ACTH is derived from a 266 amino acid precursor, pro-opiomelanocortin (POMC: Figure 1). POMC is encoded by a single-copy gene on chromosome 2p23.3 over 8 kb (3). It contains a 5’ promoter and three exons. Apart from the hydrophobic signal peptide and 18 amino-acids of the N-terminal glycopeptide, all of the rest of POMC is encoded by the 833 bp exon 3.

Figure 1. POMC and its derivatives

The promoter of POMC has most extensively been studied in the rat. Common transcription elements such as a TATA box, a CCAAT box and an AP-1 site are found within the promoter (4, 5). Corticotroph and melanotroph-specific transcription of POMC appears to be dependent on a CANNTG element motif synergistically binding corticotroph upstream transcription element-binding (CUTE) proteins (6). These include NeuroD (7), Ptx1 (8), and Tpit (9). Interestingly, Ikaros transcription factors, which had previously been characterized as being essential for B and T cell development, have recently been demonstrated to bind and regulate the POMC gene in mice. Moreover, Ikaros knockout mice demonstrate impaired corticotroph development in their pituitaries, as well as reduced circulating ACTH, MSH and corticosterone levels (10).

POMC transcription is positively regulated by corticotrophin releasing hormone (CRH: vide infra). CRH acts via its G-protein coupled receptor to activate adenylate cyclase, increase intracellular cAMP and stimulate protein kinase-A (11). Transcription stimulation is mediated by an upstream element (PCRH-RE) binding a novel transcription factor (PCRH-REB) containing protein kinase-A phosphorylation sites (12). CRH also stimulates the transcription of c-fos, FosB and JunB, as well as binding to the POMC AP-1 site (13). The pituitary adenylate cyclase-activating peptide (PACAP) also stimulates cAMP synthesis and POMC transcription, presumably through a common pathway with CRH (14).

POMC mRNA transcription is negatively regulated by glucocorticoids (15). The glucocorticoid effect appears, in the rat POMC promoter, to be dependent on a negative glucocorticoid response element partially overlapping the CCAAT box (16). The element binds the glucocorticoid receptor as a homodimer plus a monomer on the other side of the DNA helix (17). Glucocorticoid regulation of transcription may also be indirectly mediated via down-regulation of c-jun expression and direct protein-protein mediated inhibition of CRH-induced AP-1 binding (18), as well as inhibition of CRH receptor transcription (19).

Leukaemia Inhibitory Factor (LIF), a pro-inflammatory cytokine expressed in corticotrophs, has also been shown to stimulate POMC transcription, via activation of the Jak-STAT pathway (20, 21). This stimulation is synergistic with CRH. Deletional analysis of the POMC promoter has identified a LIF-responsive region from –407 to –301. It has been speculated that this pathway might form an interface between the immune system and regulation of the pituitary-adrenal axis (22).

Prohormone convertase enzymes PC1 and PC2 process POMC (Figure 1) at pairs of basic residues (lys-lys or lys-arg). This generates ACTH, the N-terminal glycopeptide, joining peptide, and beta-lipotropin (beta-LPH). ACTH can be further processed to generate alpha-melanocyte stimulating hormone (alpha-MSH) and corticotropin-like intermediate lobe peptide (CLIP), whereas beta-LPH can be processed to generate gamma-LPH and beta-endorphin (23). In corticotrophs, POMC is mainly processed to the N-terminal glycopeptide, joining peptide, ACTH and beta-LPH; smaller amounts of the other peptides are present (24). Other post-translational modifications include glycosylation of the N-terminal glycopeptide (25), C-terminal amidation of N-terminal glycopeptide, joining peptide and alpha-MSH (26, 27), and N-terminal acetylation of ACTH, alpha-MSH and beta-endorphin (28, 29).

This 41 amino-acid neuropeptide (30) is derived from a 196-amino acid prohormone (31). CRH immunoreactivity is mainly found in the paraventricular (PV) nuclei of the hypothalamus, often co-localised with arginine vasopressin (32). CRH binds to a seven-transmembrane domain receptor (33) G-protein coupled to adenylate cyclase, stimulating cAMP synthesis and PK-A activity. Besides stimulating POMC transcription and ACTH biogenesis (vide supra), CRH stimulates the release of ACTH, leading to a biphasic response with the fast release of a pre-synthesised pool of ACTH, and the slower and sustained release of newly synthesised ACTH (34).

