Abstract
The left and right rodent hippocampi are functionally lateralized with respect to memory. Though theories of bilateral hippocampal function are beginning to incorporate hemispheric lateralization, there is a lack of data concerning how interhippocampal communication contributes to memory. One hypothesis suggests that synaptic plasticity in left CA3 facilitates acquisition and storage of new information, while synaptic stability in right CA3 is ideal for rapidly producing spatial representations. Convergence of these inputs in bilateral CA1 may bind memories to locations. Loss of interhippocampal communication would then spare the ability to acquire, store and retrieve memories, but would impair the binding of information and events to particular locations. To test this hypothesis in male and female mice, we performed split-brain surgery to transect the ventral hippocampal commissure, which contains direct interhippocampal projections. Mice underwent surgery to section the hippocampal commissure and overlying corpus callosum (HC+CC), just the corpus callosum (CC), or neither (SHAM) and then underwent a battery of hippocampus-dependent behavioral paradigms. We found deficits indicative of impaired binding of events and information to particular locations in HC+CC mice only. Despite these deficits, hippocampus-dependent contextual fear memory was unaffected by HC+CC surgery. Moreover, CC mice did not show any deficits in these tasks. These data suggest that interhippocampal communication may be needed for the memory of events at certain locations, but not for contextual associative memory. We propose that consideration of hemispheric lateralization and interhemispheric communication is necessary to formulate a more comprehensive understanding of hippocampal memory processes.
Significance Statement Hemispheric asymmetries in memory encoding and retrieval are well established in humans and rodents. However, less is known about the role of inter-hemispheric communication per se in memory function. Here, we studied “split-brain” mice in which we severed interhemispheric pathways connecting the left and right hippocampi, structures essential for spatial and episodic memory. Mice with transected inter-hippocampal pathways showed deficits in spatial, but not contextual memory, indicating that integration of the left and right hippocampus is required for spatial memory, but not hippocampus-dependent associative memory. We compare models of bilateral hippocampal function and propose that hemispheric asymmetry in function and interhemispheric communication may specifically endow the hippocampus with the ability both to store memories and to guide navigation.
Acknowledgments
We would like to thank Jeff Beeler and Dan McCloskey for discussions involving experimental design and Josè Esquivelzeta for advice on split-brain surgeries. This work was funded by PSC CUNY grants 65449-00-43, 66404-00-44 to CLP; CUNY Doctoral Student Research Grant and the Mina Rees Dissertation Fellowship to JTJ
Introduction
Damage to the hippocampus bilaterally produces profound amnesia for episodic, spatial and contextual memories in mammals (Scoville & Milner, 1957; Morris et al., 1982; Kim & Fanselow, 1992). Like many brain regions underlying complex cognitive processing, the human hippocampus is functionally lateralized with task-dependent hemispheric specializations. Interestingly, an understanding of rodent hippocampal asymmetries has lagged behind that in humans.Increasingly, it appears that that some, but not all, hippocampus-dependent memory tasks are functionally lateralized (Klur et al., 2009; Shinohara et al., 2012; Shipton et al., 2014)., However, it is not clear how the interaction between the left and right hippocampi play a role in memory.
The rodent hippocampus was long considered to be functionally symmetric due to an absence of evidence for hemispheric lateralization. For example, lesions of the left and right hippocampi in rats produced similar behavioral deficits on the Morris Water Maze (MWM) (Fenton & Bures, 1993), a widely used task to assess hippocampus-dependent learning and memory. Interestingly, an elegant study by Klur et al. (2009) reported that while indeed both hippocampi are needed for the MWM, they offer distinct contributions. Inactivation of the left hippocampus during acquisition prevented formation of an engram as evidenced by amnesia on a probe trial, while right inactivation during acquisition produced no such amnesia. Conversely, following intact acquisition, inactivation of the left hippocampus during a probe trial had no effect, while right inactivation prevented accurate searching. These data suggest complementary roles for the left and right hippocampi in, possibly being involved in different stages of engram formation, storage, and retrieval, necessary for spatial memory and navigation.
