Natural exploration is of particular importance when considering animal behavior, as it is the means by which animals gather information about their environment, discover available resources, and identify potential threats. Furthermore, a number of studies have shown that tasks utilizing natural exploratory behavior can be particularly useful and efficient paradigms to investigate specific processes involved in learning and memory (Gilbert, Kesner & Decoteau, 1998; Gilbert, Kesner & Lee, 2001; Gilbert & Kesner, 2002, 2003, 2004; Hauser, Totentino, Pirogovsky, Weston & Gilbert, 2009; Kesner, Hunsaker & Gilbert, 2005), and specifically the process of match-mismatch recognition (Lee, Hunsaker & Kesner, 2005; Save et al., 1992). A substantial body of research exists on the contributions of the hippocampus and its subregions, to a number of specific processes involved in learning and memory.
As reviewed by Gilbert and Brushfield (2009) and Kesner (2007) the DG-CA3 network has been shown to be critical in a number of processes, including: (1) the detection of novel spatial configurations of familiar objects (Hunsaker, Mooy, Swift & Kesner, 2007; Lee, Hunsaker & Kesner, 2005); (2) the formation of arbitrary associations (e.g., information from the parietal cortex regarding the spatial location of an object may be associated with information from inferior temporal cortex regarding the visual recognition and identity of the same object), specifically those that involve associations between a stimulus and a spatial location (Gilbert & Kesner, 2002, 2003; Kesner, 2007; Kesner, Gilbert & Wallenstein, 2000; Lee, Jerman & Kesner, 2005; Rolls, 1996; Rolls & Kesner, 2006); (3) pattern completion (i.e., the ability to use partial or degraded cues to retrieve a complete representation of stored information; see Gold & Kesner, 2005; Lee, Jerman, & Kesner, 2005; Lee, Rao, & Knierim, 2004; Leutgeb & Leutgeb, 2007); and (4) pattern separation (i.e., the ability to separate partially overlapping patterns of activation so that one pattern may be retrieved separately from other patterns; see Gilbert & Kesner, 2006; Gilbert, Kesner & DeCoteau, 1998; Lee, Hunsaker & Kesner, 2005; Leutgeb & Leutgeb, 2007; McNaughton & Nadel, 1989; O’Reilly & McClelland, 1994; Rolls, 1996). There are a number of anatomical and cytoarchitectural features of the DG-CA3 network that allow for these operations to occur.
Functional dissociations between anatomical subregions of the hippocampus have been reported by a number of current studies (Lee et al., 2004; Lee, Hunsaker & Kesner, 2005; Lee, Jerman & Kesner, 2005; Leutgeb & Leutgeb, 2007; Leutgeb, Leutgeb, Treves, Moser & Moser, 2004; Gilbert & Kesner, 2003; Gilbert, Kesner & Lee, 2001; Hunsaker, Tran & Kesner, 2008, 2009; Kesner, Lee & Gilbert, 2004; Rolls, 1996; Shapiro & Olton, 1994; Tanila, 1999). As previously mentioned, pattern separation refers to a hippocampal dependant process that is critical for the reduction of potential interference that can occur when different memory representations have similar elements (Gilbert & Kesner, 2006; Gilbert et al., 1998; Gilbert et al., 2001; Lee, Jerman & Kesner, 2005; Leutgeb & Leutgeb, 2007; Leutgeb et al., 2004; Save et al., 1992; Rolls, 1996). Electrophysiological studies have been able to provide significant support for the idea of regional specificity within the hippocampus with regards to the processing of spatial and temporal pattern completion and separation. Recordings from populations of CA3 pyramidal cells have found that these cells are able to create and maintain very discrete neural representations of visually similar environments (Tanila, 1999). Additionally, when rats are placed in environments with varying degrees of spatial similarity, it has been found that distinct sets of CA3 place-cells are activated for each environment, regardless of the amount of similarity (Leutgeb et al., 2004). Finally, results from a study by Leutgeb and colleagues (2007) indicate that when small changes were made to the shape of an exploratory environment, two distinct changes occurred within the rodent hippocampus: (1) the activity in the DG place cells was altered; and (2) novel subsets of CA3 neurons were activated in response to the environmental changes. These results further support the notion that the DG-CA3 network in the hippocampus may support processes involved in performing temporary pattern storage and potentially reconstructing previously stored representations from partial or degraded inputs (i.e., pattern separation; see Myers & Scharfman, 2009).
