We asked whether episodic-like memory requires neural mechanisms independent of those that mediate its component memories for “what,” “when,” and “where,” and if neuronal connectivity between the medial prefrontal cortex (mPFC) and the hippocampus (HPC) CA3 subregion is essential for episodic-like memory. Unilateral lesion of the mPFC was combined with unilateral lesion of the CA3 in the ipsi- or contralateral hemispheres in rats. Episodic-like memory was tested using a task, which assesses the integration of memories for “what, where, and when” concomitantly. Tests for novel object recognition (what), object place (where), and temporal order memory (when) were also applied. Bilateral disconnection of the mPFC-CA3 circuit by N-methyl-d-aspartate (NMDA) lesions disrupted episodic-like memory, but left the component memories for object, place, and temporal order, per se, intact. Furthermore, unilateral NMDA lesion of the CA3 plus injection of (6-cyano-7-nitroquinoxaline-2,3-dione) (CNQX) (AMPA/kainate receptor antagonist), but not AP-5 (NMDA receptor antagonist), into the contralateral mPFC also disrupted episodic-like memory, indicating the mPFC AMPA/kainate receptors as critical for this circuit. These results argue for a selective neural system that specifically subserves episodic memory, as it is not critically involved in the control of its component memories for object, place, and time.
Episodic memory refers to the recollection of an event in a particular time and place (Tulving 2005) and in animal models has been defined as an integrated representation of discrete memories for object (what) in place (where) and time (when) (Clayton and Dickinson 1998; Kart-Teke et al. 2006). Conceptually, episodic memory could be a specific and separated mnemonic process or it could merely be a consequence or sum of its component memory systems. One approach to answer this question is to ask whether impairments in episodic memory result from deficits in one or more of the individual component memories, or from a deficiency in the integration of these components, per se. If the integration of the individual component memories into episodic memory are merely a consequence of their sum (due to their simultaneous encoding), then a deficit in episodic memory would be expected to be accompanied by deficits in one or more of the component mnemonic information. In contrast, deficient episodic memory in face of intact memory components would suggests episodic memory to engage a process separated from its composing factors. Concomitantly, one can also ask whether episodic memory requires unique separate neural substrates or if it is dependent on the integrity of the individual neural mechanisms that subserve memory processing for “what,” “where,” and “when.”
The perirhinal cortex (PRC), hippocampus (HPC), and medial prefrontal cortex (mPFC) are crucial for processing information as to what, where, and when (Hannesson et al. 2004; Barker et al. 2007; Barker and Warburton 2011). Moreover, the interaction between mPFC and HPC was found to be essential for temporal order memory, but not for object memory, nor for memory of its location (Barker and Warburton 2011; DeVito and Eichenbaum 2011). Also, electrophysiological (Leutgeb et al. 2005; Komorowski et al. 2009; Neunuebel and Knierim 2014), neuroimaging (Cabeza et al. 2002; Hassabis et al. 2007), lesion (Ergorul and Eichenbaum 2004; Fortin et al. 2004), and genetics (Place et al. 2012) studies indicate the HPC to be a structure critical for the formation of episodic memory. Particularly, the HPC subregion CA3, with its specific anatomical structure forming a self-feedback circuit, has been proposed to be an autoassociative network for processing episodic memory (Treves and Rolls 1994; Rolls and Kesner 2006). Human neuroimaging findings showing higher neural activation of the CA3/dentate area when episodic memory is encoded (Eldridge et al. 2005), along with the failure of CA3 lesioned rats to exhibit episodic-like memory (Li and Chao 2008), support the critical role of CA3 region in episodic memory. In addition to the HPC, the PFC is also involved in the processing of episodic or episodic-like memory (Blumenfeld and Ranganath 2007; Spaniol et al. 2009; Li et al. 2011). Neuroimaging (Schott et al. 2011; Bonnici et al. 2012), electrophysiological (Watrous et al. 2013), and lesion (Barker and Warburton 2011) studies implicate interacting neural circuits between the PFC and medial temporal lobe, especially the HPC, in the control of episodic memory (Aggleton and Brown 2006).
