Learning about stress: neural, endocrine and behavioral adaptations

Abstract

In this review, nonassociative learning is advanced as an organizing principle to draw together findings from both sympathetic-adrenal medullary and hypothalamic–pituitary–adrenocortical (HPA) axis responses to chronic intermittent exposure to a variety of stressors. Studies of habituation, facilitation and sensitization of stress effector systems are reviewed and linked to an animal’s prior experience with a given stressor, the intensity of the stressor and the appraisal by the animal of its ability to mobilize physiological systems to adapt to the stressor. Brain pathways that regulate physiological and behavioral responses to stress are discussed, especially in light of their regulation of nonassociative processes in chronic intermittent stress. These findings may have special relevance to various psychiatric diseases, including depression and post-traumatic stress disorder (PTSD).

Keywords:

• ACTH
• adrenal medulla
• catecholamines
• corticosterone
• facilitation HPA axis
• nonassociative learning
• sympathetic nervous system

Introduction

Between 1910 and 1936, the laboratories of two pioneering physician–scientists, Walter Bradford Cannon of Harvard Medical School and Hans Selye of McGill University, established the early foundation for the field of stress research (Cannon, 1914, 1915, 1932; Selye, 1936, 1951, 1973, 1978). This foundation was comprised of two distinct pillars, the sympathetic-adrenal medullary system and the hypothalamic–pituitary–adrenocortical (HPA) axis, reflecting the research emphases of Cannon and Selye, respectively (McCarty, 2016a,b). Since those early years of stress research, studies relating to these two primary stress-effector systems have largely been maintained as separate domains, with some laboratories specializing in studies of the HPA axis, while others have focused on the sympathetic-adrenal medullary system.

In this review, I will take a comprehensive look at both stress-effector systems and their patterns of response to chronic intermittent stress. My overarching goal is to situate this voluminous body of research findings within a theoretical framework linked to non-associative learning.

Nonassociative learning as an organizing principle

Nonassociative learning is a simple form of learning that is characterized by a change in the magnitude of response to a given stimulus as a result of repeated exposure to that stimulus. Nonassociative learning does not require temporal pairing between two different sensory stimuli or between a sensory stimulus and a corresponding response and it is not explained by sensory adaptation or muscle exhaustion. Nonassociative learning is a highly conserved pattern of behavior across all animal phyla, from protozoa to primates (Eisenstein et al., 1982; Lau et al., 2013). Three types of nonassociative learning have been described:

• Habituation: A decrease in the magnitude of response to repeated exposure to a relatively low intensity stimulus.

• Dishabituation: Exposure to a novel stimulus following habituation to another stimulus often leads to an enhanced response to the original stimulus. The novel, or dishabituating stimulus may be noxious or non-noxious.

• Sensitization: An increase in the magnitude of response to repeated exposure to a noxious or painful stimulus.

The path-breaking papers of Thompson & Spencer (1966) and Groves & Thompson (1970), together with a recent update and revision by Rankin et al. (2009), provide a solid theoretical framework from which to evaluate the published findings on neural, endocrine and behavioral responses to chronic intermittent stress (Table 1). However, it is important to keep in mind that several of the 10 parameters of habituation delineated by these investigators are not typically the focus of experiments on chronic intermittent stress. Indeed, the experimental models discussed by Thompson & Spencer (1966) permit studies of habituation, dishabituation and sensitization in a greatly compressed time scale compared to the majority of experiments on chronic intermittent exposure of laboratory animals and humans to stressful stimuli. A frequently referenced model system in the paper by Thompson & Spencer (1966) was the spinal flexion reflex of the spinal cat in response to electric shocks to the skin. As an example, with this model system, habituation and spontaneous recovery can be studied in the same animal in just over 120 min. More recently, studies of the gill withdrawal reflex in the sea hare, Aplysia californica, also permit a more compressed time scale for parametric studies of nonassociative learning (Kandel, 2001).

