|Home | About | Journals | Submit | Contact Us | Français|
The prefrontal cortex (PFC) mediates higher-order cognitive and executive functions that subserve various complex, adaptable behaviors such as cognitive flexibility, attention and working memory. Deficits in these functions typify multiple neuropsychiatric disorders that are caused or exacerbated by exposure to psychological stress. Here we review recent evidence examining the effects of stress on executive and cognitive functions in rodents, and discuss an emerging body of evidence that implicates the N-methyl-D-aspartate receptor (NMDAR) as a potentially critical molecular mechanism mediating these effects. Future work in this area could open up new avenues for developing pharmacotherapies for ameliorating cognitive dysfunction in neuropsychiatric disease.
Higher-order cognitive and executive functions encompass mental processes that allow for complex, adaptable behavior and include attention, working memory, behavioral inhibition, and cognitive flexibility. Deficits in executive functions are hallmarks of various neuropsychiatric disorders, such as schizophrenia, depression, obsessive compulsive disorder, and drug addictions (Murphy et al., 2003, McLean et al., 2004, Waltz and Gold, 2007). In recent years, the field has placed a growing focus on understanding the mechanisms underlying executive deficits as an approach to designing improved treatments for these disorders (Barnett et al., 2010).
An outstanding question in this context is how exposure to environmental insults such as stress, which is a known risk factor for these disorders (Sinha, 2008, Lupien et al., 2009), impacts executive functions. This then raises the corollary issue of which molecular mechanisms are involved and whether these offer targets for new pharmacotherapies. The goal of the current article is to provide an update on recent research that has attempted to address these questions using rodent preclinical models.
We first provide a brief overview of the literature on stress effects on two relatively well-studied prefrontal cortex (PFC)-mediated executive functions - working memory and cognitive flexibility, as well as on fear extinction, an important PFC-mediated form of emotional regulation. We then discuss preliminary evidence implicating the N-methyl-D-aspartate receptor (NMDAR) as one potential molecular mechanism mediating stress effects on cognition and fear extinction.
Stress, even when relatively mild and brief, can profoundly alter structure and neuronal morphology in rodent PFC (Arnsten, 2009, Holmes and Wellman, 2009, Shansky and Morrison, 2009). This in and of itself strongly implies that stress would impact behaviors that are dependent upon intact PFC functions. Stress effects on rodent cognitive and executive functions have been directly examined using a range of well-characterized behavioral assays (for more detailed descriptions of these tasks, see Robbins, 2007, Floresco et al., 2008, Holmes and Wellman, 2009, Brigman et al., 2010, Chudasama, 2011). Some of the main findings from this literature are summarized in Table 1.
Working memory is the transient storage of recently acquired information. Working memory impairments are produced by medial prefrontal cortex (mPFC) lesions encompassing the prelimbic and infralimbic cortices and occasionally extending to the anterior cingulate cortex (Fritts et al., 1998, Schwabe et al., 2004, Chudasama and Robbins, 2006, Gisquet-Verrier and Delatour, 2006, Sloan et al., 2006) or experimentally-induced alterations in mPFC neurotransmitters such as dopamine (Cools and Robbins, 2004, Arnsten and Pliszka, 2011, Gamo and Arnsten, 2011).
A number of studies have examined the effects of stress on working memory. Mice exposed to chronic cold water stress, or rats subjected to chronic unpredictable stress (CUS) exhibit impaired working memory in the Morris water maze (MWM), radial arm maze and delayed alternation T-maze (Diamond et al., 1999, Mizoguchi et al., 2000, Woodson et al., 2003, Cerqueira et al., 2007, Park et al., 2008, Trofimiuk and Braszko, 2008, Hains et al., 2009). Exposure to a single stress episode, such as fifteen minute exposure to a predator’s odor, is sufficient to impair working memory on assays including the delayed, non-matching-to-sample object recognition task, delayed alternation water T-maze or delayed win-shift radial maze (Morrow et al., 2000, Del Arco et al., 2007, Buchanan et al., 2008, Butts et al., 2011, Del Arco et al., 2011). However there are also reports of no effect of stress (Choy et al., 2008) or even facilitation of working memory after exposure to two hour restraint stress or forced swim stress (Barha et al., 2007, Yuen et al., 2009, Yuen et al., 2011). The reason for these differences is not fully clear but may relate to differences in the type of stress, the duration and chronicity of its application, and the interval between stress and behavioral testing (Anisman and Matheson, 2005, Joels et al., 2006, Joels et al., 2007). Indeed, there remains much to be understood the degree to which these parameters determine the impact of stress on cognition in general.
Stress-induced working memory impairments can be mimicked by treatment with corticosterone or dexamethasone, implicating glucocorticoids in this effect (Cerqueira et al., 2005, Trofimiuk and Braszko, 2008). Indeed, a recent report found that acute tail pinch stress applied specifically during the working memory retention interval in a maze-based task impaired performance and, moreover, that this could be rescued by blocking glucocorticoid receptors in the mPFC (Butts et al., 2011). However, the contribution of other neural mechanisms, including NMDARs, to stress-induced working memory deficits is yet to be elucidated.
Cognitive flexibility, the ability to modify previously learned behavior following changes that alter the outcome of that behavior, can be assayed in rodents in various ways. One simple measure of cognitive flexibility is reversal learning, in which the subject is required to switch responding from one cue or location to another in order to gain a reward or avoid a punishment. By contrast, set-shifting involves more complex stimuli with distinctly perceptible features, e.g. odor and texture, and requires the subject to attend to and switch between these relevant features. Lesions studies in rats and mice have typically demonstrated a dissociation between orbitofrontal cortex (OFC) mediation of reversal and mediation of set-shifting by the mPFC (Birrell and Brown, 2000, Chudasama and Robbins, 2006, Boulougouris et al., 2007, Bissonette et al., 2008).
Rats exposed to CUS or given glucocorticoid treatment subsequently exhibit deficits in reversal learning as measured in the MWM (Hill et al., 2004, Cerqueira et al., 2005, Cerqueira et al., 2007). CUS or chronic intermittent cold exposure also impairs reversal and set-shifting in an odor-texture attentional set-shifting task, an analogue of the human Wisconsin Card Sorting Task, in a manner that is reversible by treatment with the monoaminergic antidepressants desipramine, nomifensine, fluoxetine or citalopram block these effects (Liston et al., 2006, Bondi et al., 2008, Lapiz-Bluhm et al., 2009, Bondi et al., 2010, Danet et al., 2010, Furr et al., 2011, Nikiforuk and Popik, 2011).
Extending these findings, recent work from our laboratory found that under some test conditions, stress can ‘paradoxically’ facilitate reversal learning. We found that mice exposed to three days of forced swim stress showed improved reversal learning in a touchscreen-based operant task, and that this effect was mimicked by lesions of the mPFC (Graybeal et al., 2011). Given lesions of the dorsolateral striatum (DLS) impair reversal learning in this task (Graybeal et al., 2011), one interpretation of these data is that stress-induced impairment of mPFC function disinhibits DLS-mediated learning. This hypothesis is in line with recent work from Dias-Ferreira and colleagues and Schwabe and Wolf showing that stress promotes DLS-mediated learning and the formation of habits (Dias-Ferreira et al., 2009, Schwabe and Wolf, 2009, 2011). Taken together, these studies underscore a key issue, which is that while the PFC may be a particularly sensitive target for stress, the consequences of resultant PFC dysfunction are complex and may not always simply be impairing. Rather, stress effects will manifest at the systems levels as shifts in the balance of control over behavior amongst parallel and even competing brain regions.
Fear extinction is not a measure of executive function but is very salient to the current discussion given the importance of the mPFC to fear extinction (Quirk and Mueller, 2008) and clinical evidence of deficient fear extinction in diseases such as schizophrenia, where PFC dysfunction is prominent (Holt et al., 2008).
In an early indication that stress impairs fear extinction, mice exposed to three days of forced swim were found to have slower extinction, which was coupled with dendritic retraction in mPFC pyramidal neurons (Izquierdo et al., 2006). More prolonged stress regimens involving chronic restraint or corticosterone treatment in adult rats, or maternal separation in neonatal rats, have also been shown to impair extinction and PFC-mediated neuronal encoding of extinction (Miracle et al., 2006, Baran et al., 2009, Gourley et al., 2009, Wilber et al., 2009, Green et al., 2011, Wilber et al., 2011).
Elegant work by Maier and colleagues has shown that the controllability of stress determines the nature of stress effects on the expression of learned fear (Maier and Watkins, 2010). For example, rats exposed to escapable stress subsequently exhibit reduced fear conditioning and decreased fear during fear extinction training (Baratta et al., 2007). The apparent resilience conferred by a history of escapable stress extends to other behaviors (social and depression-related) (Amat et al., 2006, Christianson et al., 2008, Christianson et al., 2009, Amat et al., 2010). Interestingly, this resiliency is dependent upon mPFC (specifically infralimbic cortex) function, as demonstrated by the finding that infralimbic inactivation during stress or testing prevents the fear-reducing effect of escapable stress on fear expression (Baratta et al., 2008), apparently due to loss of regulation of midbrain serotonergic signaling (Amat et al., 2005). However, the potential contribution of glutamatergic signaling, either in mPFC or at the midbrain level, to this effect has not been determined and remains an interesting question for future work.
Glutamate, the main excitatory neurotransmitter, binds to multiple metabotropic receptors and three families of ionotropic receptors: alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPARs), kainate receptors and NMDARs. While all three families have been found to play a role in cognitive and stress-related behaviors, NMDARs represent a particularly intriguing mechanism mediating stress-induced changes in prefrontal-mediated behaviors for a number of reasons.
Pharmacological blockade of NMDARs can have profound effects on PFC-mediated cognitive processes discussed above (Floresco et al., 2008, Brigman et al., 2010, Neill et al., 2010). A number of studies have shown that either acute or chronic treatment with NMDAR antagonists such as ketamine (Nikiforuk et al., 2010, Kos et al., 2011), phencyclidine (Abdul-Monim et al., 2006, Didriksen et al., 2007) or MK-801 (Bardgett et al., 2003, van der Meulen et al., 2003, Watson and Stanton, 2009c, b, a) impairs cognitive flexibility. For example, phencyclidine administration disrupts reversal learning (Abdul-Monim et al., 2006, Abdul-Monim et al., 2007) and ketamine treatment impairs extra-dimensional set-shifting (Nikiforuk et al., 2010, Kos et al., 2011). A similar disruptive effect of ketamine is observed in humans tested in the Wisconsin card sorting task (Krystal et al., 2000). Further reinforcing the translational relevance of the rodent data, clinically efficacious antipsychotics including sertindole, clozapine or risperidone rescue these cognitive-impairing effects NMDAR antagonists (Amitai et al., 2007, Didriksen et al., 2007, Kos et al., 2011).
Systemic treatment with NMDAR antagonists also disrupts working memory (Jentsch et al., 1997, Aura and Riekkinen, 1999, Ko and Evenden, 2009, Willmore et al., 2009, Smith et al., 2011) and produces deficits in attentional performance that reflect increased impulsivity (Higgins et al., 2003, Higgins et al., 2005, Baviera et al., 2008, Amitai and Markou, 2009, 2010, Murphy et al., 2011, Thomson et al., 2011). These effects can be recapitulated by direct infusion of NMDAR antagonists into the mPFC (Aura and Riekkinen, 1999, Baviera et al., 2008), consistent with the hypothesis that NMDARs in this brain region are key mediators of these behaviors.
An important functional characteristic of NMDARs is their heteromeric structure: obligatory GluN1 subunits and one or more GluN2 (GluN2A-GluN2D) or GluN3 subunits (Rosenmund et al., 1998). GluN2A and GluN2B are modulatory in that their inclusion influences receptor kinetics (Cull-Candy et al., 2001). NMDARs containing GluN2A and/or GluN2B are widely expressed in corticolimbic regions, including the mPFC, where they are localized on both pyramidal and interneurons (Cull-Candy et al., 2001).
The behavioral role of GluN2B-containing NMDARs has been studied using GluN2B-selective antagonists such as ifenprodil and Ro 25-6981. Studies utilizing these agents have shown, for example, that systemic GluN2B antagonist treatment impaired reversal learning as assayed in a rat lever pressing task (Dalton et al., 2011a) and in the mouse MWM test (Duffy et al., 2008), and produced impulsivity in an attentional task (Higgins et al., 2003, Higgins et al., 2005). While parallel studies on GluN2A are hampered by a lack of selective drugs for this subunit (Neyton and Paoletti, 2006), the pharmacological literature on this and the other subunits has been extended using mutant mice in which a NMDAR subunit is genetically deleted.
Mice with constitutive genetic inactivation of GluN2A exhibit impairments in operant discrimination and reversal learning (Brigman et al., 2008) and spatial working, but not reference, memory (Bannerman et al., 2008). This working memory deficit phenocopies the effects of deleting GluN1 in a subpopulation of forebrain parvalbumin-expressing GABAergic interneurons (Belforte et al., 2010) or deleting GluN2B in hippocampal principal neurons (von Engelhardt et al., 2008). More pronounced effects are produced by forebrain-wide or corticohippocampal deletion of GluN2B in principal neurons, which produces deficits in object recognition, maze-based spatial working and reference memory, and discrimination learning (Brigman et al., 2008, von Engelhardt et al., 2008).
Conversely, transgenic overexpression of GluN2B throughout the forebrain facilitates fear extinction (Tang et al., 1999) and working memory performance on multiple tasks in concert with enhanced plasticity (long-term potentiation) at PFC synapses (Cui et al., 2011). These data suggest that GluN2B can exert bidirectional control over cognition, although the precise contribution of GluN2B to these effects remains to be determined. The same limitation applies to current understanding of the role of GluN2A given the absence of studies that specifically manipulate the subunit (either genetically or pharmacologically) in the PFC. Nonetheless, given the nature of the deficits on PFC-dependent tasks it is reasonable to hypothesize that NMDARs mediate forms of cognition that are subserved by this brain region.
A number of findings establish a role for PFC NMDARs in fear extinction. Pre-extinction treatment with the NMDAR partial agonist D-cycloserine facilitates fear extinction when given systemically (Matsuda et al., Walker et al., 2002, Lee et al., 2006) and facilitates re-extinction when given directly into the mPFC (Chang and Maren, 2009). Systemic administration of NMDAR antagonists also impairs both the acquisition and consolidation of fear extinction if administered before or after extinction training, respectively (Santini et al., 2001, Lee et al., 2006, Sotres-Bayon et al., 2007, Sotres-Bayon et al., 2009). This consolidation impairment is mimicked by mPFC-infusion of a NMDAR antagonist, possibly via reductions in the post-extinction burst firing of mPFC neurons (Burgos-Robles et al., 2007).
As with the PFC-mediated cognition, an important issue around the role of NMDARs in extinction is whether specific receptor subunits can be implicated. In this context, systemic treatment with GluN2B-specific antagonists (Rodrigues et al., 2001, Sotres-Bayon et al., 2007, Sotres-Bayon et al., 2009, Dalton et al., 2011b), as well as disruption of phosphorylation sites on the subunit (Nakazawa et al., 2006), impairs fear extinction. The effect of direct infusion of GluN2B antagonists on fear extinction has not been reported, although GluN2B inactivation, via antagonism or RNA interference, in dorsal regions of the mPFC (i.e., anterior cingulate) impairs the acquisition of contextual (but not cued) fear memory (Zhao et al., 2005).
It is notable that while NMDAR antagonists such as ketamine have impairing effects on fear extinction and cognition, the same drugs have fast-acting antidepressant and anxiolytic effects in patients (Krystal et al., 1994, Berman et al., 2000, Zarate et al., 2006, Preskorn et al., 2008). When delivered systemically, these compounds also reduce anxiety-like (reviewed in Cryan and Dev, 2008, Barkus et al., 2010) and ‘depression-related’ behaviors in rats and mice at baseline and following stress (Maeng et al., 2008, Li et al., 2010, Li et al., 2011). These effects are linked to the PFC, by the finding that they occur in parallel with various changes in this region, including enhanced mammalian target of rapamycin (mTOR) signaling, increased expression of synaptic proteins (synapsin, PSD-95, AMPA GluA1) and dendritic spinogenesis (Li et al., 2010). Moreover, intra-PFC blockade of mTOR is sufficient to prevent some of these behavioral and neuronal effects of systemically-delivered NMDAR antagonists (Li et al., 2010). These findings reveal a key role for NMDARs, via effects in PFC, in mediating rodent antidepressant-like actions, expanding the role of this mechanism beyond ‘classic’ PFC-mediated extinction and cognition behaviors which we have focused on in the current review.
There is growing evidence that, as with cognition and extinction, the GluN2B subunit is critical to NMDAR antagonist antidepressant-like effects. The above mentioned effects of ketamine, including the reversal of CUS-induced depression-related behaviors, are replicated by systemic treatment with the GluN2B-selective antagonist Ro 25–6981 (Li et al., 2010, Li et al., 2011). Moreover, gene deletion of GluN2B on principal neurons in the cortex, as well dorsal CA1 hippocampus, leads to an attenuated despair-like response to repeated swim stress, but normal baseline anxiety-behavior (Kiselycznyk et al., 2011). This phenotype contrasts with the more generalized anxiolytic-like effects of either more widespread forebrain GluN2B deletion (von Engelhardt et al., 2008) or systemic GluN2B antagonist treatment (Fraser et al., 1996, Maeng et al., 2008, Li et al., 2010). This would be consistent with a preferential role for cortical GluN2B in regulating adaptive responses to stress, although this remains to be shown.
