The Relationship Between Acute Glucocorticoid Levels and Hippocampal Function Depends Upon Task Aversiveness and Memory Processing Stage

Abstract

This review evaluates the effects of glucocorticoids (GCs), the adrenal steroids released in response to stress, on memory functions requiring the hippocampus in animals and humans. The data support the hypothesis that the learning function between GCs and hippocampal-dependent memory is modulated by 1) the aversive nature of the learning paradigm and 2) stage of memory processing (acquisition, consolidation, retrieval). When tasks are minimally aversive, the glucocorticoid receptor (GR) mediates an inverted U-shaped relationship between GC levels and hippocampal function, while the mineralocorticoid receptor (MR) mediates attentional processes and/or reaction to novelty. This inverted U-shaped relationship during minimally aversive training paradigms describes GC-mediated memory processing at both acquisition and consolidation. In contrast, highly aversive paradigms activate the amygdala and elevate GCs as part of the training procedure, revealing a nonlinear inverted U-shaped relationship during acquisition and a positive linear function during consolidation. Thus, highly aversive tasks that activate the amygdala shift the memory function from an inverted U-shaped curve to a linear representation between GC levels and memory consolidation.

INTRODUCTION

This review examines how memory mediated by the hippocampus is influenced by stress hormones called glucocorticoids (GCs), steroids such as corticosterone in rats and mice and cortisol in human and non-human primates. The literature reveals that GCs demonstrate a non-linear function with hippocampal-dependent memory during the acquisition and consolidation from minimally aversive tasks: optimal performance occurs at low to moderate GC levels and impaired memory occurs at very low or high GC levels. This non-linear relationship between GC levels and hippocampal-dependent memory has been previously described (Lupien and Lepage 2001; Kim and Diamond 2002). In this review, however, the relationship between GC levels and hippocampal-dependent memory is proposed to shift from an inverted U-shaped function to a positive linear function when GC manipulations occur during memory consolidation and the task is sufficiently aversive to activate the amygdala. Amygdalar activation is proposed to be a critical determinant in the facilitation of memory consolidation by high levels of GCs.

The current review emphasizes the acute actions of GCs in the adult male, and not long-term GC actions and/or developmental influences. These “activational actions” of GCs refer to reversible events occurring after a single exposure to GCs in the adult. Brief exposures to GCs are thought to be beneficial for adaptation to the environment. In contrast, prolonged exposure to GCs can be detrimental, leading to long-lasting changes that predispose organisms to disease (McEwen et al., 1997). Thus, mechanisms that underlie the response to acute and chronic GC exposure are different (for review, see McEwen 2000), and this critique focuses on acute GC exposure.

THE HYPOTHALAMIC-PITUITARY-ADRENAL (HPA) AXIS

The HPA axis represents the anatomical regions involved in the hormonal cascade that eventually triggers the release of GCs in response to a stressor (for review, see Dallman et al., 1987; de Kloet 1991). When a stressor is initially detected, the hypothalamus releases corticotrophin releasing hormone (CRH) into the local hypophyseal portal blood system. CRH triggers the anterior pituitary to secrete adrenocorticotropin hormone (ACTH), which then stimulates the adrenal cortex, located near the kidneys, to release GCs into the bloodstream. Due to this multi-step hormonal cascade, the rise of GC levels in response to a stressor occurs relatively slowly over many minutes. GC release is regulated by potent negative-feedback at the anterior pituitary, hypothalamus, and hippocampus, a limbic structure involved in learning and memory. The hippocampus contains one of the highest concentrations of receptors for GCs in the brain (McEwen et al., 1968, 1969), which suggests that the hippocampus is sensitive to changes in GC levels and that GCs may significantly impact hippocampal function.

Two receptors mediate GC actions on brain function: the mineralocorticoid receptor (MR or Type I) and the glucocorticoid receptor (GR or Type II). Within the hippocampus, the binding affinity of GCs to MRs is nearly ten-fold higher than to GRs (Veldhuis et al., 1982). The GC occupancy of hippocampal MR is consistently high even during nonstress (approximately 70% to 90%), whereas the occupancy of hippocampal GRs fluctuates between 10% and 90% as a function of stress or the circadian rhythm (Reul and de Kloet 1985; Reul et al., 1987; de Kloet et al., 1993a). The ability of hippocampal GR to detect large differences in GC levels has led to the hypothesis that hippocampal GR mediates the GC signal for stress responses (de Kloet and Reul 1987).

PARADIGMS USED TO INVESTIGATE GC INFLUENCE ON HIPPOCAMPAL FUNCTION

The hippocampus is an integral part of spatial memory processing, whereby multiple cues are used to navigate within an environment. How the hippocampus represents the environment is debatable with several prominent theories that include: cognitive mapping (O’Keefe and Nadel 1978), configural versus elemental associations (Rudy and Sutherland 1995), and flexible relations of multiple versus individual representations (Eichenbaum et al., 1990). Regardless of how the information is represented, spatial mazes are very sensitive to hippocampal system disruptions. Examples of spatial mazes include the radial arm maze (Olton et al., 1978), Morris water maze (Morris et al., 1982), radial arm water maze (Diamond et al., 1999), and Y-maze (Conrad et al., 1996). Spatial abilities require rodents (typically rats and mice) to locate a goal by using extra-maze (distal) cues. Rats with hippocampal lesions fail to remember the goal location when extra-maze cues are essential for navigation. In contrast, rats with hippocampal lesions readily locate the goal when it is visible or when the start and goal locations are held constant. These studies show that hippocampal damage impairs place learning (complex representations), but spares response learning (simple representations).

