The REM Cycle is a Sleep-Dependent Rhythm

Transcription

1 Sleep, 2(3): Raven Press, New York The REM Cycle is a Sleep-Dependent Rhythm L. C. Johnson Naval Health Research Center, San Diego, California Summary: Two studies, using data from fragmented sleep studies from three sleep laboratories, are reviewed. These studies indicate that the REM cycle is primarily governed by a sleep-dependent oscillator. These data, however, do not rule out the potential influence of endogenous or environmental variables as factors influencing the REM cycle. Further, acceptance of the REM cycle as a sleep-dependent rhythm does not lead to the denial of the basic rest-activity cycle (BRA C) proposed by Kleitman. It is questioned, however, whether the BRAC measured during waking is an extension of the REM cycle recorded during sleep. Key Words: Sleep-REM cycle-ultradian rhythms. Some may wonder why the title of this paper does not read' 'the REM - NREM cycle is a sleep-dependent rhythm." While REM - NREM is the cycle of sleep, our goal was to see if the REM cycle continued during awake activity and thus in the absence of intervening NREM sleep. The continuation of the min REM cycle throughout the 24 hr period, regardless of state, is what we refer to as the sleep-independent model. Though we found little support for this sleepindependent model, the focus of our study was on the REM cycle and not on the REM-NREM cycle. For clarity of discussion, I have positively stated our position in the title of this paper. This rhythm, however, is not imperturbable. Our own data clearly indicate that fragmentation of sleep, which leads to shortened REM episodes, reduces the length of the REM cycle (Moses et ai., 1978). Also, as indicated by the varying positions taken by participants in this workshop and the numerous factors that have been advanced as governors of REM sleep, it is highly unlikely that a single explanation would account for all the variance. Globus (1966), in a study of daytime nap sleep, found that REM onsets tended to cluster around certain clock times, relatively independent of the time of sleep onset, and concluded that the occurrence of REM sleep is a function of the time of day. Schulz et al. (1975) reported a gradual forward shift in REM onset time from Accepted for publication April Address reprint requests to Dr. Johnson at Naval Health Research, Bldg. 36-4, Naval Regional Medical Center, San Diego, California

2 300 L. C. JOHNSON night to night, indicating that the occurrence of REM is not dependent on time of day, but is rather a function of an ongoing basic rest-activity cycle (BRAC), with a cycle length that is not a submultiple of 24 hr. Brezinovli (1974) found that the duration of the REM cycle was sensitive to factors within sleep itself, such as the amount of stages 3 and 4 within the cycle, suggesting that the timing of the REM cycle is not a simple manifestation of an ongoing BRAC. In a study of acute shifts in the sleep-wake cycle, Taub and Berger (1973) interpreted their finding of shortened REM cycles when-sleep onset was delayed past 0200 hr as a reflection of the circadian influence on the REM cycle. Before summarizing our results and conclusions, I would like to clarify our position vis a vis Kleitman's BRAC hypothesis (1963). In its simplest form, this hypothesis states that there is an ultradian oscillation in certain neurophysiological functions that continues throughout waking and sleeping. We have never stated that a BRAC is nonexistent. We maintain that there is more than one oscillatory system governing the rhythms seen during waking and sleeping. It is quite possible that certain neurophysiological functions exhibit ultradian rhythms, regardless of the state of consciousness. But we have presented evidence (Moses et ai., 1977, 1978) that at least one-rem sleep-is governed by a state-dependent oscillator. Our data come from nap studies from three different sleep laboratories. METHODS The reader is referred to the original articles by Moses et al. (1977, 1978) for details as to subjects and procedures. Briefly, the sleep of a total of 25 subjects from three independent sleep laboratories was analyzed. In addition, following a base-line night, 8 male Naval School of Health Sciences students, aged years, underwent a 60 min sleep-160 min wake schedule for 40 hr at the Naval Health Research Center (NHRC), San Diego. The 40 hr period was followed by 8 hr of recovery sleep (Moses et ai., 1975a). Ten students, 6 men and 4 women, aged 17-21, were studied at the Stanford University School of Medicine. Following 2 base-line nights, the subjects were placed on a 30 min sleep-60 min wake schedule for 5 days. The altered schedule was followed by 2-4 recovery nights, on which the subjects were allowed to sleep ad lib (Carskadon and Dement, 1975). Seven subjects, 6 men and 1 woman, aged 23-40, participated in a study at the Clinical Research Center of Montefiore Hospital. The subjects slept normally for 7 base-line nights and then underwent a 60 min sleep-120 min wake schedule for 10 days. Seven recovery nights followed the altered schedule (Weitzman et ai., 1974). All sleep recordings were scored according to Rechtschaffen and Kales (1968). In the analysis concerned with severe sleep fragmentation, the data from two additional groups of subjects from a previous study were used (Moses et ai., 1975b). These subjects underwent 3 consecutive nights of either slow-wave deprivation (S group, n = 7) or REM deprivation (R group, n = 7). RESULTS Altered sleep-wake schedules, such as in napping or multiphase sleep, provide an opportunity to test the sleep-dependent hypothesis. According to the BRAC Sleep, Vol. 2, No

