Circadian regulation of slow waves in human sleep: Topographical aspects

Abstract

Slow waves (SWs, 0.5-4Hz) in field potentials during sleep reflect synchronized alternations between bursts of action potentials and periods of membrane hyperpolarization of cortical neurons. SWs decline during sleep and this is thought to be related to a reduction of synaptic strength in cortical networks and to be central to sleep's role in maintaining brain function. A central assumption in current concepts of sleep function is that SWs during sleep, and associated recovery processes, are independent of circadian rhythmicity. We tested this hypothesis by quantifying all SWs from 12 EEG derivations in 34 participants in whom 231 sleep periods were scheduled across the circadian cycle in a 10-day forced-desynchrony protocol which allowed estimation of the separate circadian and sleep-dependent modulation of SWs. Circadian rhythmicity significantly modulated the incidence, amplitude, frequency and the slope of the SWs such that the peaks of the circadian rhythms in these slow-wave parameters were located during the biological day. Topographical analyses demonstrated that the sleep-dependent modulation of SW characteristics was most prominent in frontal brain areas whereas the circadian effect was similar to or greater than the sleep-dependent modulation over the central and posterior brain regions. The data demonstrate that circadian rhythmicity directly modulates characteristics of SWs thought to be related to synaptic plasticity and that this modulation depends on topography. These findings have implications for the understanding of local sleep regulation and conditions such as ageing, depression, and neurodegeneration which are associated with changes in SWs, neural plasticity and circadian rhythmicity.

Keywords: EEG; Forced desynchrony; Homeostasis; Slope analyses.

Graphical abstract

Fig. S1

Data processing flow chart from data collection to statistical analysis. The 12 channels were processed independently. The 20 min intervals were the basic units for the statistical analyses of the SW parameters and SWA. The data from these intervals were collapsed into larger sleep and circadian bins depending on the actual statistical analyses. See Materials and methods section for further details.

Fig. S2

A representative example of the effect of sleep timing on sleep structure and on the time course of multiple slow wave (SW) parameters across each studied EEG derivation (BB0214, Female, 31 years) presented on an interpolated color map. A. Sleep stages during FDn1 when sleep was scheduled to occur at the habitual times, and the interpolated map of corresponding time course (1 min resolution) of slow wave activity (SWA: 0.5–4 Hz), incidence, absolute amplitude, duration, absolute mean slope of the individually detected negative half-waves (> 5 μV, 0.5–4 Hz) of SWs across all studied EEG derivations. B. Sleep stages and the time course of studied SW parameters across all EEG derivations as measured during the 9 hour and 20 minute sleep period FDn4 scheduled 180^o (~ 12 h) out of phase with habitual sleep time.

Fig. S3

A representative example of the effect of sleep timing on the association between the absolute mean slope and the amplitude of the SWs during the first and second half of the sleep episode (BB0135, Female, 23 years). Fitted least square mean regression lines between the absolute mean slope and the amplitude of the SWs (negative half waves > 37.5 μV, absolute value) as measured during the first and the second half of the night when sleeping at habitual nocturnal bedtimes s (FDn1) and the sleep episode starting 12 h later, i.e. during the biological day (FDn4). The data show that when sleeping during the biological day, for any given amplitude the slope is greater, than when sleeping during the biological night.

Fig. S4

The effect of SW half-wave segment on the duration and slope measures of both positive and negative SW measures. A. The effect of half-wave segment (initial versus final) and sleep-dependent regulation on the absolute slope and duration of SW for both negative and positive half-waves. Least square mean (Lsmeans) and standard error of the mean (SEM) are presented indicating sleep-dependent estimates at each 3-hour and 6.7-minute intervals (third of the sleep period) measured across all studied circadian phases and EEG derivations for both the initial and final segments (open circles = initial segment; filled bars = final segment). Data are double plotted for a better visualization of sleep-dependent rhythmicity. The level of significance is indicated for the main effect of sleep-dependent (H) regulation and the interaction (H ∗ S) with the segment of the SW half-wave factor (*P < .005, **P < .001, ***P < .0005). B. The circadian regulation of SW half-wave duration and various types of slope measures (absolute values) as a function of half-wave polarity and half-wave segment. Lsmeans and SEM indicate circadian phase-dependent estimates at 60^o (~ 4 h) bins measured across all studied sleep intervals and EEG derivations for both the initial and final segments (open circles = initial segment; filled bars = final segment). Data are double plotted for a better visualization of circadian rhythmicity. Zero phase is set at the dim light melatonin onset. The level of significance is indicated for the main effect of circadian regulation (C) and the interaction (C ⁎ S) with the segment of the SW half-wave factor (***P < .0005). C. The main effect of the segment of the half-wave, independent of the sleep-dependent and circadian factors on the slope measures (absolute values) and duration of both positive and negative SW half-waves (open bars = initial segment; filled bars = final segment) (*P < .005, ***P < .0005).

