Abstract
Neurons in the cerebral cortex and hippocampus discharge synchronously in a brain state-dependent manner to transfer information. Published studies have highlighted the temporal coordination of neuronal activities between the hippocampus and a neocortical area; however, how the spatial extent of neocortical activity relates to hippocampal activity remains partially unknown. We imaged mesoscopic neocortical activity while recording hippocampal local field potentials in anesthetized and unanesthetized GCaMP-expressing transgenic mice. We found that neocortical activity elevates around hippocampal sharp wave ripples (SWRs). SWR-associated neocortical activities occurred predominantly in vision-related regions including the visual, retrosplenial, and frontal cortex. While pre-SWR neocortical activities were frequently observed in awake and natural sleeping states, post-SWR neocortical activity decreased significantly in the latter. Urethane-anesthetized mice also exhibited SWR-correlated calcium elevation, but in longer timescale than observed in natural sleeping mice. During hippocampal theta oscillation states, phase-locked oscillations of calcium activity were observed throughout the entire neocortical areas. In addition, possible environmental effects on neocortico-hippocampal dynamics were assessed in this study by comparing mice reared in ISO (isolated condition) and ENR (enriched environment). In both SWR and theta oscillations, mice reared in ISO exhibited clearer brain state-dependent dynamics than those reared in ENR. Our data demonstrate that the neocortex and hippocampus exhibit heterogeneous activity patterns that characterize brain states, and postnatal experience plays a significant role in modulating these patterns.
Significance Statement
The hippocampus is a center for memory formation. However, the memory in the hippocampus is gradually transferred into the cerebral cortex, and synchronized activities between the neocortex and hippocampus have been hypothesized for the basis [for hippocampus-independent memory, see Sutherland and Rudy (1989)]. However, spatiotemporal dynamics between the hippocampus and whole neocortical areas remains partially unexplored. We measured neocortical calcium activities with hippocampal electroencephalogram (EEG) simultaneously and found that the activities of widespread neocortical areas are temporally associated with hippocampal EEG. The neocortico-hippocampal dynamics is primarily regulated by animal awake/sleep state. Even if similar EEG patters were observed, temporal dynamics between the neocortex and hippocampus exhibit distinct patterns between awake and sleep period. In addition, animals’ postnatal experience modulates the dynamics.
Introduction
The sleep–wake cycle is a fundamental rhythm of the brain. Ample evidence from psychological and behavioral studies shows that experience acquired during wakefulness is consolidated to long-term memory during sleep (Squire and Alvarez, 1995; Born and Wilhelm, 2012). According to the “two-stage model” of memory, experience is first stored in a labile representation by the neocortico-hippocampal interface and later is reactivated in the hippocampus to be consolidated into the higher-order cortex (Buzsaki, 1989; Frankland and Bontempi, 2005; Mitra et al., 2016). Two major distinct hippocampal EEG patterns, theta oscillations (8–15 Hz) and large irregular activity (LIA; Buzsaki, 1989; Pavlides and Winson, 1989; Wilson and McNaughton, 1994; Mizuseki and Miyawaki, 2017), have been linked to the encoding (online) and consolidating (offline) stages of memory formation, respectively. In particular, synchronized rhythmic neocortical activity with hippocampal theta oscillations has been implicated in working memory and spatial navigation (Siapas and Wilson, 1998; Jones and Wilson, 2005; Siapas et al., 2005). Moreover, sharp wave ripples (SWRs, ∼200 Hz) during LIA have been proposed to integrate hippocampal output with activities of various cortical areas including the prefrontal (Wierzynski et al., 2009), retrosplenial (RS; Battaglia et al., 2004), and visual cortex (Norman et al., 2019).
Theta and SWRs are observed both in sleep and wakefulness (Jarosiewicz and Skaggs, 2004; Jadhav et al., 2012; Roumis and Frank, 2015). Theta states dominate during attentive behavior and REM sleep while LIA represents consummatory behavior and slow-wave sleep (Buzsaki, 1989; Foster and Wilson, 2006; Carr et al., 2011; Mizuseki and Miyawaki, 2017). A number of electrophysiology studies have been conducted to find the temporal relationship between the hippocampus and particular neocortical regions (Mizuseki et al., 2011; Tang et al., 2017). While these studies have described distinct brain state-dependent temporal coordination of cortico-hippocampal activity, the spatial and temporal extent to which hippocampal activity affects the entire cerebral cortex remains partially understood.
The postnatal environment strongly influences the animal's brain function and behavior. For instance, animals reared in an enriched environment (ENR) develop higher cognitive abilities while isolated and impoverished (ISO) rearing induces anxiety-like behavior (van Praag et al., 2000; Medendorp et al., 2018). In the neocortex, ENR rearing strengthens the synchrony of local field potential (LFP) patterns across multiple areas (Mainardi et al., 2014). Both neocortical and hippocampal principal neurons increase their morphological complexity by ENR (Volkmar and Greenough, 1972; Moser et al., 1994). We previously reported that theta-associated gamma oscillations are facilitated in ENR animals compared with ISO animals (Shinohara et al., 2013; Hirase and Shinohara, 2014). In spite of detailed documentation of neural circuit alterations after ENR or ISO in individual brain regions, little is known how experience sculpts neocortico-hippocampal network dynamics.
Here we investigated how spatiotemporal activity between the neocortex and hippocampus is orchestrated in distinct behavioral states and postnatal experience by combined hippocampal LFP recording and neocortex-wide Ca2+ imaging. We used G-CaMP7 expressing transgenic mice (Ohkura et al., 2012) that allowed monitoring of neural activity through the skull (Monai et al., 2016). This configuration enabled the assessment of neocortical activity before and after SWRs as well as the determination of hippocampal theta-modulated cortical activity. We found that distinct neocortical activity patterns coincided with hippocampal SWRs are observed between sleeping versus awake state. Neocortical activities correlated with sleeping mice are predominantly observed in pre-SWR periods, whereas those correlated with awake mice exhibited prolonged activities which covers post-SWR periods. Intriguingly, such neocortico-hippocampal activity coupling is enhanced in ISO reared mice than that in ENR mice. Hippocampal theta oscillations were also well correlated with neocortical oscillations, and the rhythm was dependent on animal states. Animal postnatal rearing conditions modified both SWR-associated and theta correlated neocortical oscillatory activities.
Materials and Methods
Animals
G7NG817 mice (Monai et al., 2016) were weaned at postnatal day (P) 19 and reared for 4–4.5 weeks prior to imaging (Fig. 1A). Mice were reared under either ISO or ENR conditions. For ISO, mice were caged individually after weaning and raised in standard cages (length, 32 cm; width, 22 cm; height, 13.5 cm). For ENR, 5–8 male littermates were housed together in a larger cage (length, 44 cm; width, 27 cm; height, 18.7 cm) with a ladder, running wheels, tunnels, and toys, the location of which were changed every 5 d. Both rearing environments had a 12 h light/dark cycle, and water and food were given ad libitum. Prior to an imaging session, mice were anesthetized with ketamine–xylazine, an aluminum head plate with an imaging window was fixed to the mouse skull with dental cement, and a 16-channel silicon probe was implanted in the left CA1 region of the hippocampus. The cranial window and implanted silicon probe were protected from scratching or other physical damage by a plastic cover with aluminum taping. After surgery, mice were returned to their designated cages (i.e., ISO or ENR), and a habituation procedure for the head-restrained imaging/recording apparatus was performed every other day for the 7–10 d recovery period. All procedures involving animal care, surgery, and sample preparation were approved by the Animal Experimental Committee of Nagoya City University and performed in accordance with the guidelines of the Animal Experimental Committee of Nagoya City University.
