Neuropathology of stress

Whenever an endogenous or exogenous challenge is perceived as unpleasant, aversive or threatening, a series of systems and processes is activated that generates a coordinated response to that particular challenge, or stressor. This so-called stress response, an integral part of any adaptive biological system, is conserved throughout evolution. It is particularly active when an individual’s homeostasis, well-being, overall health or survival is threatened. An unpleasant surprise, relational or financial problems, the loss of a loved one, bereavement, unpredictability, an acute threat, e.g., when faced with an animal predator, or with psychosocial demand in humans, can all initiate a stress response. The same is true for perturbations of a more biological nature, such as an energy crisis, physical injury, hemorrhage or inflammation.

According to the popular press, stress is ever present in our modern, performance oriented and demanding society [109]. Stress contributes to several disabilities worldwide and as such represents a severe economical burden. The WHO expects that mental disease, including stress-related disorders, will be the second leading cause of disabilities by 2020. In the US, job stress alone, e.g., has been estimated to cost several hundred billions of USD every year in absenteeism, turnover, diminished productivity and medical, legal and insurance costs.

Stress, however, is no single entity and several different types of stressors can be distinguished: stressful challenges can be acute (being confronted with a predator or giving an important oral presentation) or of a chronic nature (living in poverty or in a broken family). It may occur only once, or may rather take place in a repetitive manner, that can be anticipated. Conversely, stress can be unpredictable and uncontrollable, mild or severe, and occurring in or out of context, e.g., of a learning experience [93, 102]. In addition, how stress exposure is actually perceived by an individual varies greatly, as does the persistence of its consequences.

Importantly, physiological ‘stress’ responses also occur following rewarding, “positive” and appetitive stimuli (e.g., winning a competition, sexual activity). Although they are often not considered to be stressors in classic, generally “negative”, terms, the physiological responses elicited by them can be as large as those seen after more aversive stimuli [104]. For the purpose of definition, a stressor in the context of this review will refer to any environmental demand that exceeds the physiological regulatory capacity of an organism, in particular during situations of unpredictability and uncontrollability. They are characterized by the absence of an anticipatory response (unpredictable), or a reduced recovery (uncontrollable) of the neuroendocrine reactions to stress. Taken together, depending on the type of stressor, various signals converge to orchestrate an integrated stress response that ‘resets’ many peripheral and central processes and allow an individual to adapt to the changes in its environment and thus to restore homeostasis.

The physiological stress response can be divided into two different time domains with a very quick response and a delayed response. The first phase of the stress response is considered the “alarm reaction” or the “fight-fright-or-flight” response, which involves the rapid activation of the autonomic nervous system (ANS) that causes the release of epinephrine and norepinephrine from the adrenal medulla. These hormones quickly elevate basal metabolic rate, blood pressure and respiration, and increase blood flow to the more vital organs that are essential for the “fight-or-flight” response, such as the heart and skeletal muscles. At a later stage, the hypothalamic–pituitary–adrenal (HPA) axis is activated as well. In this classic neuroendocrine circuit, limbic and hypothalamic brain structures coordinate emotional, cognitive, neuroendocrine and autonomic inputs, which together determine the magnitude and specificity of an individual’s behavioral, neural and hormonal responses to stress.

This second response is mediated by glucocorticoid (GC) hormones (corticosterone in rodents and cortisol in humans) which generally act in a slow, genomic manner as transcriptional regulators of glucocorticoid responsive genes. Fast (non-genomic) GC actions have also been described and their actions are mediated by putative membrane-bound receptors. It should be emphasized that other signalling pathways act in concert with the HPA axis like the gonadal axis, the adipose axis, and the immune system. All these help to direct energy resources such that attention can be focused on the most urgent and important elements of the challenge while other, less urgent functions, e.g., food intake, digestion or reproduction, are temporarily suppressed [93]. The profound change in activational patterns that is induced by perceiving a situation as stressful can nowadays be visualized by functional magnetic resonance imaging (fMRI) studies, and as can be seen in Fig. 1, encompasses a variety of limbic system and frontal lobe structures, including hippocampus, amygdala and the anterior cingulate cortex. Throughout the text, we focus on the HPA axis, because this system is most heavily investigated, but it should be emphasized that several other systems contribute to the stress response as well (see “Inflammatory changes and depression”, “Gender differences in depression: relationship with HPA axis activity”, “The glutamatergic and GABAergic systems in depression”, “Glial plasticity”, “Water content, volume changes and altered vasculature”).

