Introduction

Optimal adaptation to the environment is critical for animals’ inclusive fitness and requires the balance of behavioural inhibition and activation (Ghazizadeh et al. 2012; West and Gardner 2013). Behavioural control is highly influenced by motivational states (“impulses”). The active inhibitory mechanism, which modulates such internally or externally driven prepotent desires for reinforcement, is referred to as impulse control (Jentsch and Taylor 1999; Winstanley et al. 2006).

Deficient impulse control leads to maladaptive impulsive behaviours including inability to wait and difficulty withholding responses, generally defined as impulsive action or motor impulsivity (Brunner and Hen 1997; Dalley et al. 2011; de Wit 2009). The dominant behavioural measures of impulse control are response inhibition paradigms, such as the five-choice serial reaction time task (5-CSRTT). During 5-CSRTT performance, rats are required to withhold from premature responding to a visual, reward-predicting stimulus, generally regarded as an index of impulse control (Carli et al. 1983; Muir et al. 1996; Robbins 2002). Impulse control is based on cortico-limbic-striatal circuits, and dysfunctions of these systems are associated with several psychiatric disorders characterised by high levels of impulsivity, like ADHD (Nigg and Casey 2005), obsessive-compulsive disorder (Anticevic et al. 2013), pathological gambling (Fineberg et al. 2010), schizophrenia (Meyer-Lindenberg et al. 2002; Pantelis et al. 1997; Robbins 1990), drug abuse and other forms of addiction (Kalivas and Volkow 2005; Russo and Nestler 2013). There is evidence that frontostriatal connections are part of parallel, functionally segregated re-entrant striato-thalamo-cortical loops. In both primates and rats, frontostriatal projections are topographically organised so that functionally different subregions of the prefrontal cortex (PFC) have separate targets in the striatum (Alexander et al. 1986; Berendse et al. 1992; McGeorge and Faull 1989). The most pronounced anatomical partition of the rodent PFC is made in the medial PFC, which can be divided into dorsal (anterior cingulate and medial precentral cortices) and ventral subdivisions (prelimbic, infralimbic and medial orbital cortices) (Dalley et al. 2004; Gabbott et al. 2005; Heidbreder and Groenewegen 2003; Ongur and Price 2000). The anatomical heterogeneity of the mPFC is paralleled by functional subregional differentiation, with the network of prelimbic and infralimbic cortices, henceforth referred to as ventral medial prefrontal cortex (vmPFC), being more critically involved in impulsive behaviour (Chudasama et al. 2003; Kesner and Churchwell 2011). On the striatal level, the nucleus accumbens (NAc) as part of the ventral striatum and as core element of the mesoaccumbal dopamine (DA) system is generally implicated in reward and motivation and ideally positioned to integrate input signals of executive-cognitive information, such as impulse control, arising from the mPFC (Carlezon and Thomas 2009; Groenewegen and Trimble 2007; Mogenson et al. 1980). The vmPFC of rats is anatomically and functionally interconnected with the NAc, whereas the rodent dorsal mPFC preferentially innervates the dorsomedial striatum (Berendse et al. 1992; Brog et al. 1993; Ding et al. 2001; Gorelova and Yang 1997; McGeorge and Faull 1989; Sesack et al. 1989; Vertes 2004). Converging lines of evidence further indicate a functional relationship between mPFC and NAc in terms of behavioural inhibition, as both regions have found to be involved in inhibitory action control (Aron 2007; Christakou et al. 2004), impulsive decision-making (Costa Dias et al. 2012; Diergaarde et al. 2008), behavioural flexibility (Coppens et al. 2010; Goto and Grace 2005) and drug seeking (Bossert et al. 2012; Peters et al. 2008; Vassoler et al. 2013). Recent findings indicate that impulsive behaviour is not only top-down controlled by cortical areas, but also modulated on the subcortical level (Dalley et al. 2011). For instance, intra-NAc injections of DA receptor antagonists reverse behavioural disinhibition induced by vmPFC inactivation (Ghazizadeh et al. 2012) and block premature responding following mPFC lesions in the 5-CSRTT (Pezze et al. 2009). Findings of electrophysiological recording (Hayton et al. 2011), lesion (Chudasama et al. 2003; Muir et al. 1996; Pezze et al. 2009) and reversible inactivation (Izaki et al. 2007; Murphy et al. 2012; Narayanan et al. 2006; Paine et al. 2011) studies already implicated the rodent mPFC in impulse control. The contribution of the NAc to impulsive behaviours turned out to be even more complex, as the NAc cannot be regarded as an anatomical and functional entity (Groenewegen and Trimble 2007; Heimer 2003). Based on anatomical, neurochemical and electrophysiological criteria, the NAc in the rat brain is divided into distinct subterritories, which are also present in the human brain and show considerably different input-output features: a dorsolateral core region surrounding the anterior commissure and a shell compartment that is located ventromedially to the core (Berendse et al. 1992; Brog et al. 1993; Heidbreder and Groenewegen 2003; Meredith et al. 1996; Sokolowski and Salamone 1998; Zaborszky et al. 1985). The functional dichotomy of the NAc subregions regarding several behaviours (Bassareo et al. 2002; Corbit et al. 2001; Floresco et al. 2006; Floresco et al. 2008; Jongen-Relo et al. 2002; McFarland et al. 2004; Pothuizen et al. 2005a; Robbins and Everitt 1996; Stratford and Kelley 1997; Vassoler et al. 2013) appears also to hold true for impulse control in terms of motor impulsivity. While core lesions induce deficits in 5-CSRTT and differential reinforcement for low rates of responding tasks (DRL) (Christakou et al. 2004; Pothuizen et al. 2005b), lesions of the NAc shell lack a significant influence on anticipatory responding in response inhibition tasks (Murphy et al. 2008; Pothuizen et al. 2005b). In line with this, disconnection of the mPFC from the NAc core by lesions caused impulse control deficits in the 5-CSRTT (Christakou et al. 2004), whereas an involvement of the mPFC-NAc shell connection was not yet examined. However, DA D1 receptors in NAc shell are involved in inhibitory response control in the 5-CSRTT (Pattij et al. 2007).

