Sex differences in resilience: experiential factors and their mechanisms

Abstract

Adverse life events can lead to stable changes in brain structure and function and are considered primary sources of risk for post-traumatic stress disorder, depression, and other neuropsychiatric disorders. However, most individuals do not develop these conditions following exposure to traumatic experiences, and research efforts have identified a number of experiential factors associated with an individual’s ability to withstand, adapt to, and facilitate recovery from adversity. While multiple animal models of stress resilience exist, so that the detailed biological mechanisms can be explored, studies have been disproportionately conducted in male subjects even though the prevalence and presentation of stress-linked disorders differ between sexes. This review focuses on 1) the mechanisms by which experiential factors (behavioral control over a stressor, exercise) reduce the impact of adverse events as studied in males; 2) whether other manipulations (ketamine) that buffer against stress-induced sequelae engage the same circuit features; and 3) whether these processes operate similarly in females. We argue that investigation of experiential factors that produce resistance/resilience rather than vulnerability to adversity will generate a unique set of biological mechanisms that potentially underlie sex differences in mood disorders.

Graphical Abstract

Women are at a higher risk for certain stress-linked psychiatric disorders and accounting for sex as a biological variable in preclinical research is fundamental for developing effective therapies for both men and women. We review the neural and behavioral sequelae of stress-buffering factors in each of the sexes.

Introduction

Exposure to adverse events is a risk factor for a variety of negative health outcomes including the development of psychiatric disorders such as depression, generalized anxiety, and posttraumatic stress disorder. Although polygenic factors certainly play a role, many resilience studies have focused on factors that change how an individual experiences trauma (e.g., coping strategies, pharmaceuticals, physical health, safety learning). These factors are associated with positive mental health outcomes and are modifiable through experience, which can be leveraged by interventions designed to inculcate resilience to adversity. Unfortunately, treatments for stress-related disorders are not always effective and this is may be due to different responses to resilience factors including differences that arise from sex (Breslau, 2009). Indeed, many psychiatric disorders that have stress as a contributing factor are sex-biased in their prevalence and/or presentation. Women are over twice as likely to develop mood and anxiety disorders whereas men are nearly three times as likely to develop a substance use disorder (Steel et al., 2014). Due to the limited use of females in preclinical research (Beery & Zucker, 2011), it is not understood how the neural circuits underlying resilience factors compare between the sexes and whether they produce similar or different behavioral outcomes. The purpose of this review is to provide a current summary of the neural mechanisms and behavioral phenomena induced by stress-protective factors in each of the sexes and to identify future research avenues.

Psychological variables such as a perceived ability to cope with or control a stressor can enhance long term psychosocial health and wellbeing (Kohn et al.,1994; Yi et al., 2005; Dijkstra et al., 2016). Similarly, physical health through aerobic exercise pre or post trauma can reduce depression ratings and relapse to the same degree as many current pharmacotherapies (Blumenthal et al., 2012). Whether these protective factors lead to a resilient phenotype that differs between men and women is not fully understood. There is evidence that men and women cope with adversity differently, which may be due to the fact that certain resilience factors are more effective for one sex than the other (Hamilton & Fagot, 1998; Tamres et al., 2002). Current research efforts are directed at identifying the shared and sex-specific features, and corresponding neural processes, of experiential resilience factors in order to maximize the applicability of preventative and durable treatment strategies for stress-linked disorders.

Animal models of stress have allowed us to study possible neural mechanisms driving the behavioral sequelae of stress exposure. One approach to studying stress vulnerability/resistance is to examine natural variation in the stress response. For example, an experimental manipulation (e.g. social defeat) will cause a measurable outcome (e.g. social interaction) that follows some distribution pattern. A criterion set by the experimenter can then be used to separate these groups (stress susceptible vs. stress resilient) in order to identify the neural circuit, endocrine, and molecular mediators that underlie the difference in response.

For some outcome measures, like freezing in fear conditioning/extinction tasks, male and female rodent models have been reported to produce a similar distribution pattern, termed sex convergence (Iwasaki-Sekino et al., 2009). However, others report sex divergence in which freezing is greater in males than in females (Kudielka et al., 2005; Amat et al., 2006; Baratta et al., 2008). Indeed, sex divergence is most commonly seen when the behavioral outcome measure involves ambulation, such as in the sidman avoidance, open field, and emergence (sheltered to exposed) tasks (Archer, 1975) in which female rodents are more active in avoiding threat to a greater degree than males. Higher ambulatory activity in females is often interpreted as a blunted fear response, despite the possibility that this is a female specific fear behavior (Gruene et al., 2015). Divergence in the opposite direction has also been reported. In response to chronic variable stress, females show stress induced sequelae in a vast array of behavioral tests in which males do not (Hodes et al., 2015). In addition, females exhibit a greater physiological response to footshock and handling. Given that male and female responses to stress in rodents do not necessarily parallel the sex differences in susceptibility to stress related disorders, future studies would benefit from a better understanding of sex specific behavioral responses to stress and the experiential factors that influence these responses.

An alternative approach to studying the neural mechanisms underlying resistance/resilience to stress is to manipulate experiential factors that are associated with resilience in humans. Manipulating the function of neural elements (gene to circuit) during an event known to produce resilience has helped to identify resilience mechanisms, with the majority of work conducted in males. We are only beginning to understand how stress resilience/resistance models work in females and how the underlying elements compare to males. Many experiential factors either occurring before, during, or after stress have been identified as stress protective. Here, we will focus on three that have been reported to be protective or predict protection from stress outcomes in humans: the use of coping strategies, ketamine and physical exercise. In preclinical models these factors are protective, provide enduring protection to a diverse set of stressors and engage distinct neural mechanisms.

Coping with stress

In humans, many of the factors associated with the ability to withstand (resistance) or facilitate recovery (resilience) from the impact of an adverse life event revolve around coping factors. Strategies aimed at eliminating or blunting the cause of the stressor (problem-focused coping) or reducing the emotional impact of the stressor (emotion-focused coping) are viewed as key processes in exerting actual or perceived control over negative circumstances in an individual’s life (Billings & Moos, 1981; Folkman et al., 1984). Behavioral control over some aspect of the stressor (Pearlin & Carmi, 1978; Baker & Berenbaum, 2007; Diehl & Hay, 2010) represents one of the few aspects of coping that can be experimentally manipulated in animals in order to better understand how resilience is mediated at a neural level.