3 THE REM CYCLE 301 hypothesis, REM should occur at approximately equidistant times throughout the altered schedule period. This, however, was not the case. The total time (wake plus sleep time) between REM onsets was extremely variable, both within and between subjects, ranging from 21 min to more than 24 hr. Although the amount of REM in all these studies was greatest between 0200 and 1000 hr, REM onset had no consistent relationship to time of day. REM onset appeared to be related to the amount of sleep since the last REM onset. If the REM cycle clock is sleep-dependent, stage REM onset should occur only after min of sleep has accrued since the last REM onset, regardless of the amount of wake time between or within the scheduled periods. Figure 1 illustrates the sleep-dependent and sleep-independent models. According to the sleep-dependent hypothesis, the REM cycle length (RCL) in naps should not differ significantly from the base-line night RCL. All analyses began with the first REM onset of the sleep episode. RCL was measured from the beginning of one REM episode to the beginning of the next REM episode. A REM episode was a series of consecutive sequences of REM sleep separated by less than 15 min of NREM sleep, movement time, or waking. We computed the average difference between nap and base-line RCL for each subject, and the overall average difference was tested for significant (0.05 level) deviation from zero with a t-test. None of the mean differences differed significantly from zero (Table 1). Since publication of the data in Table 1, we have collected data on 4 subjects from a subsequent study of fragmented sleep. Following a full night of sleep, the subjects maintained a continuous 44 hr testing schedule, followed by two 2 hr naps taken 8 hr apart. They were then allowed a full night of recovery sleep. When the sleep time obtained during base line, naps, and recovery was combined and viewed as one continuous sleep period, the average REM-to-REM interval was SLEE P - DEPENDENT MINUTES, SLEEP - INDEPENDENT MINUTES, I <.'.' FIG. 1. Hypothetical examples of REM onsets occurring during a 30-min sleep/60-min wake schedule. In the s'leepdependent model, total sleep time between REM onsets is 90 min (total real time is 270 min). The "c1ock" governing the time interval between REM onsets runs only during sleep; it stops on awakening and resumes at the next sleep onset. In the sleep-independent model, total real time, between REM onsets is 90 min (total sleep time is 30 min). The "clock" runs during both sleep and waking. Studies of nap sleep from three independent sleep laboratories support the sleep-dependent model. (From Moses et al., 1977.) '

