Urethane anesthesia suppresses hippocampal subthreshold activity and neuronal synchronization

Urethane, an anesthetic utilized for animal experiments, induces neocortical slow oscillations in which a large number of neurons emit rhythmic synchronized activity. However, it remains unclear how urethane affects neuronal activity in the hippocampus. In this study, we obtained in vivo patch-clamp recordings from dorsal hippocampal CA1 neurons in mice and found a reduction in the fluctuation of subthreshold membrane potentials during urethane anesthesia, implying reduced synaptic activity in the hippocampus. We then performed spike unit recordings from dorsal hippocampal CA1 neuronal ensembles in rats and found prominent reductions in the spike rates of the majority of hippocampal units, especially spatially selective units, during urethane anesthesia, whereas a subset of nonspatial units exhibited increased spike rates. The overall reductions in neuronal spike rates induced by urethane led to prominent decreases in spike synchronization across neuronal units. Consistently, the magnitude of hippocampal sharp wave ripples was also reduced by urethane. The suppression of hippocampal neuronal synchronization by urethane may lead to the disruption of offline memory reactivation mechanisms.


Introduction
Urethane, ethyl carbamate, is an anesthetic to acutely induce longterm immobilization for animal experiments while it has chronic toxic effects including mutagenic, carcinogenic, and hepatotoxic effects (Maggi and Meli, 1986;Flecknell, 2016). Especially in the field of cerebral physiology, urethane has been utilized as a tool not only to induce animal's immobility but also to study unique brain activity patterns triggered by urethane anesthesia. Across widespread areas of the neocortex, urethane typically triggers slow (typically, < 1.0 Hz) oscillations and a prominent large-amplitude and slow frequency (1 Hz) rhythm, which is similar to the oscillatory patterns in deactivated states during nonrapid eye movement (non-REM) sleep (Steriade et al., 1993;Amzica and Steriade, 1995). Neuronal mechanisms underlying such oscillatory patterns have attracted interest from the point of view of neurophysiology and have been extensively studied using acute in vivo electrophysiological recordings and optical imaging from individual neurons. By time-locking to slow oscillations, a large number of neocortical neurons exhibit synchronized membrane depolarization with superimposed action potentials (up states) followed by hyperpolarization with cessation of action potentials (down states). Such urethaneinduced up/down subthreshold oscillations are generated from an altered balance between excitation and inhibition (Haider et al., 2006), possibly through the integration of complex molecular mechanisms https://doi.org/10.1016/j.brainres.2020.147137 Received 9 June 2020; Received in revised form 4 September 2020; Accepted 23 September 2020 such as the potentiation and suppression of neurotransmitter-gated ion channels (Hara and Harris, 2002;Accorsi-Mendonca et al., 2007;Tian et al., 2012).
The hippocampus, a brain region that plays a central role in memory functions through an intricate communication between the neocortex, exhibits considerably different oscillatory patterns independent of the neocortex; while the hippocampus generates transient slow oscillationlike activity patterns, these oscillations do not necessarily co-occur with neocortical slow oscillations (Wolansky et al., 2006;Clement et al., 2008;Sharma et al., 2010). This fact means that, even while the neocortex generates such large synchronization by urethane, this information does not entirely propagate to the hippocampus. Indeed, the majority of hippocampal neurons, except a subset of hippocampal interneurons and dentate granule cells, do not exhibit apparent up/down subthreshold oscillations (Hahn et al., 2006(Hahn et al., , 2007Isomura et al., 2006). Due to the strong independence of hippocampal activity, further studies are needed to examine how urethane affects hippocampal neuronal activity at both subthreshold and suprathreshold levels.
To date, several pieces of evidence has demonstrated that urethane exerts specific effects on spike patterns of hippocampal neurons that are different from those of neocortical neurons; urethane prominently decreases the excitability of hippocampal neurons both in vivo (Mercer et al., 1978;Suzuki and Smith, 1987;Kamondi et al., 1988;Shirasaka and Wasterlain, 1995) and in vitro (Tian et al., 2012). However, recent studies noted that hippocampal neurons are physiologically heterogeneous; spike patterns and participation rates for synchronization differ considerably across cells (Graves et al., 2012;Mizuseki and Buzsaki, 2013), and not all but a certain fraction of cells participate in spatial information processing, as typically represented by place cell ensembles that selectively fire when rats visit a specific location (O'Keefe and Nadel, 1978). Consistently, the degree of the urethaneinduced reduction in hippocampal spikes is also heterogeneous across hippocampal neurons (Mercer et al., 1978). It remains unknown how urethane affects spike patterns of those hippocampal neuronal ensembles that encode different information. In addition, hippocampal neurons generate synchronized spikes time-locked to sharp wave ripples (SWRs), which are considered crucial in memory reactivation during sleep (Lee and Wilson, 2002;Girardeau et al., 2009). Considering the urethane-induced changes in spike patterns of single neurons, SWR-associated neuronal synchronization may be altered during urethane anesthesia.
To address these questions, we first tested how membrane potential dynamics in hippocampal neurons undergo changes during urethane anesthesia using in vivo whole-cell patch-clamp recordings. We then recorded the spike patterns of hippocampal cell ensembles, including spatial units, in rats using a tetrode assembly and analyzed how their spike rates and synchronous spike patterns were altered by urethane anesthesia. The main findings were that urethane reduced subthreshold potential fluctuations and synchronized spikes in hippocampal neuronal units, which potentially may be a neurophysiological mechanism underlying urethane-induced memory deficit.

