Structure and regulation of sleep

    Sleep of terrestrial mammals consists of two main sleep stages: rapid eye movement (REM) sleep and non-REM sleep which can be further divided into light and deep non-REM sleep. These stages can be recognized by their typical characters in sleep polygraphy, which includes electrophysiological recording of brain cortex activity by electroencephalography (EEG), eye movements by electro-oculography (EOG) and skeletal muscle tonus by electromyography (EMG). During light non-REM sleep there are typical activity bursts called sleep spindles in EEG and deep non-REM sleep can be recognized by synchronized EEG activity with high amplitude and low frequency delta waves, and for this reason deep non-REM sleep is also called slow wave sleep (SWS). Widely used human sleep stage classification criteria categorizes non-REM sleep into four stages (S1-4) where S2 is light and S4 deep non-REM sleep. S1 represents transition between wake and light sleep and S3 is an intermediate stage between light and deep non-REM sleep (Rechtschaffen and Kales 1968). Electrophysiologically REM sleep differs from non-REM sleep in several ways. During REM sleep, EEG is highly desynchronized corresponding to the EEG recorded during wakefulness. There is also almost a total loss of tonus in the skeletal musculature interrupted by occasional twitches especially in the facial muscles which can be recognized in EMG. In EOG there are bursts of rapid eye movements which has given the name for this vigilance stage. Because non-REM and REM sleep differ in several characteristics, these sleep stages can be regarded as two independent vigilance states together with wakefulness.

    According to a current and widely accepted model of sleep regulation, sleeping is controlled by two separate components: circadian process C which affects the appropriate timing of sleep and homeostatic process S which accounts for a sufficient amount of sleep (Borbély 1982). This is a two process model. The circadian process C is mainly controlled by the rhythmic activity of the suprachiasmatic nucleus in the hypothalamus. For the homeostatic process S, no single locus has been found in the CNS. It seems rather to be controlled by several neural systems including monoamines, neuropeptides, and cytokine transmitters which are localized in the hypothalamus, basal forebrain and brain stem nuclei (Borbély and Tobler 1989). There is another theory, which models the rhythm between REM sleep and the other vigilance states. According to this reciprocal interaction theory, REM sleep is controlled by feedback regulatory connections between monoaminergic and cholinergic nuclei in the pontine brainstem (Hobson et al. 1975). Monoaminergic nuclei of this regulatory system consist of the noradrenergic locus coeruleus and serotonergic raphe nuclei. These nuclei are most active during wakefulness, decrease their activity during non-REM sleep and are almost totally inhibited during REM sleep. Cholinergic laterodorsal and pedunculo-pontine tegmental nuclei (LDT, PPT) are most active during REM sleep. Several factors such as neuropeptides may also have modulatory effects on this REM sleep regulatory system.

Association of GH secretion with sleep

    Growth hormone (GH, somatotropin) is a polypeptide hormone consisting of 191 amino acids in humans and is secreted by somatotropic cells of the anterior pituitary. Unlike the other pituitary hormones, which affect a specific target organ, GH affects several tissues throughout the body. The main effects of GH are stimulation of growth in bone metaphyses during growing, anabolic effect in the muscles, conservation of proteins and carbohydrates, and mobilization of fat for energy sources (lipolysis). The effects of GH are partly mediated by somatomedins of which the most important are insulin-like growth factors (IGF) 1 and 2. Although somatomedins are secreted locally by several tissues, the plasma IGF-1 content is mostly originated from the liver and kidneys.