Asymmetrical disconnection designs have successfully been used to show a functional interaction between mPFC and NAc in a variety of behavioural paradigms, including effort-based decision-making (Hauber and Sommer 2009), Pavlovian conditioning (Parkinson et al. 1999), behavioural flexibility (Block et al. 2007), working memory (Floresco et al. 1999) as well as inhibitory and attentional control (Christakou et al. 2004). Bilateral projections from the vmPFC to the NAc are predominantly ipsilateral (Berendse et al. 1992; Brog et al. 1993; Gabbott et al. 2005; McGeorge and Faull 1989; Montaron et al. 1996; Sesack et al. 1989). Thus, combined unilateral lesioning or reversibly inactivating both structures in opposite hemispheres results in the disruption of the neuronal circuitry in both hemispheres (Gaffan and Wilson 2008). Chemical agents like the GABAA agonist muscimol allow repeated acute and reversible inactivations of distinct brain regions, and hence, within-subject designs accompanied by increased test-retest reliability (Lomber 1999).

Hence, in the present study, we used an asymmetrical disconnection approach combining unilateral temporary inactivations of the vmPFC and the contralateral NAc core or shell by muscimol to investigate the relevance of the vmPFC-NAc connectivity to impulse control in the 5-CSRTT in rats. Interestingly, previous results from our laboratory revealed that transient bilateral inactivation of the vmPFC (Feja and Koch 2014) as well as the NAc shell, but not the core (Feja et al. 2014), by muscimol induced impulsive over-responding in the 5-CSRTT. The present study aimed to extend these findings hypothesising, firstly, that the vmPFC and NAc functionally interact in motor impulsivity; secondly, that specific frontostriatal connections differentially affect 5-CSRTT performance and, thirdly, that motor impulse control preferentially requires an intact vmPFC-NAc shell circuitry.

Methods

Subjects

A total of 22 adult male Lister hooded rats (Harlan, Venray, Netherlands) weighing 260–340 g at the beginning of the experiments were assigned to two testing cohorts, defined as vmPFC-NAc core (n = 12) and vmPFC-NAc shell (n = 10) groups. The animals were housed in groups of four to six in standard Macrolon cages (type IV) under controlled ambient conditions (21–22 °C, 45–55 % humidity, 12-h light/dark cycle, lights on at 7:00 a.m.). The rats were kept on their experimental body weights of approximately 85 % of those under free feeding conditions by controlled feeding of laboratory rodent chow (Altromin Standard Diet 1324 19 % crude protein, 4.0 % crude fat, 6.0 % crude fibre, 7.0 % crude ash; Altromin Spezialfutter GmbH & Co. KG, Lage, Germany) and received tap water ad libitum. Behavioural testing took place between 8:00 a.m. and 6:00 p.m. The experiments were performed in accordance with the National Institutes of Health ethical guidelines for the care and use of laboratory animals for experiments and were approved by the local animal care committee (Senatorische Behörde, Bremen, Germany).