In the present study, we asked whether an integrating mechanism that determines episodic memory exists independent of those that underlie the component memories for what, where, and when. We used an episodic-like memory test which assesses the integrated memory for object identity (what), object place (where), and temporal order of object presentation (when) (Kart-Teke et al. 2006), plus novel object recognition, object–place and temporal order tests to measure memory for these components individually (Dere et al. 2007). We targeted the circuit comprising the mPFC and hippocampal CA3 as a possible critical substrate for episodic memory. Initially, we found that disconnection of the mPFC and CA3 via contralateral N-methyl-d-aspartate (NMDA) lesions prevented the expression of episodic-like memory, but left the component memories for object, place, and temporal order intact. Since recent studies have implicated the participation of ionotropic glutamate receptors of the mPFC in object recognition memory (Barker and Warburton 2008, 2015), we sought to test whether glutamate receptors in the mPFC are critical components of this circuit. Accordingly, we injected AMPA/kainate- and NMDA receptor antagonists into the mPFC together with a contralateral CA3 NMDA lesion and tested the influence on episodic-like memory.
Materials and Methods
Male Wistar rats (Tierversuchsanlage, University of Düsseldorf, Germany) weighing between 250 and 300 g were used. They were grouped 5 per cage (60 × 38 × 20 cm) and housed under standard conditions under a reversed light–dark rhythm (light off from 07:00 to 19:00). They had free access to food and water through the whole studies. Each animal was handled for 3min per day for 5 days before behavioral testing. All experiments were in accordance with the Animal Protection Law of Germany and of the European Communities Council Directive (86/609/EEC).
NMDA was used as a neurotoxin for cell body lesions. The competitive NMDA receptor antagonist D-(−)-2-amino-5-phosphonopentanoic acid (AP-5) and the competitive AMPA/kainite receptor antagonist 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) were dissolved in 0.9% saline in the concentrations of 5 and 1 mg/mL, respectively. Drug solutions were prepared in the volume of 0.5 mL and preserved at −80°C until use. Drugs were purchased from Sigma Aldrich (Steinheim, Germany).
Animals were anesthetized with pentobarbital (Narcoren; Merial GmbH, Germany, 50 mg/kg, i.p.), and placed in a stereotaxic frame (David Kopf, Tujunga, CA, USA). A heating pad beneath their bodies was used to maintain normothermia. Incisions were made and the scalp was retracted to expose the skull. Holes were drilled into the skull over the mPFC and HPC subregion CA3. For the disconnection, NMDA solution was microinjected into the target sites with a 10-μL Hamilton syringe connected to a CMA 100 microinfusion pump. Surgical details are described below.
The disconnection approach relies on the assumption that, if a function is dependent on a neural interaction between 2 areas of the brain, then a unilateral lesion/inactivation of one area combined with a unilateral lesion/inactivation of the other in the opposite hemisphere, disconnects the circuit bilaterally at 2 different levels and thus, leads to a functional deficit. The same lesion/inactivation of the areas located in the same hemisphere spares an intact circuit in the contralateral hemisphere, in which case the function may be preserved in part (Geschwind 1965, 1965).
Experiment 1: Disconnection of the mPFC-CA3 Circuit
Animals received a unilateral NMDA injection into the mPFC combined with an ipsi- or contralateral NMDA injection into the hippocampal CA3 region. Thus, they were assigned into 2 experimental groups: mPFC + CA3 ipsilateral lesions (n = 7) and mPFC + CA3 contralateral lesions (n = 7). The neurotoxic agent NMDA (2 μg dissolved in 0.2 μL phosphate buffer saline, flow rate 0.1 μL/min) was injected into either the right or the left mPFC, counterbalanced within groups (coordinates relative to skull at bregma: AP = 3.0 mm; ML = ± 0.7 mm; DV = −3.0 mm; AP = 3.0 mm; ML = ± 0.7 mm; DV = −4.0 mm; AP = 4.0 mm; ML = ± 0.7 mm; DV = −3.0 mm) with a 26-gauge steel cannula. The cannula was left in situ for an additional 1 min to prevent reflux. Within the same surgical session, NMDA (2 μg dissolved in 0.1 μL phosphate buffer saline, flow rate 0.1 μL/min) was also injected into 3 locations of the ipsi- or contralateral hippocampal CA3 region (AP = −2.8 mm; ML = ± 3.0 mm; DV = −3.9 mm; AP = −3.3 mm; ML = ± 3.4 mm; DV = −3.9 mm; AP = −4.1 mm; ML = ± 4.2 mm; DV = −4.0 mm). After each injection, the needle was left in place for additional 2 min before retracting it. Then, the scalp was sutured and the wound was disinfected with 70% ethanol. The animals were allowed to recover for 2 weeks.
