Stress

Huda A. Akil and M. Inés Morano

Stress, though a nebulous concept, is often evoked as an important trigger in the expression of several psychiatric illnesses, including major depression, anxiety disorders, and schizophrenia. The term generally refers to any physical or psychological change that disrupts the organism's balance or homeostasis. Psychological stressors are typically emphasized as key elements in psychiatric illness, although physical stressors (e.g. viral infections, autoimmune disorders) are becoming increasingly recognized as potential triggers.

Stress activates many systems in the body, including the adrenal medulla and the autonomic nervous system. It was first proposed by Cannon as the integrator of fight-or-flight responses (15). Of particular interest, however, is the role of the brain in orchestrating responses to stress. The interface between the emotional or limbic system and the stress control circuits may indeed represent the site(s) of interaction between the control of stress and the expression of various psychiatric diseases. In the following sections, we focus on some of the major actors that participate in the expression and termination of stress responses at the level of the brain and briefly summarize what is known about the circuits involved. We also describe what is currently known about aging and the stress system, both because the aged animal has been proposed as a model of a somewhat dysregulated stress system and because, in clinical studies, there appears to be an interaction between aging, indices of dysregulation of the stress axis, and some psychiatric illnesses such as severe depression. Finally, we present a brief overview of current research on the relationship between stress and depression, as an example of one psychiatric illness whose relationship to stress has been extensively investigated. This discussion also illustrates the impact of the basic science findings on the choice of clinical research approaches. The overview in the last section attempts to emphasize the conceptual and integrative questions at the interface between stress biology and affective disorders (see also Physiological and Anatomical Determinants of Locus Coeruleus Discharge: Behavior and Clinical Implecations, Noradrenergic Neural Subtrates for Anxiety and Fear: Clinical Associations Based on Preclinical Research, Neuroendocrine Interactions, The Serotonin Hypothesis of Major Depression, and The Role of Acetylcholine Mechanisms in Mood Disorders).

All organisms, from bacteria to humans, have evolved mechanisms to deal with significant changes in their external or internal environments, that is, stressors. In mammals, this function is carried out, in part, by the limbic–hypothalamo–pituitary–adrenal (LHPA) axis. This system integrates various inputs indicative of stress, converging on a final common path in the brain, the neurons of the medial parvocellular division of the paraventricular nucleus of the hypothalamus (mpPVN). These neurons synthesize corticotropin-releasing hormone (CRH) and arginine vasopressin (AVP), and project to the external layer of the median eminence (86). Activation by stressors leads to release of the peptides into the portal blood, carrying these secretagogues to the anterior pituitary. In turn, CRH and AVP receptors on the anterior pituitary corticotropes are responsible for the release into the total circulation of adrenocorticotropin hormone (ACTH) and related peptides derived from the common precursor proopiomelanocortin (POMC). ACTH activates the biosynthesis and release of glucocorticoids, corticosterone in rodents, and cortisol in primates by the cells of the adrenal cortex. These steroids possess extremely broad actions mediated by specialized receptors affecting expression and regulation of genes throughout the body, and readying the organism for the changes in energy and metabolism required for coping (60).

Corticotropin-releasing hormone is a 41-residue peptide that plays the major role in regulation of pituitary corticotrope trophic activity (34), POMC gene transcription (10), and ACTH secretion (8). In addition to CRH, various other factors have also been shown to induce ACTH secretion during stress. These intrinsically weaker secretagogues include AVP, oxytocin (OT), angiotensin II, cholecystokinin, vasoactive intestinal peptide (VIP), and catecholamines (11). However, most of them have been shown to depend upon the presence of CRH for their modulatory action in LHPA axis activation. Among these other secretagogues, the physiological importance of AVP in the regulation of the stress response should be highlighted, because of the colocalization and corelease of CRH and AVP in mpPVN (86) and because of the synergistic effects of both peptides on the corticotropes. Although AVP is a weak ACTH stimulator by itself, it markedly potentiates the effect of CRH (8).

The CRH-stimulation of ACTH secretion, POMC transcription, and POMC mRNA levels in the corticotropes is thought to be mediated by cyclic adenosine monophosphate- (cAMP) coupled CRH receptors. In contrast, the corticotrope AVP receptors are thought to be coupled to phosphatidyl inositol turnover (8) and to mediate primarily effects on ACTH secretion. Moreover, AVP does not further affect CRH-induced POMC transcription or messenger ribonucleic acid (mRNA) expression (53).

Proopiomelanocortin (POMC) is one of the three opioid peptide precursor genes. This precursor encodes, not only the stress hormone ACTH, but also the opioid peptide b-endorphin 1–31 and three copies of the active core of melanocortin (a, b, and gMSH). Although the major site of expression of the POMC gene is the pituitary gland, it is also expressed in several brain regions and peripheral tissues. In the pituitary, it is expressed in anterior lobe (AL) corticotropes, and in every cell of the intermediate lobe (1). The precursor POMC undergoes an orderly series of enzymatically mediated posttranslational processing steps to produce its tissue-specific final products. In the corticotropic cells, the cleavage of double basic residues converts POMC into equimolar amounts of a carboxy-terminal glycopeptide termed 16K, ACTH 1-39, and b-LPH. Approximately one-third of b-LPH molecules are further processed in these cells to produce the opioid active b-endorphin 1–31. By contrast, in the intermediate lobe, all of the b-LPH is converted to b-endorphin and, ACTH, b-endorphin, and 16K undergo further posttranslational modifications to produce highly altered final products. In particular, ACTH is converted to aMSH and CLIP (1).