HPA axis activation is triggered by corticotropin-releasing hormone (CRH) in the paraventricular nucleus (PVN) that induces adrenocorticotropic hormone (ACTH) release from the pituitary, which in turn releases GCs from the adrenal. Regulation occurs through negative feedback after GC binding to high-affinity mineralocorticoid (MR) and lower affinity glucocorticoid receptors (GR) [47]. HPA axis activity is not only affected by stress, (see “Inflammatory changes and depression”) and a bi-directional communication exists, e.g., between the immune and neuroendocrine stress systems.

The GR helps to maintain GC levels within physiological limits [57, 104]. Aberrant GR expression has been implicated in stress resistance, anxiety and depression [47, 172, 234]. GC plasma levels are under circadian and ultradian control [115, 163]. Together, MR and GR determine sensitivity of the brain to stress [77, 135, 160, 203] and thereby modulate attention, vigilance, behavior and memory formation and eventually adaptation and coping with stress.

Upon their release in the periphery, GCs enable an individual to engage in an adaptive response to a stressor by affecting energy and lipid metabolism, among others. The large numbers of GRs in the brain, and particularly the hippocampus, make it vulnerable to elevated GC levels [47, 120, 209]. The GR occurs in at least two isoforms (GRalpha and GRbeta), with the GRalpha isoform being the most predominant one in brain. In addition to rodent studies, [171] several human brain regions express GR and MR [77, 162, 189, 233]. In humans, considerable diversity of GR and MR transcripts exists [49, 100], which includes 13 exon 1 mRNA variants and 8 N-terminal variants, that arise from the GRalpha isoform. In the human hypothalamus, GRalpha protein is selectively expressed in CRH-containing parvocellular, but not magnocellular, neurons of the PVN [231]. Application of exogenous, synthetic GCs strongly suppresses CRH and vasopressin production in neurons of the human PVN (Fig. 2) [59], whereas oxytocin (OXT) neurons are not affected.

In contrast to the relative paucity of GR in the rhesus monkey [175], the human hippocampus shows abundant GR protein expression in CA1 and DG neurons, although the CA3 subregion has lower levels. GR is additionally expressed in astrocytes and overall GR levels remain stable with age. Increases have been found in the hippocampus and amygdala of patients with major depression (MDD) [229, 231, 232]. Thus, even though less is known about the MR [162], the prominent expression of GR in human brain makes these regions important targets for stress exposure [2, 93, 100]. Whereas acute and short-term stress is generally considered beneficial and adaptive, chronic stress may cause an MR/GR imbalance or down-regulation [47, 162] which can alter HPA feedback and result in overexposure of the brain and peripheral tissues to these powerful steroids.

When considering neuropathological aspects of stress, earlier animal experiments had reported brain damage after prolonged periods of stress, especially in aged animals and mainly in the hippocampus. For example, Landfield et al. [107] showed that cumulative GC exposure influenced hippocampal viability and compromised cognition. Subsequently, Sapolsky et al. [178, 182] reported that chronic stress causes a loss of pyramidal neurons in the hippocampus, accompanied by cognitive deficits in rats. Another study described that training rats for 6 months in a two-way shuttle escape task, using low intensity foot shock stress (4 h/days) resulted in endogenous hypercortisolism and CA1 pyramidal neuronal loss in senescent rats [98], while others reported reactive gliosis, reduced dendritic branching and reductions in volume and in CA1/3 cell numbers [181, 183].

