discrimination stage. A two-way repeated measures ANOVA (discrimination × drug), followed by appropriate multiple comparison post hoc analysis (Bonferroni’s t test), was used to investigate statistical differences between test groups (P < 0.05).

For amphetamine-induced hyperactivity, data are presented as the mean activity counts over the 60-min test session. A one-way ANOVA, followed by appropriate multiple comparison post hoc analysis (Bonferroni’s t test), was used to investigate statistical differences between test groups (P < 0.05).

Results

Activities of GABAA modulators on α5β2γ2 and α2β2γ2 receptors determined by the FLIPR® Membrane Potential Assay Kit

GABA concentration–response curves on the α5β2γ2 receptor subtype were determined in the absence and in the presence of the negative allosteric modulators—RO4938581 and DMCM—and the positive allosteric modulator—diazepam—using the FLIPR® membrane potential assay (Fig. 1a). The strong non-selective negative modulator, DMCM, shifted the curve to the right; the EC50 value for GABA was increased from 3.8 ± 0.25 to 7.8 ± 0.52 μM. RO4938581 was shown to be a weaker negative modulator, shifting the GABA curve to a lesser extent (from 3.8 ± 0.25 to 5.8 ± 0.44 μM). In contrast, the positive modulator, diazepam, shifted the GABA curve to the left—from 3.8 ± 0.25 to 2 ± 0.15 μM (Fig. 1a). The selective profile of RO4938581 (Ballard et al. 2009; Knust et al. 2009) was confirmed in the FLIPR® membrane potential assay setup since no shift in the GABA concentration–response curves was seen in cells expressing the α2β2γ2 receptor subtype in the absence (EC50 = 17 ± 1.4 μM) or presence of RO4938581 (EC50 = 19 ± 0.59 μM; Fig. 1b).

Fig. 1
figure 1

Activities of GABAA modulators on α5β2γ2 and α2β2γ2 receptors determined by the FLIPR® Membrane Potential Assay Kit (Molecular Devices Corporation). A representative dataset (out of n = 4) shows changes in voltage across the cell membrane induced by GABA without or with saturating concentrations (10 μM) of allosteric modulators to the α5β2γ2 (a) and α2β2γ2 (b) GABAA receptor subtypes

Full size image

Activities of GABAA modulators on α5β3γ2 and α1β3γ2 receptors determined by electrophysiological recordings in oocytes

GABA (3 μM), applied for 1 min every 5 min, induced the expected inward currents in oocytes injected with cRNA encoding GABA α5β3γ2 and α1β3γ2 receptors. When a stable response was obtained with GABA, DMCM, RO4938581 and diazepam were tested at 0.1 μM for modulation of the GABA response on both α5β3γ2- and α1β3γ2-containing receptors. DMCM showed negative modulation at both α5- and α1-containing receptors, with an inhibition equal to 54 ± 8% and 65 ± 7%, respectively (Fig. 2b, e). RO4938581 showed a weaker negative modulatory effect at α5-containing receptors, with an inhibition equal to 27 ± 5%, whilst no effects were seen at α1-containing receptors (1 ± 1%; Fig. 2a, d). In contrast, diazepam showed positive modulation at both α5- and α1-containing receptors, with an increase in current equal to 154 ± 23% and 194 ± 63%, respectively (n = 3–5 for all test groups; Fig. 2c, f).

Fig. 2
figure 2

Activities of GABAA modulators on α5β3γ2 and α1β3γ2 receptors determined by electrophysiological recordings in X. laevis oocytes (n = 3–5). White bar Control stimulation with 3 μM GABA. Black bar Application of test compound followed by co-stimulation with 3 μM GABA. Scale bar, horizontal, 1 min; vertical, 0.05 μA (a); 0.02 μA (b, c); 0.2 μA (d–f)

Full size image

Off-target binding profile

RO4938581 was studied at 10 μM in a 100 standard receptor binding profile (CEREP) and found to be devoid of activity (inhibition lower than 50%) for the ion channels, neurotransmitter or peptide receptors tested (data not shown).