The process of pattern separation is supported and made possible by the anatomical and connective structure of the CA3 region and the circuits that comprise the DG-CA3 network (granule cell/mossy fiber pathway; Rolls, 1996, 2010). It is known that the granule cells of the DG project to CA3 pyramidal cells through the mossy fiber pathway (MFP; Amaral & Witter, 1995). In the rat, the MFP from the DG provides very sparse but extremely powerful connections to the 3 x 105 CA3 pyramidal cells. Rolls (2010) indicates that there are approximately 46 synapses from the DG that innervate CA3 pyramidal cells via the MFP; however, as the DG axons are very large and terminate in extremely close proximity to the soma of the pyramidal cells, they are extremely effective in eliciting depolarization. Layer II of the entorhinal cortex (EC) provides a second source of external axonal inputs to CA3. In the rat, EC layer II provides approximately 3.6 x 103 (3,600) synapses on to the 3 x 105 (300,000) CA3 neurons. It also has also been well established that CA3 pyramidal cells are predominantly connected to themselves via the recurrent collateral connections. Rolls (2010) indicates that approximately 1.2 x 104 (12,000) CA3 pyramidal cell synapses create feedback loops that are internally specific to the CA3 region (Gilbert & Brushfield, 2009; Rolls 1996, 2010; Rolls & Kesner, 2006; Treves & Rolls, 1992).
These recurrent collateral connections that create the recursive CA3 network, essentially allow the pyramidal cells to function as a single autoassociative network (for review see Gilbert & Brushfield, 2009). This attribute enables the firing of any set of CA3 neurons representing one part of a memory to be associated with the firing of any other set of CA3 neurons representing another part of the same memory (i.e., a memory that contains both a spatial and an object component will activate a different set of CA3 pyramidal cells for each specific set of stimuli, yet the two sets will be associated together; see Gilbert & Brushfield, 2009; Marr, 1971; Myers & Scharfman, 2009; McClelland, McNaughton & O’Reilly, 1995; Rolls, 2010; Rolls & Kesner, 2006). As previously mentioned, the associatively modifiable synapses within the CA3 recurrent collaterals allow for memories to be encoded and then later retrieved from partial or degraded cues (i.e., pattern completion; Rolls & Deco, 2002). For example; studies have shown that rats with localized CA3 lesions show significant deficits, comparable to rats with complete hippocampal lesions, on object-place (Gilbert & Kesner, 2002, 2003; Rolls, Miyashita, Cahusac, Kesner, Niki, Feigenbaum & Bach, 1989) and odor-place (Gilbert & Kesner, 2003) association tasks. These unique patterns must be encoded such that there is not an extensive overlap between representations. This process is critical in order to insure that unique representations activate separate patterns of CA3 pyramidal cells, and are, therefore, able to be retrieved separately from one-another (Myers & Scharfman, 2009).
The network between the dentate gyrus and the CA3 region requires critical evaluation based on its cellular structure and connectivity. The dentate gyrus is the primary target of the efferent connections of the entorhinal cortex (via the perforant pathway; Swanson & Cohen, 1977; Witter, 1993). In the rat, cortical inputs to the superficial layers [I, II, and III] of the EC originate in the olfactory region of the telencephalon, perirhinal cortex, presubiculum, and parasubiculum (Witter, 1993). Within the hippocampus, the dentate granule cells project to CA3 (via the granule cell/mossy fiber pathway; Amaral & Witter, 1995). Additionally, it has been established that CA3 is integral in the processes of pattern completion. Therefore, based on its intermediary location between the entorhinal cortex and CA3, the dentate gyrus has been shown to be a critical structure when considering the aforementioned processes carried out by region CA3.