Declarative (explicit) memory is proposed to be a broader domain of hippocampal-dependent memory that encompasses spatial memory (Cohen and Eichenbaum 1991; Squire 1992) in humans (Zola-Morgan et al., 1986) and non-human primates (Zola et al., 2000). Declarative memory refers to the conscious recall of everyday facts and events (Cohen and Eichenbaum 1991) and involves a temporal component (Eichenbaum et al., 1994). As suggested by Eichenbaum, the hippocampus is required during the intermediate period when the relationship between events is processed, but is not necessary for short- or long-term storage of this information. For instance, hippocampal damage does not disrupt immediate recall of declarative memory, nor the long-term storage and recollection of facts learned before (retrograde) hippocampal damage. However, hippocampal damage impairs the long-term storage of newly-learned facts (anterograde amnesia). Hippocampal damage also disrupts working memory, which is the short-term representation of information required for only the current trial, while sparing reference memory, the long-term representation of information required over many trials. Hippocampal lesions impair working memory when complex representations of the environment are required to navigate (place learning), but not when simple representations are used (response learning). Immediate recall in humans parallels response learning in rodents, whereby information lacks complex representations and hence is hippocampal-independent.

Hippocampal integrity is also necessary when animals must recognize environments that were previously paired with an aversive event. For instance, a previously neutral environment can be paired with footshock to produce an association between the environment and the aversive stimulus. The training environment acts as a conditioned stimulus in both passive avoidance (Sahgal 1993; Lorenzini et al., 1996) and contextual fear conditioning (Kim and Fanselow 1992; Phillips and LeDoux 1992; Maren and Holt 2000). For passive avoidance, rodents are given a brief footshock upon entering the preferred dark side of a two-compartment chamber. After a delay, memory for the aversive event is determined by reluctance (increased latency) to enter the dark compartment. For contextual fear conditioning, rodents are exposed to brief footshocks without the opportunity to escape. After a delay, memory for the aversive event is determined by enhanced freezing in the previously aversive environment. A potential problem is that memory for the aversive environment can be supported by both hippocampal and non-hippocampal systems (Penick and Solomon 1991; Kim and Fanselow 1992; Phillips and LeDoux 1994; Young et al., 1994; Gale et al., 1996; Hall et al., 1996; Logue et al., 1997; Maren et al., 1997). Specifically, the hippocampus is involved when processing of the environment uses complex representations (place learning), but not when a single cue representation (response) is used. The difficulty is determining when the hippocampus contributes to performance because both strategies (place and response) have the same behavioral outcome: enhanced latency (passive avoidance) and freezing (contextual conditioning). Thus, findings from multiple paradigms are helpful in interpreting the literature.

Hippocampal function is also necessary for trace conditioning, when a temporal gap occurs between a neutral and an aversive stimulus (Moyer Jr. et al., 1990; Weiss et al., 1999). The trace conditioning studies in this review presented rats with white noise (250 ms) as the conditioned stimulus, followed 500 ms later by periorbital eyeshock (100 ms) as the unconditioned stimulus (Beylin and Shors 1998; Beylin and Shors 2003). Hippocampal function is required when the delay between the conditioned and unconditioned stimulus is 250–500 ms, but not when the delay is removed.

Hippocampal function can be investigated at the cellular level using long-term potentiation (LTP), long-term depression (LTD), and primed burst potentiation (PBP). LTP, PBP and LTD exhibit long-lasting changes in synaptic activity after high frequency stimulation of neuronal afferents. These paradigms are believed to model some aspects of learning and memory (Teyler and DiScenna 1987; Bliss and Collingridge 1993) because they are rapidly induced, correlate with behavioral learning (Berger 1984; Morris et al., 1986), and share some common cellular mechanisms with spatial learning (Garcia 2001). This model can be studied in vivo and in vitro within the hippocampus, allowing the influence of GCs to be observed in the whole system (e.g. anesthetized or behaving animals) and in isolation (e.g. hippocampal slices).