The hippocampus was previously thought to inhibit CRH activity directly, and given its high MR and GR density [171, 189, 231], damage to the hippocampus was expected to cause a disinhibition of CRH activity, thus increasing the drive on the HPA axis, which in turn would further stimulate GC levels and aggravate hippocampal damage. This feed-forward “glucocorticoid cascade hypothesis” was proposed to be a pathogenic mechanism underlying stress effects on the brain, and considered relevant for human disorders associated with peripheral HPA axis changes that were paralleled by structural changes in the hippocampus, like Alzheimer’s disease (AD), post-traumatic stress disorder (PTSD), Cushing’s disease and depression [182].

These interpretations were supported by studies on subordinate, wild-born vervet monkeys that had experienced prolonged, severe social stress in captivity and that displayed post-mortem adrenal hypertrophy. Although the coincident hippocampal degeneration in these animals was consistent with the glucocorticoid cascade hypothesis, the morphological alterations and neuron loss in these, and also in other non-human primates implanted with cortisol pellets in their hippocampi [183], were most pronounced in CA neurons [217], brain regions particularly sensitive to pressure artifacts; after death, the soft tissue of the brain can be compressed by its own weight or by manipulation during brain collection, which refers to “post-mortem compression”.

While the initial studies had used rather extreme, physical stressors or extremely high pharmacological GC concentrations [101, 181], later studies in non-human primates and rodents that used more relevant psychosocial stressors [112, 219, 223], added conflicting data insofar that they did not observe massive neuronal loss or obvious neuropathology following chronic stress when this was assessed using unbiased stereological tools [157, 201, 223]. For example, when adult male rats were chronically treated with the selective GR agonist dexamethasone, with dexamethasone plus the selective MR agonist aldosterone, or with supraphysiological doses of corticosterone, only dexamethasone treatment reduced DG granule and CA3 pyramidal cell numbers [202]. However, this synthetic steroid differs from the endogenous corticosterone in terms of brain penetrance and retention and although it can induce cognitive deficits in aged mice [238], it did not influence the CA1 or hilar subfields. In contrast to dexamethasone, animals injected with corticosterone failed to reveal any change in cell number in any of the hippocampal subfields, although volume reductions in hilus and CA3 were observed [201]. Also in tree shrews exposed to chronic subordination stress, the CA1 and CA3 pyramidal neuron numbers were not different from controls, despite elevated cortisol levels [223]. Chronic studies in pigs, non-human primates and chickadees yielded similar results while changes in apoptosis were also not demonstrated [112, 157, 219].

It is now known that the tonic inhibitory control on HPA axis activity [82] is exerted through several, often indirect, neural pathways including the bed nucleus of the stria terminalis (BST), the amygdala and the endocannabinoid system [212, 221]. The GC-mediated negative feedback of the HPA axis thus takes place at several levels, including the hypothalamus and pituitary, and not only at the hippocampus [82, 84]. Also, although GCs target hippocampal cells, they can increase vulnerability to, but not directly cause, subsequent insults [33]. Moreover, an absence of massive structural changes after GC exposure does not necessarily imply that no functional, molecular or changes in responsivity [43] are induced. Furthermore, variables like genetic risk factors, early life programming and structural plasticity were not considered at the time. As a consequence, the GC cascade hypothesis has been refined and rephrased over time [31, 119, 134, 154, 160]. Together with the corticosteroid receptor hypothesis of depression [87] and the realization that GCs are on the other hand, essential for, e.g., dentate viability [197], it has sparked the development of drugs selectively targeting specific stress system components [88, 188, 243].