Immunohistochemical quantification of parvalbumin-positive cells in the prefrontal cortex of sub-PCP- and neo-PCP-treated rats

The analysis demonstrated that both sub-PCP- (69 ± 6.19%, n = 10; Fig. 3a) and neo-PCP-treated rats (69 ± 6.87%, n = 9; Fig. 3b) displayed a statistically significant reduction of parvalbumin-positive cells in the prefrontal cortex when compared with vehicle-treated sub-PCP (100 ± 11.17%, P < 0.0001, Student’s t test, n = 10; Fig. 3a) or vehicle-treated neo-PCP rats (100 ± 23.20%, P < 0.0001, Student’s t test, n = 10; Fig. 3b).

Fig. 3
figure 3

Immunohistochemical quantification of parvalbumin-positive cells in the prefrontal cortex following administration of sub-chronic vehicle (sub-V-treated; 1 ml/kg i.p., b.i.d, 7 days) or PCP (sub-PCP-treated; 5 mg/kg i.p., b.i.d, 7 days plus a 7-day washout) (a) and early postnatal administration of vehicle (neo-V-treated; 10 ml/kg, s.c., PND 7, 9 and 11) or PCP (neo-PCP-treated; 20 mg/kg, s.c., PND 7, 9 and 11) (b). Data are expressed as per cent parvalbumin-positive-stained cells in prefrontal cortex + SD. ***P < 0.0001 versus vehicle-treated group (Student’s t test, n = 9–10)

Full size image

Effect of QH-II-066 in the NOR task

The overall study included a vehicle-treated control group and three QH-II-066 dose groups, each containing 12 rats. Following video scoring by a trained observer, blinded to the treatment groups, two rats from the vehicle-treated control group, four from the 0.3-mg/kg, two from the 1-mg/kg and two from the 3-mg/kg QH-II-066-treated groups were excluded from the experiment as they did not reach inclusion criteria (see “Novel object recognition task: apparatus and behavioural testing”). Animals explored each object equally during the acquisition session (data not shown). As shown in Fig. 4a, vehicle-treated rats spent significantly more time exploring the novel object during the 3-min test trial (P < 0.01, t = 4.11, paired t test, n = 10). Animals treated with a low dose of QH-II-066 (0.3 mg/kg) also preferred the novel object compared with the familiar object (P < 0.05, t = 2.37, paired t test, n = 8); however, increasing doses (1 and 3 mg/kg, n = 10) disrupted the preference for novel object exploration, i.e. rats spent approximately equal amounts of time investigating both novel and familiar objects. Administration of QH-II-066 did not affect the total exploration time (Fig. 4b) or distance travelled (Fig. 4c) at the doses employed.

Fig. 4
figure 4

Effects of QH-II-066 on exploration time (a), total exploration time (b) and distance travelled (c) in the NOR task. Rats were given acute administration of vehicle or QH-II-066 (0.3, 1 or 3 mg/kg, s.c.) prior to behavioural testing. Data are presented as the mean+SEM. *P < 0.05; **P < 0.01 versus novel object exploration time (paired t test, n = 8–10)

Full size image

Effect of RO4938581 in the sub-PCP-NOR task

The overall study included a vehicle-treated control group (sub-chronic saline twice daily for 7 days, followed by a 7-day washout and acute vehicle on test day), a PCP-vehicle group (sub-chronic PCP twice daily for 7 days, followed by a 7-day washout and acute vehicle on test day) and two PCP-RO4938581 dose groups (sub-chronic PCP twice daily for 7 days, followed by a 7-day washout and acute RO4938581 on test day). Each experimental group consisted of 12 rats. Following video scoring by a trained observer, blinded to the treatment groups, one rat from the vehicle-treated control group and two rats from the PCP-vehicle group were excluded from the experiment as they did not reach inclusion criteria (see “Novel object recognition task: apparatus and behavioural testing”). Animals explored each object equally during the acquisition session (data not shown). As shown in Fig. 5a, vehicle-treated rats spent significantly more time exploring the novel object during the 3-min test trial (P < 0.01, t = 3.35, paired t test, n = 11). Vehicle-treated animals, which had previously received PCP for 1 week, 7 days prior, spent approximately an equal amount of time exploring both objects (n = 10). Administration of RO4938581 (0.3 and 1 mg/kg, n = 12) dose-dependently attenuated PCP-induced deficits in novel object exploration, an effect that reached statistical significance for the 1-mg/kg treatment group (P < 0.001, t = 6.23, paired t test). RO4938581 did not affect the total exploration time (Fig. 5b), but did reduce the distance travelled (Fig. 5c) at the highest dose tested (1 mg/kg; one-way ANOVA: F(3,44) = 2.94; Bonferroni’s t test: P = 0.04), but not the lower dose (0.3 mg/kg, P = 0.07).

