Neuropeptide Alterations in Mood Disorders

Paul M. Plotsky, Michael J. Owens, and Charles B. Nemeroff

The past several decades have witnessed an explosion of knowledge about the central nervous system (CNS). Dozens of peptide neurotransmitter and neuromodulator candidates have been identified and characterized, their CNS distributions mapped, and their genes cloned. Dale's principle of one neuron–one transmitter has been overturned (22) with numerous demonstrations of neurons containing multiple peptides or combinations of peptides and nonpeptides (32; see also General Overview of Neuropeptides). Additionally, the past ten years have yielded an embarrassment of riches in the form of neurotransmitter receptor diversity, diversity of receptor–effector coupling, and neurotransmitter transporters. These recent discoveries have not yet been fully integrated into our concepts of normal or aberrant CNS function, although dysfunction at any level could conceivably lead to neurological and cognitive disorders. Thus, although there are many choices for discussion, in this chapter, we have chosen to review three of the many peptide neurotransmitter systems: corticotropin-releasing factor (CRF), somatostatin (SRIF), and thyrotropin-releasing hormone (TRH). These have been selected on the basis of clinical interest in the potential consequences of dysregulation of these circuits for the pathophysiology of specific mood disorders. There has accumulated a considerable amount of basic and clinical research on these systems since the publication of the previous volume in 1987, but limitations of this volume prevent extensive citations of the original literature and, thus, the reader is referred to recent reviews.

After a search spanning three decades, CRF was isolated and characterized by Vale's group in 1981 as a 41-amino acid peptide (83; see also Corticotropin-ReleasingFactor: Physiology, Pharmacology, and Role in Central Nervous System and Immune Disorders). Corticotropin-releasing factor is a primary and an obligatory hypothalamic adrenocorticotropic hormone (ACTH) secretagogue in most species (61); it also appears to function as a putative neurotransmitter in a CNS network and may coordinate global responses to stressors. Along with its homologs, CRF represents an ancient family of peptides subserving numerous functions. In the higher organisms, including mammals, evidence supports the hypothesis that CRF plays a complex role in integrating the endocrine, autonomic, immunological, and behavioral responses of an organism to stress (20, 58). Of particular concern in this chapter are the clinical ramifications drawn from the remerging evidence implicating dysregulation of central CRF circuits in association with mood and anxiety disorders, as well as anorexia nervosa and Alzheimer's disease, disorders also commonly associated with comorbid mood alterations. A more complete review of CRF neurobiology, with citations, is presented in Corticotropin-ReleasingFactor: Physiology, Pharmacology, and Role in Central Nervous System and Immune Disorders) and reviews by Dunn and Berridge (20) and Owens and Nemeroff (58).

The synthesis and secretion of glucocorticoids represents the final step in a neuroendocrine cascade beginning in the CNS. Physical and psychological stressors, circadian drive, and humoral influences initiate the cascade by releasing multiple hypothalamic ACTH secretagogues into the hypophyseal–portal circulation (61). Elaborated by perikarya in the hypothalamic parvocellular paraventricular nuclei (pPVN), CRF plays an integral and obligatory role in the regulation of adenohypophyseal ACTH and b-endorphin production and secretion, as well as expression and synthesis of their precursor, proopiomelanocortin (POMC).

The negative feedback loop necessary for hypothalamic–pituitary–adrenal (HPA) axis regulation is completed by ACTH-induced glucocorticoid secretion (16). Many stressors have been shown to increase CRF messenger ribonucleic acid (mRNA) concentrations in the pPVN, whereas glucocorticoids negatively regulate CRF mRNA concentrations in the PVN. It has been suggested that the adaptive function of the HPA axis is critically dependent on glucocorticoid feedback mechanisms to dampen the stressor-induced activation of the HPA axis and to shut off further glucocorticoid secretion. In animal models, exogenous CRF or endogenous hypersecretion of CRF leads to reproductive failure, altered locomotor activity, reduced food intake, sleep disruption, hypercortisolemia, and dexamethasone resistance (20, 58); these effects are similar to the changes observed in patients with major depression.

Preclinical studies by Meaney et al. (50) and Plotsky and Meaney (62) on the long-term consequences of the neonatal environment support the hypothesis that early, severe, or chronic stress may be transduced into neurobiological changes, which then increase an organism's probability of exhibiting clinically significant mood disorders (63, 64). In animal models, chronic elevation of CRF secretion can lead to increased corticotrope numbers (26), which may underlie the observed increase of pituitary size in depressed patients (see ref. 43 or review). Furthermore, transgenic mice, which overproduce CRF (81) or lack glucocorticoid receptors (59) exhibit behavioral and endocrine signs consistent with major clinical depression. Hypersecretion of hypothalamic CRF is clearly associated with down-regulation of adenohypophyseal CRF receptor numbers (30), thus giving rise to a blunted ACTH and normal or exaggerated glucocorticoid responses to exogenous CRF administration (see ref. 68 for review).

