Differential modulation of global and local neural oscillations in REM sleep by homeostatic sleep regulation

Bowon Kim; Bernat Kocsis; Eunjin Hwang; Youngsoo Kim; Robert E. Strecker; Robert W. McCarley; Jee Hyun Choi

doi:10.1073/pnas.1615230114

New Research In

Edited by Nancy Kopell, Boston University, Boston, MA, and approved January 5, 2017 (received for review September 23, 2016)

Significance

This study demonstrates that slow and fast cortical oscillations undergo different adaptations to homeostatic challenge of chronic sleep deprivation, which may benefit different functions of sleep. When mice sleep only 6 h/d for 5 d, rapid eye movement (REM) sleep settles on a persistently elevated level, even though sleep debt continues to accumulate. Using high-density EEG, we found that different forms of slow oscillations follow this general pattern, whereas all high-frequency oscillations showed progressive daily increases. Slow and fast oscillations play distinct roles in coordination of brain cell activity on different scales, and thus our results help to reconcile two seemingly opposite functions of sleep in synaptic homeostasis and sleep-dependent memory consolidation.

Abstract

Homeostatic rebound in rapid eye movement (REM) sleep normally occurs after acute sleep deprivation, but REM sleep rebound settles on a persistently elevated level despite continued accumulation of REM sleep debt during chronic sleep restriction (CSR). Using high-density EEG in mice, we studied how this pattern of global regulation is implemented in cortical regions with different functions and network architectures. We found that across all areas, slow oscillations repeated the behavioral pattern of persistent enhancement during CSR, whereas high-frequency oscillations showed progressive increases. This pattern followed a common rule despite marked topographic differences. The findings suggest that REM sleep slow oscillations may translate top-down homeostatic control to widely separated brain regions whereas fast oscillations synchronizing local neuronal ensembles escape this global command. These patterns of EEG oscillation changes are interpreted to reconcile two prevailing theories of the function of sleep, synaptic homeostasis and sleep dependent memory consolidation.

There has been substantial recent progress in understanding the neuronal mechanisms of two seemingly unrelated but, more likely, complementary functions of sleep. The first, conceptualized as the synaptic homeostasis theory (1), produced experimental evidence for a global downscaling of synaptic strength during sleep to offset the unsustainable upscaling associated with neuronal activation during the preceding period of wakefulness (2). The second line of research keeps accumulating data to support an active role of sleep in offline memory processing (3) and argues that certain synapses not only escape global downscaling in sleep but instead are potentiated; firing rate and synchrony in select neuronal ensembles representing newly acquired and deemed relevant information are increased. The mechanisms of how these two proposed functions of sleep are reconciled on the network or ensemble level are poorly understood.

From the very beginning, both processes were linked to specific oscillatory patterns of neuronal activity. The synaptic homeostasis hypothesis, and sleep homeostatic regulation in general, was correlated with EEG delta power (4) and memory consolidation was associated with episodic fast oscillations during non-REM (NREM) sleep (5). Recent evidence from network level investigations concur. On the one hand, different measures of homeostatic downscaling on the neuronal level (synaptic depotentiation, decrease in firing rate, decreased synchrony, etc.) were shown to correlate with the global decrease in delta power (6). On the other hand, reactivation of memory traces primarily occur during spindle-ripple events (7) and requires the temporal organization provided by local fast oscillations for coordination of firing within local neuronal ensembles.

Slow and fast oscillations, respectively, support global and local processing, not only in NREM sleep but also in any brain state, in general. Slow oscillations (delta, theta, alpha, beta1) allow synchronization of neuronal activity over large networks, whereas fast oscillations (beta2, gamma, ripple) coordinate neuronal firing locally. Here, we test the hypothesis that global processes, such as homeostatic regulation, may differentially affect fast versus slow oscillations, in accordance with each category of oscillations appearance and different functional roles in sleep-related processing. In fact, in NREM sleep the synaptic downscaling was shown to entirely materialize in delta-dominated interripple periods, whereas within-ripple firing and synchrony actually increased over the course of sleep (8). In rapid eye movement (REM) sleep, slow and fast oscillations occur simultaneously and are strongly coupled. However, as we demonstrate in this study using a chronic sleep restriction (CSR) paradigm, slow and fast oscillations undergo entirely different adaptations to homeostatic challenge which may benefit different sleep functions.

REM sleep shows strong homeostatic regulation; acute losses are compensated by rebounds proportional to the lost amount of REM sleep (9, 10). However, recent studies using the CSR paradigm, which better resembles common human sleep loss (11, 12), revealed limitations of this mechanism, demonstrating that after the first day of sleep loss, REM sleep rebound settles on a persistent elevated level and does not show a progressive increase over days as REM sleep debt keeps accumulating. The goal of this study was to test how this behavioral/state level regulation of sleep rebound after deprivation translates to population activity of different cortical networks. Using high-density EEG as depicted in Fig. S1 (13), we investigated the dynamics and topography of slow and fast oscillations in REM sleep during CSR in which mice were only allowed to sleep 6 h/d for 5 consecutive days. We found an overall increase in oscillatory synchronization in different frequency bands over wide areas of the cortex and the underlying hippocampus. However, there was a clear separation in the pattern of activation between slow and fast oscillations across days, such that slow oscillations followed the pattern of persistent elevation paralleling the behavioral adaptation of REM sleep, whereas power at higher frequencies showed cumulative increase, starting with slight or no change on the first day of sleep restriction (SR1) and reaching a maximum on SR5. This rule was common for different areas of the cortex even though specific frequency bands of both low and high-frequency components and their cross-frequency coupling (CFC) patterns varied by cortical region.

Fig. S1.

(A) Electrode montage. The black lines indicate the coronal and lambdoidal sutures. (B) Sample EEG signals. FrA, frontal association cortex; M1, primary motor cortex; M2, secondary motor cortex; S1, primary somatosensory cortex; V1, primary visual cortex; V2, secondary visual cortex.

Results

Persistent Increase in REM Sleep Time Rebounds After Repetitive Sleep Deprivation.

Sleep-wakefulness architecture during the daily 6-h periods from Zeitgeber time (ZT) 0 to ZT6 was analyzed for a baseline (BL) day, CSR days SR1, SR3, and SR5 and days R1 and R3 of the recovery that followed the 5 d of CSR. The timeline for the experiment is depicted in Fig. S2. During the daily sleep opportunities on CSR days, total sleep time gradually decreased throughout CSR, whereas there was no change on SR1 compared with BL (two-way ANOVA, P = 0.380; Fig. 1A). Substantial sleep loss reached 16 ± 13% on SR3 (7.0 ± 5.6 min of sleep loss per hour, P = 0.006) and 27 ± 14% by SR5 (11.1 ± 6.4 min, P = 0.001). Total sleep time fully recovered after 2 d of recovery sleep (two-way ANOVA, P = 0.051 for R1 and P = 0.804 for R3). These results were consistent with previous observations using this paradigm (11, 14).

Fig. 1.

Changes in sleep architecture across CSR. (A) Grand average of total sleep time during sleep opportunity, which is a cumulative NREM and REM states during ZT0 to ZT6. BL, SR, and R refer to baseline, sleep restriction, and recovery days, respectively. (B) Grand average of total duration of wake, NREM, and REM states within ZT0 to ZT6 plotted with respect to days. (C) Percentage of NREM and REM sleep stages. (D) Significant increase in REM episode duration within ZT0 to ZT6 in SR days compared with BL. Note that the increase was not progressive. Colors indicate individual mice. I Time-resolved spectrum in BL (top) and SR5 (bottom) in an individual mouse. Gray traces (Movement) show the signal of an accelerometer. Time is referenced to ZT0. (F) Representative sleep-wake patterns in BL, SR1, and SR5 of the same mouse shown in E. (G) Scatter plot of REM sleep episode duration in each mouse. The size of marker corresponds to the occurrence rate. The order of column in each day block corresponds to the order of animals, representing the same order for different days. (Inset) Grand average of REM sleep episode duration. Error bars indicate the SEM of the mice. Only sleep across ZT0 to ZT6 was counted. All comparisons were carried out using nonparametric ANOVA. *P < 0.05.

Fig. S2.

Experiment timeline. Mice are nocturnal and they preferentially sleep when room light is on. The mice had sleep opportunity in the first half of daytime (from ZT0 to ZT6), and their sleep was restricted in the rest of the day by placing them in a slowly moving wheel. For convenience, the day started at ZT6 and ended at ZT6 in the following day because sleep restriction started at ZT6. Recordings lasted from 1 d before to 3 d after sleep restriction. Screw EEG and motion sensor were continuously recorded throughout experiment, and high-density EEG was collected only in from ZT0 to ZT3.

Decomposition of total sleep to NREM and REM sleep states showed that the reduction of total sleep time during CSR was due to reduction in NREM sleep (Fig. 1B), whereas the time spent in REM sleep actually increased. Importantly, REM sleep rebound immediately appeared during the first sleep opportunity on SR1 which occurred immediately after the first 18 h of acute sleep deprivation and continued throughout the sleep deprived nights on a nearly constant level (2.2 ± 1.0, 1.8 ± 1.2, and 2.3 ± 1.2 min/h REM sleep increase on SR1, SR3, and SR5, respectively). The amount of this rebound REM sleep increase was between 40–70% of normal REM sleep time (cf., 3.8 ± 1.2 min/h on the day before CSR). REM sleep rebound disappeared as soon as CSR ceased, in R1 (two-way ANOVA, P = 0.632). Persistent increase in REM sleep and progressive reduction of NREM sleep resulted in a cumulative homeostatic increase in REM/NREM ratio (Fig. 1C).

We further analyzed the details of REM sleep time to verify the pattern of persistent REM sleep elevation as opposed to a progressively increasing rebound due to accumulation of homeostatic REM sleep drive over 5 d. First, analysis of REM sleep rebound in individual mice revealed an SR1 enhancement in all mice, with no coherent additional enhancements on the subsequent days (Fig. 1D; null-hypothesis test for slope of a regression line, P = 0.859) [i.e., four of nine mice showed positive trends (R² = 0.368 ± 0.126), whereas the remaining five showed negative trends (R² = 0.739 ± 0.119)]. Second, we analyzed REM sleep episode lengths in individual mice and found that longer REM sleep episodes occurred more frequently after sleep deprivation. Fig. 1 E and F compares BL and SR5 in a representative experiment on an individual mouse; the duration of each REM episode for each mouse is plotted in Fig. 1G. Average REM sleep episode duration increased significantly from 77 ± 10 s during BL by 24 ± 9 s, 26 ± 11 s, and 21 ± 22 s in CSR days 1, 3, and 5, respectively (two-way ANOVA: P = 0.0001 for SR1, P = 0.0001 for SR3, and P = 0.02 for SR5) with no positive trend from SR1 to SR5 (Fig. 1G, Inset). Regression analysis on median and maximal values of REM sleep episode duration was performed; neither of these parameters correlated with CSR days but both showed statistically significant, persistent increase compared with the BL.

Persistent Increase in REM Sleep Low-Frequency Oscillations During CSR.

The Fig. 2 A and B heat maps show the topographies of relative EEG spectral power during REM sleep after sleep deprivation compared with baseline power at each frequency in the range of 1–90 Hz. Color represents the relative power change and black dots on each map indicate the channels with significant change in power (paired t test, P < 0.05) compared with REM sleep in BL. The hertz-by-hertz relative power maps indicate that sleep deprivation affects the EEG rhythms in a frequency and region dependent manner. Potentially, some of these observed changes in different EEG components may reveal progressive homeostatic modifications from SR1 to SR5, in contrast to the persistent REM sleep time rebound. We used two statistical measures to identify these patterns in the EEG power in different frequency bands. First, spectral power in different frequency bands were averaged within cortical regions on SR1, SR3, and SR5 and compared with BL (Table S1). Second, linear regression of EEG spectral power in these frequency bands was calculated for individual mice and the slope was statistically tested for the group using the sign test to distinguish the patterns of progressive increase or decline from persistent level throughout SR1 to SR5 (Table S2).

Fig. 2.

