Abstract
Brain cells are continuously exposed to corticosteroid hormones, although the levels vary (e.g., after stress). Corticosteroids alter neural activity via two receptor types, mineralocorticoid (MR) and glucocorticoid receptors (GR). These receptors regulate gene transcription but also, as we now know, act nongenomically. Via nongenomic pathways, MRs enhance and GRs suppress neural activity. In the hypothalamus, inhibitory GR effects contribute to negative feedback regulation of the stress axis. Nongenomic MR actions are also important extrahypothalamically and help organisms to immediately select an appropriate response strategy. Via genomic mechanisms, corticosteroid actions in the basolateral amygdala and ventral-most part of the cornu ammonis 1 hippocampal area are generally excitatory, providing an extended window for encoding of emotional aspects of a stressful event. GRs in hippocampal and prefrontal pyramidal cells increase surface expression of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors and strengthen glutamatergic signaling through pathways partly overlapping with those involved in long-term potentiation. This raises the threshold for subsequent induction of synaptic potentiation and promotes long-term depression. Synapses activated during stress are thus presumably strengthened but protected against excitatory inputs reaching the cells later. This restores higher cognitive control and promotes, for example, consolidation of stress-related contextual information. When an organism experiences stress early in life or repeatedly in adulthood, the ability to induce synaptic potentiation is strongly reduced and the likelihood to induce depression enhanced, even under rest. Treatment with antiglucocorticoids can ameliorate cellular effects after chronic stress and thus provide an interesting lead for treatment of stress-related disorders.
I. Introduction: from Stress to Changes in Neural Function
Potential threats to homeostatic processes (stressors), physical or psychological in nature, are registered in the brain and give rise to a well orchestrated response via activation of the autonomic nervous system and the hypothalamopituitary-adrenal (HPA1) axis. The stressors do not have to be present; in anticipation of potentially threatening situations, the same cascade is triggered. The actual or anticipated situations of threat are subjectively experienced as “stress.”
A. Stress and Its Mediators: Focus on Corticosteroids
Signals caused by physical stressors, such as respiratory distress or pain, are conveyed to the brainstem and from there can activate preganglionic cells in the intermediolateral cell column of the spinal cord (Kvetnansky et al., 2009; Ulrich-Lai and Herman, 2009). This allows an almost immediate response to the stressor via the sympathetic nervous system, of which the release of adrenaline and noradrenaline from the adrenal medulla is one of the most powerful effector tools. The parasympathetic nervous system is also activated to prevent the reaction from overshooting.
In most cases, though, a much more extensive network of brain regions becomes activated in response to (potentially) dangerous situations, in which the paraventricular nucleus (PVN) of the hypothalamus plays a key role. Neurons located in the dorso- and ventromedial part of this nucleus can activate preganglionic sympathetic neurons, adding a more central and integrated aspect to activation of the sympathetic nervous system (Sawchenko et al., 2000). Peripheral release of adrenaline can indirectly change noradrenaline levels in the brain, via the vagal nerve and the nucleus tractus solitarii (Williams and McGaugh, 1993). However, the PVN also harbors cells that are essential to activation of another system. Thus, parvocellular neurons in the middle part of the PVN produce corticotrophin-releasing hormone (CRH) and vasopressin, which upon stress are released in high amounts from terminals at the median eminence into the portal vessels. Through these vessels, CRH and vasopressin reach the anterior pituitary, where their synergistic actions lead to optimal secretion of adrenocorticotropin hormone into the circulation (Gillies et al., 1982). In turn, this gives rise to the synthesis and release of hormones from the adrenal cortex, primarily cortisol in humans and corticosterone in rodents (but also aldosterone; see section I.B). These actions together constitute the HPA axis (Fig. 1).
The activity of the HPA axis is under strong control of many limbic regions, so that perceptual aspects, emotional elements, and reference to earlier experiences can be integrated with information from lower brain areas, usually concerning physical aspects (Ulrich-Lai and Herman, 2009). Although some of the limbic projections impinge directly on parvocellular neurons in the PVN, most inputs relay through GABAergic neurons surrounding the PVN or located in the bed nucleus stria terminalis. Excitatory projections from the hippocampus—particularly from the ventral subiculum—and the prelimbic prefrontal cortex thus indirectly inhibit CRH-producing cells. By contrast, neurons in the infralimbic prefrontal cortex and amygdalar nuclei activate the HPA axis; the latter is explained by the GABAergic nature of the amygdalar efferent projections, causing disinhibition of CRH neurons.
Exposure to stressful situations thus involves release of many neurotransmitters and hormones, including catecholamines, CRH, vasopressin, and corticosteroid hormones, which collectively and in concert help the organism to adapt to the changing environment. This review will focus on only one element in this complex response (i.e., the effect of corticosteroid hormones on electrical activity of forebrain neurons under nonstressed and stressed conditions).
B. Release of Corticosteroid Hormones
Upon activation by adrenocorticotropin, the adrenal cortex secretes the glucocorticoids corticosterone and cortisol. The name “glucocorticoid” refers to the important contribution of these hormones to gluconeogenesis in the liver, as opposed to “mineralocorticoids,” another class of adrenal steroids, with a prominent function in maintaining the mineral balance in the body, primarily by acting on the kidney. Aldosterone, a steroid synthesized in the zona glomerulosa of the adrenal cortex, is the main mineralocorticoid hormone. It circulates in a >100-fold lower concentration than corticosterone or cortisol and does not play a major role in limbic areas. Because the referral to glucocorticoid activity gives rise to confusion when discussing the brain – most of the ‘glucocorticoid’ actions in the brain bear no direct relevance to glucose metabolism (see also sections I.C and I.D)– we will in the remainder of the text use the term ‘corticosteroid’ hormones when discussing the actions of corticosterone and/or cortisol in the forebrain.
In mammals, daily corticosteroid release follows a circadian pattern, with hormone levels peaking at the end of the resting phase in anticipation of the increased metabolic demand of the active period (Young et al., 2004). Circadian corticosteroid release is principally controlled by the hypothalamic suprachiasmatic nucleus and is further orchestrated by efferent projections controlling the CRH/vasopressin-containing neuroendocrine cells in the PVN and median eminence (Kalsbeek and Buijs, 2002; Engeland and Arnhold, 2005), and adrenal gland sensitivity mediated by the autonomic nervous system (Ulrich-Lai and Herman, 2009).
High resolution blood sampling methods have shown that these circadian fluctuations actually overlay a highly oscillatory ultradian pattern (Fig. 2). Typically, the adrenal gland releases bursts of corticosteroids into the blood with a periodicity of approximately 60 min, with circadian modulation in amplitude. These pulsatile patterns are maintained across the blood-brain-barrier and persist in corticosteroid target regions such as the hippocampus (Droste et al., 2008). Ultradian corticosteroid pulsatility has been described in many species (Weitzman et al., 1971; Tapp et al., 1984; Jasper and Engeland, 1991; Windle et al., 1998b; Cook, 2001). Recent studies demonstrated that corticosterone pulsatility significantly contributes to physiology by maintaining normal GR signaling and HPA axis responsiveness to stress (Lightman and Conway-Campbell, 2010; Sarabdjitsingh et al., 2010a). These results have important implications for experimental studies but also for therapeutic application, because continuous corticosteroid administration is likely to be less effective than pulses. The high efficacy of pulses was already appreciated decades ago in clinical therapy with respect to growth hormones and estrogen replacement therapy (Amato et al., 2000; Shoupe, 2001), but no such administration protocols have yet been designed for corticosteroids. A better understanding of pulsatile glucocorticoid release and the underlying nuclear receptor mechanism may importantly contribute to the prognosis and treatment of (stress-related) diseases.
Until recently, the origin of ultradian corticosterone pulse generation was unknown. Walker et al. (2010) provided elegant biomathematical evidence that systems with a delay between reciprocally connected feedforward and feedback pathways such as that of the HPA axis, by definition have no other choice than to oscillate. The authors proposed a model describing that rapid feedforward activity via pituitary adrenocorticotropin release and a slightly delayed feedback loop via adrenal corticosterone results in self-sustaining rapid ultradian corticosterone oscillations, even in the absence of hypothalamic input. Augmentation of the pulsatile pattern can occur by alterations in adrenal gland sensitivity and steroidogenesis (Ulrich-Lai et al., 2006; Son et al., 2008; Spiga et al., 2011).
Ultradian corticosteroid amplitude and frequency can be remarkably plastic and vary throughout transitions in the life span (e.g., puberty, lactation, and ageing) (Windle et al., 1997; Lightman et al., 2000; Evuarherhe et al., 2009). Differences in corticosteroid patterns also exist for instance between sexes or strains that differ in their susceptibility to stress (Windle et al., 1998a; Seale et al., 2005). Deregulation in pulse characteristics have been linked to changed neuroendocrine and behavioral responsiveness to stress (Sarabdjitsingh et al., 2010a,c) but also to early-life stress in rodents and to various (stress-related) pathological conditions in humans such as inflammation, depression, and Cushing's syndrome (Young et al., 1994; Shanks et al., 2000; Windle et al., 2001; van Aken et al., 2005). Deviations from the normal physiological pulsatile pattern might enhance vulnerability to (psycho)pathology (Young et al., 2004; Lightman and Conway-Campbell, 2010; Sarabdjitsingh et al., 2012). Collectively, these data illustrate that circulating levels of corticosteroid hormones not only vary as a result of stress exposure but also intrinsically as a result circadian and ultradian processes.
C. Factors Determining the Availability of Corticosteroids in Brain
The effect of corticosteroid receptor signaling in target tissues such as the brain is largely determined by the local availability of the steroids (Fig. 1). Although corticosteroid secretion is the main regulatory system that determines ligand concentrations, multiple factors “downstream” to the adrenals also influence corticosteroid levels. These include corticosteroid binding to carrier proteins in blood [i.e., corticosteroid-binding globulin (CBG) and albumin] that affect uptake by target tissues and transport across cell membranes, including the blood-brain barrier. Moreover, intracellularly, corticosteroids are prone to enzymatic conversion, which influences receptor signaling more directly.
CBG is a low-capacity, high-affinity plasma protein and functions as the principal carrier for circulating corticosterone and cortisol. Together with albumin, it binds approximately 95% of the corticosteroid pool. The remaining unbound steroid fraction is free to diffuse across target cell membranes and has ready access to receptors. CBG also affects metabolic clearance rates, because only the unbound fraction is subject to degradation by the liver (Hammond, 1990; Rosner, 1990). Under normal conditions, CBG thus acts as a tissue buffer against the potentially detrimental effects of high corticosteroid levels, by regulating the availability of free steroid access to the brain. Under particular circumstances (e.g., ultradian/circadian peak, stress, disease, enzyme activity, or CBG genetic variance), CBG saturation is exceeded or binding properties and expression levels are affected, increasing the corticosteroid bioavailability (Breuner and Orchinik, 2002; Gagliardi et al., 2010; Henley and Lightman, 2011). CBG is a prime target for specific classes of proteinases (e.g., neutrophil elastase). Elastase cleavage of CBG results in the irreversible loss of steroid binding, thereby presumably elevating steroid concentrations at the sites of inflammation (Klieber et al., 2007).
Before corticosteroids can reach and enter neuronal target cells they have to pass the blood-brain barrier. This highly dynamic, physical, and metabolic barrier involves specialized endothelial cells and maintains brain homeostasis by protecting the brain from compounds in the circulation (Bradbury, 1993). Natural corticosteroids, particularly corticosterone, can readily pass the blood-brain barrier by passive diffusion, because of their small size and lipophilic nature. In accordance, the pulsatile pattern of corticosterone is maintained across the blood-brain barrier, suggesting that neurons are exposed to highly fluctuating corticosterone levels (Droste et al., 2008). This contrasts with the endogenous hormone cortisol (which in rodents circulates in much lower concentrations than corticosterone) and the synthetic steroid dexamethasone, which have hampered penetration because they are a substrate of the multidrug resistance (mdr) 1a P-glycoprotein (Meijer et al., 1998; Karssen et al., 2001). Mdr1a P-glycoprotein is also thought to expel corticosteroids from brain cells (Pariante, 2008). It is noteworthy that this membrane transporter has been proposed as a promising therapeutic target for treatment of depression and the presumed steroid resistance in this disease (Pariante, 2008). For instance, the antidepressant desipramine requires the mdr1a P-glycoprotein to up-regulate GRs in the mouse brain (Yau et al., 2007), potentially normalizing HPA-axis function, which is often aberrant in depressives. Genetic polymorphisms in the mdr1a P-glycoprotein were found to predict clinical response to antidepressants (Pariante, 2008).
After crossing the blood-brain barrier, intracellular levels of corticosteroids are additionally modulated by 11β-hydroxysteroid dehydrogenase (HSD) enzymes. These exist in two isoforms and tightly regulate the interconversion of corticosteroids in cells expressing these enzymes (Wyrwoll et al., 2011). Type 2 (11βHSD-2) catalyzes inactivation of glucocorticoids, so that in the kidney, for example, corticosterone and cortisol are no longer available to bind receptors for which they have very high affinity, allowing the less prevalent hormone aldosterone to bind. Type 2, however, is not substantially expressed in most brain regions (with the exception of the nucleus tractus solitarii), so that corticosterone and cortisol are now the main ligands for the corticosteroid receptors (Wyrwoll et al., 2011). Type 1 is ubiquitously present in multiple tissues, including the brain and pituitary and promotes the reduction of inactive 11-keto-derivatives, such as dehydrocorticosterone and cortisone, into active corticosterone and cortisol, respectively (Seckl and Walker, 2004). These enzymes are essential in determining intracellular corticosteroid levels. The relevance of the enzyme for cognitive function was illustrated by the fact that 11βHSD-1 deficiency protects against cognitive decline during ageing (Wyrwoll et al., 2011).
