Introduction
Schizophrenia is a devastating mental disorder characterized by positive, negative, and cognitive symptoms. Cognitive deficits are a hallmark of this condition, and understanding the mechanisms of these symptoms is of particular significance as they are highly predictive of long-term disease prognosis (Green
1996). Cognitive symptoms include deficits in working memory and behavioral flexibility (Forbes et al.
2009; Leeson et al.
2009), two processes of executive function that are essential for normal cognition. Both the Measurement and Treatment Research to Improve Cognition in Schizophrenia initiative (MATRICS) and the Cognitive Neuroscience Treatment Research to Improve Cognition in Schizophrenia group (CNTRICS) highlight cognitive deficits as a core feature of this disorder (Nuechterlein et al.
2008,
2009). These deficits are resistant to treatment and the neural mechanisms that underlie them are not well understood (an der Heiden and Hafner
2000; Riedel et al.
2006).
Dysfunction of the glutamatergic system, specifically
N-methyl-
d-aspartate (NMDA) receptor hypofunction, has been hypothesized to play a central role in schizophrenia pathogenesis (Kantrowitz and Javitt
2010; Olney et al.
1999). Thus, NMDA receptor antagonists, such as phencyclidine (PCP), mimic deficits in executive function and working memory in humans (Cosgrove and Newell
1991; Krystal et al.
1994). Sub-chronic treatment with an NMDA antagonist in rodents similarly produces deficits in cognitive tasks including novel object recognition (Bado et al.
2011; Grayson et al.
2015). Critically, PCP-induced deficits in rodents are still present after prolonged drug-free periods (1 week and longer) and thus reflect enduring functional changes rather than acute drug effects. PCP-pre-treated animals exhibit impaired performance on novel object recognition, spontaneous alternation, and pre-pulse inhibition, which are believed to model cognitive and negative symptoms of human schizophrenia (Castane et al.
2015; Grayson et al.
2015; Jorgensen et al.
2015). The mechanisms through which PCP produces behavioral impairments are largely unknown, but neurophysiological and immunochemical results point to the involvement of prefrontal cortical circuits (Kargieman et al.
2012; Piyabhan et al.
2013; Young et al.
2015).
Medial prefrontal cortex (mPFC) provides a major input to nucleus accumbens (NAc), a structure widely implicated in motivational and cognitive tasks (Bolding et al.
2012; Pezze et al.
2009). Specifically, prelimbic cortex is implicated in behaviors relevant to negative symptomatology, such as goal-directed behavior (Chudasama et al.
2001; Killcross and Coutureau
2003). Cortico-accumbal inputs play a role in cognitive function, such as working memory and attention, and a recent patient fMRI study showed synchrony deficits between mPFC and accumbens in schizophrenics (French and Totterdell
2003; Richter et al.
2015). Limbic structures including NAc have been hypothesized to underlie some of the behavioral abnormalities in schizophrenia and show abnormalities in patient brain scans (Grace
2000; van Erp et al.
2015). Consistent with this, NAc shell but not core was hypothesized to mediate some of the therapeutic effects of antipsychotics (Shilliam and Dawson
2005). In rats, acute NMDA receptor blockade causes abnormal local field potential oscillations in NAc (Goda et al.
2015). These observations suggest that deficits in cortico-accumbal network activity may underlie some aspects of compromised behavioral function (Bolding et al.
2012).
To address this here, we recorded single-unit activity in NAc shell and mPFC in awake, behaving rats pre-treated with PCP. Cognitive deficits in NMDA antagonist-pre-treated animals are widely studied using the novel object recognition (NOR) task (Grayson et al.
2015; Rajagopal et al.
2014), yet the neural dysfunction underlying NOR deficits is not well understood. Consistent with the hypothesized role of prefrontal cortex in negative symptomatology, rodent mPFC is implicated in working memory and visual recognition memory (Akirav and Maroun
2006; Yoon et al.
2008). Here we found that synchrony between mPFC and NAc was disrupted by PCP pre-treatment and overall activity in the NAc shell was reduced. In addition to this global deficit, we found that phasic responses in both structures in response to ongoing behavior were affected: PCP pre-treatment disrupted the increase in mPFC and NAc activity associated with novel object exploration in control animals.