Changes of slow and fast EEG rhythms in REM sleep. Topographic changes of slow (A) and fast (B) EEG powers in SR days with respect to BL power in REM sleep. The numbers next to each map show the corresponding EEG frequency in Hz. The channel locations of significantly increased or decreased EEG power with respect to baseline power are marked as black dots. The rhythms of interests were defined as follows: delta (1–4 Hz), theta1 (5–7 Hz), theta2 (8–10 Hz), beta1 (12–16 Hz), beta2 (18–30 Hz), gamma1 (30–50 Hz), and gamma2 (70–100 Hz). Note increased power (yellow to red to dark red and black dots) in slow rhythms in A in both SR1 and SR5 over different regions in contrast to increases in high-frequency power (B) limited to SR5. (C) The location of EEG electrodes over different cortical regions and power ratio (±SEM) of significantly changed EEG rhythms across SR and R days. Note persistent increases in slow rhythms and progressive increases in fast rhythms. The cortical regions of brain are shown as the representation of EEG channels. A regression analysis was performed on EEG powers with respect to SR days in each mouse, and P values show the null-hypothesis test on the slope sign. The r value was calculated for significantly increasing rhythm. Note persistent elevation for slow rhythms (upper row) and progressive rise in power for fast rhythms (lower row).

View this table:

Table S1.

Statistical analysis of changes in EEG power compared with BL

View this table:

Table S2.

Statistical analysis of progressive changes in EEG over CSR

Fig. 2A compares low-frequency oscillations on SR1 and SR5, and Fig. 2B compares high-frequency oscillations, derived from the same EEG signals. At low frequencies, statistically significant increase in power was observed from SR1 through SR5 in different frequency bands (black dots in the maps in Fig. 2A) accompanied with modifications of the dominant frequencies within these bands. In contrast, fast oscillations were not altered in SR1, but distinct high-frequency EEG components significantly increased in SR5 at specific locations (Fig. 2B).

Delta power (1–4 Hz).

Persistent elevations in the delta range were concentrated in specific narrow-band oscillations in the motor and sensory cortex at 1 Hz and in the prefrontal cortex (PFC) at 2 and 4 Hz (Fig. 2A). When averaged over the motor and the sensory cortex regions, 1-Hz power showed significant elevations after SR1 (P = 0.045 and P = 0.004), SR3 (P = 0.027 and P = 0.031), and SR5 (P = 0.021 and P = 0.038; Table S1) compared with BL with no significant trend of increase or decrease over SR days in either the motor or sensory cortex (P = 0.169 and P = 0.512; Table S2). A widely distributed significant increase at 4 Hz (P = 0.043 and P = 0.011 in motor and sensory cortex) after SR1 became restricted in the following CSR days to a strong tendency of increase, selectively localized in the two most anteriorly located leads (channels 1 and 2) (P = 0.092 and P = 0.090 on SR3 and SR5, respectively), however, again with no significant trend over the SR days (Table S2). The spatial and frequency distribution of narrow-band delta oscillations during REM sleep localized at specific regions were markedly different from wide-band delta activity of NREM sleep which showed uniform increase during CSR at all delta frequencies over the entire cortex (Fig. S3).

Fig. S3.

The hertz-by-hertz topographies of changes in EEG powers during NREM states analogous to Fig. 2 A and B. The power ratio corresponds to grand average over all of the NREM episodes from nine animals. (A and B) Power changes in slow rhythms (A) and in fast rhythms (B). The numbers next to each map show the corresponding EEG frequency in hertz. The channel locations of significantly increased EEG power with respect to baseline power are marked with black dots. Note that power increased broadly over the frequency bands of delta, theta, and beta rhythms in SR1 and further to low gamma in SR5. The anatomical area of statistically significant increase in power was broad and similar for different rhythms.

Theta power (5–10 Hz).

Theta power (5–10 Hz) showed similar dynamics to the alterations observed in delta power [i.e., an initial wide-band (5–10 Hz) theta increase on SR1 compared with BL over the entire cortex and the underlying hippocampus was followed by a robust redistribution of power within the theta band after repetitive sleep deprivation]. Theta2 power (8–10 Hz) followed the pattern of the persistent elevation throughout the next 4 d of SR [P < 0.05 for SR1, SR3, and SR5, over all cortical regions (Table S1) and no significant trend in alterations over subsequent SR days (Table S2)]. In contrast, low-frequency theta (5–7 Hz) showed a reduction in power starting on SR3 and reaching its maximum reduction on SR5 (decrease on SR5: P = 0.002; and negative trend: R² = 0.72 ± 0.13, P < 0.001).

Progressive Increase in REM Sleep High-Frequency Oscillations During CSR.

Beta power (15–30 Hz).

Overall, beta activity after SR1 was similar to BL but then gradually increased in subsequent days in the somatosensory cortex (not significant on SR1 and SR3; P = 0.009 on SR5; Table S3). This trend was verified by linear regression (R² = 0.43 ± 0.11, P = 0.022; Table S4). However, post hoc detailed analysis of frequency distribution within the beta power band (Fig. 2C) showed that homeostatic activation of beta power in the somatosensory cortex was limited to beta2 frequencies (18–27 Hz; strongest at 27 Hz) (R² = 0.57 ± 0.11, P < 0.001), whereas changes in beta1 power (12–16 Hz) followed the pattern of persistent elevation throughout SR1-SR5 (tendency of persistent increase: P = 0.031, P = 0.109, and P = 0.054 in SR1, SR3, and SR5, respectively; with no trend, P = 0.729).

View this table:

Table S3.

Changes in traditional wide-band frequency ranges [delta (1–4 Hz), theta (5–10 Hz), and beta (20–50 Hz)]

View this table:

Table S4.

Statistical analysis of the direction of progressive changes in EEG over repetitive SRs

Gamma power (30–90 Hz).

Homeostatic changes were observed in specific EEG gamma band frequencies during REM sleep responses to CSR (Fig. 2B). Similar to beta2 power alterations, gamma power showed a slight but nonsignificant increase after SR1 which was followed by a progressive rise after repetitive days of SR. After SR5, enhancements in the gamma1 range (30–50 Hz) compared with BL were observed over a wide area including prefrontal (P = 0.025), motor (P = 0.039), and sensory (P = 0.090) cortices, with positive trends verified by regression analysis in all regions (R² = 0.65 ± 0.11, P < 0.001 in PFC; R² = 0.60 ± 0.12, P = 0.008 in motor; and R² = 0.52 ± 0.12, P = 0.022 in sensory cortex). In the gamma2 range (70–90 Hz) EEG power significantly increased in sensory cortex (and possibly in the underlying hippocampus) in SR3 and SR5 but not in SR1 (P = 0.782 in SR1, P = 0.049 in SR3, and P = 0.010 in SR5) and an increasing trend was observed in the motor cortex (R² = 0.42 ± 0.13, P = 0.013).

Table 1 and Table S5 summarizes the statistically significant changes in distinct EEG rhythms during REM sleep throughout the periods of SR and recovery sleep compared with the BL. A persistent increase (i.e., a pattern similar to increase in REM sleep time) was found at low frequencies, including selective delta range oscillations (1 Hz in somatosensory motor and 4 Hz in PFC), theta2 (over all regions indicating its primary origin in the underlying hippocampus), and beta1 (over the somatosensory cortex) rhythms. In contrast, high-frequency oscillations showed progressive increase over the CSR days, specifically in the beta2 (sensory cortex), gamma1 (prefrontal and somatosensory-motor cortex), and gamma2 band (centroparietal leads, suggesting somatosensory cortex and volume conduction from the hippocampus).

View this table:

Table 1.

Summary of significant relationship between EEG power and sleep pressure

View this table:

Table S5.

Summary of spatial–frequency relationship between EEG power and sleep pressure

Interaction Between Slow and Fast Oscillations in REM Sleep During CSR.

It was also noticeable in the topographies (Fig. 2 A and B) that progressive increase in fast oscillations at different frequencies colocalize with increases in low-frequency oscillations characteristic for specific areas. Thus, progressive increases in beta2 (18–27 Hz) coincide with enhanced narrow-band oscillations at 1 Hz in motor cortex; progressive increase in gamma1 power (30–50 Hz) coincide with narrow-band 4-Hz oscillations persistently increased in PFC, and the region of progressive increase in fast gamma power (70–100 Hz) overlaps with the region of permanent increase in theta2 oscillations. Therefore, we next investigated the possibility of specific cross-frequency relationship between these slow and fast oscillations and their topography over different cortical regions. It was suggested that CFC facilitates the coordination of activity between larger networks oscillating at lower frequencies and local neuronal ensembles oscillating at higher frequencies (15).

The modulation index (MI) between the phase of slow EEG rhythms (1–20 Hz) and the amplitude of high-frequency EEG components (20–100 Hz) was calculated for each day, for all signals and the results, averaged across experiments, are presented as maps of comodulograms (Fig. 3). Fig. 3A is the array of comodulograms of BL REM sleep and Fig. 3B shows the changes in SR5 relative to this baseline. Three topographically distinct CFC patterns can be clearly distinguished. First the most prominent CFC was observed between theta2 and gamma2 oscillations in centro-parietal recordings. Differential comodulograms (Fig. 3B) show a strong enhancement of this CFC component by CR5, along with a spatial expansion covering a wider area. Second, less prominent but significant MI was observed between 1 Hz and beta2 in motor and somatosensory cortex in BL which also expanded to the most lateral EEG leads in SR5. Finally, in SR5, a highly localized CFC pattern between delta phase and gamma1 amplitude appeared in the two most anterior EEG leads (channels 1–2) over the PFC.

Fig. 3.

Increased CFC of slow and fast rhythms in the REM sleep of SR5. (A) Group averages of comodulogram at BL. MI quantifying modulation of high-frequency oscillations (f_A, 20–100 Hz) by the phase of low-frequency oscillations (f_P, 1–13 Hz). Strong phase-frequency coupling was observed between theta phase and gamma2 amplitude in broad region of centro-parietal cortex (dark red spots). Couplings between 1 Hz and beta2 (20–27 Hz), and between delta and gamma1 were found in temporal and frontal cortex, respectively. (B) Differential comodulogram (ΔMI) of SR5 with respect to baseline comodulogram. Note strong increase in theta2–gamma2 coupling over a wider area shifted to higher f_A and f_P frequencies [i.e., a decrease at f_A ∼60-Hz, f_P ∼7-Hz frequencies (blue spots) and increase at f_A ∼80-Hz, f_P ∼9-Hz frequencies (dark red)]. Enhanced delta range (2–4 Hz) modulation of gamma1 (∼40 Hz) was restricted to the two most anteriorly located leads over the PFC, whereas no change in was seen in low-frequency modulation of beta2 oscillations in sensory-motor cortex, relative to BL.

Hippocampal Theta and Coupled Gamma Activity in REM Sleep During CSR.

Hippocampal theta–gamma nesting is the most well studied CFC phenomenon. The location of the strong comodulation between theta2 phase and gamma2 power in centro-parietal EEG leads with local maxima at approximately anteroposterior = −2.0 and mediolateral = ±1 mm in CSR (Fig. 3) matched the extent of the underlying hippocampus indicating that at least the theta component originated from the massive hippocampal theta generator. Throughout the experiment (BL, SR, and R days), gamma amplitude was modulated by theta phase (as exemplified in Fig. 4D), but the frequency band for stronger comodulation broadened and shifted to the right with SR days as shown in an individual mouse in Fig. 4E. The center frequency pairs calculated as ∫610MI(f)×fdf/∫610MI(f)df for each mouse for each day (Fig. 4F) showed a strong linear trend (R² = 0.4). Statistical comparison (P < 0.05, paired t test, n = 9) of MI values on different SR and recovery days relative to BL verified the significant increase in CFC strength starting on SR1 and lasting until SR5 at theta2–gamma2 frequencies and a progressive decline of CFC at theta1–gamma2 frequencies starting on SR3 (Fig. 4G).

Fig. 4.

Changes of theta and coupled gamma activity in REM sleep during CSR. (A) Average power spectra (±SEM) in the theta frequency band (5–11 Hz) during REM sleep in BL, SR1, and SR5. Marker on bottom shows the frequencies of significant differences in SR1 (yellow) and SR5 (red) compared with BL (gray). Note that double peaks are prominent in SR5. (B) The full-width half-maximum (FWHM) of the power spectrum across SR and R days. The broadening of power spectrum in theta rhythm was statistically significant in all SR and R days. (C) Frequency of spectral peaks from BL and SR and R days in the individual experiments. Asymptotic lines were fit to emphasize the bifurcation of peak frequency. (D) Representative signal traces of raw and filtered EEG above hippocampus. The filtered signals show a prominent gamma2 modulation corresponding to the positive peaks of theta rhythm. The modulation of gamma1 oscillation by theta phase is less prominent. (E) The MI of theta and gamma oscillations from individual mice in BL and SR and R days (sorted by baseline value). The phase frequency nesting gamma was broaden and shifted to the higher value. (F) The scatter plot of center frequency in the comodulogram showing positive correlation between phase (f_P, theta) and amplitude frequency (f_A, gamma). (G) The differential comodulogram in the theta and gamma bands obtained by subtracting the BL comodulation. The region of significant change compared to BL (P < 0.05, paired t test) was marked by the boundary. Note enhanced gamma modulation by theta2 from SR1 to SR5, and decreased modulation by theta1 in SR3 and SR5.