Behavioral lateralization may be reflected in anatomical asymmetries. Axons originating from left or right CA3 neurons project bilaterally to form synapses with CA1 neurons that have different properties (Kawakami et al., 2003; Shinohara et al., 2008; Kohl et al., 2011; Shipton et al., 2014; see El-Gaby et al. (2015) for review). CA1 dendritic spines in either hemisphere that receive left CA3 input are small and have a high density of GluN2B (Shinohara et al., 2008), a molecule associated with LTP induction (Lisman et al., 2012). CA1 spines in either hemisphere targeted by right CA3 are large, suggesting saturation of LTP, and have a high density of GluA1 (Shinohara et al., 2008). Consistent with these findings, optogenetic stimulation of left CA3 fibers produces LTP in left and right CA1, whereas stimulation of right CA3 fibers does not (Kohl et al., 2011; Shipton et al., 2014). It has been proposed that the convergence of left and right CA3 inputs in CA1 may be important for hippocampal function (El-Gaby et al., 2015). Specifically, El-Gaby et al. (2015) proposed that left CA3 alone may be sufficient for the storage of new learned associations, while input from right CA3 allows for the rapid emergence of cognitive maps to which new associations can be rapidly bound. According to this model, eliminating interhemispheric convergence of CA3 input in CA1 would impair performance on tasks that require binding of new memories stored in the left hemisphere to spatial representations of specific locations contributed by the right, while sparing performance on tasks that simply require the storage and retrieval of new memories (El-Gaby et al., 2015).
To test the role of inter-hemispheric communication in hippocampal-dependent memory and retrieval, we performed split-brain surgery on mice to sever pathways connecting the left and right hippocampi. These mice were trained on hippocampus-dependent behavioral paradigms, including the Y-Maze short-term memory task and the MWM, which require binding of memories to locations, as well as contextual fear memory and the elevated plus maze, which require no such binding. We then discuss our results in the context of bilateral hippocampal processing.
Methods
Animals
We used C57/BL6J mice (Jackson Labs, Bar Harbor, ME) that have been bred in-house for 2-5 generations. All mice were adults aged 2.5-7 months at the time of surgery. Mice were housed on a 12 h light/dark cycle, with all behavioral sessions occurring during the light phase.
Experimental Design
The ventral hippocampal commissure (VHC; Fig. 1) contains axons connecting the left and right hippocampi (Amaral & Lavenex, 2007). The term ventral hippocampal commissure is meant to distinguish this structure from the dorsal hippocampal commissure, which connects extra-hippocampal cortical areas in the left and right hemisphere, and should not be confused with the dorsal/ventral distinction used to describe hippocampal function (Kheirbek et al., 2013). Fibers in the VHC originate and terminate throughout the entire dorsal-ventral extent of the hippocampus. We performed “complete” or “partial” split-brain surgeries in mice. Complete split-brain surgery consisted of transection of both the VHC and the overlying corpus callosum (Fig. 1B, 2C), as the VHC cannot be accessed without transecting the corpus callosum. Partial split-brain surgery consisted of transection of only the corpus callosum located over the VHC (Fig. 2B) to control for possible contributions of the corpus callosum to hippocampus-dependent memory (Zaidel, 1995). Sham surgeries consisted of sectioning cortex overlying the corpus callosum (Fig. 2A). We refer to mice receiving complete split-brain surgery as HC+CC (n = 14; 6 females, 8 males), mice receiving partial split-brain surgery as CC (n = 10; 6 females, 4 males), and sham-operated mice as SHAM (n = 9; 4 females, 5 males). There were no sex differences in any behavioral measure, therefore males and females were combined for all analyses. Sample sizes of mice used were corrected following histological confirmation of surgery and then varied slightly by behavioral test, see individual sections in Results.
Surgery
In order to sever interhemispheric pathways, we modified a method developed by Schalomon & Wahlsten (1995). We used an L-shaped, sharpened piece of tungsten wire (0.25 mm in diameter) as a knife. Temperature of the mice was monitored and maintained via a temperature probe and heating pad. Mice were anesthetized with a ketamine xylazine cocktail (i.p., 90-120 mg/kg and 5-10 mg/kg, respectively) and further anesthetized for 1 minute in an isofluorane chamber (4.5% isoflurane). Mice were placed in a stereotaxic apparatus, receiving a constant flow of 1.5% isofluorane and oxygen (1.5 L/minute) and were given an injection of bupivacaine (1.25-2 mg/kg) under the scalp for local analgesia. An opening was made by drilling two adjacent 1 mm-wide holes into the skull to access the brain. To avoid the superior sagittal sinus, openings in the skull were made 0.5 mm off the midline and the side of surgery for each animal was randomly chosen (±0.5 ML, −0.8 AP from bregma for the first hole, ±0.5 ML, −1.6 AP from bregma for the second hole). To sever both the HC and CC, the short, sharpened end of the L-knife was placed on the surface of the brain along the medial side of the hole and was then slowly lowered 3.4 mm. Once lowered, the knife was translated anteriorly so that the knife moved posterior to anterior to “hook” the HC and CC fibers. The knife was then raised 3.4 mm until the short arm reached the underside of the skull. The knife was then translated back and raised out of the hole. To sever the CC only, we performed the same procedure as HC+CC transection, however the knife was only lowered 2.2 mm. For sham surgeries, we used the same procedure, but the knife was lowered 1.0 mm. Mice were administered buprenorphine following surgery (0.1 mg/kg SC). Transected and sham mice were indistinguishable by observing their behavior in the homecage.
