Spectral and network investigation reveals distinct power and connectivity patterns between phasic and tonic REM sleep

Abstract

Although rapid eye movement (REM) sleep is often thought of as a singular state, it consists of two substates, phasic and tonic REM, defined by the presence (respectively absence) of bursts of rapid eye movements. These two substates have distinct EEG signatures and functional properties. However, whether they exhibit regional specificities remains unknown. Using intracranial EEG recordings from 31 patients, we analyzed expert-labeled segments from tonic and phasic REM and contrasted them with wakefulness segments. We assessed the spectral and connectivity content of these segments using Welch's method to estimate power spectral density and the phase locking value to assess functional connectivity. Overall, we found a widespread power gradient between low and high frequencies (p < 0.05, Cohen's d = 0.17 ± 0.20), with tonic REM being dominated by lower frequencies (p < 0.01, d = 0.18 ± 0.08), and phasic REM by higher frequencies (p < 0.01, d = 0.18 ± 0.19). However, some regions, such as the occipito-temporal areas as well as medial frontal regions, exhibit opposite trends. Connectivity was overall higher in all bands except in the low and high ripple frequency bands in most networks during tonic REM (p < 0.01, d = 0.08 ± 0.09) compared to phasic REM. Yet, functional connections involving the visual network were always stronger during phasic REM when compared to tonic REM. These findings highlight the spatiotemporal heterogeneity of REM sleep which is consistent with the concept of focal sleep in humans.

Keywords: REM; connectivity; microstate; phasic REM; spectrum; tonic REM.

Graphical Abstract

Figure 1.

Patient selection flow chart.

Figure 2.

Schematic of the analysis pipeline. Patients with full-night sEEG recordings were analyzed. Experts marked wakefulness, tonic, and phasic REM segments. Displayed is a representative patient (#2) showing a few selected intracranial channels, chin electromyography (EMG), and electro-oculogram (EOG). EOG was used to identify rapid eye movements (see the dashed box). The duration of the selected segments is marked by the underline colored (yellow-wakefulness, phasic- red, tonic- blue). Segments were then analyzed for their spectral and connectivity content using Welch’s method and the phase locking value (PLV). Channels were then grouped into regions using the MICCAI 38, and the Yeo 7 network atlas respectively. Regions and networks with at least 3 patients and 5 channels were then tested for differences between tonic and phasic REM in a paired manner and then compared to wakefulness segments in a non-paired manner. LOF: Left orbitofrontal, LA: left amygdala, LL: left lingual gyrus, RA: right amygdala, Lhp: Left hippocampus.

Figure 3.

Coverage of regions and networks. Reported are the numbers of channels and patients for each region of the MICCAI 38-region atlas, and YEO 7-network atlas. Regions and networks with less than 3 patients and/or 5 channels were not analyzed and are marked in black as not available.

Figure 4.

Brain wide power differences between phasic and tonic REM. Effect sizes (Cohen’s d) are plotted as colors on each available region exhibiting significant differences between matching time periods of phasic and tonic REM. The effect sizes of significant differences are presented for each power band tested. Significance was set to 0.05 after FDR correction.

Figure 5.

Regional power analysis of high frequencies. The power distributions of every segment for tonic REM, phasic REM, and wakefulness at high frequencies are presented. Regions exhibiting a significant difference and a moderate effect size (d > 0.3) between wakefulness and phasic REM are shown. Note that the differences between tonic and phasic REM are matched paired and tests using a paired t-test design, while the differences between phasic REM and wakefulness, and tonic REM and wakefulness are based on the distributions and tests using a regular unpaired t-test. Results are FDR corrected. Significant results are considered for p < 0.05.

Figure 6.

Power trend differences between tonic REM, phasic REM, and wakefulness. The trends for significant differences between wakefulness and phasic and tonic REM have been classified along 6 possible behaviors and color coded for each power band. Every significant regional trend is represented with a color depending on which time period exhibited the highest and lowest power. For example, red represents regions where phasic REM had the highest power, followed by tonic REM, and then wakefulness (phasic > tonic > wakefulness).

Figure 7.

Network-based connectivity differences between phasic and tonic REM. Effect sizes of connectivity, which was significantly different between phasic and tonic REM are plotted as the size and color of connecting lines between each network-pair. Connectivity differences are visualized with blue indicating significantly higher connectivity during tonic REM, and red highlighting regions where connectivity was stronger during phasic REM. (A) all significant results including both low and moderate effect sizes. (B) Only connections which had significant difference with an effect size d > 0.3. Significance was set to 0.05 after false discovery rate correction. V—Visual, S—Somatomotor, DA—Dorsal attention, VA—Ventral attention, L—Limbic, FP—Frontoparietal, DM—Default mode network.

Figure 8.

Network analysis of high frequencies. The PLV distributions for wakefulness, phasic REM, and tonic REM sleep at high frequencies are presented. In this figure, we are representing and comparing with the wakefulness state, the distribution of PLV values for the network pairs that exhibited a significant differences between phasic and tonic and with a moderate effect size (d > 0.3). Note that the differences between tonic and phasic REM are matched pairs, while the differences between phasic REM and wakefulness, and tonic REM and wakefulness, are based on the distributions.