We next studied the nature and the possible origin of the theta2 rhythm which appears in CSR. Fig. 2A shows the opposing behaviors of low and high-frequency theta oscillation during CSR which could be due either to a shift of the peak frequency or a discrete switch between the 7-Hz theta in BL and a dominant 9-Hz oscillation in SR5 (Fig. 4A). Starting on SR1, the power spectra showed a permanent broadening to cover the 6–10-Hz range in all mice which persisted even after recovery sleep (Fig. 4B; tested for half-width of the spectra, P < 0.05). The peak frequency of the power spectrum in each mouse was obtained by using zero-crossing of first derivatives of the power spectrum. In most of the mice, double peaks in the power spectrum were observed and a clear separation of the two bands at ∼7 Hz and at ∼9 Hz was noticed (Fig. 4C). This frequency bifurcation suggests a transition of REM theta oscillation from unimodal to bimodal oscillations.

As fast theta oscillation normally appear during short episodes of phasic REM sleep, we also tested whether the emergence of the theta2 component in CSR is due to an increase of phasic REM sleep episodes. Phasic REM sleep episodes were identified according to the criteria of Montgomery et al. (methods section of ref. 16) during ZT1–3 every day, using high-density EEG averaged over centro-parietal cortex channels. In BL, phasic REM epochs covered 1.0 ± 0.4% of total REM sleep time. This ratio significantly decreased in SR1 (0.7 ± 0.3%; P = 0.01), but returned to the baseline in the following days (1.0 ± 0.6% in SR3, 1.5 ± 0.6% in SR5, P > 0.05). The average duration of phasic REM sleep episodes did not change either (1.24 ± 0.42 s in BL and 1.07 ± 0.17, 1.21 ± 0.24, and 1.15 ± 0.14 s in SR1, SR3, and SR5, respectively), thus opting out phasic REM as a source of theta2 rhythm in SR days.

Low- and High-Frequency Oscillations in the Frontal Cortex in REM Sleep During CSR.

Although delta activity (1–4 Hz) is the signature EEG pattern of NREM sleep, strong activity in the delta band appears in active waking as well, associated with cognitive tasks (17), and was also observed during REM sleep in this study, mainly in the frontal cortical leads (Fig. 2A). The characteristics of this activity, however, differ from the wide-band NREM delta; they appear as real, narrow-band oscillations at different frequencies within the delta range. The sinusoidal character of these oscillations in which the phase can be unequivocally identified as monotonically rising, a necessary condition for CFC analysis (18), allowed testing their capacity to modulate the homeostatically increased high-frequency oscillations in CSR.

In the PFC, EEG power increased at 2 and 4 Hz starting in SR1 and was maintained throughout SR5. As shown in Fig. 2A, after SR1, this activity became highly localized in space (channels 1 and 2) and frequency (e.g., no activity at 3 Hz; Fig. 2A). Fig. 5 A and B show an example of the raw signal recorded over the PFC and the corresponding high-resolution time-frequency plots (short-time FFT: Δf = 0.5 Hz, Δt = 100 ms), demonstrating the narrow-band, sinusoidal character of delta-band oscillations, at 2 Hz and at 4 Hz. Similar to awake animals (17), most of these oscillations were short-lived (with duration of <10 s), explaining the clear tendency but with significance at P = 0.09 (Fig. 2A and Table S1). The episodic nature and the short total length of these oscillations also put limitations on the statistical analysis of MI. A well-defined (i.e., limited in both low and high frequencies) delta–gamma1 CFC was present on the average CFC plots (Figs. 3A and 5C), but the difference shown between BL and SR5 (Figs. 3B and 5C) could not be statistically verified (P > 0.05).

Fig. 5.

Alternating signals between 2 and 4 Hz in prefrontal area, modulating gamma1 during SRs. (A) Example of EEG trace during SR5 on prefrontal channel (channel #1) showing segments of 2-Hz and 4-Hz oscillations. (B) Power spectrogram (FFT, window size = 2.048 s, 50 ms moving). Two oscillations are alternating. The color was scaled to fit the power range (300–2,100 mV²/Hz). (C) Comodulogram between delta and fast oscillation on prefrontal channel (channel 1). A weak coupling between delta (2–4 Hz) and gamma1 oscillation (30–40 Hz) in baseline sleep became strong as CSR continued; however with statistically insignificance (P = 0.29 and P = 0.28 for SR3 and SR5, respectively). (D) Comodulogram between delta and fast oscillation on motor cortex (channel 7). A broad band of fast oscillation was weakly modulated by 1-Hz oscillation in baseline sleep, and the modulated range became reduced in SR days. The color ranges from 0 to 2 × 10⁻⁴.

Oscillations at 1 Hz were strongly enhanced by CSR from SR1 throughout SR5 over the frontal cortex (Fig. 2A). CFC analysis indicated that the amplitude of ∼25-Hz beta2 activity was modulated by the phase of ∼1-Hz slow waves, both in BL and SR5 (Fig. 3). CFC at these frequencies did not increase after sleep deprivation, although its character changed (Fig. 5D). In CSR days it represented a clear peak, whereas in BL and R3 CFC values were high in a wide range (20–70 Hz) which may indicate spurious CFC generated by nonlinearities due to nonsinusoidal character of the slow waves (18).

“Delta power in NREM sleep” represents wide-band activity enhanced in CSR over the entire cortex (Fig. S3A) and is markedly different from the localized, narrow-band oscillations in REM sleep, even in SR1 when the initial activation of low-frequency EEG was less focused (Fig. 2A). It was shown recently (8) that the temporal dynamics of NREM delta, namely its progressive reduction across sleep (1), is closely related to the EEG activity of the intervening REM sleep episodes. We tested whether this relationship remained in effect during CSR by calculating NREM delta power in pairs of consecutive NREM sleep episodes separated by REM sleep and found a significant reduction of NREM delta across REM sleep epochs (paired t test, P = 5.31 × 10⁻⁶), enhanced in SR3 and SR5 (Fig. S4). The near homogeneous increase in low-frequency activity in NREM sleep (Fig. S3) indicates that further analysis, using different technics, is necessary for assessment of CSR-induced alterations of NREM activity, e.g., including nonspectral measures of local characteristics of delta waves and transient events, such as sleep spindles and hippocampal ripples. The differences in the behavioral (sleep) pattern of REM and NREM sleep (Fig. 2B) may also require different approaches which were out of the scope of this report.

Fig. S4.

(A) The effect of REM sleep on NREM delta power of NREM sleep was examined by comparing delta power in pairs of two consecutive NREM states separated by a REM state (Left) and compared with NREM states separated with arousal and short NREM episodes (Center) or waking (Right). Note the significant reduction in NREM delta power before and after REM sleep (paired t test, P = 5.31 × 10⁻⁶) and no change in the other two NREM sequences [paired t test, P = 0.65 for NREM–(NREM) –NREM and P = 0.42 for NREM–(wake) –NREM]. (B) The decrease in delta power across REM sleep episodes in the course of CSR. The percentage shown in the plot corresponds to reduction of NREM delta power after REM compared with delta before REM. The REM–NREM pairs with no or short (<30-s) arousal (bottom row) and with intervening long waking (top row) were analyzed separately. Note that statistically significant reduction in NREM delta power was only observed when NREM was interrupted by REM sleep (with no or short arousal), and the amount of reduction was larger in SR3 and SR5 days compared with BL and SR1 days. This reduction recovered to the baseline level in R days.

Discussion

We studied the spatial-temporal alteration of EEG oscillations in REM sleep during CSR (18 h daily sleep deprivation for 5 d). Our high-density EEG, covering the entire mouse cortex from frontal to occipital, revealed marked topographic differences between dominant slow and fast oscillations and their coupling in PFC, motor and somatosensory cortex, and the underlying hippocampus. The pattern of changes induced by CSR, however, followed a common rule across all areas. Slow oscillations (<15 Hz) repeated the behavioral pattern of REM sleep alterations (i.e., a persistent increase throughout SR1 to SR5, whereas high-frequency oscillations showed progressive increases over the course of CSR). CFC of slow and fast oscillations during REM sleep were preserved or even enhanced in CSR with a matching topography [i.e., a prominent (8–10 Hz) theta2–gamma2 (70–90 Hz)] nesting was detected in the areas overlaying the hippocampus, narrow-band delta range oscillations (2 and 4 Hz) were modulating gamma1 (35–45 Hz) in the PFC and a weaker slow (∼1 Hz) modulation of beta2 (20–30 Hz) was observed in the sensory-motor area. Collectively, these findings suggest that in REM sleep, slow oscillations closely following the behavioral pattern imposed by REM sleep regulation may serve to translate this top-down homeostatic control to a wide range of cortical networks whereas fast oscillations escaping this control might serve as an instrument granting relative independence for local ensembles to synchronize on shorter spatiotemporal scales.

Noncumulative REM Sleep Rebound.

Our results confirmed prior observations (11, 12, 19, 20) that REM sleep duration increases significantly after the first day, and the increase is maintained on the same level in the following days of CSR. REM sleep episode duration also showed persistent increase (ref. 12 and this study) along with other measures of REM sleep behavior, such as higher REM sleep episode number and shortened REM sleep latency (12). This pattern is different from that observed in acute sleep deprivation in which REM sleep rebound positively correlates with sleep pressure (9, 10). In CSR, daily accumulation of REM sleep loss [∼1.5 h/d (11)] did not lead to progressive increasing REM sleep rebound. Every day, during the 6 h sleep opportunity the rats only restored 10–15% of REM sleep lost in the previous 18 h, thus outlining a limited working range of the behavioral REM sleep homeostat in this paradigm (14). Organized by a complex network of cortical and subcortical structures and multiple neurotransmitter systems (21 ⇓–23), this pattern represents high level control with global effects on the animal’s state and behavior. We used this pattern as reference for the assessment of spatial, temporal, and frequency alteration of EEG in this study.

The Pattern of REM Sleep Rebound Is Mirrored in EEG Slow Oscillations.

EEG changes equivalent to the pattern of behavioral REM sleep rebound were found in the EEG low-frequency bands (<15 Hz). It is worth emphasizing two technical conditions that facilitate the accurate identification of this pattern. First, high-density EEG was necessary not only to show differences between cortical areas during REM sleep but more importantly, to show that the basic principles of homeostatic regulation are common, although expressed by locally specific slow rhythms. Second, analysis of average EEG power in traditional frequency bands is misleading; wide-band increases only occurred as part of the initial, “acute” response, i.e., after SR1 (Table S3). Later on, after CSR days 3 and 5, the spectral structure of these EEG components, showed a progressive reorganization. While maintaining power on a stable level, activity in the delta band (1–4 Hz) became more and more focused over the days of CSR in both frequency and space, i.e., by SR5 narrow-band delta range oscillations persisted exclusively in motor cortex at 1 Hz and in prefrontal at 2 Hz and 4 Hz. Similarly, theta rhythm first bifurcated to 7- and 9-Hz components from a single mode oscillation during baseline and then entirely shifted into a narrow, theta2 (8–10 Hz) band.

Subcortical structures are an integral part of slow rhythm generators and modulation of these thalamo-cortical (24) and septohippocampal networks (25) may provide direct and wide access for brainstem sleep control mechanisms to cortical function. The intimate relationship between sleep homeostatic regulation and NREM delta rhythm has long been recognized (1, 4). More recently, REM sleep’s most prominent slow rhythm, hippocampal theta was also shown to play an essential role in homeostatic sleep regulation by Grosmark et al. (8). These authors demonstrated that significant reduction in delta power, a proposed mechanism for synaptic downscaling (1), occurs from each pre-REM to post-REM episode of NREM sleep, and the decrease is strongly correlated with theta power in the intervening REM sleep. Based on these findings, an increase in REM sleep duration and especially in REM/NREM ratio in response to CSR (Fig. 1C) could intensify synaptic depotentiation when time available for sleep is limited. Enhanced theta activity within each REM sleep episode (Fig. 2) would further strengthen this effect. It is worth noticing that the increase in theta power which immediately occurred on SR1 and lasted throughout CSR was restricted to 8–10 Hz, i.e., to the range which most affected post-REM homeostatic processes (8). Narrowing of the theta band is also favorable for CFC, suggesting that at later stages of CSR, theta may also represent an instrument that remains accessible for global regulation to exert control over local circuits by modulating high-frequency oscillations (Fig. 3) which progressively escape the homeostatic restrain (see below).

