Research ArticleNEUROSCIENCE

Coordinated infraslow neural and cardiac oscillations mark fragility and offline periods in mammalian sleep

See allHide authors and affiliations

Science Advances  08 Feb 2017:
Vol. 3, no. 2, e1602026
DOI: 10.1126/sciadv.1602026
  • Fig. 1 The 0.02-Hz oscillation in sigma power in undisturbed non-REM sleep of mice and humans.

    (A to D) Sleep analysis in freely moving mice (n = 18). (A and B) Sigma (red; 10 to 15 Hz) and SWA (blue; 0.75 to 4 Hz) power time course for a single mouse, with hypnograms shown below. Gray-shaded area in (A) is expanded in (B), with aligned band-pass–filtered ECoG traces. (C) FFT of power time course for sigma (left) and SWA (right) for individual mice (gray traces, n = 18) and for the average across mice (color + shading, means ± SEM). Open circles denote FFT peaks obtained from Gaussian fits (their SD was 0.015 Hz). Vertical dotted lines indicate mean peak frequency ± 0.5 SD. Minor ticks are added to indicate the 0.02-Hz value on the frequency axis. (D) Mean peak values from (C) for sigma power, SWA, theta (6 to 10 Hz), and beta (16 to 20 Hz) bands (Friedman rank sum test; P = 4.9 × 10−8, post hoc Wilcoxon signed-rank tests relative to sigma power, P = 7.63 × 10−6 for SWA; P = 3.81 × 10−5 for theta; P = 1.53 × 10−5). (E to J) Sleep analysis in humans (n = 27). (E and F) Same as (A) and (B) for a single human subject (sigma, 10 to 15 Hz; SWA, 0.5 to 4 Hz). (G) Power spectral profiles for sigma power and SWA time course during non-REM sleep (S2 + SWS) graphed as in (C). The open circles indicate the peak of Gaussian fits, which show an SD of 0.008 Hz. (H) Left: Same as (D), calculated over all non-REM sleep (theta, 4 to 8 Hz; beta, 16 to 20 Hz; beta2, 20 to 24 Hz). Same statistics as (D): P = 3.5 × 10−4; post hoc Wilcoxon signed-rank tests with P = 9.5 × 10−6 for SWA; P = 0.009 for theta; P = 0.095 for beta; P = 4.3 × 10−4 for beta2. Right: Individual fast spindle power peaks (FSP ± 1 Hz) and adjacent frequency bands. Same statistics as in (C): P = 0.034 for Friedman rank sum test; P = 0.015 for −5 to −3 Hz; P = 0.032 for +3 to +5 Hz. (I) Same analysis as (G), restricted to S2 or to SWS and sigma power. (J) Left: Power analysis as in (H) restricted to S2 sleep revealed a prominent peak for sigma power over other frequency bands (n = 27; Friedman rank sum test, P = 8.5 × 10−6, followed by Wilcoxon signed-rank tests with respect to sigma power; P = 2.5 × 10−6 for SWA; P = 0.023 for theta; P = 0.011 for beta; P = 0.002 for beta2). Right: As left for SWS only (same statistics as the left panel: P = 1.8 × 10−4; P = 0.11 for SWA; P = 0.052 for theta; P = 6 × 10−4 for beta; P = 1.6 × 10−4 for beta2). +P < 0.1, *P < 0.05, **P < 0.01, ***P < 0.001.

  • Fig. 2 The 0.02-Hz oscillation is present in local cortical areas and predominates in somatosensory cortex.

    (A) Top view of mouse brain with indication of recording sites and with corresponding representative traces obtained during non-REM sleep scored on the basis of EEG/EMG recordings. (B) Sigma (red) and SWA (blue) power time course for a single non-REM sleep bout recorded simultaneously from all areas. The gray-shaded area indicates the time corresponding to the traces in (A). Dotted lines indicate 100%. (C) FFT of power time course for sigma (left) and SWA (right) for individual mice (gray traces, n = 6) and for the average across mice (color + shading, means ± SEM). Open circles denote FFT peaks obtained from Gaussian fits. Vertical dotted lines indicate mean peak frequency ± 0.5 SD. (D) Mean peak values from (C) for sigma power and SWA for all brain areas and EEG recordings, analyzed as in Fig. 1D. RM ANOVA with factors “area” and “frequency”; area, P = 1.25 × 10−6; frequency, P = 6.88 × 10−5; post hoc paired t tests; EEG, t = 11.19, P = 9.93 × 10−5; SI, t = 17.88, P = 1.01 × 10−5; SII, t = 5.72, P = 0.0023; AC, t = 2.83, P = 0.037; mPFC, t = 2.02, P = 0.1; *P < 0.01, ***P < 0.001. SI and SII, primary and secondary somatosensory cortex; AC, auditory cortex; mPFC, medial prefrontal cortex; Ref, reference.