Urethane reduces membrane potential fluctuations in hippocampal neurons in mice
We first tested how urethane anesthesia affects synaptic activity in dorsal hippocampal CA1 neurons by in vivo patch-clamp recordings from mice (Fig. 1A). Here, mice were selected because only mice allowed recordings of stable subthreshold signals by our in vivo patchclamp setup. Mice were head-fixed under awake or urethane anesthesia (2.25 g/kg) conditions in which mice were perfectly immobile and showed no signs of sensory responses. Whole-cell recordings were obtained from electrophysiologically and morphologically identified hippocampal pyramidal cells ( Fig. 1A and 1B; awake, n = 5 cells; urethane, n = 11 cells). The recording periods were 68.0 ± 5.6 s and 174.1 ± 50.6 s in awake and urethane-anesthetized mice, respectively. Baseline membrane potentials did not significantly differ between the awake and urethane-anesthetized mice (-58.4 ± 3.1 mV and -59.5 ± 1.3 mV, respectively; t 14 = 0.41, P = 0.69, Student's t-test). Subthreshold membrane potential fluctuations, which reflect synaptic activity, occurred with much larger magnitude and higher frequency in living animals than in slice preparations (the example traces in Fig. 1B), making it impossible to precisely extract single excitatory and inhibitory synaptic inputs. We instead quantified the amplitude of membrane potential fluctuations by computing the standard deviation of the subthreshold membrane voltage (SD mV ) without any filtering and compensation in each cell throughout the recording periods (Fig. 1B). Some of the neurons occasionally exhibited suprathreshold spikes, but our analysis excluded membrane potentials 50 ms before and after suprathreshold spikes because urethane-induced spike patterns were further analyzed from multiunit recording data in subsequent figures . Overall, the SD mV in urethane-anesthetized mice was significantly lower than that in awake mice ( Fig. 1C; t 14 = 3.37, P = 0.0046, Student's t-test). To further examine whether subthreshold traces included specific oscillatory patterns and whether entire power differed between the awake and urethane-anesthetized conditions, we applied a power spectrum analysis on these traces using fast Fourier transformation (Fig. 1D). First, no apparent peaks of power were detected within the individual spectrums, demonstrating that subthreshold activity contained no specific oscillatory patterns in both of the conditions. Second, we computed the oscillatory power of slow wave (0.5-2 Hz), delta (2-4 Hz), and gamma (30-80 Hz) bands from the spectrum, neuronal oscillatory bands that have been typically computed in the studies of sleep and arousal states (Buzsaki, 2006). Consistent with the results of SD mV , all of the power in awake mice was significantly higher than that in urethane-anesthetized mice ( Fig. 1E; slow wave: t 14 = 3.32, P = 0.0051; delta: t 14 = 2.97, P = 0.010; gamma: t 14 = 3.07, P = 0.0084, Student's t-test), confirming that the magnitude of subthreshold activity related to arousal states is prominently reduced by urethane. These results suggest that urethane reduces synaptic activity in hippocampal neurons, consistent with early observations that urethane inhibits several types of ligand-gated ion channels (Hara and Harris, 2002;Accorsi-Mendonca et al., 2007;Tian et al., 2012).