Experiment 2: The Role of Glutamatergic Receptors in the mPFC-CA3 Circuit
The surgical procedure was identical to experiments 1, except that the animals received a unilateral NMDA injection into the CA3 area and were implanted with an injection cannula (diameter 0.7 mm; length 16 mm) aiming at the contralateral mPFC (AP = 3.0 mm; ML = ± 0.7 mm; DV = −3.0 mm). A stainless steel thread (26-gauge, protruding 0.1 mm from the tip of the guide cannula) was inserted to prevent occlusion. They were secured to the skull by 2 stainless steel screws (2.6 mm diameter) and dental cement. The animals were allowed to recover for at least 7 days, before being assigned into AP-5 (n = 7) or CNQX (n = 7) injection groups.
A black-acrylic open field (60 × 60 × 30 cm) was used to assess object exploration. Illumination was provided by 2 LED lights (luminous density on the center ∼8 and ∼6 lx in the corners). A camera was mounted 2 m above the arena and connected to a computer and a DVD recorder. The open field was located in a sound-attenuated room and geometric figures (29 × 21 cm, black-white stripe and black circle in white background) were pasted around the room as spatial cues.
Different sets of objects in quadruplicate made of different material (glass and ceramic), shape (rectangle and irregular shapes), color (white and blue), and size (25–32 cm height and 7–11 cm diameter) were applied in the spontaneous object exploration tests. The objects were heavy enough to prevent a displacement by the animals. Different sets of objects were used for each object exploration test and the assignments of objects were counterbalanced for each subject. Object exploration was defined as a physical contact with the object with snout, vibrissae, or forepaws. Climbing on the object, or touching the object while looking around the environment, were excluded from this measure. Object exploration times were registered by experimenters blind to the experimental design with the Ethovision software 3.1 (Noldus, Wageningen, The Netherlands). Behavioral measurements were conducted between 10:00 and 17:00 h. Acetic acid solution (0.1%) was used to remove odor cues after each trial.
Episodic-like Memory Test
Animals were habituated to the open field by placing them into the center and allowing exploration for 10 min for 3 consecutive days. One day after the third habituation trial in the open field, the memory test was conducted according to the procedure described in detail elsewhere (Kart-Teke et al. 2006). This test was composed of 2 sample trials and 1 test trial, lasting 5min each and with an intertrial interval of 1h. Each animal was placed into the center of the open field, in which 4 copies of an object occupied 4 of 8 possible locations along the walls. The locations for placement of objects were randomly selected. One hour later, animals were again put into the arena into which 4 copies of a novel object were placed. Two of them were randomly placed at 2 of 4 possible locations, which were once occupied during the first sample trial, while the other 2 objects were randomly placed at 2 of 4 novel locations (i.e., locations that were not previously occupied). After another delay of 1h, each animal was put into the arena containing 2 copies of the object from the first sample trial, “old familiar,” and 2 copies of the object from the second sample trial, “recent familiar” objects. One object of each type was presented at a randomly chosen location where it was previously located during the respective sample trial. The other object of the same type was placed at a randomly chosen location, which was not previously occupied during the respective sample trial. Therefore, the test trial contained one old familiar object at a stationary location (OFS), one old familiar object at a displaced location (OFD), one recent familiar object at a stationary location (RFS), and one recent familiar object at a displaced location (RFD; Fig. 2A). Previous studies showed that rats explore the OFS object more than the OFD one along with exploring the RFD object more than the RFS one (Kart-Teke et al. 2006; Li and Chao 2008). This behavioral pattern suggests that there is an interaction between “place” and “recency,” which indicates that rats show the ability to simultaneously integrate “what–where–when” components (see also “dependent variables” section below).