For urethane-anesthetized (acute) experiments in Figures 3 and 6, mice were also reared in either ISO or ENR condition after weaning. All procedures involving animal care, surgery, and sample preparation were approved by the Animal Experimental Committee of RIKEN Center for Brain Science and performed in accordance with the guidelines of the Animal Experimental Committee of RIKEN Center for Brain Science.
Experimental design
Experimental design is shown in Figure 1A–C. ISO or ENR G7NG817 mice described above were used for experiments. For unanesthetized head-restrained experiments, we did surgery of implanting a head plate and a 16-channel silicon probe in the left hippocampus 7–10 d before imaging. GCaMP signals of the right side of the neocortex were transcranially imaged at 25 Hz. LFPs of the left side of the hippocampus was recorded simultaneously by the silicon probe.
For urethane anesthesia experiments, we anesthetized mice with 1.5 g/kg urethane. Then, head skin was removed, and we inserted a 16-channel silicon probe in the left CA1 region of the hippocampus and waited at least 1.5 h. Then, transcranial imaging of GCaMP signals of the right side of the neocortex was performed with LFP recording of the hippocampus.
LFP recording
LFP recording was performed by fixing animals to a stereotaxic stage via a head frame. Silicon probes for chronic implanting (A1×16-5mm-50-177-HZ16_21mm with a TdT clip; NeuroNexus) were attached to a microdrive (dDrive-m; NeuroNexus). The microdrive was installed on the skull with dental cement to target the left hippocampus through a small craniotomy above CA1 (bregma: mediolateral 1.8 mm, anteroposterior −1.8 mm). Electrode penetration was guided by monitoring the EEG waveforms and electrode depth. Mice were allowed to recover for at least 1 week before LFP recording. LFPs were recorded by an RZ2 multichannel recording system at 24.4 kHz (Tucker-Davis Technologies).
For acute experiments (Figs. 3, 6), a 16-channel linear silicon probe (interchannel distance, 50 µm; Alx15-5 mim-50-177-A16; NeuroNexus) was used. Extracellular field potentials were recorded continuously with a sampling rate of 31 kHz (Digital Lynx; NeuraLynx; Shinohara et al., 2013; Tanaka et al., 2017). Body temperature was maintained at 37°C throughout the surgery and recording sessions by a heat pad with rectal temperature feedback.
For data analysis, LFPs were resampled at 20 kHz for SWR analysis and at 1,250 Hz for theta analysis. We used Matlab code supplied by Tucker-Davis Technologies for unanesthetized mice for further resampling. For acute experiments, we used custom C program for downsampling (Sakatani et al., 2008; Shinohara et al., 2013; Tanaka et al., 2017). Downsampling of LFPs was performed by linear interpolation method.
Optical imaging
Mice were fixed to a stereotaxic stage by clamping the head frame and placed under a fluorescence stereomicroscope (MZ10F, Leica). The GFP3 filter set (excitation, 470 ± 20 nm; emission, 525 ± 25 nm; Leica) was used with an EL6000 light source (Leica). Images were acquired using ORCA-Flash 4.0 V3 CMOS cameras for chronic and ORCA-Flash 4.0 V2 CMOS cameras for acute imaging. HCImage software (Hamamatsu Photonics) was used for image acquisition (25 Hz, 512 × 512 pixels, 16 bit) and shutter control. Electrophysiological recording was synchronized with imaging by feeding the 25 Hz image acquisition TTL pulse that were fed from LabVIEW installed in another laptop PC.
Mouse behavioral state analysis
Mouse behavioral states were recorded by a digital camera during experimental sessions and analyzed thereafter. Mouse states were classified as awake, drowsy, and immobile. The awake state was defined as the presence of clear eye opening and large pupils (Reimer et al., 2016; Hulse et al., 2017). Sometimes, in awake state, mice shook their muzzles, nose, and paws. We next classified immobile state. During immobile state, mice did not move, and their eyes were firmly closed for >10 s, and behaviorally the animal appears to be asleep. Animal states that were not clearly one of these two states were defined as the drowsy state and excluded from further analysis. An example of animal behavior score is shown in Figure 2B. After behavioral state scoring, hippocampal LFPs were checked, their fluctuations were calculated, and immobile state mice in LIA state was used as sleeping state mice.
EEG data analyses and detection of brain states
EEG data analysis was carried out using MATLAB on computers running Linux or Windows.
SWR detection
Among simultaneously recorded LFPs from a silicon probe, the channel corresponding to the stratum pyramidale was identified by the presence of ripple oscillations. We found that electrode placement was stable as manifested by the channel location of ripples limited to within 100 μm (i.e., two channels of silicon probes) during chronic recordings over a week. For the analysis of SWR events, LFPs in the stratum pyramidale were first resampled to 20 kHz. Next, ripple events were detected automatically (Tanaka et al., 2017) with some improvements for noise reduction: abrupt changes in the high-frequency signal (80–250 Hz; >10 times SD) in multiple channels were removed as noise, and 96 ms LFPs around noise were excluded from further analysis by zero-padding. The LFP was bandpass filtered for the ripple frequency band (125–250 Hz for chronic, and 80–250 Hz for acute experiments), and the resultant signal was squared and then smoothed with a Hamming window of length 19.2 ms. In the first screening, ripple events were detected as the periods where the smoothed signal exceeded the mean value by 6.5 times the SD, with an interripple interval of 100 ms. In the second screening, the local minima within ±35 ms of each detected point (i.e., ripple trough) was assigned as the ripple timing, and 200 ms ripple-filtered waveforms centered around the ripple timing were extracted for further analyses. After automatic detection, detected ripples were manually confirmed for the coincidence of sharp waves in the stratum radiatum and stratum lacnosum moleculare. Sharp waves without ripples were removed from further SWR analyses.
Singlet SWR analysis
As SWRs tend to happen successively (Yamamoto and Tonegawa, 2017), we calculated intervals between SWRs. Then, only SWRs whose intervals are >280 ms were collected and further analyzed as “singlet SWRs.”
Theta detection
The channel used for the detection of theta oscillations was located 250 μm below the stratum pyramidale. Theta oscillations were detected as described elsewhere (24, 29). We computed the EEG spectrogram in the CA1 stratum lacnosum moleculare and identified periods that satisfied two criteria: (1) the ratio of the peak powers of the theta band (6.5–8.5 Hz) and the delta (2–3 Hz) band in each bin exceeded 0.6, and (2) the period was at least 10 s long. For acute experiment (Figs. 3, 6), we used the following criteria: (1) the ratio of the peak powers of the theta band (3.5–7.0 Hz) and the delta (2–3 Hz) band in each bin exceeded 0.6, and (2) the period was at least 15 s long.
For theta phase analysis (Extended Data Fig. 11-1), we assigned the phase of the theta oscillations by approximating to a sine wave by Hilbert transform and detected image frames nearest the trough, ascending, peak, and descending phases, respectively.