Urethane reduces the spike activity of hippocampal neurons in rats
The urethane-induced reduction in synaptic activity suggests that urethane alters the spike patterns of dorsal hippocampal neuronal populations. We implanted rats with eight tetrodes directed at the hippocampus for multiunit recordings ( Fig. 2A). Here, rats were selected because rats allowed synchronous recordings of larger numbers of units with larger numbers of electrodes, compared with mice. First, we tested how different dosage of urethane affects spike activity of hippocampal neurons. During recordings from hippocampal units, we intraperitoneally injected urethane with increasing dosage by 0.5 g/kg every 45 min up to 1.5 g/kg (n = 4 rats; typical 10-s periods are shown in Fig. 2B). At a dosage of 0.5 g/kg, the rats slowly walked around in the recording box and showed apparent righting reflex and average spike rates were not significantly different from those without urethane ( Fig. 2C; P = 0.10, t 40 = 2.20, paired t-test followed by Bonferroni correction). At a dosage of 1.0 g/kg, the rats showed neither righting reflex nor reflex in response to hind limb pinching but occasionally moved their hind limb and average spike rates were not significantly different from those without urethane ( Fig. 2C; P = 0.095, t 40 = 2.23, paired t-test followed by Bonferroni correction). At a dosage of 1.5 g/ kg, the rats did not show any of these behavioral patterns and average spike rates were significantly lower than those without urethane ( Fig. 2C; P = 0.026, t 40 = 2.77, paired t-test followed by Bonferroni correction). As our goal was to understand apparent effects of urethane as a perfect, not partial, anesthesia on hippocampal neuronal activity, we chose the dosage of 1.5 g/kg in following analyses.
Next, the other 5 rats were trained to perform a U track task in which they continuously ran on a U track to obtain chocolate milk as a reward at both ends of the track ( Fig. 2D and 2E). This spatial task enabled us to identify hippocampal spatial units, putative place cells, from our recorded cell populations. On recording days, spike patterns were recorded from hippocampal pyramidal neuronal units from 5 rats (Rat 1,16 units;Rat 2,15 units;Rat 3,7 units;Rat 4,8 units;Rat 5,5 units) in which five, five, six, two, and two tetrodes identified 1, 2, 3, 4, and 5 units, respectively, resulting in a total of 51 units. The locations of these tetrodes were presented in Fig. 2A. A recording day included three sessions ( Fig. 2D): (1) the prerest session, in which the rats were almost immobile in a rest box for 10-15 min and which was analyzed as a control session for comparison with the urethane session, (2) the track session, in which the rats performed the U track task for 10-15 min, and (3) the urethane session, in which the rats were anesthetized with urethane (1.5 g/kg, i.p.) after the task and showed no movements of the body or vibrissae in response to sensory stimuli. Restricting the time periods of all these sessions within 2 h enabled continuous tracking of spikes of identical units in multiunit recordings. The mortality rate of rats 4 h after urethane administration by this dose was approximately 40%.
In the U track task, of 51 recorded neuronal units, 34 (66.7%) spatial units with apparent place fields on the track were identified ( Fig. 2E). In addition, when the rats were outside the track in the prerest and urethane sessions, these units were repeatedly activated ( Fig. 2F), which are considered crucial for memory reactivation of awake experiences (Wilson and McNaughton, 1994;O'Neill et al., 2008). Overall, the spike rates of both spatial and nonspatial units were considerably lower in the urethane session than in the prerest session ( Fig. 2G). In particular, the spike-rate reductions in spatial units were more prominent (98.2 ± 0.1%, n = 34 cells) than those in nonspatial units (68.3 ± 2.0%, n = 17 cells). Contrary to the overall tendency, the spike rates of a subset (41.2%) of nonspatial units were almost 0 Hz in the prerest and track sessions and increased by > 50% in the urethane sessions (Fig. 2H, left). These results demonstrate that urethane suppresses spike activity in the majority of hippocampal neurons, while it activates the minority of neurons, demonstrating its heterogeneous effects on hippocampal neuronal spikes.

