Mammalian sleep dynamics: how diverse features arise from a common physiological framework

Abstract

Mammalian sleep varies widely, ranging from frequent napping in rodents to consolidated blocks in primates and unihemispheric sleep in cetaceans. In humans, rats, mice and cats, sleep patterns are orchestrated by homeostatic and circadian drives to the sleep-wake switch, but it is not known whether this system is ubiquitous among mammals. Here, changes of just two parameters in a recent quantitative model of this switch are shown to reproduce typical sleep patterns for 17 species across 7 orders. Furthermore, the parameter variations are found to be consistent with the assumptions that homeostatic production and clearance scale as brain volume and surface area, respectively. Modeling an additional inhibitory connection between sleep-active neuronal populations on opposite sides of the brain generates unihemispheric sleep, providing a testable hypothetical mechanism for this poorly understood phenomenon. Neuromodulation of this connection alone is shown to account for the ability of fur seals to transition between bihemispheric sleep on land and unihemispheric sleep in water. Determining what aspects of mammalian sleep patterns can be explained within a single framework, and are thus universal, is essential to understanding the evolution and function of mammalian sleep. This is the first demonstration of a single model reproducing sleep patterns for multiple different species. These wide-ranging findings suggest that the core physiological mechanisms controlling sleep are common to many mammalian orders, with slight evolutionary modifications accounting for interspecies differences.

Figure 1. Schematic of the sleep model.

Bihemispheric model (gray box), and its extension to model unihemispheric sleep, including MA and VLPO populations, and circadian (C), homeostatic (H), and cholinergic/orexinergic (ACh/Orx) drives. Arousal state feeds back to H. Pointed and rounded arrowheads indicate excitatory and inhibitory connections, respectively. To model unihemispheric sleep we add an inhibitory VLPO-VLPO connection (dotted arrows). Time series are shown alongside MA and H, showing simulated human bihemispheric (top) and dolphin unihemispheric sleep (bottom), with solid and dashed lines distinguishing the hemispheres.

Figure 2. Map of system dynamics corresponding to different mammalian species.

(A) Parameters corresponding to sleep patterns of 14 mammalian species, using data from the following sources: rat, mouse, hamster, squirrel and chinchilla , eastern mole , asian elephant , dog , jaguar , cat , fox , opossum , armadillo , common shrew , rhesus monkey , and slow loris . (B) Sleep duration for these parameters, with zones corresponding to different numbers of sleep episodes per day, as labeled.

Figure 3. Comparison of experimental data to model output.

Time series for wake vs. sleep state are shown for three species, comparing the model to experimental data. Human: (A) data from , (B) model (, h). Elephant: (C) data from , (D) model (, h). Opossum: (E) data from , (F) model (, h). Noise is added to the model to make sleep patterns less regular (see Methods for numerical details).

Figure 4. Positive correlation between homeostatic time constant and body mass.

Log-log plot of homeostatic time constant (ranges from regions in Fig. 2) vs. body mass for 17 species. Linear fits are shown for non-primates (solid, ), corresponding to a power law with exponent 0.29±0.10 (Mean±S.D. calculated using bootstrapping), and for primates (dashed, ) with exponent 0.01±0.26. A linear fit to all species () yields an exponent of 0.28±0.12.

Figure 5. Model simulation of unihemispheric sleep.

Simulated transition from polyphasic bihemispheric (BHS) to unihemispheric sleep (UHS), effected by increasing VLPO-VLPO connection strength. Raster plot of sleep for left (white) and right (black) hemispheres, with environmental light level indicated by background brightness. This simulates the behavior of a fur seal in a terrestrial environment on days 0–2 and aquatic thereafter. The VLPO-VLPO connection strength linearly increases from 0 to during the transition period on days 2–4.