THE DUAL ROLE OF GCS IN MEMORY FUNCTION: THE INVOLVEMENT OF THE AMYGDALA IN HIGHLY AVERSIVE TASKS

As reviewed by de Kloet and colleagues (1999), the manner in which GCs influence memory depends upon the context during information processing. GCs tend to facilitate memory on highly aversive paradigms, including passive avoidance, trace conditioning, and fear conditioning, while they impair memory on less aversive tasks, such as spatial mazes and non-emotional declarative memory. In specific examples, rats that were chronically stressed or exposed to stress levels of GCs showed facilitated performance on highly aversive fear conditioning (Conrad et al., 1999b; Conrad et al., 2004), and impaired spatial memory on the less aversive Y-maze (Conrad et al., 1996). Facilitated performance in highly aversive or emotionally arousing tasks has been observed across many paradigms using a single exposure to stress levels of GCs, including passive avoidance (Flood et al., 1978; Gibbs and Ng 1984), trace conditioning (Beylin and Shors 1998; Beylin and Shors 2003), and cued recall of emotionally arousing pictures in humans (Buchanan and Lovallo 2001). In contrast, stress levels of GCs impaired Y-maze spatial memory in rats (Conrad et al., 1999a) and emotionally neutral declarative memory in humans (Newcomer et al., 1994; Kirschbaum et al., 1996; Newcomer et al., 1999; Plihal and Born 1999; Wolf et al., 2001; Monk and Nelson 2002). As described earlier, hippocampal function is required for successful performance in these highly aversive tasks. However, the neural network underlying successful performance in highly aversive and less aversive tasks is not identical. Thus, the contribution of unique neural substrates under these two distinct conditions is a factor in GC modulation of hippocampal function.

The amygdala may play a greater role under highly aversive conditions than under less aversive conditions and is thought to be an underlying component in emotional responses (LeDoux 1992; Cahill and McGaugh 1996). In one study, performance in stressful (cold water) and nonstressful (warm water) conditions of the Morris water maze was compared to determine the contribution of the hippocampus and amygdala (Akirav et al., 2001). The joint activation of the amygdala and hippocampus occurred when rats performed well in the maze under the stressful condition. In contrast, only the hippocampus was activated in rats that did not learn the maze under the stressful condition, and in rats that learned the maze during the nonstressful condition. Another study found that rats tested in the cold water condition on the water maze had higher GC levels and performed better relative to rats tested in the warm water condition that exhibited lower GC levels (Sandi et al., 1997). Moreover, administering post-training GCs to rats tested in the warm water condition improved spatial memory (Sandi et al., 1997). These findings may also help to explain potential inconsistencies in the literature. For instance, passive avoidance tasks typically use footshock as an aversive stimulus, which is interpreted as being equally aversive across passive avoidance tasks. However, a study by Bohus and colleagues (1970) showed that rats exhibited impaired passive avoidance when injected with 2 mg/100 gm of GCs prior to 0.5 mA footshock, but showed functional passive avoidance when the same amount of GCs were given prior to 1.0 mA footshock. The lower footshock intensity may not have been sufficiently aversive under these testing conditions. Thus, potentially anomalous findings may be explained by how subjects perceive a given task, with tasks perceived as highly aversive causing more amygdala activation.

In another elegant series of studies, hippocampal LTP was examined after exposure to electrical stimulation of the amygdala or swim stress (Akirav and Richter-Levin 1999). Amygdala stimulation 30 seconds before testing facilitated hippocampal LTP, while amygdala stimulation 1 hour before testing impaired hippocampal LTP. Additionally, stress exposure modified the influence of the amygdala by blocking the facilitatory action of amygdala stimulation and impairing hippocampal LTP. These data show that: 1) the activation of the amygdala facilitates or inhibits hippocampal function depending upon when activation occurs, and 2) stressors modulate this process. In summary, all of these studies illustrate that animals perform well when GCs and the amygdala are activated as a consequence of the training conditions. Reviews have described these emotionally arousing conditions as “training conditions that consequently result in enhanced GC levels during the post-training period” (Sandi 1998), which contrasts with “distracting” stressors that are out of context of the original learning task (de Kloet et al., 1999) or unrelated to the task (Wolf 2003). Thus, the beneficial effects of GCs may arise under sufficiently stressful or emotionally arousing conditions during the training procedure.

Many reviews have discussed the role of the amygdala in modulating the storage of memory (Cahill and McGaugh 1996; McGaugh and Roozendaal 2002), including memory mediated by the hippocampus (Roozendaal et al., 1997; Ferry and McGaugh 2000; Richter-Levin and Akirav 2000; Roozendaal 2000; McGaugh 2002). GCs can directly influence amygdala function because the amygdala contains GRs and MRs (Warembourg 1975; Sarrieau et al., 1985; Reul and de Kloet 1986). Intra-amygdala infusions of GR agonists dose-dependently enhanced passive avoidance (Roozendaal and McGaugh 1997a; Roozendaal et al., 2002), and lesions of the basolateral amygdaloid (BLA) subnucleus or its efferents blocked the GR agonist-induced facilitation in passive avoidance (Roozendaal and McGaugh 1996; Roozendaal et al., 2001) and in the Morris water maze (Roozendaal et al., 2003). Additionally, many studies have demonstrated a reciprocal relationship between GCs and neuromodulatory factors. In passive avoidance, for example, GCs reduced or blocked the facilitatory actions of adrenaline (Borrell et al., 1984) and nootropics/cholinomimetics (Mondadori et al., 1992). Moreover, intra-amygdala infusion of β-adrenoceptor antagonists (Quirarte et al., 1997; Roozendaal et al., 2002), cholinergic antagonists (Power et al., 2000), and protein kinase A inhibitors (Roozendaal et al., 2002) blocked the facilitation of passive avoidance from post-training GC administration. Therefore, highly aversive tasks such as passive avoidance and classical conditioning have the added complexity that the training procedure enhances emotionality, amygdala contribution (Cahill 2000), and GC release (Sandi 1998), thereby facilitating hippocampal-dependent memory.