In the anterior pituitary, AVP principally binds to the seven-transmembrane domain V1b receptor, also known as the V3 receptor (35). The receptor is coupled to phospholipase C, phosphatidyl inositol generation and activation of protein kinase-C (36, 37) and not via adenylate cyclase and cAMP (11). AVP stimulates ACTH release weakly by itself, but synergises with the effects of CRH on ACTH release (38). Downregulation of protein kinase C by phorbol ester treatment abolishes the synergistic effect of AVP on ACTH release by CRH (39). AVP does not stimulate POMC transcription either by itself or in synergism with CRH (40).

Frequent sampling of ACTH with deconvolution analysis reveals that it is secreted in pulses from the corticotroph with 40 pulses +/- 1.5 measured per 24 hours, on analysis of 10 minute sampling data. These pulses temporally correlate with the pulsed secretion of cortisol, allowing for a 15 minute delay in secretion, and correlate in amplitude (76). Pulse concordance has been measured at 47% (ACTH to cortisol) and 60% (cortisol to ACTH) in one study (77), and 90% (ACTH to cortisol) in another (78). Although the pulsatility of ACTH secretion may result from pulsatile CRH release, there is evidence that isolated human pituitaries intrinsically release ACTH in a pulsatile fashion (79).

In parallel with cortisol, ACTH levels vary in an endogenous circadian rhythm, reaching a peak between 0600-0900h, declining through the day to a nadir between 2300h-0200h, and beginning to rise again at about 0200-0300h. An increase in ACTH pulse amplitude rather than frequency is responsible for this rhythm (76).

The circadian rhythm is mediated via the supra-chiasmatic nucleus (SCN). An autoregulatory negative feedback system involving cyclical synthesis of period proteins PER1-3, CLOCK/BMAL1 and CRYPTOCHROME acts as the basic oscillator (80). Entrainment of the oscillator is achieved by light input from the retina, mediated via the retinohypothalamic tract. Light-activated transcription of immediate-early genes such as c-fos and JunB (81, 82) causes activation of PER1 gene transcription as well as modification of the acetylation pattern of histone tails. The latter are implicated in the control of chromatin structure and accessibility of genes to transcription (83).

Is a circadian rhythm in CRH secretion responsible for the ACTH rhythm? Although there is a report of a circadian rhythm in CRH secretion (84), other reports do not confirm this (85). Moreover, the circadian rhythm persists despite a continuous infusion of CRH, suggesting that other factors are responsible for the modulation of ACTH pulses (86). The most likely alternative candidate is AVP: immunocytochemical studies show a circadian rhythm in AVP expression (87) and transgenic knockout mice for CLOCK show a loss in the circadian rhythm in AVP RNA expression (88).

Stress, both physical and psychological, induces the release of ACTH, particularly via CRH and AVP (89, 90), and increases the turnover of these neurohypophysiotropic factors by increasing the transcription of CRH and AVP (91). The hypoglycaemia during the insulin tolerance test is one such stressor (Figure 2), as is venepuncture (92).

Figure 2. Typical response to hypoglycaemia (≤2.2 mmol/l) induced by 0.15 U/kg Actrapid i.v. in a normal subject. Peak cortisol is ≥550 nmol/l.

Interestingly, there is evidence that different stress paradigms have differential effects on CRH and AVP. In situ hybridization with intronic and exonic probes can be used to study the transcription of heterogenous nuclear RNA (hnRNA), followed by its processing (including splicing, capping and polyadenylation) to messenger RNA (mRNA) within 1-2 hours. CRH and AVP hnRNA levels in rats subjected to restraint show significant increases at 1 and 2 hours after the induction of stress, followed by significant increases in mRNA levels at 4 hours (93). In contrast, intraperitoneal hypertonic saline causes a rapid 8.6-fold increase in CRH hnRNA and mRNA within 15 minutes, returning to basal levels by 1 hour. AVP hnRNA responses are slower, peaking at 11.5-fold increase by 2 hours, followed by a prolonged elevation of AVP mRNA levels from 4 hours onwards (94).