Specifically because of their potency and their range of action, glucocorticoids have to be maintained within an optimal range—either too little or too much is deleterious to the organism (60). Therefore, the LHPA is well controlled to produce rapid and optimal responses, which terminate promptly upon cessation of stress. This is achieved by multiple, nested, negative-feedback loops, primarily mediated by the steroids themselves (47). These mechanisms differ in terms of their timing, their underlying mechanisms, as well as their sites of action. The most rapid mechanism is the so-called rate-dependent fast feedback, which occurs within minutes of activation, that is, as soon as steroid levels begin to rise. This mechanism monitors the rate of rise of steroids in the circulation, rather than their absolute level, and serves to turn off CRH/AVP secretion. The exact receptors mediating this feedback are unknown, but the site of action is thought to be suprapituitary (47), at hypothalamic or possibly suprahypothalamic levels (25, 45 ). At the other extreme, is a genomically mediated negative feedback with a much slower time course, whereby glucocorticoid receptors, which are in fact transacting factors, negatively control gene expression, decreasing rates of transcription of critical genes such as POMC. Indeed, the negative regulation starts immediately upon receptor activation by steroids, but its consequences on the cell, in terms of mRNA and peptide levels take hours or even days to manifest, because of their intrinsic kinetics and the presence of large reserves. This genomic feedback operates at the level of the pituitary corticotrope by controlling POMC gene expression (10) as well as at the brain level. Finally, there is an intermediate feedback, with a time course between the fast and the genomic, that remains ill-understood (47). In a living organism, all these mechanisms are probably activated simultaneously, but they come into play in different time domains; fast feedback is likely to set the magnitude and duration of each response, whereas genomic feedback sets the range of stress responsiveness of an organism. In addition, these feedback mechanisms can be seen as different lines of defense, involving a hierarchy in time and space, with fast feedback being more sensitive, rapid and brain mediated, and genomic feedback being slower but having more profound effects at multiple levels of the axis.

Brain steroid receptors are thought to mediate negative feedback upon the axis, but the relative role of the various elements is not fully delineated. What is known is that glucocorticoids inhibit CRH and AVP mRNA levels (82), probably from reduced gene transcription, which would lead to diminished biosynthetic capacity and possibly result in decreased peptide release. These inhibitory effects are amplified at the level of the anterior pituitary. When steroid levels are high, the pituitary POMC cell is simultaneously under reduced secretory drive and under direct genomic inhibition, both of which dramatically inhibit biosynthesis. Conversely, when corticosteroids are removed (e.g., by adrenalectomy or steroid synthesis inhibitors), CRH and AVP mRNA expression is significantly induced in the hypothalamic neurons. These presumably fire more actively, releasing their contents. In turn, corticotropic cells receive the combined influence of increased activation by secretagogues, which is known to induce POMC biosynthesis, and of the absence of steroid feedback, leading to an overall tenfold increase in POMC mRNA levels (10). An important question is: Are the effects of steroids on CRH neurons direct or indirect? The hypothalamus is rich in steroid receptors (68) and the steroid effects could be direct. Indeed, glucocorticoid implants within the hypothalamus have been shown to suppress CRH/AVP peptide expression (48). However, negative regulation may also be mediated by neuronal pathways that are sensitive to steroids and that modulate the firing of CRH/AVP neurons. The fact that the hippocampus is the brain area richest in corticosteroid receptors (40, 68) strengthens the notion that negative steroid feedback does not reside exclusively in the PVN. The specific role of the hippocampus in negative feedback remains unclear, but the existence of an indirect hippocampal–PVN pathway that modulates CRH/AVP expression is suggestive of a circuit that plays a role in negative regulation of the LHPA (see below).

Two steroid receptors mediate negative feedback in the CNS. Why two? remains an interesting questions, as the field attempts to understand their similarities and differences at a molecular and functional level. DeKloet and his colleagues (68) using steroid receptor binding and autoradiography first described the existence of two binding sites in the hippocampus that recognize corticosteroids. These were termed type I and type II and have different pharmacological profiles in vivo: Type I has higher affinity to cortisol/corticosterone than type II, whereas type II recognizes dexamethasone with higher affinity than type I. More recently, these receptors have been cloned and identified as the glucocorticoid receptor (GR), which subserves type II binding (42), and mineralocorticoid receptor (MR), which in the kidney recognizes aldosterone and in brain corresponds to type I binding (9, 64). Both receptors have a broad distribution; although both are particularly enriched in the hippocampus, the PVN region appears to express predominantly GR or type II receptors (40).

Both receptors belong to the steroid receptor family, which has a well-known topology. The carboxy terminal domain is the ligand-binding domain and subsumes the specific recognition site for ligands. In the middle is the DNA-binding domain, which contains so-called cysteine-rich zinc fingers critical for DNA interactions, and the amino terminal domain, which is thought to be important for transcriptional modulation. These domains appear to be functionally separable, and numerous chimeras have been constructed to retain appropriate binding selectivities and DNA recognition properties (36, 78). Glucocorticoid and mineralocorticoid receptors are very similar in the ligand-binding domain, which is reasonable given their recognition of common ligands. However, they have interesting differences that reflect their possible unique functions in the context of the stress axis.