Behavior
Behavioral testing began about four weeks after surgery. Mice were habituated to handling for one day by the experimenter. On testing days, mice were transported to a designated behavior room and were allowed to acclimate for a minimum of 20 minutes before the start of the task. All behavior was scored by an experimenter blind to surgical condition and sex using the Stopwatch+ program.
Short-term spatial memory was measured using the Y-Maze, known to be equally sensitive to inactivation of either the left or the right hippocampus (Shipton et al., 2014). The Y-maze apparatus was constructed of clear acrylic and had three arms (height: 20 cm; length: 30 cm; width: 8 cm) 120 degrees apart. The room contained many spatial cues including light fixtures and furniture. In addition, a painting and a movie poster were placed on the walls in line with the axes of the familiar and novel arms, while the experimenter stood along the axis of the start arm during each trial. Each arm was marked with a black line at the entrance for determining whether the mouse was in the arm or not. Mice were considered to be in an arm if all four paws were across the entrance line. The paradigm consisted of a 2-minute encoding trial during which one arm was blocked off, followed by a 1-minute intertrial interval, then a 2-minute retrieval session (Fig. 4A). The start arm remained the same in both the encoding and retrieval trial, while the exposed arm during the encoding trial was considered the familiar arm and the blocked arm was considered the novel arm. At the start of the encoding trial, mice were placed facing outward in the start arm and were allowed to explore the start and familiar arms for two minutes, beginning when the mouse left the start arm. Mice were removed from the Y-Maze and placed back into their home cage for one minute. While mice were in the home cage, the block was removed to expose the novel arm. To remove any potential confounds from odor cues, the apparatus was wiped with 70% ethanol, rotated 120 degrees, and then wiped with a dry paper towel before the retrieval trial. After the intertrial interval, the mice were again placed facing out in the start arm and were allowed to explore the entire maze for two minutes. At the end of the retrieval session, mice were placed back in their home cages and the maze wiped and dried before the next animal was run. Y-Maze spatial memory was scored as the time spent in the novel and familiar arms during the retrieval paradigm.
Long-term spatial learning and memory was measured using the Morris Water Maze (MWM), following the protocol of Vorhees & Williams (2006). The pool was 110 cm in diameter and was filled with opaque water colored with non-toxic white paint maintained at a temperature of 25.3°C ± 0.5°C. The escape platform was white and was submerged approximately 0.5 cm under the surface of the water. The platform remained in the same location throughout training. Salient room cues were visible from the surface of the pool and included colored and patterned posters, lighting, and furniture. Training consisted of four trials per day for five days with the starting location varying on each trial. Intertrial intervals were 30 seconds, during which the mice remained on the platform before starting the next trial. Mice that did not reach the platform within 60 seconds were placed onto the platform. Twenty-four hours after training, a 60-second probe trial was given during which the escape platform was removed. To measure spatial learning, latencies to the escape platform were recorded for each training trial and were averaged across trials for each mouse on each of the 5 training days. For the probe trials, spatial memory was scored by time spent in the target quadrant versus the average of the times spent in non-target quadrants, computed by summing the time spent in non-target quadrants divided by three (following Teixeira et al., 2006; Arruda-Carvalho et al., 2011; Cancino et al., 2013).