Urethane inhibits the spike synchrony of hippocampal neurons
Hippocampal neuron populations emit synchronized spikes associated with sharp-wave ripple events (Wilson and McNaughton, 1994;Buzsaki, 2015). The highly sparse spikes in the urethane session imply that spike synchronization across hippocampal neuron populations is prominently reduced by urethane. To test this possibility, instantaneous changes in the number of coactive units were computed from spike rasterplots (bin = 200 ms) (Fig. 3A). To quantify datasets from all rats, Comparison of slow wave (0.5-2 Hz), delta (2-4 Hz), and gamma (30-80 Hz) power of awake (n = 5) and urethane-anesthetized (n = 11) mice. *P < 0.05, Student's t-test.
the number of coactive units was normalized by the total number of units recorded in each rat, which was computed as the percentage of coactive units (Fig. 3B). The percentage of coactive units in the urethane session was significantly lower than that in both the prerest and track sessions (n = 5 rats; vs prerest, U = 40, P = 0.0238; vs track, U = 40, P = 0.0238, Mann-Whitney U test followed by Bonferroni correction), whereas no significant difference was found between the prerest and track sessions (n = 5 rats; U = 30, P > 0.99, Mann- (Left) A schematic illustration of multiunit recordings from the hippocampus of a freely moving rat and a cresyl violet-stained brain section showing the tetrode location in the hippocampus (black arrowhead). (Right) Superimpositions of recording sites for all tetrodes (from 5 rats) on the dorsal HPC cell layer in sequential coronal brain sections. Each circle represents each tetrode. (B) Rasterplots in which each row represents a hippocampal unit and each dot represents a spike. Here, 19 units were simultaneously recorded from a rat and urethane was intraperitoneally injected into the rat. The values above represent the total amount of urethane injected. (C) Average spike rates of all units recorded across different urethane concentrations (n = 41 units from 4 rats). Data are presented as mean ± SEM. *P < 0.05, paired t-test followed by Bonferroni correction. (D) The experimental timeline of hippocampal multiunit recordings. After a > 10-min prerest session, the rats performed a U track task for > 10 min. The rats were then anesthetized with urethane (red arrow). In each of the three sessions, identical hippocampal neurons were recorded. (E) Averaged firing-rate distributions of hippocampal units on a trajectory from R1 to R2 (left) and from R2 to R1 (right). In each panel, spatial units followed by nonspatial units were aligned in order of the positions of place-field peaks. (F) Rasterplots showing a representative 10-s period from each session is presented. (G) Average spike rates of spatial (black, n = 34) and nonspatial (cyan, n = 17) units across the three sessions. (H) Changes in spike rates of individual units. The left and right panels show units for which urethane induced > 50% increases (n = 7 units) and decreases (n = 43 units), respectively, in their firing rates. Each line represents a unit, labeled in black (spatial unit) or blue (nonspatial unit). Whitney U test followed by Bonferroni correction). To test whether the decreased spike synchronization is explained by the overall reduction in spike rates during urethane anesthesia, we randomly downsampled spikes obtained from the prerest session so that spike rates of individual units were similar to those in the urethane session, termed prerest (downsampled) data. For each rat, we created 1000 surrogate datasets. No significant difference in the percentage of coactive units was found between the prerest (downsampled) data and the urethane session ( Fig. 3C; n = 5 rats, U = 30, P = 0.69, Mann-Whitney U test). These results suggest that the reductions in the spike rates of individual neurons account for the reductions in synchronized neuronal spikes at the population level during urethane anesthesia.
An additional prominent feature of neuronal synchronized activity is the cofiring of neurons (O'Neill et al., 2008;Takahashi et al., 2010). The degree of cofiring of a unit pair was quantified as the correlation coefficient of their spike timing (an example shown in Fig. 3D). This analysis was applied to all possible unit pairs to generate a cofiring map diagram for each rat (Fig. 3E). The number of unit pairs with pronounced cofiring was significantly reduced in the urethane session compared with the prerest and track sessions ( Fig. 3F; prerest vs urethane: D max = 0.28, P = 0.00092; track vs urethane: D max = 0.28, P = 0.00089; prerest vs track: D max = 0.063, P > 0.99; the Kolmogorov-Smirnov test followed by Bonferroni correction). Consistent with the overall tendency of decreased synchronized spikes at the population level shown in Fig. 3B, these results suggest that synchronized spikes of specific neuron pairs are prominently reduced by urethane.

Urethane reduces the hippocampal sharp wave ripple amplitude
In the hippocampus, synchronous spikes of neurons are reflected as SWR signals in local field potential (LFP) traces (Csicsvari et al., 2000;Buzsaki, 2015). We next asked how SWR signals were changed under urethane anesthesia conditions (Fig. 4A). Generally, hippocampal SWRs were detected when the envelope or the magnitude of a ripple band (150-250 Hz)-filtered LFP trace exceeded a certain threshold (mean + 3 × SD) that was determined based on the average and standard deviation (SD) of an LFP trace throughout a recording period (Fig. 4B). Here, we first determined this threshold from the distribution of an LFP trace in the prerest session (the green line in Fig. 4B), as our main focus was to examine how SWR signals were altered in the urethane session compared to the prerest session. Our analysis excluded the track session because this session involved complex behavioral patterns, such as running and reward consumption. On average, the frequency of SWRs in the prerest session was 0.58 ± 0.08 Hz (Fig. 4D; n = 5 rats), but no SWRs were detected in the urethane session (Fig. 4E,  green). This result was consistent with our observations in Fig. 3 that synchronized spikes of hippocampal neurons were significantly reduced in the urethane session compared to the prerest session. However, this The frequency of SWR events at a common threshold defined from the prerest session (green, n = 5 rats). The rightmost plot represents the frequency of SWR events at a different threshold defined within the urethane session (cyan). Each thin dot represents each rat. result does not mean that urethane perfectly abolished SWRs. As shown in the traces in Fig. 4A and the power distributions in Fig. 4B, the entire LFP power of the ripple (150-250 Hz) band was prominently reduced during urethane anesthesia. We thus reset the threshold based on the distribution of an LFP trace within the urethane session (the cyan line in Fig. 4B), which resulted in a decreased threshold. With this new threshold, SWR signals were detected in LFP traces in the urethane session (Fig. 4E, cyan), while their amplitudes were considerably lower than those of SWRs observed in the prerest session (Fig. 4C).