The conclusions of these studies are 1) that hormone levels in the brain quite accurately follow the fluctuations seen peripherally, albeit with a delay; and 2) that corticosterone and cortisol—rather than aldosterone—are the main ligands for corticosteroid receptors in the brain.
D. Corticosteroid Hormone Receptors in the Brain
Once corticosteroid hormones have entered the brain, they can in principle exert their actions on those cells expressing a receptor. Two types of receptors were demonstrated in brain tissue (Reul and de Kloet, 1985): mineralocorticoid receptors (MRs) and glucocorticoid receptors (GRs). These names refer to the main peripheral processes in which they are involved: mineral balance and gluconeogenesis, respectively. The genes encoding the receptors mediating these peripheral functions are identical to the genes in the brain (Hollenberg et al., 1985; Arriza et al., 1987).
Despite the fact that only two genes encode for these receptors, there are many isoforms known, giving rise to considerable variation in the expression levels and transcriptional activity of corticosteroid receptors in the various brain regions. A wave of corticosteroids reaching the brain may therefore lead to extensive regional differentiation in the transcriptional response, at least with nonsaturating concentrations (for discussion, see section VI.B). More specifically, corticosteroid receptors belong to the superfamily of nuclear receptors, which act as transcription factors. The human GR gene consists of nine exons; exons 2 to 9 code for the GR protein (Derijk and de Kloet, 2008; Revollo and Cidlowski, 2009; Turner et al., 2010) (Fig. 3). Exon 1 is liable to alternative splicing, each variant having its own transcriptional start site and promoter region. The variants display regional specificity, which is thought to contribute to differences in GR expression level. Exon 9 can also be alternatively spliced, resulting in the prevalent GRα or the shorter GRβ isoform. Both GRα and GRβ mRNA can give rise to eight translational isoforms, each with its own tissue expression, subcellular localization, and transcriptional activity. Transcriptional activity is further fine-tuned by post-translational modifications. Much less is known about isoforms of the MR, but for this receptor as well, different promoter regions, several splice variants, translational variants, and post-translational modifications have been described previously (Pascual-Le Tallec and Lombès, 2005; Derijk and de Kloet, 2008; Gomez-Sanchez, 2010).
The two receptor types are not uniformly distributed in the brain (Reul and de Kloet, 1985). GRs are quite ubiquitous in their distribution and expressed in neurons and glial cells alike. Nevertheless, there are brain regions with very high expression levels, such as the PVN, the CA1 hippocampal area and dentate gyrus, central and cortical amygdala, lateral septum, and nucleus tractus solitarii. Conversely, MRs are very restricted in their distribution but are, for instance, highly expressed in neurons of all hippocampal subfields and the lateral septum.
The MR and GR proteins consist of several domains that are important for their function (Revollo and Cidlowski, 2009) (see Fig. 3):
An N-terminal domain with an activation function motif through which the receptor can interact with the transcriptional machinery;
The DNA-binding domain consisting of two zinc-fingers essential for receptor dimerization and binding to recognition elements in the DNA;
A hinge that contains a nuclear localization signal; and
A C-terminal region that contains 1) a second nuclear localization signal, 2) a motif that is important for protein-protein interactions, and 3) the pocket to which ligands can bind, each with a specific affinity.
MR and GR have different affinities for endogenous corticosteroid hormones, so that variations in hormone concentrations in the brain lead to shifts in the balance between MR and GR activity. The MR has a high affinity for the endogenous hormones aldosterone, corticosterone and cortisol, with a Kd of approximately 0.5 nM (Reul and de Kloet, 1985). The affinity of the GR for corticosterone and cortisol is approximately 10-fold lower and manyfold lower for aldosterone. Given 1) the very high affinity of MRs for corticosterone, 2) the >100-fold excess of corticosterone over aldosterone, and 3) the fact that most brain cells express high levels of 11β-HSD1 instead of 11β-HSD2, these receptors in brain will almost always be substantially occupied by corticosterone (or cortisol), even with hormone concentrations during the trough of the circadian rhythm and ultradian pulses. By contrast, the affinity of GRs is such that these receptors will be only partly occupied when corticosteroid levels are low but gradually become occupied when hormone levels rise (e.g., at the peak of ultradian pulses and the circadian rhythm or after stress). Thus, the expression levels of the receptors determine the range in which neural activity can be affected and the hormone concentration the actual position within this range.
E. Transcriptional Regulation
In the inactive state, MRs and GRs are located in the cytoplasm and bound to chaperones such as the 90-kDa heat-shock protein. Upon binding of the ligand, this complex dissociates, uncovering nuclear localizations signals, so that the ligand-receptor complex moves to the nucleus. There, they can either bind as homodimers to consensus sequences in the promoter of 1 to 2% of the genes and directly change transcription of these genes or interact with other transcription factors, altering their efficacy (Datson et al., 2008) (see Fig. 1). The former pathway usually promotes gene transcription, whereas the latter can result in suppressive or facilitating effects.
Corticosteroid hormones are fundamental in the regulation of metabolism, development, inflammation as well as several other biological functions. Such a widespread spectrum of biological involvement suggests that a broad array of genes may be regulated by the receptors. Indeed, in clinical therapy, synthetic glucocorticoids are frequently used for the treatment of many pathological conditions. However such generalized glucocorticoid administration often concurs with many adverse side-effects. From a clinical and pharmacological point of view, the classification of (tissue-specific) glucocorticoid target gene-expression profiles into functional categories is necessary information to enhance clinical efficacy while minimizing potential adverse effects.
Functional categories of target genes have been explored for the brain as well. Thus, large-scale gene expression profiling has proved to be a successful strategy in the identification of central target genes underlying corticosteroid-mediated effects in the brain (Datson et al., 2001). The transcriptional response to acute GR activation is highly dynamic (Fig. 4). For instance, hippocampal slices exposed to a 20-min pulse of 100 nM corticosterone were profiled 1, 3, and 5 h after treatment; this dose of corticosterone is assumed to primarily activate GRs in slices prepared from animals killed at the circadian trough and under rest, because the MRs (but not GRs) are already substantially occupied under control conditions (i.e., before corticosterone treatment). Strikingly, the transcriptional profiles seemed to occur in a wave-like pattern, with a shift from strictly down-regulated genes within 1 h to up-regulation of the majority of the genes 3 h after GR activation. After 5 h, the response was back to baseline (Morsink et al., 2006b). Highly similar response profiles of corticosteroid responsive target genes were obtained in cultured cells, displaying kinetically complex transcriptional patterns, frequently with alternating activation and repression phases (Morsink et al., 2006a; John et al., 2009). Classification of these corticosteroid target genes into functional categories resulted in clusters of genes that mainly involve energy metabolism, signal transduction, neuronal structure, vesicle dynamics, neuronal catabolism, cell adhesion, and genes involved in regulation of corticosteroid signaling (Datson et al., 2008).
Corticosteroids have regionally different receptor-mediated transcriptional effects, which is important for diversity in physiological effects. These differential corticosteroid-dependent transcriptional responses depend on a number of factors contributing to cell- and even subregion-specific transcriptional responses in the brain, such as the local receptor expression and properties (Kitchener et al., 2004; Turner et al., 2010), differential coregulator recruitment (van der Laan and Meijer, 2008), DNA composition (So et al., 2007; Datson et al., 2011), and chromatin accessibility and remodelling mechanisms (John et al., 2011).
Accumulating evidence shows that GR-mediated transcriptional efficiency is not only governed by stress-induced release of corticosteroids but is also under control of ultradian corticosterone pulsatility (Lightman and Conway-Campbell, 2010; Conway-Campbell et al., 2012). Repetitive corticosterone exposure results in transient cyclic recruitment and exchange of GR at regulatory sites in the genome (Stavreva et al., 2009; Conway-Campbell et al., 2010), whereas MR is retained in the nucleus (Conway-Campbell et al., 2007). Experimental evidence supports the idea that pulsatile patterns are necessary to safeguard GR sensitivity and downstream signaling in target tissues (Sarabdjitsingh et al., 2010b).
In summary, when a surge of corticosteroids reaches the brain, a transient wave of primary transcriptional events takes place, targeting many genes. This changes the levels of multiple proteins, including those involved in neural activity, in a time window that commences when corticosteroid levels are starting to normalize and that outlasts the hormonal surge by at least several hours. Pulsatile hormone release seems necessary to guarantee optimal transcriptional efficiency.
F. Nongenomic Pathways
Despite the overwhelming amount of data showing that corticosteroid hormones via their receptors regulate gene transcription and the detailed knowledge that has been obtained about the signaling pathways, it has been known for decades that corticosteroids can also quickly change neural and brain function, in a time domain that is incompatible with pathways involving transcription and translation (for review, see Borski, 2000; Evanson et al., 2010). For instance, a pioneering study by Pfaff et al. (1971) showed that peripherally administered corticosterone suppresses hippocampal firing within 20 min. Almost instantaneous though prolonged effects of corticosteroids were observed in PVN neurons antidromically identified through their projections to the median eminence (Saphier and Feldman, 1988; Chen et al., 1991); of these cells, nearly three quarters were inhibited by iontophoretically applied corticosterone or hydrocortisone.
Behavioral investigations also supplied evidence for rapid corticosteroid actions in brain. For example, corticosterone was found to inhibit male reproductive behavior as well as medullary firing in newts (Rose et al., 1993), mediated by a corticosteroid receptor in the membrane with a pharmacological profile different from the “classic” nuclear receptor (Orchinik et al., 1991) and linked to G-protein coupled signaling pathways (Orchinik et al., 1992). Such an unexpected pharmacological profile for rapid corticosteroid effects has also been observed in tissues other than the brain (Lösel and Wehling, 2008). A second example of rapid behavioral modulation by corticosterone concerns a mutual rapid positive feedback between HPA axis activation and the brain mechanism controlling aggression: hypothalamic aggression was reported to be rapidly enhanced by corticosterone in adrenalectomized rats, whereas stimulation of the hypothalamus rapidly activated the adrenocortical response, even in the absence of an opponent (Kruk et al., 2004).
The rapid nongenomic corticosteroid actions expand the time window during which these hormones can change neural function. Effectively, their influence stretches from as soon as they enter the brain to many hours later.
G. From Stressor to Multiple Functional Endpoints
As is evident from the previous sections, once corticosteroids reach a neuron they can exert multiple effects via either genomic or nongenomic signaling (Fig. 1). The nature of the effect firstly depends on the type of receptor that is activated; this in turn is determined by the local receptor availability, the sensitivity of receptors to corticosteroids, as well as the hormone concentration. None of these is a stable factor, though; they are largely modulated by genetic background such as receptor haplotypes, life history of the subject, and even whether or not the organism was exposed to a stressor in recent hours. Most of these aspects will be discussed in the later parts of the review (sections IV and V).
The transcriptome analyses to date underline that corticosteroid hormones target multiple classes of genes, which in some cases (e.g., when corticosteroids target the gene encoding for another transcription factor) regulate a whole array of secondary genes. One surge of corticosteroids can thus change numerous endpoints. These can be studied at many different levels of organization, ranging from changes in gene transcripts, proteins, and biochemical pathways to neurotransmission, structural plasticity, and behavioral endpoints. In this review, we focus particularly on the actions of corticosteroid hormones on neural activity (Fig. 5): How do these steroids affect the properties of ion channels, whole-cell currents, action potentials, firing frequency of cells, and single-cell or field responses to the main excitatory and inhibitory inputs? These endpoints at an intermediate level of integration are highly relevant, because they give insight in functional connectivity between (groups of) neurons. We review the literature for both the rapid nongenomic effects (section II) and slow gene-mediated actions (section III) in the mammalian forebrain. At the end of each of these sections, the relevance of these changes in electrical activity for behavioral adaptation will be discussed.
II. Rapid Effects
Although several studies over the past 30 years reported corticosteroid-induced changes in neural activity that developed within seconds to minutes—a delay that is incompatible with the classic signaling pathway involving gene transcription and translation—extensive investigation at the single cell level started only approximately a decade ago. Rapid and presumably nongenomic actions have now been observed in several brain areas (Supplemental Table I). We will here discuss only those areas in which the rapid actions of corticosteroids are well documented.
A. Rapid Modulation of Neural Activity
1. Hippocampus.
Over the past years it has become increasingly evident that neurons in the ventral-most (20%) part of the hippocampus have electrical properties, including their response to corticosteroids and stress, different from those of neurons in the rest of the hippocampus (Maggio and Segal, 2010). The data discussed in this review pertain to the latter part (the vast majority of cells), unless stated otherwise.
Rapid effects of corticosteroid hormones on passive and active membrane properties or on specific voltage-dependent ion currents have hardly been investigated in the hippocampus. The former appear not to be affected by corticosterone (Joëls and de Kloet, 1993). An early study reported rapid inhibition of L- and N-type calcium currents in dissociated CA1 pyramidal neurons, but this required very high concentrations of cortisol, whereas corticosterone was not very potent (ffrench-Mullen, 1995). More recently, it was found that the voltage-dependence of activation of a transient K+ current (IA) in CA1 pyramidal cells is shifted to the right by corticosterone via an MR-dependent postsynaptic mechanism, so that this channel is less activated during small depolarizations (Olijslagers et al., 2008). This will result in a higher likelihood of inducing action potentials with excitatory input.
More information has appeared regarding corticosteroid actions on excitatory and inhibitory inputs to the hippocampus. Rapid excitatory transmission in the brain is primarily mediated by glutamate, acting via AMPA and NMDA receptors. Glutamate release is evoked by the arrival of action potentials and subsequent calcium release in the presynaptic terminal, but to a limited extent glutamate transmission also occurs spontaneously, even in the absence of action potentials. This spontaneous background activity is apparent (Fig. 5) from the postsynaptic response to a spontaneously released synaptic vesicle containing glutamate, a so-called miniature excitatory postsynaptic current (mEPSC). Inhibitory transmission is mostly carried by GABA. Spontaneous release of GABA-containing vesicles is postsynaptically recorded as a spontaneous inhibitory postsynaptic current (sIPSC) or, in case of action potential blockade miniature IPSC (mIPSC).