Object–Place Recognition and Novel Object Recognition Tests
One day after the previous test, animals were tested for their memories for recognizing locations and objects. Each rat was placed into the open field where 2 copies of an object were located in the 2 corners of the apparatus. Animals were free to explore the objects and the arena for 4min and, then, were returned to their own cage. One hour later, each animal was placed into the center of the arena where the same objects were presented. Whereas one of the two objects was displaced at a novel location, the other one remained at the same location. Each animal had 4min to freely explore and, then, was returned to their cages. Rodents spend more time exploring the displaced object more than the one at the same location, suggesting that they recognize the displacement and, thus, show memory for place (Ennaceur et al. 1997). After a delay of 1h, each animal had 4min to explore the arena where 2 objects were presented. This time, one object was replaced by a novel one while the other one remained the same. These 2 objects were placed at the previously encountered locations and the location for replacing the novel object was counterbalanced. Rodents prefer exploring the novel object more than the old one, indicating that they can recognize the old object (Ennaceur and Delacour 1988).
Temporal Order Memory Test
One day after the previous tests, a test for temporal order was applied. This test was composed of 2 sample trials and 1 test trial. The animal was placed into the center of the open field where 2 copies of an object were presented in the 2 corners of the arena (sample 1). Each rat was free to explore the objects and the arena for 4 min. After a delay of 1h, the animal was put into the apparatus with 2 copies of a novel object for 4min (sample 2). One hour later, one of the objects used in sample 1 and one of the objects applied in sample 2 were placed together in the arena (test trial). Each animal was allowed to freely explore the objects and the apparatus for 4min. The places used for locating objects were identical in all the trials. During the intervals, the animals were returned to their cages. Rodents explore the object previously encountered in sample 1 more than the recent presented object from sample 2, indicating a memory for recency (Mitchell and Laiacona 1998).
In experiment 2, rats were habituated to manual restraint 3min per day for 3 consecutive days before testing. Fifteen minutes prior to the test trial for episodic-like memory, animals were restrained manually and the thread in the guide cannula was removed. The infusion cannula (26-gauge, protruding 0.05 mm from the tip of the guide cannula) was connected to a 10-μL Hamilton syringe with a flexible polyethylene tube (PE-10). A CMA 100 microinfusion pump was used to infuse 0.5 μL of AP-5 or CNQX solutions into the brain with a flow rate of 0.5 μL/min. The infusion cannula was left in situ for additional 1min for diffusion. The microinjection procedure was completed within 5 min. Only the episodic-like memory paradigm was conducted in this experiment.
For all the tests, the length of time the objects were explored during the test trial was taken as dependent variable. In objet exploration paradigms, rodents explore a novel object longer than an old one (Ennaceur and Delacour 1988), a displaced object longer than a stationary one (Ennaceur et al. 1997), and a previously encountered one (more distant in the past) longer than a more recently encountered object (Mitchell and Laiacona 1998). In the episodic-like memory paradigm, rats explore the OFS object longer than the RFS one, the RFD longer than the RFS one, but they explore the OFS longer than the OFD one (Kart-Teke et al. 2006). This exploration pattern indicates that rats show concomitantly memory for when (the temporal order of objects presentation), what (a particular object) was encountered where (the location of objects), since there is an inversion in the pattern of exploration between the OFS × OFD and RFS × RFD objects. This suggests that rats exhibit an integrated memory by combing what–where–when information into an episodic-like memory (Kart-Teke et al. 2006). For statistical comparison between the ipsi- and contralateral disconnection groups, 3 ratios were calculated:
Integration ratio = (time exploring the OFD object – time exploring the OFS object)/(time exploring the OFD object + time exploring the OFS object).
Where ratio = (time exploring the RFD object – time exploring the RFS object)/(time exploring the RFD object + time exploring the RFS object).
When ratio = (time exploring the OFS object – time exploring the RFS object)/(time exploring the OFS object + time exploring the RFS object).
