Interactions Between the Nervous System and the Immune System

Adrian J. Dunn

Interaction between the nervous and immune systems provides a physiological basis for psychosomatic medicine. In approximately 200 AD, the Greek author Galen wrote that melancholic women are more susceptible to breast cancer than sanguine women. Since then, a wealth of anecdotal evidence has convinced physicians of the importance of psychological factors in the prognosis of disease. This belief is now bolstered by substantial evidence that nervous system output can indeed modulate immune function. However, the interactions are not unidirectional. The immune system can also have powerful influences on the nervous system. This should not be surprising to anyone who has ever felt sick. Anomalies of immune system function can certainly cause diseases of the nervous system, and this may be manifest in psychiatric disease. It is clear that effective defense against infections requires a complex coordination of the activities of the nervous and immune systems, and that abnormalities in the relationships between the two systems can cause disease (see also General Overview of Neuropeptides, Neuronal Growth and Differentiation Factors and Synaptic Placticity, The Role of Acetylocholine Mechanisms in Mood Disorders, and Neuroendocrinology of Mood Disorders).

The interactions of the nervous system and the immune system have been the subject of a number of critical reviews (e.g., 20) and several books, of which the most recent comprehensive text is that edited by Ader et al. (2).

The experimental evidence for nervous system– immune system interactions can be summarized:

This evidence is discussed below, but it is pertinent to consider first what the potential links between the nervous and immune systems might be. Classical medical teaching leaves little scope for interactions between the nervous and immune systems, but with the benefit of hindsight we can postulate a number of specific mechanisms by which the nervous system might affect immune function. These include glucocorticoids secreted from the adrenal cortex, catecholamines secreted from sympathetic nerve terminals and the adrenal medulla, other hormones secreted by the pituitary and other endocrine organs, and peptides (including endorphins) secreted by the adrenal medulla and autonomic nerve terminals. The network includes not only the autonomic nervous system and classical neuroendocrine mechanisms, but involves an endocrine function of the immune system. A variety of immune system products (e.g., cytokines, {1} peptides, and other factors) that function to coordinate the immune response may also provide important signals for the nervous system. Thus chemical messengers can account for a variety of interactions between the nervous system and the immune system. Fig. 1 provides a schematic of the most well-known interactions between the nervous system and components of the endocrine and immune systems.

Compelling evidence for an influence of the nervous system on the immune system arises from studies that indicate that behavioral conditioning can modify immune responses. Many early observations suggested this, but more recent studies have provided strong experimental support (see ref. 1). A landmark study by Ader and Cohen published in 1975 found that after the immunosuppressive drug, cyclophosphamide, had been paired with the taste of saccharin, subsequent ingestion of the saccharin prevented the production of antibodies in response to sheep red blood cell (SRBC) administration (1). These findings were not greeted with universal enthusiasm; many immunologists were reluctant to accept the notion that immune system function could be regulated by the brain. However, the original findings were replicated and extended by the original authors and others. The technique has been used to prolong the lives of mice with the autoimmune disease lupus erythematosus. Thus far, there have only been reports of conditioned augmentation of immune activity from one research group, and the effects have been small. There can be little doubt that conditioning can alter immune responses, but the immunological specificity of the effects is not clear, and the mechanisms are as yet unknown. It is possible that at least some of the immunosuppressive effects are from a conditioning of hormone and neurotransmitter secretion (e.g., glucocorticoids or catecholamines).

Although many studies have indicated that brain lesions have an effect on immunity, the literature is exceptionally fragmented and complex (see ref. 70). Effective lesions are most commonly located in the hypothalamus and are generally inhibitory. Lesions in other limbic areas may also be effective, notably in the septum, hippocampus, and amygdala. Some studies have indicated that cortical lesions can affect immune responses and that the effects depend upon the laterality of the lesion. Renoux et al. have reported evidence that lesions of the left cortex, but not the right, produced pronounced immune deficits in spleen cell number, lymphocyte proliferation, and natural killer (NK) cell activity (68). The lateral specificity indicates that the effect cannot be from nonspecific effects of the lesion, and it could account for the greater number of left-handed individuals who exhibit diseases of the immune system. Lesions of the central noradrenergic systems have also been shown to impair various aspects of the immune response.

Everybody knows that stress impairs the immune system, but the truth is probably much more complicated. Certainly chronic stress is unhealthy, but the mechanisms involve much more than a depression of immune function. The dogma that stress suppresses immunity is to some extent based on the well-established immunosuppressive effects of glucocorticoids. However, the supraphysiological doses of the steroids used in most of the studies do not allow simple extrapolation to the normal physiological state. In fact, endogenous glucocorticoids at physiological doses are not universally immunosuppressive and actually may enhance immune function see below). Furthermore, glucocorticoids may not even be the major mechanism by which stress suppresses immune function.

Experimental data from both animal and human studies have confirmed the immunosuppressive effect of stress (1, 43, 49). However, it is important to emphasize that there is considerable evidence to suggest that stress may also enhance immune function (1). Common human experiences suggest that under acute stress conditions (approaching examinations, grant deadlines, etc.) an impending infection can be held at bay, but resistance collapses when the pressure is relieved. Some animal experiments also suggest that mild acute stressors may actually enhance measures of immunity (e.g., see ref. 92; for a more complete review, see ref. 1; see also Neuroendocrine Interactions and Stress).

