Entrainment of Ultradian Oscillations in the Secretion of Insulin and Glucagon to the Nonrapid Eye Movement/ Rapid Eye Movement Sleep Rhythm in Humans*

The cause of ultradian oscillations in the secretion of glucagon and insulin with a period length between 70-140 min has been attributed to feedback mechanisms of glucose and insulin. Influences of the central nervous system on these ultradian glucagon and insulin os- cillations remained to be elucidated. In the present study on one occasion, concentrations of glucose, glucagon, insulin, and GH were determined at 15-min intervals from 2100-0700 h in 16 healthy subjects while they were infused with saline solution. On another occasion, concentrations of these hor- mones during nocturnal sleep were determined in 10 of these subjects while they were constantly infused with glucose (4.5 mg/kgmin). The order of the treatments (placebo vs. glucose) was balanced across subjects, and experiments were performed in a double blind manner. Significant glucagon and insulin peaks were determined by the peak detection algorithm Cluster. Sleep was recorded somnopolygraphi- tally. During the infusion of saline solution, glucagon concentrations showed spontaneous oscillations, with a mean periodicity of 107.9 -C 13.2 min. During the constant infusion of glucose, oscillations of similar periodicity

A NUMBER OF studies have shown that insulin secretion is subject to a multioscillatory process. Rapid oscillations in peripheral serum insulin concentrations, with an average periodicity of 9-15 min under basal conditions, have been identified in several animal species (1) and in humans (2). In addition, slower ultradian oscillations of insulin concentrations with a period length ranging between 70-140 min have been found in dogs (3-6) and in humans in response to meals (7,8), during continuous enteral nutrition (9), as well as during constant glucose infusion (10,111. Oscillations of similar periodicity have been found for glucagon concentrations in animals and humans (1,6,10,12,13). The mechanisms triggering the rapid and the slower oscillations of insulin and glucagon secretion are at present unclear. However, infusion of glucose has been found to decrease the relative amplitude of the rapid insulin pulses, but to increase the relative amplitude of the slower ultradian oscillations, suggesting that the two oscillatory modes have different origins (14). The slower ultradian oscillations of insulin secretion have been considered a consequence of the activity of the feedback loops linking glucose and insulin secretion, with a delay between increases in insulin secretion and the resulting reduction of glucose levels (15). However, such ultradian oscillations are not unique to insulin secretion. Ultradian oscillations with a periodicity of about 90 min have been demonstrated for secretion of a number of other hormones, such as cortisol, ACTH, GH, LH, TSH, PRL, and PRA (16). Most of these oscillations appeared to be linked to the nonrapid eye movement (non-REM)/rapid eye movement (REM) cycles of sleep, such that REM sleep coincides with decreasing plasma concentrations (i.e. decreased secretory activity), and non-REM sleep coincides with increased secretory activity of these hormones (17)(18)(19)(20).
This study aimed to evaluate a possible linkage between the ultradian secretory oscillations in concentrations of glucagon and insulin and the non-REM/REM sleep cycles during nocturnal sleep. Insulin concentrations were assessed during a constant infusion of glucose, as the amplitude of the oscillations in insulin levels increases during glucose stimulation (3). This condition also provided the opportunity to assess the effects of glucose and the associated elevation of insulin concentrations on the sleep pattern and sleep-associated GH secretion in healthy volunteers. In animal studies, insulin has been found to increase the amount of time spent in slow wave sleep (21-24) and to decrease the amount of time spent in REM sleep (21,25). Moreover, several studies have suggested an inhibitory effect of glucose and insulin on GH secretory activity (26,27 Table 1 summarizes results concerning sleep parameters. Total sleep time, percentage of sleep spent in different sleep stages, and latencies of the different sleep stages during infusion of placebo indicated normal undisturbed sleep. The infusion of 4.5 mg/kgmin glucose did not change any of the sleep parameters significantly. The mean length of the non-REM/REM cycles was 123.6 2 5.8 min during the infusion of glucose and 116.7 t 7.9 min on the placebo nights.

