What pharmacological interventions indicate concerning the role of the perirhinal cortex in recognition memory

1.1. Testing rodent recognition memory

The great majority of studies of rodent recognition memory use tasks based on the animals’ spontaneous preference for novelty: a rat (or mouse) will spend longer exploring an object if that object is less familiar or in a less familiar place than an object that is familiar (on the basis of previous exploration) and in a familiar place (Dix & Aggleton, 1999; Ennaceur & Delacour, 1988); see Fig. 1. Such tasks have the advantage of avoiding the potential confounds of using differential rewards or punishments. They have the disadvantage of greater variability and more ready disruption than explicitly reinforced behaviours. Care must be taken if a measure of object familiarity is wanted that is uncontaminated by spatial or associational factors. The most common design for measuring object recognition memory starts with habituating the animal to a display arena. Arenas are typically a square enclosure with four walls that are of the same colour and material and are sufficiently high to mask other features of the room; however, circular arenas, Y-mazes (Winters, Forwood, Cowell, Saksida, & Bussey., 2004), shuttle boxes (Mumby & Pinel, 1994) and ‘bow-tie mazes’ (Albasser, Poirier, & Aggleton, 2010; Albasser, Chapman, et al., 2010) are also used. In a sample or acquisition phase the animal is then allowed to explore (typically for several minutes) two identical copies of an object. After a memory delay (minutes to days), in the choice (or test or retrieval) phase the animal is allowed to explore a third copy of the previously explored object along with a new object occupying the same position as previously occupied by the second copy of the explored object (Fig. 1A). The expectation is that the new object will be explored more than the familiar object. Lesions have established that this task is dependent upon the perirhinal cortex (Barker, Bird, Alexander, & Warburton, 2007; Barker & Warburton, 2011b; Bussey, Muir, & Aggleton, 1999; Ennaceur, Neave, & Aggleton., 1996; Mumby & Pinel, 1994; Norman & Eacott, 2004; Winters et al., 2004).

In the majority of studies, object recognition memory has been found to be unimpaired by hippocampal lesions (Barker & Warburton, 2011b; Forwood, Winters, & Bussey., 2005; Good, Barnes, Staal, McGregor, & Honey, 2007; Langston & Wood, 2010; Mumby, Gaskin, Glenn, Schramek, & Lehmann, 2002; Winters et al., 2004, though see Clark et al., 2000, 2001). It is noteworthy that this hippocampal–perirhinal dissociation is not true for a variant of this task in which two different objects are explored in the sample phase, one of which is replaced in the choice phase, i.e., explore A+B and test A versus C or C versus B, rather than explore A+A and test A versus B (e.g., Clarke, Cammarota, Gruart, Izquierdo, & Delgado-Garcia, 2010; Mansuy, Mayford, Jacob, Kandel, & Bach, 1998; Oh et al., 2010; Pittenger et al., 2002; Barker, G.R.I. unpublished observations). Under the view advanced in this review, in this hippocampally-dependent variant of the object recognition memory task, the rodent also makes spatial or relational associations that require hippocampal processing and which are not made when the sampled objects are both the same; however, that such associations are indeed made remains to be tested experimentally. There are also indications that the involvement of perirhinal cortex and hippocampus may differ if the task employs multiple rather than single exposure trials (Albasser et al., 2011a; Forwood et al., 2005; Gaskin et al., 2010; Mumby, Tremblay, Lecluse, & Lehmann., 2005).

Memory for the familiarity of a location may be tested by starting with a sample phase with two identical copies of an object and, after a delay, measuring exploration of a third copy of the object in the same position compared to that of a fourth copy in a new location (Dix & Aggleton, 1999); see Fig. 1B. The expectation is that the object in the new location will be explored more than the object in the unchanged position. For this task intra- and extra-maze spatial cues are made available to the rat. Lesion studies indicate that this task is dependent on the hippocampus and not perirhinal cortex (Barker & Warburton, 2011b; Mumby et al., 2002), the perirhinal independency presumably reflecting the fact that the task makes no specific demands on object identification.

