Neuroendocrinology of Mood Disorders

Florian Holsboer

The term endocrinological psychiatry was coined by Laignel-Lavastine, who first used it at a psychiatric congress in Dijon in 1908 when he was encouraging his colleagues to intensify research on the interaction between personality and the endocrine system. Later, Manfred Bleuler in Switzerland established a school, and it was here that psychopathological changes in endocrine diseases such as acromegaly, Graves' disease, Cushing's syndrome, and Addison's disease were first documented as being secondary to endocrinopathy.

These seminal works are still valid and continue to fertilize psychiatric research even today. From the early studies, we learned that the brain is an endocrine target and that compounds such as hormones and pharmaceutical agents that enter the brain can induce changes in mood and behavior. Thus, the concept of endocrinological psychiatry paved the way for attempts to influence brain function with drugs, although early hormone treatments of psychiatric disorders were not successful at all. Interest has now shifted from the mental symptoms associated with endocrine disorders to the other end of the continuum, the focus now being on the endocrinological symptoms that emerge as part of psychiatric disorders. The landmark works by Gibbons, Sachar, Stokes, Rubin, Prange and other pioneers have triggered an enormous number of studies explicitly describing the nature of neuroendocrinological changes under baseline conditions and after specific probes. The rapid progress in biotechnology has opened up the possibility of translating these observations into questions that can be readily pursued in basic research laboratories. Thus, experimental neuroendocrinological research in mood disorders encompasses both clinical and basic research strategies, which in tandem promise to narrow the gap between "bench and bedside." This chapter does not attempt to be exhaustive, but rather uses selected examples to illustrate this process (see Thyrotropin-Releasing Hormone: Focus on Basic Neurobiology, Corticotropin-Releasing Factor: Physiology, Pharmacology, and Role in Central Nervous System and Immune Disorders, Neuroendocrine Interactions, Stress, and The Role of Acetylcholine Mechanisms in Mood Disorders for related topics).

The idea that those individuals with a genetic predisposition to a mood disorder who have also experienced life events are more likely to develop depression than those who have not dates back to Kraepelin, who also postulated that once triggered the disease may then progress independently of psychosocial stressors. More recently, epidemiological studies have corroborated this view; in addition a wealth of preclinical studies have emerged that have focused on acute and long-term consequences of psychosocial stress upon regulatory processes in specific brain circuitries that are commonly linked to the etiology of mood disorders.

Classically, stress is defined as a threatening of homeostasis to which the organism, in order to survive, responds with a large number of adaptive responses. According to work by Cannon, this mainly implicates the sympathetic nervous system and the hormones of the adrenal medulla. Selye was the first to suggest that neuroendocrine factors play a decisive role, and he considered the pituitary–adrenal system to be the major organizer of nonspecific responses to stress. Any type of emotional (cognitive) or physical (noncognitive, e.g., inflammatory) stressor sets into motion a cascade of processes that fine-tune the adaptive response according to specific demands. Centrally, sympathetic pathways are activated to allow for enhanced alertness and focused attention, whereas vegetative functions such as feeding, sleep, and sexual drive are decreased. Peripherally, the humoral and neural systems support the most pressing requirements by elevating heart rate, blood pressure, respiratory rate, and gluconeogenesis.

An integral part of adaptation to a stressor is the protection of the organism against an overreaction and curtailment of the response following termination of stress exposure. If the organism is incapable of terminating the response to stress at the end of stress exposure or if it is exposed to chronic stress, then the adaptive response can lead to pathological changes. The core hypothesis of this chapter is that counter-regulatory mechanisms are essential if overreaction is to be prevented, and that termination of the stress response is defunct in affective disorders. Defects in the counter-regulatory mechanisms may be either genetically encoded or acquired during premorbid life, or they may be a scar imprinted by previous disease episodes. Whatever their origin, there is plausible evidence that changes in stress-adaptive mechanisms are involved in the development, treatment, and prevention of mood disorders.

Patients with depression exhibit increased hypothalamic–pituitary–adrenocortical (HPA) activity, as evidenced by an increase in the number of adrenocorticotropic hormone (ACTH) secretory episodes and an increase in the magnitude of cortisol secretory episodes. This HPA overactivity is further reflected in elevated urinary "free" cortisol (UFC) levels, which appear to be about twice as high in depressed patients as in normal controls, but lower than in patients with Cushing's syndrome. Furthermore, salivary and cerebrospinal fluid (CSF) concentrations of cortisol, which represent the free (i.e., unbound) fraction of cortisol, are reported to be elevated in depression. Finally, the group led by Nemeroff found that the concentration of corticotropin-releasing hormone (CRH) in the CSF is elevated in depressives, but decreases after electroconvulsive treatment (46).

Whereas clinical neuroendocrine studies and preclinical behavioral studies provide compelling evidence for a key role of elevated CRH in the development of mood disorders (30), it must be noted that the CRH levels in the CSF may not be an appropriate reflection of enhanced hypothalamic CRH drive during stress and depression. A study in which CRH concentrations in CSF were serially collected every 10 min for 4 hr showed decreased CRH levels in the presence of increased plasma ACTH and cortisol concentrations (22). The value of this study is perhaps limited by the small number of subjects (n = 6) and the failure of the subjects to exhibit pituitary adrenocortical characteristics typical for depression, but it points to a possible dissociation of the secretory activity of CRH, ACTH, and cortisol. Another example is the finding in primate CSF that CRH exhibits a peak preceding the cortisol peak by approximately 14 hr and is similarly uncoupled from peripheral HPA indices (21). A similar dissociation exists between plasma levels of ACTH and cortisol, where the degree of temporal association between the hormones is 50% to 60% (36).