Stress is the most common risk factor for the development of mood disorders like MDD that are thought to result from interactions between genetic predispositions and the environment [173]. Especially stressful life events experienced during early childhood or adolescence can significantly increase the risk to develop depression [78]. Indeed, in a large proportion of depressed patients the HPA axis activation and GC feedback resistance is common. This is reflected by the high percentage of dexamethasone non-suppressors in this population as well as hypertrophy of the adrenals and pituitary and increased plasma levels of cortisol, particularly during the through of the circadian rhythm, and increases in CRH and AVP expression in the PVN [209]. Notably, depressed subjects show remarkable heterogeneity in neuroendocrine function and the proportion of depressed individuals demonstrating overt HPA axis abnormalities may range from 35 to 65 %.

Post-traumatic stress disorder (PTSD) is a multisystem disorder with multiple comorbidities. PTSD patients have experienced severe, often life-threatening, episodes of stress that cause flashbacks, nightmares and sleep problems, emotional numbness or emotional outbursts, anhedonia, inappropriate startle reflexes and problems with memory and concentration [21]. Notably, the HPA changes in PTSD indicate low cortisol levels consistent with increased glucocorticoid sensitivity. In one particularly intriguing experimental protocol, 40 pairs of identical twins were studied, one of whom participated in the Vietnam war while the other stayed at home. Of those who experienced combat, 43 % developed PTSD. They turned out to have smaller hippocampi, but their stay-at-home twin brother did too. In contrast, those who did not develop PTSD had larger hippocampi, and so did their stay-at-home twin as well. A small hippocampus thus seems to be present already prior to the stressful experience and in fact to infer an increased vulnerability to PTSD [69, 180]. This goes back to the idea of either genetic factors, or life history and shaping events during critical development periods, or an interaction between the two, determining hippocampal volume early in life, and putting the individual on a trajectory for (psycho)pathology [30, 124, 161].

Thus, it remains elusive whether hypercortisolism is indeed responsible for hippocampal shrinkage, since combat-related PTSD is associated with decreased HPA axis activity and steroid feedback super-sensitivity that often lasts for decades after the initial trauma. It has been presumed that early on in the process the HPA axis may have been strongly activated. This is based on the observation that soldiers who had undergone random bombardments in the Korean war displayed markedly increased levels of cortisol, with the highest levels of cortisol in soldiers who had been in the greatest danger. It has therefore been hypothesized that high levels of cortisol at the time of the stressor would result in damage to the hippocampal neurons that may persist for many years after the original trauma and that could lead to reductions in hippocampal volume and subsequent differences in feedback or stress responsivity. However, those victims of rape or motor vehicle accidents who later developed PTSD appeared to have—already a few hours after the traumatic event—lower cortisol levels than victims who had no subsequent psychiatric disorder, or those who developed major depression. Yet, pituitary and adrenal hyperactivity to exogenous CRH and ACTH has been demonstrated in these patients. An increased sensitivity or up-regulation of GRs in PTSD and a pre-existing smaller hippocampal volume thus seems, at present, the best explanation [180, 239].

One of the best known forms of structural plasticity is dendritic retraction that was first observed in CA3 and CA1 hippocampal neurons following chronic stress exposure. At the structural level, prolonged exposure to high doses of corticosterone reduces apical (but not basal) dendritic complexity of CA3 pyramidal neurons [237] and a prolonged exposure to various types of chronic stress results in similar changes in different species [64, 129]. These changes in hippocampal dendritic morphology generally need several weeks to develop. In addition, chronic stress also leads to a loss of mossy fiber synapses, increased surface area of the post-synaptic density, and rearrangements of synaptic mitochondria and vesicles at the presynaptic terminals [176, 213].

The effects of stress on spine density are less clear. While increase in the number of spines were reported on CA3 pyramidal dendrites and in the size of the post-synaptic densities on CA1 synapses, others could not find changes in spine density [129] or reported a decrease in spines, that was notably rapidly reversible already after a recovery period or subsequent training [1, 176, 205]. Oscillating GC plasma concentrations along the circadian rhythm appear to be important for the elimination of spines present before motor learning, and also for the maintenance of new spines along with the retention of motor memory [115]. Spine elimination, but not their formation, was shown to require MR activation [115]. Together, these studies suggest that prolonged exposure to chronic stress and glucocorticoids markedly alters the number and morphology of both pre- and post-synaptic structural elements and thus strength of excitatory synapses in the hippocampus.

