Discussion
Here we have investigated the involvement of the 5-HT2CR on egocentric discrimination and reversal learning using a T/Y maze-based task. Separable effects on perseverance and learned non-reward were revealed when the reversal learning task was dissected into its constituent cognitive components. Acute pharmacological antagonism of 5-HT2CR function attenuated perseverative responding but impaired responding to previously non-rewarded choices and, perhaps as a consequence, there was no effect on the full reversal task (Fig. 2A, B). However, genetic inactivation of these receptors in 5-HT2CR KO mice had opposite effects to those of acute 5-HT2CR antagonism and these mice also showed impaired egocentric discrimination learning in the initial phase of the maze task (Table 1). When 5-HT2CR KO mice were subsequently challenged with contingency shifts, they showed a selective increase in perseverative responding. There was no significant effect of genetic inactivation of the 5-HT2CR on either the learned non-reward or the full reversal task (Fig. 2C, D).
Interestingly, these results contrast with those from visuospatial instrumental assays in both rats [15], [30] and mice [16] suggesting that egocentric and visuospatial assays may depend upon different underlying neural systems. The inconsistent effects of genetic inactivation and acute antagonism on perseverative responding may have a parallel in the finding that 5-HT2CR antagonism has relatively small effects on nigrostriatal dopamine systems compared to those on the mesolimbic dopamine system [31] whereas the effects on nigrostriatal dopamine system function in 5-HT2CR KO mice are very much greater [32].
Also, both 5-HT2CR KO and SB242084 failed to affect early and late errors to criterion. In full reversal learning, early errors are often assumed to reflect the stability of the CS-US association, or perseverance, while late errors are assumed to be a measure of general cognitive abilities related to attention and the acquisition of an alternative CS-US association [33]. However, analyses of early and late errors are fundamentally different from the experimental manipulation of reversal learning currently used. As previously correct as well as incorrect CSs are presented in both early and late phases of full reversal learning, both associations may influence choice behaviour in both phases of learning.

Generalisation and Novelty in the Egocentric Reversal Learning Task
In some variants of the task used here, animals were challenged with choice of a previously experienced response-arm and a novel response-arm. Animals may have generalised between the 90° and 45° turns in the same direction that were used to generate novel alternatives, and such generalisation would result in the perseverance and learned non-reward tests resembling tests of full reversal learning. However the test condition-dependent effects of SB242084-treated and 5-HT2CR KO mice, as well as the significantly increased number of trials required and larger number of incorrect responses made in the full reversal test than the learned non-reward and perseverance tests of the SB242084 experiment, suggest that animals perceived a 45° shift in arm location as novel. In Experiment 3, which relied on measures of unconditioned exploratory behaviour, a 45° shift in arm location led to significant increases in time spent and entries made into that arm, also suggesting it was recognised as novel.
Treatment-dependent changes in response to novelty may also affect performance in this task. The novel response arm is correct in the perseverance test but incorrect in the learned non-reward test. Increased novelty attraction would therefore facilitate learning in the perseverance test where the novel arm is correct, and retard learning in the learned non-reward test, where the novel arm is incorrect. Thus, one potential explanation for the pattern of results in the SB242084 experiment is that 5-HT2CR antagonism enhances responding for a novel choice in the maze. However, we are unaware of prior evidence for a role of the 5-HT2CR in novelty attraction or novelty recognition. SB242084 also failed to affect performance in the novelty recognition test (Experiment 3), suggesting the observed effects on cognitive flexibility (Experiment 1) instead are related to the ability to overcome previously learned contingencies of reward and non-reward.
Although the discrepant effects in visuospatial and egocentric tasks are most likely due to the tasks tapping different brain-regions and subpopulations of the 5-HT2CR, there are substantial differences in the two types of tasks. Additional to the use of different discriminanda, the current egocentric task involves perseverance to a greater extent than visuospatial reversal learning in the mouse [16]. The current protocol also involves less discrimination training than a visuospatial protocol in the rat [15], [30].

