Stress-related learning, as modeled by classical fear conditioning, is thought to play an important role in the development of anxiety [1, 2, 3] and trauma and stress-related [4, 5] disorders, and it provides an important avenue through which traumatic memories can produce persisting changes in behavior. However, stressors are commonly encountered without resulting in permanent or pathological changes. The difference between adaptive and nonadaptive coping and whether stress has temporary or lasting effects can vary with the properties of the stressor, including whether or not it is controllable [6].

Stress can affect sleep, and traumatic life events virtually always produce at least transitory problems with sleep that may include insomnia or subjective sleep complaints [7]. Rapid eye movement sleep (REM) may be particularly vulnerable to the effects of stress, and a decrease in REM is initially observed in response to most stressors experienced during the normal sleep period [8, 9]. REM has been hypothesized to be important for processing emotion [10, 11, 12, 13], and we [14, 15] and others [14, 15, 16, 17] have suggested that it is important for adaptive responses to stress [8, 9]. Reduced and disrupted REM is also implicated in the genesis and symptomology of posttraumatic stress disorder (PTSD) [18, 19, 20], whereas greater baseline REM is associated with weaker fear-related activity in the amygdala, hippocampus, and ventromedial prefrontal cortex as well as reduced hippocampal modulation of the amygdala [21]. Thus, sleep, and REM in particular, appears to be important either as an active mediator of the stress response or as a biomarker for adaptive and maladaptive stress responses.

Animal studies have demonstrated that the amygdala can strongly influence REM. In particular, the central nucleus of the amygdala (CNA) can directly regulate REM and can mediate the effects of stress and fear memory on REM. For instance, inactivating CNA with microinjections of the $\gamma$-aminobutyric acid (GABA${}_{\text{A}}$) agonist muscimol (MUS) selectively decreases REM, whereas blocking GABAergic inhibition with the GABA${}_{\text{A}}$ antagonist bicuculline (BIC) increases REM [22]. Additionally, BIC microinjected into CNA attenuated the reduction in REM following inescapable shock (IS) [23], an uncontrollable stressor [24, 25, 26], whereas microinjection of MUS into the CNA did not [23], thus suggesting that CNA regulates IS-induced alterations in REM.

Several studies have demonstrated that training rats with IS, and re-exposure to the fearful context associated with IS [24, 25], can be followed by significant reductions in REM that can occur without recovery sleep (note that there can be individual differences in responses [15, 27]). In mice, controllable stress modeled by escapable shock (ES) and ES-related fear memories can produce significant increases in REM, whereas IS and IS-related fear memories can produce significant decreases in REM [14, 28, 29, 30]. By comparison, the effect of controllable stress on sleep in rats has rarely been explored though a few studies have examined sleep after training in an avoidance paradigm in which a signal allows the animals to prevent the presentation of a footshock [31, 32]. In the present study, we examined the effects of a controllable stressor on subsequent sleep using an ES paradigm in which an animal always receives a footshock but can terminate it by executing a simple behavioral escape response. As reminders of ES in mice produce changes in sleep similar to those observed when the footshock stressor is presented, we examined sleep after re-exposure to the ES context without footshock presentation to determine whether contextual reminders of ES would also alter sleep in rats. Finally, we assessed whether the CNA, a region implicated in the regulation of spontaneous REM [22, 33, 34, 35, 36, 37, 38] and stress-induced reductions in REM [23], regulates the effects of ES on sleep. Our goal was to establish the effects of ES learning on sleep in rats and to determine whether CNA regulates ES induced increases in REM as well as IS induced decreases as previously reported.

ES resulted in significant increases in total REM and REM%. These increases occurred across training days that also had significant decreases in light period total NREM and TS. The changes in total REM and REM% also were not due to a recovery of lost REM but were long-lasting increases over the entire recording period. Re-exposure to the shock chamber alone was followed by changes in sleep that resembled those observed following ES, with significant increases in total REM and REM% in both light and dark periods. Thus, reminders of the controllable stressor produced changes in subsequent sleep directionally similar to those observed following the original stressor. These results are similar to the increases in REM observed after ES and ES-associated memories in mice [14, 28, 29, 30].