Fig. 5
figure 5

Effects of RO4938581 on exploration time (a), total exploration time (b) and distance travelled (c) in the NOR task following administration of sub-chronic vehicle (control; 1 ml/kg, i.p., b.i.d, 7 days) or PCP (5 mg/kg, i.p., b.i.d, 7 days plus a 7-day washout). PCP-treated animals were given acute administration of vehicle or RO4938581 (0.3 or 1 mg/kg, p.o.) prior to behavioural testing. Data are presented as the mean+SEM. **P < 0.01, ***P < 0.001 versus novel object exploration (paired t test, n = 10–12); *P < 0.05 versus vehicle-treated group distance travelled (one-way ANOVA, followed by Bonferroni’s t test, n = 10–12)

Full size image

Effect of RO4938581 in the neo-PCP-ID/ED task

The overall study included a vehicle-treated control group (n = 9, early postnatal saline on PND 7, 9, 11 and test day), a PCP-vehicle group (n = 9, early postnatal PCP PND 7, 9, 11 and vehicle on test day), a PCP-RO4938581-treated group (n = 8, early postnatal PCP PND 7, 9, 11 and RO4938581 on test day) and a PCP-modafinil group (n = 5, early postnatal PCP PND 7, 9, 11 and modafinil on test day). Based on previous consistently positive in-house data, the PCP-modafinil-treated group size was kept to a minimum in the interest of reducing the number of animals used. All animals enrolled in the study reached the inclusion criteria (see “Attentional set-shifting task (ID/ED): apparatus and behavioural testing”). As shown in Fig. 6, statistical analyses revealed an overall interaction between task and treatment (two-way repeated measures ANOVA: F(18,162) = 5.96, P < 0.001). Post hoc analyses confirmed the formation of an attentional set via a comparison of performance (trials to criterion) at the ID2 versus the ED test stage for vehicle-treated control animals (Bonferroni’s t test: P < 0.001), thus verifying that the ED shift was more demanding than the ID2 shift discrimination (i.e. animals were “locked” in the relevant dimension). Furthermore, post hoc analyses revealed that the early postnatal PCP-treated rats were significantly impaired in their performance at the ED and EDR test stages when compared with the vehicle-treated control group (P < 0.001). No differences between treatment groups at any other discrimination stage (i.e. SD, CD, ID1, ID2 and ID2R) were observed, implying that the early postnatal PCP treatment regime employed specifically affected ED shift and EDR performance only. In contrast, RO4938581 (1 mg/kg) and modafinil (64 mg/kg) treatment significantly improved both ED shift (P < 0.001 and P < 0.001, respectively) and EDR (P < 0.01 and P < 0.001, respectively) performance, relative to the ED and EDR discrimination scores of the early postnatal PCP-treated rats given an acute vehicle challenge. The time taken to perform an individual trial, irrespective of discrimination, was comparable between vehicle–vehicle and PCP-vehicle-treated groups (data not shown).