The origin of the suppression of theta1 at later stages of CSR remains unknown. It is tempting to speculate that such a “blue-shift” of theta peak frequency (Fig. 4 A–C) and CFC (Fig. 3) is either a sign of increased phasic REM sleep or of the intrusion of wake-like theta into REM sleep. Phasic REM sleep is a complex process which is hard to detect reliably without pontine waves and non-EEG signs [e.g., eye movements, heart rate, etc. (22, 26)]. However, using a detection algorithm based on time-frequency criteria of the theta signal (16) in this study, we did not find changes in the occurrence or in duration of phasic REM episodes during CSR. Evidence for wake-theta intrusion is also weak. First, in wake exploration theta is correlated with running speed (among other factors) on a linear scale and does not show the abrupt switch from 7 to 9 Hz seen in CSR (Fig. 4C). Second, there is not enough data in regard of the neurotransmitters activated in waking arousal but suppressed in REM sleep. Norepinephrine, one of these neurotransmitters which was shown to specifically activate theta2 in exploration (27) did not show changes in receptor mRNA levels during CSR (14).

Prominent delta band activity, such as slow waves (∼1 Hz) and wide-band delta (1–4 Hz) normally occur in NREM sleep but was also reported in REM sleep after acute sleep deprivation (20, 28), similar to our finding on SR1, linking them to homeostatic processes. “Pure”, narrow-band oscillations were also reported in the delta band during waking, at 3 Hz in motor cortex during isometric tracking (29) and in PFC at 2 Hz and 4 Hz, associated with specific cognitive tasks (17) and stimulation of brainstem arousal pathways (30). They were shown to strongly modulate gamma activity and play a role in PFC-hippocampal communication. Narrow-band delta (2 Hz and 4 Hz, but not 3 Hz; Fig. 2A) normally missing in REM sleep (31) showed strong presence in CSR together with narrow-band gamma1 (34–40 Hz; Fig. 2B) and prominent CFC (Fig. 3), strictly constrained to the two electrodes above PFC. Their role remains unknown but the findings appear consistent with the suggestion that, similar to theta, they provide some control over the run-away high-frequency oscillations in areas involved in mnemonic function such as the PFC and hippocampus.

Progressive Increase in High-Frequency Oscillations Reflects Activation of Local Cortical Circuits.

A key new finding of this study is that despite the resistance of the global sleep homeostasis to follow cumulative REM sleep loss, CSR induces progressive alterations in cortical activity manifested by progressive increases in the level of high-frequency oscillations. These changes were absent on SR1 and only started at later days of CSR, indicating essential differences between the response of cortical activity to acute and chronic sleep deprivation. Indeed, acute sleep deprivation also elicited delta and theta rebounds with no significant change in beta2 and gamma EEG in the following all-day unlimited sleep opportunity (20), similar to the reaction on the first day (SR1) in this study.

High-frequency oscillations reflect synchronized activity of local circuits (32, 33) and focused gamma synchrony assisting local processing is specifically sharpened in REM sleep (3, 16, 34). Simultaneous EEG and multiunit recordings demonstrated that correlated firing within small cortical ensembles has the strongest relationship with beta2 and gamma amplitude at the nearest EEG electrode (35, 36). The frequency is determined by the architecture of the local network. Whereas beta2 rhythm can be maintained by pyramidal neurons in the absence of synaptic input (37), gamma activity involves parvalbumin-positive interneurons (38 ⇓⇓–41) which in turn require sufficient excitatory drive to generate oscillations. Clear separation of these rhythms was shown in vitro where recurrent glutamatergic activity generated gamma or beta2 oscillations in immediately adjacent cortical areas (42, 43) and in different layers of the same slice of somatosensory cortex (37). Although spatial resolution of surface EEG is limited, high-density EEG showed a clear fronto-caudal topography in CSR with sharp contrast between prefrontal gamma1 and somatosensory-motor beta2 and gamma2 activation (Figs. 2 and 3) which was further emphasized by area-specific low-frequency modulations (Fig. 3). Thus, CSR-enhanced activity correspond to typical EEG topography which normally reflects functional differences, i.e., beta involved in somatosensory perception (44) and motor maintenance (45) and gamma involved in various cognitive processes, including perception, attention, and memory. Distinction of gamma1 and gamma2 synchronization within the CA1 was also given functional significance in routing information flow from different sources (46) also providing distinct theta inputs to the same circuit (47).

Functional Implications: Activation of Local Cortical Circuits and REM Sleep-Related Synaptic Potentiation.

Sleep deprivation has been shown to impair memory consolidation in numerous human and rodent studies (3), including rat and mouse CSR paradigms (48, 49), suggesting that compensatory processes which may be initiated are certainly insufficient. We suggest however, that the results of the current study may open a window on the organization of cortical networks to serve diverse sleep functions. Normal memory processing can involve both REM and NREM sleep-dependent mechanisms (50) and the processes of sleep homeostasis seen after sleep loss may represent compensatory mechanisms to maintain sleep-dependent memory consolidation. Specifically, synaptic potentiation in REM sleep was proposed as part of a two-stage process of memory consolidation (7, 51) in which neuronal circuits reactivated in SWS (52 ⇓–54) are primed for long-term potentiation (LTP) which occurs in the ensuing REM sleep supported by enhanced plasticity-related gene expression (55, 56) and high-frequency “desynchronized” activity, characteristic for this state.

Although there is an extensive literature associating high-frequency cortical oscillations with cognitive function in waking, their functional role in REM sleep remains poorly understood. Beta is low in normal REM sleep but increases in animal models of motor diseases (57, 58). Gamma is high in REM sleep, and yet its possible involvement in neuronal assembly formation has little experimental evidence. The few studies reporting replay of waking discharge sequences in REM sleep (59, 60) did not investigate gamma synchrony. It should be emphasized that fast oscillations are determined by the population dynamics of the local circuits and signal a state of activation necessarily leading to engagement of inhibitory interneurons to balance excitation with inhibition (61), indicating “orthogonal to the formation of spike sequences, i.e., the discharge probability increases independent of whether a neuron is part of a spatiotemporal spike sequence or not”, as suggested for ripple busts (40, 52). For neurons participating in reverberating sequences and thus tagged in pre-REM ripple-spindle episodes (7, 51), gamma oscillation may promote temporal coactivation necessary to strengthen their synaptic weights. Low-frequency modulation, by theta in hippocampus or by 4-Hz rhythm in PFC spanning several gamma cycles, brings firing in these ensembles within the NMDA receptor time constant, necessary for LTP. Accordingly, hippocampal neurons active in novel places and reactivated during REM sleep are firing at the positive half of theta cycle, consistent with the theta phase suitable for LTP induction (62). We found in this study that enhanced gamma power in CSR was strongly modulated by theta with the phase matching this pattern implicating this mechanism in compensation for the losses of REM sleep-related synaptic potentiation.

Increased high-frequency neuronal activity is also necessary for plasticity-related gene expression, including CREB, Arc, Zif-268, and BDNF expression, which increases in REM sleep after learned experience (55, 56, 63 ⇓–65). Selective REM sleep deprivation up-regulated the levels of Zif268, Arc and BDNF during REM sleep rebound (66). In CSR, BDNF also increased in the cortex and hippocampus on SR1, immediately after 18 h sleep deprivation, but then decreased below baseline values on the following days (67). Interestingly, mRNA level of adenosine 2A receptor, critically involved in enhancing the effect of BDNF on hippocampal LTP (3, 68), progressively increased in the hippocampus after CSR reaching significance on SR5 (69).

Experimental Procedures

Animal studies were performed in compliance with animal protocols approved by the Institutional Animal Care and Use Committee of the Korea Institute of Science and Technology (KIST) (AP-2014L7002). Further details of materials and methods are provided in SI Experimental Procedures.

CSR.

As depicted in Fig. S2, sleep restriction was repeated for 5 d on daily basis. For sleep restriction, the animal was placed in a motorized habituated wheel (5.5 inch in diameter × 2.3 inch in width, Lafayette Instrument, Inc.), which was habituated before experiment. The wheel was programmed to rotate an eighth of a revolution for 4 s followed by 2 s intermission to prevent the mice from falling into sleep. The water and foods were accessible in the wheel. The sleep restriction lasted for 18 h starting from ZT6 ending at ZT24. At ZT0, the mouse was allowed to sleep until ZT6 in a habituated acrylic transparent chamber (7.8 inch in diameter; 9.8 inch in height). In the chamber, the mouse had ad libitum access to food and water. Throughout the recording section, the mouse was tethered to connector for continuous recording of screw EEG, and the high-density EEG was connected only during ZT0 to ZT3 corresponding to the first half of sleep opportunity.

Sleep Recording.

Nine mice were implanted with screw electrodes (10–50 kΩ) and microarray (100-300 kΩ) for standard EEG and high-density EEG, respectively. The monopolar screw electrodes record the frontal and parietal EEG with ground at occipital bone, and the microarray record 38 channels of cortical EEG on the exposed skull with reference at occipital bone and ground at olfactory bone. Fig. S1 shows the electrode montage of high-density EEG with exemplary signals from prefrontal, frontal, somatosensory, and visual cortices. The electrodes were secured with dental cement, and an accelerometer (weight, 15 mg; sensitivity, 1.5 G) based motion sensor was attached to the electrode connector with glue. The standard EEG and motion sensors were recorded for 8 d of study without cessation, whereas the high-density EEG was recorded for 3 h in each day at the start of ZT0.

High-Density EEG Analysis.

All of the offline analysis of high-density EEG was performed in custom-built programs developed in Matlab (Mathworks). For the high-density EEG spectral analysis, the frequency bands of interest were classified as follows: slow oscillation (< 1Hz), delta (1–4 Hz), theta1 (5–7 Hz), theta2 (8–10 Hz), beta1 (12–18 Hz), beta2 (21–27 Hz), gamma1 (34–46 Hz), and gamma2 (70–90 Hz) (70). The power spectral analysis was based on a fast FFT with frequency resolution of 0.122 Hz. The power of individual frequency of an epoch was obtained by averaging the FFT*conj(FFT) in the band of interest. The power ratio was calculated using (P_SD – P_BL)/P_BL, where P_BL and P_SD are the power in the BL and SR day epochs, respectively. The daily power ratio was obtained by averaging power ratios over epochs and then the statistical tests for any increase or decrease were performed in each EEG channel across animals.

Topographic maps of oscillations were calculated in the epoch and then averaged over epochs and animals. The maps were represented on a 3D template cortical model by cubic spline interpolation function. This model was built using the spm_surf function in SPM8 software (Wellcome Trust Centre for Neuroimaging, University College London) with the mouse magnetic resonance data from the open database of Magnetic Resonance Microimaging Neurological Atlas Group (71).

Possible CFC between slow and fast oscillations was studied using MI described by Tort et al. (72). The MI values for phase frequency in slow-frequency band (0.1–20 Hz) and amplitude frequency in fast-frequency band (20–100 Hz) were calculated and then visualized by pseudocolor map (comodulation plot; SI Experimental Procedures).

Detection of Phasic REM.

Phasic REM sleep was detected according to the criteria of Montgomery et al. (16). The criteria follow the observation on EEG during phasic REM, which are characterized by the faster and stronger theta rhythms compared with those during tonic REM. Peak-to-peak detection was applied to define the interpeak distance.

Testing Statistical Significance.

On any given REM epoch, EEG properties were computed at each channel and the significance was test in epochs from each mouse as well as epochs from all of the mice. Unless stated otherwise, data are presented as mean ± SEM, and a normality test was performed before applying variance analysis using one-sample Kolmogorov-Smirnov test. Statistical analysis was generally based on two-way repeated measures ANOVA or a nonparametric Kruskal–Wallis test. Paired t test was used to compare experimental days with BL. A P value < 0.05 was considered significant.

SI Experimental Procedures

Animals.

We raised all mice on a 12:12 light:dark cycle with ad libitum access to food and water. We used female F1 heterozygous hybrid (C57BL/6 × 129S4/SvJae) mice in compliance with animal protocols approved by the Institutional Animal Care and Use Committee of the Korea Institute of Science and Technology (AP-2014L7002).

Surgical Procedure of Microarray.