CA1 hippocampal pyramidal cells were found to respond to corticosterone within minutes with enhanced mEPSC frequency (Karst et al., 2005) (Fig. 6). Other properties of the mEPSCs, such as amplitude, rise time, or decay, were entirely unaffected by the hormone. In a paired-pulse stimulation paradigm, corticosterone diminished the second evoked response relative to the first, indicating that hormone treatment most likely boosts the release probability of glutamate containing vesicles rather than increasing the number of synaptic contacts. Upon washout of corticosterone, the mEPSC frequency rapidly returned to the pretreatment level. A second pulse of corticosterone 1 h later evoked a highly comparable response (Karst et al., 2010). The rapidly reversible response and the fact that corticosterone exerted very similar effects in the presence of a protein synthesis inhibitor support the theory that this effect is accomplished through a nongenomic pathway. Corticosterone conjugated to (membrane-impermeable) bovine serum albumin (BSA) induced very similar effects on mEPSC frequency, whereas intracellular administration was ineffective (Karst et al., 2005; Olijslagers et al., 2008). This suggests that corticosterone binds to a molecule that is accessible from the outside of the cell. Pharmacological and genetic tools supported that this involves an MR rather than GR. The MR-dependent increase in mEPSC frequency requires expression of limbic system-associated membrane protein, Lsamp (Qiu et al., 2010). A highly similar MR-dependent raise in mEPSC frequency was observed in granule cells of the dentate gyrus (Pasricha et al., 2011).
Follow-up studies in the CA1 area revealed that MRs mediating the rapid effect are probably localized on the presynaptic membrane and linked to the ERK1/2 signaling pathway (Olijslagers et al., 2008). It is noteworthy that corticosterone changes glutamate transmission in the CA1 hippocampal area not only by increasing the release probability, a presynaptic property, but also by other means. In the postsynaptic membrane, GluA2 subunits of the AMPA receptor exhibit increased lateral movement and a higher dwell time in the postsynaptic density (Groc et al., 2008). This effect as well was mediated by MRs, mimicked by corticosterone-BSA, and persisted in the presence of a protein synthesis inhibitor. Both actions on glutamate transmission are expected to increase the (spontaneous) activity of hippocampal CA1 neurons. This would fit with the enhanced population spike amplitude observed extracellularly when treating slices from adrenalectomized rats (devoid of endogenous corticosteroids) with low doses of corticosterone, presumably activating MR (Reiheld et al., 1984).
One study (Tse et al., 2011) reported that CA1 cells also respond more strongly to a slightly later excitatory input (i.e., 20–30 min after the start of corticosterone administration), although the characteristics were somewhat different. When excitatory postsynaptic currents were synaptically evoked (eEPSC), the NMDA/AMPA ratio was increased. No change was observed in paired-pulse responsiveness. This effect seemed to be mediated by GR [i.e., mimicked by dexamethasone and blocked by the GR-antagonist mifepristone (RU38486)] and was not accompanied by AMPA-receptor subunit trafficking. The fact that mifepristone was effective raises the question of whether these effects are truly nongenomic, because this drug is generally ineffective in blocking membrane-receptor–mediated events (for example, see Di et al., 2003; Liu et al., 2007; Zhang et al., 2012). Tse et al. (2011) observed an extracellular increase in the field excitatory postsynaptic potential evoked via NMDA but not AMPA receptors.
With this slightly longer delay, however, several studies reported reduced responses to synaptic input. Spontaneous firing rate of hippocampal cells was reduced by corticosterone peripherally injected 20 min earlier (Pfaff et al., 1971). This fits with more recent data showing that various types of stress impair the stability or reduce the firing rate of hippocampal place cells in this time domain (Kim et al., 2007; Passecker et al., 2011). In vitro-administered corticosterone (at a very high dose) was found to reduce the population spike amplitude in the CA1 area, an effect that reached a plateau 20 to 40 min after beginning corticosterone administration (Vidal et al., 1986). Likewise, the probability of evoking an action potential with synaptic stimulation, as well as the amplitude of the excitatory postsynaptic potential and slow inhibitory postsynaptic potential of CA1 neurons, gradually declined starting 20 min after corticosterone administration (Joëls and de Kloet, 1993). Exposure to a foot shock before preparation of slices reduced the amplitude and frequency of CA1 hippocampal mEPSCs as well as paired-pulse facilitation (Zhang et al., 2005; Gao et al., 2008). The latter may also be linked to the fact that this study involved much younger animals than most other studies. In this respect, it is interesting that corticosterone also reduced NMDA-evoked currents in neonatal cultured hippocampal neurons through a membrane-bound receptor not blocked by classic MR or GR antagonists (Liu et al., 2007; Zhang et al., 2012).
The overall activity of CA1 pyramidal cells will depend on the balance between excitatory and inhibitory transmission. Exactly how corticosteroids affect GABAergic transmission in the CA1 area is not quite clear. GABAergic inhibitory responses were reduced when recording with sharp electrodes but slightly enhanced with whole-cell recording, indicating that an intracellular molecule is important in mediating the reduced inhibition (Zeise et al., 1992; Teschemacher et al., 1996). Hu et al. (2010) reported—using whole-cell recording—that restraint stress and dexamethasone-BSA increase the frequency and amplitude of sIPSCs, although mIPSCs were unaffected. This effect involved postsynaptic G proteins and retrograde transport by nitric oxide and was not blocked by either MR or GR antagonists.
Although it is not easy to predict how corticosteroid hormones will quickly affect the overall information flow through the CA1 hippocampal area, the balance seems to be tipped toward enhanced spontaneous excitatory activity. Synaptically evoked field potentials, though, are not markedly altered by corticosterone administration (e.g., Wiegert et al., 2006; Pu et al., 2007), but this is a rather indirect signal so that subtle effects may have remained unnoticed. This is also when other transmitters and hormones released by stress are active and determine excitability. For instance, CRH is known to quickly potentiate the population spike in the Schaffer projection to the CA1 hippocampal area (Blank et al., 2002). Overall, enhanced hippocampal activity during this phase directly after stress may prevail. Slightly later (i.e., 20–30 min after corticosterone reaches hippocampal cells), inhibitory actions seem more prevalent. This is an ambiguous time-domain, rather slow for nongenomic actions but probably too rapid for genomic actions.
2. A New Concept in the Basolateral Amygdala: Metaplasticity.
In the basolateral amygdala (BLA), MRs are expressed at a lower level than the hippocampus (Reul and de Kloet, 1985). Nevertheless, corticosterone rapidly increased mEPSC frequency of principal cells in the BLA via an MR-dependent mechanism (Karst et al., 2010) (Fig. 6), similar to what was observed in the CA1 and dentate neurons. However, in the BLA, the frequency remains at a high level, even after washout of corticosterone. Although the onset is clearly nongenomic, the persistence of the response critically depends on protein synthesis and expression of both corticosteroid receptor types. This (assumed) genomic element changes the properties of BLA cells such that they respond differently to a second pulse of corticosterone, in this case with a diminishment of mEPSC frequency (i.e., exactly the opposite of what is seen upon the first exposure) (Fig. 6). The rapid suppression of mEPSC frequency requires expression of GRs and the cannabinoid receptor-1. It was also seen when animals had been exposed to stress before preparation of the brain slices. This indicates that the rapid response of BLA neurons to corticosterone depends on the recent stress history of the organism, a phenomenon that has been dubbed “metaplasticity” (Karst et al., 2010). The shift in responsiveness after a second corticosterone treatment is possibly due to a change in the MR/GR ratio in the membrane (e.g., caused by internalization of MRs after the first pulse of corticosterone).
The extent to which these modulations in mEPSC properties are reflected in the transfer of ongoing excitatory transmission is unclear. At both the single-cell and the field-potential levels, corticosterone did not quickly change AMPA or NMDA receptor-mediated synaptic responses (Liebmann et al., 2009; Pu et al., 2009).
3. Paraventricular Nucleus.
In parvocellular neurons of the PVN—including those producing CRH—exposure to corticosterone or dexamethasone was reported to consistently decrease the release probability of glutamate-containing vesicles (Di et al., 2003) (Fig. 6). Antagonists of the MR or GR did not prevent this effect of dexamethasone. Although the involvement of a nonspecific corticosteroid receptor was proposed initially, later experiments showed that rapid effect of glucocorticoids are abolished in conditional GR knockout mice (Haam et al., 2010; Tasker and Herman, 2011), suggesting a role for GRs. The suppression in mEPSC frequency involves changes in endocannabinoid levels and retrograde signaling on the cannabinoid receptor-1 (Di et al., 2003) and depends on Gαs (Di et al., 2009). Similar effects on mEPSC frequency were observed in magnocellular neurons of the PVN, via the same signaling mechanism (Di et al., 2005). It is noteworthy that in these cells corticosteroids also changed spontaneous GABAergic signaling, causing an enhanced GABAergic tone via nitric oxide and Gβγ (Di et al., 2009). Corticosterone does not affect mIPSC frequency in parvocellular PVN cells (Verkuyl et al., 2005).
As mentioned above, prior stress exposure reveals a cannabinoid receptor-1-dependent suppression of mEPSC frequency in BLA neurons that is not seen under rest. This also may apply to PVN neurons. If so, moderately stressful conditions just before onset of the experiment (e.g., transportation from the animal house to the laboratory) could explain the suppression reported for the PVN. However, the PVN may also be intrinsically different from the BLA with respect to its rapid corticosteroid actions.
B. Rapid Modulation of Synaptic Plasticity
The frequency of mEPSCs is enhanced during both the early and the late phases of synaptic plasticity (Isaac et al., 1996; Wiegert et al., 2009), and enhanced glutamate release can cause LTP-like strengthening of synapses (reviewed in Nicoll and Schmitz, 2005). Rapid modulation of mEPSC frequency by corticosteroid hormones coinciding with high-frequency input may thus have consequences for synaptic plasticity.
Rapid enhancement of LTP by corticosterone was observed in the CA1 and DG (but Filipini et al., 1991; Wiegert et al., 2006; Pu et al., 2007; Tse et al., 2011), depending on the presence of limbic system-associated membrane protein (Qiu et al., 2010). However, a series of articles showing that, in particular, exposure to a novel environment prevents induction of LTP, depotentiates earlier induced LTP, and facilitates the induction of LTD, in the CA1 area but also nucleus accumbens (Xu et al., 1997, 1998; Manahan-Vaughan and Braunewell, 1999; Hugues et al., 2003; Yang et al., 2004; Tse et al., 2011). Remarkably, these effects could be blocked by mifepristone (Xu et al., 1998; Tse et al., 2011) and were found to depend on protein synthesis in one study (Xu et al., 1998), although their rapid onset (<10 min) and the fact that they occur independent of increases in corticosteroid level almost precludes a GR-dependent genomic effect.
C. The Underlying Mechanism
The enhanced mEPSC frequency reported for rapid corticosteroid actions clearly requires MRs (Karst et al., 2005, 2010; Olijslagers et al., 2008; Qiu et al., 2010; Pasricha et al., 2011). The suppression of mEPSC or enhancement of sIPSC frequency, via NO or endocannabinoid, seems to involve GRs (Di et al., 2003, 2005, 2009; Hu et al., 2010; Karst et al., 2010), although mifepristone was usually ineffective, and solid proof for a role of the GR gene was provided only in one study (Karst et al., 2010). The inhibitory effects observed in the CA1 area in a slightly later time window after stress or corticosterone application (i.e., after approximately 20 min) were reported to be blocked by mifepristone, pointing to GR involvement (Xu et al., 1998; Tse et al., 2011), and could be overcome by serotonergic agents (Shakesby et al., 2002). Some care needs to be taken in the interpretation of effects with very high corticosteroid concentrations (e.g., Vidal et al., 1986; Liu et al., 2007), because ethanol (in which corticosterone is dissolved) is also known to acutely reduce NMDA currents (Peoples et al., 1997) and to decrease mEPSC frequency and LTP of evoked EPSCs in hippocampal cells via enhancement of endocannabinoid levels (Basavarajappa et al., 2008).
Whenever tested, steroids conjugated to BSA were as effective in mediating the rapid effects as the native hormone (Di et al., 2003; Karst et al., 2005; Liu et al., 2007; Hu et al., 2010; Pasricha et al., 2011), suggesting that the receptor is located in the membrane. Electron microscopic proof for such membrane location is still scarce (Johnson et al., 2005; Prager et al., 2010), and biochemical isolation of a “membrane receptor,” unlike earlier success in amphibian brain (Orchinik et al., 1991, 1992; Rose et al., 1993), has not yet been convincingly demonstrated in mammalian brain tissue. The pharmacological profile of these presumed membrane-located receptors is different from that described for nuclear MR and GR. For example, inhibitory effects purportedly mediated by GRs are not always sensitive to mifepristone (Di et al., 2003; Hu et al., 2010), and the dose of corticosterone required to induce rapid increases in mEPSC frequency of CA1 cells is 10-fold higher than expected for an MR-dependent phenomenon (Karst et al., 2005). These deviations in pharmacological profile are not unprecedented (Orchinik et al., 1991; for review, see Lösel and Wehling, 2008). They may point to the existence of a different receptor molecule—such as the case for some rapid effects by estrogens (Prossnitz et al., 2008)—but could also be explained by the constraints imposed by the membrane localization and, for example, the inability of chaperones to associate with the receptor.