Repetitive stress causes variable effects, enhancement or desensitization, on ACTH responses, depending on the stress paradigm involved. This appears to be positively correlated with changes in AVP binding to V1b receptors, reflecting changes in the number of binding sites and not their affinities. It is at present unclear whether this is due to changes in transcription of the V1b gene, alterations in mRNA stability, translational control or recruitment of receptors from intercellular pools (95).

Recent work has also characterised roles for endogenous nitric oxide (NO) and carbon monoxide (CO) in mediating the ACTH response to stress (96). Neuronal nitric oxide synthase co-localizes with AVP and to some extent CRH in paraventricular neurones (97, 98). Knockout mice lacking wild-type and neuronal nitric oxide synthase have much reduced quantities of POMC immunoreactivity in their arcuate nuclei and pituitaries compared to wild-type mice (97, 99). In general, inflammatory stressors appear to activate an endogenous inhibitory pathway whereby NO and CO attenuate the stimulated secretion of CRH and AVP. These effects can also be seen in terms of circulating AVP. However, the regulation of the pituitary-adrenal axis by other stressors may involve an activating role for these gaseous neurotransmitters.

Glucocorticoid feedback occurs at multiple levels: at the pituitary, at the hypothalamus, and most importantly, centrally at the level of the hippocampus, which contains the highest concentration of glucocorticoid receptors in the central nervous system. Multiple effects mediate this feedback (Figure 3), including:

Figure 3. Regulation of ACTH. Green arrows denote stimulatory influences, red arrows denote inhibitory influences.

Fast feedback occurs within seconds to minutes and involves inhibition of ACTH release by the corticosteroids. In vitro this appears to involve inhibition of stimulated ACTH and CRH release, but basal secretion is not affected. Protein synthesis is not required, implying that the glucocorticoid effect is non-genomic, e.g. by inhibition of second-messenger systems (103, 104). Recent evidence implicates the endocannabinoids in mediating this fast feedback inhibition (105). Intermediate feedback occurs within 4 hours time frame and involves inhibition of CRH synthesis and release, but does not affect ACTH synthesis (104). Slow feedback occurs over longer timeframes and involves inhibition of POMC transcription (104).

There is evidence that ACTH can inhibit CRH synthesis in the context of elevated CRH levels due to Addison’s disease or hypopituitarism, although not in the context of normal human subjects (106). Immunohistochemical studies of the paraventricular nuclei in adrenalectomised or hypophysectomised rats show a reduction of CRH and AVP positive cells when these rats are given ACTH infusions (107).

The GH locus, a 66 kb region of DNA, is located in chromosome 17q22-q24 and consists of 5 homologous genes, which appear to have been duplicated from an ancestral GH-like gene – Table I (110, 111).

Because of their origin from an ancestral GH-like gene, all five genes in the GH genomic locus share 95% sequence identity including their promoters (119): proximal elements in the promoter bind Pit-1/GHF-1 (120-123). Although Pit-1 is necessary for transcription of transfected GH1 genes in rat pituitary cells, it is not sufficient (124). Other transcription factors such as Sp1, CREB, and the thyroid hormone receptor are involved (122, 125, 126).

A placenta-specific enhancer found downstream of the CSH genes (127) as well as pituitary-specific repressor sequences found upstream of GH2, CSH-1 and -2, and CSHL-1 may serve to limit transcription of these particular genes to the placenta (128).

A locus control region consisting of two DNase-I hypersensitive regions 14.5 and 30 kb upstream of GH1 appears to be required for pituitary-specific GH1 expression (129). This region, which also binds Pit-1 (130), activates histone acetyltransferase, which controls chromatin structure and the accessibility of the GH locus to transcription factors (131, 132).