To our knowledge, this study provides the first description of sleep in rats after an “escapable” shock paradigm in which animals always experience the stressor, but are given the opportunity to learn and execute a behavioral response that terminates it. The term “escapable” has been used in the literature to describe a variety of paradigms, but most are actually “avoidance” paradigms [32]. In avoidance paradigms, the animal can learn a behavioral response, which, if executed in response to a signal, will prevent them from experiencing the stressor. Avoidance training is also a stress paradigm, but it has historically been utilized to examine the potential role of sleep in learning and memory, and studies typically have been interpreted as providing evidence that the post-training increases in REM (either % or duration) played a role in memory consolidation [46, 47, 48].

Our results with ES demonstrate substantial increases in REM duration and REM% despite the fact that the footshock stressor was always experienced. This contrasts with the commonly observed decrease in REM duration that can occur following training with IS [24, 25, 26]. In addition, in the case of ES, the increases in REM occur throughout the dark period, and in the case of IS the significant decrease in REM can occur without an apparent period of recovery REM [24]. Moreover, re-exposing the animal to the ES training context, a situation that should bring about recall of memories associated with the stressor, also increased REM. Re-exposure to the training context alone also produces alterations in sleep in IS paradigms that are similar to those observed after training with IS [24]. Thus reminders of controllable and uncontrollable stress produce alterations in sleep similar to those produced by the original footshock stressor. Indeed, parallel increases or decreases in REM can occur across multiple days with ES and IS training, respectively. It is certainly possible that memories are recalled and reconsolidated across training days, but this also involves directionally different changes in REM. Another consideration for the ES and IS training paradigms is that the changes in sleep appear disproportionally large for learning relatively simple associations and responses. The increases in REM with ES were larger than increases in REM reported after avoidance training in a shuttlebox in which the animals had to learn to respond to a cue to avoid the presentation of footshock [47, 48].

While the escape latencies as well as altered sleep after re-exposure to the ES context alone demonstrate that learning occurred, the ES paradigm also involves a strong emotional component. Our microinjection study shows that the alterations in REM following ES are regulated by CNA, a region involved in the regulation of emotional behavior [49]. Microinjections of saline into CNA following ES did not alter the increase in REM observed across the light and dark periods. However, the inactivation of CNA with microinjections of MUS following ES prevented the increase of REM. Performing the microinjection after footshock exposure was completed ensured that behavioral responses during ES were not altered and that the effect of inactivation of CNA on post-stress sleep was being tested instead of on behaviors that occur in stressful situation. The involvement of the amygdala in regulating post-stress REM suggests that differences in REM after ES and IS could reflect differences in modulating the long-term effects of stress-related emotion after the experience with controllable and uncontrollable stressors.

Other work also supports a role for CNA in the initiation and maintenance of REM in both undisturbed and post-stress conditions. For example, noradrenergic denervation of CNA significantly reduces REM rebound after sleep deprivation [37]. Our lab has demonstrated, in non-stressful situations, that inhibition of CNA by microinjections of MUS produced significant reductions in the light period [22] but not the dark period [38], REM. By comparison, activation of CNA by microinjections of BIC increased REM [22]. Additionally, optogenetic activation of CNA projections to the nucleus reticularis pontis oralis, but not pedunculopontine tegmentum or nucleus subcoeruleus, during the dark period also promotes REM [38]. As noted above, microinjections of BIC into CNA blocked REM reductions found following IS, whereas microinjections of MUS did not [23]. Taken together, these studies suggest that CNA plays a significant role in regulating REM in both stressful and non-stressful situations and that activation of CNA can promote REM, possibly via specific projections to brainstem REM generator regions.