Behavioral control has been studied in a variety of species, but the rat has been most frequent. Here, rats that receive physically identical stressors are compared, with one group having behavioral control over the termination of the stressor, and another group having no control. In order to isolate the variable of control, each subject is placed into small boxes with a wheel mounted on the front wall. The rat’s tail extends from the back of the box and is affixed to electrodes. For one group (ES; escapable stress), subjects receive a series of tailshocks each of which is terminated when the rat performs a given instrumental escape response (turning the wheel). Thus, ES subjects can exert control over one aspect of the stressor - the duration of each of the shocks. Each member of a second group (IS; inescapable stress) is paired with one from the escape group and simply receives tailshocks of the same duration as determined by its ES partner. For the IS subject, turning the wheel has no consequence. A third subject receives no tailshock, and serves as a no stress control group. Tailshock is used in rat studies to ensure that subjects with and without control receive physically identical shocks (number, intensity, duration, temporal pattern, etc.), a necessary requirement for isolating the impact of controllability. It may be the case that behavioral control would mitigate the outcome of stressors other than tailshock (e.g. social defeat, restraint), but since controllability can’t be readily manipulated in these paradigms its potential mitigating role cannot be determined.

Numerous behavioral changes are produced in subjects exposed to IS but do not occur if the subject is allowed to exert behavioral control over the stressor (ES). Behavioral changes (often termed “learned helplessness effects”) that follow IS, but not ES, include impaired shuttle box escape behavior, increased social avoidance, neophobia, potentiated morphine conditioned place preference, reduced aggression, resistance to fear extinction, and exaggerated shock-elicited freezing when tested in a novel environment 24 hr after stress treatment (Weiss, 1968; Maier & Watkins, 1998; Maier & Watkins 2005; Baratta et al., 2007; Rozeske et al., 2011). It is important to note that behavioral control doesn’t block or reduce all of the neurochemical or behavioral outcomes of IS. For instance, tail shock leads to robust changes in autonomic (Thompson et al., 2013), endocrine (Maier et al., 1986; Helmreich et al., 1999; Helmreich et al., 2012), brain immune (Frank et al., 2007), neural (McDevitt et al., 2009) and behavioral (Woodmansee et al., 1993) endpoints that are insensitive to control. Thus, it’s not the case that uncontrollable stress is simply more “aversive” or “potent” than controllable stress.

Additionally, the stress-buffering effects of ES extend beyond the stressor being controlled. That is, an initial experience with behavioral control buffers against the neural and behavioral sequelae of future inescapable stressors occurring much later (enduring) and in a different apparatus/environment from the original control experience (transsituational; Williams & Maier, 1977; Amat et al., 2006; Baratta et al., 2008; Amat et al., 2010). For example, Amat et al. (2006) showed that a previous experience of ES blocked the typical neurochemical and behavioral consequences (shuttle box escape interference and exaggerated shock-elicited freezing) produced by IS one week later, a process termed “behavioral immunization”. Prior exposure to IS had no effect on later IS, suggesting that the prior experience of control over shock, rather than shock per se, is the active ingredient in the development of resilience. It is doubtful that ES buffers against all future adversity, yet the boundary conditions of its protection remain largely unexplored. Of note, the proactive effects of ES extend to future uncontrollable stressors that don’t involve shock, namely social defeat. Similar to IS, acute social defeat leads to interference with shuttle box escape and increases social avoidance. ES occurring 7 days prior to social defeat blocked these social defeat-induced behavioral outcomes (Amat et al., 2010).

The initial work investigating the neural mediation of controllability effects focused on the mechanisms that produced the behavioral sequelae following uncontrollable stress. This has been reviewed in detail elsewhere (Maier & Watkins, 1998) and it involves a neural circuit that involves intense activation of serotonergic (5-HT) neurons in the dorsal raphe nucleus (DRN) which is critical for the production of IS-induced behavioral effects. The intense activation of DRN 5-HT neurons by uncontrollable stressors leads to a persistent sensitization of these neurons such that later testing procedures (footshock in the shuttlebox, presence of a juvenile, etc.) produce an exaggerated release of 5-HT into DRN projection regions that are the proximal mediators of the behaviors (Amat et al., 1998; Grahn et al., 1999; Bland et al., 2003; Strong et al., 2011). Activation of the DRN is both necessary and sufficient for producing many of the behaviors that follow IS.

Not surprisingly, behavioral control over shock does not lead to potent activation of DRN 5-HT and its exaggerated release in target regions of the DRN (Amat et al., 1998; Maswood et al., 1998; Strong et al., 2011). Rather, the experience of control engages the prelimbic (PL) region of the mPFC and provides top-down inhibition over DRN 5-HT activity (Amat et al., 2005), thereby preventing the shock-induced behavioral changes. This inhibition occurs because glutamatergic output neurons from the PL preferentially synapse onto GABAergic interneurons within the DRN (Jankowski & Sesack, 2004) which provide monosynaptic inhibition over 5-HT activity (Challis et al., 2014). Electrical stimulation of the mPFC (Hajos et al., 1998; Varga et al., 2001) and pathway-specific photostimulation of DRN-projecting mPFC neurons (Challis et al., 2014) both lead to DRN 5-HT inhibition through a GABAergic mechanism. In contrast, IS does not engage the PL projection to the DRN and pharmacological inhibition of the PL has no impact on the neurochemical and behavioral outcomes of IS (Amat et al., 2005; Baratta et al., 2009). These findings underscore the view that resilience mechanisms are not simply the absence of changes that drive vulnerability; rather, it engages a distinct set of neural mechanisms.