Memory for the familiarity of a specific association between a particular object and a particular place (object-in-place memory) may be tested by allowing the animal to explore four (or more in some versions) different objects in the sample phase and later testing exploration of two objects that have remained in the same positions against two objects that have exchanged positions (Dix & Aggleton, 1999); see Fig. 1C. Again, intra- and extra- maze spatial cues are made available to the rat. The expectation is that the objects that have moved positions will be explored more than the objects that have stayed in the same place. Lesions have demonstrated that this task is dependent on perirhinal cortex, hippocampus and medial prefrontal cortex (Barker et al., 2007; Barker & Warburton, 2011b; Bussey, Duck, Muir, & Aggleton, 2000).

Recency or temporal order memory may be tested by having two (or potentially more than two) successive sample phases before the choice phase. In the first sample phase two copies of an object are explored and in the second sample phase two copies of a second, different object are explored; see Fig. 1D. Differential exploration of a copy of each of the objects from the first and second sample phases is then measured in the test phase, with the expectation that the object seen first will be explored more than that seen second (Mitchell & Laiacona, 1998). Lesions have demonstrated that this task is dependent on perirhinal cortex, hippocampus and medial prefrontal cortex (Barker et al., 2007; Barker & Warburton, 2011b; Hannesson, Vacca, Howland, & Phillips, 2004a,b; Mitchell & Laiacona, 1998).

1.2. Sensory perception and memory

1.3. Pharmacological interventions

As indicated above, ablation studies have established the importance of perirhinal cortex (amongst other regions) for recognition memory. However, a finer dissection of underlying mechanisms is possible by using reversible interventions produced by systemic or localised infusions of compounds with selective actions at particular receptors. Recognition memory is well-suited to pharmacological studies as acquisition can occur in a single brief exposure and the memory formed can last for over 24 h. As in other cortex, perirhinal cortex contains multiple types of receptors for neurotransmitters such as glutamate, γ-amino butyric acid (GABA), acetylcholine and monoamines.

Taking the definition of perirhinal cortex, i.e., areas 35 and 36, from Shi and Cassell (1999), the region is located >4 mm and <6.5 mm posterior to bregma. This definition of perirhinal cortex accords with the region previously investigated in recording studies (Zhu, Brown, & Aggleton, 1995a). It is the region most closely associated with a lesion-induced visual recognition memory deficit (Albasser, Davies, Futter, & Aggleton, 2009), and is differentially activated by novel and familiar individual items in immunohistochemical imaging work (Albasser, Poirier, et al., 2010; Zhu, Brown, McCabe, & Aggleton, 1995b). So defined, perirhinal cortex is sufficiently compact, so that drug infusion via a cannula can be largely confined to it, with there being little spread to surrounding regions. In various regions of the brain where it has been measured, including perirhinal cortex, 1 μl of infusate typically spreads outwards for ∼0.5–1.5 mm from the cannula tip (Attwell, Rahman, & Yeo, 2001; Izquierdo et al., 2000; Martin, 1991).

In many studies the infusate concentration is in the order of 10–100 times the Ki value for the target receptor. In spite of a potentially high and necessarily uncertain concentration within the target region, where comparisons have been made results have proved similar to those produced by systemic injections where the concentration delivered within the brain is more readily established. In particular, for metabotropic glutamatergic and cholinergic anatagonists, and the L-type calcium channel blocker verapamil, the same temporal pattern of recognition memory impairment (see further below) has been found when administration is systemic as when it is by infusion into perirhinal cortex. Accordingly, systemic administration that necessarily affects the hippocampus as well as other brain regions in addition to perirhinal cortex does not change the recognition memory deficit found when only perirhinal cortex is targeted. This parallelism of effects between localised perirhinal and systemic administration potentially involving the whole brain strongly argues for the dominant role of perirhinal cortex in the tested recognition memory functions. Moreover, the parallelism of effects between the local perirhinal infusions and systemic administrations indicates that the infusion findings are not distorted by drug concentration gradients that are likely occur with distance from the cannula tip.

Unlike administration by a systemic route, any effects of the compound must result from actions within the perfused region (though it needs to be remembered that these might include compromise of the functioning of distal sites). Thus localised infusion has the advantage of site-specific delivery and the avoidance of potential peripheral side-effects. It is also usable with drugs that do not cross the blood–brain barrier or which would have major detrimental effects on an animal’s health or behaviour if delivered systemically. Notably, here, it is less likely that effects on recognition memory performance will be produced by impairments of global alertness, attention or movement—though the absence of such potential effects should nevertheless be checked through comparisons with the behaviour of controls. Drug effects need always to be compared to the effects of infusion of a similar volume of the vehicle solution, having a matched overall ionic composition, pH and osmolarity.