    Secretion of GH from the anterior pituitary is pulsatile. The secretion bursts are flanked by almost undetectable levels of plasma GH. GH pulses occur more frequently and the basal level of plasma GH is higher in females than males who have fewer GH pulses but which are of a higher amplitude (Van Cauter et al. 1998). In humans there is typically one high secretion pulse and a few lower ones during the 24-h day-night span (Van Coevorden et al. 1991, Van Cauter et al. 1998). Soon after the development of radioimmunological hormone assay methods in the 1960’s, it was found that the main secretion pulse was closely associated with the beginning of the sleep phase when the amount of SWS is highest (Takahashi et al. 1968, Honda et al. 1969, Sassin et al. 1969, Van Cauter et al. 1992). Sampling and measurements of plasma GH at frequent (up to 30 s) intervals combined with the deconvolution procedure, which mathematically estimates the rate of secretion of a hormone from the plasma concentrations, have revealed a close association of GH secretion with SWS phases (Van Cauter et al. 1992, Holl et al. 1991). Delay, advance or interruption of a sleep phase will shift the main GH secretion pulse correspondingly (Golstein et al. 1983, Van Cauter et al. 1992). At least in humans GH secretion is also controlled by an endogenous circadian rhythm. When the sleep period is shifted from its normal time, some GH is still secreted during the early night according to the endogenous clock (Van Cauter et al. 1992). GH secretion is highest during growing and early adulthood. In humans the secretion rate starts to decrease during the fourth decade of life. During aging the daytime secretion pulses diminish first, while the sleep associated GH pulse persists longer (Van Cauter et al. 1998).

    In animals it is more difficult to find a correlation between GH secretion and sleep because many animal species have typically several sleep phases of variable lengths during the 24-h day-night span. However, elevated plasma GH levels during sleep have been demonstrated in several mammals (reviewed by Van Cauter et al. 1998). In the rat, which is a widely used animal model in neuroscience, the GH secretion is pulsatile with an approximately 3-h cycle. This rhythm is associated with an ultradian sleep-wake rhythm with the same cycle length, so that the GH pulses precede the sleep maxima by about 24 min (Mitsugi and Kimura 1985). Short term (3 h) total sleep deprivation during the light phase resulted in a decrease of GH secretion during the deprivation in the rat (Kimura and Tsai 1984).

The effect of injected GH on sleep

    Systemic administration of GH by an intraperitoneal injection has been found to increase especially REM sleep in cats and rats (Stern et al. 1975, Drucker-Colín et al. 1975). When endogenous GH was immunoneutralized by an injection of GH antiserum, both SWS and REM sleep decreased in the rat (Obál et al. 1997). In humans, intramuscular bolus injection of GH before sleep onset was found to increase REM sleep and to decrease SWS (Mendelson et al. 1980). However, a more recent study failed to repeat this finding (Kern et al. 1993).

GH regulatory hypothalamic peptides

    Pituitary secretion of GH is controlled by two hypothalamic peptides: growth hormone-releasing hormone (GHRH) which stimulates GH secretion, and somatostatin (somatotropin-release inhibiting factor, SRIF) which inhibits it. Somatostatin was isolated and chemically characterized from the ovine hypothalamus in the early 1970’s (Brazeau et al. 1973). Somatostatin is a 14 amino acid cyclic peptide linked by an internal disulfide bridge (Fig. 5). Somatostatin is widely distributed in the body, e.g. in the brain, intestine and pancreas. In addition to GH secretion, it inhibits the secretion of several other hormones such as TSH, insulin, glucagon, and also pancreatic secretion. The somatostatin gene is 1.2 kilobases in length and have one 630 base intron within the coding sequence. Somatostatin mRNA is about 600 base pairs long and codes a 116 amino acid precursor peptide preprosomatostatin (Goodman et al. 1985). GHRH was isolated and chemically characterized from a pancreatic tumor in a patient with high GH levels in 1982 (Guillemin et al. 1982). It is a linear peptide containing 43 amino acids in rats and 44 in humans, but in the blood circulation also shorter forms exist. The rat GHRH gene spans nearly 10 kilobases in the genome and has five exons which encode a mRNA of approximately 700 nucleotides. GHRH mRNA is translated into 104 amino acid precursor peptide of the rat GHRH (Mayo et al. 1985, Gonzales-Cresbo and Boronat 1991).