Mapping of the distribution of CRF-like activity in the CNS using immunohistochemistry and radioimmunoassay revealed a wide extrahypothalamic distribution compatible with its involvement in stress responses, emotionality, and cognition. In particular, CRF activity has been reported in forebrain limbic areas and in autonomic and viscerosensory brainstem nuclei (see Corticotropin-ReleasingFactor: Physiology, Pharmacology, and Role in Central Nervous System and Immune Disorders).

Consistent with its role as a putative neurotransmitter, central administration of CRF mimics many of the behavioral and autonomic aspects of the stress response (20, 42, 58). Conversely, pretreatment with a specific CRF antagonist attenuates many of the behavioral and autonomic components of the stress response. This may be interpreted as support for the hypothesis that endogenous CRF acts on extrahypophyseal targets to produce autonomic and behavioral components of stress responses. In most, but not all, cases these effects have been demonstrated to be independent of the effects of CRF on HPA function. As reviewed in Dunn and Berridge (20) and Koob et al. (42), CRF has been found to: (a) increase the latency to begin eating in food-deprived animals provided food in a novel setting, (b) increase the acoustic startle response, and (c) decrease punished-responding. In each case, these effects are reversed by benzodiazapine treatment. The CRF receptor antagonist, a-helical CRF9-41, although certainly not an anxiolytic drug, attenuates some, but not all, of the behavioral effects of stressors.

Dysregulation of the HPA axis in major depression remains one of the most consistent findings in biological psychiatry (reviewed in ref. 68). Since the association between HPA axis dysfunction and major depression was made, neuroendocrine function tests have served as a window into the brain in attempts to elucidate mechanisms that underlie the pathophysiology. A subset of patients fulfilling DSM-III-R criteria for major depression often present with a constellation of symptoms including hypercortisolemia, resistance of cortisol to suppression by dexamethasone, blunted ACTH responses to CRF challenge as compared to controls, and elevated CRF concentrations in the cerebrospinal fluid (CSF). The pathological mechanisms underlying HPA axis dysregulation in major depression and other affective disorders remain to be elucidated.

Administration of rat, human, or ovine CRF yields blunted ACTH, b-lipotropin, and b-endorphin secretion with a normal cortisol response in patients with major depression (3, 28, 33, 44, 85). Patients with posttraumatic stress disorder (79), 50% of whom also fulfill DSM-III criteria for major depression, and patients with panic disorder (70) also show blunted ACTH secretion in response to CRF challenge. In patients with major depression, a correlation appears to exist between DST nonsuppression and a blunted ACTH response to CRF challenge (44), suggesting that these features may represent state markers for depression. According to Amsterdam et al. (3), normalization of the ACTH response to CRF challenge occurs following clinical recovery.

Two hypotheses have been advanced to account for the observed blunting of ACTH secretion in response to CRF challenge that is associated with major depression: (a) down-regulation of adenohypophyseal CRF receptors occurs as a result of hypothalamic CRF hypersecretion, and (b) increased glucocorticoid-mediated negative feedback tone at the pituitary or CNS. Although the first hypothesis has not been directly tested in humans by measurement of postmortem hypothalamic CRF secretion or mRNA content or by the assessment of anterior pituitary CRF receptor function or mRNA levels, substantial support has accumulated favoring it. Animal studies (vide supra) show down-regulation of adenohypophyseal CRF receptors under conditions of chronic stress and/or elevated CRF. Postmortem measurements of CNS CRF receptors in the frontal cortex demonstrate down-regulation in suicide victims (56) who were, presumably, suffering from depression. Finally, preliminary data from our laboratory suggests the presence of increased CRF mRNA in the hypothalamus of suicide victims versus age- and sex-matched controls. Less data is available to support the second hypothesis. In all of these studies, neuroendocrine studies represent a secondary measure of CNS activity; pituitary ACTH responses reflect the activity of hypothalamic CRF rather than that of the corticolimbic circuits, which are likely to be involved in the pathophysiology of major depression.

A potentially more direct avenue for evaluation of extrahypothalamic CRF tone may be obtained from measurements of CRF in ventricular or lumbar CSF. A marked dissociation between CSF and plasma neuropeptide concentrations has been described, thus indicating that neuropeptides are secreted directly into CSF from brain tissue as opposed to being derived from plasma-to-CSF transfer (65). Evidence that CRF concentrations in CSF originate from nonhypophyseotropic CRF has been obtained from studies in which CRF concentrations in CSF were repeatedly measured over the course of the day. Moreover, CRF concentrations in rhesus monkey CSF are not entrained with pituitary–adrenal activity (24). The proximity of corticolimbic, brainstem, and spinal CRF neurons to the ventricular system suggests that these areas contribute to the CRF pool found in CSF.

In a series of studies, Nemeroff's group has demonstrated significant elevation of CRF in the CSF of drug-free patients with major depression or following suicide. In the initial study, CRF was measured in the CSF of 10 normal controls, 23 depressed patients, 11 schizophrenics, and 29 demented patients; CSF concentrations of CRF were elevated in the depressed patients compared to all of the other groups (57). In a larger study of CRF in the CSF of 54 depressed patients, 138 neurological controls, 23 schizophrenic patients, and 6 manic patients, the depressed patients exhibited a marked twofold elevation in CSF–CRF concentrations. Collection of postmortem cisternal CSF–CRF concentrations from postmortem depressed suicide victims and sudden death controls also revealed elevated CRF concentrations in the depressed group (5). Risch et al. (67) have confirmed these findings of elevated CSF concentrations of CRF in depressed patients.