Animal handling and surgery were performed according to the guidelines of the Korean Animal and Plant Quarantine Agency (73), conforming to NIH Guide for the Care and Use of Laboratory Animals (74). The surgery tools and electrodes were sterilized before surgery. Mice were anesthetized with ketamine and xylazine mixture (120 and 6 mg/kg, respectively) by i.p. injection and positioned on a stereotaxic apparatus. If necessary, supplemental anesthetic (a third of the original dose) was given. After shaving the hair, an incision was made to expose the skull, and the skull surface was cleaned. Screws were implanted before microarray. For conventional EEG recording, three screw electrodes (stainless tapping screw, 0.8 mm × 4.8 mm, Nitto Seiko Co., Japan) were implanted: one above the frontal cortex (1.5 mm anterior and -2 mm lateral to bregma), one above the parietal area (2 mm posterior and 4 mm lateral), and one on the interparietal bone above cerebellum for reference and ground (6 mm posterior and ±2 mm lateral). Two additional screws were implanted on the skull for the purpose of firm cement. After screws, the high-density EEG microarray was placed on the skull to fit the center of fifth wing of the electrode array aligned to meet the bregma. After placement of microarray, the dental cement was carefully coated over the electrodes for fixation, and then antibiotic cream was applied around the incision. After surgery, the mice were individually housed and underwent a recovery period (>1 wk). The surgery procedure is shown in video format (75).

Nanofabrication and Characterization of Microarray.

Flexible microarrays for high-density EEG were fabricated in nanofabrication facility under the supervision of Dr Ho Kun Sung in Korea Advanced Nano Fab Center (www.kanc.re.kr/index.jsp). Briefly, the fabrication was processed on six inch silicon wafer according to the following steps: (i) aluminum coating as a sacrifice layer on the wafer, (ii) spin coating of the polyimide layer as base substrate, (iii) patterning the metal layers (Cr/Pt) for formation of electrodes, interconnection lines, and connection pads using a photolithography process, (iv) spin coating a second layer of polyimide with the same thickness of the first layer, (v) selective reactive ion etching of the polyimide layer to expose the electrical contacts and pads, and (vi) releasing the microarray by etching the aluminum layer. The fully processed microarrays were soldered to a board connector (DF 16 series, Hirose Electric Group) by using package service in a packaging company (Soojung Electrics Suwon).

After packaging the microarray, the electrodes in the microarray were electrochemically characterized by means of impedance spectroscopy (measurement amplitude; 50 mV, frequency 1–2 kHz). The measurements were performed at room temperature in physiological (0.9%) saline solution. The impedance decrease as frequency increases. For example, the impedance magnitude was 12 ± 1 kΩ at 1 kHz and 145 ± 5 kΩ at 100 Hz. Frequency response curves for impedance magnitude and phase are displayed in Fig. 1 D and E in Choi et al. (13). Any electrode with out-of-range magnitude was considered as defect and excluded in the analysis. After implantation of microarray, the electrode impedances were checked on the skull before recording high-density EEG (test frequency at 30 Hz) using Impedance menu in Neuroscan (Compumedics).

High-Density EEG.

High-density EEG signals were recorded with SynAmps2 amplifier (AC/DC amplifier, Neuroscan Inc.). Screw EEG signals were recorded with Grass analog amplifier (0.1–1,000 Hz, Grass Technologies) and digitized with an analog-digital converter (Digidata 1440A; Molecular Devices). The EEG signals were low- and high-pass filtered at 0.5 and 100 Hz, respectively, amplified, digitally converted, and sampled at 500 Hz. The motion sensor signals were digitally converted and sampled at 500 Hz. Whereas the screw EEG signals were recorded for 9 d without cessation, the high-density EEG signals were recorded only for three hours starting at ZT0 because of buffer overflow that frequently happened in long recording. Two amplifiers were synchronized by external trigger signal. Before start of experiment, the animals were habituated to the tethers in the recording wheel for a half hour in several days. Because of the long period of observation, during which the signal quality may differ significantly across days, the high-frequency components (130–250 Hz) were monitored each day. EEG is often notorious for movement artifact. Compared with human scalp EEG, mouse extracranial EEG has less affect from interface change. (Examples are shown in figure 3 of ref. 13.) Noise can be produced from the wires when the wires swing due to animal’s movement, but not in inactive period. All of the recording were performed in Faraday cage, which was ground to wall ground. Faraday cage, instruments, computer, and experimenter were electrically wired for equipotential condition. Any electrode showing significant change in the high-frequency components were excluded from the analysis. The center-to-center distance of microarray contact is ∼0.9 mm.

Sleep Scoring.

Sleep scoring was performed in 10 s epoch with the standard EEG and motion sensor signals by a trained experimenter using SleepSign software (Kissei Comtec America). Epochs containing movement artifacts were included in the state totals as wake periods but excluded from subsequent spectral analysis. Following assignment of state scores, sleep architecture were analyzed as a total time or mean period of each stage per day.

MI.

We calculated MI to investigate the relationship between slow and fast rhythms. MI calculates how much amplitude of fast waves is modulated by the phase of slow waves. Essentially, MI measures the scale of fast oscillation nested by slow oscillation with a normalized value from 0 to 1. We used comodulationroutine program developed and presented by Professor Adriano Tort in Federal University of Rio Grande do Norte. Rigorous derivation of MI is well described in supporting information of Tort et al. (72). Briefly, slow and fast rhythms were decomposed by zero-phase band-pass Butterworth filter with 1-Hz bandwidth, and designated as x_fp(t) and x_fA(t) for two frequencies of interest, f_p and f_A which refer to the phase frequency and amplitude frequency, respectively. Then, an instantaneous phase, ϕ_fp(t) of x_fp(t) was obtained by Hilbert transform. Likewise an instantaneous amplitude, A_fA(t) of x_fA(t) was calculated with the magnitude of Hilbert transform. Next, the phase of slow component was binned into 20° intervals and referred as ϕfpn for the nth bin. Then, the mean value of A_fA(ϕfpn) at each phase bin was calculated to obtain the amplitude distribution as a function of phase. The probability distribution of amplitude, p_n was obtained by normalized average of A_fA(ϕfpn), then applied to Shannon entropy according to H = Σ p_n log p_n. By definition, a uniform amplitude distribution, i.e., zero phase-amplitude coupling will lead H_max, and the stronger modulation will have the smaller value of H. MI is defined as MI = (H_max – H)/H_max. The MI as a function of (f_p, f_A) was mapped over frequency space and referred as comodulogram map. The frequency bins of 0.5 Hz and 5 Hz was used for phase and amplitude frequencies, respectively.

Acknowledgments

The authors thank Adriano Tort for his program of the modulation index and comodulogram; Ritchie Brown, Gyorgy Buzsáki, Valdyslav Vyazovskiy, and Robert Stickgold for discussions; and Mark Zielinski for critical reading of the manuscript. This work was supported by Korea National Research Council of Science and Technology Grant Convergence Research Center (CRC)-15-04-KIST; KIST Intramural Grant 2E26640; Department of Veterans Affairs Grants I01 BX000270, I01 BX002774, and I01 BX001356; and NIH Grants R01 MH100820, R01 MH039683, and P01 HL095491.

Footnotes

↵¹B. Kim and B. Kocsis contributed equally to this work.
↵²To whom correspondence may be addressed. Email: bernat_kocsis{at}hms.harvard.edu or jeechoi{at}kist.re.kr.

Author contributions: Y.K., R.E.S., R.W.M., and J.H.C. designed research; B. Kim, E.H., and J.H.C. performed research; B. Kim, B. Kocsis, Y.K., and J.H.C. analyzed data; B. Kocsis contributed new reagents/analytic tools; and B. Kocsis and J.H.C. wrote the paper. The microarrays are available by contacting J.H.C.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1615230114/-/DCSupplemental.

Freely available online through the PNAS open access option.