This is a 191 amino-acid single chain polypeptide hormone that occurs in various modified forms in the circulation. During spontaneous pulses of secretion the majority full-length isoform of 22 kDa makes up 73%, the alternatively spliced 20 kDa isoform contributes 16%, while the ‘acidic’ desamido and N-alpha acylated isoforms make up 10%. During basal secretion between pulses other forms (30 kDa, 16 kDa and 12 kDa) can also be identified which consist of immunoreactive fragments of GH (133-135).

Higher molecular weight forms of GH exist in the circulation, representing GH bound to binding proteins or GHBP (136). The high-affinity GHBP consists of the extracellular domain of the hepatic GH receptor, and this binds the 22 kDa GH isoform preferentially (137). The low-affinity GHBP binds the 20 kDa isoform preferentially (138). Binding of GH to GHBP prolongs the circulation time of GH as the complex is not filtered through the glomeruli (134). GH/GHBP interactions may also compete for GH binding to its surface receptors (139).

GHRH was originally isolated from a pancreatic tumour taken from a patient that presented with acromegaly and somatotroph hyperplasia (140). GHRH is derived from a 108 amino-acid prepro-hormone to give GHRH(1-40) and (1-44) (Figure 4), which are both found in the human hypothalamus (141, 142). The C-terminal 30-44 residues appear to be dispensable, as residues 1-29 show full bioactivity. GHRH binds to a seven-transmembrane domain G-protein coupled receptor that activates adenylate cyclase (143), which stimulates transcription of the GH gene as well as release of GH from intracellular pools (144, 145). No other hormone is released by GHRH, although GHRH has homology to other neuropeptides such as PHI, glucagon, secretin and GIP (146).

Figure 4. Hypophysiotrophic hormones influencing GH release.

Somatostatin (a.k.a. somatotropin release inhibitory factor or SRIF) is derived from a 116 amino-acid prohormone to give rise to two principal forms, somatostatin-28 and -14 (147). Both of these are cyclic peptides due to an intramolecular disulphide bond (Figure 4). Somatostatin has multiple effects on anterior pituitary as well as pancreatic, liver and gastrointestinal function:

Somatostatin binds to specific seven-transmembrane domain G-protein coupled receptors, of which there are at least 5 subtypes. Subtypes 2 and 5 are the most abundant in the pituitary (157). The somatostatin receptors couple to various 2nd messenger systems such as adenylate cyclase, protein phosphatases, phospholipase C, cGMP dependent protein kinases, potassium, and calcium ion channels (158).

Ghrelin is the most recently discovered GH regulatory factor and was isolated from stomach as the endogenous ligand of the GH secretagogue receptor (GHS-R), another member of the seven-transmembrane receptor family G-protein coupled to the phospholipase C-phosphoinositide pathway (159, 160). Ghrelin is derived from preproghrelin, a 117 amino-acid peptide, by cleavage and n-octanoylation at the third residue to give a 28 amino-acid active peptide (Figure 4). The majority of circulating ghrelin exists as the des-octanoylated (des-acyl) form: octanoylated ghrelin constitutes approximately 1.8% of the total amount of circulating ghrelin (161). Octanoylation appears to be essential for GH secretagogue activity, as des-acyl ghrelin is inactive for GH release (159). Historically, the earliest GH secretagogues discovered such as GHRP-1, GHRP-2, GHRP-6 and hexarelin were synthetic and derived from enkephalins (162).

In the circulation, ghrelin appears to be bound to a subfraction of HDL particles containing clusterin and the A-esterase paraoxonase. It has been suggested that paraoxonase may be responsible for catalyzing the conversion of ghrelin to des-acyl ghrelin (163). However, inhibition of paraoxonase in human serum does not inhibit the de-acylation of ghrelin, and there is a negative correlation in these sera between the paraoxonase activity and ghrelin degradation. Instead, it is more likely that butyrylcholinesterase and other B-esterases are responsible for this activity (164).

