When a rat runs through a familiar environment, the hippocampus retrieves a previously stored spatial representation of the environment. When the environment is modified a new representation is seen, presumably corresponding to the hippocampus encoding the new information. The medial septum is hypothesized to modulate whether the hippocampus engages in retrieval or encoding. The cholinergic agonist carbachol was infused into the medial septum, and hippocampal CA1 place cells were recorded in freely moving rats. In a familiar environment, septal activation impaired the retrieval of a previously stored hippocampal place cell representation regardless of age. When the environment was changed, medial septal activation impaired the encoding process in young, but facilitated the encoding of the new information in aged rats. Moreover, the improved encoding was evident during a subsequent exposure to the modified environment 24 h later. The findings support the role the septum plays in modulating hippocampal retrieval/encoding states. Furthermore, our data indicate a mechanism of age-related cognitive impairment.
Age-related memory decline is associated with hippocampal dysfunction (Rapp and Amaral, 1991; Jenkins et al., 2000; Zhang et al., 2007). Hippocampal pyramidal cells (“place cells”) fire strongly when the rat is in a particular location in the environment (“place field”). Place fields typically remain in the same location across repeated exposures to an environment (O'Keefe and Nadel, 1978; Thompson and Best, 1990). This “map” of the environment can remain constant despite the removal (O'Keefe and Speakman, 1987; Markus et al., 1994) or exchange of stimuli (Tabuchi et al., 2003), suggesting that a stored representation of the environment is repeatedly evoked despite slight changes in the environment (McNaughton and Morris, 1987). When the environment or task requirements are altered, place cells can “remap” (i.e., change their spatial tuning or stop firing). Unlike the recall of a previously stored representation, remapping indicates encoding or establishing a new representation of the environment (Frank et al., 2006).
In aged rats, place fields are as stable as those observed in young animals within a given recording session in a familiar environment (Markus et al., 1994; Oler and Markus, 2000; Wilson et al., 2004, 2005). However, under conditions that normally induce remapping in young rats, less remapping is found in aged animals (Oler and Markus, 2000; Wilson et al., 2004). This rigidity of the hippocampal representation may underlie the age-related impairments in spatial processing and episodic memory formation (Wilson et al., 2004).
Although major changes in the environment can induce a switch from retrieval of a representation to encoding a new representation, the mechanisms underlying the transition are unknown. The current study examined the hypothesis that the medial septum (MS) modulates the transition between retrieval and encoding states by regulation of hippocampal theta (4–12 Hz field oscillation). Theta has been related to exploration (Buzsaki, 2002) and increased neuronal plasticity (Judge and Hasselmo, 2004), and may act to synchronize activity across different neural regions during the encoding phase of learning (Hasselmo et al., 2002).
Inactivation of the MS decreases hippocampal theta and spatial task performance (Givens and Olton, 1990; Markowska et al., 1995). Similarly, the cholinergic modulation of the hippocampus is reduced during aging in rats (Gill and Gallagher, 1998), as well as humans (Bartus et al., 1982). Conversely, intraseptal cholinergic drugs increase hippocampal theta power (Lee et al., 1994; Givens and Olton, 1995; Markowska et al., 1995), and can promote memory performance in scopolamine-treated or aged memory-impaired rats (Givens and Olton, 1995; Markowska et al., 1995; Sabolek et al., 2004a), but not in young rats (Markowska et al., 1995; Frick et al., 1996; Bunce et al., 2003).
The present study examined whether increased MS activation (by intraseptal carbachol, a cholinergic agonist) would promote remapping, indicating the encoding or establishment of a new representation of the environment within the hippocampus. Our results indicate that in a familiar environment, in which the hippocampus normally retrieves a stored representation, MS activation interfered with the recall process. In a novel configuration, in which aged animals normally show reduced encoding, MS activation enhanced encoding.
Materials and Methods
The subjects were seven young adult (6–10 months old) and eight aged (22–23 months old) Fisher-344 male rats. Rats were single housed in transparent plastic tubs, in a room with a 12 h light/dark cycle. All experimental procedures were conducted in accordance with the guidelines of the University of Connecticut Animal Care Committee.
All animals were food deprived to 85–90% of their ad libitum weight. Before starting the training for the electrophysiological recordings, the rats were trained to traverse runways for chocolate sprinkle reward. The majority of rats were also trained to perform a mixed working and reference memory task on a radial arm maze for 2 weeks. Animals showing robust running were subsequently trained on the alternation task.
After finishing adaptation training, rats were placed on an automated runway and trained to alternate for food reinforcement (Noyes Precision Pellets, New Brunswick, NJ) between the ends of two arms of a seven-position maze (Fig. 1A, black arms). The rats were trained for 30 min a day until they reached a criterion of 80 trials (i.e., traversing the runway in one direction) per day on at least 2 consecutive days of training. The maze was located in the recording room, where several large stable extramaze cues were available (door, chairs, lamp, etc.).