In the absence of steroids, the majority of GR- and MR-binding sites reside in the cytoplasm and appear to be associated with a protein complex that involves several heat-shock proteins including hsp-90 (22, 23). Association with this complex is thought to be critical to the ability of the corticoid receptors to bind ligand; GR and MR in the absence of hsp-90 lose their binding affinity to glucocorticoids (43). In the presence of glucocorticoids, hormone-receptor interaction takes place in the cytoplasm. This cytoplasmic interaction, which can also be seen in cell-free systems, results in both the dissociation of the receptor–hsp-90 complex and an increased affinity of the receptor for deoxyribonucleic acid (DNA). This represents the initial step necessary for the nuclear translocation and activation of the receptor to its final, DNA-bound state. Recent work from our laboratory (Morano et al., unpublished), using GRs mutated in the hsp-90 binding region, suggests that the receptor/hsp-90 interaction is required not only for hormone binding but also for its activation to a transcriptionally functional DNA-binding form. Thus the hsp-90 association is a critical aspect of the cell cycle of these receptors, and, along with the steroids themselves, determines the state of activity of these transacting factors. This is in contrast to other members of the steroid family (e.g., thyroid hormone receptor) that dwell in the nucleus and are not complexed with a heat-shock protein.

In addition to the multiplicity of receptor types, there is heterogeneity within the MR family. We have shown that there are at least three types of mRNAs for MR, termed a, b, and g, which harbor the same protein coding domain but differ in their 5¢ untranslated regions (5¢UT) (50, 64). These variants appear to be the result of alternative splicing of a single MR gene, are differentially expressed in various brain regions and other tissues, have unique developmental patterns, and are differentially regulated (50).

Steroid receptors exert their effect on DNA by binding to specific stretches of DNA or recognition elements. The glucocorticoid recognition element (GRE) consists of 15 base pairs of DNA and is a motif found in numerous genes. Receptors of other classes of steroid hormones, progestins, and androgens, can also modulate genomic activity by binding this element (36). In the best studied cases, binding of the GR to the GRE causes an increase in transcription. This positive regulation is predominant. However, all the known genes of the stress axis—POMC, CRH, AVP—appear to be negatively regulated by glucocorticoids. If this regulation is direct (as opposed to being neuronally or secretion mediated), then these genes must possess a negative GRE. Indeed, GRE-like elements have been identified in the POMC promoter (27) and there are several examples of negative GRE's in nonstress system genes (70). However, other hypotheses exist suggesting that negative regulation by steroid receptors may be due to proteins preventing activation by another transacting factor, either because the DNA elements overlap or through protein–protein interactions with other transacting factors such as c-fos/c-jun (27, 87). An interesting synthesis was offered in a paper by Yamamoto and his colleagues (26) showing that a single element could be either positive or negative depending on the cellular milieu and the recent history of activation of the gene.

The question of gene regulation is even more difficult with regard to the mineralocorticoid receptor (MR or type I). No specific MRE has yet been defined. Evans and his colleagues (9) have shown that MR can activate a positive GRE-containing gene (MMTV) but is weak (5%) compared to GR activation. They suggested that MR may share the same DNA recognition site as GR but be weaker at transactivation. Given that MR has a higher affinity for natural corticosteroids, this receptor would represent a lower threshold, but be a low-efficiency mediator of steroid actions, whereas GR would represent a higher threshold but be a higher efficacy element. This is an interesting hypothesis, which remains to be tested.

Recently, Yamamoto's group analyzed the activities of GR and MR at a composite response element to which both the steroid receptors and the transcription factor AP1 can bind. In this case, under conditions in which GR repressed AP1-stimulated transcription, MR was inactive (66) suggesting that the differential interactions of nonreceptor factors with specific receptor domains at composite response elements may explain distinct physiological effects of MR and GR. Under this model, one can see MR, not only as a weak GR but also as a potential GR antagonist, binding to the same DNA elements but failing to produce a transcriptional effect (33, 66).

The Evans hypothesis along with the findings of Yamamoto's laboratory on the composite element, raise several interesting questions: For example, does MR work exclusively through GREs or are there unique MREs that may mediate its effects? Are both receptors equally likely to interact with other transacting factors? Even if MR is acting at a GRE, is the potency of MR relative to GR always low, regardless of the target gene? In particular, what happens for negatively regulated genes? Are some more sensitive to MR than to GR? An intriguing example is the case of AVP in CRH neurons. This gene appears to be quite sensitive to adrenal steroids, as indicated by the low basal level of AVP expression in control rats. Following removal of steroids by adrenalectomy, AVP mRNA is dramatically induced (eightfold) (77). On the other hand, the CRH gene in the same neurons, is expressed actively in the unstressed rat, and is only induced two- to threefold by adrenalectomy (77). This suggests that steroid levels in the basal circadian range (i.e., levels that primarily activate MR, see below) are sufficient to inhibit AVP but not CRH expression. Yet when dexamethasone (a GR agonist) is administered in supraphysiological doses, CRH mRNA levels are dramatically reduced. Thus, it is likely that AVP may be more MR-responsive than is CRH, suggesting that different genes, even in the same tissue, may exhibit different responses to the two types of steroid receptors.

One of the characteristics of basal steroid levels is that they oscillate in a circadian fashion. The levels are correlated with the rest–activity cycle of the animal, rather than the light cycle. Thus, in man, steroids begin to rise in the early morning hours, peak around awakening, and then fall throughout the day. In nocturnal animals such as rats, the converse pattern is seen, whereby levels peak in late evening and are at a nadir in early morning. Connections between this cycle and rest–activity, sleep, and feeding behavior have been made. The effects of stress, then, are superimposed on the basal rhythmicity, and there is evidence that stress responsiveness and the effectiveness of negative feedback may also oscillate across the cycle. Thus, at the trough of the rhythm, animals appear to be more sensitive to both stress activation and inhibition by glucocorticoids, suggesting that at this time, the axis is exquisitely responsive (21). The drive to the axis prior to awakening appears to be initiated by the suprachiasmatic nucleus (SCN), leading to enhanced tone of CRH, and resulting in increased activity throughout the LHPA (49). Interestingly, the basal levels of steroids at the peak are thought to be sufficiently high to occupy a majority of the type I or MR receptors (estimates vary from 60% to 90%) but only occupy a small proportion of type II or GR (~10%) (68). Thus, there may be elements of the axis (e.g., AVP in mpPVN) that are particularly sensitive to circadian drive and are modulated by MR, whereas others are particularly stress responsive and only modulated by GR or by both receptors.