How can MRs or GRs be translocated to the plasma membrane instead of the nucleus? Very little is known about corticosteroid receptors, in contrast to the mechanisms behind regulation of the membrane associated estrogen receptor α (ERα). ERα was shown to be inserted into the membrane via two mechanisms: binding to caveoline-1 and palmitoylation of the receptor (Acconcia et al., 2005; Pedram et al., 2007). The insertion and internalization are very dynamic processes, as was demonstrated by Dominguez and Micevych (2010). Both insertion and internalization peaked within 1 h after estradiol treatment. GR and MR also bind to caveoline-1 (Matthews et al., 2008; Pojoga et al., 2010), although it is presently unclear whether this protein is involved in trafficking of corticosteroid receptors to the plasma membrane of neurons. The GR has a conserved palmitoylation motif, as opposed to the MR (Pedram et al., 2007). For the GR, it is therefore plausible that trafficking to neuronal membranes could take place via a mechanism comparable with that for the ERα, but it is less clear how MRs could be incorporated in the membrane.
D. Behavioral Relevance
It has been argued extensively that rapid corticosteroid actions on cellular activity in the PVN contribute to rapid negative feedback regulation of HPA axis activity after stress (Tasker et al., 2006; Tasker and Herman, 2011). In this respect, it seems very likely that the endocannabinoid-dependent suppression of mEPSC frequency in the PVN occurs secondary to activation of the axis, similar to what has been observed in the BLA (Karst et al., 2010); after all, a rapid negative feedback function is meaningful only in the light of earlier activation. The rapid inhibitory effect in the BLA and the excitatory effect in the CA1 region possibly also contribute to this negative feedback, given the stimulatory and suppressive projections, respectively, of these areas to the PVN (Ulrich-Lai and Herman, 2009).
However, rapid corticosteroid actions on electrical activity may also play a role in higher cognitive processes. In the 1990s, MRs were already found to be important for the appraisal of novel situations and selection of response strategies, thus promoting acquisition of important information (Oitzl and de Kloet, 1992; de Kloet et al., 1999); this MR function seems to require the presence of limbic system-associated membrane protein (Qiu et al., 2010). It has been proposed that nongenomic actions via MR favor a shift toward simple learning strategies, which form a good behavioral alternative at the short term (Schwabe et al., 2007, 2010a,b). MRs are also important for control of emotional arousal and adaptive behaviors, because this is lost in the absence of forebrain MR, so that anxiety-related responses remain augmented (Brinks et al., 2009). In line with this supposed adaptive role of MR, administration of an MR antagonist before training in a contextual fear conditioning paradigm interfered with adequate memory formation (Zhou et al., 2010) (Fig. 7). It is noteworthy that administration of the antagonist immediately after the learning trial was ineffective, indicating that appropriate MR activation during a restricted time window linked to acquisition of the task is important for optimal performance. A highly comparable role of MRs (but not GRs) was observed during reexposure to the aversive environment (Zhou et al., 2011), a situation in which stress hormone levels rise in association with the (earlier) learning context and promote behavioral performance. Re-exposure to a conditioned stressful environment was found to be associated with a higher likelihood to induce LTD (Hugues et al., 2003), which may be due to metaplastic changes in the circuit (see section III).
Administration of stress or corticosteroids out of context before retention of earlier learned information is known to interfere with memory performance through a nongenomic pathway (de Quervain et al., 1998; Sajadi et al., 2006). This effect was shown to involve a nongenomic MR component (Khaksari et al., 2007) (Fig. 7), although interaction with other stress hormones is necessary to accomplish the full effect. The rapid depotentiation of LTP and shift toward LTD reported when animals are exposed to a (stressful) novel environment possibly plays a role in this phenomenon (Xu et al., 1997; Manahan-Vaughan and Braunewell, 1999; Yang et al., 2006; Hirata et al., 2009), although there is evidence that this is GR- rather than MR-dependent (Xu et al., 1998).
Two recent studies give insight into how glucocorticoids may promote consolidation of information in an intermediate time domain (approximately 20 min after stress), through protein-protein interactions. Thus, in the dentate gyrus acute stress starts a cascade in which ERK1/2 is phosphorylated, leading within 15 min to activation of the nuclear kinases MSK (mitogen- and stress-activated kinase)-1 and Elk-1 (Gutièrrez-Mecinas et al., 2011). This in turn results in histone acetylation and induction of the immediate early genes c-fos and Egr-1. Histone acetylation via nongenomic actions of GR on phosphorylated cAMP response element-binding protein in the insular cortex was found to be important for object recognition, whereas such a process in the hippocampus plays a role in object location memory in the hippocampus (Roozendaal et al., 2010).
III. Delayed Effects
Although rapid nongenomic corticosteroid actions have been increasingly acknowledged over the past decade, many more studies have revealed hormone actions on neural activity in a much slower time domain, fitting the gene-mediated signaling pathway of nuclear receptors. Areas that have been studied most intensively include the hippocampus, BLA, mPFC, and the ventral tegmental area (Supplemental Table II).
A. Slow Modulation of Neural Activity
1. Hippocampus.
Corticosteroids generally do not change passive or active membrane properties of CA1 pyramidal neurons, such as resting membrane potential, input resistance or characteristics of the action potential, at least not in tissue from ADX animals (e.g., Joëls and de Kloet, 1989; Kerr et al., 1989). One study in ADX animals reported a higher membrane time constant when animals received a low dose of corticosterone (mimicking a condition of predominant MR activation) for several weeks (Beckford et al., 1996). More recently, a clear difference between the ventral-most (20%) and remaining part of the hippocampus was observed with respect to corticosteroid actions (Maggio and Segal, 2009a). In the ventral-most part, corticosterone lowered the threshold for action potential generation, resulting in higher excitability; in the dorsal hippocampus, corticosterone reduced the input resistance and membrane time constant, making cells leakier.
A second category of functional targets concerns the voltage-dependent currents. Sodium and potassium currents were not much affected by the hormone (Karst et al., 1993; Werkman et al., 1997). However, voltage-dependent calcium currents form a major target for corticosteroid hormones. In particular, the amplitude of sustained high voltage-activated calcium currents was found to be enhanced by glucocorticoid treatment compared with the situation in which predominantly MRs are activated (Kerr et al., 1992; Karst et al., 1994). The enhancement in calcium current amplitude requires protein synthesis and DNA-binding of GR homodimers (Kerr et al., 1992; Karst et al., 2000). Other characteristics of the current, such as voltage dependence or kinetics, were unaltered. The enhanced calcium current amplitude was also observed several hours after stress exposure (Joëls et al., 2003). Through pharmacological isolation of the various current types, it was demonstrated that L- rather than N-type calcium currents are a target for glucocorticoids (Chameau et al., 2007). It is noteworthy that in the absence of corticosteroid hormones (or with reduced levels of GRs), sustained calcium current amplitude was also high (Karst et al., 1994; Hesen et al., 1996), revealing a U-shaped dose dependence (Joëls, 2006). A similar U-shaped dose dependence was also described for CA3 neurons, when correlating circulating corticosteroid levels throughout the day with the calcium current amplitude (Kole et al., 2001). By contrast, dentate granule cells do not show an enhanced calcium current amplitude several hours after they have been exposed to a pulse of corticosterone (van Gemert et al., 2009), although MR activation was shown to be necessary to restrain calcium influx in ADX animals (Karst and Joëls, 2001).
What happens when hippocampal cells receive amino acid-mediated synaptic input? This has been examined in three ways: 1) by exposing cells to a depolarizing step, to mimic a steady excitatory input; 2) by examining specific glutamatergic pathways; and 3) by studying GABAergic inputs. The results of these experiments are described below. Corticosteroid actions on other neurotransmitters/modulators, such as noradrenaline, are only briefly addressed in section VI.C.
When CA1 neurons are depolarized, they fire action potentials, but the frequency gradually accommodates during depolarization (Fig. 8). This is caused by activation of a slow calcium-dependent potassium current (IsAHP). Upon termination of the depolarization, the IsAHP is slowly deactivated, which results in a lingering afterhyperpolarization (AHP). Both phenomena (i.e., firing frequency accommodation and AHP) will not affect the likelihood of evoking a single spike upon depolarization but reduce the transfer of multiple spikes. High levels of corticosterone or glucocorticoids were found to enhance the amplitude of the IsAHP and AHP in CA1 pyramidal neurons, resulting in fewer spikes upon depolarization (Joëls and de Kloet, 1989; Kerr et al., 1989; Liebmann et al., 2008) (Fig. 8); this differs from the situation of predominant MR activation (such as occurs under rest), which is characterized by more spikes during a depolarizing pulse than in the absence of steroids (Joëls and de Kloet, 1990), reflecting the same U-shaped dose dependence as observed with respect to calcium current amplitude. Similar effects were also seen in the CA3 area (Kole et al., 2001). The onset of the corticosterone-dependent modulation in firing frequency is slow and requires protein synthesis (Karst and Joëls, 1991). The ventral-most part of the hippocampus reacts differently to a high dose of corticosterone than the dorsal part, showing reduced firing frequency accommodation and more spikes upon depolarization (Maggio and Segal, 2009a) (Fig. 8).
At the level of single synapses, corticosteroids change glutamatergic transmission in a manner sharing characteristics of long-term potentiation. In CA1 and cultured hippocampal neurons, a pulse of corticosteroids enhanced the amplitude but not frequency of mEPSCs recorded several hours after corticosteroid exposure, via GRs (Karst and Joëls, 2005; Martin et al., 2009) (Fig. 9). This concurs with a slow GR-dependent increase in surface expression of GluA2 subunits, a process requiring protein synthesis and occluding chemically induced LTP (Groc et al., 2008; Martin et al., 2009). The effects on mEPSC amplitude peaked between 150 and 200 min and were not seen earlier than 1 h after starting with corticosterone delivery (Karst and Joëls, 2005; Groc et al., 2008; Martin et al., 2009). It is noteworthy that inhibitory signals (i.e., sIPSC amplitude) are also enhanced in the dorsal hippocampus via GRs (Maggio and Segal, 2009a); in the ventral hippocampus, an MR-dependent reduction in sIPSC frequency was reported. The effects on sIPSCs started 25 min after onset of corticosterone administration, peaking at 55 min. This is in a much more rapid time domain than the effects on excitatory transmission.
Synaptically evoked responses recorded extracellularly in the various hippocampal areas were usually not affected by corticosterone or stress (Pavlides et al., 1996; Bramham et al., 1998; Zhou et al., 2000; Yamada et al., 2003; Chen et al., 2010), although enhancements (Avital et al., 2006; Kavushansky et al., 2006) or reduced activity (Hirata et al., 2008) were reported in a few studies. Corticosteroid actions on excitatory or inhibitory transmission are restricted to a limited number of synapses and thus not discernible at a more general level, similar to what has been found after learning (Whitlock et al., 2006). It may also relate to the dose of corticosterone that was used or the intensity of the stressor. This is suggested by a study of Rey et al. (1987), showing that low doses of corticosterone enhance the amplitude of the population spike evoked by synaptic stimulation in the CA1 area, whereas high doses decreased the population spike.
2. Basolateral Amygdala.
Slow corticosteroid actions on principal cells in the BLA resemble the responses in ventral rather than dorsal hippocampal CA1 neurons. Thus, a brief pulse of corticosterone resulted some hours later in an increased input resistance and more depolarized membrane potential (Duvarci and Paré, 2007). This was seen only in a subpopulation of neurons with very high input resistance, not in cells with a lower resistance (Duvarci and Paré, 2007; Liebmann et al., 2008). In BLA neurons, firing frequency accommodation and AHP amplitude were reduced or unaffected by corticosterone, whereas dorsal CA1 neurons measured in the same study showed a clear enhancement (Duvarci and Paré, 2007; Liebmann et al., 2008) (Fig. 8). It was argued that low expression of α1.3 calcium channel subunits in the BLA contributes to the lack of modulation in IsAHP (Liebmann et al., 2008), despite a clear GR-dependent increase in sustained high voltage-activated calcium currents (Karst et al., 2002). Corticosterone furthermore shifted the reversal potential of GABA receptor-linked chloride channels to more depolarized potentials, causing a reduced inhibitory postsynaptic potential amplitude with synaptic stimulation. Overall, these changes are expected to cause a slow enhancement in excitability after a single pulse of corticosterone. This was indeed observed at the field potential level after elevated platform stress or corticosterone injection, both in vivo and in vitro (Kavushansky and Richter-Levin, 2006; Kavushansky et al., 2006), although not 24 h after restraint stress (Rodríguez Manzanaers et al., 2005).
3. Medial Prefrontal Cortex.
Glutamatergic transmission in layer V prelimbic neurons also seems to be slowly enhanced by corticosterone and stress, similar to what has been described in the CA1 area (Fig. 9). Thus, in cells recorded >2 h after exposure to stress, the amplitude of EPSCs via AMPA or NMDA receptors was increased (Yuen et al., 2009, 2011; Liu et al., 2010). In parallel, the mEPSC amplitude but not frequency was enhanced, as was the surface expression of AMPA and NMDA receptor subunits. The slow corticosteroid effects on glutamatergic transmission are mediated by GRs.
A high dose of corticosterone (presumably activating both MR and GR) reduced mIPSC frequency but not amplitude in layer V prelimbic neurons. In view of the enhanced paired pulse depression, this effect seems to be caused by a reduction in GABA release mediated by endocannabinoids (Hill et al., 2011).
4. Ventral Tegmental Area.
Several studies show increased firing rates of dopaminergic neurons and LTP-like changes after corticosterone administration or stress exposure. Low levels of corticosterone were reported to enhance basal firing rate and glutamate-induced burst firing of presumed dopaminergic cells (Overton et al., 1996). Although this would argue for involvement of MRs, Cho and Little (1999) found effects on NMDA or AMPA induced cell firing that could be blocked with the GR-antagonist mifepristone. Restraint stress (Anstrom and Woodward, 2005; Anstrom et al., 2009) and foot shock (Brischoux et al., 2009) also increased burst firing, the latter only in the ventral tegmental area; inhibition after foot shock was observed in dorsal cells. It is noteworthy that the effect of restraint stress persisted for >24 h. This may explain why Saal et al. (2003) observed an enhanced AMPA/NMDA ratio 24 h after forced swim stress, an effect that was blocked by mifepristone and required GluA1 subunits to develop (Dong et al., 2004).
