Abstracts

Rationale: It is widely recognized that stress and sleep loss can promote seizures in individuals with epilepsy. Some studies even suggest that rapid eye movement (REM) sleep and non-REM (NREM) sleep can differ in the extent to which they contribute to this effect. Experimental studies in animal models of chronic epilepsy could help clarify these effects. To this end, we propose a method for closed-loop sensory stimulation in mice and test its utility for selective REM sleep restriction in a pilot study. Methods: All methods were performed with prior IACUC approval. Four adult C57BL/6J mice, each 4-6 weeks old, were implanted under anesthesia with a head-mounted EEG/EMG preamplifier and monitored round-the-clock after allowing sufficient time for complete recovery from surgery. Spectral band power features were extracted from a baseline recording and used to design a classifier that could score 4s epochs of data into REM, NREM, and Wake states with accuracy verified against a human scorer. The classifier was then used to detect the occurrence of REM sleep in real time and briefly trigger a sensory stimulation device to interrupt sleep; stimulation was repeated at fixed intervals until sustained arousal was achieved. Sensory stimulation was implemented using a pad placed on the cage floor fitted with vibrating micromotors. This device was activated whenever REM sleep was detected from the EEG/EMG for a minimum duration of 4s. For each animal, the REM sleep deprivation protocol was applied for a period of six-hour period starting at noon on a different day from that on which the baseline was recorded. The performance of the REM sleep restriction technique was assessed by scoring data from the experimental phase of recording offline and comparing the percent time spent in each state and the distribution of REM bout duration against their values in the baseline. Results: The automated sleep restriction system detected bouts of REM sleep in mice in real time with a mean sensitivity of 96%, mean positive prediction value of 66%, and mean latency of 7.4s (for true positive detections). The time spent in REM sleep in the baseline was about 10%, which is typical for wild type mice, but less than 5% during the experimental period: a reduction of over 50% due to closed-loop sensory stimulation. There was a dramatic reduction in REM bout duration; this was accompanied by a slight reduction in NREM bout duration due to stimulation triggered by false positive detections but possibly also due to REM sleep homeostasis. There were no noticeable changes in awake behavior except for brief attention to the stimulus without a startle response. Conclusions: Changes in sleep quality can influence seizure likelihood. We have developed a simple but effective method for selective sleep restriction in mice and tested its ability to deprive them of REM sleep. The results suggest that the feasibility of the technique is adequate for the purpose. In future work, we expect to employ it for studying the effect of selective sleep restriction in a murine model of epilepsy. Funding: NIH grant NS083218 and KSCHIRT grant 10-5A.