Table 1. The ten defining characteristics of habituation based upon the list originally presented by Thompson and Spencer (1966), and later revised by Rankin et al. (2009).

Criterion	Description
1	Repeated application of a stimulus results in a progressive decrease in some parameter of a response to an asymptotic level. This change may include decreases in frequency and/or magnitude of the response. In many cases, the decrement is exponential, but it may also be linear; in addition, a response may show facilitation prior to decrementing because of (or presumably derived from) a simultaneous process of sensitization.
2	If the stimulus is withheld after response decrement, the response recovers at least partially over the observation time (‘‘spontaneous recovery”).
3	After multiple series of stimulus repetitions and spontaneous recoveries, the response decrement becomes successively more rapid and/or more pronounced (this phenomenon can be called potentiation of habituation).
4	Other things being equal, more frequent stimulation results in more rapid and/or more pronounced response decrement and more rapid spontaneous recovery (if the decrement has reached asymptotic levels).
5	Within a stimulus modality, the less intense the stimulus, the more rapid and/or more pronounced the behavioral response decrement. Very intense stimuli may yield no significant observable response decrement.
6	The effects of repeated stimulation may continue to accumulate even after the response has reached an asymptotic level (which may or may not be zero, or no response). This effect of stimulation beyond asymptotic levels can alter subsequent behavior, for example, by delaying the onset of spontaneous recovery.
7	Within the same stimulus modality, the response decrement shows some stimulus specificity. To test for stimulus specificity/stimulus generalization, a second, novel stimulus is presented and a comparison is made between the changes in the responses to the habituated stimulus and the novel stimulus. In many paradigms (e.g. developmental studies of language acquisition), this test has been improperly termed a dishabituation test rather than a stimulus generalization test, its proper name.
8	Presentation of a different stimulus results in an increase of the decremented response to the original stimulus. This phenomenon is termed “dishabituation.” It is important to note that the proper test for dishabituation is an increase in response to the original stimulus and not an increase in response to the dishabituating stimulus (see point #7 above). Indeed, the dishabituating stimulus by itself need not even trigger the response on its own.
9	Upon repeated application of the dishabituating stimulus, the amount of dishabituation produced decreases (this phenomenon can be called habituation of dishabituation).
10	Some stimulus repetition protocols may result in properties of the response decrement (e.g. more rapid rehabituation than baseline, smaller initial responses than baseline, smaller mean responses than baseline, less frequent responses than baseline) that last hours, days or weeks. This persistence of aspects of habituation is termed long-term habituation.

Most experiments relating to habituation of neuroendocrine responses have involved exposing animals to a single stressor each day for periods ranging from several minutes to several hours. The number of daily exposures to a given stressor has varied from 4 to as many as 40 consecutive days. The duration of each daily stress exposure and the number of consecutive daily exposures to a given stressor vary over a wide range and are a reflection of the experiences of different groups of investigators to ensure, in part, that habituation of a given neural or endocrine parameter is fully established (McCarty, 1989).

Parametric studies of nonassociative learning (habituation, dishabituation, spontaneous recovery, etc.) using laboratory rats and chronic intermittent restraint as the stressor would require many control and experimental groups and a considerable investment of time and resources. A single restraint stress session is typically 30–120 min long and, in the case of one well-studied hormone, plasma corticosterone, levels of the hormone do not return to baseline levels for a prolonged period following termination of restraint stress (up to one hour).

Given these constraints, it would be difficult to subject a test animal to more than three restraint stress sessions per day and study HPA axis responses such that recovery to prestress baseline levels of corticosterone occurs after each stress session. In addition, spontaneous recovery from chronic intermittent restraint stress may take more than four weeks. This strong memory trace for the stressor is not surprising given that young adult laboratory rats exposed to a single, brief footshock during passive avoidance training demonstrate strong memory for the training footshock up to 21 days later (Gold & Korol, 2014). Few investigators have displayed an interest in conducting a systematic evaluation of the neuroendocrine responses to a given stressor or pairings of stressors against the 10 parameters of habituation (Rankin et al., 2009), and some of the parameters add little value to the aims of such experiments. However, such studies are essential to determine the extent to which neuroendocrine responses to chronic intermittent stress satisfy the 10 key characteristics of habituation, as determined from studies of simpler model systems. Time will tell if these issues will be addressed completely in freely behaving laboratory rats and mice through direct experimentation.

