Although psychological stress is a harbinger for many psychiatric and medical illnesses, the biological underpinnings of this relationship remain largely unknown. The present study sought to identify potential changes in neuroinflammation in mice subjected to psychological stress. Based on previous studies examining the stress-inflammation relationship, we hypothesized that the stressor employed would have to be perceived by the animal to be significantly threatening in order to elicit a neuroinflammatory response. We therefore modified a model of predatory stress (PS) whereby a mouse is in the presence of an aggressive predator without the ability to escape. In this model, the mouse is exposed to sensory information in the absence of direct physical contact, allowing us to be confident that any response emitted by the mouse stemmed from the psychological assessment of the situation.
To verify that the PS model led to significant deficits whereby a mediator could be examined, adolescent mice were subjected to 28 consecutive days of PS. A separate cohort of mice were subjected to the often used CUS, a model of chronic stress that has been shown to elicit depressive-like behavior , a psychological disorder modulated by inflammation , and exacerbate various inflammatory-related diseases including obesity , atherosclerosis , and Alzheimer's disease . Using these two stressors, we tested: 1) whether CUS/chronic PS could modulate basal expression of inflammation and subsequent response to an immunogenic challenge (LPS), 2) whether chronic PS leads to similar decrements as CUS in depressive/anxiety related behaviors and, 3) whether mice subjected to CUS/chronic PS show similar adaptation in the classic stress measures (CORT) following 28 days of stress.
Our first goal was to examine whether CUS/chronic PS could modulate basal expression of inflammation-related genes and subsequent response to a LPS challenge. Two structures were examined: the midbrain, a region we have shown to be susceptible to inflammation [39–41], and the hippocampus, a critical player in the stress response . Within the midbrain of PS mice we observed ~ 2-fold increase in basal TNF mRNA compared to controls. A similar trend for an increase in basal expression of TNF (hippocampus) and IL-1 (midbrain and hippocampus) mRNA was also observed in PS mice, although this effect failed to reach statistical significance. When CUS/chronic PS mice were subsequently given LPS, the expression of TNF and IL-1 was markedly attenuated within both the midbrain and hippocampus compared to control mice, suggesting that chronic stress might facilitate organizational changes within the immune system that may lead to an inadequate response to a future stimulus.
Several researchers have looked at stress-immune interactions, although methodologies often differ considerably (see studies below). Notably, there are significant differences in regards to species and stressor used, age of the animal, duration of stress exposure (within a session and number of sessions), timing of second stimulus in relation to stress, tissue and inflammatory factors examined, as well as time after final challenge tissue is assessed. This is further complicated by the individual variability that occurs when the stress response results from interactions between two organisms in social stress situations [43, 44]. Indeed, the sheer numbers of variables that exist between any two studies make comparisons difficult at best and therefore should be cautioned. That said, others have shown a decrement in responding similar to those presented herein . Similarly, we have previously shown increased basal expression of IL-1 mRNA in dominant/submissive pair-housed rats. When these same rats were given acute footshock stress, we observed an attenuated IL-1 response compared to rats where a dominant/submissive hierarchy could not be determined . By contrast, a number of studies have also demonstrated that stress sensitizes the inflammatory response to challenges such as LPS [46, 47], an effect that has also been reported in humans . The most obvious mechanism by which chronic stress might attenuate the inflammatory response is through CORT. Although CORT levels rose following acute stress, CORT levels were not different in CUS/PS mice from controls on the final day of stress and again two weeks later when mice were given LPS. Given this, and the fact that others have shown stress-induced impairment of immune function are independent of CORT , we suggest that other mechanism(s) that regulate cytokine expression including anti-inflammatory cytokines and intracellular signaling pathways (e.g., suppressors of cytokine signaling and NFκB) may play a more central role. Future studies will be required to assess this hypothesis.
Regardless as to whether sensitization or desensitization is observed following chronic stress, changes in a tightly regulated inflammatory system could potentially have devastating outcomes. For instance, increased basal expression of TNF and IL-1 such as those found in these studies might be enough to create a pro-death environment in susceptible brain regions (such as midbrain). In addition, the impaired cytokine response to LPS suggest that the immune response to infectious agents might be below the threshold of what is required to clear it. This in turn could have a variety of devastating effects. To this end, studies examining whether chronic PS can modulate disease in a susceptible animal are currently underway.
