Do Corticosteroids Damage the Brain?

Authors


Dr Joe Herbert, Department of Physiology, Development and Neuroscience, and Cambridge Centre for Brain Repair, University of Cambridge, Robinson Way, Cambridge CB2 2PY, UK (e-mail: jh24@cam.ac.uk).

Abstract

Corticosteroids are an essential component of the body's homeostatic system. In common with other such systems, this implies that corticosteroid levels in blood and, more importantly, in the tissues remain within an optimal range. It also implies that this range may vary according to circumstance. Lack of corticosteroids, such as untreated Addison's disease, can be fatal in humans. In this review, we are principally concerned with excess or disturbed patterns of circulating corticosteroids in the longer or shorter term, and the effects they have on the brain.

Corticosteroids: an essential but temporally complex neuroendocrine signal

Corticosteroids are amongst the most labile of all the hormonal systems. Thus, the brain may be exposed to levels that vary greatly across time and between individuals. Values in the saliva of humans, a good indicator of ‘free’ plasma levels and hence those in the tissues, vary by as much as eight-fold during the day (1). This diurnal rhythm tracks, but is not driven by, the daily cycle of activity and rest. Rather, the glucocorticoid cycle is an important signal connecting the central hypothalamic clock mechanism in the suprachiasmatic nucleus, synchronised by daylight, with peripheral ‘clocks’ as the target. This prepares the body for demands and opportunities offered by the solar cycle and can be termed ‘predictive’ homeostasis. Absolute levels are determined by genotype, experience, age and physiological state, but there is always a cyclical signal except under pathological conditions (Table 1). Corticosteroids also track more episodic events. This can be termed ‘reactive’ homeostasis. Perturbations in the internal environments which challenge homeostasis, such as lack of food, water, salt or excessive temperatures, all elicit increased corticosteroid secretion as part of the response to such demands or ‘stressors’ (2, 3). External events, particularly those that threaten the individual (stress), also result in exaggerated glucocorticoid levels, and these may persist in situations of chronically threatening social adversity (4) and become dysfunctional. Added to these ‘group’ effects are those related to individual differences in basal levels or the response to stressors. In both humans and rats, there are marked individual and genotypic differences (5–7), dependent upon both genetic processes and the occurrence of adverse events during foetal or neonatal life (see below). The exposure of the brain to corticosteroids in both the short and longer term rests on all the factors controlling their diurnal, episodic or life-time levels in the blood, and the way these are transmitted to the cells of the brain. As for changes in basal glucocorticoid levels, reactive homeostasis is governed by genetic, ontogenetic and experiential factors. In humans, pathological events confound physiological regulation. The excess glucocorticoids characterising Cushing's syndrome results in increased exposure of the brain to these steroids, but many other chronic conditions are also accompanied by increased glucocorticoids, including some forms of unipolar depression. Severely reduced levels, as in Addison's disease, are treated by corticosteroid replacement, a regime that implies − but does not assume − restoration of relatively normal levels and patterns of corticosteroids over the longer term (8). All these sources of variations in corticosteroids raise the question: what are their implications for brain function? In particular, is brain function imperilled by excess or dysregulated corticosteroids? In this review, we assess these questions by drawing on a wide spectrum of information. We consider the ‘normal’ biology and control of corticosteroid secretion, the phenomenon of stress and the role that corticosteroids play in it, the cellular actions of corticosteroids on neurones, and the consequences that altered corticosteroids have on the brain. We pay special attention to the effects of corticosteroids on the ability of the brain to repair itself, its vulnerability to damage or malfunction, and on functions such as emotion, memory and learning. First, however, we consider some of the known factors that determine corticosteroid levels in the brain because it is crucial to know the relation between brain levels and those in the blood, in which corticosteroids are usually measured.

Table 1.   Adrenal Steroids.
1. Three sets of steroids produced by the adrenal cortex
2. Aldosterone (mineralocorticoid) from zona glomerulosa, cortisol (corticosterone in some rodents; glucocorticoid) from zona fasciculata, dehydroepiandrosterone from zona reticularis (primates)
3. Separate control systems for the three regions. Glomerulosa: angiotensin II (via renal renin), fasciculata: adrenocorticotrophic hormone (pituitary); reticularis: largely unknown
4. Mineralocorticoids and glucocorticoids both bind to the mineralocorticoid receptor (MR) present in target tissues, including the brain. High expression in several areas, including the hippocampus, anterior hypothalamus. Glucocorticoids also bind to the glucocorticoid receptor (GR) expressed much more widely in the brain. Both MR and GR bind as dimers to a range of palindromic glucocorticoid receptor elements (GRE) following activation by adrenal steroids
5. Glucocorticoids at least may also interact with membrane bound receptors in the brain (e.g. GABAA or NMDA receptors); while recent evidence indicates that the classical steroid receptors themselves may mediate such membrane effects
6. Binding to GRE activates or suppresses a wide range of downstream genes

Circadian rhythms in corticosteroids

The circadian rhythm of glucocorticoids is well known. It corresponds to the daily activity–rest cycle, so levels are maximal at the beginning of the dark phase in nocturnal rats, but in the light phase in diurnal humans. Travel across time zones, or alteration of activity pattern (e.g. night-shifts) results in a corresponding, but slow, re-phasing of the hypothalamo-pituitary axis (HPA) axis (Fig. 1a). However, this daily profile overlays a highly irregular underlying secretory pattern (Table 2). Both rats and humans show intermittent troughs in glucocorticoid levels throughout each 24 h, interweaved between the peaks that characterise periods of behavioural and adrenal activity (9). There are variations in both amplitude and frequency of the peaks. The circadian pattern is actually the result of changes in pulse frequency. This implies at least two control mechanisms: one determining pulse height, the other pulse frequency (10). These troughs and peaks have physiological significance. The relative affinities of the mineralocorticoid (MR) and glucocorticoid (GR) receptors (see below) mean that variations in pulse height and frequency can ‘address’ these two sets of receptors differentially. Furthermore, the effect of intercurrent stressors on glucocorticoids results from an interaction between the stressors and the prevailing corticosteroid level. Chronic conditions (both experimental and clinical) may result not only in elevated corticosteroids, but also a loss of the diurnal rhythm, and both may contribute to the resultant disturbance in brain function (11).

Figure 1.

 Glucocorticoids as synchronisers of the circadian transcriptome. (a) Shiftwork acutely disrupts daily cortisol profiles. On day shifts, workers exhibit characteristic salivary cortisol profiles with morning peak. On the first night shift, this rhythm is retained but, with continuing shiftwork, the cycle progressively re-adjusts such that, after 1 week, it is inverted relative to the typical pattern (Hatfield, Hastings and Herbert, unpublished data). Long-term shift work increases prevalence of cardiovascular and metabolic disease. (b) Schematic view of the hierarchy of circadian regulation. The neuronal oscillator of the suprachiasmatic nuclei (SCN) consists of a negative feedback loop pivoted around cyclical expression of clock genes and their cognate proteins. Activation of the hypothalamo-pituitary axis (HPA) by the SCN maintains the daily rhythm of circulating corticosteroids. These have a central role in synchronising secondary, local circadian clocks in peripheral tissue. These clocks in turn direct local circadian transcriptomes which underpin circadian metabolism.

