Proceedings of The Physiological Society

Sleep Sleep and Circadian Rhythms (London, UK) (2018) Proc Physiol Soc 42, C19

Poster Communications

Modelling Local Sleep Homeostasis

C. W. Thomas1, M. Guillaumin1, L. McKillop1, P. Achermann2, V. Vyazovskiy1

1. Department of Physiology, Anatomy & Genetics, University of Oxford, Oxford, Oxfordshire, United Kingdom. 2. Institute of Pharmacology and Toxicology, University of Zurich, Zurich, Switzerland.

Sleep homeostasis refers to the process by which a need for sleep accumulates during wakefulness and dissipates during subsequent sleep. Homeostatic sleep need is typically measured using slow wave activity (SWA); oscillatory power at 0.5 - 4 Hz present in extracellular field potentials during NREM sleep, which is generated by synchronous alternating bouts of neuronal spiking activity ("on periods") and silence ("off periods"). Existing quantitative descriptions of sleep homeostasis describe the dynamics of slow wave activity only as a function of recent sleep-wake history. However, homeostatic sleep pressure is non-uniform across the brain, originates locally, and changes in association with neuronal activities. This project aims to develop quantitative models that predict homeostatic sleep need markers from multi-unit firing rate history within the same channel. Models were developed and applied to an existing dataset of electrophysiological recordings from 16-channel microwire arrays implanted in mouse frontal cortex (McKillop et al., 2018). Data were obtained continuously for several days while mice were freely behaving, including periods of spontaneous sleep and wake, and periods of 6hr sleep deprivation (by gentle handling and novel object presentation). To assess model fit quality, an error metric was defined as the sum of absolute differences between median simulated and empirical values over continuous NREM episodes at least 1 minute duration, weighted by episode length. Model parameters were algorithmically optimised to minimise this error metric. A simple model, in which homeostatic sleep need increases in proportion to multi-unit firing rate and decreases exponentially over time, can often describe the time course of SWA with high accuracy (n=28; 7 mice x 4 best quality channels per mouse). An alternative model, which employs a firing rate threshold, with a saturating exponential rise in homeostatic sleep need above threshold, and exponential decrease below, provided an even better fit to SWA, yielding a lower minimal error metric (p < 0.001, n=28, Wilcoxon signed rank test). The time course of off period occupancy shows qualitatively similar temporal dynamics to slow wave activity and is a viable alternative sleep homeostasis metric. Preliminary results using this suggest that restricting simulated homeostatic sleep need decay to detected off periods improves model fit (p < 0.05, n=6, Wilcoxon signed rank test). In conclusion, the dynamics of local homeostatic sleep need can be well described by models dependent solely on local neuronal activities, independent of any information about the animal's global wake-sleep state. The advancement of models of sleep homeostasis will likely provide means to test competing theories of its mechanistic origin within neurones and local networks.

Where applicable, experiments conform with Society ethical requirements