The 24-h growth hormone rhythm in men: sleep and circadian influences questioned

Growth hormone (GH), a major regulator of protein anabolism and tissue growth, is secreted in a series of pulses throughout the 24 h. Under usual night sleep and daytime wake routine, a major pulse occurs shortly after sleep onset in temporal association with the first slow-wave sleep (SWS) episode (Sassin et al., 1969; Takahashi et al., 1968). In normal young men, this pulse accounts for about 50% of the daily GH production and is consistently correlated with SWS duration (Van Cauter et al., 1992) and delta waves (0.5–3.5 Hz of the spectral power of the electroencephalogram) (Gronfier et al., 1996). It is widely accepted that the GH rhythm is dependent mainly on sleep processes with a weak influence of the endogenous circadian system which was best identified in a study using repeated injections of GH-releasing hormone (Jafféet al., 1995). This concept however has been recently questioned in sleep-deprived persons in whom the diminution of the sleep-related pulse is compensated for during the day, so that the amount of GH secreted over 24 h is similar whether or not a person has slept during the night (Brandenberger et al., 2000). In addition, during chronic sleep debt, the GH profiles are biphasic (Spiegel et al., 2000) which does not conform to the predictions from previous sleep deprivation studies (reviewed in Van Cauter et al., 1998), demonstrating low secretion during forced waking periods followed by a secretion rebound during recovery sleep.

In light of these findings, we examined simultaneous 24-h rhythms of GH and melatonin in permanent night workers who represent a model challenging sleep and circadian processes. We have previously shown that night workers display an incomplete adjustment of their GH rhythms (Weibel et al., 1997a) and a highly variable adaptation of their melatonin rhythms (Weibel et al., 1997b). In the present study, we determined whether the blunting of the sleep-related pulse is compensated for during waking periods, as found in sleep-deprived persons. Then, we examined whether the extent of the shift in melatonin onset could predict the degree of GH adaptation.

A control group of 10 healthy male day-active subjects (23–32 years) with a regular sleep-wake schedule were studied during their usual 24-h cycle (from 23:00 to 23:00 hours) with nocturnal sleep from 23:00 to 07:00 hours. Over the preceding weeks, the subjects had kept normal meal and sleep routines, and had experienced no time shift or sleep deprivation. Eleven healthy male night workers (25–33 years) who had been working on a permanent night shift schedule at least the previous 2 years with four to five consecutive night shifts per week, were selected. All the night workers claimed high work satisfaction without any apparent behavioral problems. Night workers were studied immediately after their last night work of the week for 24 h (from 23:00 to 23:00 h) with day sleep from 07:00 to 15:00 hours. No subject was taking any medication. Before their enrollment in the study, they underwent a medical examination. They gave their written informed consent to participate in the study. The protocol was approved by the local Ethics Committee.

The day-active controls entered the laboratory for a continuous period of about 48 h beginning with an habituation session of night sleep. At 16:00 hours, before a subject entered the sleep room, a catheter was inserted into an antecubital vein. The subjects were then studied for 24 h (from 23:00 to 23:00 hours) under their usual sleep-wake cycle with normal sleep from 23:00 to 07:00 hours.

Night workers entered the laboratory for a period of about 40 h after their last night shift of the week. They came to the laboratory in the early morning directly after their night of work and slept from 07:00 to 15:00 hours as a habituation session. At 16:00 hours, a catheter was inserted into an antecubital vein. They were then monitored for 24 h from 23:00 to 23:00 hours under their usual 24-h sleep-wake cycle with day-sleep from 07:00 to 15:00 hours.

The experiments were carried out in soundproof air-conditioned sleep rooms. To avoid the influence of repeated meal intake, the subjects received continuous enteral nutrition, which began at 16:00 hours (Sondalis ISO; Sopharga, Puteaux, France; 50% carbohydrate, 35% fat and 15% protein, 378 kJ h⁻¹). The subjects remained supine throughout the experiment. When awake, they were maintained in dim light (<100 lx) and read or watched television. During the night without sleep, the night workers were kept under continuous surveillance and conversed with members of the laboratory staff. Special attention was given to avoiding microsleeps during the night.

Polygraphic sleep recordings included two electroencephalograms, two electrooculograms, one electromyogram, and one electrocardiogram. Sleep stages were scored at 30-s intervals using standardized criteria (Rechtschaffen and Kales, 1968).