All of the animals explored all of the arranged objects during the sample and test trials. Thus, none of them had to be excluded from data analysis.
After completion of behavioral testing, each animal was euthanized with an over dose of pentobarbital (80 mg/mL, i.p.). Animals were then perfused intracardially with 0.9% phosphate buffer saline, followed by 10% formalin, and the brains were extracted. Then, the brains were stored in 30% sucrose formalin at 4°C until processed. Coronal brain sections were cut with a cryostat (50 μm; Leica, Germany) and mounted onto presubbed glass slides. The slices were stained with cresyl violet (Sigma Aldrich, USA) to determine the extent of the lesions and the site of cannulae placement.
Mixed two-way ANOVAs with “object” as the within factor and “group” as the between factor were applied to analyze times for object exploration in all the tests. If significant effects of object and/or “object × group” were found, the following tests were applied. For the episodic-like memory test, two-way ANOVAs with the 2 within factors place and recency were conducted for each group separately. Paired t-tests were then used within each group when a significant effect of place × recency” interaction was found. One-sample t-tests were then used for the ratios to compare with chance-level performance (the zero value). To compare group differences, independent t-tests were used to analyze the ratios. For the NOR, OPR, and TOM tests, paired t-tests were also applied within each group when appropriate. As the preference for the direction of exploring is important for the interpretation, the paired and one-sample t-tests were used as one-tailed comparisons. All other statistical tests were two-tailed comparisons. The level of significance was set at P < 0.05.
Similar lesion extents were found in the mPFC and dorsal hippocampal CA3 area among different groups. For the mPFC lesions, the dorsal anterior cingulate, the prelimbic, and infralimbic areas were mainly damaged and also part of the secondary motor cortex. For the hippocampal subregion lesions, dorsal CA3 area was primarily damaged, while part of the primary somatosensory cortex and parietal association cortex were also influenced. The sites for cannula implantation were targeted at the medial PFC in the animals in experiment 2. One animal was excluded because of incorrect lesions (n = 1 from the AP-5 injection group). Details are shown in Figure 1.
Experiment 1: Disconnection of the mPFC-Hippocampal CA3 Prevented Episodic-like Memory, but Not Memories for What, Where, or When, Individually
There were significant effects of object (F3,36 = 4.14, P = 0.013) and object × group interaction (F3,36 = 8.497, P < 0.001), but not of group (P > 0.05) in a 2 (groups) × 4 (object) mixed two-way ANOVA. The following within factors 2 (place) × 2 (recency) ANOVA revealed a significant interaction effect (F1,6 = 20.835, P = 0.004), but no effects of place and recency (P > 0.05) in the ipsilateral lesion group. For the contralateral lesion group, there was a significant effect only of place (F1,6 = 10.987, P = 0.016), but neither of recency nor place × recency interaction (both P > 0.05). Paired t-tests indicated that the ipsilateral lesion group explored the OFS object more than the OFD one (t6 = 2.503, P = 0.023), the RFD object more than the RFS one (t6 = −4.84, P = 0.0015), and the OFS more than the RFS (t6 = 2.42, P = 0.026; Fig. 2B). This characteristic pattern of object exploration based on a distinct spatial-temporal context expresses the integration of what–where–when information into episodic memory (Kart-Teke et al. 2006).
In the comparison of the ratios, the animals with an ipsilateral lesion had significantly lower “integration” and higher where ratios than those with the contralateral lesion (P = 0.005 and P = 0.011, respectively), while no group difference was found in the comparison of the when ratios (P > 0.05; Fig. 2C). Whereas the ipsilateral lesion animals exhibited negative integration (t6 = −3.117, P = 0.01) and positive where and when ratios (t6 = 8.736, P < 0.001; t6 = 2.777, P = 0.016, respectively) compared with zero, the contralateral lesion group did not (P > 0.05). Hence, the contralateral lesion animals did not exhibit the integrated memory of what–where–when (namely, episodic-like memory), while the ipsilateral lesion group did.