Ghrelin is present in the arcuate nucleus of the hypothalamus and in the anterior pituitary (165). Immunofluoresence studies show that ghrelin is localized in somatotrophs, thyrotrophs and lactotrophs but not in corticotrophs and gonadotrophs, suggesting that ghrelin may be acting in a paracrine fashion in the anterior pituitary (166). It stimulates GH release in vitro directly from somatotrophs (159) and also when infused in vivo, although the latter action appears to require the participation of an intact GHRH system (150). Ghrelin stimulates GH secretion in a synergistic fashion when co-infused with GHRH (63). Besides its GH releasing activity, ghrelin has orexigenic activity (167, 168), stimulates insulin secretion (169), ACTH and prolactin release (170). Knocking out the ghrelin gene in mice does not seem to affect their size, growth rate, food intake, body composition and reproduction, indicating that ghrelin is not dominantly and critically involved in mouse viability, appetite regulation and fertility (171). Ghrelin null mice show an increased utilization of fat as an energy substrate when placed on a high-fat diet, which may indicate that ghrelin is involved in modulating the use of metabolic substrates (172). GHS-R knockout mice have the same food intake and body composition as their wild-type littermates, although their body weight is decreased in comparison. However, treatment of GHS-R null mice with ghrelin does not stimulate GH release or food intake, confirming that these properties of ghrelin are mediated through the GHS-R (173).

This is not to say, however, that des-acyl ghrelin does not have any biological effects. It has been shown to inhibit apoptosis and cell death in primary cardiomyocyte and endothelial cell cultures (174), to have varying effects on the proliferation of various prostate carcinoma cell lines (175), to inhibit isoproterenol-induced lipolysis in rat adipocyte cultures (176), and to induce hypotension and bradycardia when injected into the nucleus tractus solitarii of rats (177). More controversially, intracerebroventricular or peripherally adminstered des-acyl ghrelin causes a decrease in food consumption in fasted mice and inhibits gastric emptying. Des-acyl ghrelin overexpression in transgenic mice causes a decrease in body weight, food intake, fat pad mass weight and decreased linear growth compared to normal littermates (178). These observations were not replicated by other researchers, who found no effect of des-acyl ghrelin on feeding (179). The effects of des-acyl ghrelin appear not to be mediated via the type 1a or 1b GHS-R (174-176). The effects of peripherally administered des-acyl ghrelin on stomach motility can be inhibited by intracerebrovascular CRH receptor type 2 antagonists, suggesting that CRH receptor type 2 is involved, but there is no direct evidence that des-acyl ghrelin binds this receptor (180)

As noted above, the actions of ghrelin in vivo seem to require an intact GHRH system, as immunoneutralisation of GHRH blocks GH secretion induced by ghrelin (150). The actions of GH secretagogues are blocked by hypothalamo-pituitary disconnection, which suggests that in vivo ghrelin’s stimulatory actions are indirect and mediated by GHRH (181). However, GHRH cannot be the sole mediator of ghrelin’s actions as the GH response to ghrelin is greater than that to GHRH (182), and, as noted above, ghrelin synergistically potentiates GH release by a maximal dose of GHRH (63). There is no evidence to suggest that ghrelin decreases somatostatinergic tone as immunoneutralisation of somatostatin does not block ghrelin’s ability to release GH (150). There may therefore be another mediator, the so-called ‘U’ factor, released by ghrelin, which causes GH secretion (183).

Glucocorticoid treatment has a biphasic effect on GH secretion: an initial acute stimulation in 3 hours, followed by suppression within 12 hours (184, 185). The latter is the clinically important effect, as excess endogenous and exogenous glucocorticoids are well known to suppress growth in children (186).

Leptin is a 167 amino-acid peptide primarily produced by white adipose tissue (187), which regulates body fat mass (188) by feedback inhibition of the appetite centres of the hypothalamus (189). Leptin and its receptor has been detected both by RT-PCR and immunohistochemistry in surgical pituitary adenoma specimens and in normal pituitary tissue (190, 191). However, pituitary adenoma cells in culture do not secrete GH in response to leptin treatment (191, 192).

In rats, immunoneutralisation with leptin antisera decreases GH secretion. Intracerebroventricular leptin administration reverses the inhibitory effect of fasting on GH levels in rats. However, intracerebroventricular leptin by itself does not significantly influence GH secretion (193). These observations, however, may not be extendable to humans, as the physiology of GH in humans appears to be very different from rats, e.g. GH levels in humans are increased by fasting in contrast to suppression in rats (vide infra).