Elevation of CSF concentrations of CRF has also been reported in patients with anorexia nervosa (38) and reverts to the normal range as these patients approach normal body weight. No alterations of CSF concentrations of CRF appear to be associated with other psychiatric disorders including mania, panic disorder, and somatization disorders as compared to controls (8, 36).

More critically, Nemeroff et al. (53) have shown that the elevated CRF concentrations in the CSF of depressed patients (versus normal controls) are significantly decreased 24 hr after final electroconvulsive therapy (ECT) treatment indicating that CSF–CRF concentrations, like hypercortisolemia, represent a state rather than a trait marker. Other recent studies have confirmed this normalization of CRF concentrations in the CSF following successful therapeutic intervention. Banki et al. (9) demonstrated significant reduction of elevated CRF in the CSF of 15 female patients with major depression who remained depression-free for at least 6 months following antidepressant treatment as compared to little significant treatment effect on CRF in the CSF of nine patients who relapsed in this 6-month period. The authors suggest that elevated or increasing CSF concentrations of CRF during antidepressant treatment may indicate lack of normalization in major depression despite symptomatic improvement and, thus may predict early relapse. Treatment of depressives with fluoxetine (18) is also reported to reduce CRF concentrations in CSF.

Increased central drive to the pituitary–adrenal axis is associated with an increase in the number and volume of corticotropes as well as adrenal gland hypertrophy (16, 76). Recently, Krishnan (43) reviewed magnetic resonance imaging (MRI) and computed tomography (CT) studies of changes in human pituitary and adrenal cross-sectional area and volume in patients with mood disorders. Two studies reported enlargement of the pituitary gland in depressed patients as compared to age- and sex-matched controls. Furthermore, a significant correlation existed between pituitary gland enlargement and post-dexamethasone cortisol concentrations. In a study of postmortem tissue, Zis and Zis (86) observed adrenal gland enlargement in suicide victims versus controls, a result first observed in an early CT study (4). This has been confirmed in a study comparing adrenal volume by CT in depressed and age- and sex-matched normal controls (55); however, adrenal volume did not correlate with dexamethasone suppression test results, patient age, or the severity or duration of the depressive episode. A study by Rubin et al. (71) confirmed the finding of adrenal enlargement and presented evidence of the state dependency of this change. Correlation of these depression-associated changes in pituitary and adrenal size with the results of CRF stimulation tests and CRF concentrations in CSF would be welcome.

In the relatively short span of a decade, the putative role of CRF has expanded from that of a classical hypophysiotropic factor participating in the regulation of pituitary–adrenal drive to that of a central neurotransmitter implicated in organization of counterregulatory responses to a variety of stressors. This expanded role was first suggested by immunohistochemical mapping studies, which identified a unique extrahypothalamic distribution of CRF and was further supported by electrophysiological, pharmacological, and behavioral work in animals and, subsequently, in clinical studies. From these studies, it is clear that dysfunction of central CRF systems is associated with selected mood disorders. However, neither the exact nature of these dysfunctions nor whether they represent primary or secondary defects is yet clear.

Functional alterations in the HPA axis and central CRF systems have been studied most extensively in patients with major depression. Current evidence leads to the hypothesis of CRF hypersecretion of either hypothalamic or extrahypothalamic origin. Evidence of a hypersecretory state in depression may be inferred from basic and clinical observations of down-regulation of CRF receptors in the adenohypophysis and frontal cortex, increased CRF concentrations in CSF, blunted ACTH response to a standard CRF stimulation test, and enlarged pituitary and adrenal areas and volumes as assessed by MRI or CT scans. Many, if not all, of these abnormalities revert to normal after successful treatment of depression. Therefore, in major depressive illness, the neuroendocrine system serves as a window into the brain, with abnormalities reflecting a state, rather than a trait marker of depression. Interpretation of the cerebrocortical reductions in CRF concentration in brains from patients with Alzheimer's disease is unclear. Enhanced receptor numbers in these brains support the hypothesis of reduced CRF secretion; however, no direct demonstration of degeneration of CRF-containing neurons in this disorder has been reported. Furthermore, as in depression, little work has been performed at the molecular level. Although extrahypothalamic CRF circuits are increasingly accepted as participating in cognitive processes, it is unclear which, if any, symptoms of Alzheimer's disease are secondary to the pathological involvement of CRF neurons in this disease.