Adrenalectomy has been shown to prevent the immunosuppressive effects of stress in some animal studies (20, 43), but many other studies have found that stress-induced changes in immunity persist in adrenalectomized animals (8). Adrenalectomy appears to be effective in studies that have examined acute responses to brief stressors (for which the immunosuppressive effects are rapidly reversed), but may be less important for the effects of chronic stress (20, 43). Adrenalectomy does not permit a distinction among the effects of steroids, catecholamines, or even of neuropeptides secreted by the adrenal gland. More recent studies have suggested an important role for the circulating catecholamines, derived from the sympathetic nervous system and adrenal medulla, in the chronic studies (see below).

The choice of immune parameters measured may also influence the results. Earlier studies relied heavily on mitogen-stimulated proliferation assays, which assess the responsivity [i.e., cell division measured by deoxyribonucleic acid (DNA) synthesis] to lectin mitogens [such as concanavalin A (Con A), phytohemagglutinin (PHA), lipopolysaccharide (LPS), or pokeweed mitogen] in vitro. The interpretation of such assays is questionable, because the results are susceptible to a large number of extraneous influences, and the assays are conducted after several days of in vitro incubation separated from normal physiological influences (55). Also, the data typically display high variability. A measure used more often recently has been that for NK cells. There is good evidence that NK cells are involved in the rejection of tumors (4), and therefore at least one of their immunophysiological functions is clear. Stressful treatments have been shown to suppress NK cell function in both animal and human studies (45, 77). The major effector for the stress-induced effects on NK cell function appears to be opiates (77) and catecholamines through b-adrenergic receptors (84).

Because most of the studies of stress on immune function have used ex vivo procedures (i.e., stress in vivo, immune assays in vitro), another important factor is whether or not the population of cells sampled may be altered by the in vivo treatment. Cell trafficking, the movement of lymphocytes around the body, is known to be regulated by hormones and other secretions, including those secreted during stress, and it is likely, therefore, that the stressful treatments alter the population of cells harvested for the in vitro studies (see below).

The best known mechanism for an influence of the nervous system on the immune system is circulating glucocorticoid hormones secreted by the adrenal cortex. Glucocorticoids have long been known to have immunosuppressive effects (13, 88). The data derive in part from the medical practice of using glucocorticoids postsurgically to decrease tissue inflammation and the rejection of transplanted tissues. However, considerable experimental data suggest that the effects of glucocorticoids are not exclusively immunosuppressive (47, 88).

Although it is well established, it is too often forgotten that glucocorticoids are essential for normal immune responses. Extensive experimental data from animal and human studies indicates that adrenally compromised individuals are more susceptible to infections and that the adrenal cortex is more important than the medulla in this respect (47). Replacement studies clearly implicate a major role for the corticosteroids in immune defense mechanisms. Of particular importance is the work of Kass (see ref. 47), who showed that an optimal concentration of corticosteroids was essential for normal recovery from infections in adrenalectomized animals.

Nevertheless, the extensive evidence for the immunosuppressive effects of glucocorticoids should not be ignored. It should, however, be viewed in the light that most of the data were generated using high doses of synthetic glucocorticoids, such as prednisolone, triamcinolone, or dexamethasone, which are considerably more potent than the native steroids. The concentrations of these compounds used clinically can cause lysis of immune cells, especially immature ones. The more careful studies have used natural steroids at doses in the normal physiological range; these have noted stimulatory effects of steroids at lower doses (see ref. 88). Inhibitory effects occur at higher doses, typically 10-6 M, which is close to the maximum concentration of free corticosterone or cortisol found in stressed animals after correcting for that bound by corticosteroid-binding globulins (20). It is also important that elevations of plasma glucocorticoids following acute stressors are short-lived.

Although there are direct effects of glucocorticoids on immune cells in vitro, there may also be indirect ones in vivo. One of the oldest known physiological correlates of stress is the involution of the thymus. This involution, which can decrease thymus weight by more than half, occurs largely because lymphocytes that normally reside there are driven out to the periphery. Stress-induced thymic involution is prevented by adrenalectomy and can be induced by administration of glucocorticoids (13). Thus glucocorticoids can alter the body's distribution of lymphocytes, which may in itself be an important factor marshalling the immune response to infection. Moreover, as mentioned above, the population of lymphocytes derived by harvesting tissues from animals subjected to experimental treatments may be altered by the redistribution of cells due to glucocorticoid secretion. This should be an important consideration in interpreting the results of ex vivo data.

Lymphocytes bear both a- and b-adrenergic receptors. Catecholamines appear in the circulation from both the adrenal medulla [norepinephrine (NE) and epinephrine] and from sympathetic terminals (NE). In addition, lymphocytes may be exposed more directly to neuronal secretions while they are resident in the thymus, spleen, and lymph nodes. Anatomical studies have clearly demonstrated a sympathetic innervation of immune structures, such as the bone marrow, thymus, spleen, and lymph nodes (28). Thus lymphocytes could be exposed to high local concentrations of catecholamines, as well as neuropeptides. A parasympathetic (i.e., cholinergic) innervation of these organs, has not been confirmed (see ref. 28).