Importantly, compared with ablation, the effects of infusions are potentially reversible. Though varying with the compound, the actions of many infusates will be established in the target region within 15 min and last for about an hour (e.g., Day, Langston, & Morris, 2003). Infusions may therefore be given so that the infusate is active either during acquisition (and early consolidation), after acquisition and during consolidation, or during retrieval—so allowing potential actions upon memory acquisition, consolidation and retrieval to be separately assessed. However, it needs to be appreciated that at short memory delays drugs given before acquisition are likely to be present also during consolidation and retrieval.

Indeed, drugs active during acquisition may produce recognition memory impairment when retrieval is in the absence of the drug, whereas that impairment disappears if the drug is also active during retrieval. Such an effect is called ‘state-dependency’ as the mnemonic information becomes more readily retrievable when the brain is in the same state as at acquisition. Testing for state-dependency is important when effects of a drug are seen at long but not short memory delays. This is because at a short delay (typically <∼30 min) it is likely that the drug will be active not only during acquisition but also during retrieval, whereas the drug is unlikely to be still active after a longer memory delay ($\gg$1 h).

The results of the large number of pharmacological studies of recognition memory that involve the selective targeting of specific brain regions, almost all performed in the rat, will now be summarised. Most studies have looked at effects on object recognition memory; however, as available, effects of drug delivery in perirhinal cortex upon other types of recognition memory will be contrasted with effects from targeting hippocampus and medial prefrontal cortex. The effects of drugs on acquisition, retrieval and consolidation will be separately considered. Although drugs may affect memory performance by increasing (or reducing) interference between memoranda, they may also impair (or enhance) initial registration or, by acting on consolidation mechanisms, produce (or potentially reduce) temporal decay, or disrupt or enhance retrieval processes.

2.1. Perirhinal cortex

2.2. Other areas

To investigate the extent to which perirhinal function may be dependent upon or duplicated by other brain regions, studies of the effects upon recognition memory acquisition produced by pharmacological infusions to other areas will now be considered. Other parts of the visual system are likely to provide input and may receive output from the perirhinal cortex. The hippocampal system continues to be of interest as a potential partner of perirhinal cortex in recognition memory functions. There have, however, been relatively few such studies.

3.1. Perirhinal cortex

In contrast to the impairments found when drugs are infused into perirhinal cortex so as to be active during acquisition, most compounds that have been tested do not have any effect on rat recognition memory when infused so as to be active during retrieval. This is not because perirhinal cortex is unnecessary for recognition memory retrieval processes. CNQX infusion during the choice phase produces recognition memory impairment (Winters & Bussey, 2005c). This result establishes that perirhinal cortex is necessary for object recognition memory during the retrieval process as well as during acquisition, again a result that it is not possible to establish with conventional ablation techniques. This impairment arises because CNQX shuts down perirhinal cortex activity during retrieval, so preventing the processing of any perirhinal signal at that time. Confirmation of the involvement of perirhinal cortex in retrieval processes is provided by the impairment produced by local infusion of the sodium channel blocker lidocaine (Barnes, Floresco, Kornecook, & Pinel, 2000; Hannesson et al., 2004b; Winters & Bussey, 2005b). It should, however, be noted that since lidocaine blocks axonal transmission, its infusion will also involve fibres that merely pass through perirhinal cortex rather than terminating or originating there.