After reaching criterion, rats were anesthetized with a ketamine mixture (i.m.; ketamine, 12.5 mg/ml; acepromazine, 0.25 mg/ml; xylazine, 1.25 mg/ml). Each rat received initial doses of 49.6 mg/kg ketamine, 0.4 mg/kg acepromazine, and 5.08 mg/kg xylazine, and additional boosters as needed (total dosage given did not exceed double the initial dose). They were implanted with a cannula aimed at the MS (0.5 mm anterior, 0 mm lateral, 4.5 mm ventral from bregma), as well as a microdrive for EEG and single-unit recordings aimed at the dorsal hippocampus (3.2 mm posterior, ±2.2 mm lateral, 1 mm ventral from bregma). The microdrive was composed of one 60 μm tungsten wire for EEG recordings and eight movable tetrodes (four in each hemisphere) for hippocampal unit recordings (Wilson and McNaughton, 1993). Each tetrode comprised four polyamide-insulated 14 μm nichrome wires (H. P. Ried, Palm Coast, FL) that were twisted together. Rats received 0.1 mg/kg meloxicam (Metacam; Boehringer Ingelheim, St. Joseph, MO) per os for 4–5 d after surgery.
After recovery from surgery, animals were retrained on the alternation task. The electrodes were slowly lowered into the CA1 region of the hippocampus, until single units could be isolated. Single-unit data were collected when the animal sat quietly in its home cage outside the maze room, and also on the maze, when the rat was performing the alternation task, in either a familiar or novel maze configuration.
Recordings were performed in four stages: (1) Holder preinfusion data were recorded while the rat was sitting quietly in a home cage outside of the recording room. (2) Maze preinfusion data were collected for 9 min while the rat was running on the maze. The recording was stopped and the rat was returned to its home cage, where a sham or drug infusion (saline or carbachol) was performed. During the break, the maze was carefully cleaned with 30% alcohol to remove odor cues. (3) Maze postinfusion data were collected while the rat ran the familiar alternation task for 9 min. (4) Holder postinfusion data were collected while the rat sat in the home cage for 10 min (Fig. 1A).
For each rat, up to five novel configuration recordings were conducted by moving one of the maze arms to a new location (one of the gray arm positions in Fig. 2A). The same recording apparatus and room were used for the familiar and novel configuration recordings. Recordings in a novel configuration were performed as follows. (1) Holder preinfusion data were recorded while the rat was sitting quietly in its home cage outside of the recording room. (2) Maze preinfusion data were collected while the rat ran 20 trials on familiar configuration on the maze (Fig. 2A, F and F* arms). The recording was stopped and an infusion was performed. The maze was cleaned with 30% alcohol. (3) Maze postinfusion data were collected while the rat ran a novel configuration for 12 min. (4) Holder postinfusion data were collected while the rat sat in the home cage for 10 min. The procedure was repeated 24 h later (in the same preinfusion and postinfusion arm configurations), at which time a sham infusion was performed.
The sequence was as follows. The configuration that the rat had been exposed to during training was 1–6 (the first familiar configuration), and the first two novel configurations were obtained by rotating the F* arm by 51.4° to one of its adjacent positions (i.e., 5 or 7). The next novel configuration of the maze the rats experienced was 1–3 (154.2° rotation), and always occurred in a control (saline or sham) recording. This allowed us to compare the effects of a small configurational change with those of a bigger change. Subsequently, the rats were trained to run in 1–3 configuration for several days, so that it became familiar (the second familiar configuration). Two more novel recordings were obtained on arms adjacent to arm 3 (i.e., 2 and 4). The novel configuration was obtained either by moving one arm by either 51 or 154° (to other side of the maze). There was no differential effect of the large or small rotation on any of the place field characteristics (all p > 0.1); therefore, the data were pooled.
Data recording and analysis
Electrodes were lowered slowly into the brain (<500 μm daily), until hippocampal single units could be identified. Hippocampal units were recorded when the waveforms appeared stable (and at least 30 min after moving an electrode). During the recording session, the animals wore a multichannel headstage device that was connected to the recording apparatus through a fine wire bundle, which did not limit the animal's freedom of movement. The location and the head direction of the animals were recorded by an overhead tracking system, based on the position of two arrays of infrared and light emitting diodes that were attached to the headstage. The signal from the tetrodes was amplified 2000–5000 times (Neuralynx, Tucson, AZ), filtered between 300 Hz and 6 kHz, and stored for off-line analysis. Individual units were isolated using a spike parameter cluster separation method. The clustering was based on the spike duration and the relative amplitude of the signals on the four electrode wires composing a tetrode (O'Keefe and Reece, 1993). The recording environment was divided into a 64 × 64 bin array consisting of 0.6 × 0.6 cm squares. Firing rate maps were constructed for each cell using an adaptive smoothing method (Skaggs and McNaughton, 1998). Place fields were defined as an area of at least 15 bins sharing adjacent edges, with a firing rate per bin >2 SDs above the mean firing rate of the cell on the entire apparatus.