In vitro studies have revealed adrenergic effects on lymphocytes. Early studies suggested separate a- and b-adrenergic effects; b-adrenergic receptors were largely inhibitory, whereas a-adrenergic receptors were stimulatory (35). This generalization has endured to some extent, but the detailed results are very complex. There appear to be separate a- and b-adrenergic stimulatory effects on antibody production in vitro (71), whereas NK cell activity appears to be inhibited by b-adrenergic stimulation.

The results of in vivo studies have been bewilderingly complex. Depending on the parameters used, sympathectomy has been shown to impair, enhance, or not change immune responses (54). In general, sympathectomy in adult animals depresses immune reactivity, but there are also paradoxical effects on lymphocyte proliferation and B cell differentiation. Among the confounding factors that may contribute to the complexity are compensatory increases in adrenomedullary output, redistribution of lymphocytes, compensatory changes in the number and kind of adrenergic receptors, and the coexistence in sympathetic terminals of peptides, such as neuropeptide Y (NPY).

Several recent studies have suggested that a major mechanism by which NK cell activity is regulated in vivo involves catecholamines released by the sympathetic nervous system. For example, the inhibitory effect of intracerebroventricular corticotropin-releasing factor (CRF) on NK activity is blocked by the ganglionic blocker, chlorisondamine (44), as is the immunosuppressive effect of icv interleukin-1 (IL-1) (83). There is also direct evidence that b-adrenergic receptor blockade can prevent stress-induced effects on NK cell activity (16).

Sympathetic terminals contain not only NE, but also neuropeptides, including endorphins, which may act on the immune system. Felten et al. described the presence of NPY, substance P (SP), and vasoactive intestinal polypeptide (VIP) in the thymus spleen and lymph nodes, as well as calcitonin gene-related peptide (CGRP) in the thymus and lymph nodes, enkephalin and somatostatin in the spleen, tachykinin in the thymus, and peptide histidine isoleucine in lymph nodes (28).

It has been shown that lymphocytes can synthesize and secrete certain peptides. The spectrum of peptides reported to be synthesized is large, and includes many of the known peptide hormones, as well as the hypophysiotropic factors. The peptides include adrenocorticotropin hormone (ACTH), CRF, growth hormone, thyrotropin (TRH), prolactin, human chorionic gonadotropin, the endorphins, enkephalins, SP, somatostatin, and VIP (11, 34). The quantities of the peptides produced are typically very small, and their biochemical characterization has often been perfunctory. Sometimes their existence has been inferred only from the results obtained in the very sensitive assays used to detect their messenger ribonucleic acids (mRNAs), which should not be construed as unequivocal evidence for the presence of the peptides themselves. More careful analyses have not always substantiated the original claims, especially for the endorphins (76). There is probably considerable variability in the ability of lymphocytes from different sources to produce a specific peptide, but this issue has received no serious attention in the literature.

The physiological significance of this production of peptides is not at all clear. Because in many cases lymphocytes display receptors for these same peptides, they may function as chemical messengers within the immune system. However, Blalock has suggested that the peptides may also have systemic functions, for example, ACTH could activate the adrenal cortex. Although there is no good experimental support for this specific example (see below), it is possible that there may be a local bidirectional communication between lymphocytes and other cells. One example of this communication may be in the spleen, where CRF appears to be present in the innervating neurons and CRF-receptors are present on resident macrophages (91). Another example involves endorphins; b-endorphin produced by lymphocytes in an area of inflammation may exert an analgesic action directly on sensory nerve terminals (79). Such a mechanism is attractive, because the concentrations of the peptides produced locally may be adequate to exert such effects, and the metabolic lability of peptides would ensure that the effect was localized.

The principal known actions of peptides not covered elsewhere in this chapter on immune function can be summarized briefly. Substance P is a potent stimulator of cellular proliferation and cytokine [IL-1, IL-6, and tumor necrosis factor (TNF)] production (60). Its proinflammatory properties may be important in rheumatoid arthritis and inflammatory bowel disease. Vasoactive intestinal polypeptide may be involved in the regulation of lymphocyte migration (66). The interested reader is referred to reviews of this very complex topic (2, 78).

Many other hormones are known to affect the immune system. Firstly, there are the hormones of the hypothalamic–pituitary–adrenocortical (HPA) axis, each of which has been reported to affect immune function: CRF, ACTH, and the endorphins.

Corticotropin-releasing factor itself has been reported to have a variety of effects. The reported direct effects of CRF on immune cells have generally been stimulatory. For example, CRF has been shown to stimulate B cell proliferation and NK activity, as well as IL-1, IL-2, and IL-6 production (52). Receptors for CRF have been found on immune cells (91), providing a mechanism for these effects. Although it seems unlikely that CRF in the general circulation ever achieves concentrations high enough to stimulate these receptors, it is possible that local actions may occur, for example, in the spleen (91). By contrast, CRF injected intracerebroventricularly (icv) has largely inhibitory effects on immune function. A major effect of icv CRF is evident on NK cell activity and appears to be mediated through the sympathetic nervous system (44). The footshock-induced reduction of NK activity appears to be mediated by cerebral CRF, because an antibody to CRF injected icv but not peripherally prevented the shock-related response (46).