Overall, then, many important research avenues remain open with respect to the role of CRF in mood disorders. Postmortem studies of CNS tissue from these patients at the molecular level are in progress in our laboratory and at numerous other sites; the results from these studies are bound to lead to new insights into the pathophysiology of these diseases. However, it will be important to examine tissue from numerous control groups drawn from age- and sex-matched disease-free populations as well as from populations representative of other neurological disease states. The recent cloning of the CRF-binding protein and the CRF receptor provides new tools, both specific molecular probes and antisera, for assessing changes in these entities in various disease states. On the basis of the considerable pharmacological and behavioral work cited above, it is clear that one of the most exciting areas will be the development of long-acting, CNS-potent CRF-like agents. One exciting prospect would be the possibility of using a CRF antagonist or CRF-binding protein analog for the treatment of depression and/or anxiety disorders. Perhaps the recent cloning of the CRF-binding protein and the CRF receptor will aid in the elucidation of the active portion of the peptide or the active site on the receptor. These discoveries may lead to the rational design of lipophilic drugs that will be clinically useful. Finally, design of appropriate ligands for PET scanning would permit evaluation of CRF receptor distribution and, perhaps, affinity during the course of these diseases.

Somatostatin (somatotropin release-inhibiting factor, SRIF), like a number of other neuropeptides, was serendipitously discovered during attempts to purify growth hormone-releasing factor (GRF). As the name implies, somatostatin inhibits the release of growth hormone from the anterior pituitary. Since its structural identification 20 years ago, somatostatin has been unequivocally shown to be the major inhibitory influence on endocrine growth hormone secretion. Additionally, and perhaps of more interest, somatostatin fulfills a number of criteria for status as a neurotransmitter within the CNS. The acceptance of somatostatin's role as a neurotransmitter has led to its investigation in a number of psychiatric and neurological diseases. As will be discussed below, nonendocrine somatostatin neurons may play a role in a number of illnesses including, but not limited to, depression, dementia, and epilepsy (see also Somatostatin in the Central Nervous System).

Prior to discussion of somatostatin's role in mood disorders, we briefly review several aspects of basic somatostatin neurobiology. Interested readers will find a detailed description of somatostatin neurobiology and preclinical pharmacology in Somatostatin in the Central Nervous System).

Radioimmunoassay and immunocytochemical studies reveal that somatostatin-containing neurons are heterogenously distributed throughout the CNS. High concentrations of somatostatin are found in the hypothalamus and median eminence, amygdala, hippocampus, cerebral cortex, medial preoptic area, and nucleus accumbens. Cell bodies are most numerous in the preoptic and periventricular nuclei of the hypothalamus, although they are present in significant amounts in cortical and limbic regions as well. As is increasingly the case with many neurotransmitter systems, somatostatin is colocalized within a number of other monoamine- or neuropeptide-containing neurons. Moreover, like many neuropeptide systems, the distribution of somatostatin neurons in humans is similar, but not identical, to that observed in rodents. Of the two major forms of somatostatin present in the body (somatostatin1-14 and somatostatin1-28), somatostatin1-14 is the major form found in the brain (70% to 80%).

Studies have shown that pharmacologically distinct somatostatin receptor subtypes exist in the CNS with different affinities for somatostatin analogs, differential localization, differential coupling to second messenger systems, and functional responses. These have been termed SRIF1 and SRIF2 receptors of which the SRIF1 subtype is most predominant. Recently, four distinct somatostatin receptors (SSTR1–SSTR4) have been cloned. The receptor sequences appear to be highly conserved across species and are members of the G-protein coupled family of receptors.

Electrophysiological experiments have revealed both excitatory and inhibitory actions of somatostatin. Central administration of somatostatin has been observed to alter cholinergic, dopaminergic, noradrenergic, and serotonergic neurotransmission. Moreover, not only does somatostatin regulate growth hormone release from the anterior pituitary, somatostatin can also inhibit the secretion of a number of other hormones, particularly TSH, and CRF-stimulated ACTH secretion.

Like many other neuropeptide transmitters, central administration of somatostatin produces a variety of behavioral and physiological effects. Briefly, the peptide can produce a nonopioid-mediated analgesia in animals and man. Changes in sleep patterns, food consumption, locomotor activity and memory are also altered by somatostatin administration. This wide spectrum of effects of somatostatin led to investigation of its involvement in a number of psychiatric and neurological disorders. Of particular interest is the fact that the above changes in sleep, eating, activity, and anterior pituitary hormone secretion are all altered in depression.

The greatest number of clinical studies with somatostatin have focused on its involvement in neurological disorders. Consistent decreases in tissue and CSF concentrations of somatostatin are observed in senile dementia, Alzheimer's disease, Parkinson's disease, and multiple sclerosis during relapse. Somatostatin neurons have been found to be a major source of the observed plaques and tangles associated with Alzheimer's pathology. In contrast to the decrements in central somatostatin measures, elevated CSF concentrations of somatostatin, presumably reflecting leakage from neuronal damage, are observed in patients suffering from a number of inflammatory or destructive neurological disorders including spinal cord compression, destructive cerebral disease, meningitis, and metabolic encephalopathy.

