Contrasting Existence and Robustness of REM/Non-REM Cycling in Physiologically Based Models of REM Sleep Regulatory Networks

Typical human sleep throughout the night consists of alternating periods of rapid eye movement (REM) sleep and non-REM (NREM) sleep. This ultradian rhythm of NREM/REM cycling is thought to be produced by the state-dependent activity of REM-on and REM-off brainstem and hypothalamic neuronal populations that, respectively, promote or suppress REM sleep. Synaptic interactions among these populations define REM sleep regulatory networks; however, the identity of the key neuronal populations in these networks and the dynamics of interactions among them are disputed and cannot be addressed comprehensively with current experimental techniques. The purpose of this study is to use physiologically based mathematical models to explore the dynamic implications associated with competing hypotheses for network-based REM sleep regulation. Generally, putative REM sleep regulatory networks fall into two categories: a reciprocal interaction network consisting of an excitatory REM-on population interacting with an inhibitory REM-off population, and a mutual inhibition network where both REM-on and REM-off populations are inhibitory. We focus on the generation of regular periodic cycling solutions, which would generate the NREM/REM ultradian rhythm, in these networks. By applying our understanding of network mechanisms, we develop efficient numerical criteria for the existence of stable cycling solutions in each network structure. To investigate the robustness of cycling, we systematically analyze the response of model dynamics to manipulation of the components governing network interactions. By establishing the implications of network structure for the mechanisms and dynamics of NREM/REM state transitions, this comparative analysis identifies key targets for future experimental work to distinguish the structure of the proposed physiological REM sleep regulatory network.