Apparatus

The 5-CSRTT was conducted in two operant aluminium chambers (26 × 26 × 26 cm; Campden Instruments Ltd., Loughborough, UK), wherein five apertures (2.5 × 2.5 cm, 4 cm deep) were inserted 2 cm above floor level in the concavely curved rear wall. This assembly provided five response options located equidistant to the food magazine on the opposite. Inside each hole, a light-emitting diode (LED) generated visual stimuli of variable duration. Nose-poke responses of the animals were detected by infrared photocell beams at the entrance of the apertures. The rats could be placed in the box through a Plexiglas® door on the upper part of the front wall. Underneath the door, a small Plexiglas® panel provided access to the food magazine which was lighted via two LEDs and automatically supplied with casein pellets (45 mg Dustless Precision Pellets, Bio-Serv®, UK) by an electromechanical feeder. Food collection was detected by a microswitch monitoring the movement of the hinged panel. Each chamber was illuminated by a 3-W house light mounted on the ceiling. A noise-damped fan served as ventilation and background noise of about 60 dB. The grid floor facilitated the removal of excrements. For the purpose of sound attenuation, the wooden cabinet was reinforced with an insulating plate at the interior of the door. The apparatus was controlled by specific software written in Turandot (Cambridge Cognition Ltd., version 1.23) which was run on a personal computer connected to the BNC Mark 2 System (Behavioral Net Controller, Campden Instruments Ltd., Loughborough, UK).

General procedure

Training

The animals (n = 22) were trained to detect the occurrence of brief light stimuli in one of the five rear wall apertures. The general procedure was based on the protocol of Campden Instruments and was divided into a habituation, pretraining and baseline training phase (Campden Instruments 2005). Before training and tests, the rats were acclimatised to the laboratory for at least 30 min in their home cages.

The first experimental phase comprised two daily half-hour habituation sessions. The boxes were prepared as follows: before the first training session, the tray panel was opened to facilitate access to 15 freely available pellets in order to reinforce the magazine as location of food reward. During the second session, no panel manipulation was carried out. Besides the reward in the tray, two pellets were placed in each aperture to promote exploration of these areas. The chambers were permanently illuminated by the house light during both sessions.

The daily training session lasted 30 min or was finished after completion of 100 trials. Each session started with the simultaneous illumination of the box and the food magazine and the delivery of a single pellet into the tray. Once the rat opened the panel for food retrieval, the first trial was initiated. The magazine light faded, and a fixed intertrial interval (ITI) of 5 s started. At the end of the ITI, a light stimulus of determinate duration (stimulus duration (SD)) was randomly presented in one of the five holes. The rats had to respond with a nose poke into the appropriate aperture during the stimulus presentation or within a subsequent limited hold period (LH). A correct response was followed by the supply of a pellet into the lighted food magazine. The next trial was triggered by the panel movement. Inappropriate responses led to a punishment in terms of a predefined 5-s period of darkness (time out) without reward delivery. The task procedure offered various opportunities for such reactions:

  • Incorrect responses in a hole where no stimulus appeared,

  • Omissions in the form of absent reaction to the occurrence of the stimulus within the LH,

  • Premature responses before the onset of the stimulus during the ITI in one of the apertures and

  • Perseverative responses, meaning additional responses after a correct response and before reward collection.

Every nose-poke response in one of the apertures during the time-out phase (time-out responses) prolonged the period of darkness. Following the time-out, the box and the tray were illuminated again and the next trial was started by a nose poke into the food magazine. Within a session, the visual stimuli were randomly presented in equal number in each hole. The progressive decrement of the variables SD (60 → 1 s) and LH (60 → 5 s) over eight training levels enabled the acquisition of the 5-CSRTT.

The baseline training session was determined by the conditions of the eighth training level (SD = 1 s, LH = 5 s). After showing a stable baseline performance (>80 % accuracy and <20 % omissions with <10 % variation over five consecutive training sessions), rats underwent surgery.