In addition, the circadian rhythmicity of the system needs to be kept in mind as one examines the anatomy and function of various stress-related circuits. It is important to recognize that both stress activation and stress termination mechanisms might either modulate or be modulated by this daily rhythmicity. In addition, manipulations, whether they are surgical or pharmacological, intended to look at the stress circuits might alter the rhythm. Such an interplay may also prove important in aging, which is known to alter sleep, rest–activity, and stress responsiveness. Thus circadian rhythmicity is an intrinsic feature of this system, having molecular, anatomical and integrative implications.

As we discussed above, negative feedback, which is critical to the termination of the stress response as well as to establishing the overall set point of the axis, may be mediated both directly at the level of gene regulation and by neuronal circuits. The presence of high levels of both GR and MR in the hippocampus led to the notion that it may play a key role in negative feedback (c.f., 25,45). Previous work has demonstrated that hippocampal lesions elevate circulating levels of glucocorticoids, both basally and poststress (30). Our group (Watson, Akil, and coworkers) was therefore interested in investigating the possible existence of a hippocampo–hypothalmic pathway that may play a role in negative control of the axis. We used lesions as our tool and examined CRH and AVP mRNAs and, ACTH and corticosteroid plasma levels. Our reasoning was as follows: if a brain site is critical to inhibiting the tone of the LHPA, its destruction should result in an increase in that tone, as evidenced by a chronic increase in circulating stress hormones (ACTH, corticosteroids). If steroids only have direct effects on CRH/AVP in the PVN, then the high steroid levels resulting from these lesions should lead to down regulation of the CRH/AVP message. However, if steroid feedback works in part through our target site, then lesioning that site should not only increase secretion but also prevent steroid-induced down-regulation, and actually lead to the unusual condition of increased circulating hormones, and increased CRH/AVP.

To date we have studied multiple regions of the hippocampus and have shown that, indeed, hippocampal lesions result in this unusual combination of elevated glucocorticoids and elevated levels of secretagogues (41). Although several hippocampal sites participate in controlling basal CRH expression, only very discrete regions appear involved in the termination of the stress response (Herman et al., unpublished). Lesions of the latter sites result in stress response patterns where activation appears normal but termination appears slow, a pattern seen in aged animals (see below). What are the neural pathways mediating these effects? Interestingly, the hippocampus apparently lacks significant direct projections to the PVN, suggesting the existence of one or more relays. Extensive anatomical studies in our laboratory have suggested that these are multiple redundant pathways, which connect the hippocampus to the PVN through various relays points, including one in the bed nucleus of the striaterminalis (BNST) (19). These pathways may play a role in controlling various aspects of the stress response or may serve other distinct functions (Fig. 1).

Even less is known about the neuronal circuits involved in the initiation of the stress response (Fig. 1). We, of course, know the final common path in the mpPVN, and we also know that brainstem nuclei are likely to play an important role in stress detection, for example, locus coereleus and nucleus of the solitary tract. A number of afferent inputs have been shown to participate in the activation of the HPA axis. In particular, noradrenergic and adrenergic neurons of the A2 region in the caudal medulla are known to send direct inputs to the mpPVN (20) and their modulation clearly alters the tone of the CRH neuron, presumably by a1 adrenoceptors (4, 58, 18). The nucleus tractus solitarius, which is partly catecholaminergic, has been shown to be critical in inducing ACTH release in response to hypotension (24). Similarly, serotonergic cell groups (dorsal raphe and raphe magnus) project to the parvocellular PVN (76) and appear involved in stress responsiveness (31) and circadian rhythmicity (83). The amygdala has also been implicated in stress responsiveness (3, 12) as has the medial BNST (28). More rostrally, stimulation of the preoptic area and frontal cortex can excite PVN neurons that project to median eminence (71).

Different types of stressors (physical vs. psychological, painful vs. nonpainful) are likely to engage the system from different starting points. But at what level do they converge? Is it only at the level of the CRH/AVP neuron in the parvocellular PVN? Or do they do so more proximally, for example in limbic sites, which then lead to the mpPVN? A major stumbling block for these experiments has been the need to develop a strategy for detecting stress activation in the CNS. This is a prerequisite to carrying out lesion studies to disrupt the pathway and begin to delineate its components. One can, of course, measure peripheral indices of stress—corticosterone and ACTH increases in plasma. However, the ability to monitor CNS correlates of acute activation by various stressors is critical to discerning the existence of unique pathways as well as common elements. The measure has to be rapid enough to detect the activation before negative steroid feedback dampens or even reverses the response. Several recent studies have been undertaken in an effort to map brain regions activated by various stressors, using immediate early genes (IEGs) such as c-fos, c-jun, and zif/268 as markers for early neuronal activity. To date we have examined IEG responses following restraint stress and swim stress (Cullinan et al., unpublished). The results suggest that in addition to activation of the mpPVN, a core set of cortical and subcortical limbic brain areas are activated following both stressors, as are certain brainstem nuclei (e.g., catecholaminergic cell groups) implicated in HPA activation. However, various stress-specific areas of IEG induction were also encountered. Current studies examining very early time points after stress should provide insight into the sequence of this neuronal activation.