View Abstract

References

↵
1. Tononi G,
2. Cirelli C
(2006) Sleep function and synaptic homeostasis. Sleep Med Rev 10(1):49–62.
.
OpenUrl CrossRef PubMed
↵
1. Vyazovskiy VV,
2. Cirelli C,
3. Pfister-Genskow M,
4. Faraguna U,
5. Tononi G
(2008) Molecular and electrophysiological evidence for net synaptic potentiation in wake and depression in sleep. Nat Neurosci 11(2):200–208.
.
OpenUrl CrossRef PubMed
↵
1. Rasch B,
2. Born J
(2013) About sleep’s role in memory. Physiol Rev 93(2):681–766.
.
OpenUrl Abstract/FREE Full Text
↵
1. Borbély AA
(1982) A two process model of sleep regulation. Hum Neurobiol 1(3):195–204.
.
OpenUrl PubMed
↵
1. Buzsáki G
(1989) Two-stage model of memory trace formation: A role for “noisy” brain states. Neuroscience 31(3):551–570.
.
OpenUrl CrossRef PubMed
↵
1. Vyazovskiy VV, et al.
(2009) Cortical firing and sleep homeostasis. Neuron 63(6):865–878.
.
OpenUrl CrossRef PubMed
↵
1. Diekelmann S,
2. Born J
(2010) The memory function of sleep. Nat Rev Neurosci 11(2):114–126.
.
OpenUrl CrossRef PubMed
↵
1. Grosmark AD,
2. Mizuseki K,
3. Pastalkova E,
4. Diba K,
5. Buzsáki G
(2012) REM sleep reorganizes hippocampal excitability. Neuron 75(6):1001–1007.
.
OpenUrl CrossRef PubMed
↵
1. Shea JL, et al.
(2008) Rapid eye movement (REM) sleep homeostatic regulatory processes in the rat: Changes in the sleep-wake stages and electroencephalographic power spectra. Brain Res 1213:48–56.
.
OpenUrl CrossRef PubMed
↵
1. Wurts SW,
2. Edgar DM
(2000) Circadian and homeostatic control of rapid eye movement (REM) sleep: Promotion of REM tendency by the suprachiasmatic nucleus. J Neurosci 20(11):4300–4310.
.
OpenUrl Abstract/FREE Full Text
↵
1. Kim Y,
2. Laposky AD,
3. Bergmann BM,
4. Turek FW
(2007) Repeated sleep restriction in rats leads to homeostatic and allostatic responses during recovery sleep. Proc Natl Acad Sci USA 104(25):10697–10702.
.
OpenUrl Abstract/FREE Full Text
↵
1. Leemburg S, et al.
(2010) Sleep homeostasis in the rat is preserved during chronic sleep restriction. Proc Natl Acad Sci USA 107(36):15939–15944.
.
OpenUrl Abstract/FREE Full Text
↵
1. Choi JH,
2. Koch KP,
3. Poppendieck W,
4. Lee M,
5. Shin HS
(2010) High resolution electroencephalography in freely moving mice. J Neurophysiol 104(3):1825–1834.
.
OpenUrl Abstract/FREE Full Text
↵
1. Kim Y,
2. Chen L,
3. McCarley RW,
4. Strecker RE
(2013) Sleep allostasis in chronic sleep restriction: The role of the norepinephrine system. Brain Res 1531:9–16.
.
OpenUrl
↵
1. Buzsáki G
(2006) Rhythms of the Brain (Oxford Univ Press, New York), pp 334–356.
.
↵
1. Montgomery SM,
2. Sirota A,
3. Buzsáki G
(2008) Theta and gamma coordination of hippocampal networks during waking and rapid eye movement sleep. J Neurosci 28(26):6731–6741.
.
OpenUrl Abstract/FREE Full Text
↵
1. Fujisawa S,
2. Buzsáki G
(2011) A 4 Hz oscillation adaptively synchronizes prefrontal, VTA, and hippocampal activities. Neuron 72(1):153–165.
.
OpenUrl CrossRef PubMed
↵
1. Aru J, et al.
(2015) Untangling cross-frequency coupling in neuroscience. Curr Opin Neurobiol 31:51–61.
.
OpenUrl CrossRef PubMed
↵
1. de Souza L, et al.
(2010) Age-related changes during a paradigm of chronic sleep restriction. Behav Brain Res 214(2):201–205.
.
OpenUrl PubMed
↵
1. Stephenson R,
2. Caron AM,
3. Famina S
(2015) Behavioral sleep-wake homeostasis and EEG delta power are decoupled by chronic sleep restriction in the rat. Sleep 38(5):685–697.
.
OpenUrl
↵
1. Brown RE,
2. Basheer R,
3. McKenna JT,
4. Strecker RE,
5. McCarley RW
(2012) Control of sleep and wakefulness. Physiol Rev 92(3):1087–1187.
.
OpenUrl Abstract/FREE Full Text
↵
1. Datta S,
2. Maclean RR
(2007) Neurobiological mechanisms for the regulation of mammalian sleep-wake behavior: Reinterpretation of historical evidence and inclusion of contemporary cellular and molecular evidence. Neurosci Biobehav Rev 31(5):775–824.
.
OpenUrl CrossRef PubMed
↵
1. Saper CB,
2. Fuller PM,
3. Pedersen NP,
4. Lu J,
5. Scammell TE
(2010) Sleep state switching. Neuron 68(6):1023–1042.
.
OpenUrl CrossRef PubMed
↵
1. Steriade M
(2003) The corticothalamic system in sleep. Front Biosci 8:d878–d899.
.
OpenUrl CrossRef PubMed
↵
1. Vertes RP,
2. Kocsis B
(1997) Brainstem-diencephalo-septohippocampal systems controlling the theta rhythm of the hippocampus. Neuroscience 81(4):893–926.
.
OpenUrl CrossRef PubMed
↵
1. Rowe K, et al.
(1999) Heart rate surges during REM sleep are associated with theta rhythm and PGO activity in cats. Am J Physiol 277(3 Pt 2):R843–R849.
.
OpenUrl
↵
1. Kocsis B,
2. Li S,
3. Hajos M
(2007) Behavior-dependent modulation of hippocampal EEG activity by the selective norepinephrine reuptake inhibitor reboxetine in rats. Hippocampus 17(8):627–633.
.
OpenUrl CrossRef PubMed
↵
1. Tobler I,
2. Franken P,
3. Scherschlicht R
(1990) Sleep and EEG spectra in the rabbit under baseline conditions and following sleep deprivation. Physiol Behav 48(1):121–129.
.
OpenUrl CrossRef PubMed
↵
1. Hall TM,
2. de Carvalho F,
3. Jackson A
(2014) A common structure underlies low-frequency cortical dynamics in movement, sleep, and sedation. Neuron 83(5):1185–1199.
.
OpenUrl CrossRef PubMed
↵
1. Roy AT,
2. Svensson FP,
3. Mazeh A,
4. Kocsis B
Prefrontal-hippocampal coupling by theta rhythm and by 2-5 Hz oscillation in the delta band: The role of the nucleus reuniens of the thalamus. Brain Struct Funct, in press.
.
↵
1. Kocsis B
(2014) Unique oscillatory patterns in the rat prefrontal cortex after NMDA receptor blockade. Soc Neurosci Abstr 44:844.015.
.
OpenUrl
↵
1. Buzsáki G,
2. Logothetis N,
3. Singer W
(2013) Scaling brain size, keeping timing: Evolutionary preservation of brain rhythms. Neuron 80(3):751–764.
.
OpenUrl CrossRef PubMed
↵
1. Fries P,
2. Nikolić D,
3. Singer W
(2007) The gamma cycle. Trends Neurosci 30(7):309–316.
.
OpenUrl CrossRef PubMed
↵
1. Cantero JL, et al.
(2003) Sleep-dependent theta oscillations in the human hippocampus and neocortex. J Neurosci 23(34):10897–10903.
.
OpenUrl Abstract/FREE Full Text
↵
1. Snyder AC,
2. Morais MJ,
3. Willis CM,
4. Smith MA
(2015) Global network influences on local functional connectivity. Nat Neurosci 18(5):736–743.
.
OpenUrl CrossRef PubMed
↵
1. Whittingstall K,
2. Logothetis NK
(2009) Frequency-band coupling in surface EEG reflects spiking activity in monkey visual cortex. Neuron 64(2):281–289.
.
OpenUrl CrossRef PubMed
↵
1. Roopun AK, et al.
(2006) A beta2-frequency (20-30 Hz) oscillation in nonsynaptic networks of somatosensory cortex. Proc Natl Acad Sci USA 103(42):15646–15650.
.
OpenUrl Abstract/FREE Full Text
↵
1. Cardin JA, et al.
(2009) Driving fast-spiking cells induces gamma rhythm and controls sensory responses. Nature 459(7247):663–667.
.
OpenUrl CrossRef PubMed
↵
1. Sohal VS,
2. Zhang F,
3. Yizhar O,
4. Deisseroth K
(2009) Parvalbumin neurons and gamma rhythms enhance cortical circuit performance. Nature 459(7247):698–702.
.
OpenUrl CrossRef PubMed
↵
1. Csicsvari J,
2. Hirase H,
3. Czurkó A,
4. Mamiya A,
5. Buzsáki G
(1999) Oscillatory coupling of hippocampal pyramidal cells and interneurons in the behaving Rat. J Neurosci 19(1):274–287.
.
OpenUrl Abstract/FREE Full Text
↵
1. Buzsáki G,
2. Horváth Z,
3. Urioste R,
4. Hetke J,
5. Wise K
(1992) High-frequency network oscillation in the hippocampus. Science 256(5059):1025–1027.
.
OpenUrl Abstract/FREE Full Text
↵
1. Cunningham MO, et al.
(2006) Neuronal metabolism governs cortical network response state. Proc Natl Acad Sci USA 103(14):5597–5601.
.
OpenUrl Abstract/FREE Full Text
↵
1. Traub RD, et al.
(2005) Single-column thalamocortical network model exhibiting gamma oscillations, sleep spindles, and epileptogenic bursts. J Neurophysiol 93(4):2194–2232.
.
OpenUrl Abstract/FREE Full Text
↵
1. Lalo E, et al.
(2007) Phasic increases in cortical beta activity are associated with alterations in sensory processing in the human. Exp Brain Res 177(1):137–145.
.
OpenUrl CrossRef PubMed
↵
1. Brovelli A, et al.
(2004) Beta oscillations in a large-scale sensorimotor cortical network: Directional influences revealed by Granger causality. Proc Natl Acad Sci USA 101(26):9849–9854.
.
OpenUrl Abstract/FREE Full Text
↵
1. Colgin LL, et al.
(2009) Frequency of gamma oscillations routes flow of information in the hippocampus. Nature 462(7271):353–357.
.
OpenUrl CrossRef PubMed
↵
1. Kocsis B,
2. Bragin A,
3. Buzsáki G
(1999) Interdependence of multiple theta generators in the hippocampus: A partial coherence analysis. J Neurosci 19(14):6200–6212.
.
OpenUrl Abstract/FREE Full Text
↵
1. McCoy JG, et al.
(2013) Chronic sleep restriction impairs spatial memory in rats. Neuroreport 24(2):91–95.
.
OpenUrl
↵
1. Zielinski MR,
2. Davis JM,
3. Fadel JR,
4. Youngstedt SD
(2013) Influence of chronic moderate sleep restriction and exercise training on anxiety, spatial memory, and associated neurobiological measures in mice. Behav Brain Res 250:74–80.
.
OpenUrl CrossRef PubMed
↵
1. Watson BO,
2. Levenstein D,
3. Greene JP,
4. Gelinas JN,
5. Buzsáki G
(2016) Network homeostasis and state dynamics of neocortical sleep. Neuron 90(4):839–852.
.
OpenUrl
↵
1. Ribeiro S,
2. Nicolelis MA
(2004) Reverberation, storage, and postsynaptic propagation of memories during sleep. Learn Mem 11(6):686–696.
.
OpenUrl Abstract/FREE Full Text
↵
1. Nádasdy Z,
2. Hirase H,
3. Czurkó A,
4. Csicsvari J,
5. Buzsáki G
(1999) Replay and time compression of recurring spike sequences in the hippocampus. J Neurosci 19(21):9497–9507.
.
OpenUrl Abstract/FREE Full Text
↵
1. Skaggs WE,
2. McNaughton BL
(1996) Replay of neuronal firing sequences in rat hippocampus during sleep following spatial experience. Science 271(5257):1870–1873.
.
OpenUrl Abstract/FREE Full Text
↵
1. Wilson MA,
2. McNaughton BL
(1994) Reactivation of hippocampal ensemble memories during sleep. Science 265(5172):676–679.
.
OpenUrl Abstract/FREE Full Text
↵
1. Ulloor J,
2. Datta S
(2005) Spatio-temporal activation of cyclic AMP response element-binding protein, activity-regulated cytoskeletal-associated protein and brain-derived nerve growth factor: A mechanism for pontine-wave generator activation-dependent two-way active-avoidance memory processing in the rat. J Neurochem 95(2):418–428.
.
OpenUrl CrossRef PubMed
↵
1. Ribeiro S,
2. Goyal V,
3. Mello CV,
4. Pavlides C
(1999) Brain gene expression during REM sleep depends on prior waking experience. Learn Mem 6(5):500–508.
.
OpenUrl Abstract/FREE Full Text
↵
1. Jeantet Y,
2. Cayzac S,
3. Cho YH
(2013) β oscillation during slow wave sleep and rapid eye movement sleep in the electroencephalogram of a transgenic mouse model of Huntington’s disease. PLoS One 8(11):e79509.
.
OpenUrl
↵
1. Degos B,
2. Deniau JM,
3. Chavez M,
4. Maurice N
(2009) Chronic but not acute dopaminergic transmission interruption promotes a progressive increase in cortical beta frequency synchronization: Relationships to vigilance state and akinesia. Cereb Cortex 19(7):1616–1630.
.
OpenUrl Abstract/FREE Full Text
↵
1. Louie K,
2. Wilson MA
(2001) Temporally structured replay of awake hippocampal ensemble activity during rapid eye movement sleep. Neuron 29(1):145–156.
.
OpenUrl CrossRef PubMed
↵
1. Poe GR,
2. Nitz DA,
3. McNaughton BL,
4. Barnes CA
(2000) Experience-dependent phase-reversal of hippocampal neuron firing during REM sleep. Brain Res 855(1):176–180.
.
OpenUrl CrossRef PubMed
↵
1. Merker B
(2013) Cortical gamma oscillations: The functional key is activation, not cognition. Neurosci Biobehav Rev 37(3):401–417.
.
OpenUrl CrossRef PubMed
↵
1. Poe GR,
2. Walsh CM,
3. Bjorness TE
(2010) Cognitive neuroscience of sleep. Prog Brain Res 185:1–19.
.
OpenUrl CrossRef PubMed
↵
1. Datta S,
2. Li G,
3. Auerbach S
(2008) Activation of phasic pontine-wave generator in the rat: A mechanism for expression of plasticity-related genes and proteins in the dorsal hippocampus and amygdala. Eur J Neurosci 27(7):1876–1892.
.
OpenUrl CrossRef PubMed
↵
1. Ribeiro S, et al.
(2007) Novel experience induces persistent sleep-dependent plasticity in the cortex but not in the hippocampus. Front Neurosci 1(1):43–55.
.
OpenUrl CrossRef PubMed
↵
1. Calais JB, et al.
(2013) Long-term decrease in immediate early gene expression after electroconvulsive seizures. J Neural Transm (Vienna) 120(2):259–266.
.
OpenUrl
↵
1. Ravassard P,
2. Hamieh AM,
3. Malleret G,
4. Salin PA
(2015) Paradoxical sleep: A vigilance state to gate long-term brain plasticity? Neurobiol Learn Mem 122:4–10.
.
OpenUrl
↵
1. Zielinski MR, et al.
(2014) Chronic sleep restriction elevates brain interleukin-1 beta and tumor necrosis factor-alpha and attenuates brain-derived neurotrophic factor expression. Neurosci Lett 580:27–31.
.
OpenUrl CrossRef PubMed
↵
1. Fontinha BM,
2. Diógenes MJ,
3. Ribeiro JA,
4. Sebastião AM
(2008) Enhancement of long-term potentiation by brain-derived neurotrophic factor requires adenosine A2A receptor activation by endogenous adenosine. Neuropharmacology 54(6):924–933.
.
OpenUrl CrossRef PubMed
↵
1. Kim Y, et al.
(2012) Decoupling of sleepiness from sleep time and intensity during chronic sleep restriction: Evidence for a role of the adenosine system. Sleep 35(6):861–869.
.
OpenUrl
↵
1. Pittman-Polletta BR,
2. Kocsis B,
3. Vijayan S,
4. Whittington MA,
5. Kopell NJ
(2015) Brain rhythms connect impaired inhibition to altered cognition in schizophrenia. Biol Psychiatry 77(12):1020–1030.
.
OpenUrl CrossRef PubMed
↵
1. Ma Y, et al.
(2008) In vivo 3D digital atlas database of the adult C57BL/6J mouse brain by magnetic resonance microscopy. Front Neuroanat 2:1.
.
OpenUrl PubMed
↵
1. Tort AB, et al.
(2008) Dynamic cross-frequency couplings of local field potential oscillations in rat striatum and hippocampus during performance of a T-maze task. Proc Natl Acad Sci USA 105(51):20517–20522.
.
OpenUrl Abstract/FREE Full Text
↵
Korean Animal and Plant Quarantine Agency (2014) Animal Protection Act, Act No. 12512, partial amendment 2014. Available at elaw.klri.re.kr/kor_mobile/viewer.do?hseq=32052&type=sogan&key=7. Accessed March 24, 2014.
.
↵
Committee on Care and Use of Laboratory Animals (1985) Guide for the Care and Use of Laboratory Animals (Natl Inst Health, Bethesda), DHHS Publ No (NIH) 85-23.
.
↵
1. Lee M,
2. Kim D,
3. Shin HS,
4. Sung HG,
5. Choi JH
(2011) High-density EEG recordings of the freely moving mice using polyimide-based microelectrode. J Vis Exp (47):e2562.
.