Large irregular activity and small irregular activity detection
As natural sleeping mice are sometimes hard to discern to immobile eye-closing mice, we used only SWRs during LIA states as sleeping states because the two states can be distinguished by the magnitude of LFP fluctuation. After theta state detection, non-theta state was classified into LIA or small irregular activity (SIA) states. LFP of the stratum radiatum was segmented into 1 min period, and the fluctuations of field potentials were calculated, and their histogram was computed. As previously reported (Kay et al., 2016), the histogram can be largely divided into two populations, and typically the state of LFPs with z-score of 1 SD magnitude less than the average was judged as SIA state.
Imaging data analysis
For all imaging data, ΔF/F was defined by taking an averaged baseline as F. The baseline was calculated by removing astrocytic Ca2+ surges, which are typically (1) a rise of 8% from baseline (Monai et al., 2016, 2021) with (2) slow dynamics that are synchronized across wide cortical areas (Extended Data Fig. 1-1). After the detection of SWR and theta events, images nearest the event were detected, and images during the sleep and awake states were selected according to the state of interest (i.e., awake or sleep). The calculated images for each session of the experiment were registered to a “standard image” of the right hemisphere of the cerebral cortex, which represents all the cortical images, which allowed us to compare images between the experimental sessions and between mice. Because imaging was performed at 25 Hz, the imaging data was filtered by low-pass filter of 12.5 Hz. After image registration, we reduced image resolution by applying 3 × 3 binning twice.
Analysis of imaging data around SWR events
We extracted 17 sequential images around SWR events; the SWR images were the seventh image in these sequences. The image sequences were named in terms of ΔF/F values; i.e., −360 ms, −200 ms, …, 280 ms, where the numbers indicate the time before or after SWRs. For the statistical significance map (Figs. 4A, 5A, 7A, 8A), the average intensities for the pixels in the cortical regions were calculated, and a t test (against 0) was performed for each pixel after 3 × 3 image binning. Pixels with p < 0.05 were visualized by multiplying the average ΔF/F values. The intensity transition graphs (Figs. 4B,C, 5B,C, 7B,C, 8B,C) were calculated by dividing the sum of intensities within a neocortical area by the area size at the time of interest. Each cortical area was defined as shown in Figure 1C.
For urethane-anesthetized mice, we extracted 27 sequential images (Figs. 3, 6) and performed the same analysis.
Neocortical large oscillation detection
For neocortical large oscillation (∼15 Hz) detection, neocortical areas were first divided into 84 subareas (Fig. 2C), and the averaged calcium signal intensities were calculated. As we detected large neocortical oscillations (Fig. 2D), we performed wavelet analysis at level 4 by Matlab and detected the local peaks of neocortical calcium signals (Fig. 2E). For peak detection, minimal interpeak distances were set to 25 frames (1 s).
Non-negative matrix factorization analysis
Neocortical areas were first segmented into 12 × 7 subareas (Fig. 2C), and singlet SWR-associated 17 consecutive neocortical calcium images were collected. After calcium activity map matrix (12 × 7 × 17) was linearized, non-negative matrix factorization (NNMF) was performed by Matlab NNMF function. Six main neocortical calcium activity components at the time of interest were extracted for each mouse. Sometimes, the linearized matrix was factorized into only five components in ENR mice. Neocortical calcium activation maps for ISO and ENR mice were respectively calculated by averaging most similar patterns of components (neocortical maps) across animals at each time point of interest. Statistically significant maps like SWR maps (Figs. 3A, 4A, 5A, 6A, 7A, 8A) were also calculated in this averaging process for each pixel by statistical comparisons with 0 (t test; p < 0.05). Afterward, the averaged NNMF maps (Extended Data Figs. 9-1, 10-1) were multiplied by the statistical significance maps (p < 0.05 map), resulting in statistically significant neocortical area maps (17 time points × 6 main components; Figs. 9, 10).
Statistical analysis
Statistical analysis was performed by Matlab (Mann–Whitney U test and t test) or R (ANOVA). For neocortical area mapping, statistical significance of each pixel was calculated by comparisons with 0 (t test; p < 0.05). The significance map was multiplied with the averaged signal intensities map, and the result was shown as a neocortical activity map (described above). For theta voltage versus neocortical calcium activity change analysis, the Spearman correlation coefficient was calculated by Matlab. For measuring main and interaction effects of rearing condition, animal states, and time from hippocampal SWR, R was used to conduct two-way or three-way ANOVA (with R-plugin anova-kun).
Results
Correlation between hippocampal EEG and neocortical GCaMP activity spectrogram
We imaged mesoscopic neocortical calcium activities on the right hemisphere of G-CaMP7 transgenic mice (Monai et al., 2016; G7NG817) while simultaneously performing LFP recording in the contralateral hippocampal CA1 with a silicon multisite probe. Major two distinct hippocampal LFP states, theta and non-theta states, were detected based on their spectral property (see Materials and Methods). Since spontaneous cortical and hippocampal LFP events predominantly occur in both hemispheres (Tanaka et al., 2017; Vanni et al., 2017), this setting enabled investigation of the temporal dynamics between the hippocampus and the wide extent of the cortex. After weaning, transgenic mice were reared in either ISO or ENR condition for 4 weeks before recording (Fig. 1A,B). Calcium signals from both astrocytes and neurons are imaged in this transgenic strain. We confined our analysis to periods that did not contain astrocytic calcium events, thereby focusing on neuronal activities (Extended Data Fig. 1-1). Neocortical areas were defined as Figure 1C (Vanni et al., 2017). In addition, to examine possible differences between natural sleep mice and urethane-anesthetized mice, we measured the neocortical activity from urethane-anesthetized mice after reared mice in ISO and ENR for 4 weeks.
Experimental design. A, G7NG817 mice were weaned at P19. B, Thereafter, one male mouse is reared alone (ISO condition; left column) or 5–8 male mice were reared in a large cage with toys and wheels (ENR condition; right column) for 4 weeks. C, Mice were restrained by a headstage (middle panel), and EEG (left panel), and imaging data (right panel) were obtained. Left panel, Representative EEG data from CA1 area in the hippocampus (HPC EEG) obtained from the left hemisphere. The red trace indicates the electrode in the center of the stratum pyramidale. Middle panel, Animal states were monitored during the experiments and data from awake and natural sleep were analyzed separately. Right panel, Representative image data obtained from the neocortex contralateral to electrode insertion (Cx image). The neocortical areas were subdivided into eight areas (Cx areas) and analyzed further; frontal cortex (F), motor cortex (M), somatosensory cortex (S), retrosplenial cortex (RS), association cortex (AS), auditory cortex (Au), V1, and V2. An example of neocortical calcium transients was shown in Extended Data Figure 1-1.
Figure 1-1
An example of pan neocortical calcium imaging conducted in this study. A. Pseudo-colored image of neocortical calcium signal and ROIs. Four ROIs were set to show the calcium transient for this example. B. Transition of ΔF/F taken from four ROIs of the neocortex. Averaged neocortical calcium transients taken from four ROIs. As reported in Monai et al., calcium surge 0.8% above baseline (baseline was taken from 50 images prior to surge) which propagated in the entire neocortical areas were regarded as astrocytic calcium surges and removed from this study. Download Figure 1-1, TIF file.