Table 2.   Corticosteroid Secretion.
Corticosteroids are secreted in distinct pulses. There is a rapid fall in plasma levels after each pulse
This pulse generator increases its activity in the morning (in humans) or in the evening (in rats)
The pulse generator is controlled by both genetic and epigenetic factors
Early life exposure to adversity (either physical or social) has persistent affects on later hypothalamo-pituitary axis (HPA) axis function
Pulsatile frequency is increased with age
There are sex differences in HPA activity. Some of these are reproduced in rats by manipulating neonatal testosterone
Gonadal steroids in adulthood also alter HPA activity

Circadian secretion of adrenocorticotrophic hormone (ACTH) from the pituitary is controlled by the suprachiasmatic nuclei (SCN). The molecular generation of rhythmic activity in the SCN is now quite well understood (12) (Fig. 1b). A suite of ‘clock genes’ including three Period (Per) genes and two Cryptochrome (Cry) genes, are activated at the start of circadian day by complexes containing CLOCK and BMAL, basic helix–loop–helix PAS-domain transcription factors that act on E-box DNA regulatory sequences. PER and CRY proteins associate in the cytoplasm, enabling CRY to shuttle PER into the nucleus. At the end of circadian day, the nuclear accumulation of PER::CRY complexes is sufficient to oppose the actions of CLOCK::BMAL and Per and Cry genes are inactivated. Over the course of circadian night, degradation of PER and CRY ultimately releases the feedback inhibition and the cycle can start again. The period of this self-sustaining transcriptional/translational feedback loop is determined by various factors, including rates of protein degradation, so that mutations that facilitate breakdown of PER2 are associated with accelerated circadian periods in rodents (13) and familial sleep disorder in humans (14).

Output from the loop is initiated by clock-controlled genes that also carry E-boxes and are subject to periodic expression and repression by PER::CRY (15). Ultimately, this transcriptional programme directs circadian changes in SCN firing rate and neurosecretion that co-ordinate circadian functions in its efferent targets.

This circadian activity is synchronised to solar time by input from intrinsically photoreceptive retinal ganglion cells that project directly to the SCN. The output from the SCN is less well defined (16). In terms of sleep and wakefulness, there are direct and indirect projections to basal forebrain, preoptic area orexinergic (orexin A and B expressing) neurones in the hypothalamus and the brain stem reticular formation (17). In the context of corticosteroid secretion, there are axonal projections from the SCN to the region of the paraventricular nucleus of the hypothalamus (PVN), and indirect routes via the dorsomedial hypothalamus. Chemical signals from the SCN (such as vasopressin and prokineticin-2) may reach the PVN either directly or indirectly (18–20). Prokineticin-2 is a member of a class of peptides with a range of biological actions, including eating, immune function and reproduction, as well as circadian rhythms. Damage to or other interference with the SCN both disrupts the normal diurnal pattern of ACTH (and hence the corticosteroid rhythm), and interferes with re-phasing of the corticosteroid rhythm by ambient light. Interestingly, the SCN − unlike most other parts of the brain − does not express GR, implying that circadian information from circulating glucocorticoids acts solely as an ‘output’ signal but has no direct reciprocal feedback action on the central clock (21).

It is important to recognise that cells in other parts of the brain (e.g. the hippocampus) express ‘clock’ genes, as do peripheral tissues, such as the liver, in a rhythmic fashion (22). Such local rhythms are the final determinants of circadian behaviour and metabolism (Fig. 1b). In peripheral tissues, the local clocks are so robust that they can continue to oscillate in culture. Circulating corticosteroids play a major role in synchronising these lesser clocks with that in the SCN, and thus to the daily pattern of light/dark activity throughout the body. For example, phasic administration of dexamethasone, a synthetic glucocorticoid (GC), can re-synchronise diurnal gene expression patterns in the liver of intact or SCN-ablated animals (21). The molecular basis of synchronisation of the circadian transcriptome involves multiple pathways − both direct actions on local clock feedback loops (some clock genes have glucocorticoid-response elements; GREs), and indirect effects via local glucocorticoid-dependent transcription factors (Fig. 1). Although not yet defined, it is likely that glucocorticoids have a similar entraining effect on the temporal programmes of gene expression in diverse brain regions that will complement their neural regulation by circadian inputs. It is not difficult to envisage how inappropriate mismatch between neural and endocrine circadian patterns would compromise local brain functions. For example, disruptions of the diurnal corticosteroid rhythm are a feature of several disorders, including Cushing's disease and unipolar depression (see below), whilst Alzheimer's disease can lead to temporal mismatch between sleep/wakefulness and corticosteroid rhythms (23). Temporally inappropriate exposure of the brain to GC may compromise neural functions and exacerbate mental disorder.

Alterations in central neural rhythmic events thus play a part in suboptimal brain function, implying that changes in peripheral corticosteroid patterns may contribute to this malfunction. For example, persistence of either evening hypercortisolaemia or resistance to dexamethasone negative feedback in unipolar depression predicts recurrence or relapse (24, 25). The complex nature of the diurnal cycle of corticosteroids as well as their episodic responses to external events means that assessing corticosteroid function at a single time point gives a highly incomplete picture of their true activity.

Corticosteroid secretion varies between individuals, sexes and with age

The secretion of corticosteroids changes during the life span. A circadian rhythm develops during the first 18 months of life in humans. Prior to this, children of both sexes show rather unpredictable change to minor events such as washing, nappy change and feeding. Following the establishment of a diurnal rhythm, the HPA axis of young children (under 5 years) may be hyporesponsive to stressful challenge (26), but alterations from familiar to unfamiliar social circumstances in this age range can elicit marked increases in circulating cortisol levels (27).

Old people may show increasing levels of cortisol (28), and this is reflected (in some) in increased levels in the CSF (29). There are also marked individual differences in cortisol levels. Some are related to sex (Table 2). Both female rats and humans tend to have higher peak levels of glucocorticoids than males (1, 30). In humans, there is some evidence (still not entirely certain) that this sexual dimorphism may appear or be accentuated at puberty. In rats, prenatal administration of the antiandrogen flutamide both increases basal levels of glucocorticoids in adulthood, and increases the glucocorticoid response to stress, normally greater in females (31). This suggests that, in rats at least, prenatal testosterone, known to influence subsequent development of a number of sex-differentiated endocrine functions and behaviours in males, has parallel actions on the HPA axis (9). Comparable data from humans are still awaited. The sex difference in glucocorticoids may play a part in the relative incidence of some disorders (e.g. depression, Alzheimer's disease − see below).

Studies on twins show that approximately 50% of the variance in cortisol levels in humans is genetically determined, though the genes responsible have not been identified (32, 33). There are several polymorphisms in the human GR and MR that may be responsible for individual differences in neural responses to corticosteroids (34, 35). Some (e.g. the N363S polymorphism in exon 2 of GR, which occurs in approximately 8–10%) may increase the brain's sensitivity to glucocorticoids. Carriers are hypersensitive to glucocorticoids. This polymorphism is not pathological, but may explain the carriers' unfavourable body composition as well as their metabolic and mental status, such as increased body fat and higher risk of unipolar depression. A less frequent second polymorphism (ER22/23EK), in the same exon 2, has the opposite effect. These carriers appear glucocorticoid resistant and have a healthier metabolic profile and lower risk for dementia (35). An intron-2 polymorphism (bcl1) is more common (approximately 20–25%) and is associated with elevated cortisol. A three-marker haplotype has been identified in the GR which predicts sensitivity to dexamethasone and thus, perhaps, to endogenous cortisol feedback and/or tissue responsiveness (36–38). Whether these polymorphisms contribute directly to the development of some of the pathological actions of corticosteroids under adverse circumstances is still unknown, but is clearly a question of interest.

Can we define optimal levels of corticosteroids?

There is a distinction between ‘basal’ and ‘optimal’ levels of glucocorticoids. The definition of ‘basal’ is itself problematical: glucocorticoids are so reactive to internal or external events that basal conditions may be more theoretical than observable. Furthermore, there are several parameters to take into account: maximum and minimum levels, the peak and trough of the daily rhythm, the height and frequency of individual glucocorticoid pulses, variations across days and between individuals, the sensitivity of target tissues to prevailing levels, etc. It is difficult to define the external or internal (homeostatic) conditions that represent a ‘basal’ state, even if this had biological relevance. Defining ‘optimal’ levels is another problem. Patients with unipolar depression can have cortisol levels approaching those in Cushing's disease, yet show few obvious Cushingoid signs and symptoms (osteoporosis may be an exception) (39). Those taking replacement corticosteroids at doses that are adequate under normal conditions need more during episodes of illness (i.e. stress). Older people may have increased levels, although whether or not this is adaptive or damaging is still debated (see below). It is clear that defining optimal values for corticosteroids, particularly glucocorticoids, is a relative, not an absolute, exercise. The most reliable and biologically relevant method would be to devise some measure of tissue response to corticosteroids and ensure that this fell into an optimal range, although this might differ across tissues. So far as the brain is concerned, such a measure does not currently exist.