In general, alpha-adrenergic pathways stimulate GH secretion, by stimulation of GHRH release and inhibition of somatostatinergic tone, while beta-adrenergic pathways inhibit secretion by increasing somatostatin release (194, 195). The alpha2-adrenoceptor agonist clonidine can therefore be used as a provocative test of GH secretion (196, 197) although clinical experience suggests that this is an unreliable stimulatory test for GH secretion in practice. L-dopa stimulates GH secretion; however, this action does not appear to be mediated via dopamine receptors as specific blockade of these receptors with pimozide does not alter the GH response to L-dopa (198). Instead, L-dopa’s effects appear to depend on conversion to noradrenaline or adrenaline, as a-adrenoceptor blockade with phentolamine disrupts the GH response to L-dopa (199).

Muscarinic pathways are known to stimulate GH secretion, probably by modulating somatostatinergic tone (200). Pyridostigmine, an indirect agonist which blocks acetylcholinesterase, increases the 24 hour secretion of GH by selectively increasing GH pulse mass (201). On the other hand, the muscarinic antagonist atropine is able to blunt the GH release associated with slow wave sleep (202) and that associated with GHRH administration (203). Passive immunization with anti-somatostatin antibodies abolishes the pyridostigmine induced rise in GH in rats, but not immunization with anti-GHRH antibodies, supporting the central role of somatostatinergic tone in mediating this response (204).

Endorphins and enkephalins are able to stimulate GH secretion in man (205), and blockade with opiate antagonists can attenuate the GH response to exercise (206). Passive immunization against GHRH in rats inhibits GH release in response to an enkephalin analogue, which argues for stimulation of GHRH in response to these compounds (207). This cannot be the only mechanism, however, as the met-enkephalin analogue DAMME is able to increase GH release over and above the levels released during maximal stimulation by a GHRH analogue (208). It is possible that the actions of endogenous opioids occur via an interaction with the GHS-R, as the original GH secretagogues characterised were derived from the enkephalins (162).

As with ACTH/cortisol, the endocannabinoids may also influence the release of GH. Somatotroph cells bear the CB1 receptor (56). The administration of THC for 14 days suppresses the GH secretion in response to hypoglycaemia in healthy human subjects (59). Oddly enough, THC and anandamide appear to have opposed effects on GH levels in ovariectomized rats: THC increases and anandamide decreases GH secretion in this context (209). However, the treatment of anterior pituitary cells in primary culture with THC does not seem to influence the release of GH and prolactin to GHRH and TRH, suggesting that the effects of THC are mediated via the hypothalamus and not directly on the anterior pituitary (210), perhaps by stimulating somatostatin release (211).

Many neuropeptides, including the ones in the following paragraphs, have been shown to influence GH secretion in various contexts. For the most part, however, their physiological role in man is not well characterised.

Circulating GH levels are pulsatile, with high peaks separated by valleys where the GH is undetectable by conventional RIAs or IRMAs (Figure 6). The recent development of sensitive chemiluminescent assays for GH with high frequency sampling and deconvolution analysis has allowed the detailed study of GH secretion. This shows that there are detectable levels of basal GH secretion in the ‘valleys’ (235). On average, there are 10 pulses of GH secretion per day lasting on average 96.4 mins with 128 mins between each pulse (236).

Figure 6. Pulsatility of circulating GH levels in adult men and women.

There is a dynamic interplay of pulsatile GHRH and somatostatin secretion:

However, continuous GHRH administration does not affect the pulsatility of GH secretion (238). Moreover, patients with an inactivating mutation of the GHRH receptor continue to show pulsatile GH secretion, suggesting that somatostatin pulsatility is sufficient to determine GH pulsatility (239).