Surgery

The rats were anaesthetised with chloral hydrate (360 mg/kg; Sigma-Aldrich, Steinheim, Germany) and fixed in a stereotaxic frame (David Kopf Instruments, Tujunga, CA, USA). Stainless steel 21-gauge guide cannulae were implanted unilaterally 1 mm above the target injection site into the vmPFC (anteroposterior +2.7 mm, mediolateral ±0.8 mm, dorsoventral −4.0 mm from bregma) and 2 mm above the intended injection sites into the contralateral NAc core (anteroposterior +1.2 mm, mediolateral ±1.8 mm, dorsoventral −6.8 mm from bregma) or shell (anteroposterior +1.2 mm, mediolateral ±0.5 mm, dorsoventral −7.3 mm from bregma). The sides of the implantations were counterbalanced, resulting in approximately equal numbers of rats with microinfusions in the left or right hemispheres at the level of vmPFC and NAc. Jeweller screws were anchored in the skull serving to fix the cannulae which were embedded in dental cement and closed by removable 26-gauge stylets of the same length. After surgery, the rats were kept individually for 3 days with free access to food and water.

Following a total recovery period of 5 days, the animals were reintroduced to the baseline training until they re-established the presurgical baseline performance.

Microinfusion

The test design comprised four 4-day sessions for the animals. Each session started with an injection day, followed by a day without training. The second and third post-testing days were used to achieve the baseline performance and to ensure the washout process of the drug. Before infusion, the stylets were exchanged for injection cannulae (vmPFC 27 gauge; NAc 26 gauge) connected with microlitre syringes (SGE Scientific Glass Engineering, Darmstadt, Germany) via polyethylene tubes. The rats received four sets of combined unilateral microinjections of the GABAA agonist muscimol (0.05 μg/0.3 μl) and 0.9 % saline as vehicle (0.3 μl) into the vmPFC and the contralateral NAc core or shell according to a pseudorandom Latin square design. The subject groups were divided as follows:

  • Disconnection group I (vmPFC + NAc core; n = 12): vehicle + vehicle, vehicle + muscimol, muscimol + vehicle and muscimol + muscimol.

  • Disconnection group II (vmPFC + NAc shell; n = 10): vehicle + vehicle, vehicle + muscimol, muscimol + vehicle and muscimol + muscimol.

The injection rate was 0.1 μl/30 s. The injectors were left in place for 1 min to guarantee diffusion and to avoid reflux of the solution. Ten minutes after the microinjection, the rats underwent behavioural testing. The sequence of the test sessions matched with the baseline training.

Drugs

The GABAA agonist muscimol was purchased from Tocris Bioscience (Ellisville, MO, USA) and dissolved in 0.9 % saline. Aliquots of stock solutions (0.5 μg/0.3 μl) were prepared and stored at −20 °C until use. On the treatment day, aliquots were further diluted to a dose of 0.05 μg/0.3 μl. Doses were based on previous studies (Diederich and Koch 2005).

Histology

Upon termination of the experiment, the rats were euthanised with a lethal dose of chloral hydrate. The brains were removed from the skull and immersion fixed in a 4 % formalin/30 % sucrose solution for 48 h. Coronal 50-μm sections of the mPFC were cut on a cryostat (Jung CM 3000; Leica Instrument GmbH, Nussloch, Germany), mounted on gelatine-coated glass slides and Nissl stained with thionin. Then, the sections were analysed using a light microscope and injection sites plotted on standardised coronal sections of a rat brain stereotaxic atlas (Paxinos and Watson 1998).

Data analysis

The descriptive statistics is based on means and variance and is indicated by the standard error of the mean (±SEM). The statistical analyses were conducted by the software IBM SPSS Statistics (version 20 for Windows).

The drug effects within the testing group on the following behavioural parameters were investigated using separate two-way split-plot-factorial analysis of variance (ANOVA; within-subject factor: drug treatment, between-subject factor: disconnection group): percentage of correct responses (accuracy; 100 × number of correct responses/number of correct and incorrect responses), percentage of omitted responses (100 × number of omitted responses/total number of correct, incorrect and omitted responses), number of premature responses, number of perseverative responses, number of trials completed, number of time-out responses, latency of correct response [s] and latency of reward collection [s].

In the case of significant main effects (P < 0.05), one-way repeated measures ANOVA and post hoc Bonferroni tests for the factor drug treatment as well as independent t tests between the disconnection groups were conducted separately for each behavioural parameter.

Results

Histology

In total, 22 rats received unilateral microinjections into the vmPFC combined with contralateral microinfusions into NAc core (n = 12) or shell (n = 10). The histological analysis revealed, as indicated in Fig. 1, that 18 rats (n = 9 in each group) had acceptable injection sites accurately located in the target structures with minimal tissue damage.