Fig. 6
figure 6

Effects of RO4938581 on the number of trials required to reach a criterion of six correct consecutive trials for each discrimination following early postnatal administration of vehicle (control; 10 ml/kg, s.c., PND 7, 9 and 11) or PCP (20 mg/kg, s.c., PND 7, 9 and 11). PCP-treated animals were given acute administration of vehicle, RO4938581 (1 mg/kg, p.o.) or modafinil (64 mg/kg, p.o.) prior to behavioural testing. Data are presented as the mean+SEM. $ P < 0.001 versus vehicle + vehicle-treated group (V+V) at ID2 discrimination. ***P < 0.001 versus vehicle + vehicle-treated group (V+V) at ED and EDR discrimination; ## P < 0.01, ### P < 0.001 versus PCP + vehicle-treated group at ED and EDR discrimination (two-way RM ANOVA, followed by Bonferroni’s t test, n = 5–9)

Full size image

Effect of QH-II-066 and RO4938581 on basal and amphetamine-induced hyperactivity

As shown in Fig. 7a, QH-II-066 did not significantly affect basal horizontal motor activity at any of the doses tested when measured during the habituation period. Treatment with amphetamine significantly increased total horizontal motor activity when compared with vehicle + vehicle-treated rats (one-way ANOVA: F(4,39) = 5.43; Bonferroni’s t test: P < 0.01), an effect that was significantly attenuated by 1 and 3 mg/kg, but not 0.3 mg/kg, QH-II-066 (Bonferroni’s t test: P < 0.05; Fig. 7b).

Fig. 7
figure 7

Effects of QH-II-066 on basal locomotor activity during habituation phase (a) and amphetamine-induced hyperactivity (b). Rats were given acute administration of vehicle, QH-II-066 (0.3, 1 or 3 mg/kg, s.c.), amphetamine (0.5 mg/kg), or a combination thereof prior to behavioural testing. Data are presented as the mean+SEM. ## P < 0.01 versus vehicle + vehicle-treated (V+V) group; *P < 0.05 versus amphetamine + vehicle-treated (AMPH+V) group (one-way ANOVA, followed by Bonferroni’s t test, n = 8)

Full size image

As shown in Fig. 8a, RO4938581 significantly reduced total basal horizontal motor activity at the highest dose tested (10 mg/kg) when measured during the habituation period (one-way ANOVA: F(3,31) = 4.23; Bonferroni’s t test: P < 0.05). Treatment with amphetamine significantly increased total horizontal motor activity when compared with vehicle + vehicle-treated rats (one-way ANOVA: F(4,39) = 11.20; Bonferroni’s t test: P < 0.001), an effect that was significantly attenuated by RO4938581 (0.1, 1 and 10 mg/kg; Bonferroni’s t test: P < 0.05; Fig. 8b).

Fig. 8
figure 8

Effects of RO4938581 on basal locomotor activity during habituation phase (a) and amphetamine-induced hyperactivity (b). Rats were given acute administration of vehicle, RO4938581 (0.1, 1 or 10 mg/kg, p.o.), amphetamine (0.5 mg/kg), or a combination thereof prior to behavioural testing. Data are presented as the mean+SEM. a *P < 0.05 versus vehicle-treated group (one-way ANOVA, followed by Bonferroni’s t test, n = 8). b ### P < 0.001 versus vehicle + vehicle-treated (V+V) group; *P < 0.05 versus amphetamine + vehicle-treated (AMPH+V) group (one-way ANOVA, followed by Bonferroni’s t test, n = 8)

Full size image

Plasma and brain levels

Groups of satellite rats were used to assess the plasma and brain levels of QH-II-066 and RO4938581. The free fractions for QH-II-066 were 16% and 4.5% in rat plasma and brain, respectively. For RO4938581, the free fractions were 14% and 21% in plasma and brain, respectively. Both compounds were shown to have good brain penetration with sustained exposure up to 120 min after dosing. As shown in Table 2, favourable free brain concentrations, with respect to published GABAA α5 receptor K i values (Ballard et al. 2009: Knust et al. 2009; Huang et al. 1996), were obtained at doses of QH-II066 and RO4938581 shown to modulate behaviour in the sub-PCP-NOR, neo-PCP-ID/ED and amphetamine-induced hyperactivity paradigms.