Discussion
PCP induces psychotic episodes in humans (Javitt and Zukin
1991), and its administration in rodents is widely used as an animal model of schizophrenia. Sub-chronic PCP treatment in rodents produces a number of behavioral abnormalities, which model cognitive and negative symptoms observed in schizophrenia. This study represents the first analysis of deficits in the cortico-accumbal circuit induced by sub-chronic NMDA antagonist administration in rats. As done by a number of different labs reporting on phencyclidine-induced deficits in rats, the current study used female animals (e.g., Martinez et al.
1999; Miyauchi et al.
2016; Sahin et al.
2016). This makes our results directly comparable to this previous work; importantly, however, previous studies also suggest female rats show higher vulnerability in the phencyclidine model (Snigdha et al.
2011), suggesting they are better suited for characterizing the deficits incurred by phencyclidine. Hormonal influences may affect sensitivity to PCP and this has been previously investigated in a rat model using acute PCP (Sutcliffe et al.
2008); however, in our study, both PCP- and saline-pre-treated animals were group-housed in the same holding room and tested in parallel. Thus, we do not suspect systematic differences in oestrous cycles to be a contributing factor for the differences we report here.
We found that repeated PCP treatment resulted in a tonic reduction in NAc shell single-unit firing rates and disrupted cortico-accumbal synchronization. These two regions are connected anatomically, and prelimbic cortex is implicated in behaviors relevant to negative symptomatology, such as goal-directed behavior and attention (Chudasama et al.
2001; Coutureau and Killcross
2003). Furthermore, nucleus accumbens shell but not core is hypothesized to mediate the therapeutic effects of antipsychotics (Shilliam and Dawson
2005). We also observed that phasic responses in both NAc shell and mPFC in response to ongoing behavior were affected: PCP disrupted the increase in mPFC and NAc neural activity, which was observed in vehicle-treated animals during exploration of a novel object. Consistent with the role of cortico-accumbal synchrony in memory and motivation, these neurophysiological deficits may drive widely reported cognitive abnormalities after sub-chronic NMDA antagonist treatment.
Previous studies by us and others have found impairments in visual working memory, pre-pulse inhibition, and NOR (Grayson et al.
2015; Mandillo et al.
2003; Sood et al.
2011). Consistent with these studies, here we observed a NOR impairment in PCP-pre-treated rats. Using an administration protocol, which produces reliable behavioral effects, we also observed a reduction in cortico-accumbal synchrony in PCP-pre-treated rats. These results are consistent with the role of PFC in schizophrenia and with previous work showing PCP-induced mPFC deficits. Thus, PFC is implicated in the neuropathology of schizophrenia with a notable increase in pyramidal cell density found in schizophrenic patients’ postmortem (Cullen et al.
2006) and abnormal cortical gamma-band oscillations (McNally et al.
2013; Uhlhaas and Singer
2010; Woo et al.
2010). In rodents, recent research shows disruption in single-unit responses in mPFC and abnormal synchrony after PCP treatment (Kargieman et al.
2007,
2012; Young et al.
2015). mPFC electrophysiological and neuroanatomical deficits in the sub-chronic PCP model suggest that abnormalities in a neural network encompassing mPFC may mediate novel object recognition deficits in this model (Castane et al.
2015; Young et al.
2015). The short retention intervals (1 min) tested in the NOR task in the PCP model are consistent with the role of mPFC in working memory (Grayson et al.
2007; Rosemann et al.
2010; Snigdha et al.
2011; Yoon et al.
2008). Interestingly, however, when longer retention delays are used (5 min or more), mPFC lesions do not impair object recognition (Barker and Warburton
2011; Sutcliffe et al.
2007). On the other hand, mPFC manipulations immediately after exposure impair long-term retention of the task (Akirav and Maroun
2006) further implicating mPFC on shorter retention timescales consistent with the role of the structure in working memory. Furthermore, lesion studies in both primates and rodents implicate mPFC in memory paradigms (Pezze et al.
2009; Rossi et al.
2009). Thus, prefrontal neurophysiological deficits observed in the PCP model may drive the behavioral abnormalities. Our results extend this by suggesting that abnormalities in cortico-accumbal projections may mediate some of the deficits. Consistent with the involvement of NAc, recent human imaging data suggest NAc abnormalities in schizophrenic patients (van Erp et al.
2015). In rats, one recent study reports altered local field oscillations in NAc in the methylazoxymethanol acetate (MAM) neurodevelopmental rodent model of schizophrenia (Goda et al.