Hippocampus-independent learning was assessed using a version of the MWM in which the escape platform was visible, as described by Vorhees & Williams (2006). Mice were tested in the same pool as in the spatial paradigm. However, the platform was above water level and a red disk was placed on top to contrast the platform with the white pool and water. Training consisted of three trials per day over five days. Mice were placed in the water facing the wall on the opposite side of the pool from the target platform. After finding the platform, mice were left for 15 seconds before being moved to the home cage for the intertrial interval and the platform was moved to a new spatial location. The intertrial interval had no set time and ended when the platform was moved and the water settled (approximately 30 seconds). If mice did not find the platform within 60 seconds, they were placed onto it by the experimenter and remained for 30 seconds.
Short- and long-term contextual fear memory was tested using a one-shock conditioning paradigm that is particularly sensitive to hippocampal manipulations (Wiltgen et al., 2006). Conditioning took place in a fear conditioning chamber housed in a sound-attenuating cubicle (Med Associates, Fairfax, VT). Before each session, the fear chamber was wiped down with 70% ethanol and dried. On the first day, mice were placed in the fear chamber and allowed to explore freely for 3 minutes, then given a mild foot-shock (2 s, 0.75 mA), and removed 15 seconds later (total conditioning session time = 3 minutes, 17 seconds). Two 3-minute retrieval sessions occurred 1 and 24 hours after conditioning (Fig. 6A), a protocol previously used to dissociate molecular contributions to short-and long-term contextual fear memory (Schafe et al., 1999). Fear expression was scored as time spent freezing during each minute of the three-minute retrieval trial (absence of all movement, except breathing).
Anxiety was measured using the elevated plus maze, a paradigm sensitive to both dorsal and ventral hippocampal manipulations (Kjelstrup et al., 2002; Kheirbek et al., 2013). The elevated plus apparatus consisted of four arms (30.5 cm long, 6.4 cm wide), two of which were enclosed on three sides with walls (20.3 cm high). Mice were placed in the center of the apparatus and were allowed to explore freely for 5 minutes (Fig. 7A) and were then returned to their home cage. The apparatus was wiped with 70% ethanol and dried both before and after each trial. Anxiety was scored as the time with all four paws on an open arm.
Tissue Processing and Surgical Verification
Mice were euthanized with 0.3 mL of euthasol. Brains were extracted and post-fixed in 4% paraformaldehyde. Brains were then cryoprotected in 30% sucrose before cryosectioning at 60 µm. To assess transection to the hippocampal commissure, we stained sections with luxol blue and cresyl violet. Slides were dried overnight at 37°C. Histology began with a de-fat step in which sections were serially dehydrated and then placed in xylene (twice for 5 minutes each). Sections were then rehydrated and then incubated in 70% ethanol for one hour at room temperature. Sections were then incubated in a 0.1% luxol blue solution in 95% ethanol overnight at 56°C. Myelin was differentiated via rinses in deionized water, followed by 0.05% lithium carbonated in deionized water, followed by 70% ethanol (2 minutes each; differentiation was repeated as necessary). Sections were then stained in 0.1% cresyl violet in deionized water, serially dehydrated, cleared in xylene and then coverslipped using Krystalon.
Statistical Analysis
To assess short-term memory during the Y-Maze retrieval trial, we planned comparisons for whether each treatment group exhibited a preference for the novel arm over the familiar arm, but not whether treatment groups differed in total time exploring these arms as all retrieval sessions were 2 minutes in duration. Therefore, we compared time spent in the novel arm to time spent in the familiar arm within each group using Student’s paired t-tests. To assess cued and spatial learning on the MWM, we performed a two-way mixed model ANOVA (treatment x training day, with training day as the repeating factor) on escape latency, followed by post hoc tests (Tukey’s HSD). We used Student’s paired t-tests within each treatment group to assess spatial memory on the 60-second MWM probe trial comparing time spent in the target to time spent in an average of the other quadrants (Teixeira et al., 2006; Arruda-Carvalho et al., 2011; Cancino et al., 2013). We used a one-way ANOVA to determine whether minimum swim latencies differed across groups on the final day of visible MWM training. To assess short- and long-term contextual fear memory, we performed a two-way mixed model ANOVA (treatment x minute, with minute as the repeating factor) on levels of freezing during each of the two retrieval trials. To assess anxiety during the elevated plus maze test, we performed a one-way ANOVA on time spent in the open arms across all groups.