Sources and entry of corticosteroids into the brain

The major source of corticosteroids in the brain is from the blood, despite the fact that the brain may make small amount of steroids such as pregnenolone and its derivatives (‘neurosteroids’) from cholesterol (40). Because much of what we infer about brain levels comes from measuring corticosteroids in the blood, the dynamics of transfer from blood to brain are important. Glucocorticoids in the blood are bound to corticosteroid-binding globulin (CBG). CBG represents a low-capacity, high-affinity reservoir in the blood. Under some circumstances, blood albumin may act as a higher-capacity, lower-affinity source. Because corticosteroids are highly lipid-soluble (nonpolar), this plasma corticoid-binding protein seems at first glance to be the only obstacle to direct entry into the brain. CBG levels are relatively stable, although they may decline somewhat in old age in humans, but there are marked sex differences (females have higher levels) (1, 41, 42). This is a direct response to oestrogen (30) because levels increase further during pregnancy, when oestrogen levels peak, and are replicated during therapeutic oestrogen treatment. Part, at least, of the sex difference in glucocorticoids may be determined by exposure to testosterone early in life, because perinatal testosterone treatment reduces glucocorticoid levels in adulthood (31, 43) (Table 2). This recalls the ‘organisational’ role of testosterone in other sex-differentiated contexts (e.g. sexual behaviour). Approximately 95% of blood cortisol in humans is bound mostly to CBG, and direct measure of cortisol levels in the CSF confirms that they represent approximately 5% of those in the blood (29). It is important to recognise that, if blood levels of glucocorticoids increase sharply (as in stress, or in pathological situations such as Cushing's disease), then the limited binding capacity of CBG will be exceeded and brain levels will increase disproportionately. This has obvious implications for potentiating any beneficial or damaging effects on the brain.

Although steroids are highly lipid soluble, the blood–brain barrier contains the multi drug resistance P-glycoprotein (mdr-1a) transporters which, in rodents, export exogenous compounds such as cortisol and the synthetic glucocorticoid dexamethasone. Retention of these two steroids is therefore much lower than the endogenous glucocorticoid corticosterone, which is not recognised as substrate for the transporter. However, if cortisol and dexamethasone are administered to mdr-1a knockout mice, both steroids are retained and accumulate at hippocampal sites that express MR and GR abundantly, as is the case with endogenous corticosterone (44, 45). A similar Pgp pump operates at the human blood–brain barrier.

Corticosteroids and stress

Exposure to threatening or demanding events, whether external to the animal (e.g. a dangerous location, an agonistic encounter) or internal (e.g. challenges to homeostasis, such as prolonged food or water deprivation) stimulate the adrenal secretion of glucocorticoids (2). The concept of stress has been much criticised as being too general and imprecise to be useful. However, if stress is regarded as having two components, a generic tone (implying unusual demand that requires action) and a specific one that defines the nature of the required response (both physiological and behavioural), then the term regains its value (46). Glucocorticoids (together with the autonomic nervous system) seem to be mainly associated with the first (generic) component: they signal a situation of unusual demand (47). The second (specific) component will vary according to the nature of the demand, and this is heavily dependent on specific patterns of peptide expression in the brain (46). For example, dehydration will stimulate glucocorticoids, as well as a set of endocrine and cardiovascular activations that compensate for lack of water, and a behavioural one that drives the animal to drink. Food deprivation, also associated with glucocorticoid activation, is followed by a different set of physiological and behavioural responses. Together, the two sets (generic and specific) make up the coping or adaptive response. The exact function of glucocorticoids in stress has been much debated. They seem to feed back on components of the specific response (e.g. inflammation or immune reactions) in such as way as to dampen these down and prevent maladaptive overshoot, and thus encourage a return to homeostasis (48).

The animal's social environment is a potent source of generic demand (e.g. competition for limited resources). Some monkeys at the bottom of the social hierarchy have higher cortisol levels (and flattened diurnal cortisol rhythms) compared to more dominant animals (49), although the nature of the social structure in which they live may moderate this effect (50).

A major distinction needs to be made between generic acute and chronic (persistent or repeated) stress responses to external demands. Acute stress is adaptive and results in a rapid response of ACTH and cortisol to several times normal values and a short-lived increase in corticotrophin-releasing hormone (CRH) and a lesser one in vasopressin (AVP) mRNAs in the hypothalamus. On the other hand, chronic stress, which is maladaptive, results in a decrease in hypothalamic CRH mRNA and portal CRH levels and a compensatory increase in AVP mRNA and portal blood AVP levels (51, 52). Blood levels of glucocorticoids are often lower than in the acute phase (although they may still be elevated). The hypothalamic switch in peptide coding between CRH and AVP during protracted stress is dependent on raised glucocorticoids (53). There is therefore a remarkable glucocorticoid-dependent plasticity of this system which is further demonstrated by changes during lactation. At this time, there is a flattening of the diurnal rhythm, a markedly reduced plasma corticosterone and ACTH response to stress and reduced CRH and preproenkephalin mRNA response to stress (54, 55). Bearing in mind the persistent and deleterious effects of disturbances in maternal behaviour on the offspring (see below), the hyporesponse to stress during lactation seems to be a significant adaptation.

How does the brain regulate its exposure to corticosteroids?

Once corticosteroids enter the brain, they encounter a second set of regulatory factors: the activity of 11β-hydroxysteroid dehydrogenase (11βHSD) (Table 3). This exists in two isoforms. The NADPH-dependent type 1 (11βHSD-1), which is widespread in the adult rodent and human brain as well as in liver and adipose tissue (56, 57), promotes the reduction of 11-keto-derivatives (e.g. cortisone in man, 11-dehydrocorticosterone in rats and mice) to active cortisol or corticosterone (respectively). 11βHSD-1 knockout mice are phenotypically normal, but show reduced tissue actions of endogenous glucocorticoids (e.g. a lower plasma glucose and free fatty acid during fasting or in response to stress) (58, 59). They also show increased basal levels of corticosterone, a reduced circadian nadir, and exaggerated stress responses, which are results of their requirement for increased adrenal glucocorticoid production (to make up for the lack of splanchnic regeneration of corticosterone) and reduced negative feedback on the HPA axis (60). These results indicate that local production of active steroid is a significant factor determining the brain's sensitivity to corticosteroids. A substantial proportion of rats become less able to learn a hippocampal-dependent task (e.g. the water maze) as they get older (a glucocorticoid-dependent effect: see below), but this is not observed in 11βHSD-1 knockout mice, which, despite elevated corticosterone levels have lower intrahippocampal glucocorticoid concentrations (61, 62). These data reinforce the idea that intracellular metabolism is a major influence upon glucocorticoid action within the cells of the brain. This also suggests that reduced activity in 11βHSD-1 may benefit the brain under some circumstances; indeed 11βHSD inhibitors improve cognitive function in elderly men and patients with type 2 diabetes. We discuss the role of corticosteroids in age-related changes in the brain further below.