Aging offers a model for studying the way in which the LHPA axis can become mildly dysregulated and the manner in which the multiple elements can interact to establish the most optimal homeostasis in the face of mild but persistent damage (see Table 1). The findings on aging are particularly relevant here, because they parallel in many ways some of the patterns of dysregulation observed in clinical conditions, such as severe depression (see below). Sapolsky and his collaborators (72, 74) first described an abnormality in the stress axis related to aging: whereas aged animals responded normally to restraint stress, they failed to terminate the stress response in a timely fashion, suggesting an aberrant feedback mechanism. This has led to speculations regarding the role of various brain sites, including the hippocampus, in this aberrant pattern (45, 52, 73). Although there are many controversies in this area, it appears clear that aged animals do have a decreased number of corticosteroid receptors in their hippocampi (29, 44, 72, and our own work).

However, the possible consequences or even correlates of this alteration on other levels of the LHPA are not well established. Therefore, we have examined in the same group of aged animals every level of the LHPA, looking at the expression and regulation of each of the key molecules in the system (Morano and Akil, in preparation). Our findings can be briefly summarized as follows: GR and MR receptors are decreased in the hippocampus, both at the mRNA and at the protein-binding levels (59). It might be expected that this receptor loss would result in an upregulation of CRH mRNA at the level of the PVN (41); however, we were unable to detect such a change, suggesting that control at this level may require the massive drop in receptor numbers induced by hippocampal lesions. On the other hand, we have obtained clear evidence of increased CRH secretion from the median eminence and of increased releasability of this peptide by NE. That such an increase in CRH tone does take place in these aged rats is further supported by our finding that these same animals, when compared to young controls, exhibit hyporesponsiveness of CRH receptors at the level of the pituitary, whereas their AVP responsiveness remains unaltered. This decreased CRH receptor tone leads, as expected, to a decrease in POMC mRNA levels. Interestingly, however, this is compensated for by an improvement in the translational efficiency of the POMC message.

Surprisingly, this series of compensatory changes at each level of the axis is associated with completely normal basal tone or circadian rhythmicity in these aged animals, as well as normal initial responses to stress, with the only evident functional defect being the aberrant termination of stress. These findings allow us to draw several conclusions: (1) The loss of hippocampal receptors must be primary, rather than being secondary to high circulating levels of glucocorticoids as has been previously suggested. This idea is supported by our finding of completely normal circadian patterns in the face of profound decreases in hippocampal receptors. (2) The system is capable of using a remarkable range of mechanisms to control basal and stress responsiveness at the level of circulating glucocorticoids, even when the tone of each of the individual elements is altered. (3) It may be difficult to detect even complex and extensive alterations in the axis by simply examining peripheral steroid levels. Designing appropriate challenge paradigms is of paramount importance to examining the brain elements of the axis. This will become relevant to our discussion of the stress axis dysregulation in depression (below).

Another model that has been extensively utilized to study the interplay of the various components of the stress axis has been repeated or chronic stress. This model offers the opportunity to examine the plasticity of the system, uncover the various ways in which it can adapt to continue to achieve an optimal circadian basal tone, and maintain its exquisite responsiveness to stressors. In addition, chronic stress has been proposed by Sapolsky and colleagues as a model for what happens during aging. Finally, thinking about chronic stress appears particularly relevant to thinking about the changes seen during psychiatric illnesses, as these are usually chronic and unquestionably stressful.

We shall not detail here the body of research describing the consequences of repeated stress. The effects vary based on the nature of the stressor, the duration of the regimen, the lag time between when stress is administered and the LHPA evaluated, and the age of the animal. Few laboratories have examined the entire LHPA profile in a single animal, as we have done for the aged Fisher rat. Thus, it is often difficult to make direct comparisons between studies. Nevertheless, there are some findings that appear often enough to be thought hallmarks of a chronically stressed LHPA. These have been recently summarized by Dallman (21) and are highlighted in Table 1. Note that many of the changes seen in chronic stress parallel those seen in aging. Some notable differences include the fact that CRH mRNA is elevated in chronically swum animals (López et al., unpublished), an effect not seen in aged rats. In addition, certain chronic stressors, such as footshock, increase POMC mRNA in the pituitary (80), whereas POMC mRNA in the aged rat appears decreased, although its translatability is more efficient (Morano et al., in preparation). Finally, some studies have not been carried out in exactly the same fashion in aged versus chronically stressed rats. For example, we have shown changes in the pituitary responsiveness to dexamethasone following chronic stress (89), but are unaware of parallel studies in the aged animals. Similarly, while studies on aging have focused on the natural time course of stress termination as an indicator of fast feedback, chronic stress studies have looked at the effect of preinjection with corticosterone to test fast feedback (88).

The two models described above show that repeated or long-term demands on the stress axis can certainly alter it, resulting in some aberrant responses, such as defects in feedback mechanisms or in circadian rhythmicity. However, these defects are relatively limited, and the axis can function adequately to maintain both its basal tone and it ability to respond to stress, thanks to the intricate checks and balances that it possesses. Such a pattern of partial dysregulation, with changes in set points in key brain and peripheral structures, may also take place in humans following conditions of chronic stress or in patients suffering from certain types of psychiatric illnesses.

There are numerous indications of HPA axis dysregulation in a number of psychiatric syndromes, including eating disorders, anxiety and panic disorders (see The Serotonin Hypothesis of Major Depression and The Role of Acetylcholine Mechanisms in Mood Disorders). However, the best studied case of a stress axis interface with a psychiatric illness is that of major depressive disorder (MDD). Although a detailed description of this interaction is beyond the scope of this chapter (see The Role of Acetylcholine Mechanisms in Mood Disorders, this volume), we shall briefly describe how an understanding of the HPA axis has informed the study of this syndrome and how it can continue to suggest new avenues for exploring the causes and sequelae of depression.