Dendritic spines are of particular relevance as they are critically involved in the storage of information and their density can be increased by stress under specific conditions [212]. Pyramidal cells of the hippocampus and PFC respond with reduced dendritic complexity and spine loss to chronic stress [71, 164], both of which are reversible following a recovery period [71, 165], but other regions appear resistant [194]. Consistent with circuit-specific effects of stress (Fig. 5), also increases were found in the dendritic arborization of orbital frontal cortex neurons, an effect opposite to what is observed in other cortical neuron populations. Interestingly, in the basolateral amygdala and nucleus accumbens (NAc), stress generally results in hypertrophy of dendritic arborization and increases in spine density (Fig. 6). In the amygdala, these changes do not normalize following a recovery period after stress [137, 225]. In addition to changes in spine density, chronic unpredictable stress reduces density of synapses in the rat PFC [94]. Similar reductions in synaptic density and a lower expression of synaptic function-related genes occur in the DL–PFC of MDD subjects [94].

According to the glucocorticoid cascade hypothesis, neuronal apoptosis may underlie hippocampal volume shrinkage observed after stress [182]. Neuronal death has also been implicated in the cerebral shrinkage that occurs following prednisone administration and/or in the inflammatory changes in the cortex in depression [119, 134, 196, 209]. However, initial histological and neuropathological examination of the hippocampus of depression models or from patients that had been depressed or were exposed to synthetic corticosteroids, could not support this notion. In hippocampi from established depressed patients, no indications for obvious neuronal loss or for significant neuropathology could be found using a variety of relevant architectural, synaptic and glia markers [29, 117, 119, 142] (see Figs. 7, 8). In a very recent study, Boldrini et al. [17] reported fewer granule cells in the dentate gyrus of unmedicated depressed patients (without fewer neuronal progenitor cells) suggesting that cell maturation or turnover defects in this plastic hippocampal subregion might be related to the duration of the illness. Notably, the hippocampal volume reductions present in Cushing’s disease were reversed after cessation of steroid exposure [204]. Similar findings exist on ventricular enlargements that are reversible in certain conditions (e.g., following recovery from alcoholism or prolonged steroid use), thereby challenging the hypothesis that ventricular enlargement predicts neuronal loss [210]. This agrees with the general clinical experience with depressive or Cushing’s patients, in which treatment or operation can relieve their depressive symptoms, several of the HPA alterations, and even the hippocampal shrinkage, findings that would be hard to interpret had irreversible damage or massive cell loss been induced.

The incidence of apoptosis or cell loss in rodent stress models is rare and also seen after acute stress [79, 119, 241]. Chronic stress changed apoptosis in tree shrews only in specific hippocampal subareas and was normalized by tianeptine treatment [118]. These effects also occurred in associated cortical areas and were consistent with a general anti-apoptotic mode of antidepressant action [118, 134]. This bears considerable relevance for the interpretation of structural and/or neuropathological studies on the hippocampus in depression, where almost all patients generally receive antidepressant treatment [90, 117] and hence, effects of the disorder per se might be masked by the concomitant drug treatment. Yet, in drug-free, depressed patients [17], little differences were reported when compared to patients on treatment [17], and hippocampal volume reductions were then present as well.

Another, relatively novel, form of neuroplasticity is neurogenesis that has been implicated in stress-induced hippocampal volume changes. Adult neurogenesis refers to the production of new neurons, an event that continues to occur in the adult brain of several mammalian species, including humans (Figs. 4, 9) [58, 120]. Neuron formation in the adult hippocampus received considerable attention during recent years and was proposed to contribute to depression etiology [96]. Although later studies suggested that neurogenesis was rather implicated in antidepressant drug actions [177], we still lack a coherent functional theory explaining how exactly newborn neurons in the hippocampus can contribute to mood, or to specific symptoms of depression, besides their cognitive deficits, which are related too, but not specific to mood disorders. Although a reduced rate of neurogenesis may reflect impaired hippocampal plasticity, reductions in adult neurogenesis alone are unlikely to produce depression. Lasting reductions in the turnover rate of DG granule cells, however, will alter the average age and overall composition of the DG cell population and thereby influence properties and vulnerability of the hippocampal circuit.