Engagement of the PL by behavioral control is also required for its enduring stress-buffering effects against future adverse events, a protection that lasts up to two months (Amat et al., 2006; Kubala et al., 2012). Efforts directed at understanding how an initial experience of coping with stress produces resistance to uncontrollable stressors experienced at a much later point in time have focused on N-methyl-D-aspartate receptor (NMDAR)-dependent processes. Activation of the extracellular signal-regulated kinase (ERK) cascade, an essential component of NMDAR signaling, is involved in the encoding and consolidation of an experience that leads to long-term behavioral change (Malenka et al., 1993; Thomas & Huganir, 2004; Morris, 2013). Interestingly, controllable stress increases phosphorylated ERK1 and ERK2 within the PL and blockade of NMDAR and ERK-dependent signaling in the PL during the experience of control eliminates behavioral immunization (Christianson et al., 2014). Although synaptic plasticity in the PL has never been directly assessed, behavioral control induces PL layer 5 pyramidal cell excitability by increasing intrinsic membrane excitability (Varela et al., 2012).

The reviewed work above clearly demonstrates that the PL regulates the DRN 5-HT response to shock in the presence of control, but is the mPFC also involved in the acquisition/detection of the instrumental wheel-turn controlling response? In the appetitive domain, there are two distinct instrumental learning systems with largely nonoverlapping circuitries mediating the acquisition and performance of instrumental action. One system termed “action-outcome” is mediated by a corticostriatal circuit involving the dorsal medial striatum (DMS) and the PL. The action-outcome system records the association between the action (wheel-turn) and the outcome (stress termination) and is sensitive to contingency (e.g. the probability of the outcome given the response integrated with the probability of the outcome given no response). The other system termed “habit” is mediated by the dorsolateral striatum (DLS) and sensorimotor cortex and records stimulus response associations in a contingency-independent manor (Yin et al., 2005; O’Hare et al., 2016). Recent data demonstrate that both the acute and long-term protective effects of behavioral control require the PL and the DMS (action-outcome circuit), but not the DLS (habit circuit; Amat et al., 2014).

These aforementioned data suggest that the PL is important for both the detection of the instrumental controlling response and regulation of the DRN, yet whether these processes occur through the same or different PL ensembles is unclear. The existing data are sparse, but it appears that the PL pyramidal neurons that project to the DMS and the PL neurons that project to the DMS are intermixed within layer 5 and are largely non-overlapping (Baratta & Maier, 2017). Thus, it appears that embedded within the PL are two separable ensembles associated with two distinct features of behavioral control: its ‘detection’ the subsequent ‘use’ of that information to modulate downstream structures (Figure 1).

Figure 1.

Simplified illustration of the circuits engaged by controllable stress, ketamine and exercise in males and females. (A) Proposed overlapping circuits involved in the two separable features of controllable stress in males (blue) and females (pink). ‘Detection’ of the instrumental wheel-turn response through the corticostriatal action-outcome system (light blue). The ‘use’ pathway (dark blue) showing prelimbic (PL) top-down inhibition over dorsal raphe nucleus (DRN). In females, ‘detection’ of the wheel-turn response may be acquired by the instrumental habit learning system (light pink), which does not lead to PL-DRN engagement (dark pink). (B) Schematic of ketamine’s direct engagement of the PL-DRN pathway in males and females. In both males and females, evidence for ketamine-induced spine formation specific to the PL-DRN pathway is unknown. (C) Motor structures engaged by exercise-induce plasticity in stress-responsive sites. The circuitry underlying the stress protective effects of exercise in females is unknown. Amy, amygdala; DS, dorsal striatum.

Recent mandates introduced by the National Institutes of Health and the Canadian Institutes of Health Research require investigators to examine sex effects in preclincal research in order to address health disparities (Clayton & Collins, 2014). The work described above was conducted entirely in male rodents, and prior to implementation of these policies, it was largely unknown whether behavioral control operates in females as it does in males, and if so, whether the same circuit mechanisms are involved. Few studies have manipulated controllability in females, and interestingly, control was without an effect (Dalla et al., 2008). There are a number of published female studies that utilize the term “learned helplessness”, however the term is often used in the literature in a way that differs from its definition (Maier & Seligman, 1976). Learned helplessness refers to the behavioral and neurochemical outcomes of the stressor that depend on the uncontrollability of the stressor. That is, to qualify as a learned helplessness effect a behavioral change (e.g., poor escape behavior) must follow exposure to IS, but not physically identical ES. This is what distinguishes a helplessness effect from a generic stress effect, and without this distinction, all effects of a stressor would be labeled as a learned helplessness effect. Indeed, it has been shown that not all outcomes that follow an uncontrollable stressor are learned helplessness effects. For example, IS produces reduced daily activity in running wheels, but ES produces the same reduction (Woodmansee et al., 1993) and so this reduction in activity is not produced by learned helplessness but is instead a simple stress effect. Prior female learned helplessness studies have not manipulated the controllability of the stressor (e.g., Jenkins et al., 2001), and when they have, there has often been no difference between the uncontrollable stress group and no stress controls. In either case, the impact of behavioral control cannot be determined.

Baratta et al. (2018) investigated the impact of controllability in female rats using the identical conditions that produce the differential behavioral consequences observed in male rats. In sharp contrast to what has previously been observed in males, the beneficial effects of behavioral control were absent in females, that is, both female ES and IS groups show identical levels of stress-induced sequelae when compared to no stress controls. ES and IS females exhibited greater c-Fos activation in DRN 5-HT expressing neurons, fail to escape in a shuttle box, and reduced juvenile social exploration (JSE; Baratta et al., 2018). That is, the controllability of the stressor does not determine resilience in females. Furthermore, c-Fos expression is absent in DRN-projecting PL neurons. Pharmacological activation of the PL reverses this phenotype, effectively blocking the effects of tail shock in uncontrollable and controllable stress females. These data suggest that mPFC mediated regulation of DRN 5-HT is not engaged by control in female subjects. It is currently unknown whether the lack of protection is due to sex differences in the detection of behavioral control or the use of control information to initiate regulation of the DRN.