To observe overall neocortical activities along temporal window in large scale, we first divided neocortical hemisphere into 12 × 7 segments and chose one neocortical segment as a representative of each neocortical functional area (Fig. 2C). We used frontal cortex segment as representative areas for spectrogram analysis but similar oscillations were observed in other areas of the neocortex. Hippocampal EEG states (theta and non-theta) and states with their correlation with cortical GCaMP intensities of frontal cortex were calculated from a sleeping mouse (Fig. 2A). During non-theta states, we observed wider peaks of spectrogram ∼3–7 Hz, and occasionally high intensity signal ∼12 Hz was observed. During theta states, spectrogram peaks shifted into higher frequencies which correspond to theta bandwidth, and the data indicate that both the hippocampus and neocortex oscillates similar frequencies in theta state. We also examined urethane-anesthetized mice (Fig. 2B). The spectrogram exhibited monotonous frequency over time and mild change during theta state compared with natural sleeping mice. Delta frequency signals (2–3 Hz) and ∼9 Hz signals were dominant.
Neocortical calcium activities of urethane-anesthetized and head-restrained mice. A, A representative example of spectrogram of neocortical calcium activities (bottom row) of head-restrained mouse (bottom row). Mouse behavioral states were classified into awake, drowsy, and immobile by mouse movie (top row). Mice in immobile state are closing their eyes, immobile for longer than 10 s, and look asleep. In addition, hippocampal theta and non-theta states were indicated in the top row. B, An example of spectrogram of neocortical calcium activities (bottom row) of urethane-anesthetized mouse. Top row indicates the state of hippocampal LFPs. Theta and non-theta indicate hippocampal theta and non-theta state, respectively. C, Subdivision of neocortical map (7 × 12 subareas), and representative subarea of the neocortical areas for frontal (F), retrosplenial (Rs), and V1 area of the neocortex. The averaged calcium activities of F, Rs, and V1 (one square surrounded by black solid line) were used for the following analysis (D and E). D, Examples of calcium activities of neocortical subareas (F and V1) obtained from an ISO mouse in immobile state. Hippocampal SWRs are shown in red asterisks. Neocortical fast and large calcium activities are observed, and they often occurred independently of hippocampal SWRs. E, Extracted peaks of neocortical fast calcium oscillatory activities of D by wavelet analysis. The time lag between hippocampal SWR and neocortical oscillatory activities are shown in Extended Data Figure 2-1.
Figure 2-1
Cumulative histogram of the time between the peak of large neocortical calcium oscillatory activities and hippocampal SWRs. The histogram data was first obtained for each ISO mice (n = 5). Then, the summed histograms across the mice were shown in the figure. The histogram of frontal cortex (FC; upper row), retrosplenial cortex (Rs; middle row), and V1 (V1; lower row) were shown on the left (sleeping) and right (awake) column. Download Figure 2-1, TIF file.
To observe more detailed calcium activities of the neocortex, we computed the averaged ΔF/F values of F and V1 segments as representative subareas because they are located on the anterior and posterior extremes of the neocortex. Large oscillatory calcium activities (∼15 Hz) were frequently observed in sleeping mice (Fig. 2D) in both F and V1 segments. Wavelet analyses successfully detected the peak of neocortical oscillatory activities and the oscillation occurred independently of hippocampal SWRs (Fig. 2E). However, these two phenomena occasionally occurred in close proximity in time (Extended Data Fig. 2-1).
Neocortical activation associated with hippocampal SWR was observed in urethane-anesthetized mice
A study by Abdach et al. reported that neocortical activations are observed with hippocampal SWRs during both urethane anesthesia and natural sleep (Karimi Abadchi et al., 2020). We examined if urethane-anesthetized mice could express SWR-associated calcium elevations in both ISO and ENR mice (n = 797, 649, 2,853, and 3,091 SWRs from 4 ISO mice; n = 334, 918, 1,750, 2,637, 2,154, and 1,285 SWRs from 6 ENR mice). As shown in Figure 3A,B, neocortical calcium elevation of ISO mice is observed from 440 ms before SWR (described as −440 ms) and lasts to −160 ms. ENR mice exhibit more prolonged elevation of neocortical calcium (−520 to −40 ms; Fig. 3B). Calcium activities were observed in widespread areas of the neocortex, but the anterior part of the frontal cortex and lateral part of the motor and somatosensory areas were spared. Activation areas of the neocortex are relatively unchanged over time, while the intensities of increase and decrease of the neocortical activities vary across neocortical areas. Across the brain areas, temporal patterns of the neocortical calcium elevation were similar in respective groups.
Neocortical calcium activity transition around SWR in urethane-anesthetized mice. A, Top rows, Averaged ISO mouse neocortical activity transition from −760 to +280 ms around SWR (n = 4) is shown as a percentage in color maps. In these maps, areas with a statistically significant increase or decrease from 0 (p < 0.05) are shown in the time-lapse sequence. Bottom rows, Neocortical activity transition maps around SWRs are shown for ENR mice (n = 5). B, Transitions of GCaMP ΔF/F value signal from total area of the neocortex for ISO mice (left column) and ENR mice (right column). The horizontal axis shows elapsed time (ms) with respect to SWR, and the vertical axis shows average ΔF/F values for neocortical area. Time of SWR is shown in the filled triangles. Error bars indicate SEM. For ΔF/F values of individual mice, see Extended Data Figure 3-1.
Figure 3-1
Individual values of ΔF/F around SWRs are shown in urethane anesthetized mice. Upper row: Each point of the graph indicates ΔF/F value of ISO mice (left) and ENR mice (right) panel. The data shown indicated ΔF/F value of total neocortical areas of urethane anesthetized mice (Fig. 3B). Lower row: Each point of the graph indicates ΔF/F value of ISO mice (left) and ENR mice (right) panel. Data in singlet SWRs of urethane anesthetized mice (Fig. 6) are shown. Download Figure 3-1, TIF file.
Neocortical activation associated with hippocampal SWR was also observed with natural sleeping mice, but the timescale was shorter
Next, we examined if these SWR-associated neocortical elevations are also observed in natural sleeping mice. Electrodes were implanted in the ENR or ISO mouse brain at least 1 week before the LFP measurement. Among immobile states of mice, periods with hippocampal LIA activities were used as sleeping states. For neocortical activity analyses, we performed similar analyses for sleeping mice as we did on urethane-anesthetized mice, but we shortened the time windows to −360 to +280 ms around SWR because fast large neocortical calcium elevations were observed in natural sleeping mice (Extended Data Fig. 2-1) and they often masked SWR-associated neocortical calcium elevations. As shown in Figure 4A,B, neocortical activities can be observed from −280 to −80 ms around SWR in ISO, and from −240 to −40 ms around SWR in ENR mice (461, 1,064, 730, 1,976, 682, 992 SWRs from 6 ISO mice; 1,682, 701, 236, 1,807, 215, 1,261 SWRs from 6 ENR mice). We plotted the neocortical areas which are functionally relevant to the hippocampus. All three areas exhibited similar activation patterns in both ISO and ENR animals (Fig. 4C).