Furthermore, neurogenesis is regulated by exercise, inflammation and antidepressants (Fig. 4) [37, 120, 146, 177, 247] while stress potently inhibits neurogenesis in several species (Fig. 9) [120, 187]. Both psychosocial [74] and physical stressors [224] inhibit at least one or more phases of the neurogenesis process [74, 79]. Stress and GCs further interfere with most stages of neuronal renewal, proliferation, maturation and survival (Fig. 9) [79, 148, 187, 236].

The neurogenic hypothesis of depression proposes that prolonged reductions in neurogenesis, e.g., induced by stress, may affect hippocampal structure and volume in depression, and that successful antidepressant treatment would require increases in neurogenesis [96, 120]. In preclinical studies, the stress-induced suppression of DG neurogenesis can be prevented by antidepressant treatments, which can also have direct neurogenic effects in naive animals. They can block effects on depressive-like behavior and may also restore plasticity outside the hippocampus [24, 133, 146].

Inflammation also affects neurogenesis. Microglia are considered instrumental given their homeostatic role in inflammatory signalling that may become maladaptive in the chronically stressed brain [72, 86, 155]. Under physiological conditions, microglia exhibit a resting, ramified phenotype associated with the production of anti-inflammatory and neurotrophic factors, but when primed, e.g., by early life stress [51] or challenged by pathogens or damage during adult life, microglia can switch to an activated, amoeboid phenotype, that can initiate tissue repair, or rather produce cytokines that are detrimental for neuronal function and viability [52, 56]. Specific subsets of cytokines can be proneurogenic [10] while others decrease neurogenesis through IL-1 [247]. Pro-inflammatory mediators can further restrict neurogenesis [56, 247].

The exact underlying cellular mechanisms mediating the inhibitory effect of stress or inflammation on neurogenesis are largely unknown. Adrenal glucocorticoids have been suggested as key players and MR, GR and NMDA receptors have been identified on progenitor cells [74]. At the same time, several examples exist of a longlasting inhibition of neurogenesis after an initial stressor, despite later normalized GC levels [187]. This suggests that while glucocorticoids may be involved in the initial suppression of cell proliferation, particularly in early life, when neurogenesis is abundant, they are not always necessary for the maintenance of this effect [74].

Stress also affects levels of various neurotransmitters implicated in the regulation of neurogenesis: GABA [67], serotonin, noradrenalin, dopamine [6], cannabinoids, opioids and nitric oxide (see [6]). Stress further reduces the expression of several growth factors, such as brain-derived neurotrophic factor (BDNF), insulin-like growth factor-1 (IGF-1), nerve growth factor (NGF), epidermal growth factor (EGF), and vascular endothelial growth factor (VEGF), that all can influence neurogenesis (see, e.g., [186] while gonadal steroids should not be neglected either [63]. The proximity of the precursors to blood vessels further suggests a strong interaction with the vasculature and it is this population that is particularly sensitive to stress [80].

Although there is extensive preclinical evidence that antidepressants affect hippocampal neuroplasticity, information regarding the possible impact of medication on hippocampal volume in MRI studies of MDD patients is limited. A recent study by Huang et al. [90], found smaller DG volumes in unmedicated depressed patients and a post-mortem analysis reported the same [17] which would be consistent with the neurogenic hypothesis of depression. Interestingly, both subfield and posterior hippocampal volume reductions were reported that were only seen in unmedicated depression but were absent in patients treated with antidepressants. Although it is so far not possible to reliably detect ongoing neurogenesis in vivo [35, 131], these data are consistent with preclinical studies demonstrating subregional specific effects of stress and its modulators on neurogenesis, and/or on hippocampal functionality [99].