There are several obvious explanations for this discrepancy. One is the specificity of the radioimmunoassay used in most studies. If other peptides are released in excess in depression or if the precursor molecule of ACTH, proopiomelanocortin (POMC), is cleaved differently, then the antibody might cross-react with some of these peptides, giving values for ACTH that are too high. Another caveat stems from equating immunoassayable ACTH with the adrenocortical-stimulating activity of ACTH. Poland et al. (51) showed that the amount of cortisol released from dispersed adrenocortical cells corresponds more closely to bioactive ACTH than to the immunoassayable amount of ACTH. One source of variance between plasma ACTH and cortisol levels in depression is the gradual development of adrenal hyperplasia during ongoing HPA overactivity, rendering the zona fasciculata of the adrenocortical gland hypersensitive to the tropic hormone ACTH. This hypothesis, derived by Amsterdam et al. (1) from testing with synthetic ACTH, was recently confirmed by Nemeroff et al. (47), who showed adrenal gland enlargement in major depression by computed tomography. Another possibility is the existence of non-ACTH mechanisms, perhaps involving neural sympathetic factors or humoral factors from the immune system. For example, interleukin-1 may activate the HPA system not only by increasing hypothalamic CRH and pituitary ACTH secretion but also by initiating a direct adrenocortical action, mediated by prostaglandin.

Recent studies with more sophisticated computer-aided deconvolution programs have revealed that the secretory pattern of pituitary hormone release is more complex than previously thought, and such techniques need to be employed in patients with major depression to describe more explicitly the changes in HPA-drive under baseline conditions (see Corticotropin-Releasing Factor: Physiology, Pharmacology, and Role in Central Nervous System and Immune Disorders, Stress, and The Role of Acetylcholine Mechanisms in Mood Disorders).

In healthy subjects, the release of ACTH and cortisol can be suppressed for approximately 24 hr by a single oral dose of 1 to 2 mg dexamethasone, a synthetic glucocorticoid. This suppressive action is less pronounced in patients with major depression, 50% to 70% of whom escape from dexamethasone suppression (72). The phenomenon is not merely a reflection of ACTH and cortisol hypersecretion, as one might infer from the analogy with Cushing's syndrome, and it occurs quite frequently among depressives with normal cortisol secretion at baseline. A number of studies by Carroll (12) concluded that nonsuppression on the dexamethasone suppression test (DST) is specific for the diagnosis of melancholia. Although this conclusion has since proven to be invalid, the use of the DST in psychiatric research still has considerable merit. For example, serial DST monitoring of depressed patients undergoing drug treatment showed that in treatment responders DST nonsuppression gradually turned into suppression. Patients whose DST remained abnormal or who were initially suppressors but became DST nonsuppressors during an observation period had a poorer prognosis. At a long-term follow-up, those depressed patients who were DST suppressors at baseline had a better outcome than the nonsuppressors (27).

The plasma dexamethasone concentrations are highly variable and were initially regarded as a major confounder after observation of a negative correlation between post-DST plasma cortisol and dexamethasone concentrations. However, pharmacokinetic studies of the test drug showed that plasma dexamethasone concentrations in the early biophase, which determine the pharmacodynamic effect, are identical in suppressors and nonsuppressors (26). Furthermore, when dexamethasone is administered intravenously to normal controls, the plasma dexamethasone levels are extremely low at the time when samples are drawn for measurements of plasma cortisol levels. Despite these low plasma dexamethasone levels, the cortisol concentrations are adequately suppressed (27). This set of studies indicates that the low plasma dexamethasone levels, measured at the conventional sampling times, in depressed patients do not account for DST nonsuppression.

After CRH became available for clinical studies, several groups measured the ACTH and cortisol response following injections of this neuropeptide, and they consistently reported that the ACTH response was blunted regardless of whether ovine or human CRH was used (28). In some studies, an inverse relationship was found between baseline cortisol secretion and post-CRH ACTH secretion, and it was concluded that elevated baseline cortisol accounts for ACTH blunting via negative feedback. This view was substantiated by studies in which depressed patients were pretreated with metyrapone, which suppresses cortisol biosynthesis: the subjects had normalized ACTH output after CRH stimulation (30).

These data suggest that circulating cortisol is indeed the main determinant preventing an adequate ACTH response to a CRH challenge. Of course, other factors, such as CRH receptor desensitization of corticotropes, altered processing and storage of ACTH precursors, and alternative processing of POMC may also contribute to this phenomenon. For example, Rupprecht et al. (58) reported a dissociation of the ACTH and b-endorphin responses after CRH in depression. Another dissociation was observed at the adrenocortical level, where aldosterone but not cortisol release was found to be blunted in depression (30). The latter observation is in line with the finding that the ratio of ACTH to cortisol decreases with ongoing HPA excess. The functional hyperplasia is apparently limited to the zona fasciculata of the target cell, which regresses after hypophysectomy, an effect that is counteracted by the injection of ACTH. Finally, endogenous CRH may enhance the adrenal response to endogenous and exogenous ACTH in rats, perhaps through a synergistic action on adrenal blood flow. The role of the intra-adrenal CRH system may also be altered in mood disorders, resulting in altered ACTH-to-cortisol ratios. The dissociation between pituitary and adrenocortical suppression by dexamethasone has also been noted. Young et al. (82) found b-endorphin nonsuppression to be much more frequent than cortisol nonsuppression, which suggests faulty feedback at corticotropes that is not reflected in increased adrenocortical activity.

A surprising finding emerged when patients with depression were given a DST and were then challenged with CRH during the afternoon of the next day. One would expect that inadequately suppressed cortisol levels and dexamethasone together would be additive in blunting ACTH release. In fact, normal volunteers showed an ACTH and cortisol response to CRH that gradually decreased with increasing dosages of dexamethasone before CRH administration (27). In contrast, depressed patients showed a paradoxical pattern insofar as dexamethasone pretreatment resulted in increased ACTH and cortisol responses to CRH despite combined endogenous (elevated cortisol) and exogenous (dexamethasone) glucocorticoid levels (77). This abnormality disappears after successful antidepressant treatment (27, 31). The mechanism underlying this phenomenon is not yet fully elucidated, but vasopressin appears to play a role as a possible synergizer of the CRH effects on corticotropic cells. In dexamethasone-pretreated normal controls even high dosages of CRH do not result in plasma ACTH and cortisol levels like those seen among depressed DST nonsuppressors who were not CRH-stimulated (30). This suggests that CRH alone may not account for DST nonsuppression in depressives. When dexamethasone-pretreated normal controls received vasopressin and CRH in combination, ACTH and cortisol secretory patterns emerged that strongly resemble those seen in depressives (30).