Although, ACTH has been shown to have some direct effects on immune function, including an inhibition of antibody production and modulation of B cell function (38), the effects have not been striking. On the other hand, the endorphins have been shown to exert a plethora of effects on immune function (10, 38). Lymphocytes possess binding sites for opiates, but at least some of these are not sensitive to the opiate antagonist, naloxone (see the review by Carr in ref. 10). Interestingly, binding sites have been found for N-acetyl-b-endorphin, which is the commonest form of b-endorphin secreted from the anterior pituitary and has no opiate activity (75). b-Endorphin and other opioid peptides can exert effects on lymphocytes in vitro (10). By and large, the effects are facilitatory. Such effects have been observed on NK activity as well as on proliferative responses (38). Opioid peptides are also chemoattractants for lymphocytes. By contrast with the enhancing effects in vitro, in vivo opiates are largely inhibitory, especially on NK activity. This apparent contradiction can be explained, because, at least in the case of morphine, the site of opiate action appears to be in the central nervous system (CNS) (see ref. 90). Moreover, the effects appear to be mediated by the adrenal gland, most probably by catecholamines (4).

Perhaps the most interesting effect of a pituitary hormone on the immune system is that of prolactin. As summarized by Bernton et al. (5), its effects are largely stimulatory. Reduction of pituitary prolactin secretion (e.g., by dopaminergic agonists or opiate antagonists) impairs immune function and increases susceptibility to infections, such as Listeria monocytogenes, whereas stimulation of prolactin secretion (e.g., by D2 dopaminergic antagonists or opiates) can enhance it. Bernton et al. postulate that prolactin may be the counter-regulatory hormone to glucocorticoids, and opposing interactions between these two hormones on immune function can be demonstrated in vivo (5). Direct effects of prolactin on lymphocyte function have been difficult to demonstrate, but prolactin antibodies do impair proliferative responses in vitro. Lymphocytes can produce a prolactinlike protein, although its identity with prolactin has not been demonstrated. Thus it appears that prolactin is yet another example of a multifunctional peptide produced by both the pituitary and the lymphocytes.

Few would challenge the notion that sickness is stressful. In his autobiography, Hans Selye indicates that it was the common characteristics of sickness regardless of the underlying disease "the syndrome of just being sick" that first interested him in stress research and led him to advance his much maligned proposal of the nonspecificity of stress (74). That the HPA axis is activated following infections has long been known. During World War I, it was noted that fatalities from infections were associated with striking morphological changes in the adrenal cortex (47). It was later discovered that endotoxin (lipopolysaccharide, LPS), a potent stimulator of the immune system, stimulated the HPA axis. Subsequently, it was shown that infection of rats with Escherichia coli increased the secretion of ACTH.

The concept was expanded considerably when Besedovsky et al. showed that increases of plasma corticosterone accompanied the appearance of cells producing antibodies to such commonly used antigens as SRBCs. The observation, replicated by Besedovsky et al. and Saphier (see review in ref. 72), complemented earlier Russian observations of the electrophysiological activation of cells in the medial hypothalamus accompanying an immune response. Besedovsky also showed that SRBC inoculation changed the apparent turnover of NE in the hypothalamus.

There now seems to be a consensus that the primary physiological responses in stress are the activation of the HPA axis and of peripheral and central catecholamines. To the extent that infections and immune challenges exert the same physiological effects, they can be regarded as stressful (Table 1).

The foregoing has indicated that infections and immune challenges can activate the HPA axis, but is this a specific response or merely the reaction of the organism to a disturbance of its homeostasis? Work by Besedovsky and others implicates cytokines produced by the immune system as mediators of the HPA response, which suggests that it is specific. The initial observation was that supernatants of immune cells challenged in vitro with mitogens, such as Con A, had the ability to activate the HPA axis. The active factor synthesized and secreted in response to the mitogens was suggested to be IL-1, because the supernatants of Newcastle disease virus- (NDV-) treated lymphocytes could be neutralized with an antibody to IL-1, and because injection of purified recombinant human IL-1b was found to be a potent stimulator of the HPA axis (6). It is notable that IL-1 is a considerably more potent activator of ACTH and glucocorticoid secretion than CRF itself. A scheme for this arrangement is depicted in Fig. 2.

It is important to distinguish the HPA activation that occurs at different stages in the immune response. The HPA activations related to treatment with SRBC and other antigens occurred 5 to 8 days following treatment and coincided with the peak production of antibody. Similar observations have been noted by others with myelin basic protein (53) or keyhole limpet hemocyanin (81). However, there is also an acute HPA response to immune stimulation that occurs within the first few hours. This acute response is observed following LPS and NDV administration (22, 81). It seems likely that this early response is related to cytokine secretion (see below), but whether cytokines are responsible for the later response is not known.

The mechanism of the activation of the HPA axis by immune stimuli has been the subject of intense investigation. Speculation has largely centered on the role of IL-1. However, Blalock (7) has suggested that lymphocytes may synthesize and secrete their own ACTH following immune stimulation and that this ACTH may be sufficient to activate the adrenal cortex. This mechanism has been excluded because rigorous hypophysectomy prevents the responses to NDV (26). Evidence exists for IL-1-induced activation of the axis at every level. However, the bulk of the evidence strongly favors the need for hypothalamic CRF. Deafferentation of the hypothalamus and lesions of the paraventricular nucleus (PVN) prevent the ACTH response to IL-1, and hypophysectomy prevents the effect of IL-1 and LPS (24). Moreover, antibodies to CRF attenuate or block the ACTH and corticosterone responses to IL-1 (e.g., ref. 89). The data that conflict with this are largely based on in vitro studies, which are notoriously subject to artifact. For example, although several authors have reported that IL-1 can stimulate hormone release from the pituitary in vitro, the results are remarkably inconsistent (see ref. 21), and, in general, prolonged incubation in vitro is necessary to observe such responses. Similar considerations apply to the in vitro studies on adrenal tissue.