CNQX and lidocaine (Hanneson et al., 2004a,b; Winters & Bussey, 2005b,c) impair both acquisition and retrieval, but most drugs that have been tested impair acquisition but not retrieval. An exception is provided by the blockade of voltage-dependent calcium channels by verapamil: blockade of such channels at either acquisition or retrieval produces recognition memory deficits (Seoane et al., 2009). A disjunction between acquisition impairment and a lack of effect on retrieval has been demonstrated for antagonism of NMDA (Barker et al., 2006a; Winters & Bussey, 2005b, though see Abe et al., 2004), kainate (Barker et al., 2006a), and metabotropic glutamate receptors (Barker et al., 2006b), and muscarinic cholinergic receptors (Tinsley et al., 2011; Warburton et al., 2003; Winters, Saksida, & Bussey, 2006, though see Abe et al., 2004 and, using mice, Botton et al., 2010), and for the benzodiazepine lorazepam (Wan et al., 2004). These findings strongly indicate that it is easier to disrupt recognition memory by interfering with perirhinal processes during acquisition than when the same interference occurs during retrieval. Accordingly, either acquisition processes are more readily disrupted than retrieval processes or perirhinal cortex is in some way less critical for retrieval than for acquisition. The former is more likely to be the case. First, in other brain regions and for various tasks acquisition is also found to be more readily disrupted than retrieval (see for review, Squire, 2006). Second, the impairments produced by CNQX and verapamil establish the involvement of perirhinal cortex in retrieval processes (Seoane et al., 2009; Winters & Bussey, 2005c): the lack of effects of other drugs is not because some other region has rendered perirhinal activity unnecessary. Indeed, the disruptive effects of the drugs upon acquisition rather than retrieval are most easily explained by their interference with the conditions necessary for synaptic plasticity to occur, with the consequent impairment of the laying down of the memory.

It is worth noting that there has not been determination of the maximum length of time for which information concerning a specific example of prior occurrence is held in perirhinal cortex. For memories that depend upon the hippocampus at acquisition, there is accumulating evidence that at least some of them become independent of the hippocampus over a period of days to weeks (see for recent review, Winocur & Moscovitch, 2011). However, the potential storage capacity of perirhinal cortex for familiarity discrimination of objects is sufficiently high that such storage might reside permanently within the perirhinal cortex (Androulidakis et al., 2008; Bogacz & Brown, 2003).

3.2. State dependency

Where investigated in studies of localised drug action within perirhinal cortex, no evidence of state-dependent effects has been found for rodent recognition memory. In particular, neither NMDA (Barker et al., 2006a) nor metabotropic glutamate receptor (Barker et al., 2006b) antagonism demonstrated state-dependency. Therefore for these antagonists the lack of impairment at shorter delay (15–20 min) compared to the impairment at longer (24 h) delay is not explained by state-dependency. However, a state-dependent recognition memory impairment has been reported for infusions of a nitric oxide synthesis inhibitor (nomega-nitro-l-arginine) into the hippocampus (Blokland, Prickaerts, Honig, & de Vente, 1998).

In human studies state dependency has been reported for the cholinergic muscarinic antagonist scopolamine (Petersen, 1979) and cholinergic agonist nicotine (Warburton et al., 1986).

4.1. Perirhinal cortex

Recognition memory impairments are found when CNQX or lidocaine is infused into perirhinal cortex very shortly after the sample phase so as to be active during early consolidation (Winters & Bussey, 2005b,c). No impairment is found if infusion is >40 min after acquisition (Winters & Bussey, 2005b,c). Accordingly, shutting down perirhinal activity shortly after acquisition causes impairment. This suggests either that activity-dependent processing (extracellular signalling) continuing in perirhinal cortex during the first hour after acquisition is necessary for consolidation, or that silencing the neurons during this time causes disruption of intracellular signalling, thereby interfering with normal consolidation. No tests have been made of the effects of infusions preventing perirhinal activity over many hours.

The effect of post-acquisition intra-perirhinal infusion of an NMDA receptor antagonist (AP5) on consolidation is equivocal as impairment was not seen in one study (Barker et al., 2006a) but was in another (Winters & Bussey, 2005c). Impairment was also found when AP5 was present during both consolidation and retrieval when the memory delay was 25 min (Abe et al., 2004). It should be noted that any impairment with post-acquisition infusion undermines the argument that NMDA receptor antagonism produces impairment by acting solely at acquisition. However, studies in perirhinal cortical slices maintained in vitro indicate AP5 blocks the induction of artificially produced synaptic plasticity rather than its maintenance (Ziakopoulos et al., 1999), so the predicted impairment is of acquisition. Kainate and metabotropic glutamate receptor antagonism was without effect on consolidation (Barker et al., 2006a,b). Again, metabotropic glutamate receptor activation has been shown to be important for perirhinal plasticity mechanisms (Cho et al., 2000).