STAGE OF MEMORY PROCESSING

The influence of GCs on memory also depends upon the specific phases of memory processing, including acquisition, consolidation, and retrieval (for review, see Lupien and McEwen 1997; Roozendaal 2002). Acquisition refers to the accumulation and integration of information. Consolidation involves strengthening of the learned information and assumes that processes underlying information storage are initiated at training and continue for some time following the completion of training (for review, see McGaugh 1989). Finally, retrieval is the recollection of previously learned information. The presence of GCs during each phase of memory processing determines whether GCs enhance, impair or have no effect on memory.

Stress levels of GCs prior to acquisition impair memory on all hippocampal-dependent tasks, regardless of task aversiveness. Adrenalectomized rats injected with the GR agonist RU362 before training showed impaired spatial memory on the Y-maze (Conrad et al., 1997). The highest dose or stress levels of GCs prior to training impaired passive avoidance in rats (Bohus et al., 1970; Kóvacs et al., 1977) and chicks (Sandi and Rose 1994). A problem with administering GCs prior to acquisition is that separating the effects of GCs on acquisition and consolidation becomes difficult: GCs remain elevated during acquisition and for some time afterward. Consequently, some effects of GCs on acquisition may carry over to influence consolidation. The process of elimination can be used by comparing pre-and post-training effects of GCs on hippocampal function.

In contrast to acquisition, the effects of stress levels of GCs on memory consolidation are dependent upon the aversiveness of the task. On minimally aversive tasks, immediate post-training injections of GC agonists in adrenalectomized rats impaired Y-maze performance (Conrad et al., 1997). GR antagonists given to intact rats before or after training impaired water maze performance (Oitzl and de Kloet 1992). In humans, post-training infusions (i.v.) of cortisol impaired declarative memory for information learned before GC administration (retrograde impairment, Wolf et al., 2001). Conversely, in highly aversive passive avoidance, the majority of studies find that post-training stress levels of GCs improve memory. Post-training intra-hippocampal infusions of GC agonists facilitated passive avoidance in rats (Roozendaal and McGaugh 1997b). GC injections up to one hour post-training enhanced passive avoidance in chicks (Sandi and Rose 1994) and rats (Flood et al., 1978). Stimulating GC secretion by injecting ACTH also enhanced passive avoidance in rats immediately following training (Gold and Van Buskirk 1976). In humans, elevating GCs by immersing the forearm into cold water enhanced declarative memory for emotionally arousing facts, but not declarative memory for neutral facts (Cahill et al., 2003). The opposing actions of GCs on spatial mazes and highly arousing tasks may be due to the recruitment of the amygdala as described earlier. For instance, BLA lesions blocked the facilitatory actions of intra-hippocampal-infusion of GC agonists on passive avoidance (Roozendaal and McGaugh 1997b). Thus, GCs impair memory consolidation on minimally aversive tasks and facilitate memory consolidation on highly aversive tasks.

Additionally, memory retrieval is impaired by stress levels of GCs. Stress levels of GCs or metyrapone administration impaired retrieval in the water maze in intact rats (Roozendaal et al., 1996a; de Quervain et al., 1998) and declarative memory in humans (de Quervain et al., 2000). The impairment of memory retrieval by GCs appears to be time-dependent (de Quervain et al., 1998) and selective for hippocampal systems because GCs do not disrupt immediate recall or recognition memory in humans (de Quervain et al., 2000). Adrenalectomized rats injected with selective MR and GR agonists prior to retrieval exhibited enhanced retrieval in the Y-maze (Conrad et al., 1997). However, adrenalectomized rats injected with vehicle also performed well, suggesting that the injection procedure was arousing and benefited performance regardless of treatment condition. Another study discovered that administration of MR but not GR antagonists before retrieval altered water maze search strategies (Oitzl and de Kloet 1992). Impaired memory retrieval occurs when GCs are elevated approximately 30 to 90 minutes prior to retrieval assessment. Comparisons between low and highly aversive tasks are not yet available for memory retrieval as the influence of GCs on retrieval is a relatively recent discovery.

Overall, the literature reveals an interesting pattern for how stress levels of GCs influence hippocampal-dependent memory. The majority of the findings indicate that stress levels of GCs impair memory at acquisition regardless of task aversiveness and at retrieval, at least for less aversive tasks. Unfortunately, there is no data regarding the effects of GCs on retrieval for highly aversive tasks. However, stress levels of GCs have opposing actions during memory consolidation with GCs impairing consolidation on less aversive tasks and facilitating consolidation on highly aversive tasks. The emphasis of the studies described so far has been on a single dose of stress levels of GCs. However, the amounts of GCs released by stressful experiences do not always reach maximum levels. Thus, stress levels of GCs are just one component from a continuum of GC levels that can influence hippocampal-dependent memory.