From these data and a vast number of preclinical investigations, suggesting a profound synergy between CRH and vasopressin at the pituitary level, the following hypothesis was derived: In depression as in chronic stress, there is a shift to a gradually intensifying vasopressinergic regulation of the pituitary adrenocortical system. Dexamethasone, which does not bind to corticosteroid-binding globulins, thus exerts its effect on the HPA system primarily at the level of the pituitary corticotropes and does not suppress hypothalamic CRH and vasopressin as effectively as endogenous corticosteroids. A transient corticosteroid receptor desensitization gradually develops, and vasopressin expression, which reacts differently to changes in glucocorticoid concentration than does CRH expression, becomes less effectively suppressed than CRH by circulating corticosteroids. As a result, in a depressed dexamethasone-pretreated patient, secretion of ACTH in response to exogenous CRH would be greater than in a control because of the synergistic interaction of the administered CRH bolus with the larger amount of vasopressin present at corticotropes due to insufficient suppression by dexamethasone.

The idea that there is an increased vasopressinergic drive in depression is also indirectly supported by the observed effect of age upon plasma cortisol levels after a combined dexamethasone-CRH test (78). The older the patient, the more cortisol is released in response to a dexamethasone-CRH test, which agrees with findings that with increasing age vasopressin release is enhanced, possibly because of a gradually decreasing capacity of vasopressin receptors. This effect of age is amplified by excessive physical stress such as marathon running (25).

Recently, the above hypothesis received substantial support from preclinical studies showing that repeated stress activates hypothalamic CRH neurons, resulting in increased vasopressin stores and colocalization in CRH nerve terminals. In these parvocellular neurosecretory neurons of the hypothalamic paraventricular nucleus, repeated stress evokes increased CRH and vasopressin synthesis through afferent excitatory neuronal input overriding negative feedback of glucocorticoids upon CRH and vasopressin biosynthesis (13). Vasopressin and CRH are produced during stress, and depending on the severity and duration of stress more and more CRH neurons start to produce vasopressin. When, secondary to chronic stress in the rat or depression in humans, the enhanced vasopressinergic activity is no longer suppressed by dexamethasone, then a synergy with exogenous CRH may occur (30). Furthermore, additional functional changes in the interaction of glucocorticoid receptors with the regulatory elements of the POMC gene may occur.

Similarly, hypoglycemic stress induced by intravenous insulin results in a blunted ACTH response in depression. However, the amounts of ACTH released are much higher than in the CRH test and would never be achieved by CRH alone. Here again, a synergistic interaction between CRH and vasopressin at the corticotropes is likely to be involved. At present, there is no ready explanation for the depression-related blunted ACTH response to insulin, but this finding is certainly caused by a combination of many different factors. Taken together, the results from these and many other dynamic probes of the HPA system support the view that negative feedback is disturbed in depression. By infusing cortisol in depressed patients and measuring the short-term effect upon b-endorphin and b-lipotropin (both are POMC cleavage products, as mentioned earlier) Young et al. (81) showed a decrease in b-endorphin and b-lipotropin release concomitantly with rapidly increasing plasma cortisol levels, indicating the presence of fast-feedback effects of cortisol upon bendorphin and b-lipotropin. Such a fast feedback could not be demonstrated in depressed patients, which points to a defect at the level of the corticosteroid receptor-mediated inhibition of synthesis and release of ACTH secretagogues. It is presently not known exactly how glucocorticoids exert their suppressive effect upon the expression of genes coding for several proteins that are likely to be involved in negative feedback (65) at all the various levels. However, the data base established so far indicates clearly that negative feedback regulation is defunct at the level of the limbic rather than the pituitary–adrenocortical system.

What is not known is whether the disturbance is primary, at the level of a corticosteroid receptor whose function of fine-tuning ACTH secretion via CRH, vasopressin, and POMC regulation is disturbed, or whether other factors drive expression of CRH and vasopressin and their release into hypophyseal portal blood, resulting in an ACTH and cortisol excess that secondarily decreases corticosteroid receptor capacity and function. The pertinent question of whether the corticosteroid receptor disturbance is the cause or the consequence of a CRH hyper-drive that progresses into a combined CRH–AVP hyper-drive cannot yet be answered. Too many other factors that we are only beginning to identify are likely to play a role. One of these is the excitatory amino acids involved in a variety of physiological and pathophysiological processes and also in the neuroendocrine regulation of the HPA system. A further potential modulator of CRH in the hypothalamus is nitric oxide.

Another interesting aspect that has received renewed attention is the possibility of corticotropin-inhibiting factor (CIF) action. Recently, Kellner et al. (34) observed that atrial natriuretic factor (ANF) can suppress the CRH-elicited release of ACTH in normal human controls, providing further evidence for what had been suggested by animal studies (19). These clinical and basic studies demonstrating in vivo and in vitro inhibition of ACTH as well as the neuroanatomical evidence that ANF neurons project to the external layer of the median eminence suggest that ANF may indeed represent a physiologically relevant corticotropin-inhibiting factor.

Still another possible modulator is the recently discovered CRH-binding protein (53), which may modify the synaptic and hormonal actions of CRH at selected sites in the central nervous system and the pituitary. Such interactions may be relevant in the pituitary, where this binding protein is extensively expressed and where the effects of centrally released CRH may be blunted under certain conditions. Future pharmacological research should attempt to clarify whether CRH-binding protein can aid in designing peptides that are capable of blocking the action of CRH. Such compounds, as well as nonpeptidergic drugs, that mimic the ANF effects, would be of potential therapeutic significance. Finally, the recent cloning of the CRH receptor promises the design of drugs that inhibit the potentially anxiogenic effect of hypersecreted CRH.

In response to stressful environmental signals processed by the nervous system, several hormonal adaptations occur that modify the function of specific brain areas. Two principal mechanisms, by which corticosteroids act on neurons, are involved: (a) genomic actions, where the steroid enters the nervous cell and binds to cognate receptors that transform into transcription factors, enhancing or suppressing expression of steroid hormone-regulated genes; and (b) nongenomic actions, where steroids bind to sites at synaptic membranes, affecting ion conductance. Steroids that exert their activity on neurons primarily at membrane sites are often called neuroactive steroids, although neuroactivity of steroids can be conferred through genomic actions, too (48).