Neocortical calcium activity transition around SWR in natural sleeping mice. A, Top row, ISO mouse cortical activity transition from −360 to +280 ms around SWR (n = 6) is shown as a percentage in color maps. In these maps, areas with a statistically significant increase or decrease from 0 (p < 0.05) are shown in the time-lapse sequence. Bottom row, Neocortical activity transition maps around SWRs are shown for ENR mice (n = 6). B, Transitions of GCaMP ΔF/F value signal from total area of the neocortex for sleep ISO mice (left column) and ENR mice (right column). The horizontal axis shows elapsed time (ms) with respect to SWR, and the vertical axis shows average ΔF/F values for neocortical area. Time of SWR is shown in the filled triangles. Error bars indicate SEM. C, Transitions of GCaMP ΔF/F value signal from FC, RSC, and V1 are plotted for ISO mice (top row) and ENR mice (bottom row). The horizontal axis shows elapsed time (ms) with respect to SWR, and the vertical axis shows average ΔF/F values for specific cortical areas. Time of SWR is shown in the filled triangles. Error bars indicate SEM. For ΔF/F values of individual mice, see Extended Data Figure 4-1.
Figure 4-1
Individual values of ΔF/F around SWRs are shown in natural sleeping mice. Upper row: Each point of the graph indicates ΔF/F value of ISO mice (left) and ENR mice (right) panel. The data shown indicated ΔF/F value of total neocortical areas of sleeping mice (Fig. 4). Lower row: Each point of the graph indicates ΔF/F value of ISO mice (left) and ENR mice (right) panel. Data in singlet sleeping SWRs (Fig. 7) are shown. Download Figure 4-1, TIF file.
Neocortical activation associated with hippocampal SWR was also observed in awake mice, but the temporal patterns were distinct to sleeping mice
Given the neocortical calcium elevation in pre-SWR periods, we tested whether this neocortical calcium dynamics is also true in awake periods. Interestingly, the neocortical activation maps exhibited prolonged elevation of calcium in both ISO and ENR mice compared with sleeping condition (Fig. 5A; 1,427, 365, 2,838, 2,033, 1,135, 544 SWRs from ISO; 529, 766, 1,063, 4,106, 3,843, and 353 SWRs from ENR).
Neocortical calcium activity transition around SWR in awake mice. A, Top row, ISO mouse cortical activity transition from −360 to +280 ms around SWR (n = 6) is shown as a percentage in color maps. In these maps, areas with a statistically significant increase or decrease from 0 (p < 0.05) are shown in the time-lapse sequence. Bottom row, Neocortical activity transition maps around SWRs are shown for ENR mice (n = 6). B, Transitions of GCaMP ΔF/F value signal from total area of the neocortex for ISO awake mice (left column) and ENR mice (right column). The horizontal axis shows elapsed time (ms) with respect to SWR, and the vertical axis shows average ΔF/F values for neocortical area. Time of SWR is shown in the filled triangles. Error bars indicate SEM. C, Transitions of GCaMP ΔF/F value signal from FC, RSC, and V1 are plotted for ISO mice (top row) and ENR mice (bottom row). The horizontal axis shows elapsed time (ms) with respect to SWR, and the vertical axis shows average ΔF/F values for specific cortical areas. Time of SWR is shown in the filled triangles. Error bars indicate SEM. For ΔF/F values of individual mice, see Extended Data Figure 5-1.
Figure 5-1
Individual values of ΔF/F around SWRs are shown in awake mice. Upper row: Each point of the graph indicates ΔF/F value of ISO mice (left) and ENR mice (right) panel. The data shown indicated ΔF/F value of total neocortical areas of awake mice (Fig. 5). Lower row: Each point of the graph indicates ΔF/F value of ISO mice (left) and ENR mice (right) panel. Data in singlet awake SWRs (Fig. 8) are shown. Download Figure 5-1, TIF file.
In ISO mice, neocortical calcium elevation showed clear bimodal patterns. The results were similar when we analyzed the frontal, restrosprenial cortex, and V1 (Fig. 5C). Neocortical calcium elevation is observed from −280 to −160 ms but dwindles at approximately −120 to −80 ms. Calcium signals is elevated again from −40 ms and showed sustained elevation to +280 ms. On the other hand, ENR mice showed sustained elevation of calcium from −240 to +280 ms around SWRs.
Next, we conducted a three-factor ANOVA (ISO vs ENR and sleep vs awake and time) for total neocortical areas. Significant interaction of rearing environment and time was found at −360 ms (F = 8.249; p = 0.0093), −240 ms (F = 9.882; p = 0.0051), −200 ms (F = 7.7474; p = 0.0115), 0 ms (F = 5.8539; p = 0.0252), and +200 ms (F = 7.7202; p = 0.0115). Significant interaction of animal states and time was found at −360 ms (F = 8.2197; p = 0.0095), −120 ms (F = 6.2618; p = 0.0211), −40 ms (F = 10.0224; p = 0.0049), 0 ms (F = 12.7374; p = 0.0019), 40 ms (F = 9.8901; p = 0.0051), +160 ms (F = 6.5884; p = 0.0184), and +200 ms (F = 12.1023; p = 0.0024).
These results indicate that rearing condition and animal state both modify SWR-associated neocortical calcium dynamics. Effect of rearing condition is observed before SWR (−360 and −200 ms). On the other hand, effect of animal states tends to be found around hippocampal SWRs (−40 to 40 ms and +160 to +200 ms).
Singlet SWR analyses revealed similar tendencies with total SWR analyses
As SWRs tend to happen consequently (Yamamoto and Tonegawa, 2017), we asked if the results are reproducible in singlet SWRs (see Materials and Methods). So, we performed similar analyses after we chose only singlet SWRs. Neocortical activity associated with SWRs was observed in urethane-anesthetized mice between −560 and +280 ms in both ISO (383, 286, 1,849, 1,223) and ENR mice (n = 223, 511, 1,030, 1,271, 1,096, and 676; Fig. 6). The neocortical calcium elevation was observed between −480 and −200 ms in ISO mice and between −600 and −80 ms in ENR mice, respectively. Importantly, the neocortical activation patterns were similar to total SWR results in Figure 3. As observed in total SWRs in Figure 3, clear decline of calcium activity was observed after SWRs.
Cortical activity transition of urethane-anesthetized mouse around singlet SWRs. SWRs used in Figure 3, whose intervals are longer than 280 ms are collected as “singlet SWRs” and analyzed further. A, Top row, ISO mouse neocortical activity transition from −760 to +280 ms around SWR (n = 4) is shown as a percentage in color maps. In these maps, areas with a statistically significant increase or decrease from 0 (p < 0.05) are shown in the time-lapse sequence. Bottom row, Neocortical activity transition maps around SWRs are shown for ENR mice (n = 5). B, Transitions of GCaMP ΔF/F value signal from total area of the neocortex for ISO mice (left column) and ENR mice (right column). The horizontal axis shows elapsed time (ms) with respect to SWR, and the vertical axis shows average ΔF/F values for specific cortical areas. Time of SWR is shown in the filled triangles. Error bars indicate SEM. For ΔF/F values of individual mice, see Extended Data Figure 3-1.