B. Slow Modulation of Synaptic Plasticity
Numerous studies have shown that the induction of LTP in the CA1 hippocampal area is severely hampered several hours after administration of corticosterone in vitro or in vivo or after exposure to stress, especially stress of an uncontrollable nature (Foy et al., 1987; Shors et al., 1989; Diamond et al., 1992; Shors and Thompson, 1992; Diamond and Rose, 1994; Kim et al., 1996, 2001; Zhou et al., 2000; Xiong et al., 2004; Kavushansky et al., 2006; Hirata et al., 2008, 2009; Li et al., 2008; Yang et al., 2008a; Cazakoff and Howland, 2010; Ryan et al., 2010; Ooishi et al., 2012; for review, see Kim and Diamond, 2002) (Supplemental Table II). This is a GR-dependent phenomenon (Pavlides et al., 1996); depends on NMDA-receptor mediated transmission (Kim et al., 1996); depends on phosphatidylinositol 3-kinase (Yang et al., 2008b); requires the basolateral amygdala, particularly ERK1/2 phosphorylation (Kim et al., 2001; Yang et al., 2008a); and can be rescued by estradiol administration in vitro (Ooishi et al., 2012). Enhanced LTP was reported for selective MR activation (Pavlides et al., 1996), particularly in the ventral-most part of the CA1 hippocampal area (Maggio and Segal, 2007, 2009b). GR activation not only impairs LTP induction but also slowly promotes LTD (Xiong et al., 2004; Yang et al., 2004, 2006; Kim et al., 2006; Wong et al., 2007; Gao et al., 2008; Niehusmann et al., 2010). Spillover of glutamate as a result of inadequate reuptake, thereby activating extrasynaptic GluN2B subunits, seems crucial (Yang et al., 2005; Wong et al., 2007). Stress also enhances a metabotropic glutamate receptor-dependent type of LTD (Chaouloff et al., 2007).
Reduced LTP after stress via GRs was also observed for the mossy fiber projection to the hippocampal CA3 region, requiring the conversion of cAMP into adenosine (Chen et al., 2010). Likewise, stress was found to impair LTP in the projection from the ventral subiculum or basolateral amygdala to the prelimbic area, again involving GRs (Maroun and Richter-Levin, 2003; Rocher et al., 2004; Mailliet et al., 2008; Qi et al., 2009; Richter-Levin and Maroun, 2010). The latter was not seen when the BLA had already been stimulated earlier or when animals had recently experienced stress (Richter-Levin and Maroun, 2010), another example of metaplasticity. In contrast to these findings, most studies agree that stress facilitates LTP in projections to the (baso)lateral amygdala, originating in the entorhinal cortex, external capsule or prelimbic cortex (Vouimba et al., 2004; Rodríguez Manzanares et al., 2005; Kavushansky et al., 2006; Maroun, 2006; but see Kavushansky and Richter-Levin, 2006).
Results in the dentate gyrus have been somewhat ambiguous. There is evidence that LTP is enhanced via MR activation and reduced or even turned into LTD after GR activation (Pavlides et al., 1993, 1994, 1995; Smriga et al., 1996; Yamada et al., 2003; Avital et al., 2006; Kavushansky et al., 2006; Vouimba et al., 2007; Yarom et al., 2008), similar to what has been described for the CA1 area. Abrari et al. (2009) reported facilitated induction of LTP in association with improved memory when rats were treated with a moderately high dose of corticosterone directly after training in a fear conditioning paradigm. However, others found no change after inescapable or cold stress (Shors and Dryver, 1994; Vouimba et al., 2004) or could not link stress-induced changes in LTP/LTD to known corticosteroid receptor types (Spyrka et al., 2011). Stress exposure after LTP induction has also been studied in the dentate gyrus. Low-stress conditions such as handling of the animal were found to impair weak LTP, whereas exposure to a high-stress condition (forced swim) prolonged the late phase of LTP via an MR-dependent process (Korz and Frey, 2003). The latter required an intact BLA (Korz and Frey, 2005). The relevance of the BLA for LTP in the dentate was also evident from work by Akirav and Richter-Levin (1999, 2002). They showed that stimulation of the BLA in the short term facilitates LTP in the perforant path-dentate projection but in the long-term impairs LTP. The latter was mimicked by stress exposure, and stress interfered with BLA-induced facilitation of LTP in the dentate gyrus. The modulatory effects of BLA stimulation on LTP in the dentate gyrus depends on noradrenergic and corticosteroid actions in the BLA.
C. Underlying Mechanisms and Functional Consequences
One striking observation is that corticosterone via GRs slowly promotes glutamatergic transmission (at least in part of the synapses), involving phosphorylation of specific AMPA-receptor subunits (Caudal et al., 2010). As a consequence, mEPSC amplitude is enhanced in both the hippocampus and mPFC (Karst and Joëls, 2005; Yuen et al., 2009), and surface expression of GluA2 subunits increased (Groc et al., 2008; Martin et al., 2009; see Krugers et al., 2010); stimulation of glutamate release via a presynaptic SNARE-dependent mechanism (Musazzi et al., 2010; for review, see Popoli et al., 2012) may also contribute to the overall higher glutamatergic tone, but whether these effects are due to a genomic action has not been proven yet. The gradually developing enhanced glutamatergic transmission after stress shares properties with those seen after LTP (Ho et al., 2011). This would be in line with the view that stress induces LTP-like phenomena (Shors and Dryver, 1994; Saal et al., 2003; Diamond et al., 2004; Huang et al., 2005; Karst and Joëls, 2005; Groc et al., 2008), partly impinging on the same signaling pathways, thus leading to occlusion (Groc et al., 2008). Prior activation of these pathways would raise the threshold for subsequent LTP induction and facilitate induction of LTD, which was indeed consistently observed in nearly all brain regions after stress; notable exceptions are the ventral-most part of the hippocampus and the BLA. This follows the principles of metaplasticity as it was proposed for LTP (Bear and Abraham, 1996). Metaplasticity in relation to stress was indeed mentioned over a decade ago (Kim and Yoon, 1998), and experimental evidence for metaplasticity is accumulating, both in the rapid time domain (Karst et al., 2010) and in the genomic window (Richter-Levin and Maroun, 2010).
Exactly how these slow GR-dependent effects on glutamatergic transmission develop is not fully understood. In the mPFC, evidence was supplied for a role of the serum and glucocorticoid-inducible kinase SGK, which subsequently regulates the activity of Rab4, a small GTPase, controlling recycling of, for example, GluA2 subunits between the endosome and plasma membrane (Liu et al., 2010; Yuen et al., 2011). An interesting analogy can be found in the formation of fear memory. A recent article described that a single fear stimulus promotes GluA2-containing AMPA receptors in mouse cerebellar stellate cells via noradrenaline (Liu et al., 2010). The subsequent rise in intracellular calcium and hence activation of the calcium-sensitive ERK/mitogen-activated protein kinase signaling pathway triggered new GluA2 gene transcription over the course of hours and thus a more permanent shift toward GluA2-containing receptors in the membrane. Although the specific properties of this process clearly differ from what is seen with stress (i.e., different area, no obvious transcriptional regulation of GluA2 by stress in the hippocampus), it nevertheless shows that early events may trigger a cascade developing into more lasting changes. Along these lines, rapid effects of corticosterone could lead to more glutamate release and associated postsynaptic changes, which subsequently through a slower GR-dependent mechanism promote synaptic localization of GluA2 subunits. In this model, those synapses activated during the stressful event and exposed to elevated levels of stress hormones will be strengthened for a considerable time, raising the threshold for subsequent synaptic strengthening. This could protect strengthened synapses from retrograde interference and promote higher cognitive processes in the PFC and dorsal hippocampus such as consolidation of relevant contextual information. There is indeed abundant evidence that stress/glucocorticoids either directly (Oitzl and de Kloet, 1992; Oitzl et al., 2001) or conditionally (Roozendaal et al., 2004) promote consolidation of spatial information, as well as exert beneficial effects on working memory (Yuen et al., 2009), via a GR-dependent mechanism (Fig. 10).
The strengthening of certain synapses does not imply that all excitatory information reaching hippocampal or mPFC cells some hours after stress is facilitated; the contrary is true. For instance, a larger spike frequency accommodation and AHP amplitude (Joëls and de Kloet, 1989; Kerr et al., 1989; Kole et al., 2001) could serve as a brake to steady excitatory inputs, particularly when unrelated to the stressful event. The increased AHP amplitude probably develops secondary to enhanced calcium influx. The latter pathway is one of the best resolved examples how GR activation can change neuronal function: GR-homodimers bind to DNA of responsive genes (Karst et al., 2000) and transcriptionally up-regulate many genes (Morsink et al., 2006b), including the β4 calcium channel subunit (Chameau et al., 2007; van Gemert et al., 2009), resulting in higher β4 protein levels (van Gemert et al., 2009) and a higher availability of L-type calcium channels in the membrane (Chameau et al., 2007).
The dorsal CA1 and CA3 hippocampal regions as well as the prelimbic PFC seem to respond in a very comparable manner to stress or corticosterone acting via GRs. This clearly differs from the ventral-most CA1 region, BLA, and, to a lesser degree, the dentate gyrus. In the BLA, GR-dependent pathways maintain the rapid increase in mEPSC frequency induced by corticosterone; in the ventral hippocampus, sIPSC frequency is reduced. In both regions, LTP is enhanced by stress or corticosterone, and mechanisms to restrain responses to excitatory transmission, such as firing frequency accommodation, are not in place. This suggests that behavioral (e.g., emotional) aspects involving activation of these regions have a prolonged window in time for encoding. This could contribute to the fact that cortisol generally promotes the memory of emotional rather than neutral aspects of information (Buchanan and Lovallo, 2001; Kuhlmann and Wolf, 2006; van Stegeren et al., 2010; but see Abercrombie et al., 2003 and Rimmele et al., 2003, who report effects on neutral information).
Delayed effects of stress or corticosterone administration are generally mediated by GRs. Notable exceptions are the excitatory effects observed in the ventral-most part of the CA1 hippocampal area (Maggio and Segal, 2007, 2009a,b), and facilitating effects on LTP (be it induction or maintenance of late-phase LTP) in the dentate gyrus (Pavlides et al., 1993, 1994; Korz and Frey, 2003, 2005; Avital et al., 2006). Other than that, the importance of intracellular MR occupation is only revealed against an ADX background. Under these conditions, actions via intracellular MRs are important for maintenance of excitability and confining calcium influx (Karst and Joëls, 2001), most likely contributing to a viable state of the cells. Most MR-mediated actions are opposite those exerted via GRs, with regard not only to the slow gene-mediated processes but also to rapid nongenomic actions (Karst et al., 2010).
IV. Chronic Corticosteroid Exposure
Although brief exposure to stress—as discussed in sections II and III—usually serves an adaptive role, chronic exposure to stress or high levels of corticosterone is considered to be less beneficial. In humans, prolonged episodes of stress—particularly when experienced as uncontrollable and unpredictable—have been reported to increase the vulnerability to all kind of diseases in the genetically predisposed, including psychiatric and neurological illnesses (see section VI.D).
Many animal models have been developed to examine the effects of chronic stress in more detail under very controlled experimental conditions. In these models, animals are usually exposed to 3 weeks of restraining once per day or a mixture of mild unpredictable physical or psychological stressors (Supplemental Table III). A model that is almost exclusively based on psychological stressors involves exposure of animals to repetitive social defeat, sometimes in combination with isolated housing. Some studies have made use of repeated corticosterone injections, to probe the contribution of corticosteroids in comparison to other stress mediators. Because these models differ from each other in many respects, including the neuroendocrine properties, such as HPA-axis reactivity, we have categorized the available literature in Supplemental Table III primarily on the basis of the paradigm that was used.
Many studies demonstrated changes in dendritic complexity after exposure to chronic stress and/or overexposure to corticosteroid hormones, regardless of the paradigm that was used (for reviews, see McEwen and Magarinos, 1997; Fuchs and Bode, 2006; Holmes and Wellman, 2009; Roozendaal et al., 2009; Shansky and Morrison, 2009). In most areas (e.g., CA3 or infralimbic cortical pyramidal neurons), dendritic complexity was found to be reduced. However, increased dendritic complexity has also been reported in principal cells of the BLA and the orbitofrontal cortex (for review, see Roozendaal et al., 2009). Moreover, there is evidence that chronic exposure to high levels of stress hormones affects spine density. In general, spine density is increased by chronic stress in those areas where dendritic complexity is also increased (e.g., in the basolateral amygdala; Mitra et al., 2005) and decreased if dendritic complexity is decreased (e.g., in neurons of the PFC) (Li et al., 2011). Besides affecting dendritic complexity, chronic stress has been reported to suppress proliferation, differentiation and survival of progenitor cells in the dentate gyrus (for reviews, see Joëls et al., 2007; Leuner and Gould, 2010; Lucassen et al., 2010), which may or may not be linked to dendritic remodeling in the dentate gyrus (Bessa et al., 2009). Moreover, the expression levels of many molecules involved in neurotransmission—including subunits for glutamate or GABA receptors—are modulated by chronic stress (for review, see McEwen et al., 2007).
All of these changes are expected to affect neural activity. Overall, this led to a theory proposing that chronic stress imposes a condition of increased excitability, at least in the CA3 hippocampal area, which, through excitotoxicity, may lead to dendritic atrophy (McEwen, 1999). The latter could then be interpreted as a “protective” mechanism by which cells, through reduction in the number of their synaptic contacts, would restrain the enhanced excitatory input. In support of this theory, treatment of animals with, for example, NMDA receptor blockers was found to prevent dendritic remodeling in the CA3 area and mPFC (McEwen and Magarinos, 1997; Christian et al., 2011; Li et al., 2011).