Acute 5-HT2CR Antagonism and Egocentric Reversal Learning
Acute 5-HT2CR antagonism by SB242084 facilitated the ability to overcome perseverance, observed as a decrease in trials and incorrect responses to criterion. SB242084 also caused a concurrent impairment in the ability to overcome learned non-reward by increasing the number of trials to criterion. These opposing effects appear to have summated in the full reversal task, leading to no overall effect. It is very likely that these effects reflect 5-HT2CR blockade, rather than effects on another receptor mechanism. SB242084, especially at the relatively low dose used here, is highly selective for the 5-HT2CR, acting as a full antagonist or inverse agonist [34].
The observed effect of SB242084 in the perseverance test is in agreement with previous studies indicating that acute 5-HT2CR antagonism attenuates perseverative responding [15], [30]. It has been suggested that the SB242084-induced facilitation of operant lever reversal learning in the rat is related to decreased perseverance, as systemic or OFC-specific infusions of SB242084 can decrease repetitive responding towards the previous CS+ [30] or incorrect responses made early in reversal when responding is biased towards the previous CS+ [15]. In the learned non-reward test SB242084 impaired performance, seen as an increase in trials to criterion, in contrast to the effect seen in the perseverance test. This effect differs from the facilitating effects of SB242084 on learned non-reward in an instrumental analogue of the current protocol [16], indicating that egocentric and visuospatial reversal learning may involve different neural mechanisms.
Although little is known about the pharmacology of learned non-reward, work has been done in the closely related paradigm of latent inhibition, which like learned non-reward, could be thought of as the persistence of non-reinforced associations. In this paradigm, SSRIs and atypical antipsychotics can both elevate and attenuate latent inhibition depending on the number of pre-exposures and strength of the non-reinforced association [35], [36]. Interestingly, these two classes of compound do have 5-HT2CR antagonism as a common pharmacological property in addition to their other quite disparate effects [37], [38].
The effects of 5-HT2CR antagonism on visuospatial reversal learning have previously been discussed in relation to altered 5-HT and dopamine signalling [15], [39]. The 5-HT2CR receptor tonically inhibits dopamine (and noradrenaline) signalling in the PFC [40] and dorsal [32] and ventral striatum [41], However, this same group of studies show that the effects of 5-HT2CR antagonists on tonic serotonin signalling is much less evident. The implication may be that it is phasic release of serotonin that is responsible for the effects of these antagonists in the behavioural context of reversal learning [42].

Genetic Inactivation of the 5-HT2CR and Egocentric Reversal Learning
5-HT2CR KO mice showed impaired egocentric spatial discrimination learning and, contrary to the effect of SB242084, they showed selective deficits in the subsequent perseverance test, observed as increased attrition rates, trials to criterion and incorrect responses to criterion, that could not be accounted for by the initial learning deficit. A recent study also reported opposing effects of SB242084 and constitutive loss of the 5-HT2CR in the 5-choice serial reaction time task [43].
Targeted mutations causing constitutive loss of specific components in 5-HT systems often cause adaptations additional to the mutation, leading to behavioural effects which differ from those of acute pharmacological blockade [44]. For example, the 5-HT2CR KO mutant show markedly elevated levels of dialysate DA in the dorsal striatum [32] while pharmacological inactivation is without effect on DA levels in this area [31], [45], [46]. Importantly, it has been speculated that perseverative responding can be produced by dorsal striatal DA elevations [47]. In a probabilistic reversal task, dopamine-agonist treated Parkinson patients show impaired performance compared to unmedicated patients [48] and increased dopamine activity at the D2R and D3R in the caudate nucleus, observed as an increase in methylphenidate induced [11C]-raclopride displacement in human volunteers, correlates negatively with reversal performance [49]. The selective increase in perseverative responding following genetic but not pharmacological inactivation could therefore be explained by the selective increase in dorsal striatal DA levels in the 5-HT2CR KO mouse.
Moreover, rodent egocentric spatial learning has repeatedly been shown to selectively depend upon the integrity of the dorsal striatum. For example, lesioning or local inactivation of the dorsal striatum impairs egocentric spatial but not allocentric visuospatial discrimination learning [25], [50] and working memory [26], [28] and dorsal striatal inactivation also impairs egocentric reversal learning [51]. Thus, the discrepant effects of 5-HT2CR inactivation across visuospatial [15], [16] and egocentric tasks could be explained by a greater involvement of the dorsal striatum in egocentric relative to allocentric spatial learning.