Taken together, the literature discussed thus far comes raises the intriguing possibility that stress acts via NMDARs in the PFC to produce changes in a variety of behaviors that are regulated by this brain region. There remains, however, a paucity of evidence directly testing this hypothesis.
What has been shown is that stress produces changes in the expression of NMDARs that parallel alterations in PFC-mediated behaviors. For example, rats subjected to neonatal maternal separation have decreased levels of GluN1 in the infralimbic cortex, and show impaired fear extinction, as adults (Wilber et al., 2009). Along similar lines, chronic corticosterone administration in adult rats impairs fear extinction, in tandem with decreased mPFC expression of the GluN2B NMDAR subunit (Gourley et al., 2009). Acute forced swim stress (also in rats), on the other hand, has been found to transiently (one day) increase the cell surface expression of GluN1, GluN2A and GluN2B in mPFC and improve working memory in a T-maze task (Yuen et al., 2009). Thus, changes in mPFC NMDAR expression appear to be closely (and bidirectionally) tied to stress-induced shifts in behaviors mediated by the region.
As mentioned above, chronic stress causes dendritic retraction in PFC neurons. Dendritic atrophy is also seen in CA3 hippocampus following chronic stress (Roozendaal et al., 2009). Systemic administration of drugs blocking glutamate release (phenotyin) (Watanabe et al., 1992), or NMDARs (but not AMPA) can block this morphological effect (Magarinos and McEwen, 1995). Deletion of GluN1 on CA3 principal neurons mimicked this effect but, importantly, was not sufficient to prevent stress-induced increases in anxiety-like behavior – indicating the behavioral effects were not a consequence of dendritic changes in this region and likely resided elsewhere (Christian et al., 2011). Indeed, if GluN2A (but not AMPA GluA1) is deleted brain-wide, restraint-stress-induced changes in anxiety-like behavior can be prevented (Boyce-Rustay and Holmes, 2006, Mozhui et al., 2010).
Studies that examine the consequences of blockade or deletion of NMDARs specifically within the PFC for stress-induced behavioral alterations have not yet been conducted. One recent study found that the reduced PFC synaptic potentiation observed in mice exposed to postnatal footshock stress could be partially ameliorated by systemic injection of the NMDAR partial agonist D-cycloserine (Judo et al., 2010). In addition, an important recent study by Martin and Wellman found that rats given systemic NMDAR antagonist treatment failed to show mPFC dendritic atrophy (Martin and Wellman, 2011). These authors also showed that NMDAR antagonistic treatment in the presence of stress (but not NMDAR antagonist treatment alone) produced dendritic hypertrophy. This demonstrates that NMDARs mediate stress-induced morphological changes in PFC and suggest that neuronal morphology in this region is under tonic control of NMDAR activity.
The current literature sets up a number of critical questions going forward. For example, would systemic NMDAR blockade be sufficient to prevent stress-induced alterations in PFC-mediated behaviors, such as cognitive flexibility and fear extinction? If so, would these effects be recapitulated by blockade or gene deletion of NMDARs specifically within the PFC, or are they instead driven by NMDARs localized in other brain regions? What are the specific NMDAR subunits and critical downstream signaling processes involved in these effects? The development and refinement of pharmacological and genetic experimental tools should greatly aid preclinical research addressing these questions.
Ultimately, the question will be whether the findings made in rodents can illuminate novel therapeutic targets for alleviating cognitive deficits resulting from stress. A number of issues already present themselves in translating the rodent work to potential clinical application. One issue relates to the fact that on the one hand NMDAR antagonists have cognitive impairing effects, but have antidepressant actions likely via effects in the PFC. Preclinical evidence that chronic stress impairs cognition via reductions in mPFC NMDAR might predict that stimulation, rather than blockade, of NMDARs would work to rebalance mPFC functions. This could turn out to be the case for certain cognitive processes (e.g., see fear extinction and the case of D-cycloserine). In terms of antidepressants effects, one pertinent factor may be timing, with cognitive deficits occurring acutely and then waning, but antidepressant activity being more sustained due to long-lasting plastic changes in PFC. Indeed, patients given ketamine scored high on the Brief Psychiatric Rating Scale for positive symptoms of psychosis only 40 minutes after infusion, while alleviation of depressive symptoms occurred 2 hours to 7 days after treatment (Zarate et al., 2006). This echoes to some degree data in rats showing ketamine changes in spine growth and synaptic protein expression were delayed until 24 hours after injection (Li et al., 2010). Precisely delineating the temporal dynamics of NMDAR antagonist effects will be one important area for future study as this exciting field moves forward.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.