Post-acquisition intra-perirhinal infusion of the muscarinic receptor antagonist scopolamine results in improved performance (Winters et al., 2006). This improvement was interpreted as arising from decreased interference (Winters et al., 2006), as has been reported for alcohol, nitrous oxide, and benzodiazepine -induced amnesias in human subjects (Brown et al., 1982; Parker et al. 1980; Summerfield & Steinberg, 1957).

The current leading model of memory formation presumes that activity-dependent release of neurotransmitters and their consequent binding to their receptors leads to changes in intracellular concentrations of calcium ions that in turn trigger biochemical cascades that lead to synaptic plasticity (e.g., Bading, Ginty, & Greenberg, 1993; Kotaleski & Blackwell, 2010; Citri & Malenka, 2008), i.e., they result in changes in synaptic efficacy and the maintenance of that change. Indeed, there is a report of synaptic remodelling in perirhinal cortex after object recognition memory training (Platano et al., 2006). One established mechanism of changing synaptic efficacy is to alter the number of AMPA glutamate receptors at the synapse (Collingridge, Isaac, & Wang, 2004; Nicoll and Malenka, 1999); increasing the number will produce strengthening (as for example seen in long-term potentiation); reducing it will produce weakening (as for example seen in long-term depression). By hypothesis, the memory trace is represented by the maintained changes in synaptic efficacy across the whole of the network involved (see for discussion, Martin, Grimwood, & Morris, 2000).

Infusing into perirhinal cortex drugs that target intracellular processes related to synaptic plasticity mechanisms (potential consolidation mechanisms) has been shown to produce recognition memory impairments. Most such investigations have concerned object recognition memory. Typically effects have been seen on memory measured after a 24 h but not a 20 min delay. The lack of effect on memory after a 20 min delay may be because either the consolidation mechanisms being studied are unnecessary for memory at such a short delay or that the underlying synaptic plasticity mechanisms differ (as discussed above).

Calcium–calmodulin-dependent kinases (CamKs) are enzymes that are activated as a result of increases in intracellular calcium ions. In turn, these kinases phosphorylate other enzymes involved in intracellular signalling processes necessary for memory formation, as has been shown for other types of memory dependent on the hippocampus or amygdala (e.g., Giese, Fedorov, Filipkowski, & Silva, 1998; Lisman, Schulman, & Cline, 2002; Rodrigues, Farb, Bauer, LeDoux, & Schafe, 2004; Wayman et al., 2008). Recognition memory impairments are produced by infusing into perirhinal cortex inhibitors of these enzymes so as to interfere with consolidation mechanisms. Infusion of an inhibitor of CamK kinase (CamKK) immediately post-acquisition but not 20 min after acquisition impaired object recognition memory (Tinsley et al., 2012). Thus CamKK is important for perirhinal consolidation mechanisms necessary for recognition memory within the first 20–30 min after acquisition. CamKK phosphorylates CamKI. As expected, the CamKK inhibitor also reduced the phosphorylation of CamKI; however, CamKI phosphorylation was similarly reduced in response to the viewing both of novel and of familiar stimuli. This suggests that the effect on CamKK is not selective for the registration of the occurrence of the novel stimulus (Tinsley et al., 2012). The effect of the inhibitor may thus be a generalised downgrading of cellular function or it might act on reconsolidation (of the occurrence of the familiar object) as well as consolidation (of the occurrence of the novel object) mechanisms. Inhibition of CamKII within perirhinal cortex 20–100 min after acquisition impairs object recognition memory (Tinsley et al., 2009). Such inhibition of CamKII prevented the increased phosphorylation of CamKII produced by viewing novel rather than familiar images, suggesting that phosphorylation of CamKII is selectively related to the processing of information concerning novel stimuli. These experiments also indicated that CamKIV did not compensate for the inhibited CamKII phosphorylation in long-term object recognition memory (Tinsley et al., 2009). Thus the evidence suggests that CamKII phosphorylation is more closely related to the registration of novel information in perirhinal cortex than either CamKI or CamKIV.