CA3 has an extremely efficient mechanism by which to encode information via the DG- CA3, granule cell/mossy fiber pathway from the dentate gyrus (Martinez & Kesner, 1998; Rolls & Kesner, 2006). There are a number of attributes of this DG-CA3 pathway that allow the dentate granule cells to create discrete representations within populations of CA3 pyramidal cells. First, the dentate gyrus is in a crucial anatomical position in the processing circuit as it is the recipient of the major extrinsic input to the hippocampus from the entorhinal cortex (Hjorth- Simonsen & Jeune, 1972; Swanson & Cowan, 1977). The dentate gyrus then projects to CA3 by way of the mossy fiber pathway. CA3 not only provides a major hippocampal output to the septum but also projects to CA1 via the Schaffer collaterals as well as to the contralateral CA1/3 fields (Swanson, 1977). These connections reiterate the notion that the CA3 subfield may be a principal focus of convergence for extrahippocampal activity (Blackstad, Brink, Hem & Jeune, 1970; Swanson & Cowan, 1977). Second, it also has been suggested that the connections that comprise the perforant path (from the entorhinal cortex to the dentate gyrus) may act as a “competitive learning network” (Gilbert & Brushfield, 2009; Rolls, Stringer & Elliot, 2006). Rolls et al. (2006) suggest that interneuron activity in the dentate gyrus could potentially facilitate the process of pattern separation. Activated dentate granule cells would provide excitation to dentate interneurons, and the excitation within those interneurons would subsequently provide inhibition to neighboring granule cells (see also McAuliffe, Bronson, Hester, Murphy, Dahlquist-Topala, Richards & Danzer, 2011). This ‘competition’ in the dentate gyrus would ultimately reduce synaptic redundancy and ensure that the outputs along the DG- CA3 pathway are more highly organized. This competitive organization is a natural consequence of the DG granule cells having to overcome inhibition (provided by the interneurons) in order to undergo long-term potentiation (see Gilbert & Brushfield, 2009; Myers & Scharfman, 2009; Rolls et al., 2006).
As reported by Kesner & Rolls (1996), the organization and discretion of these connections is a direct product of the quantity and distribution of cells and synapses in the DG- CA3 network. The granule cells of the DG outnumber the CA3 pyramidal cells approximately 100:3 in the rat; based on this relatively sparse connectivity there is a large reduction in redundant connections. Specifically, the dentate granule cells number at approximately 106, while the pyramidal cells of the CA3 region total about 3 x 105 (Rolls, 2010). As noted by Rolls (1996, 2010), CA3 neurons receives approximately 46 mossy fiber (mf) inputs from the dentate granule cells. The calculated probability that a given dentate granular cell contacts any individual CA3 neuron is approximately 0.0046% [46 mf synapses/106 granular cells]. The extremely small probability of pyramidal cell innervation via the MFP insures that the DG-CA3 connections are both highly sparse and as discrete as possible (Gilbert & Brushfield, 2009; Rolls & Kesner, 2006). However, as noted by Lee et al. (2005), both the DG and CA3 region are likely crucial in monitoring and responding to spatial changes in the environment.
Other models of hippocampal function have suggested that CA3 acts as the match- mismatch comparator based on its reception of two primary inputs: (1) cortical input, and (2) input from the medial septal nucleus and nucleus of the diagonal band (Vinogradova, 2001). The first input is received from cortical association areas that reach the hippocampus predominantly through superficial layers [II and III] of the entorhinal cortex (EC) via the perforant pathway (Witter, 2007). Specifically, EC layer II projects to the subiculum with projections also innervating the dentate gyrus and CA3. EC layer III projects along the perforant path to CA1 and the subiculum. The dentate gyrus, the primary recipient of the perforant path, then extends mossy fibers to region CA3. The second input to CA3 comes by way of the medial septal nucleus and nucleus of the diagonal band, which provides efferent connections from the reticular formation via the lateral septum by way of the medial forebrain bundle. Based on these two sets of inputs, it is hypothesized that CA3 is able to act as an environmental comparator by matching the signals in both sets; where the detection of a mismatch between the sets would indicate the detection of something novel (for review see Vinogradova, 2001).