Increasing evidence suggests that inflammation may play a critical role in depression. This stems from initial studies showing a pro-inflammatory profile in patients that suffer from major depression (see  for review). As described above, researchers have used the CUS model to examine the relationship between inflammation and depression . Therefore, to determine whether chronic PS, a model we have shown increases the basal expression of TNF, could also elicit behavioral deficits in disorders mediated by inflammation, depressive- and anxiety-like behaviors were assessed and compared to the CUS model. Consistent with what has been reported elsewhere, CUS led to a modest increase in anhedonia and anxiety- like behaviors [1, 50–52]. PS also precipitated increased depressive-like behaviors (sucrose preference test and tail suspension test) compared to controls. The most notable effect was observed in the marble-burying test as chronic PS mice displayed considerably more anxiety-like behaviors than both control and CUS mice. Specifically, chronic PS mice buried on average 14 out of 20 marbles, whereas CUS and control mice buried an average of 6 and 2, respectively. This particular difference in anxiety-like behavior is novel as previous studies have found that CUS has little effect in eliciting an anxiogenic response in C57/BL6 mice [51, 52], though it should be noted that anxiety was assessed using different tests. Neither CUS and chronic PS mice spent a greater amount of total time immobile in the tail suspension compared to controls, although this is likely due to a ceiling effect as total time spent immobile was ~330 s out of a total of 360 s for all groups. Our finding is not surprising as the C57BL/6 strain is known for their high levels of immobility in the tail suspension test . Collectively, these behavioral data imply that chronic PS is at least as effective as CUS for inducing depression and may have unique utility for inducing anxiety, two disorders that are often co-morbid.
Finally, we addressed the widely accepted belief that repeated exposure to the same stressor results in a decrease in the CORT response (habituation) whereas repeated exposure to different stressors does not . However, after 28 consecutive days of stress, the CORT response to both PS and CUS had returned to baseline levels. Given that CORT was not assessed after every session, the time for CORT to return to baseline levels in CUS and chronic PS mice is unclear. Despite this, Figueiredo and colleagues  found that the CORT response was not resolved after 14 days of predatory-prey (cat/rat) stress. Taken together, these data suggest that CORT may not be the driving factor governing the changes in inflammation and behavior following CUS and chronic PS for two reasons: 1) the total CORT response to both chronic stress models habituate by the final day of stressor exposure and 2) CUS/chronic PS and control mice had similar levels of CORT in response to Vehicle or LPS 2 weeks after the final day of stress.
It is important to note that outcomes between CUS and chronic PS mice are highly influenced by the nature of the stressors imposed upon the mice. Encounters with a predator are likely to result in wounding and or death. In order to keep the organism alive, a preparatory inflammatory response would be initiated to deal with potential wounding, infection, etc. Indeed, previous studies have reinforced this idea, demonstrating that immune cells migrate to areas close to the surface in order to quickly deal with impending injury following stress . Had PS not been included in the CUS model, it is possible that inflammatory differences between PS and CUS mice might have been greater. Therefore, when examining potential inflammatory consequences of chronic stress, model selection will likely be a critical factor, and we propose the PS model is best suited for these types of analyses.
In order to more fully understand our PS model, we conducted a second series of experiments whereby the inflammatory response was examined following a single acute session of PS. For these studies, adult mice were used in order to determine whether PS has inflammatory consequences for mice of all ages. In these studies, we expanded the number of structures to include the midbrain, hippocampus, hypothalamus, prefrontal cortex, and spleen. These structures were chosen based on prior work showing their responsiveness to stressful stimuli [7, 55]. Our data demonstrate that PS increased the expression of TNF and IL-1 in hypothalamus, hippocampus, and midbrain while having little effect in the prefrontal cortex and spleen. The effects of stress on IL-1 have been shown by others [7–9], although few, if any, studies have found changes in TNF following stress. Of note, CD45 mRNA was increased by stress but not until 8 hrs after PS ended. Often used as a marker for microglial activation , our data may indicate delayed immune cell activation. One possibility is that the increase in the expression of TNF and IL-1 observed shortly after PS facilitated the activation of microglia hours later. In this scenario, microglia would then be primed to increase the output of inflammatory factors should the organism encounter another stressful stimulus. Although this particular hypothesis is inconsistent with what we observed with chronic PS, others have shown that acute stress can potentiate an inflammatory response whereas chronic stress impairs it . When and how the inflammatory response goes from sensitization following acute stress to desensitization following chronic stress is a critically important question that remains unresolved.
Taken together, our results demonstrate that PS, an ethologically relevant stressor, can elicit changes in neuroinflammation and behavior that are comparable, in some measures, and greater in others, to CUS. We further propose that the PS model may be useful in elucidating mechanisms by which psychological stress modulates diseases with an inflammatory component. The significance of psychological stress being an effector of inflammation in the brain has far-reaching implications for neurological diseases with an inflammatory component.