The central role of corticosteroids in the maintenance of basic functions and in adaptation and survival under stressful conditions has led to the development of two distinct receptor systems, the mineralocorticoid receptors (MRs) and the glucocorticoid receptors (GRs) (14). The need for this dual system becomes evident if one considers that in a healthy individual plasma concentrations of cortisol may undergo circadian fluctuations from 0.5 nM to 50 nM, and under stressful conditions the maximum may well be above 150 nM. An adequate physiological response to such a wide range of hormone concentrations could not be achieved by a single receptor system only (18). However, two types of receptor, the MRs with high affinity to cortisol and the GRs with low affinity, provide sufficient control over tonic (MR) and stress response (GR) mechanisms in the hippocampus (20). Under baseline conditions the MRs are 90% occupied by cortisol (or corticosterone in the rat), whereas the GRs are only 50% occupied. In the early morning, when circadian HPA activation occurs, or following stress the GRs become more fully occupied in order to curtail HPA-activating mechanisms. Under pathological conditions, for example, severe major depression or Cushing's syndrome, continuing corticosteroid hypersecretion leads to overexposure of GRs and MRs which, in turn, are down-regulated. In nonhuman primates, the hippocampal formation plays an important role in the inhibition of the pituitary–adrenocortical system (61), and sustained stress results in progressive pyramidal neurodegeneration, mainly in the CA3 region. This morphological change results in neuroendocrine changes, because the decreased number of corticosteroid-receptor–bearing neurons in the hippocampus results in a weakened capacity to shut off stress-elevated plasma glucocorticoid levels, thus propelling forward the neurodegenerative effect in this brain area. It is important to note that glucocorticoids are not generally neurotoxic but rather create a condition that makes neurons less able to survive coincident insults such as hypoglycemia, hypoxia, and excitatory amino acids. Another hippocampal region, the granule cells of the dentate gyrus, needs the presence of corticosteroids to survive, underscoring the complexity of GR-neuronal interaction (64). Glucocorticoid receptor density in brain regions other than the hippocampus is also sensitive to chronic overexposure, and a fine-tuned mechanism maintains the capacity of the acute stress response even in the presence of chronic stress. The nature of this particular mechanism is still unknown but may involve a pathway proximal to the CRH–vasopressin neurons in the hypothalamus.

The implications of the dual corticosteroid receptor system for affective disorders are manifold. Administration of glucocorticoids to normal volunteers results in cognitive impairment, and, given the important role of the hippocampus in maximizing memory processes, it is tempting to speculate that excessive corticosteroid exposure decreases the effectiveness of the hippocampus in filtering out behaviorally irrelevant stimuli and maintaining selective attention. The inability to discriminate relevant, important information from irrelevant or unimportant information appears to be a common neuropsychological disturbance in various forms of hypercortisolemia, regardless of whether they are due to Cushing's syndrome or major depression. Yet no specific attribution of steroid effects on neuropsychological function is available. It is also of note that the hypercortisolemia in major depression is driven by central processes such as hypersecretion of CRH, vasopressin, and possibly other ACTH secretagogues. Effects of exogenously administered synthetic corticosteroids or of those corticosteroids secondary to Cushing's syndrome suppress these neuropeptides, which makes comparison of neuropsychological findings between drug-induced Cushing's syndrome, Cushing's disease, and major depression valuable only to a very limited extent.

Further evidence for a corticosteroid-receptor–mediated behavioral effect comes from sleep EEG studies with several corticosteroid probes. In general, these studies found that cortisol enhances slow-wave sleep (SWS) and suppresses rapid-eye-movement (REM) sleep, which suggests that GRs mediate REM sleep and MRs mediate SWS (8). The latter conclusion is derived from the SWS suppressive effect of canrenoate, an MR antagonist. It is not yet clear which mechanisms are involved, particularly because cortisol administration during sleep enhances GH release and suppresses central CRH and vasopressin by negative feedback. Corticotropin-releasing hormone suppresses SWS (29), vasopressin suppresses REM sleep (9), and growth hormone-releasing hormone (GHRH) stimulates SWS (68), suggesting that the effects found following steroid administration are mediated through their effects on the expression and release of neuropeptides, whose involvement in sleep regulation is firmly established (16).

In view of the evidence for hypercortisolism as the precipitator of affective symptoms in patients with Cushing's syndrome, it seems plausible to ask whether the overexposure of the brain to corticosteroids is causally related to the development of affective disorders. Corticosteroids have profound effects on the biochemistry and survival of neuronal brain cells, which may culminate in changes detectable by cranial computed tomography (CT) in HPA-altered depressives (62). If the total quantity of glucocorticoids released is enhanced and if the composition of various corticosteroids other than cortisol is profoundly altered, the pertinent question becomes whether suppression of corticosteroids would be a straightforward approach to treating depression. Some small exploratory studies suggest such a possibility (44). Because of the adverse effects of the drugs employed (metyrapone, aminogluthetionide, and the potentially dangerous drug ketokonazol) this approach cannot be recommended at present as either the sole or an adjunct treatment for depression.

The considerations regarding cortisol-suppressive therapies are based on the assumption that excessive corticosteroids precipitate mood disorders by cytosolic receptor-mediated effects. The cortisol-suppressive treatments used so far disturb the negative feedback, which then leads to enhanced secretion of CRH, vasopressin, and other ACTH secretagogues, all of which have behavioral and possibly anxiogenic or depressogenic effects of their own. The use of selective antagonists of MR and GR at the start of treatment with antidepressants (which later readjust the availability of these two corticosteroid receptors) appears to be the most promising approach for future studies.

In addition to excessive total adrenocortical hormone output in major depression, changes in adrenocortical steroid metabolism have been reported repeatedly (44). These changes have attracted renewed interest because of the recently reported existence of nongenomic effects of a certain class of steroids, termed neuroactive steroids (40, 48).