The technical developments in sensitive detection of GH referred to above have elucidated differences in secretion between men and women. Women have higher mean GH levels throughout the day than men due to higher incremental and maximal GH peak amplitudes (Figure 6), but show no significant difference in GH half-life, interpulse times or pulse frequency (240). The higher basal GH levels may underlie the higher nadir GH levels seen in normal women after GH suppression with oral glucose (241).

Differences in GH secretion patterns between the sexes, with male ‘pulsatile’ secretion versus female ‘continuous’ secretion, can cause different patterns of gene activation in target tissues, e.g. induction of linear growth patterns, gain of body weight, induction of liver enzymes and STAT 5b signalling pathway activity (242).

The secretion rate of GH shows a circadian pattern, with peak rates measured during sleep. These are approximately triple the daytime rate (243). GH secretion is especially associated with slow wave sleep (SWS – stages 3 and 4) (244). The decline in GH secretion during aging is paralleled by the decreasing proportion of time spent in SWS, although it is unclear which is cause and which is effect (245). In early data from a clinical trial, GH deficient patients have increased sleep fragmentation and decreased total sleep time, and it is conjectured that such alterations in sleep patterns may be responsible for excessive daytime sleepiness in such patients (246).

Sleep deprivation, in the laboratory or due to travel causing ‘jet lag’, causes two alterations in the GH secretory pattern: the magnitude of secretory spikes is augmented: the return to pre-travel levels takes at least 11 days and is slower to recover after westward travel. The major pulse of GH secretion occurring in early sleep is also shifted to late sleep (247).

Adminstration of a GHRH antagonist reduces the nocturnal GH pulsatility by 75% (248). Normal subjects remain sensitive to GHRH boluses during the night, however, and the lowering of somatostatinergic tone during the night may be responsible for the increase in GH secretion rate (249). Recent work, however, has also demonstrated that ghrelin levels rise through the night in lean men (250). It is likely, therefore, that a combination of increased GHRH, decreased somatostatin and increased ghrelin levels underlie the circadian variation in GH secretion.

Adminstration of GHRH augments the increased nocturnal GH release and promotes SWS. Somatostatin does not change nocturnal GH release, does not affect the proportion of SWS but may increase rapid eye movement (REM) sleep density (251). Ghrelin has been shown to promote slow wave sleep at the expense of REM sleep, accompanied by an increase in GH and prolactin release when administered exogenously (252).

Exercise is a powerful stimulus to secretion of GH (253), which occurs by about 15 min from the start of exercise (254). The kinetics may vary between subjects, an effect which is likely to be related to differences in age, sex and body composition (255). Ten minutes of high-intensity exercise is required to stimulate a significant rise in GH (256). Anaerobic exercise causes a larger release of GH than aerobic exercise of the same duration (257). Acetylcholine, adrenaline, noradrenaline and endogenous opioids have been implicated in exercise-induced GH release (200). However, ghrelin levels do not rise in acute exercise, indicating that ghrelin may not have a role to play in exercise-induced GH release (258).

Insulin-induced hypoglycaemia is another powerful stimulus to GH secretion (Figure 7) (259, 260). The peak GH levels achieved during insulin stress testing correlate well with those achieved during slow wave sleep (261). The hypoglycaemic response is mediated by a2-adrenergic receptors (262) to cause inhibition of somatostatin release (200), although other evidence argues for a role of stimulated GHRH release, as a GHRH receptor antagonist significantly suppressed hypoglycaemic GH release (263). Ghrelin is unlikely to be involved in the GH response to insulin-induced hypoglycaemia as ghrelin levels are suppressed by the insulin bolus (264).

Figure 7. Normal response of GH to insulin-induced hypoglycaemia (≤2.2 mmol/l). Peak GH secreted is ≥20 mU/l.

Other physical stresses such as hypovolaemic shock (265) and elective surgery (266) cause increased GH release. a-adrenergic dependent mechanisms are thought to underly this, as blockade with phentolamine inhibits the response (266).

