Introduction
Purposeful goal-directed responses may require an organism to flexibly adapt to changing situational demands by overcoming previously learned associations. This form of learning includes the ability to adjust responding following altered stimulus reward contingencies and is often assessed by reversal learning tests. Schizophrenia is characterised by cognitive deficits that precede and outlast other symptoms and predict long-term outcome [1]. These cognitive deficits include impaired reversal learning [2] observed as behavioural perseveration with patients showing inappropriate repetitive responding following a task contingency shift. Such perseveration may be produced by diverse cognitive impairments although the term is often associated with a potential explanation in terms of inappropriate stability of previous stimulus-reward associations. Available neuroleptics’ failure to treat these deficits severely limits treatment and can contribute to a poor long-term outcome [3].
Altered serotonin (5-Hydroxytryptamine, or 5-HT) signalling has been linked to a range of anatomical and cognitive abnormalities in schizophrenic patients. Lower brain levels of 5-HT in schizophrenic patients correlate with severity of cognitive impairment [4], hypofrontality during attentional set-shifting [2], [5], and poor long-term social and clinical outcome [3], [6]. These serotonergic abnormalities may involve altered signalling at the 5-HT2C receptor (5-HT2CR). Schizophrenic patients show aberrant 5-HT2CR binding in the prefrontal cortex (PFC) [4], [7], decreased 5-HT2CR mRNA in the PFC [8], and altered PFC 5-HT2CR pre-mRNA editing [9].
Serotonin and the 5-HT2CR are also implicated in reversal learning. Acute tryptophan depletion can impair probabilistic reversal learning in healthy human subjects [10] and PFC or orbitofrontal cortex (OFC) specific 5-HT depletion retards visual reversal learning in the marmoset [11]. Similarly, systemic 5-HT depletions retard bowl-digging [12], go/no-go [13] and instrumental probabilistic reversal learning [14] in the rat. Improved allocentric visuospatial reversal learning is also seen in rodents systemically treated with the 5-HT2CR antagonist SB242084 [15], [16] and in 5-HT2CR knock-out (KO) mice [16].
Altered reversal learning performance may be caused by changes in the ability to overcome prior associations of either or both positive and negative valence. A rewarded two-choice discrimination can be reduced to an excitatory conditioned stimulus (CS) – unconditioned stimulus (US) association, eliciting approach, and an inhibitory CS – ‘no US’ association, eliciting withdrawal. Following a contingency shift, the CS initially predicating the US becomes associated with ‘no US’, a process opposed by perseverance. Conversely, the CS initially predicating ‘no US’ now predicts the US, a process opposed by learned non-reward [17].
Although behavioural perseveration defines a range of behaviours related to the excessive maintenance of activities, including inappropriate responding in the context of reversal learning [18], [19], it does not define the valence of the association that is inappropriately maintained. The term perseverance is used here to specify excessive responding towards previously rewarded stimuli in a task that attempts to dissect the underlying cognitive components of behavioural flexibility.
One approach to understanding the relative contributions of perseverance and learned non-reward has been to dissect tasks of cognitive flexibility into separate tests assessing these two processes by pairing a novel CS either with the previously correct CS or with the previously incorrect CS [11], [16], [17], [20], [21]. Here we investigate the role of the 5-HT2CR in reversal learning dissected into perseverance and learned non-reward using a spatial maze procedure. The task used egocentric discriminanda, no exteroceptive cues were provided to accurately guide responding. All testing took place in the dark using a radial-arm maze in multiple T- or Y-configurations in order to reduce the influence of any residual allocentric cues (Fig. 1).
10.1371/journal.pone.0077762.g001 Figure 1  Diagram depicting the four types of discrimination.
Example of the spatial discrimination (A) full reversal test (B), perseverance test (B) and learned non-reward test (C). Other maze arms not shown for clarity.   Egocentric tasks have been used to assess the roles of dopamine (DA) and 5-HT signalling in reversal learning and discussed in relation to schizophrenia [22]–[24]. However, there have rarely been attempts to explore and replicate neuropharmacological manipulations across egocentric and allocentric spatial tasks of reversal learning. This becomes particularly pertinent considering that egocentric and allocentric spatial learning may require different underlying neural systems. For instance, rodent egocentric but not allocentric spatial learning has repeatedly been shown to be dependent upon the integrity of the dorsal striatum [25]–[28].
Experiment 1 assessed the effects of the 5-HT2CR antagonist SB242084 and Experiment 2 compared 5-HT2CR KO and wild type (WT) mice. The test conditions composed full reversal, or reversal in which either the previously incorrect or correct arm was replaced by a novel alternative, thus providing tests of perseverance and learned non-reward, respectively. A further experiment investigated the effect of a novel arm on unrewarded choice behaviour to demonstrate that the changed maze configuration is treated as a novel alternative by mice in this task and also to investigate potential effects of 5-HT2CR antagonism on responses to this novel alternative.