Fig. 1
figure 1

Location of the unilateral injection sites in a the ventral medial prefrontal cortex (circles corresponding to nucleus accumbens core, n = 9; triangles corresponding to nucleus accumbens shell, n = 9) and in b the contralateral nucleus accumbens core (circles, n = 9) or shell (triangles, n = 9) in the 5-choice serial reaction time task depicted on schematic drawings from the rat brain atlas of Paxinos and Watson (1998). Rostral distance (in mm) to bregma is indicated by numbers

Full size image

Effects of inactivation of vmPFC-NAc core and vmPFC-NAc shell connections by muscimol on rats’ performance in the 5-CSRTT

Before testing, the rats performed at a stable baseline with high levels of accuracy (disconnection group I 91.64 ± 1.06 %; disconnection group II 92.19 ± 0.92 %), fast correct response (disconnection group I 0.69 ± 0.01 s; disconnection group II 0.68 ± 0.02 s) and reward collection latencies (disconnection group I 1.10 ± 0.05 s; disconnection group II 1.05 ± 0.03 s), low percentages of omissions (disconnection group I 12.68 ± 1.02 %; disconnection group II 9.72 ± 1.25 %) as well as low numbers of premature (disconnection group I 8.10 ± 0.90; disconnection group II 8.28 ± 1.05) and perseverative responses (disconnection group I 1.98 ± 0.35; disconnection group II 2.00 ± 0.54). Analysis of the training data demonstrated no significant differences in the pre- and post-operative sessions and the ‘drug-free days’ between testing excluding any carry-over effects of drug treatment or surgery (data not shown).

Two-way split-plot-factorial ANOVAs on the 5-CSRTT performance showed main effects of drug treatment [F (3,51) = 6.119; P = 0.001] and disconnection group [F (1,51) = 5.71; P = 0.03] as well as a statistically significant treatment × disconnection group interaction [F (3,51) = 7.704; P < 0.001] for premature responses, suggesting impaired motor impulse control. The functional heterogeneity between both frontostriatal systems (vmPFC + NAc core vs. vmPFC + NAc shell) was further substantiated by main effects of disconnection group for omissions [F (1,51) = 8.228; P = 0.011], completed trials [F (1,51) = 4.754; P = 0.045] and accuracy [F (1,51) = 5.259; P = 0.036] as well as by a statistically significant treatment × disconnection group interaction [F (3,51) = 3.223; P = 0.031] for accuracy.

Further, one-way repeated measure ANOVAs and post hoc Bonferroni tests revealed that simultaneous unilateral inactivation of vmPFC and the contralateral NAc shell specifically enhanced impulsive behaviour reflected by a significant increase in premature responding compared to vehicle (P = 0.042), while no other measured parameter was affected (Fig. 2 and Table 1). Unilateral intra-NAc shell injection of muscimol as well as combined deactivation of vmPFC and NAc shell appeared to augment time-out responses, but this effect did not reach statistical significance (Fig. 2b). By contrast, neither unilateral NAc core nor combined vmPFC and NAc core inactivation had any effect on 5-CSRTT performance. Independent t tests between disconnection groups showed significant differences between the vmPFC-NAc core and vmPFC-NAc shell connection following combined muscimol injection regarding premature responses (P = 0.008) and accuracy (P =0.006) (Fig. 2a, c), substantiating the differential impact of the two distinct frontostriatal circuits on impulse and attentional control. Further, unilateral inactivation of NAc core significantly increased the omission rate compared to NAc shell (P = 0.021) (Fig. 2d).

Fig. 2
figure 2

Effects of combined local unilateral infusions of the GABAA agonist muscimol (M; 0.05 μg/0.3 μl) and 0.9 % saline as vehicle (V) into the ventral medial prefrontal cortex (vmPFC) and the contralateral nucleus accumbens (NAc) core (n = 9) or shell (n = 9) on the rats’ performance in the 5-choice serial reaction time task. V + M signifies vehicle injection into the vmPFC + muscimol injection into the NAc core/shell; M + V signifies muscimol injection into the vmPFC + vehicle injection into the NAc core/shell. Data of a premature responses, b time-out responses, c accuracy and d omissions are means ± SEM. Statistically significant differences between drug treatment compared to vehicle are indicated by asterisk (one-way repeated measures ANOVA, post hoc Bonferroni test, P < 0.05) and between disconnection groups (vmPFC + NAc core compared to vmPFC + NAc shell) by circles (independent t test, P < 0.05)