Discussion
While urethane anesthesia has been shown to induce slow (1 Hz) oscillations and rhythmic up state-triggered synchronous spikes of neuronal populations in the neocortex (Steriade et al., 1993;Amzica and Steriade, 1995), it remains unclear how activity patterns of hippocampal neuronal populations are altered during urethane anesthesia. Our results demonstrated that urethane reduces spontaneous membrane potential changes in mice, reflecting decreased synaptic activity, and diminishes large fractions of hippocampal synchronized spikes and large amplitude sharp wave ripples in rats, all of which are distinct from those reported in the neocortex. Taken together with early studies (Wolansky et al., 2006;Clement et al., 2008;Sharma et al., 2010), our results suggest that hippocampal activity patterns during urethane anesthesia are specifically isolated from widespread synchronized activity throughout the neocortex. We note that cellular and network activity found in this study may differ between mice and rats and further studies are required to address this issue.
General anesthesia utilized in animal experiments acts on various neurotransmitter systems (Hemmings et al., 2005); urethane potentiates GABA receptors (Hara and Harris, 2002) and inhibits NMDA and AMPA receptors (Hara and Harris, 2002) and spontaneous glutamate release from presynaptic sites (Tian et al., 2012), while ketamine mainly inhibits NMDA receptors (Gideons et al., 2014;Zorumski et al., 2016) and isoflurane suppresses neurotransmitter release (Wu et al., 2004). The complex integration of such molecular mechanisms by each anesthesia should lead to unique effects on neuronal activity. Indeed, at more macroscopic levels, the anesthesia differentially alters hippocampal activity; our study demonstrated that urethane reduced in neuronal spikes and SWR-related synchronization, while ketamine specifically increases high gamma oscillations (Caixeta et al., 2013) and isoflurane decreases high gamma oscillations (Hudetz et al., 2011) and induces SWR-like events (Lustig et al., 2016). These observations suggest that hippocampal neuronal mechanisms considerably differ across different anesthesia although their superficial actions (e.g. decreased animal's arousal states) appear almost identical. Further studies are needed to bridge the gaps of these observations between the molecular and network levels.
Our results from in vivo whole-cell recordings demonstrated that the amplitudes of spontaneous changes in membrane potentials were reduced in urethane-anesthetized compared with awake mice. This subthreshold effect of urethane inhibits neuronal membrane potentials to exceed a spike threshold, potentially accounting for overall decreases in spike rates. Subtle differences in such intrinsic neuronal mechanisms across neurons may lead to the heterogeneity of spike sensitivity to urethane across different types of neurons, such as place cells and nonplace cells. Our simulation analyses with random downsampling of spike datasets demonstrated that the urethane-induced reductions in synchronized spikes are almost fully explained by the decreases in the spike rates of individual neurons, suggesting that the ability of hippocampal neurons to synchronize their spikes was not strongly attenuated by urethane.
Our LFP analysis demonstrated that urethane nearly perfectly diminished SWR events detected at a threshold determined from the absolute LFP power in awake conditions. Notably, both transient SWR signals and the entire power of ripple (150-250 Hz) bands throughout the recording periods were prominently reduced during urethane anesthesia. These observations indicate that the threshold for SWR detection needs to be reset to extract SWR signals under urethane conditions. As expected, SWR events with smaller amplitudes could be detected at a threshold defined within the urethane condition. In accordance with our insights from the spike pattern analysis, the results suggest that urethane does not perfectly attenuate SWR generation in hippocampal networks but instead reduces the amplitude of SWR signals, possibly due to overall decreases in the spike rates of individual neurons.
It has been demonstrated that neocortical spike activity observed during urethane anesthesia differs from that during natural sleep, despite the similarity of slow oscillations in both conditions (Steriade et al., 1993;Isomura et al., 2006;Clement et al., 2008). In addition, our results suggest that memory consolidation mechanisms in the hippocampus underpinned by neuronal spike synchrony may be attenuated by urethane compared with natural rest/sleep conditions. Knowledge of such unique effects of urethane on cortical information processing will be useful for interpreting the results of animal experiments performed under urethane anesthesia.