    Peptide transmitter is synthesized by translation from mRNA in ribosomes and modified from precursor peptides in the Golgi apparatus. Transmitter peptide is then transported to axon terminals by axonal transport mechanisms. Except in the case of small molecule transmitters, there are no re-uptake mechanisms for peptide transmitters in presynaptic terminals and new transmitter is generated by translation from mRNA (Brownstein 1994). Although translation and modification of active peptide transmitter from its precursors are regulated by many cellular mechanisms, the amount of specific transmitter mRNA in a cell can be regarded as a rough estimate of the synthesis rate of the peptide transmitter. The rate of synthesis is in turn normally coupled to the release rate of a transmitter and to the activity of a peptidergic neuron.

Somatostatin cells in the CNS

    Somatostatin containing neurons are present widely in the CNS, e.g. in many hypothalamic nuclei, preoptic area, hippocampus and cortex. Somatostatin secreting cells are also found in the wall of the gastrointestinal tract. The widespread distribution of somatostatin cells in the CNS strongly suggests that somatostatin may also have functions other than the inhibition of GH secretion. The somatostatin cells that control GH release are located in the anterior part of the periventricular nucleus of the hypothalamus (Ishikawa et al. 1987, Merchenthaler et al. 1989), from where they project to the median eminence. There somatostatin is released to the pituitary portal vein system from neurosecretory axon terminals and finally to its receptors in the somatotrophic cells of the pituitary gland (Fig. 1). To date five main types of somatostatin receptors have been cloned (SSR1-5). Due to alternative mRNA splicing, there are two variants of SSR2, named SSR2A and B (Florio et al. 1996). All SSRs reduce the production of cyclic AMP. There are also several other intracellular transduction mechanisms which may couple to SSRs (Florio et al. 1996). Although all types of SSRs are present in the pituitary, inhibition of GH secretion is mediated mainly by SSR2 in humans and rats. This action involves inhibition of cytoplasmic Ca2+ influx. There is a strong expression of SSR1 in the periventricular area and the median eminence. This receptor type is possibly an auto-receptor in hypophysiotropic somatostatin cells, where somatostatin reduces its own release by an ultra short-loop feedback mechanism (Helboe et al. 1998). Somatostatin cells located in the dorsomedial part of the arcuate nucleus probably inhibit GHRH cells in the same nucleus but do not project into the median eminence (Willoughby et al. 1989, Merchenthaler et al. 1989) (Fig.1). SSRs are found in the noradrenergic cells of the locus coeruleus where they inhibit the activity of these neurons (Inoue et al. 1988, Pérez et al. 1994, Mounier et al. 1996). The locus coeruleus has projections to the hypothalamus where noradrenaline stimulates the release of somatostatin in the periventricular nucleus (Liposits 1993, Mounier et al. 1996). This mechanism may partly mediate the attenuated GH release during wakefulness when the neurons in the locus coeruleus are active (Fig. 15).

GHRH cells in the CNS

    Unlike somatostatin, the distribution of GHRH neurons in the CNS is restricted to the hypothalamus (Sawchenko et al. 1985). The most important GHRH cell group controlling the GH secretion is in the arcuate nucleus (Daikoku et al. 1986) (Fig.1). GHRH containing cells have also been found in the paraventricular nucleus, dorsomedial nucleus, and in the periventromedial hypothalamic area (Sawchenko et al. 1985). Hypophysiotropic GHRH cells in the arcuate nucleus also express tyrosine hydroxylase (TH) and have dopamine as a co-transmitter. Dopamine is released into the portal circulation of the pituitary, where it regulates at least the secretion of prolactin and TSH (Niimi et al. 1992). There are possibly autoregulatory connections within the arcuate GHRH cells (Horváth and Palkovits 1988) (Fig. 1). Two types of GHRH cells exist which have partly different morphology and possibly also different projections (Daikoku et al. 1986). Most of arcuate GHRH cells project to the median eminence and are hypophysiotropic, while the GHRH cells outside of the arcuate nucleus, especially in the periventromedial hypothalamic area project to other hypothalamic nuclei and neighboring brain areas such as the preoptic hypothalamic area and the basal forebrain (Sawchenko et al. 1985). These cells may account for the possible neuroregulatory effects of GHRH. A G-protein mediated receptor for GHRH has been isolated and cloned (Mayo 1992). In addition to the pituitary gland, the GHRH receptor is expressed in several hypothalamic nuclei and in the preoptic area (Takahashi et al. 1995).