Full size image
Table 1 Effects of combined local unilateral infusions of the GABAA agonist muscimol (0.05 μg/0.3 μl) and 0.9 % saline as vehicle (V) into the ventral medial prefrontal cortex and the contralateral nucleus accumbens (NAc) core (n = 9) or shell (n = 9) on the rats’ performance in the 5-choice serial reaction time task
Full size table

Discussion

In terms of impulsivity, the vmPFC is considered to be primarily implicated in impulsive action while there is only limited evidence for an involvement of the NAc, which is more associated with impulsive decision-making. Previous studies revealed that NAc shell lesions showed no effect on premature responding and core lesions merely tended to increase motor impulsivity in the 5-CSRTT. However, the present data show that both the vmPFC and the NAc are involved in the neural network mediating impulse control in the 5-CSRTT in rats, with a predominant role for the connection of vmPFC and NAc shell. In contrast, the vmPFC-NAc core connection appears to be more involved in attentional and motivational aspects of behaviour.

The main findings of this study are that acute disconnection of the vmPFC and NAc shell by simultaneous contralateral inactivation by the GABAA agonist muscimol considerably enhanced premature responding indicating deficits in impulse control. In contrast, transient disruption of the serial communication between vmPFC and NAc core did not affect impulsive action. Lesion studies have already documented the involvement of the rodent mPFC and NAc in inhibitory response control and revealed differential contributions regarding different aspects of inhibitory control and specific subregions of the mPFC and NAc (Christakou et al. 2004; Chudasama and Muir 2001; Chudasama et al. 2003; Muir et al. 1996; Murphy et al. 2008; Pezze et al. 2009; Pothuizen et al. 2005b). By use of the GABAA agonist muscimol, we and other groups have recently shown that the vmPFC, including the prelimbic (PL) and infralimbic (IL) cortices, is critically involved in controlling premature responding in the 5-CSRTT in rats (Feja and Koch 2014; Murphy et al. 2012; Paine et al. 2011). On the subcortical level of the NAc, lesions of the core but not the shell region increased anticipatory responding in response inhibition tasks (Christakou et al. 2004; Murphy et al. 2008; Pothuizen et al. 2005b). Coherently, disconnection lesions of the vmPFC and the NAc core enhanced premature and perseverative responding in the 5-CSRTT, whereas the vmPFC-NAc shell connection was not investigated (Christakou et al. 2004). However, the latest work from our laboratory highlighted the role of the NAc shell in terms of motor impulsivity and revealed for the first time that transient deactivation of the shell, but not the core, reduced impulse control in the 5-CSRTT in rats (Feja et al. 2014). The present study verifies our previous findings and confirms that in particular, the connection of vmPFC and NAc shell is implicated in the maintenance of impulse control during 5-CSRTT performance.

Asymmetrical inactivation of vmPFC and NAc shell also increased the number of time-out responses, although not reaching statistical significance. Time-out responses represent another aspect of inhibitory control, more related to cognitive flexibility (Robbins 2002). The increase of time-out responses substantiates the role of the vmPFC-NAc shell connection in behavioural inhibition. This is further supported by previous findings from our laboratory showing that bilateral injection of muscimol into the vmPFC or the NAc shell increased time-out responses in the 5-CSRTT (Feja and Koch 2014; Feja et al. 2014). Cognitive constructs such as impulsivity and behavioural flexibility are closely interrelated executive processes in the context of inhibitory control, hierarchically top-down, mediated by the PFC (Bari and Robbins 2013; Wise 2008). In this regard, vmPFC lesions or inactivations result in behavioural inflexibility in reversal learning tasks in rats (Kosaki and Watanabe 2012; Ragozzino et al. 1999; Ragozzino 2007). The neural network contributing to behavioural flexibility involves both the mPFC and NAc (Coppens et al. 2010). Set-shifting tasks indicated that the mPFC projection to the NAc is important for suppressing inappropriate responses, and asymmetrical inactivation of these structures impaired the ability to switch from one discrimination strategy to another (Block et al. 2007). Interestingly, the shell region apparently had a greater impact on the number of time-out responses than the vmPFC, as revealed by unilateral deactivations of the respective structure. Admittedly, inactivation of NAc shell, in contrast to core, does not impair performance in a set-shifting task in rats, but it was pointed out that the shell mediates the suppression of irrelevant or no-reward behaviours (Blaiss and Janak 2009; Floresco et al. 2006; Floresco et al. 2008). Thus, unilateral inactivation of NAc shell might have contributed to behavioural disinhibition during 5-CSRTT performance.