Fig. 1

The forced desynchrony protocol. A. Double raster plot of the 28-h forced desynchrony protocol with a representative example of sleep timing and melatonin profile (BB0102, male, 23 years, in vivo circadian period: 24.24 h). Consecutive 24-h periods are plotted next to and below each other. After an 8-hour adaptation night (ADn) followed by an adaption day (ADd) participants were scheduled to a 28-h sleep–wake cycle, in which 9 h and 20 min were scheduled for sleep (blue bars) and 18 h and 40 min for wakefulness (yellow bars). Thus, sleep and wake timing were shifted by 4 h every ‘day’ while at the same time the ratio of sleep and wakefulness remained 1:2, just as during a normal 24-hour day. Melatonin was assessed at baseline (FD1), FD4 and FD7 in order to assess phase and period of the central circadian pacemaker. Blood samples for melatonin concentration assessment were scheduled to be taken hourly but occasionally samples could not be collected due to technical or logistical problems. B. The forced desynchrony protocol represented on a 24-h angular plot indicating each individual baseline (FDn1) sleep period (red lines) against the circadian phase and relative clock time, which corresponds to the habitual sleep timing, duration and circadian phases covered by usual sleep studies. Grey lines indicate all sleep periods scheduled across the circadian cycle (N = 231) during the 10-day protocol for the 34 participants. Each sleep period indicates the time course between the scheduled bedtime (internal light blue circle) and wake-up time (external dark blue circle). The dotted green line indicates the average clock time at circadian phase zero, corresponding to the timing of the dim light melatonin onset.

Fig. 2

Baseline sleep and SW detection. A. Time course of sleep stages during the participant's (BB0214, Female) baseline night (FDn1) (upper panel), the time course of slow wave activity (SWA: 0.5–4.0 Hz) (middle panel) and the incidence of the individually detected slow waves (SWs) (bottom panel). The grey bars represent all the SWs > 5 μV. The red bars represent the SOs > 37.5 μV. Artefact segments appear as gaps in activity. B. A representative 15-second example of NREM sleep during baseline sleep (FDn1) of the same participant at ~ 1 h (upper panel) and at ~ 8 h after sleep onset, respectively. The thin black line represents the original raw EEG signal, the bold red lines indicate negative half-waves of EEG signals filtered between 0.5 and 4.0 Hz (see Materials and methods). C. Schematic representation of detection of SWs and extraction of relevant properties from a single half-wave in a 0.5–4Hz band-pass filtered signal. The negative half-wave enclosed between zero crossings A and B has its main peak (P) at time O. The dotted line tangential to the wave in I has the steepest slope (maximum slope) in the initial (AO) phase, while the one tangential to the wave in F has the steepest slope in the final (OB) phase. Mean initial and final slopes are calculated as the ratios PO/OA and PO/BO, respectively. The numbers of peaks, two in the example given, are also counted. The dashed lines indicate the ± 5 μV and ± 37.5 μV voltage threshold levels used as amplitude criteria for detecting SWs. D. The effect of time in bed (quarter of sleep period) and brain topography on SWA in NREM sleep and the raw SW measures during the first 8 h of baseline sleep (FDn1). The results for negative half-waves are presented. Least square (Ls) means (absolute values) and standard error of the mean (SEM) derived from the mixed model analyses are indicated for all studied sleep intervals (2 hourly bins) and main brain regions. The brain topography factor comprises three main brain regions each including weighted averages over the Frontal (Fp1, Fp2, F3, F4), Central (C3, C4, T3, T4), and Posterior (P3, P4, O1, O2) areas. The three brain derivations are indicated with different circles (Frontal derivation = dark blue square; Central derivation = dark green circle; Posterior derivation = light green triangle). For statistical results of the SW parameters please refer to Table 1.