In contrast to hypoglycaemia, ingestion of an oral glucose load causes an initial suppression of plasma GH levels for 1-3 hours (Figure 8), followed by a rise in GH concentrations at 3-5 hours (267). The initial suppression could be mediated by increased somatostatin release as pyridostigmine, a postulated inhibitor of somatostatin release, blocks this suppression (268). Circulating ghrelin levels also fall following ingestion of glucose (269). The GH response to ghrelin and GHRH infusions is blunted by oral glucose, an effect that is probably mediated by somatostatin (270). The later rise in GH levels is postulated to be due to a decline in somatostatinergic tone plus a reciprocal increase in GHRH, leading to a ‘rebound’ rise (200).

Figure 8. GH response to 75g oral glucose in 8 non-acromegalic, non-diabetic women, given at time 0. Error bars denote SD. Note the high variability of the baseline GH level due to the pulsatile nature of GH secretion. GH levels fall to <0.5 mU/l at 120 minutes.

In type I diabetes mellitus, GH dynamics are disordered, with elevated 24 hour release of GH (271). Deconvolution analysis shows that GH pulse frequencies and maximal amplitudes are increased. The latter is accounted for by higher ‘valley’ levels (272). Better glycaemic control appears to normalize these disordered dynamics (273). The pathophysiological mechanism appears to involve reduced somatostatinergic tone (200).

There is conflicting evidence for increased, decreased or normal GH dynamics in type II diabetics. It is likely that this reflects two factors acting in opposite directions: (1) the confounding factor of obesity in these patients, which leads to hyposecretion of GH; and (2) the hyperglycaemia, which leads to hypersecretion (200).

Chronic malnutrition states such as marasmus and kwashiorkor cause a rise in GH levels (274). A voluntary 5-day fast also leads to significant increases in discrete GH pulse frequency, 24 hour integrated GH concentration and maximal pulse amplitude (275). On the other hand, obesity is known to be associated with lower GH levels, partially due to decreased levels of GH binding protein and partially due to decreased frequency of GH pulses (276). Visceral adiposity, as assessed by CT scanning and dual energy X-ray absorptiometry, seems to be especially important, and correlates negatively with mean 24 hour GH concentrations (277).

The mechanism of decreased GH release in obesity has been ascribed to increased somatostatinergic tone, as pyridostigmine is able to reverse this, to some extent, by suppressing somatostatin release (278-280). However, this cannot be the full explanation, as pyridostigmine is not able to fully reverse the hyposomatotropinism of obesity, even when combined with GHRH and the GH secretagogue GHRP-6 (281).

Although leptin has been shown to be influential on GH secretion in rats (193), this may not be so in humans. Leptin-deficient subjects have been compared with obese non-deficient control subjects in their GH responses when stimulated with GHRH plus GHRP-6. Both these groups have decreased GH peaks compared to non-obese control subjects, as expected. There was no significant difference in mean GH peaks between leptin-deficient and leptin-replete controls, suggesting that leptin does not play a significant role in the GH suppression seen in obese humans, and that the lower GH secretion of obesity is mediated via other mechanisms (282).

Another candidate for the mechanism linking obesity to GH secretion is ghrelin. Its levels correlate negatively with body fat content (283). A comparative study between 5 lean and 5 obese men employed rapid sampling and pulse analysis of ghrelin levels over 24 hours. Ghrelin levels increased at night in the lean controls but did not in the obese group (250). Weight loss caused circulating ghrelin levels to rise in two studies (284, 285). Contradicting this, however, Lindeman and colleagues found that ghrelin levels paradoxically correlated positively with visceral fat area, in contrast with 24 hour GH secretion, which correlated negatively. Moreover, in their study, weight loss increased GH secretion but did not affect ghrelin levels (286). The role of ghrelin in linking nutritional status to GH secretion is therefore at present unclear.

GH release is stimulated by a protein meal (287). L-arginine, an essential amino acid, can be used as a provocative test for GH secretion (288). Evidence that L-arginine acts through inhibition of somatostatin release includes the observation that L-arginine can still enhance the GH response to GHRH despite the use of maximal doses of GHRH (289). However, a specific GHRH antagonist blunted the GH response to L-arginine, an observation that supports the notion that L-arginine also acts through stimulation of GHRH secretion (263). Unlike oral glucose, L-arginine does not modify the GH response to ghrelin infusion (270).