Table 3.   Control of Corticosteroids by 11β-Hydroxysteroid Dehydrogenase (HSD).
Two isoenzymes: HSD1 and HSD2
 Drive enzyme reaction in opposite directions
HDS1
 Acts as reductase in vivo
 Generates (in rats) corticosterone from inactive 11-keto precursor
 Highly expressed in hippocampus, cerebellum, cortex
 Increases expression following chronic stress
 Knockout mice show no phenotypic abnormalities
 Have increase glucocorticoid levels and hypertrophic adrenals stress responses exaggerated
 Aged brain protected against cognitive decline
HSD2
 Inactivates glucocorticoids by generating ketone product (e.g. cortisone from cortisol)
 Highly expressed in aldosterone-selective tissues (e.g. kidney, colon, etc.)
  Knockout mice (and humans with HSD2 mutations) show apparent mineralocorticoid excess limited distribution in the brain (subcommisural organ, amygdala)
 May protect foetal brain against excess glucocorticoids
 Null mice show cerebellar growth impairment

NAD-dependent type 2 (11βHSD-2) has the converse action and ensures that glucocorticoids entering cells expressing the enzyme are inactivated before they can bind to receptors (Table 3). High amounts are present in the epithelial cells of the kidney, colon, and salivary glands. Inactivation of 11βHSD-2 renders glucocorticoids able to act as mineralocorticoids in the kidney. In the adult brain, 11βHSD-2 is scarcely expressed but is found in a few nuclei and may underpin selective central effects of aldosterone upon salt appetite and blood pressure (63). Certain nonepithelial cells also express MR abundantly but, in this case, 11βHSD-2 is absent and the MR recognises aldosterone in the presence of excess naturally occurring glucocorticoids, because these also display high affinity for this receptor. Prominent examples are the MR in blood vessels and heart. However, neurones in limbic brain structures − the hippocampal pyramidal cell fields and dentate gyrus, the amygdala, lateral septal area and frontal cortical areas − express this apparently ‘nonselective’ MR (64). During development, the high expression of 11βHSD-2 in the placenta and prenatal brain protects these vulnerable structures against the high levels of glucocorticoids typical of gestation (62, 65, 66). Excess glucocorticoids during development have multiple adverse effects. These include malformation of parts of the brain (e.g. the cerebellum) (67), increased activity of the HPA axis in adult life (68), and increased likelihood of corticoid-sensitive disorders such as hypertension, increased anxiety and exaggerated stress responses (see below) (66). Over-expression of 11βHSD-2 in adult hippocampus reduces kainate-induced damage (69).

Corticosteroid receptors and gene expression in the brain

Neurones ultimately respond directly to corticosteroids in one of two ways: either as a result of the steroids binding to cytoplasmic steroid receptors or by interaction with membrane-bound receptors that mediate the actions of neurotransmitters such as GABA or glutamate. The two categories of cytoplasmic receptors, glucocorticoid (GR or type 2) and mineralocorticoid (MR or type 1) have distinctly different distributions in the brain (64). GRs are widely expressed, although the hippocampus has particularly high amounts. MRs are much more restricted: again, the hippocampus contains high concentrations (there are also appreciable amounts in the amygdala and lateral septum), whereas expression is sparse elsewhere (70). The curious fact is that MR (despite their name) have a high affinity for glucocorticoids and mineralocorticoids, whereas GR have about a ten-fold lesser affinity for glucocorticoids. A second curiosity is that both MR and GR, following binding to their ligands, translocate to the nucleus and bind to the same consensus sequence on DNA (usually called the glucocorticoid-receptor element: GRE) (Fig. 2). However, extensive evidence shows that each type of receptor has its own spectrum of action even in the same cell as well as in different tissues and, it seems, in different parts of the brain. One difference depends on corticosteroid concentration: because of their respective binding affinities, MRs will be fully saturated at low concentrations, whereas GRs (in addition to MRs) remain available to higher concentrations such as those occurring during exposure to physiological or psychosocial stressors. A second regulatory factor limiting access to the MR and GR is the enzymatic screen represented by the 11βHSD isoforms described above. This is relevant for epithelial cells in, for example, the kidney, but not for the limbic brain that expresses the type 1 isoform which regenerates the bioactive glucocorticoid.

Figure 2.

 Modulation of corticosteroid action. The naturally occurring glucocorticoids corticosterone and cortisol bind in brain with high affinity to mineralocorticoid receptors (MR) and lower affinity to glucocorticoid receptors (GR) (199). GR, but not MR, interacts as monomer with various transcription factors (TF). MR and GR both bind as dimers to glucocorticoid response elements (GRE) on DNA while interacting with different patterns of coregulators in control of gene expression (200). Recent data have demonstrated that, at the membrane level, the same nuclear receptors mediate fast nongenomic effects of the naturally occurring glucocorticoid corticosterone. This has been shown for the MR which, upon activation, rapidly enhance in hippocampal CA1 neurones the frequency of miniature excitatory postsynaptic potentials (mTIFP) and reduce paired pulse facilitation (201). Such mTIFPs are caused by the enhanced presynaptic release of glutamate. The effects mediated by MR may be targeted directly to the presynaptic terminals or be exerted trans-synaptically involving, e.g. endocannabinoids. Evidence for the latter route was found by Di et al. (202) in PVN neurones, where a GR-like receptor mediates rapid suppressive corticosteroid effects on mTIFP, suggesting a substrate for fast glucocorticoid feedback (202, 203). Other studies point to G-protein coupled receptors as target for the corticosteroids (202). Corticosteroid action can be modulated by factors controlling access to the receptor: these include the role of corticosteroid binding globulin (CBG) in releasing glucocorticoids from the blood compartment, multidrug resistance P glycoprotein (MDR-Pgp) to export exogenous steroids from the blood–brain barrier and 11βHSD-2 for conversion of naturally occurring glucocorticoids within the cell.

The common GRE for both MR and GR has been a puzzle because this, in itself, would not predict their distinct physiological actions. It now seems that the N-terminal part of the two receptors (the least homologous region) may moderate distinct interactions between the receptors and cofactors that together activate or repress downstream genes. Both differences in patterns of response between tissues (and possibly parts of the brain), as well as those between the two receptor types, may be determined by these moderators. They include interactions with other transcription factors such as fos and jun-1, NF-κB via Ikk-alpha, and the family of STAT proteins (71). Such interactions may be regionally specific within the brain and by protein–protein interactions that operate by essentially a nongenomic pathway (72). Whether variations in these and other cofactors − either between individuals or in response to extraneous events − alter these interactions and have implications for brain function remain subjects for future work, although there is evidence that splice variants of p160 steroid receptor cofactors can alter sensitivity to glucocorticoids (73).

There are also nongenomic mechanisms for the actions of corticosteroids, and these may play a part in brain malfunction. A number of steroids, including corticosterone but also neurosteroids, are known to interact allosterically with membrane-bound receptors, and rapidly affect cell properties, including activities via GABA and NMDA-type glutamate receptors and L-type calcium channels (74). Excess glutamate, and overactive Ca2+ are known to damage neurones, and may be one way that corticosteroids enhance this damage, either directly or indirectly through regulating glutamate uptake by astrocytes (75). NMDA receptor blockers prevent the dendritic atrophy induced in CA3 by excess glucocorticoids (76; see below).

Early life experience regulates corticosteroids during adulthood

The effects of corticosteroids on brain structure and/or function may vary across the life span. Current evidence suggests considerable interplay over time between cortisol and cognitive functions with both bottom-up (glucocorticoids effects on cognitive function), and top-down (effects of cognitive processing on glucocorticoid secretion) effects in the human population (77). It is also increasingly clear that the social environment influences this cortisol–memory relationship. There is evidence that the quality of interpersonal experiences during infancy and childhood can have longstanding consequences during later life, perhaps through programming the sensitivity of the HPA axis (78). In addition, interplay between cortisol and cognitive functions also have implications for psychiatric disorders. For example, patients with severe unipolar depression show increased sensitivity of declarative memory to glucocorticoids (79).

Experimental evidence also shows that exposure of the brain to increased amounts of corticosteroids early in life can have long-lasting and dramatic consequences on later corticosteroid function as well as on other glucocorticoid-sensitive processes such as metabolism and cardiovascular function. Experiencing adverse events in early life (which may raise corticosteroid levels) may have similar results. Rats have altricial (immature) young, and interruption of normal maternal care (e.g. by periods of separation) results in altered corticosteroid response to stress in later adult life, as well as, in some cases, higher basal levels than in corresponding controls (80–83). Glucocorticoid-sensitive neural functions, such as those associated with the hippocampus (see below), may also be impaired (81). There are persistent behavioural consequences of periods of maternal deprivation experienced by the pup (84). The length of the separation is a crucial determinant. A daily brief separation − a procedure called handling − produces in later life animals that have reduced corticosterone responses to a mild psychological challenge (85, 86). By contrast, prolonged separation (e.g. 24-h deprivation of maternal care) has the opposite result and these offspring show enhanced stress responsiveness at adult life (84).