Autoregulation of GH-axis in the brain

    Plasma GH content regulates the activity of hypophysiotropic somatostatin and GHRH cells. An elevated GH level in plasma increases the synthesis and release of somatostatin and decreases the synthesis and release of GHRH (Zeitler et al. 1990). Hypophysectomy decreases the activity of somatostatin cells and increases GHRH cell activity (Minami et al. 1993). GH receptors are present in periventricular somatostatin cells (Burton et al. 1992). Only a minority (<10%) of GHRH cells in the arcuate nucleus expresses the GH receptor, which may indicate that the effect of GH on GHRH is mediated indirectly by somatostatin, IGFs, or neuropeptide-Y (NPY) (Burton et al. 1995, Korbonits et al. 1996, Chan et al. 1996). (Fig. 1)

    Somatostatin and GHRH cells also affect each other via intrahypothalamic connections without the involvement of GH (Horváth et al. 1989, Zeitler et al. 1991) (Fig.1). Somatostatin cells in the periventricular nucleus inhibit the activity of GHRH cells in the arcuate nucleus, and also inhibit the release of GHRH from the axon terminals in the median eminence (Dickson et al. 1994). Arcuate somatostatin and GHRH cells possibly have a reciprocal regulatory connection within the nucleus, where somatostatin cells inhibit GHRH cells and GHRH stimulates somatostatin (Horváth et al. 1989). GHRH cells in the arcuate are possibly connected to the periventricular nucleus where they stimulate cells, but this regulation pathway may be mediated indirectly by NPY or galanin (Chan et al. 1996) because very few arcuate GHRH cells have been found to project directly to the periventricular nucleus (Willoughby et al. 1989). Nevertheless, GHRH fibers and GHRH receptors are present in the periventricular nucleus (Horváth et al. 1989, Takahashi et al. 1995), which suggests that also other than arcuate GHRH cells may regulate the somatostatin cells in periventricular nucleus.

    Feedback regulation by GH and intrahypothalamic connections generate rhythmic alteration in the expression of mRNA in hypophysiotropic somatostatin and GHRH cells, so that the maximal mRNA expression in somatostatin cells is 180° out of phase with the maximal expression of GHRH mRNA (Zeitler et al. 1991). This oscillation may explain the GH secretion rhythm in rats and GH pulses in humans.

Fig. 1  Schematic illustration of the hypothalamic regulatory system of the pituitary GH release. SS: somatostatin, PeV: periventricular nucleus, PaV: paraventricular nucleus.

The effect of injected GHRH on sleep

    I.c.v. injection of GHRH increases the amount of SWS, slow wave activity and REM sleep in rats and rabbits (Ehlers et al. 1986, Nisticò et al. 1987, Obál et al. 1988). A similar effect has also been found after systemic injection of GHRH in rats (Obál et al. 1996). The observed stimulatory effect of systemic GHRH on REM sleep is abolished by hypophysectomy. This does not affect the SWS promoting effect, which suggests that the effect of GHRH on SWS is primary and not mediated by GH (Obál et al. 1996).

    In humans repeated i.v. injections of GHRH have been found to increase the amount of SWS during the early sleep phase (Steiger et al. 1992, Marshall et al. 1996) or after a single injection in the middle of the sleep phase after the third REM sleep episode (Kerkhofs et al. 1993). A single i.v. injection in the early sleep phase increases the amount of REM sleep but has no effect on SWS (Kerkhofs et al. 1993). Continuous infusion (Sassolas et al. 1986, Kern et al. 1993, Marshall et al. 1996), bolus injection before or at the sleep onset (Garry et al. 1985, Kupfer et al. 1991) and repeated i.v. boluses during the late sleep phase (Schier et al. 1997) failed to affect the sleep structure although these treatments increased plasma GH.