One of the targets of phosphorylated CamKII is calcium response element binding protein (CREB). CamKII activates CREB by phosphorylating it (forming pCREB). pCREB then becomes active by dimerising. pCREB can be prevented from dimerising by using viral transfection to produce a dominant negative construct that binds pCREB. When this is done in perirhinal cortex, long-term recognition memory is impaired (Warburton et al., 2005). As viral transfection rather than an infusion was used, the time after acquisition at which consolidation might be disrupted could not be established (indeed, that the effect is on consolidation must be surmised).

In turn, one of the eventual targets affected by CREB phosphorylation is the immediate early gene c-fos. Changes in Fos, the protein products of c-fos, have provided a reliable marker for changes in neuronal activity related to recognition memory across many studies (see for review, Aggleton et al., this issue). Fos has been implicated in synaptic plastic mechanisms that produce reductions in efficacy (Lindecke et al., 2006; Nakazawa, Karachot, Nakabeppu, & Yamamori, 1993). Reductions in synaptic strength have been predicted to underlie familiarity discrimination (Brown & Aggleton, 2001; Brown & Xiang, 1998). Infusion into perirhinal cortex of an oligodeoxynucleotide (ODN) that prevents the production of Fos also produces an impairment in long-term (but not shorter-term) recognition memory (Seoane et al., 2012). A similar impairment was produced by perirhinal infusion of an ODN against brain-derived neurotrophic factor (BDNF). Additionally, expression of BDNF has been reported to be increased in perirhinal cortex after performance of an object recognition memory task (Griffin, Bechara, Birch, & Kelly, 2009; Hopkins, Nitecki, & Bucci, 2011; Muñoz, Aspé, Contreras & Palacios, 2010).

As mentioned above, it has been proposed that the primary plastic change underlying object recognition memory is synaptic weakening (Brown & Aggleton, 2001; Bogacz & Brown, 2003; Brown & Xiang, 1998). Such weakening would explain why responses to a new stimulus are reduced when that stimulus is seen again, as is found in perirhinal and neighbouring cortex (Brown & Xiang, 1998). Indeed, computational modelling has established that efficient familiarity discrimination must employ some element of synaptic weakening (Bogacz & Brown, 2003; Lulham et al., 2011). Activity-dependent removal of AMPA glutamate receptors from the synaptic membrane is the most obvious (but not the only) way in which to reduce excitatory synaptic strength. Such removal can be prevented by blocking the appropriate intracellular machinery—specifically by preventing AP2 (clathrin adaptor protein 2) from binding to the GluR2 subcomponent of AMPA receptors, a necessary step for the receptor’s removal from the membrane (Androulidakis, Lulham, Bogacz, & Brown, 2008). This impairment across the memory delays indicates that although there may be more than one method for inducing synaptic change in perirhinal cortex, these methods are dependent upon the same expression mechanism, namely internalisation of AMPA receptors.

Interestingly, this prevention of AMPA internalisation did not impair object-in-place memory (though this was only tested at a short memory delay). This lack of impairment indicates that object-in-place recognition memory does not share the same underlying synaptic plastic change with object recognition memory in perirhinal cortex. Preventing the activity-dependent removal of AMPA receptors by blocking AP2’s interaction leaves normal synaptic transmission as measured in vitro unaffected (Griffiths et al., 2008). Thus it is only the learning-related, long-term changes in transmission that are prevented. The important implication of this selective effect is that the responses of perirhinal neurones to stimuli should continue to be transmitted to other brain regions as normal (with the exception of any change that would otherwise have been produced by learning-related synaptic plasticity). Hence perirhinal non-mnemonic, perceptual functions should be normal after such selective, ‘plasticity lesions’. In particular, it is to be expected that object-related information would be transmitted to the hippocampus for use in an object-in-place task. Thus one interpretation of the lack of impairment of the object-in-place task by interfering with AP2 mechanisms is that the registration (plasticity and storage mechanisms) of where the particular objects had been seen previously are located in the hippocampus and only perirhinal transmission of object identity is necessary for successful performance of the task. An alternative interpretation for the lack of impairment is that the object-in-place task relies on synaptic strengthening rather than weakening processes in perirhinal cortex. The above findings strongly imply that it is possible to impair perirhinal-based recognition memory in the absence of a major perceptual deficit (object-in-place memory is more perceptually demanding than object recognition memory); however, perceptual performance studies have not been conducted after such interventions and there is a need to investigate effects on object-in-place memory at longer delays.