  • Fig. 3 Regional cortical topology of the 0.02-Hz oscillation in humans.

    (A) Top: Color scale that indicates the mean normalized power values calculated from the average 0.02-Hz oscillation band (±0.5 SD around average peak values) during non-REM sleep. Bottom: The power spectral profiles for the FSP band (FSP ± 1 Hz, ~13 Hz, left) and the SWA band (right) averaged across subjects (color + shading, means ± SEM) displayed for representative midline electrodes (FZ, CZ, and PZ); analysis as in Fig. 1G. Coloring for power spectral profiles and each subject’s 0.02-Hz oscillation peak (filled circles underneath the power spectral profiles) corresponds to normalized peak values in the color scale. Insets show human head with an approximate topography of the mean normalized peak power values for all nine EEG electrodes (F3, FZ, F4, C3, CZ, C4, P3, PZ, and P4). (B) Mean (±SEM) normalized peak values for FSP band and adjacent frequency bands (FSP −5 to −3 Hz and FSP +3 to +5 Hz), as well as sigma power (10 to 15 Hz), SWA (0.5 to 4 Hz), theta (4 to 8 Hz), beta (16 to 20 Hz), and beta2 (20 to 24 Hz) bands separate for the three midline electrodes (FZ, CZ, and PZ); analysis as in Fig. 1H with data from the participants of the memory study (n = 24). Additional Friedman rank sum test between three midline electrodes for FSP band (P = 3.5 × 10−8), sigma (P = 6.35 × 10−9), and SWA (P = 2.8 × 10−6) (top three horizontal lines), with post hoc–paired comparisons along decreases from Pz to Cz as well as Cz to Fz separate for those three frequency bands (Wilcoxon signed-rank test, all Ps < 0.0036). For consistency with the core study analyses, which relied on nonparametric statistics, the same statistical tests were performed here. +P < 0.1, *P < 0.05, **P < 0.01, ***P < 0.001 for Wilcoxon signed-rank test relative to FSP band (left bar groups) and relative to the sigma band (right bar groups).

  • Fig. 4 The 0.02-Hz oscillation imposes periods of high and low fragility to acoustic noise.

    (A) Top: Acoustic stimulation protocol. Bottom: Percentage (means ± SEM) of wake-up and sleep-through trials (n = 10 mice). (B) Representative EEG (ECoG) (upper trace)/EMG (lower trace) traces from wake-up and sleep-through trials. Gray-shaded area indicates period of noise exposure. Scale bars, 400 and 80 μV for EEG(ECoG)/EMG. (C) Time course of sigma power for the 40 s of non-REM sleep before noise onset for a wake-up (violet) and a sleep-through (orange) trial [same data as (B)]. Insets show corresponding band-pass–filtered (10 to 15 Hz) EEG (ECoG) traces. Scale bars, 200 μV. (D) Left: Means ± SEM sigma power time course for wake-up and sleep-through trials (n = 9 and 10, respectively; RM ANOVA for factors “time” and “behavioral outcome”, Greenhouse-Geisser–corrected, P < 0.0042). Right: Overlay of the traces from the left, once unfiltered (continuous line) and once band-pass–filtered (dotted lines) for the frequencies corresponding to 0.02-Hz oscillation peak ± 1 width (see Fig. 1C). (E) Means ± SEM SWA time course as in (D) [same statistics as in (D), P = 0.11]. (F) Projected time course of sigma power during noise exposure for wake-up and sleep-through trials. (G) Waveforms for average wake-up (n = 4) and sleep-through (n = 8) trials obtained through sinusoidal fits. (H) Polar representation of sigma power phases decoded from (G) (shaded area; 0°, peak), with shading of corresponding intervals for high (purple) and low (orange) responsiveness to stimulation.