In the analyses of object exploration in the OPR, NOR, and TOM tests, there were significant effects of object (F1,12 = 31.225, P < 0.001; F1,12 = 51.454, P < 0.001; F1,12 = 21.474, P = 0.001, respectively), but not group and object × group (P > 0.05). Both the ipsilateral and contralateral lesion groups explored the displaced object more than the stationary one in the OPR test (P = 0.0025 and P = 0.0065, respectively), the novel object more than the old one in the NOR test (P = 0.002 and P < 0.001, respectively), and the old-familiar object more than the recent-familiar one in the TOM test (P = 0.0125, P = 0.006, respectively; Fig. 2D). This indicated that both groups had intact memory for OPR, NOR, and TOM. Finally, there were no significant group differences in the total times the objects were explored in any of the test trials (Ps > 0.05; Table 1). Thus, the disconnection of the mPFC-CA3 circuit left the individual memories for what, where, and when, per se, intact, but impaired episodic-like memory.
|mPFC + CA3||Ipsi-||49.29 ± 4.36||48.60 ± 6.79||56.66 ± 7.71||37.80 ± 5.29|
|mPFC + CA3||Contra-||53.06 ± 5.59||41.34 ± 5.37||49.89 ± 5.06||27.11 ± 5.47|
|mPFC + CA3||Ipsi-||49.29 ± 4.36||48.60 ± 6.79||56.66 ± 7.71||37.80 ± 5.29|
|mPFC + CA3||Contra-||53.06 ± 5.59||41.34 ± 5.37||49.89 ± 5.06||27.11 ± 5.47|
ELM, episodic-like memory; OPR, object–place recognition; NOR, novel object recognition; TOM, temporal order memory; ipsi-, ipsilateral lesions; contra-, contralateral lesions.
Experiment 2: The Functional Circuit of mPFC-CA3 Is Dependent on the AMPA but Not NMDA Receptors in the mPFC for Integrating the What–Where–When Components
The 2 (group) × 4 (object) mixed two-way ANOVA revealed significant effects of object (F3,33 = 4.986, P = 0.006) and group × object (F3,33 = 6.801, P = 0.001) but not group (P > 0.05). The 2 × 2 ANOVAs with 2 within factors place and recency showed that a significant effect of “place × recency” interaction (F1,5 = 123.789, P < 0.001) was found, but no effects of place and recency (P > 0.05) in the AP-5 injection group. In contrast, there were significant effects of place in the CNQX injection group (F1,6 = 8.11, P = 0.029, respectively), but no effects of recency and place × recency interaction (P > 0.05; Fig. 3A). The animals with AP-5 injection preferred to explore OFS object more than OFD one (t5 = 6.991, P < 0.001), RFD object more than RFS one (t5 = −4.536, P = 0.003) and OFS object more than RFS one (t5 = 4.092, P = 0.0045; Fig. 3A). Thus, the animals exhibited episodic-like memory.
There was no group difference in total exploration times of objects (P > 0.05; Fig. 3B). Thus, application into the mPFC of CNQX, but not AP-5, disrupted episodic-like memory in the context of disconnecting the mPFC-CA3 functional pathway.
The interhemispheric disconnection of the mPFC and dorsal hippocampal CA3 subregion prevented episodic-like memory, but left the memories for what, where, and when, per se intact. The NOR, OPR, and TOM tests were conducted to test the hypothesis that the observed deficits in episodic-like memory could be caused by impairments of the individual what, where, or when memory components. Our results suggest that the establishment of an integrated episodic memory requires neural processes that are distinct from those that establish memory for its components. Therefore, we hypothesize that episodic memory requires the integrity of not only individual neural systems for processing information about objects and their location in “place” and “time,” but also an independent neural mechanism responsible for the “integration” of these component representations into a memory system. If the integration of the individual component memories into episodic memory were merely a consequence of their sum (due to their simultaneous encoding), then a deficit in episodic memory would be expected to be accompanied by deficits in one or more of the component memory systems. Conversely, deficient episodic memory in face of intact memory components, for which we have provided evidence, suggests episodic memory to engage a meta-system that is separate from its composing factors.
Our data also provide evidence that episodic-like memory is dependent on a functional interaction between the mPFC and CA3 area, but that object and spatial recognition, and temporal order memory of previously encountered objects are not dependent on such a functional interplay. Whereas the PRC, HPC, and mPFC are proposed to be cardinal neural substrates for the distinct what, where, and when memory systems (Hannesson et al. 2004; Barker et al. 2007; Barker and Warburton 2011), convergent findings suggest an interacting system between the PFC and HPC for the processing of episodic memory (Schott et al. 2011; Bonnici et al. 2012; Watrous et al. 2013). Here, we provide evidence that the neural circuits comprising the mPFC and CA3 region is inclusively, but not exclusively, an integral part of such a network.
Inspired by the anatomical features of the recurrent collaterals of the HPC CA3 area, several computational models have proposed that the CA3 acts as an autoassociation network for the formation and storage of episodic memory (Treves and Rolls 1994; Knierim et al. 2006; Rolls and Kesner 2006). Consistent with these models, higher neural activation of the CA3/dentate area was found during episodic memory formation imaged by high-resolution functional magnetic resonance (fMRI) (Eldridge et al. 2005). The CA3/dentate area in patients with mild cognitive impairment exhibited higher activity, suggesting that the episodic memory deficit in this population was related to changes of this area (Yassa et al. 2010). In rats, CA3 lesion disrupted episodic-like memory (Li and Chao 2008) and pharmacological inactivation of the CA3 area was found to disrupt episodic retrieval (Zhou et al. 2012).
Evidence for the involvement of the PFC in episodic encoding and retrieval is also provided by neuroimaging studies (Blumenfeld and Ranganath 2007; Spaniol et al. 2009). In episodic memory tasks, aged humans scanned by fMRI showed stronger PFC activity as compared with young ones, implying that the PFC compensated for age-related cognitive deficits (Cabeza et al. 2004; Rajah and D'Esposito 2005). Rats with mPFC lesions showed disrupted retrieval of a specific what–where–when” contextual event in an episodic-like memory test featuring fear conditioning (Li et al. 2011). In an episodic-like memory paradigm, similar to the one we applied here, mice with lesion of the mPFC exhibited a memory deficit akin to the loss of the integration ratio we describe here, indicating that the mPFC is responsible for memory of where an event happened (DeVito and Eichenbaum 2010). Taken together, these results indicate that the CA3 area and mPFC are likely to be cardinally involved in episodic memory.
Anatomically, the HPC CA3 subregion indirectly projects to the mPFC through the HPC CA1 area (Thierry et al. 2000; Vertes 2006). The mPFC sends projections back to the HPC via other regions, such as the nucleus reuniens (NRs) (Xu and Sudhof 2013) and the lateral entorhinal cortex (LEC) (Apergis-Schoute et al. 2006). Both the mPFC–NR–HPC and the mPFC–LEC–HPC circuits might participate in episodic memory, as suggested by evidence that the NR modulates contextual memory (Xu and Sudhof 2013) and that the LEC is critical for associations of object/context (Wilson, Langston et al. 2013; Wilson, Watanabe et al. 2013). A recent review also proposed that the LEC provides information to the HPC as to what happened in the environment (Knierim et al. 2014). Long-term potentiation may provide a mechanism for connecting the HPC and PFC (Laroche et al. 2000), while electrophysiological findings suggest that the mPFC biases object–location memory retrieval from the HPC (Navawongse and Eichenbaum 2013). The EC has been considered to process a more “general” memory representation of interrelated information, while the HPC enhances the information in “detail” further through pattern separation and pattern completion processes (Morris and Frey 1997; Viskontas et al. 2009). Since, the LEC anatomically projects to both the CA1 and CA3 areas, while the CA3 projects to CA1 via the Schaffer collaterals (van Strien et al. 2009), one can hypothesize that global information for object/context associations from the LEC reaches the CA1/CA3 regions and then, integrative detailed information, processed by CA3, is sent to the CA1. Hence, CA1 should also contain an episodic memory representation, as electrophysiological studies evidently showed that CA1 (Takahashi 2013) or CA1/CA3 (McKenzie et al. 2014) neurons respond in a specific manner to episodic-like memory traces. Recent studies have shown higher CA1 neural activation in processing spatial and nonspatial memory representations, while CA3 activation is mainly related to the processing of spatial information (Beer et al. 2013; Beer et al. 2014). The CA1 region, which is not principally recruited by spatial-related information like the CA3, may play a role in processing mnemonic information in a broader extent. When facing different environments, rodent CA1 neurons fired with a substantial overlap across environments, whereas CA3 neurons showed little overlap (Leutgeb et al. 2004). When the locations of proximal and distal cues in a single room were rotated, CA3 place fields remained coherent, whereas a remapping appeared in CA1 neurons (Lee et al. 2004). Thus, small or large geometrical context alternations may determine the way CA1 and CA3 neurons respond (Vazdarjanova and Guzowski 2004). The interplay between the mPFC and HPC, in terms of the comparative neural properties of CA1 and CA3, in the processing of episodic memory is still unexplored.