The failure of control to provide protection in females is striking given that females rapidly acquire the wheel-turn controlling response at a rate comparable to males. As mentioned above, there are two separable systems involved in the encoding of instrumental responses and one untested possibility is that females acquire the escape response with the PL-independent habit system rather than the corticostriatal action-outcome system. Elevated catecholamine levels in the mPFC, such as those that may occur during stress exposure, can disrupt prefrontal top-down function (Arnsten, 2009) as well as strengthen habit circuit systems (Fournier et al., 2017). Stress-evoked release of catecholamines have been reported to be greater in females (Mitsushima et al., 2006; Staiti et al., 2011) across several brain regions. Moreover, stress-induced corticosterone levels, which have been reported to be higher in females (Heck & Handa, 2018), can shift the acquisition of instrumental learning to the habit system (Schwabe & Wolf, 2011). Thus, one possibility is that stress-evoked catecholamine release in the mPFC biases females to acquire the wheel-turn controlling response with the habit learning system, evading engagement of the PL and top-down regulation of the DRN during tail shock.

The absence of a modulating effect of control in female ES subjects also extends to behavioral immunization. Baratta et al. (2019) investigated immunization effects by giving male and female rats ES, IS, and no stress home cage (HC) one week prior to an uncontrollable stressor. As is typical, ES males were resistant to the behavioral outcomes of the later experienced uncontrollable stressor, whereas ES females were not. In addition, structural plasticity in the PL-to-DRN pathway following ES and IS differed between the sexes. In males, IS lead to non-pathway-specific alterations in spine enlargement, while ES elicited spine changes only in DRN-projecting PL neurons. Thus, structural plasticity in the PL-to-DRN pathway following ES is consistent with prior work implicating the involvement of this pathway in the long-term protective effects of controllable stress (Amat et al., 2006; Varela et al., 2012; Christianson et al., 2014). In females, the pattern of spine changes differed from males. ES-elicited broad, non-specific spine changes in the mPFC while IS led to only minor alterations. Whether these sex-specific patterns of structural plasticity are related to the differential behavioral outcomes remains to be determined.

Stressor controllability produces sex-divergent behaviors in rodents; males given a controlling instrumental response over stress termination are resistant to the behavioral outcomes of that stressor, whereas in corresponding females the controlling response is without benefit. Future work dissecting the neural circuit components that respond to coping with stress, and how they differ between the sexes, may serve as an important model for understanding sex differences in susceptibility to stress-related disorders.

Stress-buffering and restorative effects of subanesthetic ketamine

Depression is estimated to be among the most burdensome disorders worldwide affecting more than 300 million people (Friedrich, 2017). Current pharmacotherapies for moderate and severe depressive disorders are only partially effective, at best, and treatment-resistant individuals require alternative therapeutic approaches with novel mechanisms of action (Wiedemann, 2011). Over the last two decades, clinical studies have found that acute treatment with a sub-anesthetic dose of the NMDAR antagonist, ketamine, produces rapid (i.e. within hours) and sustained (up to two weeks) antidepressant effects in individuals with treatment-resistant depression (Berman et al., 2000; Zarate et al., 2006; Price et al., 2009; Zarate et al., 2012; Murrough et al., 2013; Lai et al., 2014). The antidepressant effects of ketamine are observed similarly in both men and women (Carrier et al., 2013; Saland et al., 2017; Freeman et al., 2019), though a few report the effects last longer in men (Coyle et al., 2015; Rybakowski et al., 2017). The mechanisms underlying the antidepressant effects of ketamine appear to go beyond NMDA antagonism (Newport et al., 2015) and preclinical studies suggest they are also sex-differentiated.

Although ketamine is the prototype for a new generation of antidepressants for treatment-resistant depressive patients, ketamine has properties that limit its utility as a therapeutic. Ketamine has a potential for abuse and can be accompanied by dissociative and psychotomimetic side effects. This has led to efforts directed at testing and developing structural analogs of ketamine with a safer profile (Lener et al., 2017) and to understand the mechanism(s) by which ketamine is effective. In preclinical studies, which have mostly been conducted in males, low dose ketamine in male rats and mice facilitates fear extinction (Girgenti et al., 2017) and blunts the behavioral outcomes of stressor exposure (e.g. chronic corticosterone administration, inescapable tail shock, chronic mild stress) when administered prior to (Maeng et al., 2008; Dolzani et al., 2018) or after stress treatment (Li et al., 2011). The effects of ketamine are sufficiently diverse, buffering against and reversing stress-induced changes in forced swim, tail suspension, juvenile social investigation, shuttle box escape, and elevated plus maze (Armario et al., 1991; Maeng et al., 2008; Engin et al., 2009; Autry et al., 2011; Dolzani et al., 2018).

Low dose ketamine also provides stress-protection in female rodents, and the effective dose and time course of its stress-reducing effects appear to be sex divergent (Carrier & Kabbaj, 2013; Franceschelli et al., 2015; Dossat et al., 2018). In female rats, systemic administration of low dose (2.5 mg/kg) ketamine prior to testing (30 min) reduces time spent immobile in a forced swim test and decreases time to feed in a novelty suppressed feeding task. Males require at a higher dose (5 mg/kg) for similar effects (Carrier & Kabbaj, 2013). Further, ketamine delivered following 10 days of chronic mild stress can ameliorate stressed-induced behaviors for up to 24 hrs in females and 7 days for males (Franceschelli et al., 2016). Notably, some human studies have also reported that the effects of ketamine last longer in males (Coyle & Laws, 2015; Rybakowski et al., 2017). The stress-buffering effects of low dose ketamine are diminished in ovariectomized females and restored upon estrogen and progesterone supplementation (Carrier & Kabbaj, 2013). This suggests that gonadal hormones may play an important role in ketmaine’s mechanism of action in females.

Ketamine has been shown to modulate neurochemical activity in the prefrontal cortex, with some effects comparable between the sexes, in both clinical and preclinical studies. Ketamine administration in men and women with or without major depressive disorder produces an increase in glutamate transmission in the mPFC (Abdallah et al., 2018). Similarly, in male rodents, systemic administration of ketamine stimulates the release of glutamate resulting from acute, transient inhibition of GABAergic activity (Duman et al., 2019) and pharmacological inhibition of the mPFC prevents ketamine’s protective effects against stress (Amat et al., 2016). Interestingly, intra-mPFC administration of ketamine leads to dopamine release (Lorrain et al., 2003) and optogenetic inhibition of dopamine receptor 1-expressing mPFC neurons eliminates the protective effects of ketamine in both male and female mice (Hare et al., 2019). Collectively, these data implicate the induction of dopaminergic and glutamatergic signaling in the prefrontal cortex for ketamine’s antidepressant effects.

