Standard procedures to obtain meaningful data about the sleep/wake behavior of laboratory animals generally include recordings of the EEG and the EMG. The techniques of chronic EEG and EMG electrode implantations are routinely utilized in many laboratories. Sophisticated algorithms and the experience of the researchers in the handling of data being obtained in basic research with laboratory animals allow differentiation between wakefulness, NREM sleep and REM sleep. In the present study we provide a complementary element of sleep data analysis, based on a highly standardized evaluation of eye movements (EMMA). We found that REM density recorded across three nights showed remarkable intra-individual trait-like stability. REM density was significantly higher during REM sleep than during NREM sleep and this sleep-stage specific distribution corresponds with human REM density. While REM density during REM sleep showed an ultradian course, during NREM sleep REM density peaked at the beginning of the dark period. The majority of single REM and REM bursts were associated with micro arousals during NREMS, but not during REMS.
The present configuration of the EOG electrodes was designed to record eye muscle activities per se with the electrode placed between the musculus rectus lateralis and musculus rectus superior. The gold wires used in our study could be replaced by multiple wires such as those used to record with tetrodes. Single wires could then be brought individually to the musculus rectus superior and the left and right musculus rectus lateralis. Although not part of our study, this extension of the present electrode configuration would allow differentiating between dorsal and lateral eye movements. Provided that the morphological components of the recorded potentials from separated dorsal and lateral eye movements are similar to the potentials we obtained, the EMMA algorithm could be applied with only minor adaptations in threshold detection. We doubt that the musculus rectus inferior could be equipped solidly with electrodes to record ventral eye movements with the present approach.
We found REM not only during REM sleep but also during NREM sleep in mice. One possible concern could therefore be that REM during NREM sleep are due to the EEG signals, in particular the large slow potentials, contaminating the EOG signal. We judge this as unlikely for several reasons. First, we designed a wavelet filter that eliminated lower frequencies (0.25 to 2 Hz) from the EOG signal before REM detection. Second, EOG electrodes where fully isolated with coating against ohmic contact on the skull's surface, leaving only the non-isolated electrode tip embedded fully in the eye muscle without contact to the bone. Third, detection of REM was based on the concept of singularities i.e. discontinuities in the signal that are not continuous on a particular derivative. To be considered a REM candidate, a local maximum at the finest scale (fastest frequency) of the wavelet transform modulus maxima was a necessary (though not sufficient) condition. Therefore only high frequency noise can trigger false detection. We would expect slow EEG components leaking into the EOG signal to also affect neighboring signal points and our threshold based on the standard deviation of the local wavelet transform modulus maxima was chosen to protect against this possibility.
Given that the EOG electrodes are chronically implanted together with the EEG and EMG electrodes, any experimental recording situation in freely moving animals is conceivable. We applied the present electrode configurations for several weeks in all of our animals and detected neither any behavioral irritations, such as scratching the eye lids, nor inflamed tissue post mortem.
One possible limitation is the low sampling rate of 64 Hz in the present study. We can not exclude the existence of REMs with higher frequencies (> 29 Hz) and therefore our results may be an underestimation of the frequency of REMs during sleep in mice. Future studies with higher sampling rates are needed to clarify this issue.
In the present study, single REM and REM bursts were associated with micro-arousals, which we defined as transient EMG increases during NREMS, but not during REMS. The definition found in the literature for arousals is inconsistent. Single events of arousals within a 4 s epoch or a waking episode lasting less than 16 seconds were defined as sleep fragmentations and separated from clear transitions between vigilance states which were longer than 16 seconds . Other definition criteria for clear changes to another behavioral state comprised eight or more consecutive 4s epochs scored as one behavioral state, which were followed by eight or more consecutive 4s epochs scored as a different behavioral state . In other studies, micro-arousals were defined as events in NREM sleep lasting 5-15s and including a drop of at least 50% in the EEG power in the δ band . Short, single episodes in NREMS with increased power in the θ band and EMG activation were counted as wake . In the present study, transient EMG activity (< 4s) during sleep was taken to define micro arousals. When evaluating micro-arousals we found EMG increases both with and without changes in the EEG. Both types of micro-arousals were frequently accompanied with either single REM or REM bursts (Figure 5 gives examples for both events). In addition to what is defined as a state of arousal in the literature, the present data suggest differentiating micro-arousals in more detail. We would suggest reserving the concept of "micro-arousal" for events less than 8 seconds. Within this time frame, micro arousals could be separated between micro-arousal with EMG changes and micro arousal without EMG changes. Additionally, micro-arousals could be divided into events with single REM and with REM bursts. The occurrence of REM in the majority of micro-arousals in the present study may also support the need to implement the eye movements per se as a general feature in the detection and declaration of micro-arousals. Since arousals are deeply involved in the pathophysiology of human sleep disorders , the establishment of well-defined micro-arousals in mice may also help to better understand the origin of increased micro-arousals in particular sleep disorders and to characterize their sleep more fully.
Rapid eye movements, which gave name to the prominent vigilance state of REM sleep in humans, had never really been in the focus of in-depth investigations in animals, although, general references to eye movements during sleep and wakefulness were obtained in several non-mammalian species and mammalian species (please refer to the introduction). This obviously fundamental and phylogenetically diverse component of REM sleep seems to be predestined to serve as a tool in translational approaches between basic and clinical research.
In humans REM density is an important parameter that has been associated with sleep satiety , learning and memory [46–48] and psychiatric disorders . Importantly, increased REM density has been proposed to be an endophenotype for depression [54, 55] and is sensitive to treatment with antidepressants. Characterization of sleep in mouse models of depression has shown REM sleep facilitation and increased sleep fragmentation that resemble endophenotypes of depressed patients [84–88].
Analysis of REM density is expected to complement and significantly extend sleep characterization in these mouse models. Importantly, we found a high intra-individual stability of REM density across several days with almost perfect trait-like stability that was substantially higher than the stability of REM sleep proportion. REM density might therefore be better suited to characterize phenotypic variations in sleep patterns in mice.