Place cells are directionally selective on linear track mazes (Markus et al., 1995). Accordingly, analyses of place fields were performed separately for inward and outward headings. Rate map correlations were calculated only for those regions of the apparatus that were visited at least three times at a minimum speed of 2.4 cm/s in both rate maps. The velocity filter was used to ensure that all data were recorded while the rat was moving at a robust speed (i.e., in theta).
The identification of the type of cell (i.e., complex spike or interneuron) was based on the firing rate (<2.5 Hz for complex spike cells) and the spike width (>0.256 ms). Furthermore, only place cells with a minimum firing rate of 0.25 Hz and showing spatial selectivity (i.e., at least one field) were included in the analyses. For both the familiar and novel configurations, two types of analyses were performed as follows:
Firing and place field characteristics, such as mean firing rate, in-field firing rate, spatial information per spike, and sparsity. The spatial information per spike (information per spike) was calculated in terms of the spatial information content (in bits) a single spike conveys about the animal's location. Spatial information content of spike discharge was calculated using the following formula: information content = ΣPi(Ri/R)log2(Ri/R), where i is the bin number, Pi is the probability of occupancy of bin i, Ri is the mean firing rate for bin i, and R is the mean firing rate of the cell. The sparsity measures the fraction of the environment in which a cell is active (Skaggs et al., 1996).
Place field stability was calculated by measuring the correlation between the firing rate maps of a cell before and after infusion, or across different intervals of a recording. To analyze the timeline of the place field changes, the 9 min recordings in the familiar configuration were divided into three intervals (a = 0–3 min, b = 3–6 min, c = 6–9 min), and the 12 min recordings in a novel configuration were divided in four intervals (a = 0–3 min, b = 3–6 min, c = 6–9 min, d = 9–12 min). A prerequisite for performing a rate map correlation was a minimum of one place field (based on criteria above) during at least one of the intervals of a recording. The correlations were computed only if the cell fired at least 10 spikes in each interval and only for locations that were visited at least three times during the considered interval. The different rate maps were compared using Pearson correlations, computed on a bin-by-bin basis.
At the end of the maze recordings, the effect of the septal carbachol infusion on hippocampal EEG was assessed as the rats were drinking (a behavior associated with low theta power) (Buzsaki et al., 1983; Bland, 1986; Stewart and Fox, 1990). Data for each rat were recorded over five sessions. Recordings were performed while the rats were drinking chocolate (Nesquik) in their home cage, and the power in the theta band was compared before and after carbachol infusion. The signal was amplified 200 times and filtered between 1 Hz and 3 kHz. The data were stored for off-line analyses using MatLab software. Theta epochs during drinking were detected by calculating the ratio of the theta (5–10 Hz) and delta (1–4 Hz) frequency bands in 3.0 s windows (Csicsvari et al., 1999). Ten 3 s epochs (30 s total) with similar power in the delta band were analyzed for each recording session. The energy in the 6–12 Hz (theta) band was calculated for each 3 s epoch and averaged across the 10 epochs to obtain the average theta power for each recording; then, the five recordings from each rat were averaged.
All data were analyzed using parametric tests. The place field characteristics for the familiar recordings were analyzed using repeated-measures ANOVAs, with session (preinfusion and postinfusion) as the repeated factor, and age and drug as the between-subjects factors. For each day of novel recordings, the place field characteristics were analyzed using a two-way ANOVA (age by drug). The rate map correlations across different intervals of a recording were compared using two- or three-way ANOVAs. In addition, post hoc or t tests were performed to confirm the cause of significant interactions. Only effects that were significant (p ≤ 0.05) or showed a trend for significance (0.05 < p < 0.1) are presented. All nonsignificant effects had p > 0.1.
Carbachol was dissolved in saline to a concentration of 0.25 μg/μl, resulting in a total dose of 0.125 μg. Infusers 1 mm longer than the cannula were connected through polyethylene tubing to a 10 μl Hamilton syringe. A Harvard Apparatus (Holliston, MA) 505 microinfusion pump was used to generate a constant flow of 0.125 μl/min. Rats were infused for a total of 4 min; the injector was left in place for 1 more minute to allow for diffusion. A total volume of 0.5 μl was infused. The order of drug administration (carbachol and saline) was randomized between rats and infusions were separated by at least 3 d, to allow for recovery. We also performed sham infusions as additional controls and to prevent additional tissue damage. The procedure was similar, with the difference that an infuser shorter than the cannula was used, and no fluid was injected. An analysis of the place field characteristics of cells recorded in the familiar and novel configuration of the maze showed no difference between saline or sham infusions (all p > 0.1); therefore, the data from saline and sham infusions were pooled.