As indicated above, Besedovsky first reported changes in cerebral NE metabolism associated with immune activation (see ref. 26). Injection of SRBCs caused an apparent decrease in NE turnover as determined following blockade of synthesis with a-methyl-p-tyrosine. This change coincided in time with the peak antibody response at 5 to 8 days. Subsequent studies have confirmed changes in NE metabolism associated with immune stimulation, but in all cases the changes have suggested increased rather than decreased metabolism (26). Administration of SRBCs decreased the NE content of the PVN 4 days, but not 2 or 6 days, later. The decrease in NE in the PVN could reflect increased NE release. Subsequent postmortem neurochemical studies using a variety of different challenges have demonstrated acute increases in cerebral NE catabolites to challenges such as NDV (26) and LPS (22, 40). These neurochemical changes are greatest in the hypothalamus; changes in other brain regions are significantly smaller. Dopamine (DA) metabolism is affected to a lessor extent and only by LPS (22). Thus the changes contrast with those typically observed for behavioral stressors, which cause similar changes in NE metabolism in most brain regions and exhibit a pronounced activation of dopaminergic systems, especially in the prefrontal cortex. The latter is rarely observed following immune stimulation. The increases in biogenic amine catabolites (accompanied in a few cases by decreases in the parent amines) suggest increased synaptic release; studies with in vivo microdialysis reinforce this possibility. Like the HPA responses, those in the catecholamines have been best characterized after acute stimulation, but some studies suggest that similar changes occur later during the primary immune response, and this has been confirmed by studies of metabolites following SRBC administration (94).

The neurochemical changes are not confined to the catecholamines; indolamine metabolism is also affected. Decreases of 5-HT have been observed in the paraventricular and supraoptic hypothalamic nuclei. Increases in the serotonin catabolite, 5-hydroxyindolacetic acid (5-HIAA) were observed acutely following NDV (26). These increases in 5-HIAA accompany increases in tryptophan (22, 40). Interestingly, the catecholaminergic and serotonergic responses can be distinguished. In contrast to the hypothalamic emphasis of the noradrenergic responses, the indolamine responses have no apparent regional specificity (22). Moreover, they follow different time courses; the increases in DA and NE catabolism peak after 2 hours, whereas the indolaminergic ones are much slower, reaching a peak at approximately 8 hours (22, 40). Interestingly, these changes in brain tryptophan and 5-HIAA appear to be dependent upon the sympathetic nervous system, because ganglionic blockade with chlorisondamine prevents changes in these neurochemicals normally induced by IL-1 or LPS (27).

Recently, we have demonstrated a complete dissociation in the responses to LPS. Endotoxin-resistant mice (C3H/HeJ strain) do not exhibit significant noradrenergic responses (in parallel with the diminished HPA response), but do exhibit normal indolaminergic ones; however, nitric oxide synthase inhibitors can block the indolaminergic responses without affecting the noradrenergic (or HPA) responses.

Administration of IL-1 mimics the neurochemical responses to LPS and NDV very closely. The increased NE metabolism is focused on the hypothalamus, with a regionally nonspecific increase in tryptophan and in 5-HT metabolism (19, 22). The only distinction is that LPS exerts a modest effect on DA. Thus the neurochemical data complement the HPA data implicating IL-1 as a mediator of the responses.

Because NE is considered to be a major regulator of CRF secretion, it is reasonable to ask whether NE is involved in the response to IL-1. Studies in rats lesioned with 6-hydroxydopamine (6-OHDA) suggest that the ventral noradrenergic bundle input to the hypothalamus is indeed necessary for HPA responses to IL-1 (12), but studies with adrenergic blockers have been less successful in demonstrating adrenergic control of the HPA response to IL-1 (69), although we have observed a partial blockade of the HPA response to IL-1 by the a1-adrenergic antagonist, prazosin, but not by a2- or b-adrenergic receptor antagonists.

The activation of the HPA axis associated with immune responses has been interpreted to indicate that the immune system can act as a sensory system, signaling the brain to indicate the presence of a threat from the external environment and triggering a classical stress response (7). The role of the HPA activation has been suggested to be a negative-feedback mechanism provided by the immunosuppressive activity of the glucocorticoids. The inhibitory activity of the glucocorticoids limits inflammatory responses and prevents the immune system from overreacting and causing autoimmunity (6, 62). Consistent with this result, adrenal corticosteroids have been shown to play a critical role in the recovery from experimental allergic encephalomyelitis (EAE). Rats treated with myelin basic protein produce antibodies to this protein 11 to 14 days after immunization, at the same time that paralysis of the tail and hind limbs appears and plasma corticosterone is appreciably elevated (53). Intact rats normally recover within a few days, but the spontaneous recovery does not occur in adrenalectomized rats, unless glucocorticoid replacement therapy is instituted.