Similar results are obtained if young rats experience an immunological challenge (e.g. lipopolysaccharide) (87). The brains of such rats are therefore exposed to different (higher) amounts of corticosteroids during adulthood as a direct result of early life experience. Parallel evidence exists in humans (78). It appears that, in both species, there is a ‘critical’ neonatal period in which disruption of maternal care (and possibly the occurrence of other stressors) can re-set or ‘programme’ the HPA axis (88, 89). These early adverse events may have the potential for long-lasting and detrimental actions on subsequent brain function because exposure of the brain to corticosteroids is enhanced. The underlying mechanism for this early ‘programming’ of the HPA axis is still not very clear, although a site such as the hypothalamic paraventricular nucleus is a prime candidate. One way in which early experience can modulate the HPA system is through epigenetic modulation of GR by methylation of DNA or methylation and/or acetylation of DNA-associated histones. In general, DNA methylation silences genes, whereas histone-acetylation facilitates activation of transcription. Lower levels of maternal care in rats are associated with increased methylation of the consensus sequence for NGFI-A (a transcription factor) on the brain-specific exon 17 promoter of GR (90), and this results in persistently decreased GR expression in the brain (91).

Similar ‘programming’ of the HPA axis in the young can follow adversity experienced during pregnancy (92). This shows that maternal stress can also be transmitted to the young in a way that has long-lasting effects (93). These include increased liability of developing hypertension, and altered metabolism (e.g. insulin resistance) that may have secondary actions on the brain. There are also behavioural consequences of prenatal stress (84, 94–97). Offspring whose mothers were stressed may themselves show increased levels of anxiety during adulthood as well as infancy, and this is not prevented by cross-fostering, suggesting a direct effect on the foetus (98). Anxiety is an essential ingredient of an adaptive response to a threatening situation. However, excess or inappropriate anxiety may lead to a highly maladaptive state. The sensitive period for the programming actions of adversity thus begins in utero (in rats) and extends into the neonatal period (99). Lowered levels of cortisol have been observed in the babies of pregnant women exposed to a traumatic experience and who developed post-traumatic stress disorder (PTSD: see below) compared to those who did not (100). This, together with the evidence that postnatal adversity modulates later cortisol levels (78) suggests that similar sensitive periods may exist in humans.

In humans, early environmental influences may interact and moderate the impact of later ones. It has been found that children from lower socio-economic strata had elevated cortisol levels compared to those from higher ones (although the nature and direction of this association is uncertain) (101, 102). Particular aspects of the social environment may play a role in determining individual cortisol levels, as may certain internal characteristics such as shyness, which contributes to elevated cortisol levels in the face of uncertain and novel circumstances (27, 103). Neither psychological factors (such as higher emotionality and lower self-esteem), nor intercurrent social adversities were related to individual cortisol levels in a community study of adolescents or adults (104, 105). This suggests that specific aspects of current life style may be a contributing factor to relatively stable individual differences in corticosteroids (78).

However, others have suggested that individual differences in self-esteem are an important predictor of reactivity to stress in humans. Self-esteem can be described as the generic value we place on ourselves, and previous studies have documented effects of self-esteem on health, life expectancy, and life satisfaction across the life span. Self-esteem has been reported to be a potent predictor of the cortisol response to stress, so that people with lower self-esteem do not habituate their stress response upon repeated exposure to a stressor compared to those with higher self-esteem (106). In two more recent studies in young and older adults, self-esteem has been found to be associated with both basal levels of cortisol and hippocampal volume (107, 108). Self-esteem is highly correlated with depressive symptoms in most studies and it remains unclear if the associations with cortisol are concomitants of current mental distress rather than reflecting a core construct of self-percept with physiological consequences. Disaggregating the notion of self-esteem is leading to a more precise understanding of the relations between self-psychology and individual differences in cortisol. This is particularly apparent in studies of memory (formation, consolidation and recall), self percept and clinical depression (109, 110).

How corticosteroids change with age

The HPA axis changes with age. Basal levels of corticosterone are found to decrease in some strains of aged rats, which also tend to lose their circadian rhythm; in others, levels increase (26). Stress responses may become exaggerated, although this is not a consistent finding. Humans show a similar trajectory. Ageing humans tend to show increased cortisol levels, particularly during the evening, although morning levels may decrease. The diurnal rhythm may not only be flattened, but also phase-advanced (111, 112). However, this age-related change is highly variable, with some even showing decreased cortisol levels, and such individuality may be highly significant (28, 113). This picture is consistent with animal studies indicating that, in the rat, increased glucocorticoid levels are not a necessary consequence of ageing, but are significantly more prevalent in those selected for spatial memory deficits than those that remain cognitively unimpaired (114). Longitudinal studies have shown that approximately 38% of older adults present a significant increase of cortisol levels over the years, 19% decreased their cortisol over time, and 43% had a moderate increase with adequate levels of cortisol (28).

The human hippocampus may be vulnerable to increased glucocorticoid action in aged people, resulting in attendant deficits in the cognitive function associated with this part of the brain. There is a negative correlation in both rats and humans between plasma glucocorticoid levels and hippocampal volume (86). Verbal recall memory is relatively impaired in those with higher levels of cortisol, and this may relate to the smaller hippocampal volume characteristic of this group (77). Administering additional cortisol impaired memory further in those with moderate (but not higher) cortisol levels, further reinforcing the view that increased glucocorticoid is directly responsible (115). This is directly reflected in their ability to learn or recall spatial information, a property long associated with an intact hippocampus. In both rats and humans, hippocampal volume and spatial or episodic learning ability decreases with age, although with considerable individual variability. In both, there is a negative relation between corticosteroid levels in older age and hippocampal structure [functional magnetic resonance imaging (fMRI) volume] and function (spatial learning). Adrenalectomy followed by low amounts of corticosterone prevents some of these age-related changes in the hippocampus in rats (86). It seems clear therefore that excess corticosteroids contribute to this age-related decline, whereas low amounts of the steroid sufficient to occupy MR are protective. Whether excess glucocorticoid is simply a matter of increasing levels with age or flattening the diurnal rhythm (as a proximate factor) or whether the duration of exposure to corticosteroids over the life span is primarily responsible for this decline is still not clear.

In humans, at least, other factors that moderate the actions of glucocorticoids may also change with age, thus potentiating corticosteroid action even in the absence of obviously altered levels. Dehydroepiandrosterone (DHEA) is present in very high concentrations in human blood, but declines markedly with age (although this is also individually variable) (116). DHEA can protect the hippocampus against the neurotoxic actions of glutamate analogues (e.g. NMDA) (117). It can also moderate glucocorticoid actions on peripheral tissues as well as on the brain (118). This suggests that complete consideration of the steroidal environment of the ageing brain in man and its implications for declining neural function need to take into account not only corticosterone levels and secretory patterns, but also those of other factors, such as DHEA, that together may determine how this environment encourages or minimises age-related decline in central neural structure and function.

Corticosteroids and the hippocampus

The hippocampus, with its high concentrations of both GR and MR, seems particularly susceptible to the controlling or even damaging effects of corticosteroids (Fig. 3). It is interesting that the nature of these effects differs in various parts of the hippocampus. The pyramidal cells of CA1 are highly sensitive to the neurotoxic actions of agents such as glutamate or anoxia (119). Excess glucocorticoid increases this susceptibility, probably by increasing Ca2+ influx (119, 120). This action, termed ‘neuroendangerment’, implies that although higher corticosteroid levels may not in themselves damage cells, their vulnerability to other, adventitious, agents (e.g. anoxia or neurotoxins) may be accentuated. Introducing a dominant-negative GR into the rat hippocampus protects it against kainate-induced damage (69). In CA3, the effects seem somewhat different. Here, excess glucocorticoid for a prolonged period of time induces atrophy of the apical dendrites of the pyramidal cells, thus potentially impeding their access to incoming information (121, 122). Curiously, CA2 seems to be resistant to glucocorticoids: at least, there appear to be no reports of the action of either toxins or anoxia being accentuated by corticosteroids in this region. These regional differences in response to corticosteroids within Ammon's horn are difficult to reconcile with the distribution of MR or GR. MR is plentiful throughout, but the amount of GR in CA3 is markedly less than in the other areas, although the afferent neurones express abundant GR (123, 124). This apparent paradox needs further exploration. Interaction between the two sets of receptors may provide a possible avenue of explanation.