  • Fig. 5 Sleep benefit in episodic memory correlates with the strength of the 0.02-Hz oscillation in the FSP band (FSP ± 1 Hz).

    (A) Correlation of episodic memory recall (that is, recall of objects in their spatiotemporal context) with normalized peak values of FSP band. Pearson’s r values are given in all panels (*P = 0.027). (B) Same for SWA band (P = 0.26). (C) Normalized power in FSP band was positively associated with the density of fast spindles (*P = 0.011). Analyses were performed on the average of all parietocentral EEG electrodes (C3, CZ, C4, P3, PZ, and P4) for the FSP band, and across frontal electrodes (F3, FZ, and F4) for the SWA band, as these sites correspond to the locations with the highest overall power in the respective bands.

  • Fig. 6 The 0.02-Hz oscillation aligns with heart rate changes in both mice and humans.

    (A) Representative non-REM sleep bout with simultaneous recording of sigma power (red trace) and heart rate [black trace; in beats per minute (bpm)]. Insets show 1-s period of corresponding raw data (squared) to illustrate R-wave detection in EMG traces. (B) Cross-correlogram between sigma power and heart rate for traces in (A). (C) Same as (A) for a single human subject. (D) Corresponding cross-correlogram as (B). (E to F) Mean cross-correlogram for mice (n = 12) (E) and humans (n = 27) (F). Shadowing represents means ± SEM.

Supplementary Materials

  • Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/3/2/e1602026/DC1

    fig. S1. The 0.02-Hz oscillation is prominent for sigma power throughout both short and long non-REM sleep bouts in mice.

    fig. S2. Scheme of analysis for 0.02-Hz oscillations in mice.

    fig. S3. The 0.02-Hz oscillation is robust against the choice of non-REM sleep bout length for analysis and does not result from an 1/f power dependence.

    fig. S4. The sigma power dynamics in both mice and humans show a periodicity on a 0.02-Hz time scale, as assessed through autocorrelations.

    fig. S5. Scheme of analysis for 0.02-Hz oscillations in humans.

    fig. S6. Sleep parameters for the participants of the studies in humans and predominance of 0.02-Hz oscillations in S2 sleep.

    fig. S7. The 0.02-Hz oscillation is prominent for sigma power throughout early non-REM sleep in humans.

    fig. S8. Sleep in head-fixed animals reproduces the three major vigilance states and their spectral characteristics found in freely moving animals.

    fig. S9. Acoustic stimuli causing early or late wake-ups fall onto late or early portions of the declining sigma power phase, respectively.

    fig. S10. Wake-up and sleep-through trials do not depend on previous sleep duration.

    fig. S11. Ripple power increases precede sigma power elevations.

    fig. S12. Nuchal EMG recordings faithfully detect the R-waves of the heartbeat in mice.

  • Supplementary Materials

    This PDF file includes:

    • fig. S1. The 0.02-Hz oscillation is prominent for sigma power throughout both short and long non-REM sleep bouts in mice.
    • fig. S2. Scheme of analysis for 0.02-Hz oscillations in mice.
    • fig. S3. The 0.02-Hz oscillation is robust against the choice of non-REM sleep bout length for analysis and does not result from an 1/f power dependence.
    • fig. S4. The sigma power dynamics in both mice and humans show a periodicity on a 0.02-Hz time scale, as assessed through autocorrelations.
    • fig. S5. Scheme of analysis for 0.02-Hz oscillations in humans.
    • fig. S6. Sleep parameters for the participants of the studies in humans and predominance of 0.02-Hz oscillations in S2 sleep.
    • fig. S7. The 0.02-Hz oscillation is prominent for sigma power throughout early non-REM sleep in humans.
    • fig. S8. Sleep in head-fixed animals reproduces the three major vigilance states and their spectral characteristics found in freely moving animals.
    • fig. S9. Acoustic stimuli causing early or late wake-ups fall onto late or early portions of the declining sigma power phase, respectively.
    • fig. S10. Wake-up and sleep-through trials do not depend on previous sleep duration.
    • fig. S11. Ripple power increases precede sigma power elevations.
    • fig. S12. Nuchal EMG recordings faithfully detect the R-waves of the heartbeat in mice.

    Download PDF

    Files in this Data Supplement: