Effects of Stress on Different Memory Systems

There is a general agreement that there is not a single memory system in the brain but multiple interacting memory systems. Abundant literature exists on this topic, and it is not the goal of the current review to provide an in-depth overview of different models and ideas; the important question of the comparability of different memory systems in different species is not addressed here either.

I focus instead on declarative or explicit long-term memory because this domain has been, until now, investigated most intensely with respect to stress effects. In addition, I briefly outline the findings for procedural memory, working memory, and conditioning (or associative learning); these three memory systems have been sparsely investigated, at least in humans.

Declarative Memory. Acute effects Declarative memory is the explicit storage of facts and events, which can later be intentionally retrieved. It is a relational and flexible memory system. This type of memory is tested in the laboratory with word lists, paired associates, or short stories in humans. In animals, spatial maze tasks are used most often, the most famous one being the Morris water maze (MWM).

Long-term memory can be further divided into different memory phases: acquisition (or initial learning), consolidation (or storage), and retrieval (or recall). For the successful completion of declarative tasks, an intact medial temporal lobe region (the hippocampus and surrounding cortical structures) is essential for the acquisition of the task.

The role of the hippocampus in retrieval is under debate, and it might fulfill a time-limited role only. Especially in humans, it is the prefrontal cortex regions that are of great importance for (effortful) retrieval. The literature regarding the effects of stress on declarative memory is quite diverse and confusing, with groups reporting enhancing as well as impairing effects of GCs on this form of memory. However, it has become apparent that this is largely due to the fact that the different memory phases are modulated by GCs in an opposing fashion.

GCs enhance memory consolidation, and this aspect represents the adaptive and beneficial mode of GCs central nervous system action. Figure 1a depicts the chain of events underlying this phenomenon. This process has been conceptualized as the beneficial effects of stress within the learning context or intrinsic stress, emphasizing that a stressful episode is remembered better.

This effect is mediated by the action of stress-released GCs on the hippocampal formation, and it is very well documented in animals. Studies by Roozendaal and McGaugh showed that adrenergic activationinthe basolateral amygdala (BLA) appears to be a prerequisite for the modulating effects of GCs on other brain regions (e.g., the hippocampus). Lesions of the BLA, as well as ß blockade, abolish the enhancing effects of posttraining GC administration.

Although the enhanced memory consolidation is adaptive and beneficial, this process appears to occur at the cost of impaired retrieval (see Figure 1b). Using a 24-h delay interval, deQuervain and colleagues were the first to show that stress or GC treatment shortly before retrieval testing impairs memory retrieval in rats in the MWM.

Further studies revealed that, again, an intact BLA and adrenergic activation appear to be necessary for the occurrence of this negative GC effect. Roozendaal summarized these findings as indicating that stress puts the brain into a consolidation mode, which is accompanied by impaired retrieval. Such a retrieval reduction might facilitate consolidation by reducing interference.

In humans, the literature on the acute effects of cortisol on declarative memory is somewhat divergent. Few studies reported beneficial effects, some reported negative effects, and several others found no acute effects.

We recently conducted a meta-analysis in order to provide a quantitative summary of the current state of the field. The results supported the model derived from animal studies showing that cortisol impairs memory retrieval. Similar observations have been made in humans who were exposed to a psychosocial laboratory stressor shortly before retrieval testing took place.

To date, the beneficial effects on consolidation could not be clearly established using a meta-analysis, which in part might reflect a power problem. At least some studies, using emotionally arousing learning material and a sufficiently long retention delay in order to assure that GCs can influence consolidation and in addition are back to baseline at the retrieval testing, observed enhanced memory consolidation for emotionally arousing material.

Interestingly, the beneficial effects on consolidation as well as the impairing effects on retrieval in humans are more pronounced for emotionally arousing material. This observation fits nicely with the observation in animals that GCs can exert effects on memory only in the presence of adrenergic activity in the amygdala. This arousal could be the result of specifics of the learning material and/or specifics of the testing conditions.