Behavioral sequelae induced by stress exposure are often associated with changes in structural plasticity, such as increased elimination or decreased formation of dendritic spines, across numerous brain regions (Moda-Sava et al., 2019). In both male and female naive rodents, low-dose ketamine can enhance synaptic plasticity and spine formation (Ruddy et al., 2015; Duman et al., 2016; Phoumthipphavong et al., 2016; Pryazhnikov et al., 2018). At least in males, spine formation and stable maintenance is critical for ketamine’s stress-reducing effects. Using a novel photosensitive tool, activated synapse-targeting photoactivatable Rac1 (AS-PaRac1), Moda-Sava et al. (2019) demonstrated that spine development by ketamine is necessary for its long-lasting buffering effects on stress outcome. Synaptic remodeling appears to be a critical feature of ketamine’s ability to reverse the impact of stress. The mammalian target of rapamycin (mTOR) pathway, an important player in spine formation and maturation, is necessary for ketamine-induced increases in spine formation in the mPFC and it’s ability to blunt stress effects on forced swim, inescapable stress, and the novelty suppressed feeding test (Li et al., 2010). Interestingly, brain-derived neurotrophic factor (BDNF), but not mTOR, signaling is necessary for the acute antidepressant effects (Autry et al., 2011). These data suggest mTOR signaling and spine formation is most likely important for sustaining the protective effects of ketamine whereas other mechanisms may underlie the rapid protective effects.

The impact of ketamine on prefrontal structural plasticity may be sex-divergent (Table 1), particularly in the mPFC (Sarkar & Kabbaj, 2016). In males, ketamine increases prefrontal spine density and synaptic protein expression, which is not observed in females (Sarkar & Kabbaj, 2016). An increase in synaptic plasticity related proteins following ketamine administration is observed in males and only in females that are in diestrus 1 and proestrus (Dossat et al., 2018). Future work regarding the effects of ketamine on long-term changes in neuronal architecture and signaling would benefit from including both sexes and closely examining the effect of estrus cycle phase.

Table 1. Female studies relevant to behavioral control, exercise, and ketamine.

Non-exhaustive list of female rodent studies relevant to the impact of behavioral control, ketamine, and exercise on adverse events. Note that references for learned helplessness studies in which the endpoint measure was unaffected by uncontrollable stress and/or did not include a comparison controllable stress group, are not listed.

Author, Year	Resilience Factor	Sex	Main Phenotypes
Baratta et al. (2018)	Behavioral control	F	Behavioral control in female rats does not mitigate stress-induced behaviors - social avoidance, enhanced freezing, impaired escape. Prefrontal regulation of dorsal raphe serotonin is absent.
Baratta et al. (2019)	Behavioral control	F | M	Unlike males, behavioral control in female rats does not buffer against outcomes of future adverse events. Behavioral control in males, but not females, induces pathway-specific spine changes in dorsal raphe-projecting prefrontal neurons.
Carrier and Kabbaj (2013)	Ketamine	F | M	Females respond to a lower dose of ketamine in the forced swim test than male rats, an effect that is eliminated following ovariectomy and restored following estrogen and progesterone supplement.
Dolzani et al. (2018)	Ketamine	F	In female rats, ketamine produces persistent changes in prefrontal-to-dorsal raphe circuit activity such that the serotonergic response to and behavioral outcome of future stressors are blunted.
Dossat et al. (2018)	Ketamine	F | M	Proestrus female mice exhibit behavioral sensitivity to doses of ketamine that are ineffective in males and diestrus females.
Franceschelli et al. (2015)	Ketamine	F | M	Female mice displayed differential sensitivity to the rapid (30 min) and sustained (24 hr) effects of low dose ketamine in the forced swim test compared to males.
Goodwill et al. (2019)	Ketamine	F | M	Female, but not male, mice exposed to early life stress exhibit decreased grooming and elevated resting behavior in adulthood. Both phenotypes are reversed following acute ketamine administration.
Hare et al. (2019)	Ketamine	F | M	The rapid-acting effects of ketamine on novelty suppressed feeding and forced swim immobility require activity in Drd1-expressing pyramidal cells in the mouse medial prefrontal cortex.
Newman et al. (2019)	Ketamine	F	Ketamine reverses chronic defeat-induced social avoidance in female mice 24 hr after administration.
Phoumthipphavong et al. (2016)	Ketamine	F | M	A single subanesthetic dose of ketamine enhances dendritic spine formation that persists for up to two weeks in the medial frontal cortex of female and male mice.
Thelen et al. (2019)	Ketamine	F | M	A single subanesthetic dose of ketamine induces synaptic plasticity-related markers and sustained dendritic spine formation in the medial prefrontal cortex of male, but not female, mice.
Sarkar and Kabbaj (2016)	Ketamine	F | M	Chronic social isolation reduces medial prefrontal cortex spine density in both female and male rats. Ketamine rapidly reverses this effect in males but not females.
Bouchet et al. (2017)	Exercise	F | M	Phase of estrous cycle, but not exercise condition, blocks cued fear renewal in rats. Proestrus and estrus, but not metestrus and diestrus, during extinction training protects against later fear renewal.
Dishman et al. (1997)	Exercise	F	Chronic wheel running in female rats enhances shuttle box performance and reduces amygdala serotonin content following uncontrollable stress.
Jones et al. (2016)	Exercise	F	Brief exposure to wheel running attenuates restraint-induced corticosterone in ovariectomized female rats only if estrogen is replaced.
Kannangara et al. (2011)	Exercise	F	Wheel running in aged (17–18 months old) female mice reduces corticosterone levels in socially housed subjects and promotes hippocampal cell proliferation independent of housing condition.
Lalanza et al., (2015)	Exercise	F | M	Long-term treadmill exercise enhances shuttle box avoidance/escape performance in female and male rats. In females, prior exercise did not modify the endocrine response to shuttle box testing.
Tanner et al. (2019)	Exercise	F	Six weeks of wheel running in female rats buffers against behavioral sequelae of uncontrollable stress.
Robinson et al. (2019)	Exercise	F	Following stress treatment, four weeks of wheel running reduces impact of prior stress on open field, elevated plus maze, and cued fear conditioning in female rats.