Conversely, a decreased responsivity of the HPA axis is associated with the susceptibility to arthritis in Lewis rats. Lewis rats show an arthritic response that mimics human rheumatoid arthritis in response to administration of a streptococcal cell wall peptidoglycan polysaccharide (SCW), whereas the histocompatible Fischer rats do not. The arthritic response in Lewis rats can be prevented by dexamethasone and can be induced in Fischer rats by the glucocorticoid-receptor antagonist RU 486 (82). The deficit in Lewis rats appears to be associated with a deficient activation of the HPA axis to SCW, IL-1, and CRF (compared to Fischer rats), and may be associated with an inappropriate regulation of the CRF gene. These two examples indicate the physiological significance of the glucocorticoid feedback on the immune response.

The most likely messengers from the immune system to the nervous system are the cytokines. In this sense, cytokines, the hormones of the immune system are immunotransmitters. Although the cytokines play a major role in coordinating the immune response, they also have substantial effects on other tissues, including the nervous system. A large number of cytokines has been identified, including thirteen interleukins, several interferons, and a variety of other factors. Those currently known to have the most relevance for the nervous system are IL-1, IL-6, TNFa, and the interferons, but many others may soon be recognized. Shortly following most immunological challenges, macrophages produce TNFa, followed closely by IL-1 and then IL-6 (95). Surprisingly, each of these three cytokines is capable of inducing the others, so that IL-1 administration induces TNFa synthesis and vice versa (18); IL-1 and TNFa induce IL-6, and so on.

All of the interferons (IFNa, IFNb, and IFNg) are secreted by lymphocytes in response to viruses and RNA (IFNb2 has been renamed IL-6). Each of them has a variety of immunological effects that are regarded as antiviral. They include inhibition of T-cell proliferation, enhancement or suppression of antibody synthesis, and enhancement of NK activity (30).

These cytokines have a plethora of effects, many of which are related to the physiology of infections. Interleukin-1 is not the only cytokine that can affect the HPA axis; other cytokines, such as IL-6 and TNFa, can have similar effects (58), although they are significantly less potent (23). Interestingly, the production of TNFa and IL-1 is inhibited by glucocorticoids, so that the HPA activation elicited by the cytokines provides feedback regulation of cytokine synthesis (95). Also, IFNa affects the HPA axis, its effects being largely inhibitory and apparently mediated through opiate receptors. Interleukin-1 and -6 have a variety of other endocrine effects (21) and along with TNFa are each pyrogenic (18).

Interferon a is used therapeutically as an antiviral agent, although its utility has often been limited by its neurological and psychiatric side effects. It can cause an opiate-like euphoria, polyneuropathies, and behavioral and motor deficits and can alter the electroencephalogram (EEG) (61). The neurotoxic effects of both IFNa and IFNg have been postulated to depend on the altered metabolism of tryptophan. This effect may be mediated largely by an induction of indolamine 2,3-dioxygenase, a catabolic enzyme for tryptophan (39), so that plasma concentrations of tryptophan are reduced, for example, in patients with autoimmune deficiency syndrome (AIDS) (31). There is also increased metabolism of tryptophan to kynurenine and quinolinic acid (39). Quinolinic acid is a potent neurotoxin, which may be involved in neurodegenerative disorders of the CNS. Its presence in the brain is correlated with inflammatory responses from a variety of sources, including viral infections (41). The potential effects of the IFNs on serotonin and its functions are largely unknown.

Both IL-1 and IFNa are electrophysiologically active and can alter the EEG (72). Many of the effects of IFNa are prevented by naloxone, suggesting an interaction with opiate receptors.

The cytokines are also behaviorally active. Thus, IL-1 and TNFa are somnogenic (65); IL-1, TNFa, and IFNa are anorexic (15, 59); locomotor activity is decreased by IL-1 (67) and IFNa (15); and IL-1 decreases exploratory activity (25). In most cases, these responses can be elicited by both central and peripheral administration, and some may be mediated by CRF (25). Each of these responses is characteristic of sickness, and a case can be made that IL-1, perhaps in combination with other cytokines, accounts for many if not most of the physiological and behavioral responses to infections. These responses can clearly be rationalized in terms of what has been termed "sickness behavior" (37, 48). The fever may be instrumental in fighting invading organisms, and the behavioral responses (sleep, anorexia, hypomotility) may cause the animal to hide and thus escape predators (37).

The immunotransmitter activity of cytokines has been tested to some extent with antagonists. Thus, the IL-1 receptor antagonist (the IL-1ra, a naturally occurring polypeptide often synthesized and secreted along with IL-1) can prevent the HPA-activating and neurochemical effects of some immune challenges, such as NDV (24). The IL-1ra also blocked the effects of LPS on social behavior in rats but not the anorexia (48) or the HPA activation (23). Thus, IL-1 may not be the only factor mediating responses to LPS. A TNFa antibody was also ineffective against LPS, even in combination with IL-1ra (23). Because immune challenges produce a "cocktail" of cytokines, it seems likely that differences in these cocktails account for the different patterns of responses to different immune stimuli. As more suitable cytokine antagonists become available, this hypothesis can be tested.