    Blocking the action of GHRH by i.c.v. injection of a GHRH antagonist (Obál et al. 1991) or antibody (Obál et al. 1992) decreased the amount of spontaneous sleep in rats, and administration of a GHRH antibody also decreased sleep rebound after sleep deprivation. The amount of GHRH mRNA is highest in the morning when the amount of non-REM sleep is usually at its highest in rats (Bredow et al. 1996). A GHRH-GH-IGF deficient dwarf mouse strain has a reduced amount of non-REM sleep (Zhang et al. 1996), which also suggests that GHRH facilitates non-REM sleep.

The effect of injected somatostatin on sleep

    A long term intracerebroventricular (i.c.v.) infusion of somatostatin was found to increase and depletion of somatostatin stores by cysteamine to decrease REM sleep in rats (Danguir 1986). However, i.c.v. or epicortical administration of a high dose (10Fg) of somatostatin in rats increased wakefulness during the first two hours after the injection (Havlicek et al. 1976, Rezek et al. 1976). Intraperitoneal (i.p.) or subcutaneous (s.c.) injection of the long-acting somatostatin analog octreotide (SMS 201-995, OCT, Fig. 5) causes a delayed facilitation of REM sleep in rats (Danguir and De Saint-Hilaire-Kafi 1988b, Beranek et al. 1997). A selective REM sleep facilitatory effect has also been found in aged rats after oral or i.p. administration of octreotide (Danguir 1989). I.p. injection of octreotide counteracts the suppression of REM sleep caused pharmacologically by scopolamine (Danguir and De Saint-Hilaire-Kafi 1988b) or by desipramine (Danguir and De Saint-Hilaire-Kafi 1989). Immunoneutralization of endogenous somatostatin by i.c.v. infusion of somatostatin antibody decreases REM sleep (Danguir 1988) and microinjection of somatostatin antibody into the nucleus of the solitary tract antagonizes carbachol-induced REM sleep in rats (Danguir and De Saint-Hilaire-Kafi 1988a).

    Somatostatin appears to stimulate REM sleep in laboratory animals, whereas human studies have yielded inconsistent results of the effects of somatostatin on sleep. Intravenously (i.v.) infused somatostatin at the beginning of sleep did not affect the structure of human sleep, although it suppressed the GH secretion peak (Kupfer et al. 1992). Repeated i.v. boluses of somatostatin caused only a non-significant tendency of enhanced REM sleep in young males (Steiger et al. 1992). When somatostatin was injected into elderly subjects according to the same protocol, REM sleep decreased and the sleep structure deteriorated (Frieboes et al. 1997).


    GH-secretagogues (GHS) are a class of synthetic oligopeptides, such as growth hormone-releasing peptide 6 (GHRP-6), GHRP-2 and non-peptide compounds (MK-677, L-692,429, L-629,585), which are potent stimulating factors of GH release (Bowers et al. 1984, Bowers 1994). The GHS receptor, which differs from the GHRH receptor, has been cloned (Howard et al. 1996) and it is found to be expressed in the pituitary, arcuate nucleus and ventromedial hypothalamus (Guan et al. 1997). Approximately 25% of the GHRH cells in the arcuate nucleus and periventromedial hypothalamus express GHS receptor mRNA (Tannenbaum et al. 1998). GHSs stimulate GH secretion both directly from the pituitary (Cheng et al. 1993) and indirectly by affecting the activity of GHRH cells in the arcuate nucleus (Dickson et al. 1995, Dickson and Luckman 1997). The present hypothesis is that the GHS receptor is a receptor for an unknown endogenous substance controlling GH secretion.