4 302 L. C. JOHNSON TABLE 1. REM cycle length for nap and base-line sleep conditions REM cycle length Sleep San Diego Stanford Montefiore condition (n = 8) (n = 10) (n = 7) Nap %.6 ± ± ± 30 Base line 97.1 ± ± ± 12 Nap minus base line" ± ± ± 31 a None of the differences between nap and base-line RCL are statistically significant. Values are means ± SO ± 11 min, well within the range of REM intervals in normal nocturnal monophasic sleep. While an examination of REM intervals gives an indication of cyclicity, it does not address the question of rhythmicity per se. To determine rhythmicity, an autocorrelation analysis developed by Naitoh et ai. (1973) was applied to the base line, compressed nap sleep, and recovery sleep of the three groups of nap subjects. This method estimates the period of the REM cycle and yields r2, the variance in REM cycle periodicity accounted for by a cosine wave and, thus, a measure of the strength of that periodicity. For the details of this analysis, see Moses et ai. (1978). The average base line, nap, recovery sleep (min), and r2 values are presented in Table 2. The r2 values were compared with t-tests for correlated means. Except for the Stanford group, the length of the nap REM cycle did not change significantly from base line during the compressed nap sleep. The overall decrease in the strength of rhythmicity during nap sleep (Table 2) reflects the increased variability in REM cycle length during the altered schedules. These nap r2 values, however, were significantly larger than those obtained from a random distribution of sleep stages. When the stages were randomized, all the r2 values fell to near zero. There were no consistent trends across nap segments in the Stanford and Montefiore groups; the RCL and r2 values remained relatively stable throughout the schedule. There were too few REM cycles in the NHRC data for trend analysis. An illustration of the REM cycle rhythmicity in one subject for the base line, compressed nap, and recovery night is illustrated in Fig. 2. Since the Stanford group's sleep was relatively more fragmented than the other two groups (twice as many sleep periods per 24 hr), we explored the possibility that RCL was influenced by the degree of sleep fragmentation. In this analysis, we used two groups of subjects from our previous study on sleep stage deprivation (Moses et ai., 1975b) to compare to the nap groups. We chose these subjects for comparison because their sleep was highly fragmented: an average of 55 awakenings per night for the 7 subjects of the R group and 44 awakenings for the 7 subjects of the S group (this between-group difference was not significant).

5 TABLE 2. REM cycle length and r2 in base-line, nap, and recovery sleep Sleep/wake Duration of Mean REM cycle lengths (min) cycle altered Groups (min) schedule Base line Nap Recovery Base line San Diego 60/ hr ± ± ± ± 0.28 (n = 6'1' Stanford 30/60 5 days ± ± ± ± 0.15 (n = 10) Montefiore 60/ days ± ± II ± II ± 0.09 (n = 7) Significant difference from base line, p < b Two NHRC subjects had only one REM cycle during the naps and thus were excluded in this analysis. Values are means ± SO. Mean r2. Nap ± ± ± 0.06 Recovery tl'j ± 0.23 ::r.:, tl'j ± 0.11 () ± 0.09 t'-< tl'j "..::.w c v",

6 304 L. C. JOHNSON BASE-LINE NIGHT PERIOD: 9B MIN,2 = _ Z 0 :::.. 0 u 0 ; C _ COMPRESSED NAPS SEGMENT PERIOD: 92 MIN,2 =.785 Z Q.. 0 :0 C 1.0 RECOVERY NIGHT PERIOD: 86 MIN,2 = _ I I 20 60, I I I 220 LAG IN MINUTES FIG. 2. Example of REM cycle period (length) rhythmicity in one subject under three experimental conditions (Montefiore group). Top: Total elapsed time, 301 min; total sleep time, 300 min; time of day range, ; local time of REM onsets: 2354,0138,0310,0440; average sleep time between REM onsets, 95 ± 8 min. Middle: Total elapsed time, 1,423 min (23 hr, 43 min); total sleep time, 300 min obtained in seven I-hr naps at 0000,0300,0600,0900, 1200, 1500, 1800; time of day range, ; local time of REM onsets, 0055, 0628,1233,1853; average sleep time between REM onsets, 87 ± 12 min. Bottom: Total elapsed time, 301 min; total sleep time: 300 min; time of day range, ; local time of REM onsets, 0006, 0113, 0239, 0401; average sleep time between REM onsets, 80 ± 9 min. (From Moses et ai., 1978.) The autocorrelation analysis was applied to the Rand S groups to determine the length of the REM cycle and the strength of rhythmicity during base line, sleep stage deprivation, and recovery. REM cycle analysis was possible for the R group because the deprivation procedure, while severely shortening each REM episode, did not completely eliminate REM sleep. As with the three nap groups, all wake time within sleep sessions was subtracted. REM deprivation significantly reduced the RCL from a base-line average of98 min to an average of73 min, and the r2 was reduced from 0.77 base line to 0.34 during REM deprivation. Furthermore, the r2 during REM deprivation was significantly lower than the average r2 in the three nap groups. Despite the frequent awakenings, slow wave sleep deprivation had no effect either on REM cycle length or r2.