Preclinical studies showing that central administration of somatostatin can alter sleep patterns, appetite, locomotor activity, and cognition have created interest in investigating somatostatin's role in mood disorders. The clearest evidence for involvement of somatostatin in psychiatric illness has come from studies of major depression. A consistent decrease has been reported in CSF somatostatin concentrations in depressed individuals. In primates, CSF somatostatin has been reported to undergo circadian variation, with concentrations varying 10% over 24 hr with the highest concentrations observed at night (6). Although the reported differences between depressed and normal patients are substantially greater than 10%, this circadian variation emphasizes the need for attempts at CSF sampling at uniform times. Although Rubinow et al. (72) did not find any time-of-day differences in patients, he did note large differences (increases and decreases) in subjects who were sampled in both morning and evening. Research over the past 15 years on a number of neuropeptides in CSF have revealed that they are almost exclusively of central origin, although the actual sites are unclear (65, 66). Decreases in CSF somatostatin concentrations are proposed to be the result of decreased neuronal synthesis and release. Whether this is a primary or secondary effect of the illness is unknown (see below).

Gerner and Yamada (25) first reported CSF somatostatin changes in psychiatric patients. In their study, medication-free patients with either depression or anorexia nervosa had decreased somatostatin concentrations versus either normal controls or healthy young women, respectively. This was replicated shortly thereafter in a large study by Rubinow and colleagues (73). Although there was considerable overlap, CSF somatostatin was significantly decreased in depressed patients, whether unipolar or bipolar. Moreover, whereas CSF somatostatin levels did not correlate with severity of depression, CSF values in clinically improved patients rose toward concentrations in normal subjects. In nine bipolar patients followed longitudinally, CSF somatostatin concentrations were significantly lower during the depressed state than during improved or manic states. In this study, CSF somatostatin concentrations were similar in men and women as well as across age in the subjects. Following the addition of a number of patients to his original study, Rubinow (72) again reported lower levels CSF somatostatin in depressed patients than in normal subjects, improved depressed patients, manics, dysthymics, or schizophrenics. No correlations between somatostatin and age among the depressed patients, severity of depression, or time of day were observed. In a large study lacking controls, Ågren and Lundqvist (1) reported significant correlations between severity of depression and CSF somatostatin concentrations. Moreover, CSF somatostatin was significantly lower in patients studied during their worst week of depression compared to those studied more than two months following their most severe week of depression. Black et al. (13) also reported decreased ventricular CSF somatostatin concentrations in medicated depressed patients referred for cingulotomy compared with lumbar CSF from normal controls. Bissette et al. (12) have also observed decreased CSF somatostatin in depressed patients. Like the patients of Ågren and Lundqvist (1), CSF somatostatin concentrations were not correlated to post-dexamethasone cortisol concentrations.

In contrast to this finding, Doran et al. (19) reported a significant negative correlation between post-dexamethasone plasma cortisol concentrations and CSF somatostatin concentrations in depressed patients. Kling et al. (41) have also reported decreased CSF somatostatin concentrations in depressed patients. Rather than a simple linear relationship, these investigators observed an inverted U-shape relationship between post-dexamethasone cortisol and CSF somatostatin concentrations in the depressed patients, suggesting a complicated relationship between central somatostatin neurons and hypothalamic CRF neurons. As we noted earlier in the chapter, neuropeptide systems, many of which we are more familiar with as hypothalamic releasing factors involved in behavior, are likely to be of both extrahypothalamic and hypothalamic origin. Sunderland et al. (82) reported similar decreases in CSF somatostatin in older depressed patients with no correlation between CSF concentrations and severity of depression. Davis et al. (17) also observed decreased CSF somatostatin concentrations in older depressed patients versus elderly controls. Molchan et al. (51) observed similar findings in elderly depressed and control populations. Moreover, CSF somatostatin from depressed Alzheimer's disease (AD) patients was significantly lower than the already low nondepressed Alzheimer's population. Although CSF 5-hydroxyindole acetic acid (5-HIAA) did not differ among diagnostic categories, CSF somatostatin and 5-HIAA were positively correlated among both AD and non-AD depressed groups (i.e., the lower the 5-HIAA concentrations, the lower the somatostatin concentrations). Whether these are related or separate results and/or causes of depression is unclear, although a number of monoamine transmitter systems can alter somatostatin neuronal activity (see Somatostatin in the Central Nervous System).

Somatostatin concentrations in CSF have also been examined in several other psychiatric disorders. Kaye et al. (39) found no differences in anorexics, but observed a small increase in CSF somatostatin concentrations in normal weight bulimics when they stopped binging. Berrettini et al. (11) found no differences in CSF somatostatin between controls, unmedicated euthymic bipolar patients, and lithium-treated bipolar patients. Altemus et al. (2) reported significantly higher CSF somatostatin concentrations in patients with obsessive–compulsive disorder (OCD) compared to controls. Although a normal control group was lacking, Kruesi et al. (45) also reported an increase in CSF somatostatin levels in a small group of children with OCD versus a group of children with conduct disorder.