Next, we performed a singlet SWR analysis for natural sleeping mice (Fig. 7; n = 288, 486, 326, 713, 356, and 505 for ISO; n = 839, 426, 129, 988, 102, and 744 for ENR). Neocortical activity is prominent between −240 and −40 ms SWR in both ISO and ENR mice (Fig. 7A,B). We also found smaller additional activation from +40 to 240 ms for ISO and +40 to +200 ms in ENR, but the second rise is not apparent on the maps (Fig. 7A). We also examined data of singlet SWRs in awake conditions (760, 163, 1,208, 884, 544, and 249 SWRs for ISO; 301, 415, 693, 1,833, 1,835, and 214 SWRs for ENR). As shown in Figure 8, neocortical activities are detected at −240 ms, dwindled at −120 to −80 ms in ISO mice, but they showed prolonged activities that last up to +280 ms. In ENR, neocortical activation is observed from −280 to 160 ms, but no clear sag is observed.
Cortical activity transition of sleeping mouse around singlet SWRs. A, Top row, ISO mouse neocortical activity transition from −360 to +280 ms around SWR (n = 6) is shown as a percentage in color maps. In these maps, areas with a statistically significant increase or decrease from 0 (p < 0.05) are shown in the time-lapse sequence. Bottom row, Neocortical activity transition maps around SWRs are shown for ENR mice (n = 6). B, Transitions of GCaMP ΔF/F value signal from total area of the neocortex for ISO mice (left column) and ENR mice (right column). The horizontal axis shows elapsed time (ms) with respect to SWR, and the vertical axis shows average ΔF/F values for specific cortical areas. Time of SWR is shown in the filled triangles. Error bars indicate SEM. C, Transitions of GCaMP ΔF/F value signal from FC, RSC, and V1 are plotted for ISO mice (top row) and ENR mice (bottom row). The horizontal axis shows elapsed time (ms) with respect to SWR, and the vertical axis shows average ΔF/F values for specific cortical areas. Time of SWR is shown in the filled triangles. Error bars indicate SEM. For ΔF/F values of individual mice, see Extended Data Figure 4-1.
Cortical activity transition of awake mouse around singlet SWRs. A, Top row, ISO mouse neocortical activity transition from −360 to +280 ms around SWR (n = 6) is shown as a percentage in color maps. In these maps, areas with a statistically significant increase or decrease from 0 (p < 0.05) are shown in the time-lapse sequence. Bottom row, Neocortical activity transition maps around SWRs are shown for ENR mice (n = 6). B, Transitions of GCaMP ΔF/F value signal from total area of the neocortex for ISO mice (left column) and ENR mice (right column). The horizontal axis shows elapsed time (ms) with respect to SWR, and the vertical axis shows average ΔF/F values for specific cortical areas. A sag of neocortical signal at −80 ms of ISO mice is shown in a dark arrow. Time of SWR is shown in the filled triangles. Error bars indicate SEM. C, Transitions of GCaMP ΔF/F value signal from FC, RSC, and V1 are plotted for ISO mice (top row) and ENR mice (bottom row). The horizontal axis shows elapsed time (ms) with respect to SWR, and the vertical axis shows average ΔF/F values for specific cortical areas. Sags of neocortical signal at −80 ms of ISO mice are shown in dark arrows. Time of SWR is shown in the filled triangles. Error bars indicate SEM. For ΔF/F values of individual mice, see Extended Data Figure 5-1.
Next, we performed three-way ANOVA to detect possible differences and interaction between condition, state, and time (pre-SWR, SWR, and post-SWR periods). Statistical significances were found in animal states (awake vs sleep: F = 6.2136; p = 0.0216). Slightly significant difference was found in animal rearing condition (ISO vs ENR: F = 3.9919, p = 0.0595). In addition, there are significant interactions between states and time (F = 4.5393; p = 0.000), between rearing condition and time (F = 1.7260, p = 0.048). Interactions between rearing condition and time were significant at −200 ms (F = 5.3166; p = 0.0320) and −160 ms (F = 4.5395; p = 0.0459). Interaction between animal states and time was significant at 0 ms (F = 11.0335; 0 = 0.0034), +40 ms (F = 13.7837; p = 0.0014), +80 ms (F = 14.1330; p = 0.0012), +200 ms (F = 4.8345, p = 0.0398), and +240 ms (F = 7.9588; p = 0.0106). In summary, differences between ISO versus ENR and awake versus sleep conditions detected in total SWRs were also true in singlet SWRs and become clearer than total SWRs.
Spatiotemporal patterns of neocortical calcium elevation are different between ISO and ENR mice
We found consistent differences of SWR-associated neocortical calcium elevations between sleeping and awake mice by analyzing total and singlet SWRs. Awake mice show prolonged elevation of neocortical calcium around SWRs than sleeping mice, and ISO mice show a tendency of greater calcium elevation than ENR. However, the magnitudes of SWR-associated neocortical activities vary across animals (Extended Data Figs. 3-1–5-1). So, we wondered if the neocortical activities can be temporary and spatially more organized. We performed NNMF analyses on the individual singlet SWRs because singlet SWRs might comprise simpler neocortical activities than total SWRs.
In ISO sleeping mice, neocortical activities were separated into six major patterns. The averaged activation areas were calculated across mice, and, like Figures 3A–7A, statistically significant areas maps (t test; comparisons to 0; p < 0.05) multiplied by the averaged activated area maps were shown in Figure 9. The maps were classified central cortex activation pattern (No. 1), frontal cortex activation pattern (No. 2), motor area activation pattern (No. 3), RSC and V1 activation pattern (No. 4), somatosensory activation pattern (No. 5), and V1 and V2 activation patters (No. 6). Averaged activation maps across mice showed similar pattern separation of SWR-associated neocortical activity (Extended Data Fig. 9-1A). On the other hand, ENR mice showed smaller consistent activation patterns. In both ISO and ENR groups, the classification patterns are similar across time points. Neocortical activation maps for averaged intensities across mice show that ENR mice show smaller numbers of SWR-associated neocortical activation patterns (Extended Data Fig. 9-1).
NNMF analysis of neocortical areas around singlet SWRs of sleeping mice. Top panel, NNMF analysis of neocortical calcium activities around sleeping ISO mice. The neocortical activities of each time point were separated into six major patterns for each mouse. After pattern separation, most similar activity maps between animals were matched for each time point, and the maps were averaged across mice. Statistical significance of the averaged value compared with 0 was also calculated for each pixel. Neocortical activation pattern maps with statistical significance (p < 0.05) were shown. Bottom panel, NNMF analysis of neocortical calcium activities around sleeping ENR mice. For averaged signal maps (without statistically differences), see Extended Data Figure 9-1.
Figure 9-1
NNMF analysis of neocortical areas around singlet SWRs of sleeping mice. Upper panel: NNMF analysis of neocortical calcium activities around sleeping ISO mice. The neocortical activities of each time point were separated into 6 major patterns for each mouse. After pattern separation, most similar activity maps between animals were matched for each time point, and averaged maps of mice were shown. Statistical significance maps (p < 0.05) of averaged values are shown in Fig. 9. Lower panel: NNMF analysis of neocortical calcium activities around sleeping ENR mice. Averaged maps across the mice were shown. Download Figure 9-1, TIF file.