According to this theory, enhanced excitatory transmission would precede dendritic retraction rather than occur simultaneously or as a consequence. However, neither experimental nor mathematical evidence seems to support this. In a biologically realistic computational model, Narayanan and Chattarji (2010) demonstrated that dendritic retraction increases input resistance, which is directly translated into higher spiking frequencies in response to both somatic current injections and synaptic inputs. It is noteworthy that studies examining basal neuronal properties or synaptic responses after chronic stress very often find enhanced activity (Supplemental Table III). Thus, in the hippocampus, chronic stress was found to enhance NMDA receptor-mediated EPSCs in CA3 neurons (whereas intrinsic properties were largely unchanged) (Kole et al., 2002, 2004a,b), to reduce the threshold for induction of fEPSPs in the CA1 area (Kerr et al., 1991), and to increase the amplitude of AMPA receptor-mediated synaptic currents in DG granule cells, the latter in combination with only acute exposure to corticosterone (Karst and Joëls, 2003) (Fig. 11). In the dorsomedial PFC, cell-firing activity is enhanced starting 4 days after stress onset, although a reduction is seen in the ventromedial PFC some time later (Lee et al., 2011); in agreement with the latter, 5 to 7 days of restraint or unpredictable stress in young (4 week-old) rats was reported to cause reduction of both AMPA- and NMDA-receptor mediated synaptic responses in pyramidal PFC cells, in association with ubiquitin/proteasome-mediated degradation of GlA1 and NR1 subunits (Yuen et al., 2012). In the nucleus accumbens, brief swim stress enhanced AMPA receptor-mediated mEPSC amplitude (Campioni et al., 2009), whereas social defeat caused increased mEPSC frequency in medium spiny neurons (Christoffel et al., 2011). Increased basal firing and reduced afterhyperpolarization was observed in the locus ceruleus after cold stress (Mana and Grace, 1997; Jedema and Grace, 2003). Several studies reported altered GABAergic activity after chronic stress, indirectly causing enhanced excitability in the dentate gyrus (Holm et al., 2011), anterior cingulate cortex (Ito et al., 2010), nucleus accumbens (Wang et al., 2010), and PVN (Verkuyl et al., 2004; Wamsteeker et al., 2010). Nevertheless, some studies reported reduced basal or synaptic activity after chronic stress (Hu et al., 2010; Li et al., 2011; Quan et al., 2011a,b; Zhang et al., 2011). Most studies using field potential recordings found no change in basal synaptic responses (Bodnoff et al., 1995; Gerges et al., 2001; Pavlides et al., 2002; Alfarez et al., 2003; Aleisa et al., 2006a,b,c; Kessal et al., 2006; Krugers et al., 2006; Holderbach et al., 2007; Goldwater et al., 2009; Srivareerat et al., 2009; but Dumas et al., 2010; Tran et al., 2011), possibly because the effects are too subtle to be noticed extracellularly or depend on GABAergic transmission (which is also not well captured with field potential recordings). Overall, nearly all studies demonstrate that several weeks of stress either enhance intrinsic/synaptically evoked cell activity or are ineffective in this respect, regardless of the area under investigation.
This contrasts with the effects of chronic stress on synaptic plasticity. Irrespective of the models that have been used, chronic overexposure to stress hormones has generally been found to reduce the ability to induce or maintain LTP and to enhance the likelihood to induce LTD, even when circulating corticosteroid levels at the time of recording were very low. This was observed for projections to the (dorsal) hippocampal CA1 area (Bodnoff et al., 1995; Gerges et al., 2001; Von Frijtag et al., 2001; Alfarez et al., 2003; Zheng et al., 2004; Aleisa et al., 2006a,b,c; Artola et al., 2006; Krugers et al., 2006; Holderbach et al., 2007; Ma et al., 2007; Srivareerat et al., 2009; Kamal et al., 2010; Sterlemann et al., 2010; Tran et al., 2011), the CA3 area (Kole et al., 2002, 2004; Pavlides et al., 2002; Maggio and Segal, 2011), the PFC (Cerqueira et al., 2007; Goldwater et al., 2009; Lee and Goto, 2011; Lee et al., 2011b; Quan et al., 2011a,b; Zhang et al., 2011), and the bed nucleus stria terminalis (Conrad et al., 2011). The reverse, however, was seen in the ventral-most part of the hippocampus (Maggio and Segal, 2011). The effects in the DG were variable (Gerges et al., 2001; Pavlides et al., 2002; Alfarez et al., 2003; Aleisa et al., 2006a,b,c; Dumas et al., 2010; Spyrka and Hess, 2010), and reduced LTD was observed in the bed nucleus stria terminalis (McElligott et al., 2010) and nucleus accumbens (Wang et al., 2010).
How these chronic stress-induced changes in basal transmission and synaptic plasticity develop is still far from being resolved, in part because it is very difficult to monitor cellular changes over the course of weeks. Some of the above-mentioned studies demonstrated that particular molecules are critical for the electrophysiological changes to occur, such as calcineurin and Ca2+/calmodulin-dependent protein kinase II (Aleisa et al., 2006a,b,c), NR2B subunits, inhibitor of nuclear factor-κB (Christoffel et al., 2011), and brain-derived neurotrophic factor (Zhou et al., 2000; Radecki et al., 2005; Aleisa et al., 2006c). Compounds that prevented or normalized the development of changes in neural activity include nicotine (Aleisa et al., 2006a,b,c), antidepressants (Von Frijtag et al., 2001; Kole et al., 2002, 2004; Kessal et al., 2006; Holderbach et al., 2007; Wang et al., 2010; Holm et al., 2011), ketamine (Li et al., 2011), memantine (Quan et al., 2011), antiglucocorticoids (Krugers et al., 2006; Karst and Joëls, 2007; Spyrka and Hess, 2010), and phenytoin (Zheng et al., 2004), whereas β-amyloid exacerbated the effects of chronic stress (Srivareerat et al., 2009; Tran et al., 2011). However, it is unclear whether these compounds 1) directly interfere with chronic-stress-dependent signaling pathways, 2) tap indirectly into the same pathways, or 3) compensate for/counteract the effects of chronic stress through independent mechanisms.
Clearly, investigation of the development of electrical activity over time is called for, but this has rarely been done (e.g., Lee et al., 2011). A nice exception is a study by Spyrka and Hess (2010), who examined LTP in the dentate gyrus at 1, 3, 7, 14, and 21 days of restraint (neck) stress. They observed that the induction of LTP was reduced after 3 and 7 days of restraint, but enhanced after 14 or 21 days, through a GR- and MR-dependent mechanism, respectively. Although this time course may be specific for the dentate gyrus, the study illustrates that much information may be lost when investigating animals only at the end of a period of 3 or more weeks. It is noteworthy that the effects of chronic stress on synaptic plasticity under basal—i.e., unstressed—conditions (e.g., in the hippocampal CA1 area) are quite similar to what is seen in naive animals shortly after an acute period of stress. The signaling pathways involved in the latter may give insight in how synaptic plasticity may become dysregulated on a more permanent basis after chronic stress.
Another aspect that has rarely been addressed is the importance of the circulating corticosteroid levels just before and during recording. For instance, the amplitude of L-type calcium currents in CA1 pyramidal neurons is enhanced after chronic stress compared with control neurons when circulating corticosteroid levels during and several hours before the moment of investigation are low (Karst and Joëls, 2007). However, the current amplitude is low in the stressed compared with control group when measuring 1 to 4 h after a stress-like surge of corticosterone. In dentate granule cells, AMPA receptor-mediated synaptic responses and calcium-current amplitude were increased in chronically stressed animals only in combination with a brief exposure to high levels of corticosterone 1 to 4 h before recording (Karst and Joëls, 2003; van Gemert and Joëls, 2006) (Fig. 11). This indicates that the influence of chronic stress sometimes will become apparent only after the HPA axis has been acutely activated. The extent to which this applies to the articles summarized in Supplemental Table III is unclear, because the issue was usually not specifically addressed. Tissue may have been prepared from animals transported in an unfamiliar cage (causing stress due to novelty) or sacrificed at the peak of the circadian cycle, which is the case, for example, when the animals are maintained on a reversed day-night rhythm (Hu et al., 2010). Regardless, the data indicate that the responsiveness to corticosterone may change depending on the prior history of chronic stress, a phenomenon also observed when animals have been exposed to stress early in life (see section V).
What are the consequences of all these changes in electrical activity after chronic stress for brain function in general? It is a big step from electrophysiological phenomena (frequently examined in vitro) to the systems level in vivo. By concentrating on electrical activity only, the influence of many other important factors is disregarded. For instance, chronic stress will certainly affect the synthesis, release, and reuptake of essential transmitters such as glutamate and GABA (see, for instance, Grønli et al., 2007; Elizalde et al., 2010; Tordera et al., 2011). The bioavailability of these transmitters is as important for their overall effect as the processes downstream of receptor activation that eventually cause the changes in electrical activity measured with electrophysiological methods.
With respect to global consequences, chronic stress-induced changes in neural activity have in some instances been directly linked to the function of the HPA axis itself, supporting disinhibition or impaired negative feedback of the HPA axis after chronic stress (Verkuyl et al., 2004; Wamsteeker et al., 2010). Although the consequences of chronic stress for various behavioral tasks have been addressed in detail in numerous studies (for reviews, see Shors, 2006; Conrad, 2010; Marin et al., 2011), we here highlight only those studies combining electrophysiological measurements with behavioral performance. These showed that chronic stress reduces hippocampal synaptic plasticity in parallel with impaired spatial memory, including the reversal learning aspect (Bodnoff et al., 1995; Zheng et al., 2004; Aleisa et al., 2006a; Ma et al., 2007; Quan et al., 2011a,b), even many months after stress exposure (Sterlemann et al., 2010). Chronic stress also exacerbates cognitive decline due to amyloid-β treatment (Srivareerat et al., 2009; Tran et al., 2011). Furthermore, working memory (Cerqueira et al., 2007) and reward anticipatory behavior (Kamal et al., 2010) were found to be disturbed. However, this selection of articles gives only a glimpse of the complexity associated with cognitive outcome after chronic stress, which seems to be task-specific (Conrad, 2010). For instance, chronic stress impairs spatial learning on appetitively motivated tasks, such as the radial arm maze or holeboard, tasks that evoke relatively mild to low arousal components from fear; but under testing conditions that evoke moderate to strong arousal components from fear, such as during radial arm water-maze testing, chronic stress seems to have minimal impairing effects or may even facilitate spatial learning. Moreover, the effects in female animals may differ from those observed in male animals (Luine, 2002). These behaviorally relevant issues have not been addressed so far with electrophysiological approaches.
Overall, chronic stress has profound effects on electrical activity in the brain. In general, the ability to induce LTP is impaired and glutamatergic activity is increased, although the latter is less consistent than the former. Evidently, these changes will have consequences for behavioral functions that critically depend on synaptic transmission and synaptic strengthening. In accordance, chronically stressed animals do perform differently in many behavioral tasks. However, the step from altered electrical activity to disturbed behavior after chronic stress is still heavily understudied.
V. Perinatal Stress
Stress during the perinatal period can have powerful and long-lasting consequences for brain function (Pryce et al., 2005; Fenoglio et al., 2006; Sullivan et al., 2006; Maccari and Morley-Fletcher, 2007; Schmidt, 2010; Zhang and Meaney, 2010). Corticosteroids may exert their influence by changing connections in the brain at a time that these are being established and pruned (e.g., Sullivan et al., 2006), thus affecting brain organization in a similar way, as has been recognized for decades already in the case of gonadal hormones. A special case of such developmental influences pertains to the circuit involved in the stress response itself. By targeting this circuit, those exposed to perinatal stress might respond differently to stress later in life than naive control subjects. The lasting changes in stress responsiveness refer not only to, for example, activation of the HPA-axis itself but also to other pathways downstream of brain corticosteroid receptors, which in part depend on (intra)cellular proteins.
An extensive number of studies reported that perinatal stress changes synaptic properties in adulthood, both under “basal” conditions (i.e., when animals are not stressed) and when brain tissue of adult animals is exposed to a surge of corticosteroid hormones (for review, see also Ali et al., 2011) (Supplemental Table IV). How perinatal stress can induce such lasting effects on excitability has not yet been addressed. From what is presently known about other brain properties, though, it seems likely that epigenetic programming plays an important role (Meaney and Szyf, 2005; Franklin and Mansuy, 2010). Potential molecules mediating changes in brain function include the NMDA receptor (Kamphuis et al., 2003; Son et al., 2006; Yaka et al., 2007; Ryan et al., 2009; Judo et al., 2010; but see Lee et al., 2011a), AMPA receptors (Yaka et al., 2007; Ryan et al., 2009), cell adhesion molecules (Aisa et al., 2009), and brain-derived neurotrophic factor (van Hasselt et al., 2012a; Yeh et al., 2012). We will here discuss only some of the major variables and the principles that emerge from the currently available data set on electrophysiological effects after early-life stress (summarized in Supplemental Table IV).