Nonetheless, activation of the PL-to-DRN pathway appears to be important for the long-term protective effects of ketamine in both males and females (Figure 1). Using an activity-dependent marking system (robust activity marking system, RAM), Dolzani et al. (2018), demonstrated that ketamine administered induces activity in PL ensembles that are later reactivated during exposure to uncontrollable stress. In addition, inhibition of the PL-DRN pathway using a Cre-inducible inhibitory designer receptor activated by designer drugs (DREADDs) elminitates the protective effects of ketamine (Dolzani et al., 2018). These results suggest that ketamine alters the PL-to-DRN pathway in such a way that future uncontrollable stressors that normally don’t activate the PL-to-DRN pathway, now do so, leading to a blunting of the DRN 5-HT and preventing stress-induced behavioral sequelae. Whether this process involves stable spine changes, and whether it differs between the sexes, is unknown (Figure 1).

Exercise-induced stress resistance

Increasing physical activity, either before or after stressor exposure, is another way to reduce stress outcomes. Recent longitudinal studies following large numbers of subjects, for example, find strong associations between exercise and mental health (Chekroud et al., 2018; Harvey et al., 2018). Both of these studies indicate that outcomes are not influenced by exercise intensity, age or sex, and both report similar optimal durations of exercise of an hour or less. In fact, Harvey et al. (2018) report that only 60 min of exercise per week is associated with lower incidence of new-onset depression. In addition to these protective effects of exercise, exercise can reduce symptoms of existing stress-related disorders. Exercise, either alone or as an adjunct to conventional treatment strategies, can reduce symptoms of depression and anxiety in both males and females (King et al., 1991; Bartholomew & Linder 1998; Broocks et al., 1998; Blumenthal et al., 1999; Goodwin, 2003; De Moor et al., 2006; Blumenthal et al., 2007; Ströhle et al., 2007; Merom et al., 2008; Smits et al., 2008; Asmundson et al., 2013; Wegner et al., 2014). Exercise is also emerging as an effective strategy to overcome drug addiction in both males and females (Zhou et al., 2016; Lynch et al., 2017). Some important themes emerging from these clinical studies are that maintaining regular exercise seems to be more effective than exercising at a high intensity, and exercise benefits both sexes.

Rodent models provide a useful tool to investigate features of, and mechanisms underlying, exercise-induced stress resilience (Greenwood, 2019b). Rats allowed voluntary access to running wheels demonstrate resilience against a variety of stressors, such as morphine withdrawal (Miladi-Gorji et al., 2012), maternal separation (Masrour et al., 2018), forced swimming (Duman et al., 2008), olfactory bulbectomy (Van Hoomissen et al., 2011), chronic mild stress (Solberg et al., 1999), and inescapable stress (Dishman et al., 1997c; Greenwood et al., 2003; Greenwood et al., 2007a). It is this latter stressor, inescapable stress, with which exercise-induced stress resilience has been most well characterized (Greenwood & Fleshner, 2008; Greenwood & Fleshner, 2011). Male rats allowed voluntary access to running wheels for six weeks prior to exposure to inescapable tail shock are protected from the shuttle box escape deficit (Greenwood et al., 2003), exaggerated freezing (Greenwood et al., 2003), social avoidance (Greenwood et al., 2012a), circadian disruption (Thompson et al., 2016), amnesia (Fleshner et al., 2014), and potentiation of the rewarding effects of morphine (Rozeske et al., 2011), typically produced by exposure to inescapable stress in sedentary rats. Although forced treadmill training generally fails to produce stress resilience, stress resilience can be produced by forced wheel running of a pattern similar to voluntary running (Greenwood et al., 2013). In humans, the intensity of exercise is not a key factor in maintaining mental health. Similarly in rats, the protective effects of wheel running against inescapable stress are seldom related to running distance (Greenwood et al., 2003; Greenwood et al., 2013). Rather, it is the duration of wheel access that appears to be a critical factor. In male rats, six weeks of wheel running enables stress resilience, but three weeks is insufficient (Greenwood et al., 2005a; Greenwood et al., 2005b). Stress resilience produced by six weeks of wheel running is not permanent, but does persist about 15 days following forced cessation of exercise (Greenwood et al., 2012a).

Although stress resilience produced by exercise has been well characterized, the above studies focus on the effects of exercise in male rodents. Much less is known about exercise-induced stress resilience in females. Wheel running can decrease vulnerability to substance use disorders in both sexes (Zlebnik & Carroll, 2015; Lynch et al., 2017), with exercise appearing to benefit females more than males (Lynch et al., 2017). Similarly, four weeks of wheel running has been reported to reverse anxiety-like behavior following exposure to a complex stressor in female rats (Robinson et al., 2019), but another study reported that wheel running reverses depression-like behavior produced by perinatal ethanol exposure in males only (Brocardo et al., 2012). These studies provide evidence that the stress-protective effects of exercise may be sex divergent.

One component of stress resilience is protection from fear relapse following fear extinction. In male rats, an acute bout of wheel running during (Mika et al., 2015) or immediately following (Siette et al., 2014; Tanner et al., 2018) fear extinction can enhance fear extinction retrieval and reduce relapse. This beneficial effect of exercise appears to be sex dependent. An acute bout of wheel running immediately after auditory fear extinction fails to enhance extinction retrieval or reduce relapse in female rats (Bouchet et al., 2017). However, many of the rats in this experiment were in the proestrus phase of the estrous cycle during fear extinction training. Estrogen levels are high during proestrus, and estrogen is known to support strong fear extinction memory relative to females in estrous phases with low levels of estrogen, such as metestrus (Velasco et al., 2019). In fact, females in the Bouchet et al. (2017) study exposed to fear extinction training during estrous phases with high estrogen were protected from fear renewal, one form of relapse. It is possible that exercise could not further enhance fear extinction in these females already benefiting from high levels of estrogen. Similarly, Jones et al. (2016) found that 2 days of wheel running was able to attenuate stress-induced corticosterone in ovariectomized female rats only if estrogen was replaced. Running actually potentiated corticosterone increases following stress in ovariectomized females in the absence of estrogen replacement (Jones et al., 2016). Together, these data indicate that the effects of exercise in females could depend on cycling ovarian hormones.