The first behavioral observations related to these steroids date back to Selye, who over 50 years ago reported that progesterone and deoxycorticosterone (DOC) have a strong sedative action through their A-ring–reduced metabolites. These two steroids, termed allopregnanolone (THP) and allotetrahydrodeoxycorticosterone (THDOC), bind at g-aminobutyric acidA (GABAA) receptors to enhance GABA-induced chloride currents. In rats, THDOC and THP are elevated in cortical and hypothalamic tissue after stress (48), and they have been shown to be anxiolytic and hypnotic, respectively, as predicted by electrophysiology, where a benzodiazepine-like action was demonstrated. Administration of DOC to normal human controls did not evoke effects upon the sleep electroencephalogram (EEG) that would suggest a hypnotic effect after hepatic or central metabolization into THDOC (70). Furthermore, metyrapone treatments, which elevate DOC and THDOC by C-11 steroidhydroxylase inhibition, were not reported to be hypnotic.

Several other neuroactive steroids have the opposite effect. For example, the sulfated form of pregnenolone has been observed to be proconvulsant, as one would expect from electrophysiological experiments, where this steroid has been reported to antagonize GABAA-receptor–mediated chloride currents by reducing the channel-opening frequency (40). Steiger et al. (69) administered pregnenolone to normal healthy controls and analyzed the sleep EEG changes. Pregnenolone increased SWS and decreased sigma power without altering delta power. These spectral analytical data and the absence of any in vivo or in vitro evidence for genomic effects of pregnenolone at GRs or MRs suggest that this steroid acts as a partial inverse agonist at the GABA receptor complex. If this holds true, several clinical implications emerge: Drugs with similar pharmacological profiles (e.g., b-carbolines) are currently being tested as potential memory enhancers, and pregnenolone has already been found to be memory-enhancing in mice (20). Furthermore, the group led by Baulieu (6) discovered a biosynthetic pathway of steroidal compounds in the central neurons and glial cells, possibly rendering the brain independent of peripheral sources. If the bioavailability of pregnenolone is endangered by other drugs such as centrally acting cholesterol synthesis inhibitors, then this might be an explanation for the frequently occurring mental disturbances in patients treated with such compounds.

Whereas it seems attractive to differentiate genomic from nongenomic actions upon neurons, a recent study by Rupprecht et al. (59) showed that transitions may exist between these two modes of action. Steroids that are believed to have limited effects at membrane sites, such as THP and THDOC, can be oxidized intracellularly and then exert genomic actions through progesterone receptors. Thus, the steroid molecule provides a rather flexible structure, which can be modified depending on the tissue to satisfy specific demands (see GABA and Glycine).

Longitudinal studies of patients with depression showed that not only psychopathology but also symptoms of HPA dysfunction respond to antidepressants, and there is a temporal association between changes in mood and hormones. Specifically, in DST nonsuppressors normalization of the suppression response is associated with good treatment response, whereas if plasma cortisol levels remain refractory to dexamethasone suppression this indicates either persistent depression or liability to relapse (27). Studies with more sensitive neuroendocrine function tests such as the combined dexamethasone–CRH challenge yielded similar results and showed that neuroendocrine changes precede psychopathological changes toward both remission and relapse (31). These findings suggested a causal role of HPA alterations in the pathogenesis of depression and challenged several research groups to investigate whether antidepressants elevate mood in depressives through their long-term effects on HPA regulation.

During chronic stress in animals and possibly also in patients with major depression secretion of CRH and the activity of the locus coeruleus (LC) are increased. Both systems are reciprocally innervated, since tyrosine hydroxylase- (TH-, the rate-limiting enzyme in the norepinephrine biosynthesis) positive cells and processes overlap CRH-immunoreactive fibers. This was functionally confirmed by Valentino et al. (74), who showed that CRH applied to LC neurons increases the spontaneous discharge rate of LC cells. This experiment and many others strongly suggested that noradrenergic fibers, projecting from the LC to the paraventricular nucleus (PVN), are stimulators of CRH, which in turn serves as a neurotransmitter to activate the LC. In fact, acute and chronic stress increase immunoreactivity to CRH in the PVN, and infusion of CRH directly into the LC increases catecholamine levels in the cerebral cortex, plasma cortisol levels, and certain stress-related behaviors (11). When rats were treated with imipramine (5 mg/kg i.p.) for 8 weeks, the CRH-mRNA levels were decreased by 37% in the PVN. Moreover, the induction TH was reduced in the LC. Thus, the mutual activation of the two brain centers (PVN and LC) that coordinate the response to stress is dampened through antidepressants. Reul et al. (56) studied the time course of corticosteroid receptor concentration changes in rats treated with amitriptyline or moclobemide (a reversible inhibitor of monoamine oxidase-A). They focused on both MRs and GRs, which, as detailed earlier, dually control HPA system activity and whose equilibrium is possibly defunct in mood disorders. Both amitriptyline and moclobemide transiently elevated hippocampal MR by 40% to 70% between 2 and 5 weeks after the start of treatment. At 7 weeks, the increments in MR had largely disappeared. Hypothalamic GR was increased by 20% to 25% at 5 weeks of treatment, suggesting that antidepressant-induced changes in brain corticosteroid capacity may underlie the observed decrease in circulating ACTH and corticosterone levels and the decreased adrenal size. Furthermore, when challenged by a stressor, rats treated with antidepressants showed a decreased ACTH and corticosterone response, which is understandable in terms of an attenuated LC-CRH response, possibly through enhanced effectiveness of negative corticosteroid feedback from reestablished GR and MR capacity (56).