Sleep recordings
Subjects were, in general, unable to correctly identify the treatment they had received.
Blood glucose, insulin, glucagon, and GH concentrations Table 2 summarizes concentrations of blood glucose, insulin, glucagon, and GH averaged across total sleep time and separately for the first and second halves of sleep time during the infusion of placebo and 4.5 mg/kgmin glucose. During the placebo nights, insulin concentrations were very low (49.2 Ifr 4.2 pmol/L), and only l-3 significant insulin pulses/ night were detected by Cluster analysis (mean pulse amplitude, 16.8 + 4.2 pmol/L). On 4 of these nights no significant insulin peaks occurred. However, insulin concentrations were distinctly elevated during the infusion of glucose. On these occasions, 41 significant insulin pulses were detected overall, with a mean pulse interval of 110.1 i: 10.3 min. The mean pulse amplitude was 99.0 t 29.4 pmol/L, corresponding to 53.8% of the average concentration of insulin during these nights. In 5 subjects, C peptide levels were also measured. On the nights during constant glucose infusion, all significant insulin peaks were accompanied by significant C peptide peaks.
On the 16 placebo nights, 65 significant glucagon pulses were detected with a mean pulse interval of 107.9 2 13.2 min. The mean pulse amplitude was 15.8 + 3.5 rig/L, corresponding to 36.2% of the average glucagon concentration (44.9 -C 4.1 rig/L) during these nights. The infusion of glucose significantly suppressed glucagon secretion. On only 2 of these nights were significant glucagon pulses with a mean pulse amplitude of 1.9 ? 1.5 rig/L (5.7% of mean) detected.
Analysis of the relation between non-REM/REM sleep and insulin concentration was restricted to nights of glucose infusion, because during placebo infusion, insulin concentrations were minimal, and regular oscillations were not detectable. Conversely, oscillations in glucagon concentrations were assessed only for placebo nights, because during glucose infusion, glucagon concentrations were, as expected, minimal, and oscillations were not discernible.
Analysis of times of increasing and decreasing glucagon and insulin concentrations revealed that for both hormones, the time of decreasing concentrations (i.e. diminished secretory activity) exceeded that of increasing concentrations (i.e. activated secretion; Table 3). Therefore, statistical comparisons of the time of increasing VS. decreasing hormone concentrations during a particular sleep stage referred to percentages, with the time of respectively increasing and decreasing concentration set at 100%. This analysis indicated that sleep stages were not randomly distributed over times  of increasing and decreasing insulin and glucagon concentrations. REM sleep significantly coincided with decreasing concentrations of insulin and glucagon, i.e. a diminished synchronous secretory activity of the a-and p-cells of the pancreas and non-REM sleep significantly coincided with increasing concentrations of both hormones, i.e. an increased synchronous secretory activity of the endocrine pancreas. Moreover, the peak onset of 40 (representing 97.6%) of the 41 si ( 3 nificant insulin peaks fell into phases of non-REM sleep = 7.32; P < 0.01; 2 testing is based on a comparison of apriori expected numbers of pulses, i.e. -80% for non-REM sleep, and actual numbers). Similarly, the onset of 60 (92.3%) of the 65 significant glucagon pulses fell into a phase of non-REM sleep (2 = 5.23; P < 0.02; see Fig. 1 for examples of individual profiles; Fig. 2). Although periods of intermittent wakefulness appeared to be generally short, these periods were significantly associated with decreasing concentrations of insulin.
To further evaluate the temporal relationship between the secretory activities of the (Y-and p-cells of the pancreatic islets and central nervous system sleep processes, the distribution of onsets of secretory pulses across the non-REM/REM cycle was determined for glucagon and insulin (Fig. 2). For this purpose, non-REM/REM cycles were divided into 10 periods of equal length, each representing 10% of a cycle. As indicated in Fig. 2, the onsets of secretory pulses were not evenly distributed across the sleep cycle. For both hormones they were almost absent during REM sleep, but the greatest number of secretory pulses started in the beginning of the non-REM epoch of a sleep cycle [glucagon: 2 = 8.69; P < 0.01; insulin: 2 = 15.76; P < 0.001; for the comparison of REM sleep (representing about 20% of a non-REM/REM sleep cycle) and the non-REM sleep phase immediately after REM sleep (which also comprised 20% of a sleep cycle)].