There is little question that the stress axis is somehow altered in major depression, at least in a significant number of patients suffering from this illness. However, there is little agreement beyond this point. Is the dysregulation of the axis primary to the disease or is it secondary to the stress of being profoundly disturbed emotionally? Is this simply a state-dependent dysregulation, or is it an underlying imbalance exacerbated when the patient is in episode? Do all patients with MDD manifest disturbances in their LHPA axis, or is this a problem seen in a subset of the subjects? If the latter is true, then can they be distinguished on the basis of clinical diagnosis or symptomatology (manic–depressive illness, melancholia, etc.), course and history of the illness, or individual variables such as age, sex, or severity of the episode? Finally, if a defect does exist, regardless of its primacy, does it reside in the periphery, the pituitary or even the adrenal, or in the brain, perhaps the hypothalamus or the limbic system?

The correlation between stress and depression has face validity, not only because being profoundly depressed is unquestionably stressful, but because external stressful events often precipitate depressive episodes (67). Epidemiological studies strongly suggest a role for both genetic factors and psychosocial stressors in unipolar depression. For example, adverse life events are likely to occur around the onset of affective illness (65) and the presence of psychosocial stressors is associated with poor outcome in depressed patients (75). Even in cases where no precipitating events may be apparent, the occurrence of a depressive episode in itself can lead to substantial psychosocial stress, secondary to the disruption of family and work relations.

A brief historical overview of research on LHPA axis and depression reveals an interesting course. The association between stress and depression has led to studies of peripheral indices of the stress axis in depressed subjects. The assumption appeared to be that the defect being sought is certainly within the brain, but that studying the pituitary–adrenal components offers a window into the brain. Thus began a number of studies that described cortisol levels in depressed subjects using either urinary free cortisol or plasma cortisol levels during the course of the day (15, 16, 39). Carroll and his colleagues introduced the Dexamethasone Suppression Test (DST) in the mid-1970s (15). The 1980s were marked by a move toward hormonal challenges that focused more clearly on the peripheral elements of the axis. Implicitly or explicitly, these studies recognized that "the window into the brain" was a complex set of organs (adrenal and pituitary) that needed to be studied in their own right, in an attempt to discern their biological status in MDD subjects. This led to challenge studies with ACTH and with CRH, and these have revealed abnormalities at both the adrenal and pituitary levels (see The Serotonin Hypothesis of Major Depression and The Role of Acetylcholine Mechanisms in Mood Disorders, this volume).

While this brief overview underscores the recent tendency to be more analytical about the individual components of the axis, it also points out that these studies have not directly tackled the role of the brain elements of the axis in the dysregulation. This is of course very difficult to do, as we only have access to peripheral or indirect measures of brain activity. Nevertheless, armed with knowledge about the basic biology of the stress axis in animals, it may be possible to devise and validate more brain-related paradigms for studying the axis in normal and depressed human subjects.

As described above, the CRH and AVP neuron (in the mpPVN) is the final brain integrator of numerous inputs which determine stress-responsiveness. The amount and frequency of release of the secretagogues by the mpPVN neurons represents a balance between positive forces that increase synthesis and activate secretion and negative or restricting forces that limit their biosynthesis and inhibit secretion. Numerous factors have been identified that perform these activating functions (e.g., serotonin, acetylcholine) or the inhibitory controls (e.g., glucocorticoids, GABA). In considering the dysregulation of the LHPA axis in depression, we can conceive of an increase in drive, or a decrease in inhibition, or both, as contributing to the disturbance.

Another feature of the axis—its circadian rhythmicity—is also important to consider in depression. Recall that circulating cortisol is highest in humans as they awaken, and falls throughout the day, reaching a nadir in the late evening. Glucocorticoids begin to rise again by 2 a.m. to 3 a.m. moving toward their peak. While MDD subjects have a normal looking rhythm or circadian pattern of steroid secretion (albeit more active) they may exhibit more subtle disruptions and differential responsiveness to stress across the course of the day. The above notions of altered drive, altered restraint, and their interaction with daily rhythm need to be kept in mind when considering major depression.

Several lines of evidence, each with its own limitations, have converged to suggest that there is increased drive in the LHPA of MDD subjects. The first is the work of Nemeroff and his colleagues (61) showing that CRH levels are high in the CSF of depressed patients, although not all workers have confirmed this observation (69). In a collaborative study with Nemeroff and Fink, we have found that both b-endorphin (derived from brain sources) and CRH are elevated in the CSF of depressed subjects and become significantly reduced following electroconvulsive therapy (62). Although there are many CRH cell groups in the brain beyond the PVN that may contribute to the observed increase in CSF CRH, it is thought that these groups may also contribute to stress perception and responsiveness. This is based on behavioral and physiological studies suggesting that CRH mediates many behavioral and autonomic symptoms of stress and negative effect (c.f. Corticotropin-Releasing Factor: Physiology, Pharmacology, and Role in Central Nervous System and Immune Disorders, The Serotonin Hypothesis of Major Depression, and The Role of Acetylcholine Mechanisms in Mood Disorders, this volume). Thus, the notion that brain CRH systems are overly active in depression has face validity. A second line of evidence pointing to increased central drive derives from our postmortem studies on the pituitaries of suicides and controls showing an increase in POMC mRNA and peptides (55). This is consistent with changes observed in animals that have been exposed to increased central drive or chronic stress (80). Finally, our metyrapone studies have been revealing in this regard. Metyrapone alone given to normal controls does not alter release of POMC products in the evening; however, the same treatment in MDD subjects results in a significant increase in POMC products in the blood. This suggests that at the nadir of the rhythm, MDD subjects have an actively driven axis which is being restrained by circulating glucocorticoids (Young et al., unpublished). Whether this increased drive can be seen at the peak of the rhythm is currently being tested. In addition, the type of steroid receptor(s) that keep this increased drive in check is of interest.