2015). In drug-naïve animals, NAc single-unit activity increases in response to novelty during entry into a novel compartment (Wood and Rebec
2004) which is also consistent with our observations in the NOR task. Previous research implicating interactions between NAc and PFC in memory control supports the role NAc–PFC synchrony implicated in the current study. For example, a PFC–NAc disconnection significantly impaired auto-shaping in a stimulus-reward learning task, and attentional deficits in the 5-choice serial reaction time task produced by a PFC lesion are rescued by a dopaminergic antagonist in NAc (Parkinson et al.
2000; Pezze et al.
2009).
The deficits in mPFC–NAc communication observed here may involve more distributed circuits involving potentially hippocampus. Ventral hippocampal and mPFC projections to NAc show complex interactions, which were compromised in the MAM model, potentially involving ventral subiculum hyperactivity and mPFC hypoactivity (Belujon et al.
2014). NAc is a central structure for the integration of mPFC and hippocampal input. Interestingly, ventral hippocampal hyperactivity, as found in the MAM rodent model of schizophrenia (Lodge and Grace
2007), is proposed to disrupt NAc responses, providing a potential mechanism for the reduced NAc firing in the current study (O’Donnell et al.
1999). Also, hippocampal electrical high-frequency stimulation depotentiates the mPFC–NAc pathway (Goto and Grace
2005), and thus similarly plastic responses to hippocampal hyperactivity may underlie the reduced mPFC–NAc synchronization observed here. The precise systems interactions responsible for these effects must await further investigation. Interestingly, the MAM model results in some behavioral deficits which are similar to those reported in PCP-pre-treated rats, including impairments in reversal learning, and hypersensitivity to acute NMDA blockade (Belujon et al.
2014; Hradetzky et al.
2012). The involvement of distributed circuits is also suggested by the morphological abnormalities in limbic structures including anterior cingulate cortex and striatum observed in sub-chronic PCP-pre-treated rats (Barnes et al.
2014; Ingallinesi et al.
2014). Disruption in synchrony in the current study was observed between putative pyramidal, cortical and medium spiny neurons. Other neuromodulatory systems and/or striatal interneurons may mediate these effects. For example, prefrontal dopamine or NMDA blockade modulates NAc dopamine release (Del Arco and Mora
2008). Furthermore, NAc infusion of a dopamine antagonist has been found to rescue attentional deficits produced by prefrontal lesions (Pezze et al.
2009). Another potential source of modulation is NAc cholinergic interneurons, which modulate glutamatergic input as well as dopamine release in striatum (Gonzales and Smith
2015; Sofuoglu and Mooney
2009).
We also observed an increase in neuronal firing in mPFC and NAc shell associated specifically with novel object exploration in vehicle-treated animals in comparison to the familiar object consistent with previous studies showing novelty-induced activity in these structures recorded separately (Weible et al.
2009; Wood and Rebec
2004). Here this effect was abolished by PCP pre-treatment. These findings further support the conclusion that aspects of mPFC and NAc function, specifically related to memory recognition to novel and salient stimuli, are a central component of PCP-induced cognitive dysfunction.
Methods
Subjects
Twelve female Lister-Hooded rats, obtained from Charles River (Cambridge, UK), weighing between 225 and 250 g on arrival were housed in pairs prior to surgery on a 12-h reversed light–dark cycle (lights on at 1900 h) at an average temperature of 21 °C and humidity of 40–70%. All testing was carried out during the dark phase. Water and food (LabDiet 5LF5, PMI Nutrition Intl, Brentwood, MO) were freely available. The experiments were carried out under institutional ethics approval and appropriate project and personal license authority granted by the UK Home Office under the Animals (Scientific Procedures) Act 1986.
Phencyclidine treatment
One week after arrival, rats received pre-treatment of PCP (Sigma-Aldrich, Gillingham, Dorset, UK; product nr. P3029) (2.0 mg/kg;
n = 6) or saline (
n = 6), in a final volume of 1 ml/kg i.p. twice daily for 7 days. Following PCP or saline treatment, the animals were given a 1-week drug-free period prior to surgery. The PCP dosing regimen was based on previous work demonstrating robust deficits in exploratory and memory paradigms and neurochemical deficits (Grayson et al.
2015; Sood et al.
2011).