A recent report has suggested that the recognition memory impairment produced by blocking AMPA receptor removal occurs at retrieval (Cazakoff & Howland, 2011). Infusion into perirhinal cortex of Tat-GluA2(3Y), a compound that prevents AMPA receptor removal, produced impairment if infusion was before retrieval, but not if it was before or immediately after acquisition. Whereas a potential effect upon consolidation has a clear theoretical underpinning, an effect upon retrieval remains to be explained.

Intra-perirhinal infusion of an inhibitor of the activity of protein kinase Mzeta (PKMzeta), whose activity is necessary for the maintenance of synaptic strengthening (Sacktor, 2008) impairs object recognition memory when infused after acquisition (Outram, Tinsley, Henley Warburton, & Brown, 2010). In contrast, interfering with PKMzeta in the hippocampus is without effect on single object recognition memory—though it does impair object location memory (Hardt et al., 2010).

4.2. Hippocampus

Object recognition memory impairment is produced by blocking nitric oxide synthesis by an infusion into the hippocampus after acquisition, so as to be active during consolidation (Blokland et al., 1998; Furini et al., 2010). Increasing cyclic guanosine monophosphate (cGMP) -related activity by infusion during early consolidation improved rats’ object recognition memory performance (Prickaerts, de Vente, Honig, Steinbusch, & Blokland, 2002). Beta noradrenergic antagonism and inhibition of BDNF function in the CA1 region of the hippocampus after acquisition also impaired memory in the hippocampally-dependent (explore A+B, test A versus C) version of the object recognition task (Furini et al., 2010). The involvement of hippocampal (subfield CA1) cannabinoid receptors in recognition memory consolidation was established using the same task: infusions of the non-selective cannabinoid receptor agonist WIN-55,212-2 or the endocannabinoid membrane transporter inhibitor VDM-11 immediately after acquisition impaired performance of the task after long but not short delays (Clarke et al., 2008).

Using the same hippocampally-dependent version of the object recognition task, impairments in long-term but not short-term memory are produced by hippocampal (CA1) interventions that block protein synthesis in the first few hours following acquisition (Lima et al., 2009), or inhibit the activity of mTOR, a protein kinase involved in the initiation of mRNA translation (Myskiw et al., 2008), or prevent functioning of phosphorylated CREB (Pittenger et al., 2002). A similar impairment was found when protein synthesis was blocked in entorhinal cortex (Rossato et al., 2007). Three of these studies (Lima et al., 2009; Myskiw et al., 2008; Rossato et al., 2007) also found effects upon potential reconsolidation mechanisms when infusions were made immediately after viewing a novel with a familiar object (there was subsequent amnesia relating to both objects). Evidence for involvement of regions of the mouse brain in reconsolidation as well as consolidation mechanisms has been provided by changes in BDNF and egr-1 gene expression in several areas (including perirhinal, entorhinal and prefrontal cortices and hippocampus) after training or reactivation sessions at various time delays following initial acquisition (Romero-Granados et al., 2010). Again, the hippocampally-dependent object recognition memory task was used, together with retraining sessions at long delays.

There are a number of reports of modifications of hippocampal synaptic plastic processes in genetically modified mice that show object recognition memory deficits; however, the genetic modifications are not restricted to the hippocampus and modifications of perirhinal function were not sought (e.g., Jones et al., 2001; Malleret et al., 2001; Mansuy et al., 1998). The studies establish a role for the phosphatase calcineurin (Malleret et al., 2001; Mansuy et al., 1998) and immediate early gene Zif268 (Jones et al., 2001) in object recognition memory, but not which region or which memory phase might be affected. Again, these studies used the hippocampally-dependent version of the object recognition task.

In summary, the large majority of studies that have found an involvement of the hippocampus in the consolidation of recognition memory have used the variant of the object recognition memory task that lesion studies have shown to be dependent on the hippocampus as well as perirhinal cortex (e.g., Clarke et al., 2010; Mansuy et al., 1998; Oh et al., 2010; Pittenger et al., 2002; Barker, unpublished observations).