    Pharmacological studies of the effects of GHSs on human sleep suggest that these compounds could facilitate sleep. Repeated i.v. boluses of GHRP-6 during early night increased the amount of light S2 sleep, although the amount of wakefulness, SWS or REM sleep was not affected (Frieboes et al. 1995). An i.v. bolus of GHRP-2 after the third REM sleep episode did not affect sleep parameters significantly while causing a marked GH secretion pulse (Moreno-Reyes et al. 1998). Oral administration of MK-677 at bedtime for 7-14 days was found to increase the amount of deep SWS (S4) and REM sleep in young (18-30 years) and REM sleep in elderly (65-71 years) subjects (Copinschi et al. 1997). GH secretion was not affected while IGF-1 levels increased as a consequence of GHS treatment in the young subjects. In the old subjects administration of MK-766 increased both GH secretion and IGF-1 levels (Copinschi et al. 1997). The effects of systemic or local injections of GHS on sleep have not been studied in experimental animals according to the current literature. I.c.v. injection of GHRP-6 paradoxically decreased GH secretion in rats (Yagi et al. 1996). This may be due to stimulation of hypothalamic somatostatin cells because pretreatment with somatostatin antibody blunted this effect. I.c.v. injection of GHSs stimulated electrical activity and Fos protein production in the arcuate GHRH cells in rats (Dickson et al. 1995). Immunoneutralization of endogenous GHRH strongly decreased the effect of injected GHRP on GH secretion (Yagi et al. 1996), which suggests that the effect of GHS is mostly mediated by GHRH and not by its direct effect on somatotrophic cells in the pituitary.

    There is a special interest in GHS as a possible remedy against age related attenuation of GH secretion (Merriam et al. 1997). Spontaneous secretion of GH decreases strongly during the fourth decade of life and this could partly explain the simultaneous reduction of SWS (Van Cauter et al. 1998). As presented earlier, many of the known GHSs are effective when administered orally. Since they stimulate both GHRH and GH secretion they could improve SWS and REM sleep and also have such beneficial effects as conservation of muscle mass and decreased accumulation of fat.

The role of galanin as a possible sleep regulating neuropeptide

    Galanin is a 29 amino acid peptide isolated first from the porcine intestine (Tatemoto et al. 1983). The human galanin has 30 amino acids. The rat galanin mRNA is about 900 base pairs and it encodes a 124 amino acid precursor peptide, from which the 29 amino acid galanin is finally cleaved (Vrontakis et al. 1987). Galanin is present in the gastrointestinal tract and widely in the CNS where it has been found to participate in many brain functions such as feeding, memory consolidation, nociception and hormone release (Bartfai et al. 1993). Galanin is usually co-localized with a non-peptide or with another peptide transmitter: with acetylcholine in the preoptic area, with GHRH and dopamine in the hypophysiotropic cells of the arcuate nucleus, with CRH in the paraventricular nucleus, with histamine in the posterior hypothalamus, and with noradrenaline in the locus coeruleus (Melander et al. 1986, Skofitsch and Jacobowitz 1985). The post-synaptic effects of galanin are usually inhibitory and mediated by three types of galanin receptors (Smith et al. 1998). Systemic or intracerebroventricular injection of galanin increased GH secretion while immunoneutralization or i.c.v. injection of galanin antagonist decreased it (Bauer et al. 1986, Ottlecz et al. 1988, Gabriel et al. 1993). Since galanin does not affect GH release from pituitary cells in vitro, it is probable that galanin affects the GH regulatory system in the hypothalamus (Ottlecz et al. 1988). Galanin immunoreactive fibers are found in the periventricular nucleus (Liposits et al. 1993) and periventricular somatostatin cells express galanin receptors (Chan et al. 1996). This suggests that galanin increases GH secretion by inhibiting somatostatin release (Tanoh et al. 1993) (Fig.1). Since very few GHRH cells in the arcuate nucleus express galanin receptors (subtype 1), it is possible that GHRH is indirectly involved in this mechanism (Chan et al. 1996).

    Since galanin is expressed in brain areas known to regulate vigilance states, e.g. in the preoptic area, posterior hypothalamus, and locus coeruleus, galanin might directly affect sleep as a sleep modulatory neuropeptide or by the hypothalamic GH regulatory system. The noradrenergic locus coeruleus is one possible site for a direct effect of galanin on sleep. Galanin immunoreactive fibers are found in the locus coeruleus (Skofitsch and Jacobowitz 1985) and locus coeruleus cells express galanin receptors which inhibit noradrenergic cells like somatostatin in the same nucleus (Sevcik et al. 1993). Recent findings in human subjects indicate that i.v. administration of galanin increases sleep (Murck et al. 1999).