Table 2 Free plasma/brain concentrations of QH-II-066 and RO4938581 corresponding to behavioural analysis time points post-administration
Full size table

Discussion

The data from the present studies demonstrate, for the first time, that negative modulation of GABAA α5 receptors with RO4938581, a selective negative modulator (Ballard et al. 2009; Knust et al. 2009; present data), ameliorated both sub-PCP- and neo-PCP-induced cognitive deficits in rat novel object recognition and attentional set-shifting performance, respectively. Importantly, the combination of two extensively validated behavioural assays with two disease-like symptomatic animal models, with some relevance to schizophrenia, i.e. sub-PCP (NOR) or neo-PCP (ID/ED), lends strong support to the present data. In contrast, positive modulation of GABAA α5 receptors with QH-II-066, a moderately selective positive modulator (Huang et al. 1996, 2000; Skolnick et al. 1997), significantly and selectively impaired cognitive performance in the NOR task when administered to vehicle-treated animals. Interestingly, additional novel studies revealed that RO4938581 and QH-II-066 attenuated amphetamine-induced hyperactivity in rats at doses shown to improve or impair cognitive performance, respectively.

In an attempt to align research efforts and aid the clinical evaluation of potential pro-cognitive therapeutics, the Measurement and Treatment Research to Improve Cognition in Schizophrenia and, more recently, the Cognitive Neuroscience Treatment Research to Improve Cognition in Schizophrenia initiatives have identified several cognitive domains thought to be negatively affected in schizophrenia (Green et al. 2004; Barch et al. 2009). One of the preclinical assays suggested as a useful tool for the investigation of domain-specific deficits is the novel object recognition (NOR) task (Ennaceur and Delacour 1988). This test paradigm is thought to reflect visual learning and recognition processing (Young et al. 2009) and has been compared with, among others, the visual paired comparison task in humans (Clark et al. 2000; Manns et al. 2000). Interestingly, schizophrenia patients show deficits in 2D object recognition tasks very similar to the NOR paradigm (Clare et al. 1993; Aleman et al. 1999; Heckers et al. 2000; Tek et al. 2002), which is of importance for the task’s validity in relation to preclinical studies of schizophrenia. In the present study, positive modulation of GABAA α5 receptors by QH-II-066 was shown to impair cognitive performance of vehicle-treated rats in the NOR task without concurrent effects on motor activity. Exposure analysis of the highest dose tested (3 mg/kg), conducted in a group of satellite animals, indicated that plasma/brain levels were well within the range thought to enable target engagement based on previously published in vitro (Huang et al. 1996, 2000) and in vivo (Skolnick et al. 1997) data (see Table 2). In contrast, negative modulation of GABAA α5 receptors by RO4938581 (1 mg/kg) was shown to attenuate a sub-PCP-induced deficit and to reduce the total distance travelled in the NOR task. The reduction in horizontal motor activity induced by RO4938581 did not adversely impact upon total exploration time, which was comparable to that displayed by vehicle-treated rats and may therefore reflect a reduction in exploration of the test arena. Once more, exposure studies indicated that selective target engagement was most likely reached, based on previous in vitro and in vivo data (Ballard et al. 2009; Knust et al. 2009), at the dose that attenuated the sub-PCP-induced deficit (see Table 2). Interestingly, the approximate free concentrations of RO4938581, obtained in brain that induced efficacy, corresponded to a negative modulation of 27% (at 100 nM) in our oocyte electrophysiology experiments, within range of the previously reported maximum negative modulation in HEK293 cells and to elicit pro-cognitive effects in rodents (Ballard et al. 2009; Knust et al. 2009). In both NOR studies, certain animals were excluded from the data analysis as they did not reach the predetermined inclusion criteria. However, the few animals excluded from the analyses were equally represented across vehicle- and drug-treated groups, so a drug-induced “exclusion effect” is unlikely.