At the end of the recordings, the rats were killed with CO2 and transcardially perfused with 0.9% saline, followed by 3.7% paraformaldehyde. The brains were removed and stored in paraformaldehyde for at least 48 h. Coronal sections (50 μm) were cut using a vibratome, and then stained with thionin. Only rats with a good placement of the MS injector were included in the carbachol analysis. Electrode locations in the CA1 area were identified based on histological placement and electrode advancement records. Injector locations and electrode positions are shown in Figure 3. A summary of the number of cells recorded from each rat is presented in Table 1.
The effect of MS infusions of 0.125 μg of carbachol on hippocampal EEG was assessed while the animals were drinking chocolate-flavored water in their home cage (a behavior associated with low theta). There was a significant increase in theta power during drinking postinfusion, compared with the preinfusion EEG, for both young and old rats (t tests, t(6) = 3.282, p = 0.017; t(5) = 3.297, p = 0.03, respectively), although no age effect was observed (p > 0.1) (Fig. 4). Carbachol infusions into the MS also caused a decrease in running speed in the familiar, but not the novel maze configuration (supplemental Fig. 1, available at www.jneurosci.org as supplemental material).
The data were analyzed by cell; in addition, the data were collapsed and analyzed by animal. The by-animal analysis showed similar results despite reduced statistical power (supplemental Table 1, available at www.jneurosci.org as supplemental material).
Age and carbachol effects in the familiar configuration
Place cells were recorded as the rats alternated between feeders at the end of two maze arms (Fig. 1A). This procedure provided information on the hippocampal representation of the familiar configuration of the maze. Rats were then removed from the maze room and given a carbachol or control infusion, after which they were placed back on the maze. Examples of preinfusion and postinfusion CA1 place field activity are shown in Figure 1B–E.
The firing rate (Fig. 5A) was higher in the postinfusion recording session, as revealed by a session effect [repeated-measures ANOVA (RM-ANOVA), F(1,304) = 28.362, p < 0.001]. In addition, firing rates were higher in the young animals (F(1,304) = 10.211, p = 0.002). The age by session interaction was not significant (F(1,304) = 3.477, p = 0.063). This effect of age on CA1 neurons firing rate has been found in some studies (Shen et al., 1997), but not in others (Oler and Markus, 2000; Wilson et al., 2005). There were no other age-related differences in the familiar configuration (all p > 0.1) (Table 2).
Carbachol disrupted the place-specific firing of cells in both young and old rats. Carbachol did not affect the mean firing rate, but decreased the in-field firing rate of neurons, particularly in young rats (RM-ANOVA, session by drug interaction, F(1,304) = 26.534, p < 0.001; t test, t(52) = 2.945, p < 0.005 for young rats; p > 0.1 for old). Carbachol also decreased the spatial information carried by each spike (RM-ANOVA, session by drug interaction, F(1,304) = 46.547, p < 0.001; t test, t(52) = 5.996, p < 0.001 for young rats; t(66) = 4.735, p < 0.001 for old rats) (Fig. 5B). The total firing area (i.e., sparsity) was larger after carbachol than after control infusions (session, F(1,304) = 26.076, p < 0.001; session by drug interaction, F(1,304) = 21.133, p < 0.001; also t tests comparing carbachol to control, t(52) = 4.303, p < 0.001 for young, and t(66) = 4.487, p < 0.001 for old rats) (Table 2). These results suggest that the firing characteristics of hippocampal CA1 neurons were mostly unaffected by aging, and that carbachol decreased the spatial tuning of CA1 neurons.
The correlation between the firing rate maps before and after infusion was calculated for each cell to assess the stability of the place field representation (Fig. 5C) (also Fig. 1B–E). Under control conditions, the correlation was high for both old and young rats (correlations of r = 0.75, and r = 0.7, respectively). Carbachol decreased the correlations (ANOVA, F(1,304) = 26.779, p < 0.001; t test for young rats, t(153) = 3.408, p = 0.001; t test for old rats, t(171) = 3.933, p < 0.001). There was no age difference and no age by drug interaction with respect to pre–post infusion correlations (ANOVA, both p > 0.1; t tests comparing young and old rats for the control and carbachol conditions, p > 0.1), which suggests that carbachol decreased the stability of fields similarly in both age groups. When the data were analyzed by 3 min intervals, we found that place fields remained unstable over the whole duration of the postcarbachol recording (see supplemental Results and supplemental Fig. 2, available at www.jneurosci.org as supplemental material).