Clearly, more work needs to be done characterizing potential sex differences in exercise-induced stress resilience in animal models. Based on the favorable results of clinical exercise studies that include female subjects, we expect that exercise can enable stress resilience in female rodents as it can in males. Because exercise-induced stress resilience has been well characterized in males using the inescapable stress model, we set out to determine if wheel running could prevent the behavioral consequences of inescapable stress in females. We found that adult, female Sprague Dawley rats allowed access to running wheels for six weeks are indeed protected from both the exaggerated fear and reduction in JSE produced by IS (Tanner et al., 2019). Although males were not included in this experiment, the magnitude of the protective effect produced by wheel running in females is similar to that of males. Exposure to inescapable stress produces a large (between 170 – 250% above control levels) increase in shock-elicited freezing in both males and females (Greenwood et al., 2013; Baratta et al., 2018; Tanner et al., 2019). In contrast, levels of shock-elicited freezing in male and female runners exposed to inescapable stress have been reported to be 97.3 +/− 19.7 % (Greenwood et al., 2013) and 94.2 +/− 15.7 % (Tanner et al., 2019) of the freezing levels observed in non-stressed male and female runners, respectively (t-test between males and females: p > 0.05). Similarly, exposure to inescapable stress reduces exploration during a 3 min JSE test to approximately 75% of the levels displayed by non-stressed male and female rats (Baratta et al., 2018; Greenwood et al., 2012a; Tanner et al., 2019), but exercise maintains social exploration after stress to 113.7 +/− 5.6 % (Greenwood et al., 2012a) and 92.3 +/− 8.9 % (Tanner et al., 2019) of non-stressed male and female runners, respectively (t-test between males and females: p > 0.05).

As in prior work in males, the distance run in females did not influence the strength of the stress-protective effect of exercise in females, despite the fact that female rats run about twice the nighty distance as males (Tanner et al., 2019). Since the duration of wheel running is an important factor in determining exercise-induced stress resilience in males, it is critical to determine if similar durations of wheel running are required to enable stress resilience in females. In male rats, six weeks of wheel running prevents and reverses the behavioral consequences of inescapable stress, while three weeks of running is unable to produce these stress resilience effects (Greenwood et al., 2005a; Greenwood et al., 2007a). It is unknown if three weeks of wheel running is sufficient to protect females from the negative outcomes of inescapable stress. Therefore, we sought to determine the effect of three weeks of prior voluntary wheel running on the behavioral outcomes of inescapable stress in females. Adult, female Sprague Dawley rats were randomly assigned to either voluntary wheel running (“run”) or sedentary conditions. Run rats were given access to unlocked running wheels for three weeks, while the wheels in cages of sedentary rats remained locked. Following three weeks of running or sedentary conditions, rats were either left undisturbed in their home cages (HC) or exposed to IS as in our prior work (Tanner et al., 2019). Twenty-four hours after stress, all rats (n = 8/group) were tested for JSE and shock-elicited freezing, as previously described (Tanner et al., 2019). Run rats not exposed to stress were not included in the study, because we already know that wheel running does not alter social exploration or freezing behavior in non-stressed females (Tanner et al., 2019). All experimental procedures were approved by the University of Colorado Denver Institutional Animal Care and Use Committee and raw data are available upon request.

Results of the experiment are shown in Figure 2. Sedentary (179.6 ± 3.4g) and run (178.5 ± 3.2g) rats weighed similar amounts prior to the start of wheel running, but run rats gained less weight than sedentary rats over the three weeks of exercise (exercise by time interaction: F(3, 42) = 8.5; p = 0.0002; data not shown). The average daily running distance increased over the 3 weeks (F(7, 20) = 17; p < 0.0001; Figure 2A). A technical malfunction resulted in the loss of four video recordings of social exploration, thus reducing group sizes used for social exploration. IS reduced social exploration in sedentary females (n = 5), whereas the levels of social exploration in IS female runners (n = 7) were indistinguishable from those of HC females (n = 8; F(2, 17) = 4; p < 0.05; Figure 2B). Similar results were observed for shock-elicited freezing (Figure 2C). Freezing prior to the two foot shocks was minimal and did not differ between groups (Figure 2C, “pre”). Freezing behavior increased in all groups following the two foot shocks, and then subsequently decreased over the 20-minute scoring session (main effect of time: F(9, 189) = 54.3; p < 0.0001). Repeated measures ANOVA revealed an interaction between group and time (F(18, 189) = 2.7; p < 0.001), such that sedentary IS rats displayed impaired within-session extinction compared to rats in other groups (Figure 2C). As in our prior work (Tanner et al., 2019), neither distance run nor phase of estrous cycle during stress or testing (shown for JSE in Figure 2B) impacted the behavioral outcomes, although the experiment was not adequately powered to determine whether the effects of exercise depend on estrous phase. These data indicate that a shorter duration of voluntary exercise can enable stress resilience in females than in males. Thus, exercise seems to be a highly effective stress resilience factor in females. These results justify more intensive investigations into sex-dependent mechanisms of exercise-induced stress resilience.

Figure 2.

Three weeks of wheel running enables stress resilience in females. (A) Distance run over the duration of the 3 week experiment. (B) Time spent interacting with juvenile conspecific during a 3 minute juvenile social exploration test. Diamonds represent individual rats in metestrous or diestrus at the time of behavioral testing. Triangles represent individual rats in proestrus or estrus at the time of behavioral testing. (C) Levels of freezing prior to (pre) or following 2 foot shocks. Data represent group means ± SEM. Sed, sedentary; Run, voluntary running; HC, Home cage controls; IS, inescapable stress. * p < 0.05 relative to all other groups; ϕ p < 0.05 relative to Sed HC only.