Influence of Memory Processing Stage and Task Aversiveness on GC Dose-Response Function and Memory

The dose-response function that represents GC levels and hippocampal-dependent memory depends upon the stage of memory processing and the aversive nature of the task. During acquisition, the effects of GCs on hippocampal function are described by an inverted U-shaped relationship, which is independent of task aversiveness. In contrast to acquisition, the function representing the effects of GCs on hippocampal ability during memory consolidation depends upon task aversiveness: during minimally aversive tasks, the inverted U-shaped curve continues to represent GC levels and memory, whereas during highly aversive tasks, GCs and memory exhibit a positive linear relationship. Because less aversive tasks demonstrate an inverted U-shaped function at every stage of memory processing, separating the effects of GCs on each stage of memory processing is not necessary. However, the presence of GCs at each stage of memory processing is critical for interpretation only for highly aversive tasks.

Many studies support the hypothesis that GCs have nonlinear effects on hippocampal function on tasks that are minimally aversive. For spatial tasks, low to moderate levels of GCs can benefit memory. Long-term adrenalectomized rats replaced acutely with moderate levels of GCs in the drinking water during water maze training showed improved or restored performance (Conrad and Roy 1995; McCormick et al., 1997). Another study found that low levels of GCs given continuously via drinking water enhanced performance in the Morris water maze early in training, despite serum GC levels that were indistinguishable from controls (3.7 μg/dl, Bennett et al., 1996). Perhaps the continual presence of GCs provided a strong negative feedback signal that prevented a stress-induced GC surge, allowing for optimum performance. In contrast, the impairing effects of GCs occur when GCs reach extremely high or low levels, such as those achieved by high stress, adrenalectomy, or pharmaceutical manipulation. In adrenalectomized rats replaced with stress levels of corticosterone, spatial memory was impaired on the water maze (Roozendaal et al., 1996b) and Y-maze (Conrad et al., 1999a). Adrenalectomy or metyrapone administration without GC replacement impaired spatial memory on the radial arm maze (Vaher et al., 1994), water maze (Oitzl and de Kloet 1992; Roozendaal et al., 1996a; Roozendaal et al., 1996b), and Y-maze (Conrad et.al., 1997). Moreover, pre-training intra-hippocampal infusions of GR antagonists impaired spatial memory on the water maze (Roozendaal and McGaugh 1997b). In young human adults, metyrapone increased the rate of forgetting on a declarative memory task (Lupien et al., 2002). Altogether, these reports show an inverted U-shaped function between GCs and hippocampal-dependent memory on minimally aversive tasks.

Yau and colleagues (1995) published the first behavioral findings showing a non-linear relationship between corticosterone and spatial memory in young and aged rats. For young rats, morning serum corticosterone levels had a tendency to be positively related to spatial memory performance, whereas evening levels were unrelated to performance. In contrast to the findings with young rats, aged rats showed a negative correlation between morning corticosterone levels and spatial memory. The investigators found that aged rats had greater diversity in serum corticosterone levels and spatial memory ability. Furthermore, aged rats that demonstrated the best spatial memory had lower serum corticosterone levels that more closely resembled a young rat profile, and aged rats with the highest corticosterone levels exhibited the worst spatial ability. The average morning peak of serum corticosterone levels for the aged rats was 6 μg/dL, which was significantly higher than for the young rats (2 μg/dL). Thus, nonstress levels of corticosterone (2 μg/dL) and very mild stress levels of corticosterone (up to 6 μg/dL) allowed for functional spatial memory.

Models of hippocampal synaptic plasticity that are believed to parallel hippocampal learning show an inverted U-shaped function similar to that described for GC levels and spatial memory. In a seminal study by Diamond and colleagues (1992), adrenalectomized rats replaced with corticosterone showed maximum PBP induction when serum corticosterone levels were between 11 and 20 μg/dL. In contrast, a positive correlation between PBP and serum corticosterone was observed with low levels of corticosterone (0–10 μg/dL), and a negative correlation between PBP and serum corticosterone occurred with high levels of corticosterone (20 μg/dL or greater). In a complementary study, intact rats received stress levels of corticosterone or the corticosterone synthesis inhibitor metyrapone, and induction of LTP was investigated (Kerr et al., 1994). Rats with moderate levels of serum corticosterone (15 μg/dL) exhibited the most robust LTP induction compared to rats with serum corticosterone levels that were very low (7 μg/dL) or high (25 μg/dL). Thus, different techniques (replacement versus chemical blocking) to manipulate serum corticosterone levels found that moderate levels (11–20 μg/dL) enhanced PBP/LTP, while lower or higher levels impaired PBP/LTP.

