Non-rem Sleep Stage Transitions Control Rem Sleep Rhythm

1423 Stage Transition Mechanism for Ultradian Rhythm—Kishi et al neurons themselves. Although further modifications of the original reciprocal interaction model, e.g., the incorporation of selfinhibitory cholinergic autoreceptors and excitatory interactions between cholinergic and non-cholinergic REM-on neurons, have been proposed over the past 30 years with new findings of synaptic details of the model, none of these appear to have produced substantial changes in the basic framework of the original cholinergic–monoaminergic (REM-on–REM-off) reciprocal interaction.5,6,8 It is therefore straightforward to observe the shortening of REM-onset intervals by a cholinergic agonist and the prolongation by a cholinergic antagonist.12 However, according to various monoaminergic/cholinergic models,2 the monoaminergic antagonist should shorten REM-onset intervals, because decreased neurotransmission of monoaminergic neurons should result in the activation of cholinergic neurons via decreased inhibition of cholinergic neurons, while actual experimental observations in rats13 and humans14 show the opposite effect. On the other hand, γ-aminobutyric acid (GABA)-ergic neuronal populations have also been shown to be involved in REM sleep regulation.15,16 A flip-flop switch model for REM sleep alternation has recently been proposed, in which the ultradian REM sleep alternation is controlled by mutually inhibitory interactions between GABA-ergic REM-on and REM-off neuronal populations.15 As this model assumes that the GABA-ergic REM-off neurons have excitatory inputs from monoaminergic REM-off neurons, it is also expected that the monoaminergic antagonist should shorten the REM-onset intervals. This is because decreased neurotransmission of monoaminergic neurons should lead to the decreased firing of GABA-ergic REM-off neurons, resulting in the disinhibition (i.e., activation) of GABA-ergic REM-on neurons. This is again inconsistent with actual experimental findings.13,14 INTRODUCTION One of the critical moments in sleep research came in 1953 with the discovery of REM sleep.1 Since then sleep has been regarded as an active process rather than merely the cessation of activity. Sleep is, by nature, not a static but a dynamic phenomenon, resulting from complex interactions of the behavior of central neurons, mainly in the hypothalamus and brainstem.2-8 Above all, the cyclic sequence of NREM and REM sleep, or the so-called ultradian rhythm, is a highly characteristic feature of sleep.9 The ultradian REM sleep rhythm is probed through REM-onset intervals, defined as the intervals between the onset of one REM period and the beginning of the next.10 For many years, the well-known reciprocal interaction model by McCarley and Hobson5,6,11 has been used to describe ultradian periodicity, the approximately 90-min sleep cycle, indicating that NREM-REM cycles are controlled by both cholinergic and monoaminergic neuronal systems. In the original reciprocal interaction model, cholinergic neurons act as REM-on cells and send excitatory synapses to monoaminergic neurons (REM-off cells) and to the cholinergic neurons themselves, while monoaminergic neurons act as REM-off cells and send inhibitory synapses to cholinergic neurons and also to the monoaminergic NON-REM SLEEP STAGE TRANSITIONS CONTROL REM SLEEP RHYTHM