Few postmortem studies of CNS somatostatin in depressed patients have been reported to date. Charlton et al. (15) found no differences in temporal or occipital cortex somatostatin concentrations or somatostatin receptor affinity or number in a small group of controls (N = 7) versus depressed patients who died of coincidental physical illness while inpatients at a psychiatric hospital (N = 9). In another study of seven depressed patients and twelve controls, Bowen et al. (14) measured somatostatin concentrations in the pars opercularis and orbital gyrus of the frontal lobe, the parahippocampal gyrus and pole of the temporal lobe, and postcentral gyrus and superior lobule of the parietal lobe. Significant decreases (30%) in somatostatin concentrations were observed only in the pole of the temporal lobe. Although these studies are complicated by wide ranges of postmortem delay until sampling and freezing of the tissue and further complicated by the finding that somatostatin concentrations decrease within the first 6 hr following death (80), studies of this nature are exceptionally useful and sorely needed.

As mentioned in Somatostatin in the Central Nervous System, preclinical studies show that a variety of neurotransmitter systems can alter CNS somatostatin neurons. Of the centrally active drugs used clinically, serotonin-selective uptake inhibitors increase somatostatin concentrations in rat brain (37). In a group of depressed patients, carbamazepine was found to produce a significant decrease in CSF somatostatin concentrations (74). The same study found that the selective serotonin (5-HT) uptake inhibitor, zimelidine, significantly increased CSF somatostatin in five of five patients. In a small group of patients, neither desipramine nor lithium had any effect on CSF somatostatin concentrations. The further reductions in CSF somatostatin produced by carbamazepine were not correlated to worsening or improvement of symptoms. This finding suggests that (a) somatostatin may be implicated in the anticonvulsant mechanism of action of carbamazepine, and (b) together with the neurological disorders characterized by decreased CSF somatostatin concentrations without necessarily having concomitant changes in mood, decreases in CSF somatostatin are not responsible for the changes in affect seen in depression.

Although the data is still relatively limited, an overview of the extant literature suggests that decreases in CSF somatostatin are a consistent state-dependent finding of depression. Indeed, next to the hypercortisolemia associated with depression, this may be one of the more consistent findings in biological psychiatry. However, it probably does not possess any diagnostic usefulness, because similar changes are observed in a number of neurological disorders without psychiatric comorbidity. However, it appears to be associated with impairment in cognitive function. Evidence is beginning to appear to suggest that the decrease in CSF somatostatin may be related to the overactivity of the HPA axis seen in depression. Whether one is responsible for the other or both are responses to dysregulation of other neurotransmitter systems associated with depression is unknown. Finally, somatostatin-active drugs may not be of therapeutic usefulness as changes in CSF somatostatin apparently do not affect mood. Nevertheless, rational design of peptide- or nonpeptide-based drugs selectively active at different receptor subtypes will certainly aid in understanding somatostatin's role in behavior and may lead to therapeutic benefits.

The early availability of adequate tools (i.e., assays, synthetic peptides) coupled with observations that primary hypothyroidism can cause depressive symptomatology ensured extensive investigation of the involvement of the hypothalamic–pituitary–thyroid (HPT) axis in mood disorders (see ref. 48 for a review). Interested readers will find a detailed description, along with citations, of TRH neurobiology and preclinical pharmacology in Thyrotropin-Releasing Hormone: Focus on Basic Neurobiology). Thyrotropin releasing hormone, a pyroglutamylhistidylprolinamide tripeptide, was the first of the hypothalamic releasing hormones to be isolated and characterized. In its role as a hypophysiotropic factor, TRH is released from hypothalamic nerve endings in the median eminence into the primary capillary plexus of the hypophyseal–portal circulatory system where it is transported to thyrotropes in the adenohypophysis. Then, TRH diffuses out of these capillaries and binds to specific membrane receptors to facilitate the release and synthesis of thyroid-stimulating hormone (TSH). Circulating TSH acts at the thyroid gland to evoke release of L-triiodothyronine (T3) and thyroxine (T4).

Early studies established the hypothalamic and extrahypothalamic distribution of TRH (reviewed in ref. 52). This extensive extrahypothalamic presence of TRH quickly led to speculation that TRH might function as a neurotransmitter or neuromodulator. Subsequent studies established the necessary foundation required to seriously consider TRH in such a role. Thyrotropin-releasing hormone is concentrated in nerve terminal regions and presumably stored in synaptic vesicles. Enzymes were identified in the CNS which inactivate TRH and, thus, could curtail its action following release. Specific, high-affinity TRH receptors were characterized and found to be widely and selectively distributed throughout the CNS. Electrophysiological studies provided evidence that TRH directly altered neuronal activity. Thus, a large body of evidence supports the involvement of TRH as a hypophysiotropic factor and as a neurotransmitter or neuromodulator.

Administration of exogenous TRH is associated with a variety of physiological and behavioral effects including alterations in cardiovascular function, respiratory rate, body temperature, gastric secretion, colonic motility, and electroencephalographic activity (52; see also Thyrotropin-Releasing Hormone: Focus on Basic Neurobiology). Interest in putative CNS actions of TRH were stimulated by studies of the thyroid axis and depression by Prange and colleagues (see Thyrotropin-Releasing Hormone: Focus on Basic Neurobiology). Utilizing the mouse L-dopa potentiation test to screen for putative antidepressant effects, Plotnikoff et al. (60) found that TRH produced enhancement of motor activity, a behavioral marker of compounds possessing putative antidepressant efficacy. Indeed, the effects of TRH in this test were similar to those observed in imipramine-treated mice, suggesting that TRH might act as an antidepressant compound. Importantly, TRH was equally effective in intact and hypophysectomized mice, thus indicating that the observed effects were not dependent upon activation of the HPT axis but instead reflected a direct CNS action of TRH.