Figure 3.

 The sites of glucocorticoid action in the adult hippocampus. The diagram shows the basic circuitry of the hippocampus. Input from various sources (the entorhinal cortex, together with aminergic fibres from the brain stem and basal forebrain) enters mainly through the dentate gyrus. This consists of small granular cells. The progenitor cells from which neurogenesis derives lie on the innermost surface. As they proliferate and then mature, these new neurones migrate outwards. Axons from the dentate granule cells project to the pyramidal cells of CA3, which in turn project to CA1 via the Schaffer collaterals. Glucocorticoids have distinct actions of the dentate granule cells (inhibiting progenitor proliferation), and the pyramidal cells of CA3 (encouraging dendritic atrophy) and CA1 (accentuating neurotoxic actions). The output from the hippocampus comes mainly from CA1.

The most striking actions of corticosteroids occur in the dentate gyrus. This region, together with the periventricular area, is the only confirmed source of new neurones in the adult brain in all mammalian species so far examined (125, 126). Approximately 80% of the progeny of dividing progenitor cells in the inner layer of the gyrus have a neuronal phenotype, and are incorporated into the adult connections with CA3. Despite attempts to show that neurogenesis is needed for certain forms of hippocampal-dependent learning (127, 128), the precise significance of this extraordinary phenomenon is still debated. Adrenalectomy increases the proliferation rate of these cells, and excess glucocorticoid decreases it sharply, as do psychosocial or physical stressors (129, 130). As well as inhibiting proliferation, excess glucocorticoid reduces the survival of immature neurones in the dentate gyrus, and inhibits the expression of the mature phenotype in those that do survive (131). It seems clear that glucocorticoids exert a major controlling influence on the progenerative process in the granule cell layer of the dentate gyrus. Whether mineralocorticoids also play a role is less certain, although there are indications that they do. The downstream actions of glucocorticoids in the dentate responsible for these actions are still mysterious. Although decreased proliferation is prevented by GR (or MR) blockade, pointing to a ‘classical’ route via conventional cytoplasmic macromolecular receptors (132), NMDA-type glutamate receptors also alter proliferation rates, but their interaction with glucocorticoids is still unclear (76). Current work focuses on glutamate receptors, growth factors (such as brain-derived neurotrophic factor) or nitric oxide formation, all of which can also regulate progenitor cell proliferation in the dentate gyrus (125); all are also regulated by corticosteroids. Until we really understand the functional significance of neurogenesis in the adult hippocampus, we cannot fully assess the significance of the potent role of corticosteroids in the adult dentate gyrus.

Neurogenesis and depression

There has been a recent surge of interest in the possibility that altered neurogenesis underlies either unipolar depression or the therapeutic response to antidepressants such as selective serotonin reuptake inhibitors (SSRIs) (133). This may well indicate another way that corticosteroids can adversely affect brain function. SSRIs such as fluoxetine increase proliferation rates in the dentate gyrus. The anxiolytic actions of fluoxetine are not observed in 5-HT1A receptor knockout mice, and neurogenesis is not increased (134). Because the maturation and connection of new neurones takes approximately 21 days, this also offers an explanation for the long-recognised paradox that SSRIs take around this time to be therapeutically effective, despite the immediacy of their biochemical actions on the brain. However, it must be acknowledged that the evidence available to date is based on correlations, and a direct causal relationship between neurogenesis and depression has yet to be established. Excess glucocorticoids are reported to be one of the risk factors for subsequent unipolar depression (see below). Interactions in the brain between glucocorticoids and serotonin are prominent and well known (135, 136): glucocorticoids regulate tryptophan hydoxylase, and the expression of several 5-HT receptors, including 5-HT1A and 5-HT2C. Similar interactions are being found in the control of hippocampal neurogenesis. Serotonin may regulate the sensitivity of the progenitor cells in the dentate gyrus to glucocorticoid (137). There may also be converse interactions. Approximately 50% of adults with depression have flattened circadian rhythms in cortisol, with evening levels being elevated (138, 139). Replication of this in rats prevents fluoxetine from stimulating progenitor cell proliferation, which suggests one reason why some patients fail to respond to these drugs (140). However, the role of the hippocampus in depression is still somewhat speculative.

Corticosteroids and memory: the amygdala as target

An essential part of the generic response to an external threat, whether from a conspecific, a predator, or a physical event, is an appropriate behavioural reaction. The immediate response is one of appraisal that facilitates coping with the threat (e.g. ‘fight’ or ‘flight’), but it is equally important to form, encode and consolidate memories about the nature and context of the experience itself, thus enabling the individual to reduce the chances of it happening again (through avoidance strategies), or to cope with it better next time. However, this behavioural response becomes maladaptive when either it is suboptimal in the context in which it occurs, or the experience itself leads to inappropriate or disabling behaviours in other contexts. Although the exact neural mechanisms responsible may be obscure, this points to brain dysfunction and thus brain damage.

Encoding experiences, good or bad, is most effective when an emotional component is evoked. Thus, memory formation for threatening events occurs most effectively in the context of high negative emotion (e.g. fear, disgust). The amygdala is well-known to be closely associated with the generation of fear (and other emotions) and the ability to respond to fear in others (141, 142). Learning about associations between environmental events and their consequences such as danger or punishment (i.e. conditioning) is impeded by damage to the amygdala (143). Increases in systemic glucocorticoids at the point of exposure enhance such learning, and this is prevented by lesions of the basolateral amygdala (144). Direct infusions of glucocorticoids or specific GR agonists into the basolateral amygdala replicate their systemic action (145) (Fig. 4). As in rats, corticosteroids in humans may enhance memory formation for emotionally arousing events (146). It seems likely that this is part of an adaptive (generic) emotional–cognitive reaction to situations that may be stressful or even dangerous.

Figure 4.

 Schematic summarising the interactions of glucocorticoids with the noradrenergic system of the basolateral amygdala. Interactions occur at both presynaptic and postsynaptic sites. Noradrenaline (NA) is released following training in aversively motivated tasks and binds to both β-adrenoceptors and α1-adrenoceptors at postsynaptic sites. The β-adrenoceptor is coupled directly to adenylate cyclase to stimulate cAMP formation. The α1-adrenoceptor modulates the response induced by β-adrenoceptor stimulation. Glucocorticoids may influence the β-adrenoceptor-cAMP system via a coupling with α1-adrenoceptors. In addition, glucocorticoids may activate the noradrenergic system by activation of GRs in brainstem noradrenergic cell groups. α1, α1-adrenoceptor; β, β-adrenoceptor (types 1 and 2); cAMP, adenosine 3′,5′-cyclic monophosphate; GR, glucocorticoid receptor; LC, locus coeruleus; NTS, nucleus of the solitary tract. Reproduced from (204) with permission from Elsevier.

These mnemonic effects of corticosterone, however, seem to depend on interactions with noradrenaline (norepinephrine) within the basolateral amygdala (147). Interactions between these two sets of hormones have been known in the periphery for decades. For example, glucocorticoids enhance the release of noradrenaline in the spleen, lungs and other peripheral organs, and a rapid action of glucocorticoids facilitates the actions of noradrenaline on fight and flight behaviour. Systemic noradrenaline improves memory in a wide variety of emotionally arousing learning tasks (143, 148). This effect, which relays via the vagal afferents to the solitary nucleus, depends on consequent noradrenaline release in the amygdala and is itself enhanced by glucocorticoids (149). Such critical interactions between glucocorticoids and arousal-activated noradrenergic mechanisms within the basolateral amygdala may explain why glucocorticoids selectively enhance memory for emotionally arousing, and not emotionally neutral, information (150) (Fig. 4). In this context, it seems clear that glucocorticoids have an important and positive role to play in coping with adverse experience. They allow an enhanced ‘emotionally relevant’ stamp on events. Together with the hippocampus, which is concerned with encoding the events themselves (see below), glucocorticoids are concerned with complementary adaptive modifications of mnemonic functions at these two sites.