Fig. 3

The circadian and sleep-dependent regulation of sleep efficiency, NREM and REM sleep. A. The sleep-dependent regulation of sleep efficiency (F_2,220 = 12.43, P < .0001; Cohen's f² = 0.11) , NREM sleep (F_2,387 = 30.12, P < .0001; Cohen's f² = 0.16) and REM sleep (F_2,387 = 30.12, P < .0001; Cohen's f² = 0.16). Least square mean (Lsmeans) and standard error of the mean (SEM) are presented indicating sleep-dependent estimates at each 3-hour and 6.7-minute intervals (third of the sleep period) measured across all studied circadian phases. Data are double plotted for a better visualization of sleep-dependent rhythmicity. B. The circadian regulation of sleep efficiency (F_5,369 = 12.07; P < 0.001; Cohen's f² = 0.17), NREM sleep (F_5,438 = 4.13, P = 0.0011; Cohen's f² = 0.05) and REM sleep (F_5,438 = 4.13, P = 0.0011; Cohen's f² = 0.05). Lsmeans and SEM indicate circadian phase-dependent estimates at 60^o (~ 4 h) bins measured across all studied sleep intervals. Data are double plotted for a better visualization of circadian rhythmicity. Zero phase is set at the dim light melatonin onset indicated by the vertical green dashed line. The grey area in the background represents average melatonin profiles.

Fig. 4

The sleep-dependent and circadian regulation of the slow wave (SW) parameters are presented for the negative half waves. For statistical results of the SW parameters please refer to Table 2. A. The sleep-dependent regulation of SWA (0.5–4 Hz), incidence, absolute amplitude, duration and the absolute slope of the individually detected SWs. Least square mean (Lsmeans) and standard error of the mean (SEM) are presented indicating sleep-dependent estimates at each 3-hour and 6.7-minute intervals (third of the sleep period) measured across all studied circadian phases and EEG derivations. Data are double plotted for a better visualization of sleep-dependent rhythmicity. The circadian regulation of SWA (0.5–4 Hz), incidence, amplitude, period and slope of the individually detected SWs. B. Lsmeans and SEM indicate circadian phase-dependent estimates at 60^o (~ 4 h) bins measured across all studied sleep intervals and EEG derivations. Data are double plotted for a better visualization of circadian rhythmicity. Zero phase is set at the dim light melatonin onset indicated by the vertical green line. The grey area in the background represents the average melatonin profile. C. The interaction between the sleep-dependent and circadian regulation of SWs. Lsmeans indicate estimates across 3 sleep intervals (third of the night) and 60^o (~ 4 h) circadian bins averaged across all EEG derivations. The oblique purple line indicates the trajectory during baseline night. The 3D representation indicates that for several SW parameters the sleep-dependent and circadian effects interact. D. Effect size (Cohen's f²) of sleep (red) and circadian phase (blue) dependent regulation of SWA and the studied SW parameters for each studied EEG derivation. The ~ symbol indicates EEG derivation in which circadian effect size is comparable with the sleep-dependent effect size, whereas > symbol indicates EEG derivation in which circadian effect size exceeds the sleep-dependent effect size (> = Circadian f² > sleep-dependent f² up to five times; >> = the Circadian f² > sleep-dependent f² more than five times).

Fig. 5

Topographical variation of the sleep dependent and circadian modulation of SW parameters. A. The sleep-dependent regulation of the incidence, amplitude, duration and slope of the individually detected SWs across all studied EEG derivations on an interpolated color map. Data are double plotted for a better visualization of sleep-dependent effects. B. The circadian regulation of the incidence, amplitude, duration and slope of the individually detected SWs across all studied EEG derivations on an interpolated color map. Data are double plotted for a better visualization of circadian rhythmicity. The grey line in the background represents the average melatonin profile. C. The relative magnitude of the sleep dependent and circadian modulation of SW parameters. Values represent ratio of the effects size (Cohen's f²) between the circadian effects size and the sleep dependent effect size. Warmer colours indicate a dominance of the circadian effect. S. = segment.