Together, the data suggest no age-related differences in the firing properties in a familiar configuration. Septal activation with carbachol decreased the spatial tuning and stability of place fields for both young and old rats.
Carbachol effects during the first exposure to a novel configuration
Place cells were recorded as the rats ran in a novel configuration of the maze (Fig. 2A).
Running in a novel configuration resulted in decreased place-specific firing compared with running on the familiar maze configuration (see supplemental Results and supplemental Fig. 1, available at www.jneurosci.org as supplemental material). This decrease could be attributable to place fields appearing, disappearing, or changing their firing location over the course of the recording.
Under control conditions, neurons of aged rats had lower spatial information content per spike (t test, t(137) = 4.41, p < 0.001) (Fig. 6B), higher sparsity (t(137) = −3.727, p < 0.001), and lower in-field rate (t(137) = 2.031, p = 0.043), suggesting that old rats had less specific place fields than young rats in the novel configuration. A two-way ANOVA revealed that the mean firing rate was higher in the aged rats (F(1,251) = 4.774, p = 0.03), and there was no effect of carbachol or interaction between age and carbachol (Fig. 6A). Carbachol decreased the in-field rate in young rats (age by drug interaction: F(1,251) = 9.688, p = 0.002; t test, t(136) = 2.985, p = 0.003). The spatial information content was lower for old rats than for young rats (ANOVA, F(1,251) = 9.249, p = 0.003); although there was a significant age by drug interaction (F(1,251) = 3.948, p = 0.048) (Fig. 6B) for the spatial information content, post hoc analyses showed that carbachol did not significantly change the information content for either young or old rats. The spatial sparsity was higher for old rats than for young rats during the first exposure to the novel configuration (age effect on sparsity, F(1,251) = 9.87, p = 0.003), and carbachol did not affect the sparsity (all p > 0.1) (Table 3). Together, these results suggest that carbachol had limited effects on the firing properties of CA1 neurons in a novel configuration, and young rats were more affected than old rats.
To examine the temporal dynamics of the place field response, the 12 min of novel configuration recording were divided into four intervals (a = 0–3 min; b = 3–6 min; c = 6–9 min; d = 9–12 min), and place field correlations were computed between adjacent intervals (Fig. 7A,B) (for examples, see Fig. 2B). In young animals after control infusions, the correlation was high and did not change over the course of the entire recording (ANOVA, p > 0.1). Conversely, place fields of old rats were initially unstable and became more stable over the course of the recording. After control infusions, the a to b correlation was lower for aged rats than for young rats (ANOVA, F(1,142) = 4.342, p = 0.039), although it continued to improve over the course of the recording (ANOVA, F(2,167) = 5.205, p = 0.006). For young rats, the stability of fields was lower after carbachol than after control infusions (ANOVA, F(1,481) = 21.854, p < 0.001), particularly in the first 6 min in the novel configuration (t test for ab interval pair, t(157) = 4.336, p < 0.001; for cd interval pair, t(159) = 2.06, p = 0.041). For old rats, carbachol tended to increase the place field correlations (ANOVA, F(1,293) = 3.476, p = 0.063), specifically during the beginning of the recording (t test for ab interval pair, t(96) = 1.852, p = 0.067). Therefore, carbachol had little effect on the basic firing properties of cells, while altering the place field dynamics. There was an age-related differential effect: carbachol decreased the stability of fields in the young rats, whereas it tended to increase the stability of fields for the old rats.
No significant correlation between the running speed and the firing rate maps correlations
As previously noted, carbachol increased the latency per trial for young rats during the alternation task in the novel configuration. Therefore, the low place field stability during the first 3 min of the postcarbachol infusion in young rats could be attributable to the decreased running speed at the beginning of postinfusion recording. To investigate this, we calculated Pearson correlations of the number of trials completed in the first 3 min to the cells' firing map correlation values for the ab interval pair. All correlations were nonsignificant and <0.203; therefore, the decreased stability of place cells of young rats after carbachol infusions does not seem to be the result of changed running speed (Fig. 8).
To summarize, the spatial firing characteristics and the place field stability data suggest that place cells of old rats show a smaller and slower response to the novel configuration; carbachol impairs place-specific firing in young rats, but facilitates it in old rats.