During acquisition in highly aversive tasks, GC levels and hippocampal function exhibit an inverted U-shaped function as well. The possibility that footshock intensity has an inverted U-shaped relationship with learning was shown nearly 100 years ago (Yerkes and Dodson 1908). Only one mouse learned to discriminate between two chambers after 15 trials when one chamber was paired a mild footshock, whereas all mice learned in less than 10 trials following a higher footshock intensity. The highest footshock intensity, however, delayed learning in all mice. One-trial learning paradigms that have been used since then show an inverted U-shaped relationship with GC levels and memory. Pre-training injections of GCs at 1 and 5 mg/kg enhanced passive avoidance, while high doses (10 and 25 mg/kg) impaired performance (Kóvacs et al., 1977). A dose-response function of GCs given before passive avoidance training in chicks found moderate doses of GCs facilitated performance, whereas the highest and lowest doses did not (Sandi and Rose 1994). In contextual fear conditioning, adrenalectomy impaired performance and GC replacement in the drinking water restored performance (Pugh et al., 1997b). Thus, intermediate levels benefit and very low or high levels of GC impair memory acquisition on highly aversive tasks.

The influence of GCs during memory consolidation on highly aversive tasks appears to be characterized by a positive linear function. In one study, a positive correlation was observed among footshock intensity, GC levels, and contextual fear conditioning (Cordero et al., 1998). The higher the footshock intensity, the more GCs were secreted, which facilitated contextual fear conditioning. Other studies found that post-training GR agonists (s.c.) facilitated passive avoidance in sham-operated rats (Roozendaal and McGaugh 1996) or post-training intra-amygdala or hippocampal GR agonist infusions dose-dependently facilitated passive avoidance (Roozendaal and McGaugh 1997a, b). However, not all reports show a linear function between GC levels and memory consolidation. As early as 1976, Gold and Van Buskirk found that post-training injections of ACTH, the hormone that signals GC release, enhanced passive avoidance at moderate doses and impaired it at high doses. Contextual fear conditioning showed a nonlinear function with dose of GCs injected (Pugh et al., 1997b): adrenalectomized rats given moderate doses of GCs (0.25 mg/kg or 1 mg/kg) post-training showed the best contextual conditioning, while the lowest and highest dose of GCs (0 and 2.5 mg/kg, respectively) impaired contextual conditioning. Serum levels of GC within 30 minutes after injection were 15 μg/dL, 40 μg/dL, and 80 μg/dL for the rats injected with 0.25 mg/kg, 1 mg/kg, and 2.5 mg/kg, respectively. A possible explanation for these contradictory results may be the perceived aversiveness of the task. First, higher footshock intensity produces better retention, which correlates with high GCs (Cordero et al., 1998). These data show that better retention occurs when the task is sufficiently aversive, but presenting a shock does not guarantee that the task will be perceived as aversive. In the passive avoidance study by Gold and Van Buskirk, the footshock was high (.7mA), but was presented for only .35 s, which may be an insufficient duration to be perceived as highly aversive. Second, the contribution of the amygdala may help to determine the aversiveness of a task. BLA lesions blocked the memory-enhancing effects of GCs on passive avoidance (Roozendaal and McGaugh 1997b). Thus, for conditions in which passive avoidance was perceived as aversive, GCs demonstrate a linear relationship with memory consolidation.

In conclusion, GCs have a complex relationship with hippocampal-dependent memory, and this function depends upon the stage of memory processing and the aversive nature of the task. Under minimally aversive conditions, GCs have an inverted U-shaped relationship with hippocampal-dependent memory at all stages. For highly aversive tasks, GCs exhibit an inverted U-shaped function with hippocampal-dependent memory during acquisition and a positive linear function during consolidation. Potential inconsistencies may arise when subjects do not perceive the task as aversive or when the amygdala is not sufficiently involved. Thus, GC levels exhibit an inverted U-shaped relationship with memory during minimally aversive tasks when the amygdala is not activated. When the task is sufficiently aversive to activate the amygdala during memory consolidation, the relationship between GCs and memory becomes a positive linear function.

INVOLVEMENT OF THE GR AND MR IN MEDIATING THE NON-LINEAR RELATIONSHIP BETWEEN GCS AND HIPPOCAMPAL FUNCTION ON MINIMALLY AVERSIVE TASKS

Both GR and MR have the potential to mediate the inverted U-shaped relationship between GCs and hippocampal function in minimally aversive tasks and in models of hippocampal plasticity and learning. The GR is critical for the impairing effects of GCs on hippocampal function, which are represented by the negative slope of the inverted U-shaped curve. Several studies have demonstrated that GRs negatively regulate hippocampal excitability, LTP, and PBP. Direct intra-hippocampal application of dexamethasone, a GR agonist, decreased neuronal activity (Michal 1974). Stress and the GR agonist RU362 enhanced LTD (Pavlides et al., 1995a; Coussens et al., 1997; Xu et al., 1998), while the GR antagonist RU486 blocked LTD (Coussens et al., 1997; Xu et al., 1998). Injections of the GR agonist RU362 in adrenalectomized rats suppressed LTP in vivo (Pavlides et al., 1995b) and in vitro (Pavlides et al., 1996), an effect that was prevented by pre-treatment with the GR antagonist RU486 (Pavlides et al., 1995b). These studies demonstrate that GRs within the hippocampus mediate the actions of GCs on hippocampal LTP. Behavioral studies support these electrophysiological findings: injections of the GR agonist RU362 impaired spatial memory performance on the Y-maze in adrenalectomized rats (Conrad et al., 1997; Conrad et al., 1999a). Thus, hippocampal function is reduced when GRs are highly occupied by stress levels of GCs.