As discussed by Lee et al. (2005) a number of models have supported that CA1 may also be a critical region for potential match-mismatch computations (Gilbert, Kesner & Lee, 2001; Hasselmo, Fransen, Dickson & Alonso, 2000; McClelland, McNaughton & O’Reilly, 1995). Thus, it may be likely that in addition to processing in the DG-CA3 network, there could be critical comparisons occurring in CA1 between the direct entorhinal inputs and the inputs coming from CA3. Additionally, it could be hypothesized that CA1 activity may occur specifically to detect mismatch between an internal representation and an external environment when there is a significant enough change in the external environment (Shapiro & Olton, 1994). Specifically, it may be possible that the DG-CA3 network is involved in comparing the current environment to recently constructed representations (e.g., minutes old), whereas CA1 may be more involved in comparing the current environment with representations that have undergone long-term consolidation such as hours or perhaps days (see Lee, Hunsaker & Kesner 2005). Evidence for these hypotheses may be provided by the perforant path input to CA3. Rolls (2010) suggests that the perforant path does not provide a sufficient amount of excitation as to direct efficient information storage within the CA3 network (Treves & Rolls, 1992). However, even though the perforant path has been shown to be unable to catalyze the network dynamics of the CA3 recurrent collaterals, there is culminating evidence that it may play a critical role in relaying the cues that initiate pattern retrieval in CA3 (Rolls, 2010; Treves & Rolls, 1992). For example, if the perforant path (PP) were to initiate retrieval, a numerically large input would be advantageous for CA3; the advantage of PP activity initiating retrieval is that a partial cue would be sufficient to assist the recurrent collaterals in taking over the retrieval process with as minimal stimulation as possible (Rolls, 2010). The PP could provide tonic excitation such that the recurrent collaterals have an overall increased sensitivity to novel stimuli.
Further evidence for some CA1 activity in match-mismatch recognition can be elaborated by the relationship between CA1/3. The primary efferent projections from the hippocampus originate from the CA1 region through the subiculum, with CA3 contributing via the perforant path (Witter, 1993). However, the cytoarchitecture of the hippocampus allows for CA3 to completely bypass region CA1 (Gilbert, Kesner & Lee, 2001; Gilbert & Kesner, 2002; Swanson & Cowen, 1977). As suggested by Gilbert et al. (2001) and Gilbert and Kesner (2002), information could be passed through the CA3 extrahippocampal connections. CA3 has direct projections to the medial and lateral septal nuclei (Amaral & Witter, 1995). The lateral septum has connections with the medial septum (Jakab & Leranth, 1995), and the medial septum has efferent projections to the subiculum and EC (Amaral & Witter, 1995; Jakab & Leranth, 1995). As reviewed by Hunsaker et al. (2009), the differential targets of the CA3 and CA1 subcortical efferents suggests that there may be functional heterogeneity between the two regions. CA3 projects to the septofimbrial nucleus, throughout the rostral lateral septum, the dorsal portion of the medial septum, and the dorsal-most portions of the vertical limb of the diagonal band of Broca. The CA1 efferents project through the dorsal fornix to the lateral septum, medial septum and diagonal band of Broca and terminate in the mammillary bodies and other structures along the Papez circuit (for review see Swanson & Cowen, 1977). A number of CA1 lesion studies have been conducted to more thoroughly examine the CA1-CA3 relationship. Lee and Kesner (2002) found a dissociation between CA1 and CA3 in control animals, which may suggest that CA1 does not act as an intermediary in the transport of information from CA3 to neocortex and that the functioning of the two areas may be dissociable. In sum, CA1 may serve some function in the match-mismatch process. However, it is unclear based on current literature whether CA1 principal cells adjust their firing rate in response to novel stimuli, or are simply responsive to the mismatch signal from adjacent networks, such as DG-CA3 (see Fyhn, Molden, Hollup, Moser & Moser, 2002).