An important question is the mechanism by which IL-1 activates the HPA axis. Given that activation of CRF-containing neurons in the hypothalamus appears to be required, does IL-1 penetrate the brain and directly activate CRF-containing neurons in the PVN or are other intermediates involved? The answer to this apparently simple question may be quite complex. When injected directly into the brain, IL-1 does indeed activate the HPA axis; thus it is possible that IL-1 acts directly in the brain. However, a number of observations are not consistent with a direct action of peripherally administered IL-1 on the hypothalamus: IL-1 is a relatively large protein (molecular weight 17,500) so that it is unlikely that it readily penetrates the blood–brain barrier, although it could act on one or other circumventricular organs, such as the median eminence (ME) or the organum vasculosum laminae terminalis (OVLT) that lack a blood–brain barrier. Moreover, studies of IL-1 binding in the brain have not been encouraging. In the rat, there has been no clear demonstration of cerebral binding sites for IL-1, except for those in the capillary endothelium and choroid plexus (36, 85). Messenger RNA for IL-1 receptors has been demonstrated in some studies but not in others, and the presence of mRNA does not necessarily indicate the existence of functional receptors. In the mouse brain, binding studies with IL-1 indicate very few, if any, receptors in the hypothalamus, especially in the PVN (36, 85). In addition, although lower doses of IL-1 are necessary to activate the HPA axis when administered icv compared to those effective when administered peripherally, the concentrations may be higher than those likely to reach the hypothalamus after intraperitoneal (ip) or subcutaneous (sc) injections. Lastly, the time course of the HPA response to icv IL-1 is quite slow; maximal plasma concentrations of ACTH and corticosterone are reached only after 2 hours, like the response to peripheral injections. Such a slow response seems unlikely to reflect a direct action of IL-1 on receptors within the medial hypothalamus.

Nevertheless, several groups of researchers have reported that IL-1 can stimulate release of CRF from hypothalamic tissue in vitro (87). Although this response is observed at very low doses of IL-1 (10-13 M), the relationship of this in vitro stimulation of CRF release to that observed following peripheral injections of IL-1 is unclear for the reasons mentioned above: the lack of IL-1 receptors in the hypothalamus and the slow response to intracerebral administration of IL-1. It may be that the HPA responses to peripheral and intracerebral IL-1 occur by different mechanisms.

A specific brain uptake system for IL-1 has been proposed, based on the accumulation of radioactively labeled IL-1 in excised brain tissue (3). However, it has not been proven that the IL-1 is taken up into brain tissue, and it is possible that the apparent accumulation occurs because the IL-1 binds to sites in the capillary endothelium and choroid plexus, which are known to be rich in IL-1 binding sites (36, 85). A careful study in cats was unable to detect IL-1 in the cerebrospinal fluid (CSF) of normal cats before or after the peripheral injection of high doses of IL-1 (14), suggesting that IL-1 from the periphery does not readily cross the blood–brain barrier.

It has been suggested that IL-1 from the periphery acts directly on the OVLT. In the OVLT, IL-1 stimulates the synthesis of prostaglandin E2 (PGE2), which in turn may elicit CRF release by stimulating PVN neurons. This hypothesis is consistent with the observations that intravenous IL-1b administration increases hypothalamic concentrations of PGE2 as determined by microdialysis (51), that hypothalamic infusions of PGE2 can stimulate the HPA axis, and that prostaglandin synthesis inhibitors, such as indomethacin, can prevent the IL-1-induced HPA activation. An alternative hypothesis is that IL-1 acts directly on CRF-containing terminals in the ME (57). When injected directly into this region, IL-1 elicits CRF release, as indicated by increases of plasma ACTH, and these effects can be prevented by local administration of IL-1ra (56). Interestingly, this effect appears to be indirect because icv administration of 6-OHDA prevents this response to IL-1, suggesting that noradrenergic neurons in the ME normally regulate the release of CRF. Consistent with this, local (ME) injection of phenoxybenzamine or propranolol can prevent the IL-1-induced elevation of plasma ACTH (57).

Whether or not IL-1 is present in normal brain is controversial. Fontana first showed that glial cells can produce an IL-1-like substance, and showed that it was a potent growth factor for astrocytes. Several studies have suggested the presence of IL-1 in normal brain using immunohistochemical techniques, but the results have varied markedly, especially in the anatomical distributions and intensities described (see ref. 73). Moreover, to date, no study using rigorous biochemical purification and analysis has unequivocally identified IL-1 in normal brain. There is little doubt that IL-1 appears in the brain in pathological circumstances, such as following endotoxin administration (14, 29, 42) or damage such as caused by insertion of cannulae or microdialysis probes (42, 86, 93). Thus it is possible that the reports of IL-1 in normal brain reflect an unrecognized pathology in the subjects studied. Whether or not the IL-1 is present in neurons is also controversial, but, whereas the immunohistochemical studies have suggested that it might be, most believe that IL-1 is found primarily in microglia (32). An attractive hypothesis is that when the brain is infected or lesioned, the blood–brain barrier is breached. This allows invasion of macrophages from the periphery, which then proliferate in the CNS as microglia. The microglia synthesize IL-1, which acts as a potent growth factor for astroglia, causing them to proliferate, sealing off the lesion, and restoring the damaged blood–brain barrier (33). The IL-1 (and possibly other factors produced by the microglia) may also play other roles in the repair mechanisms and also activate the HPA axis. This schema attributes a major role to IL-1 in a classical pathological mechanism for protection of the brain.