While neurogenesis in the mature human hippocampus is a rare event [14, 58, 101], changes have been reported following antidepressant treatment in middle aged, but not older depressed patients [15, 16, 121]. In a recent post-mortem study of major depressed patients, the volume of the histologically defined dentate gyrus was in fact 68 % larger in SSRI-treated depressed subjects [16]. SSRI treatment also substantially increased neural progenitor cells (NPCs) in the dentate gyrus [15], although this was not replicated by others [170] and may depend on the age of the patients [121].

The glutamatergic pathway shows alterations in patients with mood disorder [9, 174]. Particularly, since glutamate stimulates the HPA axis antiglutamatergic agents, like ketamine [95, 244] are considered promising targeta for new antidepressants. In patients with MDD, glutamine synthetase (GS) transcripts were down-regulated in the anterior cingulate cortex and dorsolateral PFC [27]. The expression of glutamate transporters, the excitatory amino acid transporters (EAAT), EAAT1, EAAT2 are reduced frontal brain regions in MDD [27]. Accumulations of extracellular glutamate may not only perturb the ratio of excitatory–inhibitory neurotransmitters [34], but could also damage neurons and glia.

Regarding the GABAergic system, the majority of studies revealed lower GABA levels in the brain of depressed patients, suggesting a hypoactive GABAergic neurotransmission, including the PVN [9, 126]. GABA also inhibits the HPA axis in the hypothalamus [103] and GABAergic drugs can help maintain a positive response during antidepressant treatment [126].

During recent years, also glial abnormalities have been implicated in the pathophysiology of mood disorders. Glial cells slightly outnumber neurons in the human hippocampus and constitute a substantial volume fraction. During pathophysiological conditions, specific glial subtypes may become activated, or may die, or, similarly to the dendritic debranching documented in neurons, may retract their elaborate branching processes, and thereby reduce hippocampal volume. A considerable proportion of the astrocytes further express GRs in the rodent [66] and human hippocampus [231] and similar to neurons, these cells are stress responsive too. Typically, astrocytes are identified by their GFAP (glial fibrillary acidic protein) expression and stress modulates GFAP expression in rodents: while acute physical stress increase GFAP immunoreactivity in several brain regions [106], chronic stress reduced GFAP mRNA and protein expression levels in the prefrontal cortex and hippocampus [3]. As corticosterone lowers GFAP levels in rat brain [144], elevated glucocorticoids are likely instrumental in these effects.

Further studies demonstrated that in laboratory animals, exposure to long-term stress, either during early life or adulthood, decreased number and somal volume of GFAP-positive astrocytes in the hippocampus and in several other stress-related brain areas [38, 111]. This implied that chronic stress may cause astrocytic loss, while more recent experiments did not substantiate that [216]. For example, in vitro experiments showed that dexamethasone treatment of astrocytes, derived from hippocampal primary cell cultures, results in growth inhibition and moderate activation of caspase 3, which is not followed by apoptosis [241], suggesting that hippocampal astrocytes are resistant to glucocorticoid-induced apoptosis.

A recent in vivo study suggests that the reduced number of GFAP-positive cells after stress may not reflect cell death. By comparing the results of different labeling methods, it was found that chronic stress was associated with a decrease in GFAP-+ cell numbers, but there was no indication for astrocytic cell loss based on Nissl staining or S100ß-immunoreactivity [216]. This latter study also showed that astrocytes respond to chronic stress by reorganizing their cellular morphology and reducing the length, complexity and volume of their processes [216].