Similar analysis was performed for awake SWR singlets of ISO and ENR mice (Fig. 10). The activation maps of awake ISO resemble to those of sleep ISO, and this finding indicates that neocortical areas which are activated in accordance with hippocampal SWRs are similar regardless of animal states. Though ENR mice showed multiple patterns of neocortical activities similar to sleep case (Extended Data Fig. 10-1). Most of the activated areas (No. 2–6) disappeared in Figure 10, and ENR mice showed only consistent patterns of activation across animals (No. 1). The results indicate that there are variabilities of areas of activation between animals in awake ENR case.
NNMF analysis of neocortical areas around singlet SWRs of awake mice. Top panel, NNMF analysis of neocortical calcium activities around awake ISO mice. The neocortical activities of each time point were separated by NNMF into six major patterns for each mouse. After pattern separation, most similar activity maps between animals were matched for each time point, and then maps were averaged across mice, and statistical significance of the averaged value compared with 0 was calculated for each pixel. Neocortical activation maps with statistical significance (p < 0.05) were shown. Bottom panel, NNMF analysis of neocortical calcium activities around singlet SWRs in awake ENR mice. For averaged signal maps (without statistically differences), see Extended Data Figure 10-1.
Figure 10-1
NNMF analysis of neocortical areas around singlet SWRs of awake mice. Upper panel: NNMF analysis of neocortical calcium activities around awake ISO mice. The neocortical activities of each time point were separated into 6 major patterns for each mouse. After pattern separation, most similar activity maps between animals were matched for each time point, and averaged maps across mice were shown. Statistical significance maps (p < 0.05) of averaged values are shown in Fig. 10. Lower panel: NNMF analysis of neocortical calcium activities around awake ENR mice. Averaged maps across the mice were shown. Download Figure 10-1, TIF file.
Neocortical oscillations are temporarily correlated with hippocampal theta oscillations
So far, we have investigated the spatiotemporal coordination of neocortical activity relative to hippocampal SWRs and described the state-dependent activity patterns. Finally, we analyzed neocortical calcium dynamics during theta oscillations. Theta periods were detected from an EEG trace in the stratum lacunosum-moleculare of the CA1 (Extended Data Fig. 11-1). We asked if the hippocampal theta rhythms are correlated with neocortical activity, and this correlation can be modulated by animal states and rearing conditions. For each frame of neocortical calcium imaging, we computed the corresponding phase of theta oscillations: ascending, peak, descending, and trough phase, respectively.
We computed the averaged images from four phases of theta cycles from sleeping mice (ISO: n = 6; ENR n = 6). The phase which shows significant elevation is limited to peak phase in ISO animals (Fig. 11A, top panels), but the elevation map is stronger than other images. One-third of anterolateral part of the neocortex, especially the frontal part and posterolateral part, is significantly activated (Fig. 11A). In descending phase, activation areas shift to the posterior part of the neocortex. Peak phase in ENR mice shows elevation of calcium in the frontal cortex. In addition, ENR showed significant decrease of neocortical activities at trough phase in the anterolateral areas in a scattered pattern.
Correlation between neocortical activities in theta state versus hippocampal LFPs. A, Averaged neocortical activities of ascending (Asc), peak (Peak), descending (Des), and trough (Trough) phase of ISO (left; n = 6) and ENR (right; n = 6) sleeping mice. In both groups, statistical significance between 0 was calculated for each pixel by t test and statistical significance maps (p < 0.05) are exhibited. For an example of actual detection of theta phase, see Extended Data Figure 11-1. B, Bottom panel, Averaged neocortical activities of ascending (Asc), peak (Peak), descending (Des), and trough (Trough) phase of ISO (left; n = 6) and ENR (right; n = 6) awake mice. Statistical significance between 0 was calculated for each pixel and statistical significance maps (p < 0.05) are exhibited. C, Correlation between neocortical calcium change and theta voltage change (peak − trough) or theta voltage gradient in sleeping mice. (1) Correlation between total neocortical changes of the calcium activities (arithmetic sum of 12 × 7 subareas) versus theta voltage change (peak − trough) in ISO and ENR mice. (2) Correlation between total neocortical changes of the calcium activities (arithmetic sum) versus theta voltage gradient between peak and trough in ISO and ENR mice. (3) Correlation between total neocortical changes of the calcium activities (root sum of square of 12 × 7 subareas) versus theta voltage change. (4) Correlation between total neocortical changes of the calcium activities (root sum of square) versus theta voltage gradient. Points with different color indicate randomly selected 300 theta cycles from each subject. D, The same analysis as Figure 11C was conducted for awake mice. Correlation between neocortical calcium change and theta voltage change (peak − trough) or theta voltage gradient in awake mice. (1) Correlation between total neocortical changes of the calcium activities (arithmetic sum of 12 × 7 subareas) versus theta voltage change (peak − trough) in ISO and ENR mice. (2) Correlation between total neocortical changes of the calcium activities (arithmetic sum) versus theta voltage gradient. (3) Correlation between total neocortical changes of the calcium activities (root sum of square of 12 × 7 subareas) versus theta voltage change. (4) Correlation between total neocortical changes of the calcium activities (root sum of square) versus theta voltage gradient.
Figure 11-1
An example of theta oscillation from natural sleeping mice obtained from lacnosum moleculare of the CA1. The theta trough (blue triangle) and peak (red triangle) was calculated by Hilber transform of LFP data. Download Figure 11-1, TIF file.
In awake states, neocortical activation pattern showed distinct patterns to awake states (Fig. 11A, bottom panels). We found that at ascending phase of ISO mice, two-thirds of the lateral part of the neocortical areas exhibit significant increase of activities in a spatially dispersed pattern. Stronger calcium elevation is observed in wider areas of the neocortex (which spares the medial neocortical areas) at peak phase. The rest phases did not exhibit significant changes of neither increase nor decrease, which implies that neocortical calcium elevations are more temporarily synchronized than reductions. The areas activated by ENR condition is wider than those in ISO, and FC is strongly activated in peak phase. In summary, neocortical activation is observed mostly at peak phase and sometimes ascending phase, but the elevated areas are dependent on both animal rearing conditions and states. Even in awake condition, ISO showed strong activation in posterolateral areas, whereas ENR showed more frontal cortex-dominant activation. At peak phase, neocortical calcium elevations are consistently observed in ISO sleep and awake states.
Next, we wondered if the fluctuation of hippocampal EEG during theta period (i.e., theta peak voltage − theta trough voltage) and magnitude of neocortical calcium activities is correlated. We calculated the difference between peak and trough from hippocampal LFP and also gradient of hippocampal LFP [calculated as (peak voltage − trough voltage)/(peak time − trough time)]. In order to measure changes of neocortical activities, neocortical areas were divided into 72 areas (as shown in Fig. 2) and total increase or decrease of the neocortical area activities were calculated (Fig. 11B,C, total change). In total change, we computed arithmetic summations of neocortical activities of the 72 segments. To calculate the change of neocortical calcium for individual segments, the sum of absolute values of changes per segment were first calculated and summed afterward (segment change).
We computed possible correlation between the EEG amplitude/gradient versus neocortical activities during sleeping (Fig. 11C) and awake conditions (Fig. 11D). The Spearman correlation coefficient between EEG volume and neocortical activities for ISO mice were calculated from randomly selected 300 pairs of peak–trough EEG and neocortical activities from each ISO and ENR mouse, respectively (total 1,800). In any combination shown in Figure 11B,C, there was no correlation between EEG/EEG gradient and neocortical activities/absolute change of neocortical activities.