Another preclinical assay considered to display some translational validity, with respect to schizophrenia, is the intradimensional/extradimensional attentional set-shifting (ID/ED) task (Goetghebeur and Dias 2009; Goetghebeur et al. 2010; Rodefer et al. 2005, 2008). This task is thought to reflect fronto-cortical executive-like functioning (Birrell and Brown 2000) and has been compared with the Wisconsin Card Sort Test (Berg 1948) and the CANTAB intradimensional/extradimensional tests (Sahakian et al. 1988). With regard to the potential translational validity of the ID/ED task, first-episode schizophrenia patients show an impaired ability to shift their attentional set between perceptual dimensions (Elliott et al. 1995; Haut et al. 1996; Pantelis et al. 1999). Similarly, sub-PCP-treated (Rodefer et al. 2005; McLean et al. 2008) and neo-PCP-treated (Broberg et al. 2008, 2009) rats display a set-selective deficit at the ED shift stage of the task. In the present study, RO4938581 significantly attenuated a neo-PCP-induced cognitive deficit at both the ED and EDR discrimination stages in the ID/ED task, suggesting beneficial effects on both set-shifting and reversal learning abilities. The deficit observed at the EDR stage in neo-PCP-treated rats contrasts with the two previously reported ID/ED studies in this model system (Broberg et al. 2008, 2009) where a deficit was selectively observed at the ED shift. This new finding is interesting in a sense that it is reminiscent of the consistent deficit in reversal learning observed in first-episode schizophrenic patients (Hutton et al. 1998; Leeson et al. 2009; Murray et al. 2008). Modafinil, used here as a positive control, was also successful in alleviating neo-PCP-induced deficits to a level qualitatively similar to that of RO4938581. The effects of modafinil seen here in neo-PCP-treated rats are in agreement with, and further validate, recent studies where acute administration improved attentional set-shifting performance in sub-PCP-treated rats (Goetghebeur and Dias 2009; Goetghebeur et al. 2010) and are reminiscent of the beneficial effect of modafinil observed in the clinic (Turner et al. 2004).

To our knowledge, this is the first report demonstrating that negative modulation of GABAA α5 receptors improved cognitive performance in two well described preclinical assays (i.e. the NOR and ID/ED tasks) combined with two disease-like animal models (i.e. sub-PCP-treated and neo-PCP-treated, respectively). One could argue that a potential caveat in our studies with RO4938581 lies in the fact that we did not include a compound-alone control group in NOR and ID/ED experiments. However, given the plethora of published data suggestive of a pro-cognitive potential for GABAA α5 negative modulators in normal animals or under acute pharmacological deficit conditions, as outlined in the “Introduction”, we found it less interesting (taking animal numbers used into account) to include such a group to confirm published observations. We were more interested in the potential effects of RO4938581 in a disease-relevant setting, i.e. in animals exposed to sub-chronic or early postnatal PCP. When the aforementioned published data are considered, it is perhaps not surprising that positive modulation by QH-II-066 significantly impaired cognitive ability in the NOR task in non-PCP-treated rats. The present data with RO4938581 are suggestive of a negative modulation strategy that may be of benefit in the treatment of domain-specific cognitive deficits associated with schizophrenia.

The elucidation of the precise mechanism(s) by which negative modulation of GABAA α5 receptors attenuated both sub-PCP- and neo-PCP-induced deficits in the NOR and ID/ED tasks is beyond the scope of this manuscript; however, some tentative possibilities will now be discussed.