Corticosteroids and memory: the hippocampus as target

Glucocorticoids administered into the hippocampus after training also enhance memory consolidation, particularly of tasks with a strong contextual or spatial component. Adrenalectomy, as well as excess corticosteroids, impairs learned avoidance behaviour in rats, a deficit restored by physiological doses of corticosterone. Adrenalectomy also impairs consolidation as does the knockout of the GR (151). In the latter studies, the performance of mutants with a point mutation preventing the GR from dimerisation suggests that gene transcription needs to be activated for consolidation of memory. There are also marked effects of glucocorticoids on hippocampal long-term potentiation (LTP), generally regarded as an electrophysiological ‘model’ of memory. Basal levels are needed for effective LTP, but higher levels impair it. This U-shaped result may reflect the relative occupancy of MR (at lower doses) and GR (at higher ones).

Moreover, in both young and ageing humans, recent and delayed declarative memories are impaired by excess exogenous corticosteroids (115, 152) or if normal endogenous levels are suppressed by metyrapone. The effect of the latter was reversed by supplementary cortisol, reinforcing the conclusion that reduced corticosteroid levels were responsible for the memory deficit. Suppression of higher cortisol levels in the elderly by metyrapone also improves memory (115). This U-shaped result recalls the experimental data reviewed above, and indicates that moderate levels of corticosteroids are required for optimal cognitive function in humans.

Many studies performed in rodents have reported that the ratio of MR/GR occupation is a unipolar determinant of the direction of glucocorticoid-induced cognitive changes (153). For example, LTP has been shown to be optimal when glucocorticoid levels are mildly elevated (i.e. when the ratio of MR/GR occupation is high) (154). By contrast, significant decreases in LTP are observed after adrenalectomy, when MR occupancy is very low (155) or after exogenous administration of synthetic GC (156), which activates GRs and depletes cortisol, again resulting in low occupancy of MRs. De Kloet et al. (153) have re-interpreted the well-known inverted-U-shape function between circulating levels of glucocorticoids and cognitive performance in line with the MR/GR ratio hypothesis. Cognitive function can be enhanced when most MRs and only part of the GRs are activated (Fig. 5: top of the inverted-U-shape function; increased MR/GR ratio). However, when circulating levels of corticosteroids are significantly decreased or increased (extremes of the inverted-U-shape function; low MR/GR ratio), cognitive impairments will result. This model suggests that optimal corticosteroid levels are a requirement for efficient cognitive function (157, 158).

Figure 5.

 A model of the modulatory action of glucocorticoids on human cognitive function. The varying ratio of occupancy of mineralocorticoid receptors (MRs) and glucocorticoid receptors (GRs) at different circulating levels of glucocorticoids are depicted. The lines within each box represent the relative level of occupancy of each receptor type as glucocorticoid concentration increases.

Lesions of the amygdala are known to also block the effects of glucocorticoids on hippocampal LTP and memory (144, 159), further supporting the view of strong functional interactions between both brain regions in regulating the formation of new memories (see above). Systemically administered glucocorticoids or infusions of glucocorticoids into the hippocampus not only enhance the consolidation of new information, but also impair the retrieval of previously acquired adaptive behaviour (e.g. avoidance of an electric shock), and this effect also seems to require an intact basolateral amygdala or noradrenergic activity within the basolateral amygdala (160, 161). Although this might appear to act in opposition to the more direct actions in the amygdala, it may be that reducing recall can actually improve the consolidation of more recent memories (162). Moreover, as was pointed out previously, MR mediates effects on retrieval as was apparent from the administration of MR antagonists (163). These effects of MR seem to be relatively fast. Such interactions between glucocorticoid effects on memory retrieval and amygdala noradrenergic activity fit well with recent findings suggesting that glucocorticoids also appear to selectively impair the retrieval of emotionally arousing information (164). Working memory, attributable to the medial prefrontal cortex, is also impaired by high levels of glucocorticoids (161, 165). Again, it may be that this is part of an overall pattern of ‘focusing’ memory, ensuring that working memory is continually updated during extreme arousal following any stressor. There are plentiful interconnections between medial prefrontal cortex and amygdala (166) and evidence for functional interaction from fMRI studies in humans (167). As for the hippocampus, the effects of glucocorticoids on prefrontal cortical function require an intact basolateral amygdala and a functioning noradrenergic system (161).

Giving humans additional cortisol impairs declarative (but not nondeclarative) memory (168). In older humans, those with higher levels of cortisol, or in whom levels have increased over the past 4–7 years, show greater deficits in declarative memory tasks (113). Imaging studies have implicated the human hippocampus in the process of encoding novel or associative memories (169). One of the problems in relating glucocorticoid levels to hippocampal function in man is that the primate hippocampus seems to express less GR than in rodents (170). In young humans, reducing cortisol by giving them metyrapone impairs memory, and this is restored by replacing the cortisol deficit (152, 171). This shows that in humans as well as in rodents, there is a U-shaped relation between current glucocorticoid levels and hippocampal-related performance.

Clinical evidence for the damaging actions of excess corticosteroids

Although a coherent explanation of some of these mnemonic actions of glucocorticoids can be made in terms of adaptive responses, this may not necessarily always be the case. The biological outcome of glucocorticoid-dependent (i.e. stress-related) altered memory is not often studied in the laboratory, but clinical evidence suggests that traumatic memories can be maladaptive. PTSD in humans is the best-known example. This set of symptoms follows usually from a single, traumatic, life-threatening event, but generally in those with prior higher levels of social adversity or a past psychiatric history of clinical depression and/or anxiety. These symptoms include the inappropriate triggering of an acute emotional response by stimuli associated with the original traumatic event but now occurring in a benign context (‘flashbacks’), a malignant form of memory. There is practically no information on the role of glucocorticoids in the likelihood of PTSD occurring, which may affect up to 20% of people exposed to severe circumstances, sometimes for decades. Experimental evidence points to a possibly critical role for corticosteroids because cortisol is often elevated before, as well as during, the circumstances that lead to traumatic experiences. However, there are more data on levels of corticoids in PTSD. Paradoxically, perhaps, sufferers have lower levels of cortisol than controls if they are not depressed during the subsequent months or years of the poststress period (172, 173). This has suggested that giving glucocorticoids might be an effective therapy (174, 175).

The hippocampus, implicated in episodic memory, may also play a part. Those with PTSD have a smaller hippocampus than controls, as do the nontraumatised twins of those with PTSD (176). This suggests that a small hippocampus might be a risk factor or represent a ‘latent’ vulnerability for developing PTSD after traumatic experiences. Whether the well-known susceptibility of the hippocampus to corticosteroids either during development or in adult life is the responsible mechanism is not yet known. There are clear-cut structural and functional changes in the hippocampus and related areas within the developing brain which as yet are not related to differences in corticosteroids in cross-sectional case–control studies (177).

The hippocampus may not be the only significant neural substrate in PTSD. In view of the role of the amygdala in general, and its noradrenergic innervation in particular, in memory, it is interesting that noradrenergic β-receptor blockers are finding a place in its treatment and prevention (178). This approach is based on experimental data and human studies showing that blockade of central β-adrenergic receptors inhibits consolidation of emotionally arousing material (179, 180), whereas stimulation of the noradrenergic system enhances it (181).

Direct clinical evidence shows that excess corticosteroids can damage the brain, potentiate existing damage or increase the risk of cerebral malfunction. As already mentioned, Cushing's disease is characterised by over-secretion of glucocorticoid, either as a consequence of an adrenal tumour or in response to increased ACTH production (182, 183) (Table 4). Polymorphisms of the alpha-form of GR also cause relatively mild hypercortisolaemia. A high proportion of those with Cushing's disease also have unipolar depression and psychotic episodes which rapidly resolve once the excess cortisol is remedied. They may also show cognitive impairment (e.g. declarative memory). Cushing's disease is associated with reduced hippocampal volume (and associated impaired eye-blink trace conditioning), and this is partially reversed after successful treatment (lowered cortisol levels) (184, 185). In childhood cases, however, a decline in cognitive performance can be found 1 year after reversal of cortisol levels and brain atrophy (186), suggesting an unresolved brain-based deficit related in part to prior cortisol hypersecretion and resultant tissue atrophy.