Article Alerts

Email Article

Citation Tools

Request Permissions

Proceedings of the National Academy of Sciences: 118 (3)

Current Issue

Submit

[1] ↵
Tononi G,
Cirelli C
(2006) Sleep function and synaptic homeostasis. Sleep Med Rev 10(1):49–62.
.
OpenUrl CrossRef PubMed

[2] Tononi G,

[3] Cirelli C

[4] ↵
Vyazovskiy VV,
Cirelli C,
Pfister-Genskow M,
Faraguna U,
Tononi G
(2008) Molecular and electrophysiological evidence for net synaptic potentiation in wake and depression in sleep. Nat Neurosci 11(2):200–208.
.
OpenUrl CrossRef PubMed

[5] Vyazovskiy VV,

[6] Cirelli C,

[7] Pfister-Genskow M,

[8] Faraguna U,

[9] Tononi G

[10] ↵
Rasch B,
Born J
(2013) About sleep’s role in memory. Physiol Rev 93(2):681–766.
.
OpenUrl Abstract/FREE Full Text

[11] Rasch B,

[12] Born J

[13] ↵
Borbély AA
(1982) A two process model of sleep regulation. Hum Neurobiol 1(3):195–204.
.
OpenUrl PubMed

[14] Borbély AA

[15] ↵
Buzsáki G
(1989) Two-stage model of memory trace formation: A role for “noisy” brain states. Neuroscience 31(3):551–570.
.
OpenUrl CrossRef PubMed

[16] Buzsáki G

[17] ↵
Vyazovskiy VV, et al.
(2009) Cortical firing and sleep homeostasis. Neuron 63(6):865–878.
.
OpenUrl CrossRef PubMed

[18] Vyazovskiy VV, et al.

[19] ↵
Diekelmann S,
Born J
(2010) The memory function of sleep. Nat Rev Neurosci 11(2):114–126.
.
OpenUrl CrossRef PubMed

[20] Diekelmann S,

[21] Born J

[22] ↵
Grosmark AD,
Mizuseki K,
Pastalkova E,
Diba K,
Buzsáki G
(2012) REM sleep reorganizes hippocampal excitability. Neuron 75(6):1001–1007.
.
OpenUrl CrossRef PubMed

[23] Grosmark AD,

[24] Mizuseki K,

[25] Pastalkova E,

[26] Diba K,

[27] Buzsáki G

[28] ↵
Shea JL, et al.
(2008) Rapid eye movement (REM) sleep homeostatic regulatory processes in the rat: Changes in the sleep-wake stages and electroencephalographic power spectra. Brain Res 1213:48–56.
.
OpenUrl CrossRef PubMed

[29] Shea JL, et al.

[30] ↵
Wurts SW,
Edgar DM
(2000) Circadian and homeostatic control of rapid eye movement (REM) sleep: Promotion of REM tendency by the suprachiasmatic nucleus. J Neurosci 20(11):4300–4310.
.
OpenUrl Abstract/FREE Full Text

[31] Wurts SW,

[32] Edgar DM

[33] ↵
Kim Y,
Laposky AD,
Bergmann BM,
Turek FW
(2007) Repeated sleep restriction in rats leads to homeostatic and allostatic responses during recovery sleep. Proc Natl Acad Sci USA 104(25):10697–10702.
.
OpenUrl Abstract/FREE Full Text

[34] Kim Y,

[35] Laposky AD,

[36] Bergmann BM,

[37] Turek FW

[38] ↵
Leemburg S, et al.
(2010) Sleep homeostasis in the rat is preserved during chronic sleep restriction. Proc Natl Acad Sci USA 107(36):15939–15944.
.
OpenUrl Abstract/FREE Full Text

[39] Leemburg S, et al.

[40] ↵
Choi JH,
Koch KP,
Poppendieck W,
Lee M,
Shin HS
(2010) High resolution electroencephalography in freely moving mice. J Neurophysiol 104(3):1825–1834.
.
OpenUrl Abstract/FREE Full Text

[41] Choi JH,

[42] Koch KP,

[43] Poppendieck W,

[44] Lee M,

[45] Shin HS

[46] ↵
Kim Y,
Chen L,
McCarley RW,
Strecker RE
(2013) Sleep allostasis in chronic sleep restriction: The role of the norepinephrine system. Brain Res 1531:9–16.
.
OpenUrl

[47] Kim Y,

[48] Chen L,

[49] McCarley RW,

[50] Strecker RE

[51] ↵
Buzsáki G
(2006) Rhythms of the Brain (Oxford Univ Press, New York), pp 334–356.
.

[52] Buzsáki G

[53] ↵
Montgomery SM,
Sirota A,
Buzsáki G
(2008) Theta and gamma coordination of hippocampal networks during waking and rapid eye movement sleep. J Neurosci 28(26):6731–6741.
.
OpenUrl Abstract/FREE Full Text

[54] Montgomery SM,

[55] Sirota A,

[56] Buzsáki G

[57] ↵
Fujisawa S,
Buzsáki G
(2011) A 4 Hz oscillation adaptively synchronizes prefrontal, VTA, and hippocampal activities. Neuron 72(1):153–165.
.
OpenUrl CrossRef PubMed

[58] Fujisawa S,

[59] Buzsáki G

[60] ↵
Aru J, et al.
(2015) Untangling cross-frequency coupling in neuroscience. Curr Opin Neurobiol 31:51–61.
.
OpenUrl CrossRef PubMed

[61] Aru J, et al.

[62] ↵
de Souza L, et al.
(2010) Age-related changes during a paradigm of chronic sleep restriction. Behav Brain Res 214(2):201–205.
.
OpenUrl PubMed

[63] de Souza L, et al.

[64] ↵
Stephenson R,
Caron AM,
Famina S
(2015) Behavioral sleep-wake homeostasis and EEG delta power are decoupled by chronic sleep restriction in the rat. Sleep 38(5):685–697.
.
OpenUrl

[65] Stephenson R,

[66] Caron AM,

[67] Famina S

[68] ↵
Brown RE,
Basheer R,
McKenna JT,
Strecker RE,
McCarley RW
(2012) Control of sleep and wakefulness. Physiol Rev 92(3):1087–1187.
.
OpenUrl Abstract/FREE Full Text

[69] Brown RE,

[70] Basheer R,

[71] McKenna JT,

[72] Strecker RE,

[73] McCarley RW

[74] ↵
Datta S,
Maclean RR
(2007) Neurobiological mechanisms for the regulation of mammalian sleep-wake behavior: Reinterpretation of historical evidence and inclusion of contemporary cellular and molecular evidence. Neurosci Biobehav Rev 31(5):775–824.
.
OpenUrl CrossRef PubMed

[75] Datta S,

[76] Maclean RR

[77] ↵
Saper CB,
Fuller PM,
Pedersen NP,
Lu J,
Scammell TE
(2010) Sleep state switching. Neuron 68(6):1023–1042.
.
OpenUrl CrossRef PubMed

[78] Saper CB,

[79] Fuller PM,

[80] Pedersen NP,

[81] Lu J,

[82] Scammell TE

[83] ↵
Steriade M
(2003) The corticothalamic system in sleep. Front Biosci 8:d878–d899.
.
OpenUrl CrossRef PubMed

[84] Steriade M

[85] ↵
Vertes RP,
Kocsis B
(1997) Brainstem-diencephalo-septohippocampal systems controlling the theta rhythm of the hippocampus. Neuroscience 81(4):893–926.
.
OpenUrl CrossRef PubMed

[86] Vertes RP,

[87] Kocsis B

[88] ↵
Rowe K, et al.
(1999) Heart rate surges during REM sleep are associated with theta rhythm and PGO activity in cats. Am J Physiol 277(3 Pt 2):R843–R849.
.
OpenUrl

[89] Rowe K, et al.

[90] ↵
Kocsis B,
Li S,
Hajos M
(2007) Behavior-dependent modulation of hippocampal EEG activity by the selective norepinephrine reuptake inhibitor reboxetine in rats. Hippocampus 17(8):627–633.
.
OpenUrl CrossRef PubMed

[91] Kocsis B,

[92] Li S,

[93] Hajos M

[94] ↵
Tobler I,
Franken P,
Scherschlicht R
(1990) Sleep and EEG spectra in the rabbit under baseline conditions and following sleep deprivation. Physiol Behav 48(1):121–129.
.
OpenUrl CrossRef PubMed

[95] Tobler I,

[96] Franken P,

[97] Scherschlicht R

[98] ↵
Hall TM,
de Carvalho F,
Jackson A
(2014) A common structure underlies low-frequency cortical dynamics in movement, sleep, and sedation. Neuron 83(5):1185–1199.
.
OpenUrl CrossRef PubMed

[99] Hall TM,

[100] de Carvalho F,

[101] Jackson A

[102] ↵
Roy AT,
Svensson FP,
Mazeh A,
Kocsis B
Prefrontal-hippocampal coupling by theta rhythm and by 2-5 Hz oscillation in the delta band: The role of the nucleus reuniens of the thalamus. Brain Struct Funct, in press.
.

[103] Roy AT,

[104] Svensson FP,

[105] Mazeh A,

[106] Kocsis B

[107] ↵
Kocsis B
(2014) Unique oscillatory patterns in the rat prefrontal cortex after NMDA receptor blockade. Soc Neurosci Abstr 44:844.015.
.
OpenUrl

[108] Kocsis B

[109] ↵
Buzsáki G,
Logothetis N,
Singer W
(2013) Scaling brain size, keeping timing: Evolutionary preservation of brain rhythms. Neuron 80(3):751–764.
.
OpenUrl CrossRef PubMed

[110] Buzsáki G,

[111] Logothetis N,

[112] Singer W

[113] ↵
Fries P,
Nikolić D,
Singer W
(2007) The gamma cycle. Trends Neurosci 30(7):309–316.
.
OpenUrl CrossRef PubMed

[114] Fries P,

[115] Nikolić D,

[116] Singer W

[117] ↵
Cantero JL, et al.
(2003) Sleep-dependent theta oscillations in the human hippocampus and neocortex. J Neurosci 23(34):10897–10903.
.
OpenUrl Abstract/FREE Full Text

[118] Cantero JL, et al.

[119] ↵
Snyder AC,
Morais MJ,
Willis CM,
Smith MA
(2015) Global network influences on local functional connectivity. Nat Neurosci 18(5):736–743.
.
OpenUrl CrossRef PubMed

[120] Snyder AC,

[121] Morais MJ,

[122] Willis CM,

[123] Smith MA

[124] ↵
Whittingstall K,
Logothetis NK
(2009) Frequency-band coupling in surface EEG reflects spiking activity in monkey visual cortex. Neuron 64(2):281–289.
.
OpenUrl CrossRef PubMed

[125] Whittingstall K,

[126] Logothetis NK

[127] ↵
Roopun AK, et al.
(2006) A beta2-frequency (20-30 Hz) oscillation in nonsynaptic networks of somatosensory cortex. Proc Natl Acad Sci USA 103(42):15646–15650.
.
OpenUrl Abstract/FREE Full Text

[128] Roopun AK, et al.

[129] ↵
Cardin JA, et al.
(2009) Driving fast-spiking cells induces gamma rhythm and controls sensory responses. Nature 459(7247):663–667.
.
OpenUrl CrossRef PubMed

[130] Cardin JA, et al.

[131] ↵
Sohal VS,
Zhang F,
Yizhar O,
Deisseroth K
(2009) Parvalbumin neurons and gamma rhythms enhance cortical circuit performance. Nature 459(7247):698–702.
.
OpenUrl CrossRef PubMed

[132] Sohal VS,

[133] Zhang F,

[134] Yizhar O,

[135] Deisseroth K

[136] ↵
Csicsvari J,
Hirase H,
Czurkó A,
Mamiya A,
Buzsáki G
(1999) Oscillatory coupling of hippocampal pyramidal cells and interneurons in the behaving Rat. J Neurosci 19(1):274–287.
.
OpenUrl Abstract/FREE Full Text

[137] Csicsvari J,

[138] Hirase H,

[139] Czurkó A,

[140] Mamiya A,

[141] Buzsáki G

[142] ↵
Buzsáki G,
Horváth Z,
Urioste R,
Hetke J,
Wise K
(1992) High-frequency network oscillation in the hippocampus. Science 256(5059):1025–1027.
.
OpenUrl Abstract/FREE Full Text

[143] Buzsáki G,

[144] Horváth Z,

[145] Urioste R,

[146] Hetke J,

[147] Wise K

[148] ↵
Cunningham MO, et al.
(2006) Neuronal metabolism governs cortical network response state. Proc Natl Acad Sci USA 103(14):5597–5601.
.
OpenUrl Abstract/FREE Full Text

[149] Cunningham MO, et al.