One plausible mechanism for the pro-cognitive-like effects of RO4938581 in the sub-PCP and/or neo-PCP model systems used here is found in the notion that negative modulation of GABAA α5 receptors in the hippocampus would most likely lead to hippocampal disinhibition and increased excitatory output to certain target structures, including fronto-cortical regions (for a review, see Thierry et al. 2000), thereby attenuating sub-PCP- and neo-PCP-induced “hypofrontality” (for reviews, see Jentsch and Roth 1999; Neill et al. 2010). There is substantial evidence implicating an important role for NMDA receptors in learning and memory processing (for a review, see Rezvani 2006) and that NMDA functionality is aberrant in schizophrenia (for a review, see Gilmour et al. in press). The NMDA receptor hypofunction hypothesis originates from observations that NMDA receptor antagonists, such as ketamine and PCP, exacerbate psychosis in schizophrenia patients and induce schizophrenia-like symptoms in healthy volunteers (Hetem et al. 2000; Honey et al. 2006; Lahti et al. 1995). Sub-PCP administration to rodents, followed by a washout period, has been shown to affect hippocampal and fronto-cortical areas, thereby establishing a hypofunction of the prefrontal cortex resembling some aspects of schizophrenia pathology (Cochran et al. 2003). Correspondingly, early postnatal treatment of rats with PCP has been proposed as a neurodevelopmental model of schizophrenia (Wang et al. 2001) as this treatment paradigm produces widespread neurodegeneration in brain areas relevant to the cognitive deficits observed in schizophrenia patients, such as the hippocampus and frontal cortex (Ikonomidou et al. 1999; Wang and Johnson 2005). As performance in the ID/ED task relies heavily on fronto-cortical processing (Birrell and Brown 2000), such a mechanistic theory could be applied to the positive effects of RO4938581 on neo-PCP-induced deficits seen here. Specifically related to the sub-PCP-NOR task, it has been proposed that a system principally involving the perirhinal cortex is responsible for familiarity discrimination, particularly for single items (Dere et al. 2007; Winters et al. 2008; Brown et al. 2010). As performance in the rodent NOR task is compromised following sub-PCP, disinhibition of hippocampal functioning engendered by negative modulation of GABAA α5 receptors may induce some increased functioning of the perirhinal cortex, which may in turn be responsible for the attenuation of sub-PCP-induced deficits in this assay. However, these somewhat speculative mechanistic claims need further investigation.

Several studies have highlighted a role for GABAA α5 receptors in the modulation of network oscillatory activity and synaptic plasticity. More specifically, negative modulation of GABAA α5 receptors, by genetic or pharmacological interventions, has been shown to increase the power and stability of gamma oscillations (Towers et al. 2004; Glykys et al. 2008) and to enhance stimulated long-term potentiation (LTP; Atack et al. 2006; Atack 2010, 2011), albeit using slice preparation electrophysiological techniques. On the other hand, increased GABAA α5 receptor activity completely blocked LTP and impaired memory performance, effects that were reversed by pretreatment with a GABAA α5 receptor-preferring negative modulator (Martin et al. 2009). Network oscillatory activity and LTP are believed to play major roles in cognitive processing (for reviews, see Wang 2010; Cooke and Bliss 2005) and are thought to be perturbed in the schizophrenic brain (Ferrarelli et al. 2008; Hasan et al. 2011). The integrity of gamma-band oscillatory activity has been shown to be under the control of parvalbumin-positive interneurons in the hippocampus and cortex (Bartos et al. 2007). Importantly, schizophrenics display a reduced expression of these interneurons in the prefrontal cortex (Beasley and Reynolds 1997). Interestingly, the present data clearly show reduced parvalbumin-positive interneuron staining in the prefrontal cortex of both sub-PCP-treated and neo-PCP-treated rats, mimicking the clinical findings and underlining the relevance of these disease models. Therefore, it is conceivable that an increase of power and stabilisation of gamma oscillations could be involved in the beneficial effects of RO4938581 observed in the present study using sub-PCP-treated and neo-PCP-treated rats.

