Neural and Homeostatic Regulation of REM Sleep

Abstract

Rapid eye movement (REM) sleep is a distinct, homeostatically controlled brain state characterized by an activated electroencephalogram (EEG) in combination with paralysis of skeletal muscles and is associated with vivid dreaming. Understanding how REM sleep is controlled requires identification of the neural circuits underlying its initiation and maintenance, and delineation of the homeostatic processes regulating its expression on multiple timescales. Soon after its discovery in humans in 1953, the pons was demonstrated to be necessary and sufficient for the generation of REM sleep. But, especially within the last decade, researchers have identified further neural populations in the hypothalamus, midbrain, and medulla that regulate REM sleep by either promoting or suppressing this brain state. The discovery of these populations was greatly facilitated by the availability of novel technologies for the dissection of neural circuits. Recent quantitative models integrate findings about the activity and connectivity of key neurons and knowledge about homeostatic mechanisms to explain the dynamics underlying the recurrence of REM sleep. For the future, combining quantitative with experimental approaches to directly test model predictions and to refine existing models will greatly advance our understanding of the neural and homeostatic processes governing the regulation of REM sleep.

Keywords: REM sleep; REM sleep homeostasis; brain state; neural circuits and behavior; sleep.

FIGURE 1

Sleep in mice. (A) Example electroencephalogram (EEG) and electromyogram (EMG) recordings from a mouse during REM sleep, NREM sleep and wakefulness (Wake). (B) Top, color-coded brain state (hypnogram) for a continuous 24 h recording from a mouse during dark and light cycles. Bottom, a 2 h segment of the hypnogram shown at an expanded scale.

FIGURE 2

Brain circuits controlling REM sleep. Neural populations in hypothalamus, pons, and medulla promoting REM sleep (REM-on, cyan) or suppressing REM sleep (REM-off, gray), and their synaptic interactions. Glutamatergic neurons in the SLD are strongly inhibited by GABAergic REM-off neurons in the vlPAG/DpMe, which in turn are inhibited by GABAergic REM-on populations in hypothalamus (MCH neurons), pons (SLD), and dorsal (DPGi) and ventral medulla. Rabies virus-mediated retrograde tracing confirmed that vM GABAergic neurons innervate vlPAG GABAergic neurons, and that vlPAG GABAergic neurons synapse onto SLD glutamatergic neurons. Dashed lines indicate the existence of axonal projections whose functional role, however, has not been tested yet. Sagittal brain scheme adapted from Allen Mouse Brain Atlas (© 2015 Allen Institute for Brain Science. Allen Brain Atlas API. Available from: http://brain-map.org/api/index.html).

FIGURE 3

Homeostatic regulation of REM sleep timing. (A) Scheme illustrating the model for homeostatic regulation of REM sleep timing. Homeostatic REM sleep pressure accumulates in the absence of REM sleep during inter-REM periods and discharges during REM sleep. REM sleep is triggered once the REM sleep pressure reaches the threshold θ (dashed line). Consequently, longer REM periods precede longer inter-REM intervals. (B) Correlation between inter-REM interval and preceding REM sleep episode duration. Each point represents a single episode (n = 972 episodes from 27 mice). REM episodes and inter-REM intervals are positively correlated (linear regression: R = 0.39, p = 6.8 × 10^{– 36}). The model in (A) may explain this positive correlation. (B) Reproduced from Weber et al. (2018).

FIGURE 4

Quantitative models for REM sleep regulation. (A) Schematic of the reciprocal interaction (RI) model. Each node represents the average firing rate of the REM-on or REM-off population. The model was implemented as described in Diniz Behn et al. (2013). (B) Simulation of the RI model depicting the activity of the REM-on and REM-off population along with the resulting hypnogram. (C) Schematic of the mutual inhibition (MI) model. Inhibitory populations of REM-on and REM-off neurons mutually inhibit each other (bottom). The nodes represent the average firing rates of the REM-on and REM-off populations. Sigmoid input-output (i/o) functions describe the steady-state firing rate dependent on the inhibitory synaptic input (top). To exhibit oscillations, the model requires an oscillatory external input or parameter change. The accumulation of homeostatic REM sleep pressure, denoted as h(t), induces a right-wards shift of the REM-off i/o function throughout NREM sleep, resulting in reduced excitability of REM-off neurons. The model was implemented as described in Diniz Behn et al. (2013). (D) Simulation of the MI model. The average activity of the REM-on and REM-off population (middle) is depicted along with the resulting hypnogram (top) and the homeostatic REM sleep pressure h(t) (bottom), which determines the shape of the REM-off i/o function. The color-coded dashed lines represent the shape of the i/o functions as shown in (C).

FIGURE 5

Experimental support for the MI model. (A) Scheme depicting electrophysiological recordings from vlPAG GABAergic neurons expressing ChR2 using an optrode (microelectrodes attached to an optic fiber). Viral vectors expressing Cre-dependent ChR2 were injected into the vlPAG of GAD2-Cre mice. Optrode recordings allow the experimenter to test whether a recorded unit is reliably driven by laser stimulation and thus can be classified as the ChR2-expressing cell-type. (B) Average normalized EEG spectrogram (top) and mean firing rate (z-scored) of REM-off vlPAG GABAergic neurons (bottom, n = 11) during two successive REM episodes and the inter-REM interval. Each REM sleep episode and inter-REM interval was temporally normalized to unit length. (C) Comparison of vlPAG activity during inter-REM interval following short (≤90 s) and long (>90 s) REM episodes. Following a long REM sleep period, the firing rate was significantly higher than following a short one. (B,C) Reproduced from Weber et al. (2018).