Second exposure to the novel configuration
One day after the initial exposure to the novel configuration, the rats were run through the same sequence for a second time, but no carbachol was infused (Fig. 9). During the reintroduction to the novel maze configuration, the mean firing rate (Fig. 6A) was differently affected, depending on age and whether carbachol was administered during the initial exposure to the maze (age by drug interaction, F(1,188) = 4.966, p = 0.027). However, t tests showed no significant difference in firing rate between rats that received control or carbachol infusions, for either young or old rats. A two-way ANOVA for the in-field firing rate of pyramidal cells also yielded a significant interaction of drug and age (F(1,188) = 5.881, p = 0.016). The interaction was mainly attributable to a lower in-field firing rate in young rats that received carbachol the day before, compared with the young rats that received control infusions (t test, t(136) = 2.985, p = 0.003), whereas the old rats had similar mean firing rates regardless of the type of infusion. The main effects (age and drug) were not significant for the mean and in-field firing rates. Together, the effects on the firing rates suggest that firing of pyramidal neurons was more spatially disorganized in young rats than in old rats after carbachol infusions (i.e., lower in-field:mean firing rate ratio).
The spatial information content tended to be higher (F(1,188) = 3.143, p = 0.078) (Fig. 6B) and the sparsity tended to be lower (F(1,188) = 3.09, p = 0.08) in rats that received carbachol the previous day. There was no age difference or age by drug interaction for spatial information content and sparsity (all p > 0.1) (Table 4).
As with the first day in the novel configuration, the 12 min of the second exposure to the novel configuration were divided into four intervals, and correlations of the firing rate maps were computed between adjacent intervals. As can be seen in Figure 7, C and D, the manner in which the novel configuration was initially experienced (i.e., after carbachol or control infusions) affected the stability of fields during the subsequent exposure to the same configuration. Young rats that received carbachol during the first day of novel exposure had less reliable place fields on the subsequent exposure; conversely, if aged rats experienced the novel configuration under carbachol, they showed more stable fields the following day (age by drug interaction, F(1,1073) = 4.701, p < 0.001; t test comparing the effects of carbachol and control infusions: t(539) = 3.837, p < 0.001 for young rats, t(531) = 7.881, p < 0.001 for old rats). For young rats that had carbachol the previous day, place fields were initially unstable; stability increased over the course of the recording, and it was similar to the control condition during the second half of the recording (ANOVA for ab interval pair, F(1,85) = 3.097, p = 0.089; for bc interval pair, F(1,91) = 9.0, p = 0.003; for cd interval pair, p > 0.1). For aged rats that received carbachol the previous day, stability of fields remained high throughout the recording session (ANOVA for ab interval pair, F(1,89) = 8.022, p = 0.006; for bc interval pair, F(1,87) = 12.539, p = 0.001; for cd interval pair, F(1,87) = 13.781, p < 0.001).
To summarize, the data demonstrate that, for young rats, MS activation during the first exposure to a novel configuration impaired place-specific firing of pyramidal neurons during a subsequent exposure to the same configuration. However, MS activation during the first exposure to a novel configuration improved the spatial firing of neurons in aged rats during the subsequent exposure.
Is the remapping during the novel configuration limited to the rotated arm?
To address this question, we correlated the firing of pyramidal cells during the familiar recording that preceded the arm rotation to the firing after the arm rotation. The correlations of the firing rate maps were calculated only for the familiar arm (i.e., the arm that remained in the same location in the novel configuration). During the first exposure to the novel configuration, the correlation of spatial firing on the familiar arm of the maze was lower than during the familiar recording, suggesting that the fields on the familiar arm were also affected by the arm rotation (Fig. 10A). This was true for both age groups, after control (t tests, t(162) = 6.878, p < 0.001 for young; t(158) = 7.255, p < 0.001 for old) and carbachol recordings (t(121) = 4.137, p < 0.001 for young; t(107) = 2.524, p = 0.013 for old). The correlation of spatial firing on the familiar arm of the maze before and after the arm rotation was not different between cells recorded after control and carbachol infusions, for either the young or the old rats (both p > 0.1), suggesting that carbachol did not affect the amount of remapping on the familiar arm. However, old rats showed higher correlations after arm rotation than young rats (F(1,247) = 4.802, p = 0.029), which suggests that the place fields were more similar after arm rotation for old than for young rats. Also, the similarity of place fields on the familiar arm was higher than zero after arm rotation (0.328 ± 0.022 for young; 0.459 ± 0.025 for old); therefore, the remapping on the familiar arm was not total, but partial.
During the second day of exposure to the novel configuration, the correlation of place fields on the familiar arm before and after arm rotation was still lower than in a familiar environment for young rats (t tests, t(135) = 5.741, p < 0.001 for control; t(96) = 2.505, p = 0.014 for carbachol). For old rats, the correlation of place fields on the familiar arm after arm rotation was lower than in a familiar environment after control infusions (t(143) = 4.492, p < 0.001), but not after carbachol infusion (p > 0.1) (Fig. 10B). Moreover, the old rats that received carbachol before the first exposure to the novel configuration showed a trend for increased correlation of the firing on the familiar arm from the first to the second exposure to the novel (t(91) = 1.746, p = 0.084), whereas there was no difference in correlation for the young rats or the old rats that received control infusions. Similar to the first exposure to the novel configuration, the correlation of fields before and after arm rotation was higher for old rats than for young rats (F(1,189) = 9.347, p = 0.003).