The foregoing indicates that there is now substantial evidence for bidirectional communications between the nervous and immune systems. Communication occurs via chemical messengers, just as it does within the nervous and immune systems. Many of the messengers are already familiar as hormones, neurotransmitters, and cytokines, but presently the messages are poorly understood. Certain messengers from the neuroendocrine system appear to facilitate or inhibit the functions of immune cells, but the specificity remains to be elucidated. Cytokines, such as IL-1, are clearly potent activators of the HPA axis, but also exert a variety of other physiological effects. In all likelihood, many other messengers remain to be discovered.

Although our current understanding of the system is primitive, it may be important to distinguish local from systemic effects. Whereas circulating concentrations of catecholamines and steroids are probably adequate to exert physiological effects, and this also appears to be true for cytokines, the role of the peptides is less clear. Their systemic concentrations are very low and are unlikely to be sufficient to modulate immune system function in a general way. However, it is possible that peptides secreted by nerve terminals in the thymus, spleen, and lymphoid tissue may achieve local concentrations sufficient to affect immune cells. Such effects may also be possible locally in tissue at sites of inflammation.

Messengers that can travel more readily and are more stable metabolically may be active systemically, whereas the less stable peptides may be confined to local actions. The chemical nature of the messengers may be suited to their functions. As lipophilic molecules, the glucocorticoids can readily penetrate membrane barriers and affect cells in all bodily tissues, whereas the hydrophilic catecholamines are more labile and their action may be limited to the circulatory systems. Our present knowledge indicates that the glucocorticoids and catecholamines predominantly inhibit immune responses, whereas the peptides are largely facilitatory. When the organism is threatened, the systemic activity of the glucocorticoids to limit immune responses may be important to depress immune activity to prevent undesirable autoimmune actions. By contrast, peptides could facilitate immune responses in small areas close to the site of their release, for example, in an area of inflammation induced by infection or tissue damage. Catecholamines may occupy an intermediate position, existing in sufficient concentrations to have systemic actions but not having broad access to tissues and having relatively short durations of action, except when chronically elevated. Such an arrangement would permit focusing of the activation of immune response in local areas of inflammation, while preventing potentially damaging autoimmune actions that could be triggered by widespread activation.

There is an ongoing and growing literature suggesting abnormalities of immune function in the mentally ill and gnawing suggestions of a viral etiology in both psychoses and affective disorders (see ref. 50). The psychiatric implications of interactions between the brain and the immune system can be posed as two questions: to what extent can brain dysfunction alter immune function, and can immune effects cause psychiatric disease?

The literature contains a variety of reports to indicate that deficits in immunity are associated with psychiatric disease. The most consistent data have been obtained related to depression. Decreased immune function has been reported in depressed patients as measured by mitogenic stimulation and NK cytotoxicity (9, 80). Given that there is a strong association between depression and elevations of plasma cortisol, it is natural to suggest that the immune deficits may be related to the hypercortisolemia, however, a causal relationship has yet to be documented. Depression has also been correlated with a hyperactivity of noradrenergic systems. Thus either cortisol or NE could be responsible for the immunosuppression. The work of Irwin et al. indicating that icv CRF administration in rats decreases immunity (44), coupled with the observations that CSF CRF is elevated in major depression (64), suggests a potential mechanism. To the extent that the animal model is valid, it may be that the immunosuppressive effect of depression is related to the catecholamines rather than the glucocorticoids. Regardless of the mechanism, the chronic elevation of these humors that occurs in depression may add immunosuppression to the patients' other woes.

Evidence for deficits in immune function in other psychiatric diseases has been less consistent. Early studies found evidence of reduced immune function in schizophrenics, but this has not been substantiated in most recent studies (9). It was suggested that the deficits in some of the studies may have been due to neuroleptic medication. A recent analysis suggests increased incidence of schizophrenia following major epidemics of influenza (63), but the mechanism for this effect is unknown.

If immune-stimulated HPA activation provides a regulatory feedback mechanism, the balance between the nervous system and the immune system may be very delicate. Hyperactivity of the HPA axis to immune stimulation would impair immune defense mechanisms, and hypoactivity would run the risk of autoimmune disease. An example of the latter may be arthritis, which is associated with a hyperactive inflammatory and immune response. As discussed above, the susceptibility to arthritis in Lewis rats may be caused by a hyposensitivity of the HPA axis to normal stressful stimuli, so that the antiinflammatory effects of the glucocorticoids are diminished (82). Another example may be chronic fatigue syndrome (CFS), which has been likened to a glucocorticoid insufficiency because of the debilitating fatigue, the feverishness, arthralgias, myalgias, adenopathy, postexertional fatigue, exacerbation of allergic responses, and disturbances of mood and sleep (17). If, as postulated, CFS follows a chronic viral infection, it is possible that the infection may have caused sustained activation of the HPA axis, desensitizing the normal glucocorticoid response. Consistent with this, abnormalities of the HPA axis have been detected in CFS patients (17) (see also Neuroendocrinology of Mood Disorders, The Neurobiology of Treatment-Resistant Mood Disorders and Towards an Understanding of the Genetics of Alzheimer's Disease).

published 2000