Discussion
In this study, we combined transcranial neocortical calcium imaging with hippocampal LFP recording in anesthetized and unanesthetized mice and addressed the spatial extent of neocortical dynamics with respect to hippocampal activity. Additionally, we compared mice which underwent two extreme rearing conditions (ISO vs ENR; Hirase and Shinohara, 2014), and possible effects of postnatal stimulation on neocortico-hippocampal interactions were examined. High-speed pan-neocortical imaging (Monai et al., 2016; Vanni et al., 2017; Xiao et al., 2017; Cardin et al., 2020; Karimi Abadchi et al., 2020) leads to the identification of neocortical areas that are associated with hippocampal SWRs. As shown in previous reports (Karimi Abadchi et al., 2020; Pedrosa et al., 2022), vision-related areas such as the visual and RS cortex are highly associated with hippocampal SWRs. We found that neocortical activity relative to hippocampal SWRs is primarily dependent on brain awake and sleep state. We also unveiled the phase-locked pan-neocortical activity to hippocampal theta oscillations, and neocortical theta-synchronized activities were also dependent on postnatal experience. In summary, both hippocampal SWR-associated and theta-associated neocortical activities are modified by postnatal rearing conditions. Besides hippocampus-correlated activities, we also observed fast and large neocortical activities which do not always correlate with hippocampal SWRs (Fig. 2D,E).
One of the unquestionable advantages of the transcranial imaging method is minimal neocortical damage due to lack of electrode insertion. However, as neocortical transcranial calcium imaging cannot limit the source of the cortical layer of the calcium signals, the imaging technique has clear drawbacks. Time resolution of calcium-sensitive fluorescent proteins (GCaMP7) is less sensitive than electrodes (Ohkura et al., 2012). Though both sides of hippocampal and neocortical activities are symmetrical (Tanaka et al., 2017; Vanni et al., 2017), the side of recording and the side of imaging are opposite. Nevertheless, we detected fast synchronization between hippocampal LFPs and neocortical calcium signals in both around SWRs and theta periods. Usage of the surface recording electrode (Khodagholy et al., 2015) and multiple electrodes (Berenyi et al., 2014; Nitzan et al., 2022) would be also helpful for such kind of mesoscopic studies.
We found that widespread neocortical activity occurs shortly before (−280 to 0 ms) hippocampal SWR. While calcium elevations are evident in vision-related neocortical areas, visual inputs are unlikely the primary drive since this pattern is present not only in awake but also sleeping mice. The data suggest that cortical activities in various areas are involved in triggering hippocampal SWRs (Sirota et al., 2003; Rothschild et al., 2017). We also compared SWR-correlated calcium elevations between urethane-anesthetized and sleeping mice. These two groups showed distinct patterns in the neocortical calcium spectrogram. In addition, though both groups exhibited single peak SWR-associated elevation of calcium in the neocortex, the time consumed for increase and decrease of calcium was different, and we detected long-lasting clear decline of calcium activities around hippocampal SWRs in urethane-anesthetized mice. Though SWR-correlated neocortical calcium elevation is similar, this result is inconsistent with a previous paper (Karimi Abadchi et al., 2020) and suggests that there is a gap between urethane anesthesia and natural sleep in hippocampal-neocortical correlative activities.
In sharp contrast to the pre-SWR period, we identified distinct post-SWR neocortical calcium elevations (from 0 to +240 ms) depending on animals’ sleeping versus awake state. Considering the widely acknowledged notion of hippocampal replay of recently acquired experience (Wierzynski et al., 2009; Carr et al., 2011; Silva et al., 2015), the relatively lower post-SWR neocortical activity in the sleep state was unexpected. Our analysis indicates that SWR-associated neocortical- hippocampal information flow is generally unidirectional during sleep period in that neocortical activity precedes SWR. On the contrary, neocortical areas associated with vision (V1, V2, RSC, and PFC) exhibits post-SWR activity in the awake state hinting at a closed loop of neocortex–hippocampus–neocortex. Recently, the mechanism of memory transfer from the hippocampus to neocortex is gradually revealed (Goto et al., 2021). It is intriguing to imagine the possibility of awake state rather than sleeping state for memory transfer because neocortex–hippocampus–neocortex calcium elevation loop is more remarkable than sleep state.
The neocortex and hippocampus are reciprocally connected and physiologically tightly linked (Witter, 1993; Chrobak et al., 2000; Moser et al., 2014). Our results indicate that widespread neocortical regions might be involved in a bidirectional excitation loop on a longer timescale. The proportion of bidirectional activities depends on the cortical area, and the frontal part of the cortex has a relatively longer tendency for activation in the post-SWR period compared with pre-SWR activation. Interestingly, the three areas are arranged in the anterior-posterior axis of the cortex. This finding is suggestive of information transfer from the hippocampus to the cortex via SWR and integration to sensory-motor processing, in line with previous work that suggested a role of hippocampal SWR in immediate future spatial navigation (Jadhav et al., 2012; Pfeiffer and Foster, 2013).
Investigation of individual neocortical activities with respect to SWR in NNMF (Figs. 9, 10) revealed that SWR-associated neocortical calcium elevation is broadly distributed over time and their spatial pattern varies at each SWR event. As ISO sleep and awake mice exhibited similar patterns of activation map, functionally linked neocortical areas can be activated in single SWR event. Surprisingly, however, ENR mice exhibited more uniformed patterns of activity during SWRs in NNMF analysis (separation No. 1 in Figs. 9, 10). Indeed, NNMF analyses demonstrated that ISO mice showed multiple but relatively consistent patterns of calcium elevation between animals, whereas ENR mice showed uniform activation pattern (pattern No. 1) and exhibit more variabilities for less common activation patterns (No. 2–6) across animals. Additionally, ENR mice exhibited moderate elevation compared with ISO mice (V1 of sleeping mice and V1, FC, and RSC for awake mice). Our results clearly indicate that postnatal rearing conditions is an essential factor that should be accounted for electrophysiological studies.
Finally, theta phase modulation of neocortical activities was mapped at neocortex-wide level at submillimeter resolution. While theta phase-locked activity between the PFC and hippocampus has been well-established for spatial navigation (Siapas and Wilson, 1998; Jones and Wilson, 2005), our data indicate that theta phase-locked activity is spread across the entire neocortex. Interestingly, although higher neocortical activities in the posterior and medial part of the cortex than anterolateral part were commonly observed in both awake and sleep conditions (Fig. 11), the dynamics with respect to hippocampal oscillatory activities are dependent on the brain state. Distinct phase-locked activities in memory encoding versus retrieval (Hasselmo et al., 2002; Kragel et al., 2020) might be one of the reasons of differential state-dependent behaviors. In conclusion, in both the LIA and theta states, the brain state primarily regulates neocortico-hippocampal dynamics, and the postnatal environment significantly has impacts on their communication.
Footnotes
This work was supported by KAKENHI 17H02221, 20H03295, 26282222, 16H06404, and 16H01888. We are grateful for valuable discussions with Reiji Yamazaki.
The authors declare no competing financial interests.
- Correspondence should be addressed to Yoshiaki Shinohara at yoshinohara{at}yamanashi.ac.jp.