Animal ethics
This study was performed in strict accordance with the recommendations in the NIH Guide for the Care and Use of Laboratory Animals. All animals were handled according to the approval of the experimental animal ethics committee of the University of Tokyo (approval number: P29-9, P29-11 and A30-72).

Subjects
A total of 16 ICR mice (4-6 weeks old) with weights of 20-35 g were used for patch-clamp recordings (Fig. 1). In addition, 9 male Slc:SD rats (5-8 weeks old) with preoperative weights of 80-260 g were used for multiunit recordings (Figs. 2-4). All animals were purchased from SLC (Shizuoka, Japan) and were housed on a 12-h light/12-h dark schedule with lights off at 7:00 AM, and behavioral experiments for rats occurred in the dark phase. For patch-clamp recordings from mice, no food restriction was imposed before recordings. For multiunit recordings from rats, they were housed individually and reduced to 85% of their ad libitum weight through limited daily feeding while water was readily available.

In vivo patch-clamp recordings from mice
For patch-clamp recordings, 11 mice were anesthetized with urethane (2.25 g/kg, i.p.), whereas the other 5 mice were not anesthetized (awake condition). Anesthesia was confirmed by a lack of paw withdrawal, whisker movement, and eyeblink reflexes. The skin was subsequently removed from the head, and a metal head-holding plate was implanted. A craniotomy (2.5 × 2.0 mm 2 ) was performed at 2.0 mm posterior to bregma and 2.5 mm ventrolateral to the sagittal suture, and the neocortex above the hippocampus was aspirated (Kuga et al., 2011;Sakaguchi et al., 2012;Matsumoto et al., 2016). The exposed hippocampal window was covered with 1.7% agar at a thickness of 1.5 mm. Through the window, a borosilicate glass pipette (4.0-7.0 MOhm) was lowered slowly into the hippocampus at a depth of 100-300 μm from the alveus, and whole-cell recordings were obtained from neurons in the CA1 stratum pyramidale. The intrapipette solution consisted of the following reagents: 120 mM K-gluconate, 10 mM KCl, 10 mM HEPES, 10 mM Na 2 -phosphocreatine, 4 mM Mg-ATP, 0.3 mM Na 3 -GTP, 0.2 mM EGTA, and 0.2% biocytin. The solution was adjusted to pH 7.2-7.3 and 285-300 mOsm. The signal was amplified with a MultiClamp 700B amplifier, analyzed with pCLAMP10.3 software (Molecular Devices) and digitized at 20 kHz. At the beginning of the experiments, 1000-ms depolarizing and hyperpolarizing rectangular currents from -200 to + 200 pA were injected into the cell at steps of 50 pA to characterize the cell's intrinsic properties and spike responses (Fig. 1A). Pyramidal cells were identified by their regular spiking patterns (showing increasing inter-spike intervals with time) or burst spiking patterns (showing intermittent (> 100 ms) bursting including multiple spikes), but not fast spiking patterns (showing constant~10-ms inter-spike intervals), upon injected currents (Jensen et al., 1994;Jarsky et al., 2008;Graves et al., 2012). In addition, a final identification was performed in combination with post hoc histological analysis as described later. The liquid junction potential was nulled offline. Cells were discarded when the mean resting potential exceeded -50 mV. Moreover, recordings were truncated when the spike peak decreased below -20 mV.
After recordings, each mouse received an overdose of urethane and was perfused intracardially with 4% PFA and decapitated. The tissue containing the hippocampus was sliced coronally at a thickness of 100 μm in PBS using a vibratome (Dosaka). The slices were incubated with 2 μg/ml streptavidin-Alexa Fluor 594 conjugate and 0.2% Triton X-100 for 4 h and then with 0.4% NeuroTrace 435/455 blue fluorescent Nissl Stain (Thermo Fisher Scientific; N21479) for 2 h. The slices were analyzed with an FV1200 (Olympus, Tokyo, Japan) confocal system under a 10 × objective. Z series images were collected in 2.0-μm steps, and 5-25 Z sections (10-50 μm thick) were stacked using ImageJ (NIH). In reference to previous reports (Bannister and Larkman, 1995;Ishizuka et al., 1995;Graves et al., 2012), pyramidal cells were morphologically identified as cells that (1) were located within the CA1 pyramidal cell layer, (2) had one or two apparent apical dendritic trunks projecting nearly vertically to the stratum radiatum and stratum lacunosum-moleculare, some of which bifurcated in the stratum radiatum, and (3) had multiple basal dendrites projecting in the stratum oriens (Fig. 1A).