Save et al. (1992) developed a behavioral paradigm to examine the roles of the anterior and posterior regions of the posterior parietal cortices (aPPC, pPPC) and the hippocampal formation (HIP) in the match-mismatch process. In this paradigm (described below in the Methods section) rats were assessed on their ability to detect changes in object-place associations and the detection of a novel object placed amongst previously habituated objects. Habituation refers to a progressive decrease in exploration due to frequent exposure to unchanged stimuli. The results suggested that, compared to animals with control lesions (CNT), (a) habituation of both locomotor and exploratory activity was displayed equally across all lesion groups (CNT, HIP, aPPC and pPPC); (b) CNT and aPPC subjects reexplored displaced objects, whereas pPPC and HIP failed to significantly respond to the spatial change; and (c) introduction to a novel object elicited significant reexploration from all groups except pPPC lesioned animals. The indications of the study were threefold. First, the study supported a large body of literature indicating that the HIP is responsible for the processing of spatial information and the identification of environmental changes (Barnes, Nadel & Honig, 1980; Bizon, LaSarge, Montgomery, McDermott, Setlow & Griffith, 2009; Gallagher, Burwell & Burchinal, 1993; LaSarge, Montgomery, Tucker, Slaton, Griffith, Setlow & Bizon, 2007; Poe, Teed, Insel, White, McNaughton & Barnes, 2000; Robitsek, Fortin, Koh, Gallagher & Eichenbaum, 2008). Second, results from Save and colleagues (1992) supplemented existing studies that suggested a functional relationship between the pPPC and the visual cortices based on their anatomical proximity (i.e., a lesion to pPPC would disrupt tasks requiring visuospatial discriminations whereas subjects with aPPC lesions were not significantly impaired; see McDaniel & Wall, 1988). Finally, a progressive decrease in object exploration occurred across sessions regardless of lesion location, indicating that habituation occurs equally between lesion groups in rodents. A recent study by Lee et al. (2005) used the previously mentioned paradigm (Save et al., 1992) to examine the match-mismatch process in rats with selective lesions to the subregions of the hippocampus (DG, CA1 or CA3). The data indicated that compared to control animals (CNT), (a) habituation occurs equally across groups (CNT, DG, CA1 and CA3); (b) the detection of spatial novelty (familiar objects moved to a novel location) was significantly impaired in subjects with DG or CA3 lesions, whereas CA1 lesions only produced “moderate deficits” (while the exploration of previously habituated objects remained the same among lesion groups); and (c) the detection of a novel object was not significantly affected by subregional hippocampal lesions.
Given the results of Save et al. (1992) and Lee, Hunsaker, and Kesner (2005), tasks utilizing natural exploratory behavior have been shown to be particularly useful and efficient paradigms for studying memory, specifically this match-mismatch process. It has been shown that normal animals tend to recognize changes in a familiar environment by increasing exploration of novel stimuli or habituated stimuli moved to a novel location, relative to unchanged/habituated stimuli (Lee, Hunsaker & Kesner 2005). However, rodents with damage to the hippocampus (Honey et al., 1998; Save et al., 1992) or its subregions (Lee, Hunsaker & Kesner, 2005) do not demonstrate this increased exploration of spatial changes in the environment, suggesting that the hippocampus may be a critical substrate in the processing of novel spatial information. A number of age-related changes have been reported in various regions of the hippocampal formation and within the intrinsic circuitry of the hippocampal processing network (Barnes, Jung, McNaughton, Korol, Andreasson & Worley, 1994; Geinisman, Detoledo-Morrell, Morrell & Heller, 1995; Wilson, Beckett, Barnes, Schneider, Bach, Evans & Bennett, 2006). As a result, healthy aging has been associated with impairments on tasks that rely on intact functioning of the hippocampus and its surrounding regions.