[150] ↵
Traub RD, et al.
(2005) Single-column thalamocortical network model exhibiting gamma oscillations, sleep spindles, and epileptogenic bursts. J Neurophysiol 93(4):2194–2232.
.
OpenUrl Abstract/FREE Full Text

[151] Traub RD, et al.

[152] ↵
Lalo E, et al.
(2007) Phasic increases in cortical beta activity are associated with alterations in sensory processing in the human. Exp Brain Res 177(1):137–145.
.
OpenUrl CrossRef PubMed

[153] Lalo E, et al.

[154] ↵
Brovelli A, et al.
(2004) Beta oscillations in a large-scale sensorimotor cortical network: Directional influences revealed by Granger causality. Proc Natl Acad Sci USA 101(26):9849–9854.
.
OpenUrl Abstract/FREE Full Text

[155] Brovelli A, et al.

[156] ↵
Colgin LL, et al.
(2009) Frequency of gamma oscillations routes flow of information in the hippocampus. Nature 462(7271):353–357.
.
OpenUrl CrossRef PubMed

[157] Colgin LL, et al.

[158] ↵
Kocsis B,
Bragin A,
Buzsáki G
(1999) Interdependence of multiple theta generators in the hippocampus: A partial coherence analysis. J Neurosci 19(14):6200–6212.
.
OpenUrl Abstract/FREE Full Text

[159] Kocsis B,

[160] Bragin A,

[161] Buzsáki G

[162] ↵
McCoy JG, et al.
(2013) Chronic sleep restriction impairs spatial memory in rats. Neuroreport 24(2):91–95.
.
OpenUrl

[163] McCoy JG, et al.

[164] ↵
Zielinski MR,
Davis JM,
Fadel JR,
Youngstedt SD
(2013) Influence of chronic moderate sleep restriction and exercise training on anxiety, spatial memory, and associated neurobiological measures in mice. Behav Brain Res 250:74–80.
.
OpenUrl CrossRef PubMed

[165] Zielinski MR,

[166] Davis JM,

[167] Fadel JR,

[168] Youngstedt SD

[169] ↵
Watson BO,
Levenstein D,
Greene JP,
Gelinas JN,
Buzsáki G
(2016) Network homeostasis and state dynamics of neocortical sleep. Neuron 90(4):839–852.
.
OpenUrl

[170] Watson BO,

[171] Levenstein D,

[172] Greene JP,

[173] Gelinas JN,

[174] Buzsáki G

[175] ↵
Ribeiro S,
Nicolelis MA
(2004) Reverberation, storage, and postsynaptic propagation of memories during sleep. Learn Mem 11(6):686–696.
.
OpenUrl Abstract/FREE Full Text

[176] Ribeiro S,

[177] Nicolelis MA

[178] ↵
Nádasdy Z,
Hirase H,
Czurkó A,
Csicsvari J,
Buzsáki G
(1999) Replay and time compression of recurring spike sequences in the hippocampus. J Neurosci 19(21):9497–9507.
.
OpenUrl Abstract/FREE Full Text

[179] Nádasdy Z,

[180] Hirase H,

[181] Czurkó A,

[182] Csicsvari J,

[183] Buzsáki G

[184] ↵
Skaggs WE,
McNaughton BL
(1996) Replay of neuronal firing sequences in rat hippocampus during sleep following spatial experience. Science 271(5257):1870–1873.
.
OpenUrl Abstract/FREE Full Text

[185] Skaggs WE,

[186] McNaughton BL

[187] ↵
Wilson MA,
McNaughton BL
(1994) Reactivation of hippocampal ensemble memories during sleep. Science 265(5172):676–679.
.
OpenUrl Abstract/FREE Full Text

[188] Wilson MA,

[189] McNaughton BL

[190] ↵
Ulloor J,
Datta S
(2005) Spatio-temporal activation of cyclic AMP response element-binding protein, activity-regulated cytoskeletal-associated protein and brain-derived nerve growth factor: A mechanism for pontine-wave generator activation-dependent two-way active-avoidance memory processing in the rat. J Neurochem 95(2):418–428.
.
OpenUrl CrossRef PubMed

[191] Ulloor J,

[192] Datta S

[193] ↵
Ribeiro S,
Goyal V,
Mello CV,
Pavlides C
(1999) Brain gene expression during REM sleep depends on prior waking experience. Learn Mem 6(5):500–508.
.
OpenUrl Abstract/FREE Full Text

[194] Ribeiro S,

[195] Goyal V,

[196] Mello CV,

[197] Pavlides C

[198] ↵
Jeantet Y,
Cayzac S,
Cho YH
(2013) β oscillation during slow wave sleep and rapid eye movement sleep in the electroencephalogram of a transgenic mouse model of Huntington’s disease. PLoS One 8(11):e79509.
.
OpenUrl

[199] Jeantet Y,

[200] Cayzac S,

[201] Cho YH

[202] ↵
Degos B,
Deniau JM,
Chavez M,
Maurice N
(2009) Chronic but not acute dopaminergic transmission interruption promotes a progressive increase in cortical beta frequency synchronization: Relationships to vigilance state and akinesia. Cereb Cortex 19(7):1616–1630.
.
OpenUrl Abstract/FREE Full Text

[203] Degos B,

[204] Deniau JM,

[205] Chavez M,

[206] Maurice N

[207] ↵
Louie K,
Wilson MA
(2001) Temporally structured replay of awake hippocampal ensemble activity during rapid eye movement sleep. Neuron 29(1):145–156.
.
OpenUrl CrossRef PubMed

[208] Louie K,

[209] Wilson MA

[210] ↵
Poe GR,
Nitz DA,
McNaughton BL,
Barnes CA
(2000) Experience-dependent phase-reversal of hippocampal neuron firing during REM sleep. Brain Res 855(1):176–180.
.
OpenUrl CrossRef PubMed

[211] Poe GR,

[212] Nitz DA,

[213] McNaughton BL,

[214] Barnes CA

[215] ↵
Merker B
(2013) Cortical gamma oscillations: The functional key is activation, not cognition. Neurosci Biobehav Rev 37(3):401–417.
.
OpenUrl CrossRef PubMed

[216] Merker B

[217] ↵
Poe GR,
Walsh CM,
Bjorness TE
(2010) Cognitive neuroscience of sleep. Prog Brain Res 185:1–19.
.
OpenUrl CrossRef PubMed

[218] Poe GR,

[219] Walsh CM,

[220] Bjorness TE

[221] ↵
Datta S,
Li G,
Auerbach S
(2008) Activation of phasic pontine-wave generator in the rat: A mechanism for expression of plasticity-related genes and proteins in the dorsal hippocampus and amygdala. Eur J Neurosci 27(7):1876–1892.
.
OpenUrl CrossRef PubMed

[222] Datta S,

[223] Li G,

[224] Auerbach S

[225] ↵
Ribeiro S, et al.
(2007) Novel experience induces persistent sleep-dependent plasticity in the cortex but not in the hippocampus. Front Neurosci 1(1):43–55.
.
OpenUrl CrossRef PubMed

[226] Ribeiro S, et al.

[227] ↵
Calais JB, et al.
(2013) Long-term decrease in immediate early gene expression after electroconvulsive seizures. J Neural Transm (Vienna) 120(2):259–266.
.
OpenUrl

[228] Calais JB, et al.

[229] ↵
Ravassard P,
Hamieh AM,
Malleret G,
Salin PA
(2015) Paradoxical sleep: A vigilance state to gate long-term brain plasticity? Neurobiol Learn Mem 122:4–10.
.
OpenUrl

[230] Ravassard P,

[231] Hamieh AM,

[232] Malleret G,

[233] Salin PA

[234] ↵
Zielinski MR, et al.
(2014) Chronic sleep restriction elevates brain interleukin-1 beta and tumor necrosis factor-alpha and attenuates brain-derived neurotrophic factor expression. Neurosci Lett 580:27–31.
.
OpenUrl CrossRef PubMed

[235] Zielinski MR, et al.

[236] ↵
Fontinha BM,
Diógenes MJ,
Ribeiro JA,
Sebastião AM
(2008) Enhancement of long-term potentiation by brain-derived neurotrophic factor requires adenosine A2A receptor activation by endogenous adenosine. Neuropharmacology 54(6):924–933.
.
OpenUrl CrossRef PubMed

[237] Fontinha BM,

[238] Diógenes MJ,

[239] Ribeiro JA,

[240] Sebastião AM

[241] ↵
Kim Y, et al.
(2012) Decoupling of sleepiness from sleep time and intensity during chronic sleep restriction: Evidence for a role of the adenosine system. Sleep 35(6):861–869.
.
OpenUrl

[242] Kim Y, et al.

[243] ↵
Pittman-Polletta BR,
Kocsis B,
Vijayan S,
Whittington MA,
Kopell NJ
(2015) Brain rhythms connect impaired inhibition to altered cognition in schizophrenia. Biol Psychiatry 77(12):1020–1030.
.
OpenUrl CrossRef PubMed

[244] Pittman-Polletta BR,

[245] Kocsis B,

[246] Vijayan S,

[247] Whittington MA,

[248] Kopell NJ

[249] ↵
Ma Y, et al.
(2008) In vivo 3D digital atlas database of the adult C57BL/6J mouse brain by magnetic resonance microscopy. Front Neuroanat 2:1.
.
OpenUrl PubMed

[250] Ma Y, et al.

[251] ↵
Tort AB, et al.
(2008) Dynamic cross-frequency couplings of local field potential oscillations in rat striatum and hippocampus during performance of a T-maze task. Proc Natl Acad Sci USA 105(51):20517–20522.
.
OpenUrl Abstract/FREE Full Text

[252] Tort AB, et al.

[253] ↵
Korean Animal and Plant Quarantine Agency (2014) Animal Protection Act, Act No. 12512, partial amendment 2014. Available at elaw.klri.re.kr/kor_mobile/viewer.do?hseq=32052&type=sogan&key=7. Accessed March 24, 2014.
.

[254] ↵
Committee on Care and Use of Laboratory Animals (1985) Guide for the Care and Use of Laboratory Animals (Natl Inst Health, Bethesda), DHHS Publ No (NIH) 85-23.
.

[255] ↵
Lee M,
Kim D,
Shin HS,
Sung HG,
Choi JH
(2011) High-density EEG recordings of the freely moving mice using polyimide-based microelectrode. J Vis Exp (47):e2562.
.

[256] Lee M,

[257] Kim D,

[258] Shin HS,

[259] Sung HG,

[260] Choi JH

Main menu

User menu

Search

New Research In

Physical Sciences

Featured Portals

Articles by Topic

Social Sciences

Featured Portals

Articles by Topic

Biological Sciences

Featured Portals

Articles by Topic

Differential modulation of global and local neural oscillations in REM sleep by homeostatic sleep regulation

Significance

Abstract

Results

Persistent Increase in REM Sleep Time Rebounds After Repetitive Sleep Deprivation.

Persistent Increase in REM Sleep Low-Frequency Oscillations During CSR.

Delta power (1–4 Hz).

Theta power (5–10 Hz).

Progressive Increase in REM Sleep High-Frequency Oscillations During CSR.

Beta power (15–30 Hz).

Gamma power (30–90 Hz).

Interaction Between Slow and Fast Oscillations in REM Sleep During CSR.

Hippocampal Theta and Coupled Gamma Activity in REM Sleep During CSR.

Low- and High-Frequency Oscillations in the Frontal Cortex in REM Sleep During CSR.

Discussion

Noncumulative REM Sleep Rebound.

The Pattern of REM Sleep Rebound Is Mirrored in EEG Slow Oscillations.

Progressive Increase in High-Frequency Oscillations Reflects Activation of Local Cortical Circuits.

Functional Implications: Activation of Local Cortical Circuits and REM Sleep-Related Synaptic Potentiation.

Experimental Procedures

CSR.

Sleep Recording.

High-Density EEG Analysis.

Detection of Phasic REM.

Testing Statistical Significance.

SI Experimental Procedures

Animals.

Surgical Procedure of Microarray.

Nanofabrication and Characterization of Microarray.

High-Density EEG.

Sleep Scoring.

MI.

Acknowledgments

Footnotes

References

Citation Manager Formats

Sign up for Article Alerts

Jump to section

You May Also be Interested in

Similar Articles

Articles

PNAS Portals

Information