To investigate the timeline of changes on the familiar arm of the maze, we divided the novel recording into four 3 min intervals and calculated the correlation of the firing maps for each novel interval to the firing map of the familiar maze. We found that the correlation did not change over the course of the novel recording for either the first or the second exposure to the novel configuration (all p > 0.1), suggesting that the changes in place-specific firing occurred fast, within the first 3 min of exposure to the novel configuration.
This paper examines how the MS modulates the place cell remapping in the hippocampus. Placing the animals into a familiar environment presumably determined the retrieval of a previously stored representation of that environment. Introduction to a novel configuration of the environment induced the formation of a new representation. MS activation (with carbachol) had different effects under familiar and novel conditions. In young animals, carbachol always caused instability of the hippocampal representation, whereas in aged animals carbachol caused instability in a familiar environment but improved the development of a representation under conditions of novelty.
Familiar environment versus novel configuration
Our results are in agreement with previous findings that in a familiar/stable environment place fields remain stable across repeated exposures to that environment (O'Keefe and Nadel, 1978; Thompson and Best, 1990). Place fields were highly correlated between two sessions in the environment despite the animals being removed from the room and undergoing control infusions. Experiencing a novel configuration of the maze resulted in a rapid remapping of place fields, with even the familiar section of the maze showing some remapping (Fig. 10). These results are in agreement with previous reports of remapping in response to changes in the environment (Cressant et al., 2002), task requirements (Oler and Markus, 2000), and exposure to a new region (Wilson and McNaughton, 1993) or maze configuration (Frank et al., 2004).
Thus, in the current study, the familiar environment caused retrieval, whereas the novel configuration caused the formation of a new representation.
No age-related differences in the representation of a familiar environment
Aging is characterized by multiple changes in hippocampal neurobiology, including decreased septal cholinergic projection to the hippocampus (Gill and Gallagher, 1998), decreased entorhinal cortex (EC) projection to the dentate gyrus (Barnes and McNaughton, 1980) and CA3 (Smith et al., 2000), reduction of specific classes of hippocampal GABAergic interneurons (Potier et al., 2006), and decreased synaptic plasticity (Foster, 1999). Despite these age-related changes, place fields of old rats are stable in familiar environments (Markus et al., 1994; Oler and Markus, 2000; Wilson et al., 2004, 2005) (however, note Barnes et al., 1997). Similarly, the current study found no effect of age on place field reliability or spatial tuning in the familiar environment, suggesting that the ability to retrieve and reliably hold a previously stored representation of the environment was mostly unaffected by aging.
Age-related impaired encoding
In contrast with the preserved ability to retrieve a stored representation, we found an age-related decrease in the ability to form a new representation. In agreement with previous reports of less remapping in aged animals (Tanila et al., 1997; Oler and Markus, 2000; Wilson et al., 2004), old rats showed less stable place fields during the first exposure to the novel configuration, and less remapping of place fields in the familiar half of the maze. Place fields of aged rats were also less specific (lower information content and higher sparsity). Because this is not always the case (Wilson et al., 2005), presumably the lower spatial specificity was the result of the reduced stability of the fields in the novel configuration.
Effects of MS activation in a familiar environment
The septo-hippocampal GABAergic neurons provide rhythmic disinhibitory input regulating hippocampal theta (Chrobak and Buzsaki, 1996). MS infusions of carbachol excite GABAergic neurons directly by activating muscarinic receptors (Alreja et al., 2000; Wu et al., 2000) and indirectly by activating nicotinic receptors on septal glutamatergic neurons (Wu et al., 2003). The current carbachol dose (0.125 μg) was high enough to induce theta during a nontheta behavior (drinking), and it was below the level that induces seizures (no cases observed) (Sabolek et al., 2004b). The same MS carbachol dose disrupted memory in young rats performing a hippocampal delayed-non-match-to-sample radial maze task (Bunce et al., 2004a,b).
The increased MS activity induced by this low dose of carbachol was enough to impact place fields in both aged and young animals. MS activation decreased the ability of the hippocampus to retrieve and hold the stored representation of the familiar environment; it also increased latency per trial, possibly because of increased exploration.
There are no previous studies examining the effects of MS activation on hippocampal place fields. However, studies using inactivation or lesions of the MS (Mizumori et al., 1989; Leutgeb and Mizumori, 1999; Ikonen et al., 2002) showed little impact on CA1 place fields in a familiar environment. Thus septal input is not necessary in the normal retrieval of a stored representation of the familiar environment. Conversely, MS activation caused a decrease in the ability to retrieve and/or stabilize the correct hippocampal representation of the environment.