Table 4.   Clinical Disease.
Cushings
 Long-term exposure to excessive glucocorticoids
 Two-thirds due to adrenocorticotrophic hormone -secreting pituitary adenoma. Others include ectopic tumours or adrenal neoplasms
 Common symptoms include obesity (abdominal striae), muscle weakness
 Hypertension and hyperglycaemia are common attributes. Decreased immune competence common
 Associated with psychiatric disease, usually an agitated depression in approximately two-thirds of patients. Half of these are severe
 There is also cognitive impairment associated with cortisol excess
 These mental changes may take many months or even years to reverse after effective therapy. Cognitive impairment least likely to recover
Addisons
 Primary failure of the adrenal gland predominantly from autoimmune atrophy in Europe and North America and from tuberculosis in the developing world
 Magnetic resonance imaging shows structural brain changes, mostly in the amygdala and hippocampus, which may also reverse with treatment
 There is both glucocorticoid and mineralocorticoid deficiency
 Common symptoms include weakness, malaise, weight loss, dehydration with associated low blood pressure

Meta-analyses show that patients treated with synthetic glucocorticoids (e.g. prednisolone) tend to develop mood disorders (i.e. brain dysfunction), notably depression, and this is more likely with longer treatment and higher doses (187). Patients suffering from the psychotic subtype of unipolar depression also have elevated levels of cortisol (138). Preliminary evidence suggests that − as is the case in Cushing's disease − psychotic episodes can be ameliorated with antiglucocorticoid therapy (188). The earliest signs in Alzheimer's disease occur in the hippocampus and related areas (189), suggesting that antiglucocorticoid therapy might be useful. Alzheimer's disease is characterised by loss of normal day/night activity rhythms, but this is not associated with equivalent alterations in daily cortisol rhythms (23). This suggests that there is separable control of the different elements of the diurnal ‘clock’. Elevated cortisol levels are common after a stroke, and are associated with a poor outcome, although this needs to take into account the initial severity of the lesion (190). Animal studies suggest that the seizures and indices for tissue damage following stroke can be limited if glucocorticoid levels are lowered by metyrapone or their actions blocked by inhibiting GR. In both cases, it seems that enhancing MR over GR-mediated actions has beneficial effects (191). A recent study reported that treating patients following a head injury with prednisolone significantly worsened the death and disability rate (192). This may be attributable to a direct effect on neuronal survival or be an indirect one (e.g. glucocorticoids elevate blood glucose, which is known to be unfavourable following brain injury).

Individual differences in cortisol may become important in stress responses under certain adverse circumstances. For example, in adolescents of both sexes and in adult women, those with higher morning levels of cortisol are at greater risk of depression following a severe adverse life event (usually a severe and often permanent breakdown in a confiding relationship) (104, 105). (Fig. 6). In this context, it may be that relatively higher (periodic) cortisol predisposes to brain dysfunction after such an acute event, but may be benign in its absence. Whether cortisol enhances the risk for depression that arises more slowly from chronic life experiences remains unclear. The moderating influence of DHEA on glucocorticoids has already been mentioned. The molar cortisol/DHEA ratio predicts persistence longer than 18 months and nonrecovery from a first episode of depressive in adolescents (the median time to recovery is approximately 9 months) (193, 194). There is evidence that treatment with DHEA can be useful is some types of depression (195), a finding that might possibly be related to its ability to stimulate proliferation of progenitor cells in the dentate gyrus (196).

Figure 6.

 Risk processes for the development of unipolar major depression during the first two decades of life. The left side shows the processes involved in the determination of degrees of vulnerability to later mental illness. These derive from genetic polymorphisms (including those in glucocorticoid and serotonin-related genes) in combination with adverse early experience (including suboptimal parenting). These give rise to changes in the hypothalamo-pituitary axis (HPA) axis and to increased sensitivity (both endocrine and psychological) to adversity during childhood. The result is impaired self-appraisal and poor coping skills. This is generic vulnerability. The right side shows the how subsequent events can interact with preceding vulnerability to result in the onset of major unipolar depression. There are contributions of endocrine (HPA), cognitive (ruminations) affective (worry and dysphoria) and neural (serotonin) components, which together lead to a maladaptive reaction and the onset of depression.

It is important to point out that there are mutifactorial risks for unipolar depression that are themselves known to be heterogeneous in clinical characteristics and outcome, and cortisol represents only one of them (Fig. 6). Nevertheless, the predictive role of higher cortisol in the subsequent onset of depression and its association with persistent disorder in both community ascertained and clinical cases (104, 197) recalls other instances of ‘endangerment’ of the brain by corticosteroids. The concept of endangerment may apply to pathological dysfunction as much as to frank neurodegeneration. The increased risk for subsequent depression in those with higher cortisol levels prior to onset is not linear. Overall, the probability of disorder increases in a markedly nonlinear manner once cortisol levels in well adolescents have increased beyond the 80th centile expected for age and sex (Fig. 6) (41). For example, the increased risk of unipolar depression in females may be due to more sensitive brain-based responses to the effects of cortisol at a time of higher levels of accompanying sex hormones, which themselves may mediate the risk for depression in a nonlinear manner (198).

Corticosteroids and brain damage: a point of view

It would be foolish to conclude that corticosteroids per se damage the brain. It is clear that these steroids are required for normal brain function, and for optimal responses to demanding situations. Nevertheless, the fact that the secretion of glucocorticoids in particular are heightened by such demands raises the question of whether there is a ‘cost’ even for adequate adaptation, and whether those even with higher basal levels are also at risk for some category of brain malfunction. Defining corticosteroid-related damage to the brain is not always straightforward. It may seem evident that increased neuronal death, whether precipitated directly by corticosteroids or more indirectly by their potentiating other damaging agents, is disadvantageous. Decreased cognitive abilities may sometimes be directly related to correspondingly diminished volumes in relevant parts of the brain (e.g. the hippocampus). In other cases, the underlying neural processes responsible for corticosteroid-dependent changes in learning or mood are not yet known and have to be inferred. Furthermore, it is not always obvious whether some of these functional effects are adaptive (i.e. advantageous). Nevertheless, the literature is clear: across this wide range of data, excess corticosteroids can damage the brain or impede its function either directly or by increasing its susceptibility to other, coincident, damaging agents. Excess corticoids can also impair both cognitive and affective function, and may contribute to the decline of cognitive ability with age. The major problem is to define in whom, and under what circumstances, corticosteroid levels should be labelled as excessive (or even insufficient), and to estimate the likelihood of a particular adverse outcome. Once this is done, the evident power of these steroids on brain structure and function suggests that major preventative and therapeutic avenues lie ahead.

Acknowledgements

This review is based on a meeting held in Cambridge, UK, on 7 April 2005 to mark the retirement of J.H. It was supported by generous donations from the British Neuroendocrine Society; Gonville and Caius College, Cambridge; the Departments of Anatomy and Psychiatry, and the Cambridge Centre for Brain Repair, University of Cambridge; the Society for Endocrinology; Cambridge Electronic Design; MSD Neuroscience Research, and GlaxoSmithKline (GSK). The work of the authors is supported by: the Wellcome Trust (I.M.G., J.H., S.L.L., J.R.S.), Royal Netherlands Academy for Arts and Sciences (E.R.dK.), Cancer Research Committee of St. Bartholomew's Hospital (A.G.) NIMH (grant MH12526 to J. McGaugh: BR), Medical Research Council (M.H.H.), the Canadian Institutes of Health Research (S.J.L.) and British Heart Foundation (J.R.S.). We thank G. Burton, B. J. Everitt, A. C. Roberts and E. B. Keverne for their part in organising the meeting, under the chairmanship of M.H.H.

Ancillary