Conversely, activation of the MR facilitates hippocampal function. Injections of the MR agonist aldosterone in anesthetized, adrenalectomized rats enhanced LTP in vitro (Pavlides et al., 1996), in vivo (Pavlides et al., 1995b), and in freely moving animals (Pavlides et al., 1994). The aldosterone enhancement in LTP was abolished by pre-injections of the MR antagonist RU318 (Pavlides et al., 1995b). Again, behavioral studies support the electrophysiological findings: adrenalectomized rats injected with the MR agonist aldosterone, showed functional spatial memory on the Y-maze (Conrad et al., 1997).

The GR and MR have been proposed to have opposing, yet complementary effects that mediate the biphasic actions of GCs on hippocampal function (de Kloet and Reul 1987; de Kloet 1991; Joëls and de Kloet 1992; Conrad et al., 1997). Originally, MR activation was hypothesized to regulate the positive slope of the inverted U-shaped relationship between GC levels and hippocampal function. However, studies have recently indicated a different role for MRs. The GC occupation of GR and MR in the absence of stress should be approximately 10% and 90%, respectively (Reul and de Kloet 1985). It follows that if MR mediates the positive slope of hippocampal function, then blocking GR during non-stress should be insufficient to alter hippocampal function. This hypothesis predicts that adrenalectomized rats given MR agonists will perform similarly to intact rats given GR antagonists, as both conditions exhibit high MR activity. In contrast to this prediction, adrenalectomized rats given the MR agonist aldosterone performed well on the Y-maze, while intact rats injected with the GR antagonist RU555 performed poorly (Conrad et al., 1999a). Furthermore, acute injections of GR antagonists impaired performance on the water maze (Oitzl and de Kloet 1992; Roozendaal and McGaugh 1997b; Oitzl et al., 1998), context conditioning after moderate foot shock (Cordero and Sandi 1998), and passive avoidance (Johnston and Rose 1998). These data show that preventing GR activation, even in the presence of functional MRs, impairs hippocampal function. Therefore, the high activation or blockade of GRs determines spatial memory ability; MR activation alone is not sufficient. A review by Lupien and colleagues (in this issue) proposes that the ratio of MR to GR determines memory ability in humans with high ratios facilitating and low ratios impairing memory function. Thus, blocking GR function would reduce the MR/GR ratio and impair hippocampal function, which supports these experimental observations.

One study using a rewarded spatial maze demonstrated results contradictory to those described above (Douma et al., 1998). Intact rats injected repeatedly with MR antagonist, but not GR antagonist, showed impaired spatial reference memory. According to the hypothesis, intact rats injected with MR antagonist should exhibit intact spatial memory because GRs are functional. Compensatory changes in brain function seem unlikely to account for these findings because antagonists were administered every other day. Moreover, reference memory was altered early in training, which emphasizes the acute actions of the MR antagonist, as opposed to long-term compensatory changes. A major difference in this study was that the rats were rewarded with palatable food, which can increase arousal and motivation. Thus, the possibility that the amygdala altered the dynamics of hippocampal function cannot be discounted and may explain these contradictory results.

What role do MRs have on hippocampal function assessed by spatial memory performance? Previous investigators have suggested a modulatory role through altering attentional processing (for review, see Lupien and McEwen 1997) or reactivity to the environment (de Kloet et al., 1993a; de Kloet et al., 1993b). Such functions would be supportive, but not necessary, to spatial processing and corroborate the behavioral profile (Conrad et al., 1999a). Indeed, the MR antagonist spironolactone did not alter consolidation of spatial memory, but changed search patterns (Oitzl and de Kloet 1992; Yau et al., 1999). In fear conditioning, the MR antagonist RU318 did not affect performance (Cordero and Sandi 1998). In another study, an inverted U-shaped function was demonstrated among exploration in the open field, circulating levels of corticosterone, and MR occupation levels (de Kloet et al., 1993b). This study indicates that MRs regulate exploration in a biphasic fashion, with moderate corticosterone levels corresponding to the most interest in the environment. Thus, MR-induced enhancements in attention may benefit performance. Consequently, blocking MRs does not necessarily impair performance, indicating that the role of the MR in hippocampal function is supportive but not necessary.