Exercise impacts a variety of nervous system factors and processes that could contribute to stress resilience (Morgan et al., 2015). Based on the observation that exercise elicits stress resilience in females faster than in males, it is possible that males and females differentially engage stress resilience mechanisms. Supporting the idea that some of the effects of exercise may be bigger in females than in males are emerging data suggesting the cognitive benefits of exercise are also more powerful in females (Barha et al., 2017a; Barha et al., 2017b; Barha & Liu-Ambrose, 2018). However, the majority of mechanistic studies have so far focused on males.

In males, exercise impacts monoaminergic systems (Dishman, 1997a; Dishman, 1997b; Greenwood & Fleshner, 2011; Nicastro & Greenwood, 2016; Greenwood, 2019a), as well as growth factors and other mediators of neural plasticity including BDNF, galanin and mTOR (Cotman et al., 2007; Sciolino & Holmes, 2012; Lloyd et al., 2017; El-Sayes et al., 2019). Monoaminergic systems appear to be sensitive to exercise in females (Dishman, 1997b), but much remains unknown. Exercise increases BDNF (Neeper et al., 1995) and neurogenesis (van Praag et al., 1999; Triviño-Paredes et al., 2016; Kim et al., 2019) in both male and female rodents; both of which have been implicated in stress resilience (Duman, 2009; Levone et al., 2015). In female rats, the increase in BNDF produced by exercise is dependent on estrogen (Berchtold et al., 2001). Exercise also attenuates the hypothalamic-pituitary-adrenal response to acute and chronic stressors in males (Droste et al., 2003; Campeau et al., 2010; Sasse et al., 2013); an effect that seems to vary with estrous phase in females (Jones et al., 2016).

There are multiple mechanisms by which exercise could prevent specifically the behavioral consequences of inescapable stress (Christianson & Greenwood, 2014; Nicastro & Greenwood, 2016). Neither attenuation of the hypothalamic-pituitary-adrenal axis response to inescapable stress, nor hippocampal BDNF (Greenwood et al., 2007b) appear to be necessary for the ability of wheel running to prevent inescapable stress-induced behaviors. Because of the critical role of DRN 5-HT neurons in mediating the behavioral effects of inescapable stress, much research has focused on the effects of exercise on central serotonergic systems (Greenwood & Fleshner, 2011). In males, six weeks of wheel running attenuates activity of DRN 5-HT neurons as measured with c-Fos and increases mRNA coding for the 5-HT1A inhibitory autoreceptor (Greenwood et al., 2003; Greenwood et al., 2005b; Loughridge et al., 2013), which could enhance 5-HT1A-mediated autoinhibition of DRN 5-HT neurons during stress. Consistent with the c-Fos data, six weeks of wheel running also prevents the IS-induced potentiation of 5-HT efflux in the dorsal striatum (Clark et al., 2015), where exaggerated 5-HT release contributes to the IS-induced shuttle box escape deficit (Strong et al., 2011). 5-HT adaptations to exercise are not limited to DRN 5-HT neurons. Six weeks of wheel running also reduces the expression of 5-HT2C receptors in the dorsal striatum and amygdala (Greenwood et al., 2012b). See Figure 1 for a summary of the circuit on which exercise-induced plasticity is proposed to produce stress resilience in males. If these adaptations are also important for enabling stress resilience in females, then we would expect these exercise adaptations to occur more rapidly in the brains of females. Interestingly, and in stark contrast to ketamine and escapable stress which rely on prefrontal mechanisms, exercise appears to constrain the DRN 5-HT response to stress through mechanisms other than prefrontal cortex inhibition (Greenwood et al., 2013; Christianson & Greenwood, 2014). These findings suggest that the circuits contributing to stress resilience differ between exercise and other resilience factors, namely ketamine and behavioral control.

Conclusions

The ability of experiential factors such as coping with stress and exercise to protect against stress-induced outcomes differs between the sexes. Recent work from stressor controllability studies shows that unlike males, the acute and long-term benefits afforded by behavioral control is absent in females. These findings suggest that the neural processing of control differs in females and the lack of engagement and structural plasticity within the PL-to-DRN pathway represents a potential mechanism. Since males and females exhibit different spine changes in this pathway following behavioral control, future work should employ novel circuit-targeting strategies for labeling and manipulating spines formed in response to an experience of control. The emergence of optical tools like AS-PaRAC1, which has been used to establish a causal link between ketamine-induced spine changes and stress protection (Moda-Sava et al., 2019), will drive future mechanistic studies that explore the causal relationship between experiential factors and the stable circuit changes that produce their protective effects, and whether or not these changes are shared between factors. Identifying common circuit endophenotypes between resilience factors enhances the possibility of translating basic research findings into clinical interventions.

Women are more susceptible to stress-related disorders and factors that change how an individual experiences stress either before, during, or after the event can be utilized to develop strategies to help reduce the incidence of these disorders. Recent preclinical findings from ketamine have revealed that experiential resilience factors sexually converge at the level of behavioral outcome, yet diverge at a neural circuit level. Not all paths to resilience are concordant between the sexes nor do all resilience factors operate equally at a circuit level. Identifying these mechanistic differences represents an important step in developing sex-specific treatments for stress-linked disorders.

Abbreviations

5-HT

serotonin

AS-PaRac1

activated synapse-targeting photoactivatable Rac1

BDNF

brain-derived neurotrophic factor

BLA

basolateral amygdala

DLS

dorsolateral striatum

DMS

dorsomedial striatum

DREADD

designer receptors exclusively activated by designer drugs

DRN

dorsal raphe nucleus

ERK

extracellular signal-regulated kinase

ES

escapable stress

HC

home cage control

IEG

immediate early gene

IS

inescapable stress

JSE

juvenile social exploration

mPFC

medial prefrontal cortex

mTOR

mammalian target of rapamycin

NMDAR

N-methyl-D-aspartate receptor

PL

prelimbic cortex

PTSD

posttraumatic stress disorder

RAM

robust activity marking system