Chronic stress can also reduce the proliferation rate of glial cells. This was shown in the medial prefrontal cortex of rats subjected to 5 weeks of social defeat, or to chronic unpredictable stress or after chronic corticosterone administration [7, 39]. Similarly, prolonged and elevated glucocorticoid treatment inhibits NG2-positive cell proliferation in the adult rat hippocampus [144] reflecting changes in oligodendrocyte precursors or in a distinct mature glial type. Chronic stress also promotes significant structural remodeling of microglia, and can enhance the release of pro-inflammatory cytokines from microglia [227].

Finally, glial cells, especially astrocytes, are key components of the “neurogenic niche” that provides the necessary local microenvironment for generation of neurons in specific brain areas. They support maturation and integration of newborn neurons, both physically and by releasing a cocktail of growth factors and cytokines. Together, this implies that GCs not only are influenced by stress, but also stimulate interactions between astrocytes and neural progenitors.

In humans, post-mortem analysis of tissue from depressed patients has revealed reductions in glial numbers in the amygdala and prefrontal, orbitofrontal and cingulate cortices [20, 32, 149]. Accumulating data demonstrates that not only astrocytic cell numbers are reduced in depressed patients, but several typical structural and functional astrocytic markers as well, like GFAP, gap junction proteins, the water channel aquaporin-4 (AQP4), a calcium-binding protein S100B and glutamatergic markers including the excitatory amino acid transporters 1 and 2 (EAAT1, EAAT2) and glutamine synthetase [167].

In contrast, studies on post-mortem hippocampal samples so far failed to find significant reductions in neuron or glial cell numbers, while confirming the volume reduction in depressed patients [29]. One should add that besides astrocytes and oligondendrocytes, also microglia have also been implicated in pathological mood regulation [55, 60]. In vivo imaging studies in depressed patients demonstrate white matter abnormalities suggesting a contribution of oligodendrocytes to MDD etiology. MRI and also diffusion tensor imaging studies revealed white matter hyperintensities particularly in elderly subjects with late-life depression (reviewed in [55]) which was substantiated by post-mortem data reporting lower density of oligodendrocytes in the MDD brain and a reduced expression of oligodendrocyte-specific gene transcripts [55]. Animal models based on chronic stress paradigms also showed cellular and molecular changes in limbic structures that indicate an involvement of oligodendrocytes [39, 43, 207].

The controversial findings on hippocampal volume decrease without significant cell loss might also be explained by shifts in water content and the volumetric reduction of limbic structures is indeed often accompanied by enlarged cerebral ventricles [97]. In support, significantly shortened T1 relaxation times were found for the hippocampus, especially in elderly depressed patients, indicative of differences in hippocampal water content. A recent study used multimodal in vivo imaging, incorporating structural magnetic resonance imaging (MRI), and MR spectroscopy (¹H-MRS) in rats exposed to ethanol [242]. While MRI revealed expansion of ventricles, volume changes in dorsal or ventral hippocampi, caudate or thalamus were not detected. Also, all MR parameters returned to baseline with 7 days of recovery [242]. Thus, the rapid recovery of ventricular volume and the absence of detectable volume reductions in brain regions adjacent to the ventricles argue against neuronal or tissue atrophy as a mechanism to explain ventricular expansion but rather suggest lower tissue water content. Thus, a rapid fluid redistribution may be followed by compensatory ventricular volume changes in stress-related psychopathologies.

Water homeostasis is largely regulated by aquaporin water channels. In the CNS, their major representative is AQP4 which is expressed in the end-feet of astrocytic processes, but not in neurons [211]. In the human orbitofrontal cortex (Brodmann’s area 47; gray matter), the density of AQP4 immunoreactive astrocytic end-feet was reduced by 50 % in patients with major depression [168] while the coverage of vessels by GFAP-immunoreactive processes did not differ from controls. These data indicate possible disturbances in water homeostasis, at least in this brain region.

Finally, a recent study revealed that stress reduced numbers of microvessels and capillarization in the hippocampi of stressed rats [41, 80] which coincides with clinical studies and metareviews indicating that increased psychological distress and depression are associated with increased stroke risk and mortality [150].