Other parameters indexing attentional (omissions), compulsive (perseverative responses), motor (correct response latency) or motivational behaviours (trials completed, reward collection latency) remained unaffected following unilateral intra-vmPFC and intra-shell or combined vmPFC and NAc shell infusions of muscimol.

Taken together, the present behavioural effects on 5-CSRTT performance induced by vmPFC-NAc shell disconnection closely resemble the deficits observed following bilateral vmPFC (Feja and Koch 2014) or NAc shell (Feja et al. 2014) inactivation in the same task, while unilateral control deactivations by muscimol of the respective regions alone did not produce significant deficits. The asymmetrical manipulation method used in this study is particularly suited to investigate the interaction between components of cortico-subcortical networks (Gaffan and Wilson 2008; Peters et al. 2008). Since neuronal projections, such as frontostriatal connections from the mPFC to the NAc, are predominantly ipsilateral (Berendse et al. 1992), learned behaviours can be preserved by an intact single hemisphere and unilateral manipulations, as in our study, often lead to minor or no cognitive impairments. Via crossed unilateral inactivation of the vmPFC and NAc core or shell, the serial communication between these structures can be bilaterally impeded (Gaffan et al. 1993; Gaffan and Wilson 2008; Setlow et al. 2002). For example, a previous study showed that disconnection of the IL and NAc shell reinstates cocaine seeking in rats after extinction learning, whereas unilateral inactivation of either IL or NAc shell does not alter seeking behaviour (Peters et al. 2008). Consequently, as the effects of the vmPFC-NAc shell disconnection on premature responding in the 5-CSRTT are more pronounced than the additive effect of the single unilateral inactivations, our findings provide evidence that the control of motor impulsivity requires serial information transfer between this specific frontostriatal system.

Unexpectedly, the transient disconnection of vmPFC and NAc core as well as unilateral manipulations of vmPFC or the core region did not produce any significant behavioural effect in the 5-CSRTT compared to control treatment.

Contralateral inactivation of vmPFC and NAc core tended to increase the omission rate as well as the reward collection and correct response latencies indicating marginal attentional and locomotor deficits and a slightly reduced motivation for food. Moreover, unilateral deactivation of the core significantly augmented the omission rate compared to the respective manipulation of the shell, which might indicate an impact of NAc core on attention. However, as mentioned above, the attentional aspects of the 5-CSRTT performance (accuracy, omissions) were not significantly altered by muscimol treatment compared to control.

Previously, we have shown that the core region in contrast to the shell plays an important role in the regulation of locomotion and general responsiveness with a bilaterally inactivated core severely impaired 5-CSRTT performance (Feja et al. 2014). Particularly, the strong decrease in the number of completed trials after deactivation of NAc core, but not shell, represents a consequence of motivational dysfunction and points towards a differential role of both subregions in motivated behaviour in the 5-CSRTT. This is supported by evidence that muscimol injections into the core reduce breakpoint in a progressive ratio schedule in rats (Moscarello et al. 2010), while shell inactivation enhances motivational behaviour in that task (Stratford and Wirtshafter 2012; Wirtshafter and Stratford 2010). However, lesions of the core do not reduce food motivation in a delayed reinforcement task (Cardinal and Cheung 2005) and muscimol does not affect food intake when injected into the NAc core (Stratford and Kelley 1997) and even increases eating behaviour following infusion into the shell (Basso and Kelley 1999; Lopes et al. 2007; Reynolds and Berridge 2002; Soderpalm and Berridge 2000; Stratford and Kelley 1997; Stratford and Wirtshafter 2011).

High scores of impulsivity in the 5-CSRTT inversely correlate with attentional accuracy (Blondeau and Dellu-Hagedorn 2007; Dalley et al. 2008; Puumala and Sirvio 1998). Considering the central role of frontostriatal impairments to the pathophysiology of ADHD, incorporating attentional and impulsive dysfunctions (Nigg and Casey 2005), it seems obvious that this relationship could also be valid for the vmPFC-NAc shell connection, as simultaneous inactivation of vmPFC and NAc shell produced a significant decrease of response accuracy compared to vmPFC-NAc core disconnection. But since the effect of the vmPFC-NAc shell disconnection on accuracy did not differ from control treatment, we suppose the impaired accuracy should rather be considered a consequence of rash-spontaneous impulsive behaviour of the rats leading to some kind of ‘careless mistake’.