Alternatively, it is possible that both the impaired discrimination learning and perseverative responding seen in 5-HT2CR KO mice could be explained by altered functioning within the hippocampus. Aberrant spatial learning has previously been observed in 5-HT2CR KO mice using a water maze task [52]. Within the perforant path of the dentate gyrus, LTP-formation is suppressed both in the 5-HT2CR KO mouse [52] and by intraventricular 5,7-dihydroxytryptamine induced 5-HT depletions [53]. Since LTP-formation within the perforant path of the dentate gyrus correlate with spatial learning in the water maze [54] and blocking LTP-formation in the medial perforant path retards water maze performance [55], the observed retardation of discrimination learning could be related to the suppressed hippocampal LTP. However, there is no direct evidence to confirm that acute modulation of hippocampal 5-HT2CR function modulates either egocentric learning or the regulation of hippocampal LTP.
It should be recognised that the dichotomy that we have used between ‘visuospatial’ and ‘egocentric’ tasks may not fully reflect the differences between the types of task employed in rodent studies; particular task differences are also likely to be important. Specifically, impaired two-choice operant reversal learning has been observed in the 5-HT2CR KO mouse [43]. The reversal task described by Pennanen et al [43] is based on that described by Boulougouris et al [15], as is the one used in our earlier ‘visuospatial’ study [16]. However animals had to initiate individual trials in the two earlier studies by nosepoking into the magazine [15], [16], whereas trials were automatically initiated after a very short ITI in the recent report [43]. This is likely to have led to different behavioural strategies being used to ‘solve’ the task which themselves may be differentially sensitive to serotonergic manipulations. It may be that the perseverative impairments of 5-HT2CR KO mice in some task variants, including that used here and the one described in [43] are related to elevated dopamine dysregulation in the dorsal striatum or elsewhere.

Concluding Remarks
Taken together the present results, in conjunction with previous studies, suggest that acute 5-HT2CR antagonism is likely to enhance reversal learning in visuospatial assays by acting on receptors located within the OFC. However there is likely to be more significant involvement of other areas, including the hippocampus, in the effects of such antagonists on reversal learning when animals perform an egocentric spatial task. Constitutive loss of the 5-HT2CR has more substantial effects on performance in the present egocentric spatial task which are likely to involve disturbance of function in additional brain areas, including the hippocampus and dorsal striatum.
Notably, behavioural perseveration may be produced by underlying cognitive deficits of perseverance and learned non-reward and influenced by other factors such as motor impulsivity. The wide range of tasks used to assess reversal learning is likely to pose very different demands and involve these processes to different extents and hence heterogeneity in results is to be anticipated.
The present findings may have relevance to the pathology and treatment of the cognitive deficits of schizophrenia, as the cognitive inflexibility deficits of the disease can be produced by specific deficits in perseverance [21]. Similar perseverative deficits were observed in 5-HT2CR KO mice, suggesting that a constitutive loss of the 5-HT2CR may be relevant for understanding the cognitive inflexibility that is characteristic of schizophrenia. Moreover, SB242084 facilitated the ability to overcome perseverative responding, while causing a concurrent increase in learned non-reward. As schizophrenia has been associated with both increased perseverance [21] and attenuated latent inhibition and learned irrelevance [56], [57], a tentative suggestion would be that the 5-HT2CR might be a pharmacologically relevant target for opposing both impairments.
In conclusion, the current results show that the 5-HT2CR modulates perseverative responding in an egocentric reversal learning procedure. The pattern of results indicates that serotonergic modulation of visuospatial and egocentric reversal tasks depends on different underlying neural systems and that constitutive loss of 5-HT2C receptors leads to impaired acquisition of egocentric discriminations. An important challenge for future studies will be to specify the nature of these differences in both the tasks and experimental manipulations. This will have particular relevance preclinical tests used to characterise novel pharmacological treatments of human psychopathology.