7 THE REM CYCLE 305 Recovery Sleep Inspection of Table 2 shows that despite the severe alteration of the normal sleep-wake pattern during the nap schedules, RCL and rhythmicity on recovery were unchanged from base-line levels. This is of particular interest in the Stanford group, whose REM cycles were shortened during the naps and who were allowed to sleep ad lib during the recovery; the average total sleep time of the Stanford subjects on the first recovery night was 838 min. DISCUSSION A clear REM cycle was seen in all three studies, and with the exception of the Stanford group, the length of the REM cycle in compressed nap sleep did not change. The reduced nap r2 value, however, raises the question of what r2 value can be considered as indicative of significant rhythmicity. While the significant reduction in nap r2 for two of the three groups reflects increased validity in the nap RCL, we presently lack an adequate method of testing these r2 values for significance. Our analyses using randomized sleep stages, however, yielded near-zero r2 values, demonstrating that the nap data were nonrandom. No biological rhythm is impervious to external factors, and we have never implied, nor meant to imply, that the REM cycle is an exception. But for a biological event to be considered cyclic, it must repeat itself in time in some sort of regular fashion. If the BRAC concept is to be useful, it has to be capable of predicting the approximate timing of certain events. One interpretation of the BRAC is that REM should occur at approximately equidistant time intervals throughout the altered sleep-wake schedule. Our data clearly do not support this interpretation. Finally, there is some inconsistency in the interpretation of waking rhythms and REM sleep as manifestations of a BRAC. Stage REM during normal sleep appears to be a state of heightened physiological arousal, relative to NREM sleep, as shown by electroencephalographic changes, rapid eye movements, increased and more variable heart rate, increased body motility, and lowered arousal threshold. Some studies of waking rhythms have shown cyclic occurrences of periods of heightened arousal such as increased oral activity and rapid eye movements that are perhaps consistent with REM-like processes. Other studies have reported cyclic periods during waking of lowered arousal, expressed as increased test errors, ocular quiescence, and increased fantasy and daydreaming. It is not yet clear whether the periods of heightened arousal or lowered arousal are supposed to be the waking counterparts of REM and NREM episodes during sleep. Furthermore, to assume a continuation of the REM - NREM cycle throughout the 24 hr period would assume that the REM cycle during the awake state would have a functional significance similar to that during sleep. Many would question such an assumption, and the diversity in the physiology and behavioral states in the awake ultradian rhythms that have been identified would support skepticism