To investigate more conclusively the hypothesis that antidepressants modulate GR, Barden's group first showed that different types of antidepressants alter GR mRNA levels, a process that appears to be unrelated to monoamine uptake because increased activity of GR gene transcription is seen in mouse fibroblast cells, which do not contain any catecholamines (49). Based on the hypothesis that the apparent lack of sensitivity to corticosteroids seen in the majority of depressives results in neuroendocrine changes, which are causally linked both to pathogenesis and to the therapeutic effectiveness of antidepressant drugs, Pepin et al. (50) created a transgenic mouse with precisely this neuroendocrine deficit. An antisense RNA complementary to the 3¢ noncoding region of the GR mRNA was inserted into the mouse genome, producing an animal in which the GR gene expression is partially knocked out by formation of GR antisense RNA. These mice have increased HPA activity, a resistance to dexamethasone suppression, lethargy, and feeding disturbances, which support the idea that they are appropriate animal models for studying the neuroendocrine symptoms of depression. Treatment of these mice with antidepressants resulted in a partial reversal of enhanced ACTH and corticosteroid secretion and concurrently increased GR mRNA and steroid receptor binding. Moreover, the mRNA levels of hippocampal MR increase after long-term treatment with antidepressants (10).

The precise molecular mechanism of action of antidepressant drugs on corticosteroid receptor gene expression is unknown but may involve more basic mechanisms than synaptic transmission. For example, G-proteins, which play a key role in coupling receptors to intracellular effector systems, are targets of glucocorticoid action in the CNS. In the rat cortex, mRNA levels of stimulatory G-proteins (Gsa) are increased and the level of the inhibitory subunit (Gia) is decreased (60). This may have implications for treatment with tricyclics, which decrease Gsa and, to a lesser extent, Gia (38). Because corticosteroids modulate the actions of other transmitters that act through cyclic adenosine monophosphate (cAMP), they have been termed "permissive hormones." It is therefore very likely that the observed clinical effects upon psychopathological and neuroendocrine symptoms of depression and the involvement of these hormones at many antidepressant-induced levels of action (membrane sites, signal transduction, nuclear gene transcription) are functionally interrelated.

The effect of thyroid hormones upon mood and behavior is documented by dysphoria, anxiety, fatigue, and irritability in hyperthyroidism and by impairment of cognitive functions in milder states of hypothyroidism. In psychiatric populations, regardless of specific diagnoses, thyroid dysfunction is more common than in the general population, and among patients with mood disorders 20% to 30% exhibit some form of hypothalamic–pituitary–thyroid (HPT) abnormality. The HPT system has a hierarchical structure similar to that of the HPA system, with thyrotropin-releasing hormone (TRH) as the hypothalamic master hormone that is released from nerve endings in the median eminence, from where it enters the anterior pituitary through the portal system. There, it induces synthesis and release of thyreotropin (TSH), which enters the circulation and causes the release of the two major thyroid hormones, triiodothyronine (T3) and thyroxine (T4). Thyroid hormones feed back at the hypothalamus to inhibit TRH release and at the anterior pituitary to inhibit TSH release. Of course, many other factors are involved in fine-tuning this circuit. Among these, somatostatin (SRIF) and HPA hormones play an important role as inhibitors of HPT activity.

In the aggregate, studies that employed a TRH stimulation test in depressed patients reported that the TSH response was blunted in approximately 25% of the cases in the presence of normal T3, T4, and TSH levels at baseline (55). When Duval et al. performed TRH tests with 200 mg TRH at both 8 a.m. and 11 p.m., they found that stimulation in the evening produced greater DTSH differences between patients and controls than stimulation in the morning (15). A parameter called DDTSH (which is defined as the difference between morning and evening TSH responses) proved to detect HPT abnormality at a sensitivity rate of 89% among patients with major depression. Another indication of altered TSH secretion is the loss of the nocturnal TSH rise, which usually occurs between midnight and 3 a.m. The absence of a nocturnal surge in TSH is believed to be a more sensitive indicator of an HPT abnormality than a blunted TSH response to TRH (5). One putative mechanism to explain the decreased TSH secretion would be chronic hypersecretion of hypothalamic TRH, which would decrease the number of pituitary TRH receptors. Such a hypothesis is consonant with the finding of increased TRH in the CSF of depressed patients. The possibility that SRIF antagonizes TRH effects can be ruled out because SRIF levels in the CSF were decreased rather than elevated (57). Also, a possible inhibitory effect of thyroid hormones is not likely because serum total and free thyroid levels were reduced rather than elevated among depressives (55). It is well known that glucocorticoids inhibit TSH secretion, and the mean 24-hr serum TSH levels were fully suppressed in response to dexamethasone. In depressives, nighttime plasma cortisol and TSH levels were inversely related (5) and the TSH response to TRH was found to be linearly correlated with the ACTH response to CRH (28), suggesting that HPA activity is a modulator of spontaneous or specifically stimulated TSH surges, or both. Approximately 16% of depressed patients in remission show blunted TSH responses, but in the absence of TRH test results for premorbid patients it remains an enigma whether the TRH blunting is a true trait marker or a neuroendocrine scar acquired from preceding depressive episodes. Nevertheless, persistent TRH blunting appears to be associated with an increased likelihood of relapse (see also Thyrotropin-Releasing Hormone: Focus on Basic Neurobiology).

Several groups (7, 33) found that antidepressant treatment reduced plasma T4 levels and that responders had more pronounced reduction in plasma T4 and free T4 than nonresponders. Joffe and Singer (33) compared the potential clinical effects of T3 and T4 as adjuvants to treatment with antidepressants and found T3 to be superior to T4. Furthermore, patients with panic disorder who were treated with antidepressants appeared to profit from adjunctive T3 treatment (71). The working hypothesis formulated by Joffe and Singer (33) is that the effect of antidepressants includes reduced exposure of central neurons to thyroid hormone. The observed amplification of treatment effects by T3 does not contradict this hypothesis of a central hyperthyroidism in mood disorders. Brain cells utilize T3, which is derived from T4 by deiodination. Exogenous T3 would increase plasma T3, but, through negative feedback of T3 upon TSH, secretion of both T3 and T4 would be decreased. Thus, T3 administration would lead to decreased central bioavailability of T4, the major source of neuronal thyroid demands, and central hyperthyroidism would be compensated in this way. The antidepressant effect and possibly also the prophylactic effect of carbamazepine involves an antithyroid effect, too, because peripheral T3 and T4 decrease under this drug, whereas TSH increases.