Effects of MS activation in a novel configuration
In the novel configuration, increased septal activity disrupted the formation of stable place fields (i.e., fields emerged or stabilized slower) in young rats, whereas it accelerated this process in aged rats. Moreover, the way the environment was experienced during the first exposure affected how the same environment was represented 24 h later. Young rats that received carbachol during their initial experience in the novel configuration still had low place field reliability on their second exposure. In contrast, the reliability of place fields during the second novel configuration exposure was higher for carbachol-treated aged rats than for controls.
Implications of the findings on models of memory encoding and retrieval
CA1 neurons receive inputs from EC and CA3, both of which are modulated by the MS (Nakazawa et al., 2004). CA3 also receives input from the dentate gyrus, postulated to function as a pattern separator, allowing for the storage of multiple orthogonal representations within CA3 (McClelland and Goddard, 1996). Place fields in CA3 seem less affected by small modifications of the environment than in CA1 (Fyhn et al., 2002; Guzowski et al., 2004; Vazdarjanova and Guzowski, 2004), indicating that CA1 does not always directly reflect CA3 activity. The convergence of direct neocortical (EC) input with previously stored information (CA3) onto CA1 neurons allows CA1 to perform a comparison of a stored representation with the current environment (Fyhn et al., 2002). Place fields are constant in a familiar stable environment. However, no environment is completely stable; there are continuous fluctuations in auditory, visual, and olfactory stimuli. Under some conditions, these minor variations in neocortical input are ignored and a previously stored CA3 representation is activated. Interestingly, even major sensory changes such as turning off the lights (Markus et al., 1994) or changing the reward (Tabuchi et al., 2003) can have little effect on the location of place fields. Consequently, the degree of mismatch between EC and the stored CA3 representation cannot be the only factor for inducing encoding in CA1. Theta rhythm was proposed to be a signal for encoding (Hasselmo et al., 2002). Our results support this hypothesis, because carbachol-induced theta caused remapping even in a stable familiar environment. Normally, in a familiar environment, CA1 cells reflect the previously stored CA3 input and ignore minor sensory mismatches. Increasing the degree of mismatch by moderately modifying the environment (i.e., our novel configuration) and/or increasing septal activity in a familiar environment results in CA1 reflecting extrahippocampal input more than the previously stored CA3 input (Fig. 11A).
Septal modulation of the hippocampus is reduced during aging. In aged rats, the MS contains less cholinergic and GABAergic cells (Krzywkowski et al., 1995), and the septo-hippocampal projections are reduced (Gill and Gallagher, 1998; Potier et al., 2006). Although overall hippocampal theta power seems to be preserved (Shen et al., 1997), theta frequency potentiation declines in impaired aged rats (Tombaugh et al., 2002), and long-term potentiation (LTP) is harder to induce and decays more rapidly (Foster, 1999). Buzsaki (1989) proposed a two-stage model of memory encoding: during theta, information about the current environment reaches the hippocampus via the EC; during sharp waves, population bursts in CA3 may lead to LTP induction in CA1. Therefore, both theta and sharp wave/ripples are important for memory encoding. The predominance of theta and sharp wave/ripple states may be altered in aged rats, such that the relative time the rat is in a theta state diminishes with aging. A carbachol-induced prolonged theta state (and consequent reduced occurrence of sharp wave/ripples) would benefit the aged rats by facilitating encoding in a new environment. In young rats with preserved theta states, the reduction in sharp wave/ripple state by carbachol would impair pattern completion in CA3 and manifest as reduced spatial information content and decreased reliability of place fields. Our results suggest that the decreased response of place fields to environmental manipulations in aging reflects a deficit in the transition of the hippocampus from retrieval to an encoding state because of a decline in MS modulation (Fig. 11B).
The use of cholinergic system enhancers in aged humans and rats has given conflicting results. Our data suggest the reason for the mixed behavioral results. Recall of previously well learned information does not require MS activation, and artificially activating this system will disrupt normal recall. Conversely, encoding of new information depends on an intact septo-hippocampal system, and the age-related impairments can be reversed by MS activation.
This work was supported by University of Connecticut FRS445142 and National Institutes of Health Grant R29-A613941-01A1 (E.J.M.). We thank Dr. Monty Escabi for his help with the EEG data analysis, Gregory Peters and Mark Brandon for assistance conducting the experiments, Dr. James Chrobak for useful comments on the design of the study, and Dr. Jamie Bunce for helping with the design of the study and the revision of this manuscript.
- Correspondence should be addressed to Dr. Etan J. Markus, Behavioral Neuroscience Division, Department of Psychology, University of Connecticut, U-1020, Storrs, CT 06269.