8 306 L. C. JOHNSON as to the similarity of sleep and awake REM cycles. In a very thorough review of the issue of BRAC and ultradian rhythms, Kripke (1974) raised the question as to whether a BRAC concept is appropriate for human adults and pointed out the need for the kind of evidence we have tried to provide: "More evidence is needed to establish that the phenomena [ultradian rhythms in sleep and wakefulness] result from a common oscillatory system, an inference based presently only on the resemblance of the cycles described." As we have always maintained, we do not feel that our sleep-dependent model is adequate for the complete understanding and prediction of REM cycling. It is obvious that a more comprehensive model is needed to account for the rhythms during waking and to clarify the relationship, if any, of such rhythms to the REM cycle during sleep. Our results do not preclude a relationship between awake and the sleep ultradian rhythms, but our data strongly suggest that REM processes are processes of sleep. ACKNOWLEDGMENT This research was supported by the Naval Medical Research and Development Command, Department of the Navy, under Work Unit MF The views presented in this paper are those of the author. No endorsement by the Department of the Navy has been given or should be inferred. REFERENCES Bfezinova V. Sleep cycle content and sleep cycle duration. Electroencephalogr Clin Neurophysiol 36: , Carskadon MA and Dement WC. Sleep studies on a 90-minute day. Electroencephalogr Clin Neurophysiol 39: , Globus GG. Rapid eye movement cycle in real time. Arch Gen Psychiatry 15: , Kleitman N. Sleep and Wakefulness. University of Chicago Press, Chicago, Kripke DF. Ultradian rhythms in sleep and wakefulness. In: ED Weitzman (Ed), Advances in Sleep Research, Vol 1, Spectrum, New York, 1974, pp Moses JM, Hord DJ, Lubin A, Johnson LC, and Naitoh P. Dynamics of nap sleep during a 40 hour period. Electroencephalogr Clin Neurophysiol 39: , 1975a. Moses JM, Johnson LC, Naitoh P, and Lubin A. Sleep stage deprivation and total sleep loss: Effects on sleep behavior. Psychophysiology 12: , 1975b. Moses J, Lubin A, Johnson LC, and Naitoh P. Rapid eye movement cycle is a sleep-dependent rhythm. Nature 265: , Moses J, Naitoh P, and Johnson LC. The REM cycle in altered sleep/wake schedules. Psychophysiology 15: , Naitoh P, Johnson LC, Lubin A, and Nute C. Computer extraction of an ultradian cycle in sleep from manually scored sleep stages. Int J Chronobiol 1: , Rechtschaffen A and Kales A (Eds). A Manual of Standardized Terminology, Techniques and Scoring System for Sleep Stages of Human Subjects. UCLA Brain Information ServicelBrain Research Institute, Los Angeles, Schulz H, Dirlich G, and Zulley J. Phase shift in the REM sleep rhythm. Pj/uegers Arch 358: , Taub JM and Berger RJ. Sleep stage patterns associated with acute shifts in the sleep-wakefulness cycle. Electroencephalogr Clin Neurophysiol 35: , Weitzman ED, Nogeire C, Perlow M, Fukushima D, Sassin J, McGregor P, Gallagher TF, and Hellman L. Effects of a prolonged 3-hour sleep-wake cycle on sleep stages, plasma cortisol, growth hormone and body temperature in man. J Clin Endocrinol 38: , 1974.

9 THE REM CYCLE 307 DISCUSSION The conclusion that "'REM processes are processes of sleep" was questioned, as the data related only to the REM - NREM cycle and did not address the circadian rhythm of REM sleep. Dr. Johnson agreed that his work was related to the ultradian REM cycle and that when the data were analyzed in terms of the time of day, there was more REM sleep in the morning than in the afternoon. However, the REM cycle period length remained the same. He hypothesized that the phenomenon of sleep onset REM episodes, raised by Dr. Weitzman, probably occurred only when a sufficient amount of sleep time had elapsed since the onset of the previous REM episode. He felt this explained the very short REM sleep latencies observed in some naps and longer latencies in others. Dr. Roffwarg emphasized the similarity of results in the REM deprivation experiments and questioned whether the increased frequency of the REM - NREM cycle observed was related to REM deprivation. Dr. Spielman mentioned that he REM sleep deprived several patients with the narcolepsy-cataplexy syndrome. The awakenings required for REM deprivation continued to cluster throughout the night, at the time of expected REM episodes. He suggested this might be a good model to study the daytime period length of the Basic Rest-Activity Cycle (BRAC).