A substantial portion of patients with mood disorders have elevated TSH at baseline and following TRH stimulation tests, but normal thyroid hormone concentration. This condition defines subclinical hypothyroidism and represents a considerable risk for development of depression. Several early studies suggested that one underlying mechanism is a symptomless autoimmune thyroiditis. Nemeroff et al. (45) observed that 20% of depressed patients had antithyroid antibodies; in the normal population, the figure is only 5% to 10%. Patients with bipolar disorder have a higher rate of symptomless autoimmune thyroiditis than those with unipolar depression, and rapidly cycling bipolar patients have a particularly high rate of hypothyroidism.

The discovery by Mason et al. (41) that endogenous T3 is concentrated in presynaptic nerve terminals, where the cellular uptake is achieved and where the synthesis of T3 by deiodination of T4 takes place, points to a possible role of T3 as a neurotransmitter. In addition, the possibility that T3 neurotransmitter receptor regulation is defunct in patients with affective disorders needs to be considered. In attention-deficit hyperactivity disorder (characterized by impulsiveness, inattention, aggressiveness, intrusiveness, destructiveness, hyperactivity and purposeless motor behavior and formerly termed "minimal brain damage") such a thyroid hormone receptor deficit may be a causative factor. In a study by Hauser et al. (24), 50% of the patients with generalized resistance to thyroid hormones had an inherited (usually autosomal dominant) disorder caused by a mutation in the thyroid receptor b-gene and met the criteria for attention-deficit hyperactivity disorder as children. The mutant human thyroid receptors exert hormone resistance through a lower binding affinity and subsequently through a lower transcriptional activation of positive T3 response elements than found in normal receptors.

In light of the many findings on changes in thyroid regulation in mood disorders, it seems plausible that disturbed thyroid receptor-mediated signal transduction, possibly in concert with corticosteroid receptor dysfunction, is involved in the causation of mood disorders. The antidepressant-induced changes in HPT support this conclusion.

If healthy individuals are exposed to acute stressors they respond with increased secretion of growth hormone (GH), but if they are exposed to mild chronic stressors such as academic examinations this does not provoke continuous GH hypersecretion. In contrast, under testing conditions of mild stress patients with depression show GH secretory patterns that are clearly distinct from those of normal controls. The integrated 24-hr GH values appear to be increased in depression, whereas the GH surge around the time of sleep onset is decreased.

The use of neuropharmacological probes has made it possible to relate altered GH responses to specific neurotransmitter receptor alterations in depression. The most consistent findings have emerged from studies with clonidine, an a2-adrenoceptor agonist, namely that this drug evokes less GH in depressed patients than in controls. The original finding by Matussek et al. (42) prompted numerous studies, from which it was concluded that the reduced GH response to clonidine cannot be accounted for by drug treatment, age, or sex but may be an indicator of noradrenergic dysregulation in affective disorders (63). This view is amplified by the concomitantly decreased response of 3-methoxy-4-hydroxyphenylglycol (MHPG) and cortisol. Furthermore, a number of antidepressants that inhibit presynaptic monoamine transporters are potent GH stimuli, and some studies, but not all, demonstrated that depressed patients have a reduced GH response to these antidepressant drugs.

The earliest challenge test used in mood disorders was the insulin tolerance test, which documented a blunted GH response after insulin-induced hypoglycemia. Initially, this finding, too, was attributed to a disturbed adrenoceptor function in depressed patients. Although recent knowledge about the complex counter-regulatory humoral and neuronal responses to hypoglycemia calls into question a simple mechanistic interpretation of insulin-induced GH secretion in mood disorders, the insulin resistance of depressed patients remains a potentially interesting finding. Amsterdam and Maislin (2) have postulated that patients with bipolar disorders have GH responses to insulin that are distinct from those of unipolar patients and controls, which suggests different neuroendocrine alterations in different mood disorder subtypes.

With the clinical availability of GHRH, the hypothalamic-releasing factor specifically stimulating GH from somatotropic cells, a number of studies investigated the GH response to GHRH in mood disorders. Discordant results emerged: Whereas some studies found a reduced GH response to GHRH in depression (37), others (e.g., ref. 35) did not. One factor possibly confounding the GH response to GHRH and causing inconsistent findings is the level of circulating somatomedin C (SM-C), which is reported to be elevated in depression (37), perhaps as a result of elevated GH secretion. However, Voderholzer et al. (76) recently reported that the baseline level of GH secretory activity does not determine the amount of GH release following GHRH. Hence the question remains open whether blunted responses to GHRH are determined by the secretory activity of GH at baseline. In depressed patients and controls, positive correlations between the GH responses to a2-adrenoceptor agonists and GHRH were reported and interpreted as evidence that the GH response to clonidine is mediated by GHRH. Suri et al. (73) noted that the GH response to GHRH depends on the GH secretory status during the hour prior to challenge. Specifically, the GHRH-induced GH peak was higher when GHRH was administered while plasma GH levels were increasing than while they were decreasing. In this study, clonidine-induced GH surges were not determined by spontaneous GH patterns. Furthermore, pretreatment with clonidine did not augment the peak GH response to GHRH, but pretreatment with GHRH attenuated the peak GH response to clonidine. In another study, clonidine pretreatment increased the number of GH pulses and the total amount of GH released (39). These findings suggest that the a2-adrenoceptor agonist-induced GH stimulation occurs through pathways different from GHRH-induced GH stimulation.

The secretion of GH is regulated by an extremely complex system, in which GHRH, secreted from neurons in the arcuate nucleus, stimulates GH release from the pituitary and somatostatin (SRIF), secreted from neurons in the paraventricular nucleus, suppresses it. In turn, GH secretion stimulates transcription of the SRIF-encoding gene and SRIF secretion inhibits transcription of the GHRH-encoding gene and the release of GHRH into the portal system. However, pulse frequency and the amount of GH released per pulse are determined not only by these reciprocal effects; many other factors are also critically involved. For example, there are at least three receptors that enhance GH secretion from somatotropic cells. Ligands for these receptors are, besides GHRH, the natural but still unknown ligand through which the synthetic GH-releasing peptide (a hexapeptide, developed from an enkephalin analog) activates GH (32) and pituitary adenylate cyclase-activating peptide (PACAP). The latter peptide stimulates somatotropic activity and fulfills the requirement for a hypophysiotropic factor. It releases GH through a receptor system that can be expressed in five different splice variants, which differentially activate two distinct second messenger systems (66). Although somatostatin suppresses these effects, elevated somatostatin secretion is not likely to be the cause of the blunted GH response to GHRH because, as numerous studies have shown, somatostatin tends to be decreased in the CSF of depressed patients and to increase after clinical improvement (57). It is not known, however, whether these CSF findings reflect the secretory activity of hypothalamic somatostatin neurons.