Results
Histology
Four mice (2 male, 1 female HC+CC; 1 female CC) were excluded from analysis because surgeries missed the HC or CC fibers. Sham surgeries spared interhemispheric pathways (Fig. 2A) in all SHAM mice. Partial and complete split-brain surgeries resulted in severing the corpus callosum (Fig. 2B, C). Complete split-brain surgeries sectioned the hippocampal commissure such that the HC either remained completely severed at the time of confirmation (Fig. 2C) or resulted in a clear scar across the hippocampal commissure (Fig. 3). Due to the position of the VHC (Figure 1), damage to nearby structures, such as the fornix, fimbria, and septal nuclei likely occurred. However, in all cases we verified that these structures were not completely destroyed (Fig. 2C, 3).
Y-Maze
We performed a short-term spatial memory version of the Y-Maze (Fig. 4A) and measured time spent exploring the familiar arm versus the novel arm during the retrieval trial. One male HC+CC mouse was removed from analysis for failure to leave the start arm during the retrieval trial and therefore had scores of zero for both the novel and familiar arms. SHAM and CC mice both exhibited a preference for the novel arm over the familiar arm during the retrieval session (SHAM: t(8) = 3.502; p = 0.008; CC: t(8) = 4.641; p = 0.002; paired t-tests) mice, whereas HC+CC mice showed no such preference (t(9) = 0.205; p = 0.842; paired t-test, Fig. 4B).
Morris Water Maze
Mice were trained on a spatial version of the MWM in which the escape platform was hidden below the surface of the water and could be found by the use of distal spatial cues. One female SHAM mouse was removed from analysis as it showed signs of hypothermia following a training session.
Spatial acquisition was assessed by measuring the average escape latency on each day over the five days of training (Fig. 5A). We found a main effect of training day (F(4,25) = 8.304, p < 0.0001) and of treatment group (F(2,25) = 4.268, p = 0.025) on escape latency during acquisition, with no interaction between these factors (F(8,25) = 0.513, p = 0.845). Post hoc analyses confirmed that HC+CC mice were significantly slower on training day 4 compared to the other groups (HC+CC vs. SHAM: p = 0.026; HC+CC vs. CC: p = 0.036). CC mice did not differ from SHAM mice at any point during acquisition (p > 0.79 for each day).
Twenty-four hours after the end of acquisition, a 60-second probe trial was conducted in which the escape platform was removed and time spent searching in the target quadrant versus time spent searching in others was measured (Fig. 5B). SHAM and CC mice searched selectively, exhibiting a preference for the target quadrant over others (SHAM: t(7) = 3.924, p = 0.003; CC: t(8) = 2.480, p = 0.038). However, HC+CC mice did not appear to distinguish between the target quadrant and others (t(10) = 1.320; p = 0.216).
To rule out potential effects of split-brain surgery on vision, locomotion, or non-hippocampus-dependent procedural learning, mice were trained on a visible version of the MWM in which mice could see the platform throughout the duration of each trial (Fig. 5C). No mice were removed from this analysis. We found a main effect of training day on escape latency (F(4,26) = 98.1, p < 0.0001), but no effect of treatment (F(2,26) = 0.024, p = 0.976), and no interaction between these factors (F(8,26) = 0.766, p = 0.6332). As an additional test, we compared shortest latencies at the end of training across groups. Minimum escape latency for each mouse on the final day of training revealed no effect of treatment (F(2,26) = 0.632, p = 0.539).
Contextual Fear Memory
To determine whether surgical treatment affected contextual fear memory, we used a one-shock contextual fear conditioning protocol that is sensitive to hippocampal manipulations (Wiltgen et al., 2006). Specifically, lesions of the dorsal hippocampus produce anterograde amnesia of contextual fear memory when only a single training trial is given (Wilgten et al., 2006). We conducted retrieval sessions 1 and 24 hours after conditioning (Fig. 6A). There appeared to be no impairment of either short- or long-term contextual fear memory in HC+CC or CC mice. There was no effect of treatment (F(2,26) = 0.69; p = 0.511) or of minute (F(2,26) = 2.29; p = 0.111) on freezing during the 1-hour delay short-term retrieval test (Fig. 6B) and no interaction between these factors (F(2,26) = 1.15; p = 0.344). There was a trend towards an effect of treatment (F(2,26) = 2.66; p = 0.089) on freezing during the 24-hour delay long-term retrieval test (Fig. 6C), though this difference did not appear to suggest any memory impairment in split-brain groups compared to SHAM mice, but instead indicated greater fear expression in CC mice. There was an effect of minute (F(2,26) = 3.46; p = 0.038) but no interaction between the two factors (F(2,26) = 0.37; p = 0.839).