Although an increase in CRH/AVP drive is a viable hypothesis in depression, it is unclear whether this is due to an increase in the activating forces or a decrease in the restraining forces on this stress axis. Yet, a great deal is known about one of the factors that inhibits the LHPA both basally, across the circadian rhythm, and upon stress, that is, glucocorticoids acting via their receptors. We have described above multiple-feedback mechanisms that work together to maintain an optimal state of readiness and responsiveness of the axis. The coordination between multiple elements occurs in time and in space, using different mechanisms and monitoring different indices. It may be helpful to conceive of the fast, rate-dependent, brain-mediated feedback as the more sensitive mechanism, working against a background of slower, multilevel genomic control that constrains the overall tone of the system. It is therefore quite possible for these differing mechanisms to be dysregulated separately, to manifest this disruption more clearly under certain specific conditions (e.g., specific points on the daily rhythm or following stress) but to have consequences on the functioning of the entire axis. The classical DST is carried out in such a way that it is difficult to discern which level of feedback is being tested. The time delay between dexamethasone administration and HPA testing (approximately 14 to 24 hours) is too long for fast feedback but may not be long enough to fully reveal genomic feedback mechanisms. Recently, we have focused on studying mechanisms of fast feedback in normal and MDD subjects (90). Our results to date suggest that mechanisms of fast feedback are dysregulated in MDD subjects. However, our tests so far have examined the effect of rapidly rising cortisol levels on basal secretion. The question remains as to whether MDD subjects exhibit evidence of altered fast feedback following an acute stressor. These studies are of interest because control of rate-dependent feedback is thought to occur at suprapituitary levels, thus bringing us closer to studying brain rather than peripheral mechanisms. It should be noted however that not all fast feedback is rate dependent and that we have other results showing that the human anterior pituitary can indeed respond rapidly to a glucocorticoid stimulus. Thus, the rate-dependent nature of the challenge needs to be ascertained, if one is to use the paradigm as a probe of more central aspects of the stress axis.

The specific nature of the receptors involved in these feedback mechanisms is of interest; such information could help us ascertain the site of dysregulation in various patient populations, and inform us as to whether the dysfunction resides in the control of basal circadian rhythmicity, stress responsiveness or both. As described above, basic studies suggest an important role of MR in the control of circadian rhythmicity and a critical role of GR in controlling the magnitude and duration of responses to stress. However, these notions have not been specifically tested in humans, nor do we know whether major depressive disorders involve a dysregulation of MR-mediated feedback (hence the increased basal tone throughout the day?), GR-mediated feedback (exaggerated responses to stress?), or both. It will be necessary to employ GR-specific and MR-specific ligands (agonist and antagonists) to determine which receptors contribute to rate-dependent feedback either under basal or stress-induced conditions in normal subjects. These paradigms can then be extended to the study of the possible dysregulation of negative feedback mediated by one or both of these elements in MDD subjects. These studies are currently underway in our laboratory.

We have been discussing MDD subjects as a group; however, it is evident that a great deal of heterogeneity can exist in terms of the status of the LHPA axis in depression. We have concerned ourselves in the past with several sources of possible variance. We have seen differences between males and females (2) and we and others have reported profound differences between older and younger MDD subjects (2, 3, 38, 54, 63). Beyond issues of age and sex, it is clear that there may be multiple profiles of dysregulation of the LHPA axis among depressed subjects. For example, although most researchers agree that CRH challenge results in pituitary hyporesponsiveness in MDD subjects (6, 37, 43, 91), we have pointed out that this may not be easily explained on the basis of pre-CRH cortisol levels (91). It is true that a subpopulation of the subjects does exhibit high resting cortisol, but another has, in fact, subnormal resting levels of cortisol and yet a very weak pituitary responsiveness to CRH, indicating a very different status of the LHPA axis. Similar heterogeneity can be seen following dexamethasone, metyrapone, or cortisol administration. The weight of the evidence suggests that there are a number of ways that the LHPA axis can become dysregulated, all of which are seen in depression. Thus, adrenal hypertrophy in the face of a normal-appearing pituitary is one possibility, and another may be increased drive throughout the axis, or faulty negative feedback at the brain and/or pituitary level. The challenge before us is twofold: (a) to describe these patterns as fully as possible by carrying out multiple measures at the various levels of the axis across multiple time points and (b) to try to understand whether some of these patterns are related and represent different stages of dysregulation, for example, is adrenal hypertrophy a late stage of the increased drive pattern? Factors, such as the history of the patient, and the number and frequency of past episodes need to be considered to ascertain the possible relationship between a given LHPA profile captured at a certain point in time and the overall course of dysregulation across the lifetime of an individual.