The first principle is that rather mild disturbances early in life (such as 3-min novelty exposure for the first 3 postnatal weeks or pups being separated from the mother up to 1 h per day) generally seem to enhance excitability, for instance by facilitating LTP in the CA1 area or enhancing spontaneous firing in the PFC (Kehoe and Bronzino, 1999; Zou et al., 2001; Tang and Zou, 2002; Akers et al., 2006; Stevenson et al., 2008); there are, however, some indications for reduced activity as well (Hsu et al., 2003; Blaise et al., 2008). More severe conditions, though, are consistently linked with reduced synaptic plasticity. This is true for prenatal stress (Kamphuis et al., 2003; Son et al., 2006; Yang et al., 2006, 2007; Lee et al., 2011a; Yeh et al., 2012; but see Noorlander et al., 2008), low and/or fragmented maternal care including situations associated with impoverished environmental conditions (Bredy et al., 2003; Brunson et al., 2005; Cui et al., 2006; Champagne et al., 2008; Bagot et al., 2009, 2012; Ivy et al., 2010), prolonged maternal separation (Gruss et al., 2008; Stevenson et al., 2008; Oomen et al., 2010), or traumatic events (Akers et al., 2006; Judo et al., 2010). When investigated, basal synaptic properties or paired pulse responsiveness were not much altered (Domenici et al., 1996; Kamphuis et al., 2003; Yaka et al., 2007; Lee et al., 2011). This pattern seems more pronounced in male than in female offspring (Kehoe and Bronzino, 1999; Oomen et al., 2009 2010; van Hasselt et al., 2012a,b).
A critical factor in the changes in neuronal function is the quality and amount of maternal care. For instance, in the dentate gyrus (but less so in the CA1 area) the amount of licking and grooming received from the mother during the first postnatal week strongly predicts the ability to induce LTP during young adulthood (van Hasselt et al., 2012a,b; see Fig. 12). Another critical factor is the early overexposure to glucocorticoids, because postnatal glucocorticoid treatment either of the lactating dam or the pups results in a very similar phenotype, reducing LTP (Domenici et al., 1996; Lin et al., 2006) while enhancing LTD (Lin et al., 2006). However, other stress-mediators such as CRH are certainly also involved (Ivy et al., 2010; Wang et al., 2011). Because maternal care indirectly affects the circulating levels of stress hormones (Zhang and Meaney, 2010), these two factors are most likely interconnected.
The second principle that has emerged more recently is that the effect of perinatal stress apparent under nonstressful conditions later in life is not necessarily the same as when the (young) adult organism is tested in conditions under which corticosteroid levels are high. For instance, the effect of prenatal stress on LTP and LTD induction was particularly apparent when combined with an acute stress exposure in adulthood (Yang et al., 2007). It is noteworthy that several studies have supplied evidence that exposure to high corticosteroid levels or stress in adulthood against a background of early-life adversity may in fact result in an opposite pattern (i.e., facilitated LTP) (Stewart et al., 2004; Champagne et al., 2008; Bagot et al., 2009; Oomen et al., 2010; Fig. 12), from what is seen under basal conditions. This was also reflected at the behavioral level. Thus, early-life adversity impaired memory performance in relatively nonstressful tasks, but improved memory in stressful learning tasks (Champagne et al., 2008; Bagot et al., 2009; Oomen et al., 2010). This has led to the hypothesis that early-life adversity may adjust the development of the brain such that it is geared to optimally perform under comparable high-stress conditions later in life rather than under less stressful situations; the reverse seems to be the case in animals that grow up under quite favorable conditions. In other words, optimal performance in (young) adulthood is seen when the environment early and later in life match (Champagne et al., 2009; Oitzl et al., 2010; Nederhof and Schmidt, 2012). There are several examples regarding the link between electrical activity and behavioral performance to support the relevance of a matching beneficial environment or a mismatch in the case of early-life adversity (Kamphuis et al., 2003; Akers et al., 2006; Lin et al., 2006; Son et al., 2006; Yang et al., 2006, 2007; Judo et al., 2010; Lee et al., 2011a; but see Domenici et al., 1996; Noorlander et al., 2008), but very few studies so far have supplied evidence for the importance of a mismatch in animals without early-life stress or a match in animals with a “bad start in life” (Stewart et al., 2004; Champagne et al., 2008; Bagot et al., 2009; Oomen et al., 2010).
The data set on early-life stress is quite complex because of nonuniformity in many variables. One such variable is the moment in life at which the stressor is applied. In some studies, the stressor (or glucocorticoid treatment) was applied to the pregnant female animal, thus prenatally and indirectly affecting the pups, indirectly because corticosteroids circulating in the mother may reach the pups via the placenta (Seckl and Holmes, 2007) and because the behavior of the mother to the pups may be altered once they are born. More often, though, stress was applied directly to the pups, during the first 1 to 2 postnatal weeks. A second variable concerns the type and severity of the stressor that was applied. In the case of the mother being stressed, some studies used restraint stress for several days, others used variable stress or treatment with glucocorticoids. Postnatal stress administered to the pups ranged from exposure to a traumatic event (anoxia on PND 1 or foot shocks during PNDs 15–17) and very brief (minutes) or prolonged (ranging from 3 to 24 h) separation of pups from the dam to administration of glucocorticoids. The studies furthermore differ with respect to the delay between stress and investigation of its consequences for electrical activity. Most electrophysiological studies established the effects of perinatal stress for brain function in young adulthood. To our knowledge, none followed up the consequences for electrical activity into aging or even senescence. The outcome may very well be different for each of these stages in life, as was observed, for instance, with respect to the lasting consequences of 24-h maternal separation on PND 3 for spatial learning and memory performance throughout the lifespan (Oitzl et al., 2000). It is also important to emphasize that nearly all studies were carried out in male offspring. This is somewhat surprising, because the rationale in many cases was that early-life events are a major risk factor for psychopathology later in life, such as major depression, a condition that is much more prevalent in women than in men. And finally, although some genetic heterogeneity may occur in outbred rat or mouse strains, the relevance of gene-by-environment interactions, which is so evident from human studies, has scarcely been touched upon in rodent models (Ryan et al., 2009).
Despite these considerations, it is clear that stress experienced at an early stage of (brain) development has tremendous consequences on synaptic plasticity later in life: stress at this sensitive period may wire the brain such that it optimally functions under comparable circumstances later on. Nearly all of the currently available studies focused on consequences of perinatal stress for LTP or LTD in the Schaffer collateral projection to the hippocampal CA1 area. There is no reason to believe, however, that other projections in the brain are not equally sensitive to perinatal stress, as is in fact supported by the limited number of studies that did examine other areas. Likewise, it seems very likely that the changes are not restricted to synaptic plasticity but also apply to the underlying voltage- and ligand-gated currents (see for example Hsu et al., 2003; Lee et al., 2011a).
VI. Concluding Remarks
After more than 2 decades of research, there is ample evidence that corticosteroid hormones profoundly change neural activity, not only via the classic gene-mediated signaling pathway but also more rapidly, via G-protein-coupled pathways or protein-protein interactions. The general ideas that emerged and challenges for the future are summarized below.
A. Overarching Principles
Brain cells are continuously exposed to corticosteroid hormones, although the levels vary throughout the day (e.g., due to ultradian pulses, circadian rhythmicity, stressful experiences, but also factors determining corticosteroid transport over membranes and local enzymatic conversion). Although the hormones in principle reach every cell, they can only directly change the function of those cells expressing receptors. Nuclear MRs are nearly always substantially occupied because of their high affinity, and in those cells that express them—particularly neurons in the hippocampus and lateral septum—are thought to safeguard the viability and basal activity of the cells. More recently, it has become evident that these receptors may also reside in the plasma membrane and then serve as quick sensors for shifts in corticosteroid level. Nuclear GRs become fully activated when corticosteroid levels are very high and are therefore optimally equipped to assist in the brain's reaction to stressful conditions.
Via rapid nongenomic pathways, MRs mostly enhance neural activity, whereas GRs suppress activity (see summarizing Fig. 13). In the PVN, rapid GR-dependent effects are thought to contribute to rapid negative feedback of HPA axis activity. The role of rapid effects in extrahypothalamic brain regions is still debated, but behavioral data support the idea that nongenomic MR actions are particularly important during acquisition of relevant information, aiding the organism to select an appropriate strategy that is the best option in the short term, albeit at the cost of losing flexibility, a strategy that may not be beneficial in the long run. These rapid effects can also alter the responsiveness to subsequent corticosteroid exposure (e.g., in the BLA), but this does not occur in all brain regions.
When a wave of corticosteroids hits the brain, this not only causes rapid effects but also starts signaling pathways that change neural activity much later, ranging from approximately 20 min after steroid arrival to many hours or even days. These effects are primarily carried by GRs. In most hippocampal CA1 and CA3 neurons as well as in mPFC cells, GR activation increases surface expression of AMPA receptors and strengthens glutamatergic signaling through pathways partly overlapping with LTP. This might raise the threshold for subsequent induction of LTP and promotes LTD. In conjunction with a propensity to dampen transfer of overall excitatory information as a result of enhanced function of calcium-dependent potassium currents, GRs thus help to strengthen some synapses—presumably those synapses activated at the time of stress exposure—but protect cells against excitatory inputs reaching the cells at a later point in time. This is thought to contribute to appropriate higher cognitive functions, such as the consolidation of stress-related contextual information and working memory performance >20 min after stress. The dampening of information transfer 20 min to several hours after stress is not seen in the ventral-most part of the CA1 hippocampal area and the basolateral amygdala, providing an extended window for encoding of aspects of information that are processed via these areas. This may contribute to the fact that emotional aspects of a stressful event are extremely well retained.
It is noteworthy that the idea of corticosteroids acting in several time domains is not new; it was brought up in the 1980s by the work of Gillman and Jones (e.g., Beckford et al., 1983), showing that corticosteroids can inhibit HPA axis activity both rapidly and in a more delayed manner compatible with the classic genomic signaling pathway. In that case, the net result of both actions is the same: normalization of HPA axis activity, be it rapidly or slowly. What is new is that the two time domains in which corticosteroid hormones affect limbic cell function seem to serve two entirely different functions. The rapid mode helps the organism select the best strategy in the short term, to withstand an imminently dangerous situation, whereas the slow mode is more in line with the role of neuroendocrine negative feedback: normalization of activity, restoration of cognitive control, and preparation for the future.
When the organism is not exposed to a single stressor but to several weeks of stress, the ability to induce or maintain LTP is strongly reduced, and the likelihood to induce LTD enhanced, even under nonstress conditions. Exposing animals to stress/corticosteroids against a background of chronic stress does not cause additional suppression of LTP. Intrinsic or synaptically evoked cell activity can be either enhanced or unaffected by chronic stress. Stress experienced early in life also has strong and lasting effects on neural activity. Mild disturbances early in life generally enhance excitability, for instance by facilitating the induction of LTP. By contrast, more severe conditions were consistently linked with a reduced ability to evoke synaptic plasticity under basal HPA conditions. It is noteworthy that in these animals renewed exposure to high corticosteroid levels in adulthood causes highly efficient induction of LTP. This flip in phenotype appears to have behavioral relevance, because animals with adverse early-life conditions perform relatively poorly in learning tasks that are not very stressful but show strong memory performance in stressful learning tasks. Apparently, early-life conditions can direct brain development such that the circuits are optimally geared to perform well under comparable conditions later in life.
B. The Average Male White Rat Neuron
These overarching principles are derived from the currently available studies, which have certain limitations. Nearly all studies have been performed in young-adult rodents. Given the strong influence of early-life environment and the fact that these influences vary over the lifespan (for review, see Lupien et al., 2009), it would be very useful to also examine animals at a more advanced age. Landfield and Eldridge (1994)) have argued that many of the effects observed with corticosterone resemble what is seen in the aged brain, also with regard to electrical activity. The aged brain may thus respond differently to corticosteroids. Unfortunately, patch-clamp experiments are difficult in aged rodents and much easier to carry out in very young animals. This is probably the reason that some studies discussed in this review were even performed in prepubertal rodents. Cultured cells are also derived from very young animals and usually have not matured beyond 21 days of age. There is no guarantee that corticosteroid effects at that age are the same as later in life.
A second limitation is that nearly all experiments were performed in male rats, not in female rats. Although this is understandable, given the complicating factor of the estrous cycle in female rats, it seriously hampers the extrapolation of findings in rodent models to the human brain, where stress-related psychopathology in fact is often more prevalent in women than in men. There are numerous examples in rodent models that the effects of stress are different in male and female rodents (for reviews, see Luine, 2002; Luine et al., 2007; Ter Horst et al., 2009; Weinstock, 2011) including for electrical activity (e.g., van Hasselt et al., 2012a).
Species and strain differences have also hardly been addressed with regard to corticosteroid effects on neural activity. Although the effects were comparable for mice and rats whenever tested (e.g., compare Karst et al., 1994 and Karst et al., 2000), this does not exclude the possibility that differences may exist. Stress responsivity differs largely among strains of rats or mice (for reviews, see Henn and Vollmayr, 2005; Millstein and Holmes, 2007), and this is certainly expected to result in adaptive changes in corticosteroid receptor properties and hence their role in altering neural activity.
A pharmacologically relevant issue is the fact that corticosteroid actions have very often been examined with very high concentrations, well beyond the Kd of the receptor. Subtle differences in local expression of receptor variants or post-translational modifications of receptors, resulting in altered binding affinity or receptor capacity, will remain unnoticed with these saturating levels of the hormone. Nevertheless, in “real life,” these regional differences or changes during the life span are very meaningful and deserve more attention.
Regarding the issue of hormone concentrations, it is also relevant to consider the role of glial cells. As part of the tripartite glutamate synapse, glial cells play an important role in regulating the level of glutamate in the extracellular space (for review, see Popoli et al., 2012). In short, glial cells prevent glutamate spillover and consequent overactivation of extracellular glutamate receptors. This occurs mainly by active clearance and metabolism of glutamate into glutamine via high-affinity excitatory amino acid transporters in the plasma membrane of glial cells. This is widely accepted as the primary mechanism via which glutamatergic action in the extracellular space is terminated (Tzingounis and Wadiche, 2007). Acute stress, chronic stress, and/or glucocorticoid administration do affect glial cell glutamate uptake (Popoli et al., 2012). In general, although acute stress and corticosteroid exposure seem to induce adaptive changes in glutamate clearance, chronic stress increases extrasynaptic glutamate levels by reducing the clearance rate via excitatory amino acid transporters in hippocampal and PFC cells. Although the underlying mechanism, and especially the role of MR and GR, is still unclear, the reduced clearance rate could certainly contribute to the commonly observed disruption in neurophysiology after chronic stress.