Profound increases in CNS activity evoked by electroconvulsive shock (ECS) administered on an alternate-day schedule for 5 days (46) produced pronounced increases in TRH concentration in limbic regions, including the amygdala and hippocampus. Neither administration of a single ECS treatment nor administration of a subconvulsant electrical current altered TRH in any region assayed, indicating that the induction of a seizure was essential to produce effects on TRH. Other observations suggest that the efficacy of TRH or analogs in the treatment of seizure disorders is an area ripe for further evaluation. In summary, these animal studies have highlighted the wide range of physiological and behavioral effects exerted by TRH. One may speculate on the possibility that limbic system TRH-containing circuits may mediate, in part, the antidepressant actions of ECT and may be involved in the pathophysiological mechanisms of seizure disorders. The potential clinical implications of these studies include the possible utility of TRH in treating overdoses of sedatives or hypnotics and in treating certain forms of epilepsy.

Learned helplessness, an animal model of depression (31), can be reversed by many clinically effective antidepressants and by ECT. Interestingly, when these animals are rendered hypothyroid they exhibit resistance to the effects of tricyclic antidepressants and treatment with thyroid hormone reverses the antidepressant resistance. Over a decade ago, Whybrow and Prange (84) hypothesized that thyroid function was integral to the pathogenesis of, and recovery from, mood disorders because of the copious interactions among thyroid hormones, catecholamines, and adrenergic receptors in the central nervous system. Overall, these studies suggest a role for thyroid dysfunction in refractory depression and are consonant with clinical studies suggesting the existence of an increased rate of hypothyroidism among patients with refractory depression. Furthermore, animal and clinical studies suggest that depressive symptoms in hypothyroid patients may, in part, be determined by thyroid function before the onset of depression (reviewed in ref. 34).

The use of TRH as a provocative agent for assessment of thyroid-axis function evolved rapidly after its isolation and synthesis. A relatively standard protocol involves measurement of basal plasma TSH concentrations followed by intravenous administration of exogenous TRH (200 to 500 mg) with subsequent measurement of plasma TSH concentrations at 30-min intervals for a period of 2 to 3 hr (77). Clinical use of the TRH stimulation test to assess hypothalamic–pituitary–thyroid (HPT) axis function revealed blunting of the TSH response in approximately 25% of euthyroid patients with major depression, as reviewed by Loosen (48), Nemeroff (52), and Howland (34). These data have been widely confirmed, and it has been proposed that decreased nocturnal plasma TSH concentration may be a sensitive marker of depression (10). Using a modified TRH stimulation test in which the peptide (200 mg) is administered intravenously at 8 a.m. and at 11 p.m. Duval et al. (21) claimed a diagnostic specificity of 95% and a diagnostic sensitivity of 89%. The difference in the TSH response between the 11 p.m. and 8 a.m. tests appears to be markedly reduced in depressed patients as compared to controls. Recently, Shelton et al. (78) reported that among outpatients with major depression, 26% percent showed some abnormality of thyroid hormone concentrations; the majority of these were normalized by antidepressant treatment. Even though antidepressants did not exhibit a statistically significant effect on thyroid hormone concentrations when tested across the whole group, it must be noted that patients in this study did not exhibit a high frequency of TSH blunting to TRH. Overall, this study implies that subpopulations of the more severely depressed patients may be more likely to exhibit TSH blunting and elevations of T4 prior to therapy. The relevance of TRH to psychiatric disorders has recently been reviewed by Nemeroff (52).

The observed blunting of TSH in depressed patients does not appear to be the result of excessive negative feedback from secretion of either thyroid hormone or somatostatin hypersecretion. In fact, as noted above in detail, depressed patients exhibit reduced CSF concentrations of SRIF. It is possible, although unlikely, that the blunting is a reflection of pituitary TRH receptor down-regulation as a result of median eminence hypersecretion of endogenous TRH (65). The observation that lumbar CSF concentrations of TRH are higher in depressed patients than in controls supports a hypothesis of TRH hypersecretion but does not elucidate the origin of this tripeptide (7). The Banki study consisted of 16 control subjects (12 patients diagnosed as having only peripheral neurological disease and 4 patients having a DSM-III diagnosis of somatization disorder) and 17 patients with major depression. None of the subjects had been treated with psychotropic medications for at least 2 weeks prior to start of the study. The mean CSF concentration of TRH for the combined control group was 4.4 ± 1.8 pg/ml, whereas the concentration in the depressed group was 12.8 ± 5.7 pg/ml. In animal models, chronic administration of TRH for 2 to 3 weeks results in of the TSH response to TRH, and decreased circulating concentrations of TSH, T3, and T4 (54). Furthermore, repeated administration of TRH in humans also produces a blunted TSH response to TRH (49). Finally, these elevations of TRH concentration in CSF may be relatively specific to depression, as no such alteration has been reported in patients with Alzheimer's disease, anxiety disorders, or alcoholism (23, 69). Clearly, further studies for which CSF test concentrations are measured are needed.