Behavioral training of rats on a U-shaped track in rats
Before surgery, each rat was trained daily for at least 3 days to perform a U-shaped track task. On one training day, the rat was trained to run back and forth on a U-shaped track consisting of three 70 × 9 cm 2 alleyways (with small sides rising 0.5 cm above the surface of the arm, which was elevated 29 cm above the floor) to obtain a constant~0.2 ml of chocolate milk placed at the end of the track as a reward during a 10-15-min session. This training was repeated daily for 10 min until the rat consumed the reward at least 30 times within a 10min training period.

In vivo multiunit recordings from rats
A standard electrode assembly for multiunit recordings, called a microdrive, was prepared as described previously (Okada et al., 2017;Yagi et al., 2018;Aoki et al., 2019). Each rat was anesthetized with isoflurane gas (1.5-2.5%) and then fixed in a stereotaxic instrument with two ear bars and a nose clamp. An incision was made from the area between the eyes to the back of the head. A rectangular craniotomy with a size of 1.2 × 2.0 mm was performed above the right hippocampus (3.6 mm posterior and 3.0-5.0 mm lateral to bregma) using a high-speed drill, and the dura was surgically removed. Two stainlesssteel screws were implanted in the bone above the cerebellum to serve as ground and reference electrodes. A microdrive that consisted of 8 independently movable tetrodes, which was created using a 3D printer (Form 2, Formlabs), was stereotaxically implanted. The tip of the electrode bundle was lowered onto the cortical surface, and the electrodes were inserted into the brain at a depth of 0.75 mm at the end of surgery. The electrodes were constructed from 17-μm-wide polyimidecoated platinum-iridium (90/10%) wire (California Fine Wire), and the electrode tips were plated with platinum to lower the electrode impedances to 180-300 kΩ at 1 kHz. The microdrive was physically protected by a cone-shaped plastic cover (13 mm in height, 13 mm diameter for the top circle, 30 mm diameter for the bottom circle). To reduce its weight, the plastic cover had 36 elliptical holes that were 4 mm in diameter along the major axis and 2 mm in diameter along the minor axis. Finally, all of the wires, the microdrive and the cover were secured to the skull using stainless-steel screws and dental cement. Following surgery, the rats were housed individually in transparent Plexiglass cages with free access to water and food for at least 3 days. After recovery from surgery, food was deprived up to 85% according to body weight.
Each rat was connected to the recording equipment via a Cereplex M digitally programmable amplifier (Blackrock) close to the rat's head. The output of the headstage was conducted via a lightweight multiwire tether to the Cereplex Direct recording system (Blackrock). Electrode turning was performed while the rat rested on the pedestal. The electrode tips were advanced slowly 25-250 μm per day for 10-17 days until spiking cells in the CA1 layer of the hippocampus, which was identified on the basis of LFP signals and single-unit spike patterns, were encountered. Once the tetrodes were adjacent to the cell layer, as indicated by the presence of low-amplitude multiunit activity, they were settled into the cell layer for stable recordings over a period of several days.
During the several days of this turning period, the rats were trained again in a similar manner as in the presurgery trainings. This postsurgery training lasted for at least 10-17 days before electrophysiological recordings were obtained. Postsurgery training was performed with the recording headstage and cable attached to the rat's head so that the rat became familiar with the recording conditions. Electrophysiological recordings during the U track task began after the rats again achieved the behavioral criteria and stable well-separated unit activity was identified in the hippocampus. On recording days, each rat was first maintained on the pedestal outside the track for 10 min for a prerest session. Then, a track session in which the rat performed the U track task for > 15 min was conducted. LFP recordings were sampled at 2 kHz and filtered between 0.1 and 500 Hz. Unit activity was amplified and bandpass filtered at 600 Hz to 6 kHz. Spike waveforms above a trigger threshold (60 μV) were time-stamped and recorded at 30 kHz for 1.6 ms. To monitor the rat's moment-tomoment position, a near-infrared reflection sticker was attached to the microdrive on the rat's head, and the position of the sticker was tracked at 25 Hz using an infrared camera (MCM-303NIR, Gazo, Japan) attached to the ceiling, which was sampled by a laptop computer.
After recordings in all the sessions, the rats received an overdose of urethane and were intracardially perfused with 4% paraformaldehyde (PFA) in phosphate-buffered saline (pH 7.4) and decapitated. To aid in electrode track reconstruction, the electrodes were not withdrawn from the brain for at least 3 h after perfusion. Following dissection, the brains were fixed overnight in 4% PFA and subsequently equilibrated with a sequence of 20% sucrose and 30% sucrose in PBS. Frozen coronal sections (50 μm) were cut using a microtome, and serial sections were mounted and processed for cresyl violet staining. The slices were subsequently coverslipped with mounting agent (PARAmount-D, Falma). The positions of all tetrodes were confirmed by identifying the corresponding electrode tracks in the histological tissue sections.