CNA modulation of the effects of stress on REM could also involve projections to the locus coeruleus (LC) and dorsal raphe nucleus (DRN), which have roles in regulating REM [50] and have been linked to the effects of stress. For example, IS in rats increases norepinephrine turnover in several brain areas whereas norepinephrine utilization is reduced after the coping response is acquired with ES [51]. IS activates 5-HT (5-hydroxytryptamine) DRN neurons to a greater degree than does ES thereby increasing 5-HT in DRN and in target areas [52, 53, 54, 55]. Differential activation of LC and DRN, possibly by CNA, after ES and IS could play a significant role in the differences in amounts of REM as activation of these regions may be inhibitory and inactivation permissive to REM [50].

There are some limitations to this study that need to be considered. First, though we have previously shown that ES and IS and related memories consistently produce differential effects on REM in inbred mice [14, 28, 29, 30], the story may be more complicated in outbred rats where we have found individual differences in the REM response in Wistar strain rats [15, 27]. Vulnerable (Vul) rats show decreases in REM, and resilient (Res) rats increase in REM after the same IS and IS-related fear memories [15, 27]. We did not distinguish Vul and Res rats in the current study suggesting that an increase in REM might have occurred in some animals even without the possibility of escape. However, REM increases in Res rats do not approach the level of increases observed after ES in this study, indicating that the level of increase was impacted by ES regardless of potential differences in phenotype.

Our results also demonstrate that the ES-induced increase in REM was clearly blocked by post-training inactivation of CNA. By comparison, post-training MUS inhibition of the basolateral amygdala (BLA) in Res rats did not alter either IS-induced or fear-conditioned alterations in REM [15], whereas post-training inhibition of the BLA with MUS did not alter IS-induced alterations in REM but did block subsequent fear conditioned reductions in REM in Vul rats [15]. Indeed, several lines of research demonstrate that the different REM responses in Vul and Res rats are regulated by BLA [15, 27, 56], which projects to and through CNA [57]. We previously found that the inactivation of fibers of passage in CNA with tetrodotoxin produced a significant reduction in REM (and an increase in NREM) [38]. Those fibers likely originate in BLA and project, at least partially, to the bed nucleus of the stria terminalis (BNST) [57], which has downstream projections similar to those of CNA [58, 59]. This suggests that the amygdala may increase or reduce REM and/or reduce NREM via BLA control of CNA and BNST and their projections to brainstem REM regulatory and generator regions.

In conclusion, emotional stress can contribute to mental disorders. However, even traumatic stress can be encountered without having a permanent negative impact. Interactions between stressor parameters, sleep and learning are likely important mediators of stress outcomes. This study demonstrates that ES, and ES associated fear memory, can produce significant increases in REM as well as decreases in NREM and TS. The REM increases are regulated by the amygdala as post-training inactivation of CNA can block ES-induced alterations in sleep. These data in rats, along with similar findings in mice, demonstrate that stressor controllability is important in determining the effects of stress and stress-related memories on sleep and that the amygdala is an important mediator of these alterations in sleep.

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

LDS, LLW, and XT designed the research study. LLW and LY performed the research. LLW, AMA and LY analyzed the data. LLW, and LDS wrote the manuscript. All authors contributed to editorial changes in the manuscript. All authors read and approved the final manuscript. All authors have participated sufficiently in the work and agreed to be accountable for all aspects of the work.

All procedures were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Experimental Animals and were approved by Eastern Virginia Medical School’s Animal Care and Use Committee (Approval number: 16-002).

Not applicable.

This research was supported by NIH research grant MH64827 (LDS).

The authors declare no conflict of interest. Larry D. Sanford and Laurie L. Wellman are serving as the Guest editors of this journal. We declare that Larry D. Sanford and Laurie L. Wellman had no involvement in the peer review of this article and had no access to information regarding its peer review. Full responsibility for the editorial process for this article was delegated to Gernot Riedel.