Tetrode implantation surgery
One week after PCP treatment, rats were anesthetised with 4% v/v isoflurane (Schering-Plough) in O2, and maintained between 2–3%. Immediately post induction, an injection of glycopyrronium bromide was administered (6–8 µg/kg; i.m.; Anpharm, Warsaw, Poland) to reduce respiratory tract secretions. The animal was mounted in a stereotaxic frame and the head was adjusted so that lambda and bregma were aligned on the same horizontal plane. To prevent corneal desiccation, Lacri-Lube Eye Ointment (Allergan, Westport, Ireland) was applied to the eyes. A homoeothermic heat pad (Harvard Apparatus, Boston, Massachusetts, USA) was used to maintain body temperature between 36 and 37 °C. Glucose (5%, 3 ml/h, s.c.) was given via an infusion pump (Intec, K.D, Scientific, Holliston, Massachusetts, USA) for the duration of the surgery.
A scalp incision was made along the midline, the periosteum was retracted, and 9–10 stainless steel anchoring screws (Morris Co., Southbridge, Massachusetts, USA, part number 0 × 1/8 flat) were affixed to the skull. A left-side craniotomy was then performed above mPFC and NAc shell. Implantation co-ordinates were as follows: +3.2 mm AP; 0.5 mm ML; −2 mm DV for mPFC, targeting prelimbic cortex; and +1.2 mm AP, 1.1 mm ML; −7 mm DV for NAc shell (Paxinos and Watson
1986). The medial of the two tetrodes per structure was targeted at these locations and distance between tetrode tips was minimal (~200 micron). The dura was incised and the tetrode array was advanced into the target structures using a stereotaxic holder. Two tetrodes were implanted in mPFC and two in NAc (Fig.
1a, b). Each tetrode was made of four 12 μm tungsten wires (H-Formvar insulation with Butyral bond coat, California Fine Wire Company, Brover Beach, CA) twisted together and heated to form a bundle. The tip of each wire was gold plated to reduce impedance to 150–400 kΩ. The tetrodes were threaded through a 0.17 mm outer diameter silica tube (SGE Analytical Science; Milton Keynes, UK) to increase stability and loaded into a microdrive (Versadrive, Neuralynx, Bozeman; Montana, USA) that allowed their independent movement. A silver wire inserted into the skull above the cerebellum served as a ground. The tetrodes were sealed with paraffin wax, and the implant was built up using layers of light-curing dental cement (Flowable Composite, Henry Schein; Gillingham, UK). Antibiotic ointment (Fuciderm; Uldum, Denmark) was applied to the wound and the skin was sutured. A non-steroidal anti-inflammatory (Carprieve, 5 mg/kg; S.C; Norbrook Laboratories Ltd; Corby, UK) was administered 2–3 h before recovery and twice a day for 5 days post surgery. An antibiotic (Baytril, 2.5%, 0.2 ml/kg; S.C., Bayer; Leverkusen, Germany) was given immediately after recovery and twice daily for 5 days after surgery. The animals were given a week to recover from the surgery before behavioral testing. Animals were handled daily and remained individually housed for the remainder of the experiment to prevent damage to the implants.
Behavioral testing
PCP-induced cognitive deficits are commonly assessed using the novel object recognition (NOR) paradigm (Grayson et al.
2015). To verify drug manipulation and explore neurophysiological correlates of this behavioral deficit, we carried out recordings during NOR testing (Fig.
3a). Seven days post surgery, the electrodes were slowly advanced approximately 50 μm into the brain on three consecutive days. The tetrodes were advanced a further 50 μm approximately 30 min before each recording session. Electrophysiological recordings were obtained in an open-top black Plexiglas box (52 cm wide × 52 cm long × 40 cm high) placed within a sound-attenuated aluminum-plated chamber. The light intensity on the apparatus floor was maintained at approximately 15 Lux using indirect illumination provided by 8 LEDs evenly spread outside the testing box. The NOR protocol was based on well-established parameters (Grayson et al.
2015) and consisted of the following: After extensive habituation to the test box (20 min per day for three consecutive days), animals first underwent a short pre-exposure trial (3 min). No objects were placed in the arena during pre-exposure. This was followed by a 3-min acquisition trial, and a 3-min retention trial. All trials were separated by a 1-min inter-trial interval during which the animal was placed in a familiar holding cage (20 cm wide × 29 cm long × 38 cm high). During the acquisition trial, each rat was placed in the NOR chamber and exposed to two identical objects: the objects used were small glass jars or label-stripped food cans. On the retention trial, both objects were removed and one was replaced with an identical familiar copy and one with a novel object. The location of the novel object in the retention trial was randomly assigned for each rat. The heights of the objects were comparable (~10 cm) and they were heavy enough (475–500 g) not to be displaced by the animals. This was verified at the end of each session. Neurophysiological recordings were carried out throughout each testing trial.