Elevated Plus Maze
A previous study suggested that split-brain mice may have higher levels of anxiety which may contribute to impaired performance on a Barnes Maze spatial memory test (Shinohara et al., 2012). To determine whether split-brain surgery affected anxiety, we exposed mice to a single session of exploration of an elevated plus maze (Fig. 7A). One male HC+CC mice was removed from analysis as it fell off the apparatus during the session. Reduced exploration of the open arms of the maze would suggest increased anxiety. Surgical treatment did not appear to alter levels of anxiety as there was no effect of treatment on exploration of open arms in the elevated plus maze (F(2,25) = 0.867, p = 0.433; Fig. 7B).
Discussion
We found that sectioning the hippocampal commissure resulted in spatial learning and memory deficits, but did not affect contextual fear memory, anxiety or hippocampus-independent learning. HC+CC mice showed impaired short-term memory on a spatial Y-Maze task with a 1-minute interval between encoding and retrieval, impaired acquisition of a hidden platform in the MWM, and a lack of selective searching during a probe trial occurring 24 hours after the end of training. In contrast, CC mice did not show impairments in any task. HC+CC surgery sectioned the VHC, which contains direct projections between the left and right hippocampi, but is also situated near the fornix, fimbria, and septum. Though we cannot rule out damage to these structures as a factor, our histology showed a sparing of fornix and fimbria fibers in our HC+CC mice. Further, complete ablation of the fimbria and fornix produces anterograde amnesia for contextual fear (Maren & Fanselow, 1997) and inactivation of the lateral septum impairs the expression of conditioned contextual fear (Reis et al., 2009), neither of which appeared to occur in HC+CC mice. Therefore, we attribute the effects of HC+CC surgery to a loss of interhippocampal communication. Interestingly, our data do not support a role of the corpus callosum in hippocampal memory, as has been proposed in humans (Zaidel, 1995); however, in our study portions of the corpus callosum both rostral and caudal to the VHC were spared. Our data indicate a potential role of the mouse hippocampal commissure in some forms of memory, similar to data reported in humans (Phelps et al., 1991), though the function of the human hippocampal commissure has been debated (Wilson et al., 1987; Gloor et al., 1993; Rosenzweig et al., 2011).
Shinohara et al. (2012) found slower spatial learning in mice with both the VHC and corpus callosum sectioned. However, these mice were also monocularly deprived, which may have contributed to this finding. Contextual fear was found to be impaired in genetic split-brain mice, however these mice showed impaired hippocampal synaptic transmission (Schimanski et al., 2002) and baseline differences in freezing (MacPherson et al., 2008) that potentially confound the results. Further, it is not clear if genetic split-brain mice have similar hippocampal asymmetries as has been demonstrated in wild-type mice (Shinohara et al., 2008; Kohl et al., 2011; Shipton et al., 2014).
The left hippocampus may be a locus of engram formation (Klur et al., 2009; Shipton et al., 2014; El-Gaby et al., 2016), which is not surprising given its capacity for synaptic plasticity (Kohl et al., 2011; Shipton et al., 2014). Reversible inactivation of the left but not the right hippocampus of rats during acquisition of the MWM resulted in non-selective searching during a probe trial with both hippocampi intact. This suggests that the left hippocampus is necessary for the establishment of a long-term memory trace (Klur et al., 2009). Consistent with this idea, optogenetic silencing of left but not right CA3 impaired learning on a long-term Y-Maze test of spatial learning (Shipton et al., 2014; El-Gaby et al., 2016). However, it is not clear whether right CA3 participates in memory storage, or whether it contributes to memory-guided behaviors in a complementary way. Inactivation of the right hippocampus during learning of the MWM did not impair probe retrieval (Klur et al., 2009). However, inactivation of the right hippocampus during a probe trial after learning prevented selective searching in the target quadrant. This was interpreted as a role of the right hippocampus in spatial memory retrieval. In this model, spatial memories acquired by the left hemisphere are transferred to the right for long-term storage and retrieval. Conversely, it has been suggested that the contribution of the right hippocampus to spatial memory may degrade over time. Spatial working memory on the T-Maze was impaired by inactivation of either CA3 but was impaired to a greater degree by inactivation of right CA3 (Shipton et al., 2014). Additionally, Shipton et al., (2014) found that inactivation of both the left and right CA3 impaired performance on a short-term spatial memory version of the Y-Maze (as was used in the present study). The right CA3 was not required for a long-term 11-day Y-Maze spatial learning task. This is reminiscent of an fMRI study in humans that found degradation of right hippocampal activation correlating with the remoteness of an autobiographical memory, while left activation invariantly increased (Maguire & Frith, 2003). The right hippocampus in humans is, however, required for accurate spatial navigation in a learned environment (Spiers et al., 2001). Thus, the function and contribution of the right hippocampus to memory is currently less well understood than that of the left.