One question that remains is how GR and MR regulate hippocampal function. The GR and MR could work independently to benefit hippocampal function by concurrently activating separate systems. For example, GR activation could enhance memory consolidation, while MR activation could enhance attention; the combined activation of both GR and MR may be additive. In another scenario, GR and MR activation could work synergistically to benefit memory. Support for a GR/MR synergism leading to enhancement of hippocampal function has not been demonstrated behaviorally, but is suggested by electrophysiological studies. The GR- and MR-induced excitability of hippocampal potassium or calcium channels was greater when both GR and MR were activated together than the sum of either GR or MR activation alone (Karst et al., 1994; Hesen and Joëls 1995). The GR and MR are highly concentrated and co-localized within the hippocampus (Reul and de Kloet 1986; Van Eekelen et al., 1988), which suggests that they have the potential to act synergistically within the same neurons or in adjacent cells. Whether GR and MR synergism occurs through separate systems or within the same cell or region (e.g. hippocampus) is not clear and requires investigation.

Several other important concepts are revealed by the electrophysiological studies and behavioral paradigms. First, electrophysiological studies demonstrate that synaptic excitability occurs locally within the hippocampus in response to MR and GR manipulation. Some preparations involve the addition of steroids to hippocampal slices, showing that the actions of steroids occur within the hippocampus and not at other targets that modulate the hippocampus. Moreover, studies have shown that behaviors relying upon hippocampal function are altered by direct infusion of GR antagonists or agonists into the hippocampus (Michal 1974; Cottrell and Nakajima 1977). Second, comparing performance on hippocampal-dependent and independent tasks complements the hypothesis that GCs act locally within the hippocampal system. For instance, GR antagonists impair contextual conditioning, a form of learning that requires hippocampal function, without disturbing auditory cue conditioning, which does not require the hippocampus (Pugh et al., 1997a). Finally, several strains of transgenic mice have been developed with impaired GR activity (Pepin et al., 1992; Cole et al., 1995). These mice showed performance deficits on tasks requiring hippocampal function (allocentric or spatial strategy), with no impairment when navigation could be performed without the hippocampus (egocentric or response strategy; Oitzl et al., 1997; Rousse et al., 1997; Steckler et al., 1999; Oitzl et al., 2001). Thus, the influence of GR and MR on hippocampal function under these conditions may involve local actions within the hippocampal system.

SIGNIFICANCE AND CONCLUSION

Stress levels of GCs facilitate hippocampal-dependent memory under highly aversive conditions while impairing memory under less aversive conditions. Under highly aversive conditions that activate the amygdala, the information remembered at this time may have important life-saving repercussions. For example, a rat may have located food near a predator habitat. Remembering the environmental cues surrounding the predator will minimize future exposure to this potentially life-threatening event. Recognition of the predator should activate the amygdala to identify the situation as potentially threatening, and the degree of GC secretion should determine how well the encounter is remembered. If the predator was indifferent to the rat, moderate GC levels may moderately facilitate memory. However, if the predator pursued the rat, then high GC levels may maximize memory retention. A caveat is that remembering and avoiding all scenarios of predator exposure may be counter-productive, because the rat may never eat when constantly avoiding predators. Thus, selecting the appropriate situations to remember better (predator pursuit) than others (predator presence only) will be most cost-effective for the rat. The degree of GC secretion during a highly aversive event may help determine how well information surrounding the event is remembered.

For minimally aversive conditions that do not require amygdala involvement, stress levels of GCs impair memory. By definition, the event is not life-threatening when the amygdala is not activated, and stress levels of GCs may impair spatial memory under these conditions as a cost-effective strategy. For example, a rat foraging for food will explore many environments and may encounter a predator in one of them. The potentially life-threatening predator stimulates a stress response and GC release in the rat. The spatial representation of the previous, safe environments in the process of consolidation are susceptible to perturbations and the cost of maintaining consolidation for these environments is hypothesized to be an inefficient use of resources after encountering the predator in one area. In this scenario, the rat would benefit most by maintaining a representation of the predator habitat at the expense of hindering representation of the safe areas. Thus, stress levels of GCs will impair spatial memory of the safe environments while facilitating memory of the life-threatening predator habitat.

The purpose of this review was to evaluate the complex effects of acute GC exposure on hippocampal function. A pattern observed in the literature is that GCs have a non-linear relationship with hippocampal function, and amygdala contribution modifies this relationship. GCs facilitate memory consolidation under highly aversive conditions when GC secretion is intrinsic to the training procedure. Amygdala involvement during highly aversive conditions is hypothesized to be a critical factor in determining whether memory is facilitated by GCs. Under highly aversive conditions, GCs have a positive linear function with memory ability, which plateaus at maximum memory retention (see Fig. 2B from Sandi, 1998). In contrast, minimally aversive conditions that do not involve the amygdala demonstrate an inverted U-shaped relationship with serum GC levels, spatial memory, and synaptic plasticity models using LTP/PBP. During minimally aversive conditions, low to moderate levels of GCs (2–20 μg/dL serum blood) are beneficial, whereas extremely low or high levels are detrimental. The GRs appear to mediate the influence of GCs on hippocampal function through memory consolidation and may act locally within the hippocampus. In contrast, the MRs may benefit hippocampal function through indirect processes that influence attention or environmental reactivity. Electrophysiological data suggest that the MR and GR work synergistically to influence hippocampal function, but behavioral studies investigating this hypothesis have not yet been conducted.