In addition to these provisional hippocampal/fronto-cortical pro-cognitive mechanistic links, more substantial evidence exists suggesting that increased excitatory output from the hippocampus may lead to the dopaminergic system being driven into an overly responsive state and thus result in psychotic-like symptomatology (for a review, see Grace 2011). Such a scenario could, at first glance, be a concern if negative modulation of GABAA α5 receptors were to be considered as a viable option for the treatment of cognitive symptoms in patients already displaying psychotic symptoms, i.e. aggravation of positive symptoms by negative modulation of GABAA α5 receptors could be envisioned. In a preliminary attempt to investigate such a setting in rats, we decided to investigate the impact of positive (with QH-II-066) or negative (with RO4938581) modulation of GABAA α5 receptor functionality on amphetamine-induced hyperactivity, a behavioural assay often used to detect potential antipsychotic-like activity. Importantly, the dose of amphetamine (0.5 mg/kg) was chosen to allow room for showing inhibition or potentiation, respectively (Arnt 1995). Somewhat surprisingly, these experiments demonstrated that both RO4938581 and QH-II-066 attenuated hyperactivity induced by amphetamine at doses that did not modify basal locomotor activity. The present data with QH-II-066 are in agreement with recent studies using another GABAA α5 receptor positive modulator which was shown to attenuate amphetamine-induced hyperactivity in the methylazoxymethanol acetate developmental model of schizophrenia in rats (Gill et al. 2011), suggesting a potential beneficial effect on positive symptomatology. The present data with RO4938581 showed that negative modulation of GABAA α5 receptor function did not exacerbate, but attenuated, the effects of amphetamine, at least under the testing conditions employed. These data suggest that both strategies could have favourable profiles with respect to the alleviation of psychosis. Further studies are warranted to investigate the potential mechanisms involved in these somewhat unexpected results.

We feel that the potential limitations of the studies described herein also deserve some discussion. We decided to test the GABAA α5 receptor positive modulator, RO4938581, in the NOR assay and ID/ED task in sub-PCP- and neo-PCP-treated rats, respectively. This decision was made on the basis of knowledge gained from previous experience using these behavioural paradigms in these model systems. Although the data obtained using this approach are encouraging, we acknowledge that testing of the compound in both cognitive tasks in both disease-like animal models would have been of considerable added value, with respect to further validation of the concept.

Postmortem studies on human tissue have shown selective parvalbumin deficits in the prefrontal cortex (Beasley and Reynolds 1997) and hippocampus (Zhang and Reynolds 2002) of schizophrenic patients compared with controls. These clinical findings have more recently been back-translated to the prefrontal cortex (Pratt et al. 2008) and hippocampus (Jenkins et al. 2010) of animals treated with sub-PCP. We therefore decided to investigate the effects of sub-PCP and neo-PCP treatment on the expression of parvalbumin-positive cells in the prefrontal cortex as a surrogate marker of animal model validity. Our data showing a decrease in parvalbumin-positive cells in the prefrontal cortex of sub-PCP- and neo-PCP-treated rats are certainly interesting, especially with respect to our ID/ED data (i.e. a task reliant on fronto-cortical function; Birrell and Brown 2000), and give some support to our model validity. However, further studies investigating hippocampal parvalbumin-positive cell numbers would have been of considerable added value given the discrete localisation of GABAA α5 receptors in this brain region (Nutt 2006).

The ID/ED procedure employed here, and elsewhere (Broberg et al. 2008, 2009; Goetghebeur and Dias 2009; Goetghebeur et al. 2010; Pedersen et al. 2009), could be criticised in that animals were not randomised to both odour and media, but only media, at the SD stage of the task. However, we believe that the deficits seen at the ED stage of the task are not related to differences in saliency of the dimensions and that the observed drug effects are indeed specific to the task. Previous experiments in our laboratories showed that normal non-treated rats did not display a deficit when shifting from odour to media, or vice versa, at the ED stage versus the ID stage when only one ID stage was employed (i.e. the animals failed to form a good attentional set, unpublished observations). Only when a second ID stage (ID2) was introduced did animals find the attentional shift at ED more demanding versus ID (i.e. indicating they had indeed formed a good attentional set). If a difference between saliencies did exist, one would have expected to see a deficit at ED versus ID when only one ID stage was included.

Taken together, these novel findings suggest that negative modulation of GABAA α5 receptors may represent an attractive treatment option for the cognitive, and potentially positive, symptoms associated with schizophrenia and open up the opportunity for such a strategy to be further investigated. In contrast, the potential beneficial effect of positive modulation of GABAA α5 receptors, on dopaminergic neurotransmission, may be overshadowed by the inherent cognitive disrupting properties of such a strategy.