REM sleep is characterized by infra-slow waves propagating across the neocortical hierarchy

2588 

Abstract Submission 

Xufu Liu¹, Dante Picchioni², Yifan Yang¹, Hendrik Mandelkow², Jacco de Zwart², Jeff Duyn², Xiao Liu¹

¹The Pennsylvania State University, STATE COLLEGE, PA, ²National Institutes of Health, Bethesda, MD

Xufu Liu  
The Pennsylvania State University
STATE COLLEGE, PA

Dante Picchioni  
National Institutes of Health
Bethesda, MD

Yifan Yang  
The Pennsylvania State University
STATE COLLEGE, PA

Hendrik Mandelkow  
National Institutes of Health
Bethesda, MD

Jacco de Zwart  
National Institutes of Health
Bethesda, MD

Jeff Duyn  
National Institutes of Health
Bethesda, MD

Xiao Liu  
The Pennsylvania State University
STATE COLLEGE, PA

Infra-slow (<0.1 Hz) spatiotemporal dynamics play a crucial role in shaping resting-state functional connectivity, which has been linked to various aspects of brain function and dysfunction(1,2). Newly reported propagating waves represent a specific form of these infra-slow dynamics, which travel along the cortical hierarchical gradient between sensory/motor (SM) cortices and the higher-order default mode network (DMN)(3). This propagating direction is consistent with flows of information required for optimizing artificial neural networks(4), suggesting a possible link of these propagating waves to learning and memory. Cross-hierarchy information flow has also been hypothesized to be critical for sleep, particularly rapid eye movement (REM) sleep(5), to fulfill its known role in offline learning and memory. Nevertheless, it remains unclear how the cross-hierarchy propagating waves are modulated across sleep stages, particularly during REM sleep. Here, we examined the cross-hierarchy fMRI propagating waves across various sleep stages through whole-night fMRI/electroencephalography (EEG) recordings.

We analyzed 12 subjects, each with two nights of simultaneous fMRI/EEG(6). Among them, 18 nights showed REM sleep. We identified cross-hierarchy propagating waves(3) in bottom-up (SM to DMN) and top-down (DMN to SM) directions and compared their occurrence fraction under various sleep stages. We focused on stable sleep periods during which a specific sleep stage lasted longer than one minute. We aligned and averaged fMRI propagations according to the global mean BOLD (gBOLD) peak within each identified propagation and converted the result into t-score maps. Relative phase delay of a specific region in bottom-up propagation is the time delay between its peak activation and the propagation center defined as the gBOLD peak. We quantified eye movements based on two independent measurements: the electrooculogram (EOG) signal variation and motion signals derived from concurrent video recordings(7). We identified isolated eye movements (at least 8 seconds of stable eye position on either side, duration less than 3 seconds), and examined fMRI changes around them.

Bottom-up propagations gradually increased from wakefulness to deeper NREM sleep stages (N1 to N3), peaking during REM sleep (Fig.1A). Top-down propagations followed a similar trend during wakefulness and NREM stages but exhibited an opposite change and reached its minimum during REM sleep (Fig.1B). Hence, REM sleep featured the largest rate difference between the two types of propagations (Fig.1C). Importantly, the bottom-up propagating waves during REM sleep showed distinct dynamics in the thalamus, pons and visual cortex, key regions with established involvement in ponto-geniculo-occipital (PGO) waves of REM signature (Fig.1D and 1E). Their phase in the bottom-up propagations (i.e., relative time delay to the propagation center) was significantly advanced to earlier time as compared to other conditions (Fig.1E), and this effect is absent in the control somatosensory areas. The bottom-up propagations during REM sleep were also associated with systematic eye movements, as estimated by EOG variation (Fig.2A). Consistent with this, fMRI changes at eye movements detected via EOG showed progressively increasing time delays along the cortical hierarchical gradient, as described by a principal gradient (PG) direction (Fig.2B). This pattern manifested as a bottom-up propagation on the cortical surface (Fig.2C) and is associated with early thalamic and pontine activity (Fig.2C). In addition, fMRI changes at isolated eye movements detected through concurrent video recordings displayed similar dynamics of bottom-up propagating waves (Fig.2D-2F).

Cross-hierarchy propagating waves of fMRI activity are significantly modulated across sleep stages. REM sleep is characterized by notably dominant bottom-up propagations, which appear linked to PGO waves and eye movements.

Subcortical Structures

BOLD fMRI ²

EEG

Sleep and Wakefulness ¹

Brainstem

Electroencephaolography (EEG)

MRI

Sleep

Sub-Cortical

Thalamus