In experiment 1, a potential confounding could be that the contralateral lesioned rats exhibited intact memory in the OPR, NOR, TOM tests but were impaired in the episodic-like memory test because the environment was “simpler” concerning the amount of information (2 objects instead of 4) and would be less demanding in terms of attentional efforts or processing complexity, especially because the mPFC is known to be involved in attentional processes (Cassaday et al. 2014; Riga et al. 2014). Although one cannot completely rule out the possibility that, in our results, the contralateral lesioned rats exhibited control-level performance in the OPR, NOR, TOM tests because less attentional efforts are involved, it is very unlikely that the lesion, per se, would have caused such deficit, since in both ipsilateral and contralateral disconnection groups, only one mPFC (right or left) was lesioned, but the ipisilateral disconnected group exhibited intact memory in the OPR, NOR, TOM, and episodic-like memory tests. Likewise, unilateral mPFC lesion did not impair performance in OPR, NOR, TOM, and object-in-place tests (Chao et al. 2013). Other factors, e.g., motor, perceptual, or motivational functions, can likely be excluded as potential confounds because the total durations of object exploration in all the test trials were not significantly different. Finally, animals with sham lesion of the mPFC and CA3 showed comparable episodic-like performances to nonlesioned ones (Li and Chao 2008; Li et al. 2011).
Our findings also show that the critical functional interplay between the mPFC and CA3 area that permits an integrated episodic memory was found to rely on AMPA/kainite, but not NMDA receptors in the mPFC (Fig. 3). When CNQX was injected into the contralateral, but not into the ipsilateral mPFC before the test trial, episodic-like memory could not be observed, implicating a role for AMPA/kainate receptors in the recall of the previously learned information. These findings are compatible with previous studies showing that microinjection of CNQX, but not AP-5, into the mPFC prior to the probe trial compromised previously and newly learned paired-association memories (Tse et al. 2011). Also, the lack of effect of the pretest trial AP-5 injection in preventing the expression of episodic-like memory is corroborated by findings in an object exploration paradigm for testing memory of object–place associations, in which microinjection of AP-5 into the mPFC of rats had no effect when administrated before the test trial (Barker and Warburton 2008, 2015). Selective roles for hippocampal NMDA and AMPA receptors have also been proposed for associative memory encoding and retrieval, respectively (Morris 2006). In a paired flavor-place memory paradigm, the injection of CNQX, but not of AP-5, into the HPC prior to testing impaired performance (Day et al. 2003; Bast et al. 2005). The underlying mechanisms could be that CNQX infusions influenced fast synaptic transmission, while AP-5 infusions did not (Day et al. 2003; Bast et al. 2005). AMPA receptor insertion was also found to contribute to neural plasticity driven by experiences (Takahashi et al. 2003).
Our results 1) delineate a neural circuit that encompasses the mPFC AMPA/kainite receptors and the CA3 region as critical components of a system that determines the integration of memories for object, place, and temporal order into episodic memory and 2) imply that this integrative system for episodic memory is distinct from those employed in the representation of its component memories.
This study was supported by grant SO 1032/2-5 and Heisenberg Fellowship SO 1032/5-1 from the Deutsche Forschungsgemeischaft to M.A.S.S.
Conflict of Interest: None declared.