Meanwhile, there is a scientific consensus that impulsive behaviour is not only cortically top-down controlled but also regulated by subcortical areas (Dalley et al. 2011). Most interestingly, motor impulse control seems to be more depending on an intact NAc shell than on the vmPFC, as bilateral inactivation of the shell enhances premature responding at almost the same rate as the vmPFC-NAc shell disconnection, while bilateral deactivation of the vmPFC only produces approximately half the number of anticipatory responses (Feja and Koch 2014; Feja et al. 2014). Accordingly, we hypothesise that the NAc, particularly the shell region, might function as kind of a bottleneck for impulse control, receiving serial parallel information input from the vmPFC, integrating these input signals of impulse control with emotional (basolateral amygdala), contextual (hippocampus) and arousal contents (midline thalamus) and conveying the multiplexed information to downstream brain sites involved in feeding and drinking (lateral hypothalamus), motivation (ventral tegmental area, substantia nigra) and locomotion (caudal mesencephalon) (Carlezon and Thomas 2009; Groenewegen and Trimble 2007; Mogenson et al. 1980).

Although it may be difficult to directly compare the gained knowledge with deficits following human cortical damage or with findings from animal lesion studies, the use of reversible inactivation techniques is an effective analytical tool in the area of basic biological and pharmacological research, especially for dissecting the implications of distinct neuroanatomical structures or systems in specific brain functions. One limitation of temporary inactivation is the spatial localisation of drug effects. In our study, we could not precisely define the degree of inactivation, as we did not measure the spread of muscimol. However, previous autoradiography studies estimated the spread of muscimol in rat brains and showed diffusion of radioactive muscimol restricted to NAc core or shell following injection of similar volumes and even higher concentrations compared to ours (Martin 1991; Martin and Ghez 1999; Pothuizen et al. 2005a). These pieces of evidence suggest that muscimol diffusion in our experiments was restricted to either core or shell. Our behavioural data further indicated region specificity of injections, as they revealed clear and distinct differences between the vmPFC-NAc core and vmPFC-NAc shell groups during 5-CSRTT performance.

However, it cannot be excluded that the injections may have involved adjacent non-accumbal or non-vmPFC areas, such as parts of the ventral pallidum, the dorsal striatum or the anterior cingulate cortex (AC), especially since infusions of fluorophore-conjugated muscimol in our previous study showed an asymmetrical diffusion along the dorsoventral axis up the cannula shaft (Feja et al. 2014). In line with this, a previous study has shown that the disconnection of the AC from the NAc core produces no deficits on the 5-CSRTT performance (Christakou et al. 2001).

This suggests that the lacking impact of the vmPFC-NAc core disconnection on motor impulsivity might be attributable to a concomitantly affected AC. Regarding the vmPFC-NAc shell connection, the interaction with other structures involved in impulse control, such as the mediodorsal nucleus of the thalamus (MD), should be taken into account. Within the NAc, the shell projects predominantly to the MD and lesions of this structure have been shown to increase premature responding in the 5-CSRTT (Chudasama and Muir 2001; Groenewegen et al. 1999). Besides, the vmPFC-NAc shell disconnection might have induced an increase of DA levels in the NAc shell resulting in deficient impulse control. This is supported by a previous study showing that vmPFC inactivation results in the disinhibition of phasic excitations at the level of the NAc shell that can thereby be driven by dopaminergic input from the ventral tegmental area (VTA) promoting behavioural cue responding (Ghazizadeh et al. 2012). An increase in extracellular DA levels in the shell might also occur in consequence of inactivation of this region due to its feedback loop involving the VTA. In normal conditions, terminal DA release in the NAc is tonically inhibited via GABAA receptors in the VTA (Ikemoto et al. 1997; Rahman and McBride 2002).

Activating GABAA receptors in the shell by muscimol may hyperpolarise the MSN projecting to the VTA leading to disinhibition of DA neurons targeting the NAc shell. Consistently, blockade of GABAA receptors within the VTA increases the discharge rate of DA neurons innervating the NAc (Ikemoto et al. 1997).

In conclusion, our results extend previous findings pointing out the functional heterogeneity of frontostriatal systems and show a differential contribution of the vmPFC-NAc connection to behavioural control depending on the involved accumbal subregion. We suggest that the maintenance and regulation of motor impulse control particularly requires an intact connection between the vmPFC and the NAc shell, while the vmPFC-NAc core projection seems to be of minor importance.