Average GH concentrations during total sleep time and :::: ::j:j:j:j: ::: ZIOP during the first and second halves of sleep time did not significantly differ between the glucose and placebo conditions (Table 2). Also, peak latency (placebo, 86.3 '-' 12.3 min; glucose, 83.3 t 10.3 min) and peak values (placebo, 9.8 + 1.2 Fg/L; glucose, 10.7 2 1.9 Fg/L) for the rise in GH concentrations after sleep onset were nearly identical during the infusions of placebo and glucose. To examine whether the increase in GH secretion at sleep onset had a phase-setting effect on the synchronization of insulin oscillations with non-REM/REM sleep cycles, mean profiles of blood glucose and serum insulin, glucagon, and GH concentrations were calculated. These profiles, shown in Fig. 3 pancreatic a-cells. Stimulation of insulin secretion by constant infusion of glucose, as expected, invoked ultradian oscillations of serum insulin levels with a mean periodicity (110.1 min) comparable to that reported in several previous studies (7)(8)(9)(10)(11)30). Notably, periodicity of oscillations of insulin concentrations (during stimulation with glucose) and glucagon concentrations (under basal condition) were almost identical.
C Peptide concentrations were measured in five of the subjects. Cluster indicated that on these nights all significant insulin peaks were accompanied by significant peaks in C peptide concentrations. This finding is in line with several foregoing studies investigating the slower ultradian oscillations of insulin concentrations in humans in response to meals (7), during continuous enteral nutrition (91, as well as during constant glucose infusion (10, 11). In all of those situations, oscillations of insulin and C peptide concentrations ran in parallel. The results together suggest these oscillations to reflect a synchronous secretory activity of the endocrine pancreas rather than periodic changes in the clearance rate of insulin. It has been suggested that the ultradian oscillations in insulin concentrations are an inherent feature of the insulinglucose feedback mechanism, with a delay between increases in insulin secretion and the resulting reduction of glucose levels (15,311. However, if these oscillations were solely due to a delay in the feedback effect of glucose on insulin secretion, insulin peaks during a constant rate infusion of glucose would have been expected to be randomly distributed across episodes of non-REM and REM sleep. Actually, however, nearly all insulin secretory peaks originated during non-REM sleep, preferentially in the beginning of these epochs. Also, increasing concentrations of insulin were significantly associated with phases of non-REM sleep, indicating that the synchronized secretory activity of p-cells is linked to the presence of non-REM sleep. By contrast, decreasing insulin concentrations, reflecting a diminished secretory activity, significantly coincided with the occurrence of REM sleep. JCE & M . 1996 Volt31 . No 4 The ultradian oscillations in glucagon concentrations under basal conditions also were not randomly distributed across the night. Again, nearly all glucagon secretory peaks originated during non-REM sleep, preferentially in the beginning of these phases. In contrast, REM sleep significantly coincided with decreasing glucagon concentrations. Because under fasting baseline conditions, oscillations in blood glucose concentrations were virtually absent, it is unlikely that the oscillations of glucagon secretion reflect delays in the feedback effect linking glucose and glucagon. Rather than a mediation due to peripheral feedback interactions, the close temporal link between central nervous rhythms of non-REM/REM sleep processes and, respectively, insulin and glucagon secretory epochs strongly suggest a modulatory influence of the central nervous system on the generation of oscillations of pancreatic secretory activity. A sleep-associated pattern of glucagon and insulin release, as demonstrated here, with decreasing concentrations during REM sleep and increasing concentrations during non-REM sleep, adds to findings indicating secretory patterns similarly sleep dependent for a number of other hormones, such as cortisol, ACTH,GH,LH,TSH,PRL,.