One variable that has consistently emerged as an important source of heterogeneity is age (2, 3, 38, 54, 88). Whereas this may well represent an "end stage" dysregulation following repeated episodes and may also reflect changes in hormonal status in postmenopausal women (2), it may also result from interesting changes in the brain specifically related to mechanisms of fast feedback. We have described above a body of research in rats that suggests that aging is accompanied by significant changes in the LHPA axis. Interestingly, not all aged rats exhibit the disruption in negative feedback nor do they all exhibit the loss of pyramidal cells. The work of Meaney and his colleagues (44) has suggested an interesting correlation between hippocampal damage and feedback abnormality. In addition, Meaney et al. (57) have shown that stress early in life may be protective, decreasing the likelihood of hippocampal damage in aging. It is also interesting to note that the aged animals with the most clear-cut feedback aberration also exhibit clear-cut memory deficits (44). The above observations are extremely suggestive, especially that age alone does not appear to be sufficient to produce the damage; rather, a combination of factors (e.g., a certain developmental history and a history of repeated stress interacting with aging) needs to occur for the damage to become apparent. This is of particular relevance to the study of depression. Whereas normal aged subjects may have normal basal rhythms of steroids (Tiongco et al., personal communication) and normal patterns of turn off following a stressor (although this latter has not been tested), it is possible that older MDD patients would exhibit aberrations in the LHPA axis, particularly in the stress-termination response. If this were the case, and if indeed it were due to a loss of corticosteroid receptors in the hippocampus, then this disruption would not be state dependent. Rather, we would expect to detect altered patterns of turn-off to the stressor even when the elderly subject is euthymic. Here again, basic LHPA studies would lead us to design specific paradigms of relevance to understanding the nature of the dysregulation in depression, in this case in the aged depressed population.

We hope that we have provided the reader with a sense of the richness and plasticity of the LHPA axis and the impressive body of information we currently possess about it. Although much remains to be learned, we are in the enviable position of knowing, at the molecular level, many of the key elements involved in the control of the axis. We can also describe the broad integrative features of the system and are beginning to delineate the neuronal circuits that participate in orchestrating the molecules into an exquisitely tuned and responsive physiological system. It is against this background that we can begin to understand how the stress axis copes, or fails to cope, with extreme demands placed upon it, be they exposure to repeated stress, aging, or physical and psychiatric illnesses. Possibly the two main features we would emphasize about the system are (a) the multiple levels of control from the hippocampus (and possibly above) to the adrenal gland, and (b) the constant interplay of activation and inhibition to set optimal levels of circulating glucocorticoids, both at rest and following stress. The existence of multiple levels of control throughout the axis should explain the difficulties in pinpointing a single dysregulation to be observed in chronic stress, aging, or psychiatric disorders.

There are some hints of fascinating interfaces that we have not really fully explored, either in this chapter or in terms of research efforts in the field. One of them is the functional significance of the hippocampus in this axis. Although research has focused on the role of the hippocampus in basal control or in stress termination, only a few studies have addressed the question: why the hippocampus? The work of Levine and coworkers (18) and more recently DeKloet and his colleagues (25) and Meaney and his colleagues (57) points to an intricate interface between stress and learning and memory. Given what we know about the general functions of this structure, it may be more reasonable for us to view the hippocampus, not simply as a major controller of the stress axis, but as a critical structure that integrates information from all sources, attributing salience to it, and determining its relevance to general information processing at any one time. In this context, it would be reasonable for the hippocampus to both monitor (possibly through the steroid receptors) and modulate the status of the stress axis. This view becomes particularly plausible if one thinks of the stress axis not simply as an emergency system but also as a constant monitor of internal and external events, a function it needs to perform routinely to detect emergencies and to allocate priorities to maintenance functions (e.g., eating, drinking, reproduction) versus fight-or-flight. That this is a reasonable concept of this system is supported by the intricate controls we have alluded to not only on regulating stress responsiveness but on regulating the basal tone of the LHPA axis. To function adequately as a monitor or detector, the stress system needs to access historical information regarding any past encounters of the organism with a given stimulus (i.e., memory) and the valence or affect associated with the stimulus. The hippocampus, along with other limbic structures, would thus play a key role in determining the very assignment of a stimulus as stressful or nonstressful. In turn, whether or not a stimulus or a situation are coded as stressful would be critical to hippocampal processing in the course of learning and remembering.

Such a view of the stress axis and its interface with both metabolic and cognitive functions, may bring this system to a more central position in our concept of the biology of major depression. Of course the interface between stress and depression may be simply the result of the fact that depression is a chronic stressor. However, it may also be possible to conceive of a disruption in stress control as a critical feature in the development and/or maintenance of depression. This idea may entail a slight reframing of the standard view of major depression. Although severe depression is certainly defined by negative affect as its major feature, it can also be construed as involving a dramatic imbalance in the assignment of salience and valence to internal and external stimuli. Severely depressed people not only experience unpleasant emotions, but attend to them very strongly; their feelings appear to take precedence over other social, emotional, cognitive, or even metabolic functions. The substantial disruptions in appetitive activities, such as sleep, eating, and sex, is indicative of a dramatic shift in the priorities of the organism. Although we do not mean to imply that a dysregulation in the LHPA axis is necessarily at the root of depression, it appears reasonable to propose that a disruption in the stress axis may represent a necessary feature of severe depression. Such a disruption may represent the loss of a key control mechanism that allows humans to keep their biological and psychological priorities in working order. In someone who is vulnerable to affective disorders, an adequately functioning stress axis may serve as a break to reset matters on a more physiologically and psychologically balanced course, whereas a loss of the ability to cope or assign priorities to the various demands on the system may represent the very trigger necessary for progressing from sadness and negative affect to a serious depressive episode. Beyond that point, it is easy to envision the establishment of a vicious cycle, whereby the depressive episode further dysregulates the axis, and this, in turn would make the next episode more likely. The genetic, environmental, molecular, and neurobiological antecedents of a resilient stress axis are far from understood, and this area may represent the next challenge for both basic and clinical investigation.

published 2000