All experiments were video recorded for subsequent behavioral analysis using a HD WebCam (1050 × 720p). Exploration times of each object in each trial were scored manually using Movie Maker (Microsoft Windows) with an on-screen millisecond stop watch. Video recordings were aligned to the neurophysiological data using a trigger programmed in the software Anymaze (San Diego Instruments, California, USA). Animals were deemed to be exploring the object when the head of animal was facing it within 2 cm or touching it with any part of its body except the tail: turning around or sitting on the object was not considered exploratory behavior.
Electrophysiological recordings and data analyses
Rats were recorded during the three trials of the novel object recognition task (pre-exposure, acquisition, and retention) through a metal coil-wrapped headstage cable. We recorded from both PCP- and saline- pre-treated rats (2–4 animals/day; counterbalanced). Wideband signals were acquired continuously via an op-amp-based headstage amplifier (HST/8o50-G1-GR, 1× gain, Plexon Inc., Dallas, TX, USA), passed through a preamplifier (PBX2, 1000× gain; Plexon Inc., Dallas, TX, USA) and digitized at 40 kHz. For spike sorting, the raw signal was band-pass filtered, 300–3000 Hz, and spikes were sorted using the Matlab-based Wave_clus software to yield single-unit spike trains (Quiroga et al.
2004). Single units were detected by applying a threshold of 5× signal noise. Signal noise was estimated as the median absolute deviation of the band-passed signal (Rey et al.
2015). Spike sorting was achieved with super-paramagnetic clustering using a single parameter (‘temperature’), where in the super-paramagnetic regime, clusters of a relatively large size, corresponding to the different single units, are captured. All automatic detection thresholds and sorting solutions were examined individually and adjusted if needed. In addition to this, we inspected cross-correlograms and autocorrelograms of units obtained on the same wire as well as average cluster waveforms and ISI intervals for violations of a refractory period. PFC–NAc population synchrony was calculated as the average of synchrony values obtained across all neuronal pairs recorded from the two structures. While cross-correlograms are one possible method for analyzing simultaneously recorded spike trains, their use as a measure of spike timing synchronization has been disputed by some authors (Agmon
2012; Brody
1999). Here we applied two independent approaches to calculate synchrony. Firstly, we used a model-free spike synchronization algorithm based on counting the number of quasi-simultaneous occurrences of spikes from two concurrently recorded spike trains. The algorithm is described in detail in Quiroga et al. (
2002) and the Matlab code is available online at:
http://www.old.fi.isc.cnr.it/users/thomas.kreuz/Source-Code/Event-Sync.html. Synchronization values vary from 0 (no synchronization) to 1 (perfect synchronization) (Fig.
2a). The time lag used for detecting quasi-simultaneous occurrences is defined locally to allow for firing rate changes during recording (see Quiroga et al.
2002, Eq. 2):
$${\tau _{ij}} = \min \left\{ {t_{i + 1}^x} \right. - t_i^x,t_i^x - t_{i - 1}^x,t_{j + 1 - }^yt_j^y,t_j^y - \left. {t_{j - 1}^y} \right\}/2.$$
The lag
τ
ij
involves finding the minimum of the distances between a target spike
t
i
and the following one
t
i+1 as well as the target spike t
i and the preceding one
t
i−1 for both spike trains (
X and
Y) in a recorded pair (a total of 4 distances). In our sample, the median lag (across pairs and sessions) was equal to 25 ms. To complement this approach, we also calculated cross- and auto-spectrum-based spike coherence between PFC and NAc pairs (Halliday
2015) (Matlab code available online at
http://www.neurospec.org). Comparisons between neural responses and ongoing object exploration were carried out using cross-correlations between firing rate histograms binned over 100 ms and continuous contact data scored as 0 (no contact) or 1 (for contact; Fig.
3b). The zero time point here refers to the center of the cross-correlogram between the two activity traces (i.e., the cross-correlation calculated at zero lag between the two channels). This approach essentially allowed us to measure the similarity between the two continuous signals (behavior and firing rate; compare Matlab command
xcorr). A similar approach has recently been used successfully to assess the relationship between Purkinje cell activity and ongoing licking behavior in mice (Cao et al.
2012). All analyses were carried out using Neuroexplorer and Matlab (MathWorks, Natick, MA). Analyses of variance (ANOVA) and Wilcoxon signed-rank tests were performed using SPSS (IBM SPSS, Somers, NY, USA).