Finally, there are technical limitations linked to measuring electrical activity. Information on currents, particularly in subcortical areas, can be obtained only in vitro, in preparations lacking most afferent connections and maintained under artificial conditions. This presently precludes detailed investigation of rapid corticosteroid effects on electrical activity in association with their behavioral consequences, which would need to be carried out in vivo. Slow gene-mediated actions are easier to examine, because these seem to persist in slices prepared from animals earlier exposed to a stressor. For instance, increases in calcium current amplitude were observed several hours after stress exposure and were highly comparable with effects seen in corticosterone-treated slices prepared from naive animals (Joëls et al., 2003; van Gemert et al., 2009). A second drawback inherent to patch-clamp recording is the inevitable perfusion of the intracellular compartment. If any of the intracellular proteins is essential for the corticosteroid effects taking place but not included in the pipette solution, the hormonal effect may be lost if the cell is held in the whole-cell recording mode during the putative development of corticosteroid actions (see, for example, Teschemacher et al., 1996). In most studies, this was not a problem, though, because these compared groups of cells recorded before steroid exposure/after vehicle exposure with groups of cells that had been treated with steroids before establishing the whole-cell recording mode.
Although a direct link between corticosteroid actions on behavior and on intrinsic/synaptic currents is not (yet) possible, a tighter connection between behavior and neural firing frequency is feasible, using multielectrode recording in vivo. These advanced methods have rarely been applied so far (Kim et al., 2007; Passecker et al., 2011) but are essential to get better insight in the behavioral relevance of corticosteroid actions on neural activity.
C. In Search of Multiple Dimensions
Most studies have concentrated on one aspect only of stress effects in the brain: one area at a time; rapid or slow effects, not both or the transition from the one to the other; only one of the stress hormones (e.g., corticosterone), not the (inter)actions with other stress mediators; and one manipulation to change corticosteroid levels. Such a reductionistic approach is very useful to establish the contribution of that particular aspect to the overall consequence of stress, but it is far removed from real life events and makes it hard to piece the various bits of information together.
Although an overall pattern emerges from the current studies and shows that corticosterone affects neural activity of the BLA and ventral CA1 region (and possibly the dentate gyrus) in a different manner than the remainder of the hippocampus and mPFC, the response of multiple areas has seldom been examined in one study. Direct comparisons were made between the dorsal and ventral hippocampus (Maggio and Segal, 2007, 2009a,b), the dorsal CA1 versus dentate gyrus (Kavushansky et al., 2006; Vouimba et al., 2007; van Gemert et al., 2009), the BLA and CA1 area (Karst et al., 2010), and the BLA and DG (Vouimba et al., 2004). Most probably, corticosterone will affect connections between many areas at a time, changing the responsiveness of entire circuits. Recording the connections between all of these areas is difficult with electrophysiological methods. Such insights could be obtained by applying neuroimaging methods, using (resting state) functional magnetic resonance imaging or diffusion tensor imaging (see, for example, Ferris and Stolberg, 2010).
A dimension that has received little attention until recently is the time frame in which corticosteroids are active. The slow gene-mediated actions that develop after a delay of approximately 1 h and depend on transactivation via GRs were expected on the basis of what is known about the intracellular signaling pathways. The other end of the spectrum is formed by those effects that develop within a few minutes, involve G-protein signaling, and occur in the absence of protein synthesis. However, there also seems to be a category of events taking place in the intermediate time domain. These effects become apparent after approximately 20 min and peak at approximately 40 to 60 min after corticosteroid exposure (Pfaff et al., 1971; Vidal et al., 1986; Joëls and de Kloet, 1993; Maggio and Segal, 2009a; Tse et al., 2011). These effects are usually evoked with relatively high corticosterone concentrations and may involve rapid nongenomic protein-protein interactions between GR and phospho-ERK1/2 (Yang et al., 2004, 2008b; Gutièrrez-Mecinas et al., 2011) or phospho-cAMP response element-binding protein (Ahmed et al., 2006; Roozendaal et al., 2010), subsequently activating enzymes such as MSK (Gutièrrez-Mecinas et al., 2011) or SGK (Liu et al., 2010). Apparently, once corticosteroids hit the brain, they can change neural activity from minutes to hours.
But when exactly do corticosteroids reach the brain after stress? Although not much is known about this issue, there is evidence for a delay of up to 20 min (Droste et al., 2008). This clearly needs further study. It drives home the point, though, that other stress-released hormones and transmitters may reach relevant brain areas well before corticosteroids and that the latter do not work in isolation. In this review, we highlighted effects of corticosteroid hormones on intrinsic cell properties and the main excitatory and inhibitory inputs. Evidently, neurons also receive inputs mediated by other transmitters/modulators, many of which are enhanced in concentration after stress, such as noradrenaline, dopamine, serotonin or CRH. There is extensive evidence that stress and corticosterone are able to change not only the biosynthesis, release, and reuptake of these transmitters/modulators (Chaouloff et al., 1999; Czyrak et al., 2003; Kvetnansky et al., 2009; Bonfiglio et al., 2011), but also their influence on cellular excitability. This has been shown in detail with regard to serotonin (see, for example, Joëls et al., 1991; Laaris et al., 1999; Fairchild et al., 2003; for reviews, see Joëls et al., 2007; Haj-Dahmane and Shen, 2011), noradrenaline, and, to a lesser extent, CRH (Gallagher et al., 2008). For instance, noradrenaline (acting via β-receptors) targets the AHP (Madison and Nicoll, 1982; Faber and Sah, 2005; Oh et al., 2009), GluA2 subunit trafficking (Liu et al., 2010), and (maintenance of) LTP (Thomas et al., 1996; Straube and Frey, 2003; Gelinas and Nguyen, 2005), thereby mimicking effects described for corticosteroids. These common endpoints form potential platforms for interaction between various stress mediators (Joëls et al., 2011). There are indeed several examples of noradrenaline and corticosterone acting in concert to promote optimal neural function (Akirav and Richter-Levin, 2002; Korz and Frey, 2005; Zhou et al., 2011), which is also reflected at the behavioral level (Roozendaal et al., 2004). Yet when corticosterone is given >1 h in advance of noradrenaline or isoproterenol, the hormone suppresses noradrenergic function (Joëls and de Kloet, 1989; Pu et al., 2007, 2009; Liebmann et al., 2009). This is possibly caused by occlusion because the two compounds converge on the same functional endpoints, which would be another example of metaplastic changes induced by corticosteroids; right now this is pure speculation. This sequence of hormone exposure, of course, is pharmacological in nature, because in reality, noradrenaline will reach brain cells before and not after corticosterone. The complex interplay between all neurotransmitters, modulators and hormones released after stress and reaching specific neuronal compartments at various timepoints after stress needs further investigation to fully comprehend how the input-output relationship of neurons is altered by stress (Joëls and Baram, 2009).
Finally, an issue that has been neglected so far is the fact that fluctuations in corticosteroid level can be caused by different processes: stress, but also circadian or ultradian variations. These are not independent factors; for instance, stress exposure during the rising limb of an ultradian pulse affects transcriptional activity in the brain differently than during the falling limb (Sarabdjitsingh et al., 2010a). As illustrated by the metaplastic changes in the BLA, this interdependency may also occur in the brain: a stressor experienced at an ultradian or circadian peak may change BLA activity in another way than when this occurs at the trough. Along the same line, a number of stressors in succession could drive neural activity in the BLA in a different manner than activity in the dorsal CA1 region. These aspects need dedicated experiments to be fully understood.
D. Disease and Targets for Treatment
The functional relevance of corticosteroid effects on neural activity in rodents has been discussed in sections II and III in terms of their contribution to behavioral adaptation, helping to survive challenging situations and to apply the gathered information when experiencing comparable situations in the future. Although these actions are beneficial under “normal” conditions, they can impose a serious risk for developing pathologic conditions when the stressful conditions are 1) very severe, 2) coinciding with additional challenges to the brain, 3) uncontrollable and persistent, or 4) occurring at a highly vulnerable moment in life (e.g., during early development or in senescence) (de Kloet et al., 2005). For example, an extended window for encoding emotional aspects of a stressful situation is beneficial when it helps to remember relevant aspects of a daily life situation more vividly than less relevant information, but in humans this can turn into psychopathology when people are haunted by these details day and night, as is the case with a post-traumatic stress disorder. Another example refers to the increased calcium influx seen after stress. Normally this may help to drive calcium-dependent potassium currents, serving as a brake on excitatory input several hours after stress. However, a calcium overload—due not only to enhanced influx but also to reduced efflux (Bhargava et al., 2000, 2002)—may exacerbate the consequences of conditions associated with strong depolarization or additional calcium exposure, such as epileptic seizures. In accordance, corticosterone administration during kindling accelerated the behavioral signs of epilepsy and was associated with large calcium currents in fully kindled rats (Karst et al., 1999).
The potential risk of chronic stress for human disease is well documented in the case of major depression (for review, see Holsboer and Ising, 2010). Dysregulation of the HPA axis, showing higher trough levels of cortisol (Herbert, 2012) and an exaggerated response to a dexamethasone-CRH challenge, occurs in a substantial proportion of people with major depression, even before the manifestation of any clinical symptoms in high-risk proband with a positive family history of affective disorders (Ising et al., 2005). Normalization of HPA axis activity turned out to be a valuable predictor of relapse probability (Künzel et al., 2003; Appelhof et al., 2006; Aubry et al., 2007). Considering the added risk of periods of stress and/or dysregulation of the HPA axis, and more particularly the role of GRs in the functional consequences, the prediction would be that restricting GR effects (e.g., by antiglucocorticoid treatment or the use of steroid synthesis inhibitors) would be beneficial for the outcome.
This possibility has been specifically addressed in an animal model for chronic unpredictable stress. The reduction in LTP observed after 21 days of stress turned out to be fully reversed by treatment with mifepristone during days 18 to 21 (Krugers et al., 2006). This was also true for the chronic stress-induced increase in calcium current amplitude (Karst and Joëls, 2007). Small currents were also observed in dentate granule cells from mifepristone-treated, chronically stressed animals receiving corticosterone in vitro (van Gemert and Joëls, 2006). Similar normalizing effects of mifepristone treatment were observed with regard to neurogenesis in the dentate gyrus (Oomen et al., 2007). These cellular actions of the antiglucocorticoid possibly contribute to its beneficial effect in severe cases of psychotic depression, provided sufficiently high concentrations in plasma are achieved (DeBattista et al., 2006; Gallagher et al., 2008; Blasey et al., 2011). Whether mifepristone exerts its effects only via blockade of GRs is hard to judge, given the many adaptive actions that may occur, both peripherally and in the brain (Kling et al., 2009).
Apart from drugs directly interfering with glucocorticoid action in the brain, the functionality of corticosteroid hormones and receptors, as well as the HPA axis in general, can also serve a more prognostic goal in health care, predicting the likelihood that individual patients will respond to pharmacotherapy. Genetic variants are increasingly seen as tools to achieve such personalized medical care (Holsboer, 2008). For instance, polymorphisms of FKBP5, a chaperone molecule of the GR, were found to predict recurrence of depressive episodes and response to antidepressant treatment (Binder et al., 2004; Kirchheiner et al., 2008; Lekman et al., 2008), although this finding could not always be replicated (Tsai et al., 2007; Sarginson et al., 2010). It remains a challenge for the future to study the consequences of these genetic variants for corticosteroid actions on neural activity (e.g., when expressed in induced pluripotent stem cells). Such clinically relevant test models can provide a quick and efficient means to examine the efficacy of already existing agents to normalize aberrant function.
Acknowledgments
This work was supported by Nederlandse Organisatie voor Wetenschappelijk Onderzoek–Aard-en Levenswetenschappen (NWO-ALW) [Grant 821.02.007] (to M.J.) and [Grant 817.02.017] (to H.K.).
Authorship Contributions
Wrote or contributed to the writing of the manuscript: Joëls, Sarabdjitsingh, and Karst.
Footnotes
↵ The online version of this article (available at http://pharmrev.aspetjournals.org) contains supplemental material.
This article is available online at http://pharmrev.aspetjournals.org.
↵1 Abbreviations:
- ADX
- adrenalectomized
- AHP
- afterhyperpolarization
- AMPA
- α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid
- BLA
- basolateral amygdala
- BSA
- bovine serum albumin
- CA
- cornu ammonis
- CBG
- corticosteroid-binding globulin
- CRH
- corticotropin-releasing hormone
- DG
- dentate gyrus
- ERα
- estrogen receptor α
- ERK
- extracellular signal-regulated kinase
- GluA2
- AMPA receptor GluR2 subunit
- GR
- glucocorticoid receptor
- HPA
- hypothalamo-pituitary-adrenal
- HSD
- hydroxysteroid dehydrogenase
- IPSC
- inhibitory postsynaptic current
- LTD
- long-term depression
- LTP
- long-term potentiation
- mdr
- multidrug resistance
- mEPSC
- miniature excitatory postsynaptic current
- mIPSC
- miniature inhibitory postsynaptic current
- mPFC
- medial prefrontal cortex
- MR
- mineralocorticoid receptor
- MSK
- mitogen- and stress-activated kinase
- NMDA
- N-methyl-d-aspartate
- PFC
- prefrontal cortex
- PND
- postnatal day
- PVN
- paraventricular nucleus of the hypothalamus
- RU38486
- mifepristone
- sIPSC
- spontaneous inhibitory postsynaptic current.
- © 2012 by The American Society for Pharmacology and Experimental Therapeutics
References
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