In depression a predominance of CRH over GHRH may develop and as a result somatostatin may be suppressed by enhanced CRH in a pulsatile fashion during the daytime, giving rise to increased GH pulses during the daytime and decreased GH release at night (80). Sleep endocrinological studies provide further support for this hypothesis: when GHRH is centrally injected into rats or intravenously administered to humans in a pulsatile fashion it enhances SWS and suppresses cortisol, whereas CRH suppresses both GH and SWS and elevates ACTH and cortisol, thus mimicking some of the spontaneously occurring events seen in depression (16, 29, 68). These findings raise the possibility that alterations in GH secretion are a phenomenon secondary to enhanced HPA activity. Although cross-sectional correlations of cortisol with other hormones (e.g., thyroid hormones) at baseline or following specific probes render this unlikely, it is of note that the modulating effects of cortisol and other hormones are subject to different, and at least in humans largely unknown, time grids. Thus, cross-sectional correlations will fail to detect causal relationships. Acutely administered corticosteroids increased GH (80), whereas there were changes after 4 days in the secretion profile but not in the overall amount of GH released (54). The amount of GH released during the daytime was increased, whereas the nocturnal sleep-related GH surge was attenuated, resembling the situation seen in mood disorders (43, 67) and after administration of cortisol pulses to healthy men (80). The decreased nocturnal GH secretion after glucocorticoid administration probably involves increased SRIF release and/or negative feedback by elevated plasma GH levels and induction of SM-C (54). In a recent study by Veldhuis et al. (75), a low dosage of dexamethasone (1.5 mg b.i.d.) was administered for 1 week and found to increase integrated 24-hr plasma GH concentrations 250% and SM-C 200%. This increase was achieved by an increase in the number of secretory GH bursts and an increase in the amount of GH released per burst.

Cortisol inhibits the GH response to GHRH acutely after prolonged administration, the latter effect and also being attributed to an increase in glucocorticoid-induced SRIF (79) because pyridostigmine, a suppressor of SRIF, reverses the steroid effects (23). In agreement with this interpretation is the finding that in adrenalectomized rats the number of SRIF receptors decreases, but it normalizes again after glucocorticoid replacement. However, in pituitary cell cultures glucocorticoids desensitize SRIF receptors. Taken together, these studies suggest that the HPA hyperactivity in depression strongly influences both basal and stimulated GH secretion, probably by modulating somatostatin effects. Basic studies document a direct corticosteroid receptor-mediated effect on GH gene expression, and GH gene expression is similarly affected through ligand-activated thyroid receptors (17). In addition to these transcriptional effects on the GH-encoding gene, glucocorticoids also increase the number of pituitary GHRH receptors. Direct proof of the clinical relevance of these findings was provided by Barbarino et al. (4), who injected CRH in combination with GHRH and showed that CRH is capable of suppressing GHRH-elicited GH release, at the same time corroborating a study in which the spontaneous sleep-related GH surge was decreased by CRH administration (29) in a way similar to that seen in depressed patients with hypercortisolemia in the absence of exogenous CRH (68).

GH secretion declines with increasing age, which is best reflected by the near absence of nocturnal GH surges, accompanied by reduced SWS. Whether these effects are also secondary to increased HPA activity in the elderly is not known. In depressed patients, in whom ACTH and cortisol are enhanced and SWS and the associated GH secretion are reduced, GHRH administration does not rectify the endocrine pattern during sleep, although one would expect this from the SWS-enhancing and cortisol-suppressing effects of GHRH administration in healthy men (68).

The behavioral correlates of altered GH regulation in depression are difficult to reconcile. In patients with acromegaly, a number of cognitive and behavioral abnormalities have been noted. The best-documented effects are those on the sleep EEG. Exogenous administration of GH and pathologically hypersecreted GH (as is the case in acromegaly) attenuate SWS, most likely through suppression of GHRH, which is sleep-promoting (16, 68). After surgical removal of the GH-secreting adenoma, the amount of SWS increases (3). Animal studies have linked CRH to anorectic behavior and GHRH to increased food intake. The hypothesized increased CRH-to-GHRH ratio in depression is in keeping with the loss of appetite and weight in these patients.

Studies in which GH is administered to elderly humans or to patients after hypophysectomy have not yet systematically elaborated the potential psychotropic effects of this hormone. Therefore, more such studies should be conducted. Also of interest would be investigations with GH-releasing peptide, which is a long-acting GH releaser acting at receptors different from those for GHRH and PACAP.

The focus of this brief and selective review has been on those clinical studies that provide input for basic research, ranging from animal behavior studies to cellular and molecular experiments. The neuroendocrinology of mood disorders has gone through numerous phases and currently hormonal alterations are considered to be neuroendocrine symptoms that may lead to a deeper understanding of pathogenesis. Clinically this conclusion is drawn from the concomitance of changes in mood and behavior and in hormonal activity. Basic studies strengthen this view by showing how environmental stimuli are translated into changes in neurohormonal activity and how this leads to altered expression of genes coding for proteins that are directly or indirectly involved in mediation of mood and behavior. Likewise, encoded vulnerability for the development of mood disorders is probably subclinically apparent in enhanced HPA responses to specific probes. This inherited sensitization is then further amplified by repeated episodes, thus providing a pathogenetic rationale for early intervention strategies (52). Antidepressants interfere at various levels of hormone regulation, and with the new technology available the hypothesis that they act clinically by rectifying the perturbed HPA system can now be tested. If additional evidence can be accumulated that neuroendocrine alterations are causally involved in mood disorders, this would provide a lead for the development of more efficacious drugs.

published 2000