It should be emphasized that not every episode of non-REM sleep was associated with secretory activation of the endocrine pancreas. Thus, rather than a direct stimulatory effect of non-REM sleep on the secretory activity of the endocrine pancreas, such a temporal pattern indicates a permissive function of this sleep stage. By contrast, during REM sleep, an inhibitory central nervous system influence appears to diminish pancreatic activity. The sudden increase in the number of secretory pulses originating in the very beginning of a non-REM sleep epoch suggests inhibition to be active during the preceding REM sleep, which is released immediately upon the occurrence of subsequent non-REM sleep.
In principle, a temporal association between the non-REM/REM sleep cycle and secretory activity of the endocrine pancreas could also reflect an afferent humoral influence on central nervous system sleep processes (32, 33). Increasing concentrations of glucagon and insulin could suppress REM sleep and vice versu. However, this view is not supported by the fact that the infusion of glucose with a suppressing influence on glucagon secretion and an elevating influence on insulin secretion left central nervous system sleep processes virtually unaffected.
Thus, the present data appear to be more consistent with the idea of an efferent effect of central nervous system mechanisms coordinating both non-REM/REM sleep cycles and endocrine pancreatic secretory activity. Such an influence could be elicited via humoral factors (such as GH) or via efferent peripheral nerves (3438). However, &radian oscillations in insulin levels have been shown to persist after pancreas transplantation, suggesting that the oscillations do not essentially depend on intact neuronal input from sympathetic or parasympathetic centers (11,30,39). Nevertheless, although these findings rule out that generation of ultradian oscillations in insulin and glucagon concentrations is mediated via efferent nerves to the endocrine pancreas, they do not exclude that such innervation is involved in synchronizing pancreatic secretory activity with central nervous system sleep cycles.
In patients with noninsulin-dependent diabetes mellitus, the ultradian oscillations of glucose-stimulated insulin secretion have been shown to be irregular and reduced in amplitude (40), suggesting a physiological relevance of these fluctuations. Previous findings indicated that these oscillations may enhance the efficiency and stability of glucose disposal (5). Therefore, in light of the present findings, further examination of the association between central nervous system sleep processes and secretory activity of the endocrine pancreas in patients with noninsulin-dependent diabetes mellitus as well as in patients with sleep disturbances would be beneficial.
Besides a temporal association between sleep processes and the secretory activity of the endocrine pancreas, this study also examined the effect of glucose and the associated increase in serum insulin levels on sleep and sleep-related GH secretion. Glucose may influence brain activity via elevation of insulin, as this hormone has rapid access to the brain (41). In rats, peripheral (21,25,42) or intracerebroventricular (23) infusion of insulin increased the duration of SWS. Conversely, suppression of SWS was observed when endogenous insulin was neutralized by insulin antiserum (231, decreased by acarbose treatment (24), or depleted in streptozotocin-induced diabetes in rats (22). In some studies, a decrease (21,25,43) or an increase (42) in REM sleep after iv or ip administration of insulin was observed. In the present study in humans, none of the sleep stages was significantly affected by the increased glucose and serum insulin levels when separately analyzed for the first and second halves of sleep time. Possibly, the glucose-induced elevation of peripheral insulin levels was too small to develop a discernible influence on central nervous system activity during sleep. In line with this view is the observation that sleep-related increases in GH concentrations remained unchanged, whereas in vitro, an insulin-induced inhibition of GH secretion has been reported (26,27).
In summary, slightly elevated glucose and insulin concentrations did not affect sleep in humans. Yet, the present results demonstrate a temporal link between central nervous system sleep processes and the secretory activity of CP and p-cells, with this activity being inhibited during REM sleep and increased during non-REM sleep. This association between non-REM/REM sleep cycles and &radian fluctuations of glucagon and insulin secretory activity reflects a modulatory influence of brain structures involved in sleep regulation on pancreatic endocrine secretory activity. It is of interest to determine to what extent this temporal relation is disturbed under pathological conditions, e.g. in patients with noninsulin-dependent diabetes mellitus and sleep disturbances.