One model of the bilateral hippocampus suggests that left CA3 acquires and stores new memories via its capacity for synaptic plasticity (Shinohara et., 2008; Kohl et al., 2011; Shipton et al., 2011) and that right CA3 provides spatial representations via stable neural networks that were preconfigured during development (El-Gaby et al., 2015). Rapid emergence of bilateral CA1 cognitive maps in new environments would be contributed from right CA3. These maps could then be modified via spatial learning that engages left CA3, allowing the binding of acquired memories and learned associations to these spatial representations (El-Gaby et al., 2015). Such modifications may establish new place cells that would integrate into networks with left CA3, reducing the contribution of right CA3 spatial representations to memory over time (consistent with data reported by Shipton et al., 2014). Interestingly, in a study that did not consider hemispheric lateralization, place cells in the left hemisphere accumulated near goal locations after learning, consistent with this model (Hollup et al., 2001). Our data also support this model proposed by El-Gaby et al. (2015), as eliminating the convergence of left and right CA3 input to CA1 preserved hippocampus-dependent associative memory in the contextual fear task, while impairing spatial learning and memory in the Y-Maze and MWM. One limitation of the model, however, is that it is not clear how it would account for the requirement of the right hippocampus when searching during the probe trial of a well-learned MWM, (as seen in Klur, et al., 2009). Klur et al., (2009) explained this result by suggesting that spatial memory engrams are transferred from the left hemisphere to the right. Our data from spatial memory tasks would appear to support the model proposed by Klur et al. (2009), however, the lack of impairment seen in contextual fear memory would indicate a lack of interhemispheric engram transfer at least for this type of engram.
Our data are consistent with the proposals of Klur et al. (2009) and El-Gaby et al. (2015) that the left hippocampus specializes in new engram acquisition. However, our findings do not support the idea of interhemispheric transfer of engrams (Klur et al., 2009), unless this process varies depending on memory type. Instead, our work supports a very similar model to that of El-Gaby et al. (2015) in which memories are stored via synaptic modification in the left hemisphere while the right contributes rapidly emerging spatial representations of the environment to which memories can be bound in CA1. The contribution of the right hippocampus to memory storage and retrieval decays rapidly after acquisition (Shipton et al., 2014; Maguire & Frith, 2003). We add to this model that the right hippocampus is always needed during spatial navigation to compute and continually update distance and direction from a goal location as position along the route progresses, consistent with an invariant necessity for bilateral hippocampus in the MWM (Teixeira et al., 2006), but not in contextual fear memory (Kim & Fanselow, 1992). Data in humans (Howard et al., 2014) and in bats (Sarel et al., 2017) suggest that direction and distance computations may indeed exist in the right hippocampus. Studies of spatial navigation in humans have suggested a principle contribution of the right hippocampus (Maguire et al., 1998; Spiers et al., 2001). Such a component would not be needed when performing the long-term Y-Maze task as the goal arm merely needed to be identified to obtain the goal without any need for continual updating of distance from and direction to the goal location (Shipton et al., 2014; El-Gaby et al., 2016). Thus, we argue that the developmental configuration of synaptically stable networks in the right hippocampus, which El-Gaby et al. (2015) have suggested allows for the rapid emergence of cognitive maps in new environments, also contributes route computation functions required during navigation in learned environments.
In summary, our findings support the view of El-Gaby et al. (2015) that CA1 binds memories acquired by left CA3 to spatial representations provided by right CA3 via preconfigured cell assemblies and add the proposition that spatial navigation will always require the right hippocampus. Thus, we suggest that hippocampal lateralization may be a solution to the problem of “knowing where and getting there” (Wishaw et al., 1995; Maguire et al., 1998), whereby memories of events occurring at particular locations are stored via synaptic modification in left CA3 while right CA3 is involved in route computation.
Footnotes
Conflict of Interest: The authors no declare no conflicts of interest.