Blood was removed continuously using a peristaltic pump. Samples were collected in an adjoining room at 10-min intervals during 24 h from 23:00 hours (day 1) to 23:00 hours (day 2). Samples were taken into EDTA-K2 treated tubes (1 mg mL⁻¹ blood) and immediately centrifuged at 4 °C. The plasma was stored at −25 °C until analysis. Plasma GH was determined by radioimmunoassay (Dia Sorin, Saluggia, Italy). The detection limit was 0.2 ng mL⁻¹. The intra-assay coefficient of variation (CV) was 2% for concentrations between 2 and 20 ng mL⁻¹ and 4% for levels below 2 ng mL⁻¹. Plasma melatonin was measured by radioimmunoassay using commercial kits (Immuno Biological Laboratories, Hamburg, Germany). The detection limit was 2.5 pg mL⁻¹. The intra-assay CV was 10% below 20 pg mL⁻¹, 7% between 20 and 120 pg mL⁻¹, and 20% above 120 pg mL⁻¹. All samples collected from an individual were assayed in the same series.

The 24-h profiles of both GH and melatonin have been previously analyzed and reported separately (Weibel et al., 1997a,b). Thus, the present data emerged from further analysis of the original data. The relationship between the melatonin and the GH rhythm in the same subjects has never been studied before.

Growth hormone secretory rates were derived from the corresponding plasma concentrations by a deconvolution procedure. A one-compartment model for hormone distribution and degradation was used with a subject-adjusted half-life of between 18 and 21 min. The distribution volume was assumed to be 7% of the subject's body weight. The individual profiles of GH secretory rates were analyzed and significant pulses were identified using a modification of the pulse detection algorithm ULTRA (Van Cauter et al., 1992). This algorithm eliminates all peaks when either the increment or the decrement does not reach three times the CV. For each significant pulse, the time of occurrence and the amount of GH secreted was evaluated. The 24-h melatonin profiles were smoothed using a locally weighted regression procedure proposed by Cleveland (1979). Melatonin onset was taken as the marker of the circadian phase (Lewy et al., 1999). The start of the surge was defined as the time when the value of the best-fit curve exceeded the mean of the 10 lowest consecutive values plus two standard deviations in at least 10 consecutive samples (Weibel et al., 1997b).

Comparisons were based on the nonparametric Mann–Whitney test. Correlations were assessed using the nonparametric Spearman coefficient. All values are given as mean ± SEM.

Table 1 summarizes sleep data for both groups of subjects. Sleep parameters did not show any significant difference among night workers and day-active subjects. In particular, total sleep time, sleep efficiency, and SWS duration were similar in both groups.

In day-active subjects, melatonin and GH showed the well-known 24-h profiles with the onset of the melatonin surge occurring at 21:53 hours ± 18 min and the major GH pulse at 23:36 hours ± 10 min, shortly after sleep onset in temporal association with the first SWS episode (Fig. 1, top). This GH pulse accounted for 52.8 ± 3.5% of the 24-h hormone production and the amount secreted (expressed as percentage over the night) was correlated with the duration of the SWS episodes (r = 0.70; P < 0.05).

As previously described (Weibel et al., 1997b), night workers showed variable circadian adaptation, with the onset of the melatonin surge varying between 21:45 and 05:05 hours. Temporal organization of the GH pulses over the 24 h differed widely among night workers. Only three of them presented a well-adapted GH rhythm with the major pulse shifted to beginning of daytime sleep and accounting for 45, 48, and 82% of the daily GH production. In eight of the 11 night workers, the sleep-related pulse accounted for only 16.8 ± 3.3% (range 3.2–22.3%) of the 24-h secretion (Fig. 1, bottom). The night workers did not show a correlation between the amount of GH secreted in the first pulse, expressed as percentage over the night, and SWS (r = 0.23; NS) despite there being no differences in the duration of the first SWS episode between day-active subjects and night workers (48.2 ± 6.1 versus 35.3 ± 5.9 min; NS).

In the eight non-adapted night workers, compared with day workers, the amount of GH secreted during the sleep period was significantly lower (45.7 ± 7.9 versus 94.0 ± 15.0 μg; P < 0.01). However, large individual pulses emerged unpredictably during waking periods, so that the amount of GH secreted during waking periods was significantly increased in night workers. Consequently, the total amount of GH secreted over 24 h was similar in the day-active and night-active subjects (159.0 ± 28.2 versus 151.7 ± 16.9 μg; NS), and the mean 24-h plasma GH levels did not differ between both groups of participants (2.86 ± 0.44 versus 2.98 ±0.58 ng mL⁻¹; NS). The night workers showed a more fragmented pattern of GH secretion, with an increased number of pulses during the 24-h period (14.7 ± 1.2 versus 10.8 ± 1.1; P < 0.05). The number of pulses was significantly increased during waking periods (9.7 ± 0.9 versus 6.3 ± 0.8; P < 0.02), but during sleep the number of pulses was similar in both groups (5.0 ± 0.4 versus 4.5 ± 0.4; NS).