In the meta-analysis, time of day appeared as a second modulatory factor. Studies that administered cortisol before initial acquisition observed impairing effects on memory when they were conducted in the morning, a time of high endogenous cortisol levels in humans. In contrast, studies in the evening were more likely to observe beneficial effects. This supports an inverted U-shaped function between cortisol levels and memory in humans with both too low and too high levels at the time of acquisition being associated with impairments, especially when retrieval is tested at times when cortisol levels are still elevated.

In sum, studies in animals and humans converge on the idea that GCs acutely enhance memory consolidation of emotional arousing material while impairing memory retrieval. In addition, within this framework, emotional arousal and a nonlinear dose-response relationship are important modulatory variables.

Cortisol also influences declarative memory consolidation during sleep. Here the relationships appear to be different than in the daytime. Low cortisol levels during the first half of the night are a prerequisite for a sleep-induced enhancement of declarative memory consolidation. Moreover, blocking the rise in cortisol levels, which typically occurs during the second part of the night, leads to enhanced emotional memory facilitation.

The interesting issue of individual differences in responsiveness to GCs has received little attention so far. Such differences might be able to account for some of the variance observed within as well as between studies. Individual differences could reflect genetic factors, differences in tissue GC sensitivity, or differences in lifetime cortisol exposure. Attempts to characterize subgroups with high versus low cortisol sensitivity could be one important venue for future research.

Chronic effects. The picture is different for more chronic effects as a result of prolonged stress or GC treatment - here negative effects prevail. Chronic stress in animals has repeatedly been shown to result in impaired spatial-memory performance. The neurobiological underpinnings of those effects are still under debate. Stress-induced dendritic atrophy (or dendritic remodeling) in the hippocampus is one possible mechanism for the negative effects of chronic stress on spatial memory in rodents.

However, several additional mediators such as reduced neurogenesis and dysregulation of several catecholamines have to be considered. Recent studies demonstrated substantial plasticity in the adult rodent brain. For example, stress-induced dendritic atrophy appears to be reversible.

Interestingly, the negative effect of chronic stress on spatial memory might occur only in male rats and not in females. In fact, similarly, acute stress appears to impair spatial memory only in males. The area of sex differences in response to stress is an important and still relatively neglected field. However, authors who went through the extra effort of studying male and female animals have often observed quite striking sex differences in the effects of stress on a wide range of target systems.

These observations in combination with evidence for sex differences in emotional memory lateralization and sex differences in the effects of adrenergic manipulations on emotional memory emphasize that caution should be used when generalizing findings obtained in one sex to predict responses in the opposite sex.

In humans, the effects of chronic long-term GC treatment cannot, of course, be studied experimentally due to ethical considerations. Studies that administered GCs for several (3-10) days observed declarative memory impairments.

Moreover, studies of patients receiving GC therapy and studies of hypercortisolemic Morbus Cushing patients observed reduced declarative memory performance, which was associated with hippocampal atrophy. Some treatment studies in Cushing patients suggest that hippocampal atrophy in Cushing is reversible, which fits the observations in rodents previously mentioned.

Aging leads to alterations of HPA activity. Basal activity seems to increase and negative feedback is less efficient. Animal studies have suggested that enhanced HPA activity is associated with poorer memory. In older, otherwise healthy humans, several observational studies also reported associations between elevated or rising cortisol levels and declarative memory impairments.

Whether these associations are specific for declarative memory or are, in fact, broader (also including working memory or attention) is currently under debate. Obviously all these human studies do not allow a clear cause and effect interpretation. Moreover, in most of these studies, the acute effects of cortisol cannot be differentiated from the chronic effects.

In addition, the possible structural correlate of these hormone-performance associations remains to be firmly established. Hence, the possible association between rising cortisol levels and atrophy of the hippocampus is still not sufficiently understood, and the current empirical situation is somewhat mixed.

Finally, it has to be reiterated that other hormonal influences also might be important. With respect to aging, elevated cortisol levels could be part of the metabolic syndrome. Other aspects of this nonfavorable endocrine condition are impaired glucose tolerance and hypertension. These alterations have also been associated with memory impairment during aging, so future studies on this issue should consider taking a broader set of neuroendocrine measurements.