5.1. Object location

When a recognition memory judgement depends on spatial memory the hippocampus is critically involved. Thus, lesions in the hippocampus or transection of the fornix produce significant impairments in the object location task as described above (Barker & Warburton, 2011b; Bussey et al., 2000; Ennaceur, Neave, & Aggleton, 1997, though see also Langston & Wood, 2010), yet have no effect on novel object recognition (Barker & Warburton, 2011b; Langston & Wood, 2010). Consistent with this dissociation in the effects of hippocampal lesions, intra-hippocampal infusions of the AMPA receptor antagonist NBQX before acquisition impair object location memory tested following either a 5 min or a 1 h delay, yet have no effect on object recognition memory in rats (Barker & Warburton, 2009). Intra-hippocampal infusions of the NMDA receptor antagonist AP5 before acquisition impair object location memory tested following a 1 h delay. Thus both AMPA and NMDA receptor neurotransmission in the hippocampus are critical for recognition memory for familiar objects presented in a novel location. Selective effects upon acquisition and retrieval have not been investigated. However, temporary inactivation of the hippocampus (using the GABA receptor agonist muscimol) immediately after acquisition produced impairment in mice tested after a 24 h delay, indicating an involvement of the hippocampus in consolidation of object location recognition memory (Oliveira, Hawk, Abel, & Havekes, 2010)—interestingly, this intervention produced an enhancement of object recognition memory.

5.2. Object-in-place

Studies employing selective lesions have shown that object-in-place memory depends on multiple brain regions including the perirhinal cortex, medial prefrontal cortex and hippocampus (Barker & Warburton, 2011b; Barker et al., 2007); see Fig. 5. Importantly it has also been established that each of these regions needs to interact with the other two within the neural network for successful task performance. Moreover, insufficient information crosses over between the hemispheres to sustain performance if any two of the structures are removed unilaterally but in different hemispheres. The evidence reviewed above suggests that the contribution of the perirhinal cortex is in the processing and storage of object information, while that of the hippocampus is in spatial processing. As damage to the medial prefrontal cortex impairs object-in-place memory but neither object nor object location recognition memory, the role of medial prefrontal cortex may be in the integration of object and place information necessary to make such recognition memory judgements (Barker et al., 2007). However, it is not necessary to conclude that all the object information and all the spatial information necessary to solve the associative task is held in medial prefrontal cortex (nor as multiple copies in each of the component structures): medial prefrontal cortex may rather hold the associative links (including co-occurrence) between what is stored in the hippocampus and what is held in perirhinal cortex. In such a case, the memory would be held as a distributed engram across the network rather than as, say, an engram confined to medial prefrontal cortex that merely has to be accessed from the other two structures. This issue will only be resolved by eventually determining what type of information necessary to task performance is stored in each of the regions.

5.3. Temporal order memory

Fewer studies have explored the pharmacology of temporal order or recency memory, i.e., recognition memory that depends on judging how long ago an item was encountered or the order in which stimuli were presented. Results from studies employing permanent lesions or temporary inactivation through the use of lidocaine have shown that this form of recognition memory is also dependent upon the perirhinal cortex, medial prefrontal cortex and hippocampus (Barker & Warburton, 2011b; Barker et al., 2007; Hannesson et al., 2004a,b; Mitchell & Laiacona, 1998), and that these regions also need to interact (Barker & Warburton, 2011b; Barker et al., 2007; Hannesson et al., 2004b).

5.4. Summary

In summary, pharmacological studies of other types of recognition memory have confirmed the results of lesion studies that establish the importance of regions other than the perirhinal cortex for these other types of recognition memory. Thus object location recognition memory is dependent upon the hippocampus and does not require the involvement of perirhinal cortex as the identity of the moved object is not required for determining that a new location has been occupied. Object-in-place and temporal order recognition memory depend upon a neural circuit involving the perirhinal cortex, hippocampus and medial prefrontal cortex. Each component region of this circuit is essential, as is their interaction. Nevertheless, the indications are that the different components perform different functions. In confirmation of the work on object recognition memory, perirhinal cortex appears to provide information about objects and their familiarity, while the hippocampus provides spatial information and the medial prefrontal cortex is important for forming associational connections and for temporal information. Moreover, the roles of each region have been demonstrated to encompass both acquisition and retrieval, with acquisition being dependent upon AMPA and NMDA glutamatergic and also muscarinic cholinergic neurotransmission.