Although the majority of depressed patients readily respond to treatment with antidepressants, approximately 15% to 30% of depressed patients are treatment refractory (75). Approximately 15% of depressed patients display Grade III hypothyroidism, characterized by normal T3, T4, and TSH concentrations and an exaggerated TSH response to the TRH stimulation test, and almost 60% of these patients have detectable antimicrosomal and/or antithyroglobulin antibodies in their circulation (27). Interestingly, approximately 50% of depressed patients who are DST nonsuppressers exhibit exaggerated TSH (29). Overall, the rate of asymptomatic autoimmune thyroiditis (SAT) in depressed patients is greater than would be expected in the general population (34, 52). It may be postulated that the development of autoimmune thyroiditis gives rise to hypersecretion of hypothalamic TRH as a compensatory mechanism to maintain normal plasma T3 and T4 concentrations. A considerable literature, recently reviewed by Joffee et al. (35), exists regarding interactions between the HPT axis and mood disorders.

Preclinical studies have added to our current understanding of the physiological and behavioral effects exerted by TRH. However, clinical studies of TRH in depression have yielded mixed results with no definitive role for TRH in the pathophysiology or treatment of any psychiatric disorder. Thus, despite the initial promise offered by early preclinical and clinical studies of TRH, many subsequent investigations have yielded equivocal results. At present, the diagnostic validity of the TRH stimulation test remains open to question with respect to depression. However, several investigators have suggested that this test demonstrates prognostic value in predicting treatment responses. For instance, Langer et al. (47) presented evidence to suggest that a change in the TSH response from blunted to normal predicts a positive response to antidepressant therapy, whereas Kirkegaard (40) suggested that the TSH response to a TRH challenge appears to be of value in predicting long-term clinical outcome (remission or relapse) following completion of therapy. Obviously, additional studies are required to clarify the potential clinical utility of the TRH stimulation test in the diagnosis and treatment of depression.

Thus despite years of study, considerable gaps in our knowledge remain. Many avenues for future experimentation are available. A concerted effort should be made to perform postmortem measurements of regional brain TRH concentrations, mRNA levels of TRH, pituitary and brain TRH receptor kinetics, and mRNA concentrations of TRH, as well as postreceptor signal transduction in tissues obtained from depressed suicide victims and well-matched controls. Furthermore, studies of CSF concentrations of TRH and rhythms should be performed in populations of depressed patients, those with autoimmune thyroiditis, and well-matched controls. Finally, the synthesis of a long-acting TRH analog as well as development of a lipophilic TRH radioligand permitting PET/SPECT visualization of TRH receptor density in the CNS and pituitary would be of great utility in assessing the importance of the interaction between the HPT axis and mood disorders.

Significant progress has been made over the past decade, and many new tools have entered the basic and clinical scientist's armamentarium. Application of the RNase protection assay and of in situ hybridization will permit detection and mapping of specific mRNAs in experimental animal CNS and in human postmortem CNS tissue. Widespread use of these methods will yield a picture of the concentrations and distributions of CRF, SRIF, and TRH in mRNAs and their receptor mRNAs in normal subjects, those with mood disorders, and those with other disorders. For those mRNAs of particularly low abundance, in situ hybridization or RNase protection assays may be preceded with polymerase chain reaction (PCR) amplification of the target mRNA. Studies to assess postmortem stability of the mRNAs of interest will be necessary and might be most easily accomplished using CNS tissue removed during neurosurgical procedures and then subject to controlled processing delays. A more accurate determination of transcriptional activity may be derived by measurement of heteronuclear RNA (hnRNA) using intronic antisense hybridization probes in RNase protection or in situ hybridization protocols. However, the rapid turnover time and small pool size of nuclear hnRNA will necessitate rigid control of the postmortem delay and studies of hnRNA stability after death.

With our advancing knowledge of potential peptidinergic circuit or peptide receptor dysfunction contributing to the pathophysiology of mood disorders, the development of animal models will assume increasing importance. This may be most readily accomplished using transgenic over-expression or knock-out models or by the use of stereotaxic microinjection of sense or antisense RNA directly or carried by adenovirus or modified herpes virus vectors. Using these models, the consequences of hypo- or hyperactivity of each of these neuropeptides or receptor systems may be assessed at the neurobiological and behavioral levels, and potential therapeutic approaches may be developed.

Another exciting development is the ability to image CNS tissue using MRI, PET, and variants of these methods. It is becoming possible to monitor CNS circuits in action in healthy and diseased CNS tissue using metabolic markers and to assess receptor distribution using labeled ligands. As these techniques are refined with increases in sensitivity and resolution, they should have a major positive impact on both basic and clinical studies of the CNS mechanisms underlying mood disorders.

We would like to thank Jan Fowler for assistance in preparation of this manuscript and the support of the National Institute of Mental Health through grants MH39415, MH42088, MH40524, and MH45216, and the National Institutes of Health through grant DK33093.

published 2000