Spike sorting
Spike sorting was performed offline using the graphical clustercutting software MClust (Redish, 2009). Clustering was performed manually in 2D projections of the multidimensional parameter space (i.e., comparisons between waveform amplitudes, the peak-to-trough amplitude differences, waveform energies, and the first and second principal components of waveforms, each measured on the four channels of each tetrode). Only units with stable spike waveforms (showing < 10% changes in the amplitude of the first spikes in burst) throughout a recording period, from the prerest session to the urethane session, were included in the analysis. A refractory period of each unit was confirmed by constructing a histogram of the autocorrelation of the spike times and a unit representing a neuron was defined as a unit showing no positive correlations within a 5 ms lag and the autocorrelation functions decaying to the mean firing rate at large time lags. Here, the degree of theta modulation in the autocorrelation was not considered as a criterion because apparent theta oscillations were almost abolished in the urethane session. Finally, units with waveforms longer than 200 μs were considered units representing putative excitatory pyramidal cells and included in the analysis.

Analysis of spatial firing patterns
To analyze spike patterns, the rat's coordinates and the positions of the spikes of individual cells were projected onto a centerline of alleyways corresponding to each trajectory. The average firing-rate distribution on each trajectory was separately computed along the projected line by dividing the total number of spikes in each location bin (10 cm) by the total amount of time that the rat spent in that bin. All firing-rate distributions were smoothed by a one-dimensional convolution with a Gaussian kernel with a standard deviation of one pixel (10 cm). A cell was defined as a spatial unit, a putative place cell, based on the following two criteria: (1) the average firing-rate distribution on a trajectory in a session had a maximum firing rate of > 2 Hz (i.e., the absolute maximum firing rate) and (2) the maximum firing rate exceeded 2 standard deviations (SDs) above the mean, with the SD and the mean being computed from the series of firing rates except the maximum firing rate in that distribution. The definition of spatial units was based on previous studies (Leutgeb et al., 2007;Shin et al., 2019) and slightly modified from that in these studies so that place fields were more precisely extracted from our U track data. The other cells that did not meet the criteria were classified as nonspatial units. For each spatial unit, a place field center was defined as the position that gave the maximum firing rate in the distribution.

Analysis of synchronized events and cofiring
In each rasterplot, the number of synchronized units (or coactive units) was computed in each 200-ms time window by sliding the time window of 100 ms (Fig. 3A). To measure the degree to which a given unit pair exhibits synchronous spikes, termed cofiring (O'Neill et al., 2008), the numbers of spikes in consecutive 100-ms windows in each of the two units were counted to create N-dimensional vectors x and y, with N being the total number of windows (Fig. 3D). Pearson's correlation coefficients were computed between the two vectors as follows: This analysis was applied to all possible unit pairs.

Detection of hippocampal SWRs
The LFP signal was bandpass filtered at 150-250 Hz, and the envelope (power) of the filtered LFP trace was computed via Hilbert transformation. SWR events were recorded if the envelope exceeded a threshold for at least 15 ms. The threshold for SWR detection was set to 3 standard deviations (SDs) above the mean of the bottom 90% of all envelopes (i.e., without SWR signals) throughout the prerest or urethane sessions (Fig. 4B). The amplitude of each SWR signal was computed as the difference between the peak envelope and the threshold. The power of LFP signals at the 150-250 Hz band was calculated by fast Fourier transformation in Matlab (Mathworks).

Statistics
All data are presented as the mean ± standard error of the mean (SEM) and were analyzed using MATLAB2020a. Comparisons of twosample data were analyzed by Student's t-test. Comparisons of two distributions were analyzed by the Kolmogorov-Smirnov test. Multiple group comparisons were performed by the Kolmogorov-Smirnov followed by post hoc Bonferroni correction. The null hypothesis was rejected at P < 0.05 level.