Concerning the relationship between melatonin and GH profiles, there was a lack of correlation between the start of the melatonin surge and the quantity of GH secreted in the surge following sleep onset (r = 0.43; NS). Figure 2 illustrates this result in two night workers showing largely differing GH patterns despite a similar shift in the melatonin surge. There was also a varying time delay between the onset of the melatonin surge and the first subsequent GH surge (10–190 min).

These results demonstrate that despite similar amounts of SWS in day-active subjects and in night workers, the large sleep-related GH surge observed in subjects under a day time routine is lowered in night workers. This blunting is compensated for by the emergence of large individual surges occurring unpredictably during waking periods, which modifies the 24-h distribution of GH pulses. However, the total amount of GH secreted over the 24 h did not differ among day-active subjects and night workers. Simultaneous measurements of melatonin reveal that the extent of the melatonin surge is not related to the degree of GH adaptation, i.e. whether the major GH surge is centered around the first SWS episode and whether it accounts for the main quantity secreted over the 24-h period.

These results corroborate recent findings for sleep-deprived persons which did not conform with many previous studies (reviewed in Van Cauter et al., 1998). During acute sleep deprivation, individual GH profiles revealed the presence in awake subjects of large GH pulses that can reach similar amplitude as the nocturnal sleep-related surge and thus compensate the reduced nighttime secretion (Brandenberger et al., 2000). Similarly, in case of chronic sleep debt, the 24-h GH production was not impaired. GH secretion followed an unexpected biphasic pattern with a first pulse occurring during wakefulness around the time of sleep onset on a standard bedtime schedule, and a second pulse after the onset of sleep (Spiegel et al., 2000). All of these results indicate that sleep, whatever the conditions, does not play the dominating role on GH release that many have ascribed to it. They corroborate findings in rat that non-rapid eye movement sleep promotion and GH stimulation are independent outputs of hypothalamic GH-releasing hormone neurons (Obal and Krueger, 2001), and may explain the temporal dissociation between GH and SWS often seen in previous studies (Saini et al., 1993; Steiger et al., 1987). It should be noted that our results obtained in healthy subjects do not fit with the current view that insomnia lowers GH secretion (Lieb et al., 1999; Portaluppi et al., 1995; Vgontzas et al., 1998). Interestingly, a more fragmented pattern of GH secretion has been found in women (Antonijevic et al., 1999; Van Cauter et al., 1998). The increased daytime GH secretion was shown to be correlated with corresponding variations in endogenous estradiol levels (Ho et al., 1987).

In the case of a lesser sleep effect, is the circadian clock more directly involved in the control of GH pulsatility? We found that the extent of the melatonin shift does not predict the degree of GH adaptation. Moreover, no temporal relation could be found between the onset of the melatonin surge highly variable among night workers and the subsequent GH pulse. It thus seems unlikely that GH pulses are directly influenced by circadian processes.

It is established that a reciprocal interaction between the stimulating GH-releasing hormone and the inhibiting somatostatin is primarily responsible for GH release, and a weaker somatostatinergic influence during sleep than during awakening has been proposed (Giustina and Veldhuis, 1998; Müller et al., 1999; Van Cauter et al., 1998). Whether the large pulses observed during waking periods in night workers, and previously in sleep-deprived subjects, are attributable to interactions between the two processes is not yet known. It can be hypothesized that the large sleep-related surge in day-active subjects exerts a long-term feedback on subsequent GH secretion. One can also imagine that sleep does not trigger GH release but acts as a synchronizer of GH pulses occurring in absence of sleep. Corroboration of one or the other of these views could stem from previous results in other situations, such as short or long night conditions (Wehr, 1996) or forced desynchrony, which also revealed similar 24-h GH production despite a lowered sleep-related pulse in entrained conditions (Gronfier et al., 2001). Further examination of these data may lead to new thinking on the relationships between sleep, circadian processes, and GH pulsatility.

The metabolic consequences of changes in the temporal organization of GH pulses throughout the 24 h are not clearly established. The greater efficiency of pulsatile versus continuous GH administration has been demonstrated in animals (Bick et al., 1992) but is less apparent in humans (Jorgensen et al., 1990; Laursen et al., 2001). Exploring further the consequences of a disorganized distribution of GH pulses may have important implications for millions of individuals with sleep curtailment or involved in shift work.

We thank M Simeoni for experimental assistance, in particular with the RIA analyses, and for her contribution to the statistical analyses.