A prominent exception to this standard experimental design of repeated daily exposure to a given stressor was reported by De Boer et al (1988). They exposed laboratory rats to three consecutive 10-min periods of noise stress, with an intertrial interval of 30 min. Plasma levels of corticosterone and catecholamines showed clear signs of habituation by the third stress exposure. Importantly, plasma levels of corticosterone and the catecholamines returned to prestress basal levels following each of the three noise stress sessions. In a subsequent study, this group exposed laboratory rats to 20 brief noise stress bouts in a single day (De Boer et al., 1989). Such an experimental approach would permit a careful evaluation of additional parameters of habituation (e.g. spontaneous recovery) that have not as yet been studied.

A typical experimental design to study nonassociative processes relating to neural, endocrine or behavioral responses to chronic intermittent stress is summarized in Figure 1. Laboratory animals are often exposed to the same stressor each day for multiple days (often referred to as the homotypic stressor). Depending upon the intensity of the homotypic stressor, and to a lesser extent the environmental context in which the stressor is presented and the predictability of the stressor, the animal’s response on Day N may be diminished (habituation) or enhanced (sensitization) compared to first-time stressed controls. If an animal exposed each day to a low intensity, homotypic stressor is then exposed on Day N to a heterotypic (i.e. novel) stressor, the response is often enhanced when compared to matched, first-time stressed controls. This enhanced neuroendocrine response is referred to as facilitation.

Figure 1. Overview of typical experimental protocols to study habituation, facilitation and sensitization of the HPA axis and the sympathetic-adrenal medullary system to chronic intermittent stress. Numbers above the squares representing the Nth day of stressor exposure reflect the physiological response (e.g. plasma levels of ACTH or epinephrine) relative to first-time stressed, handled controls. In all three protocols, handled controls have a physiological response set at 100%. Open circles reflect days when animals were handled but not stressed and squares represent days when animals were stressed. Green squares reflect daily exposure of animals to a low-intensity homotypic stressor, leading to habituation of the physiological response. Orange squares represent a heterotypic, low-intensity stressor presented to animals that were previously habituated to another stressor (green squares) as well as to first-time stressed controls. Red squares indicate exposure to a high intensity stressor, leading to sensitization. **Significantly different from first-time stressed controls.

Facilitation is a term first introduced by Dallman et al (Akana et al., 1996; Dallman & Jones, 1973; Dallman et al., 1987) to reflect enhanced HPA responses of previously stressed animals following acute exposure to a heterotypic stressor. They argued that a central facilitory process must come into play during exposure of animals to a heterotypic stressor to overcome the negative feedback effects of elevated circulating corticosteroids on the HPA axis. In this review, I have adopted Dallman’s term, facilitation, to explain both enhanced HPA axis and sympathetic-adrenal medullary responses to a heterotypic stressor in animals previously exposed to several daily episodes of a homotypic stressor.

In the context of nonassociative learning, facilitation of a neuroendocrine response to a heterotypic stressor, which is similar to the effect of a dishabituating stimulus, is observed in animals previously exposed repeatedly to a homotypic stressor. If animals are again exposed to the homotypic stressor following facilitation, one could then assess the extent of dishabituation. However, I am unaware of any experiments that have explicitly studied dishabituation of neuroendocrine responses to stress in laboratory animals. This is certainly a missing link in this field of research that should be addressed in the near future (Table 1).

In the experiments on adaptation of the HPA axis and the sympathetic-adrenal medullary systems that are described in the following sections, I have included methodological details regarding stressor intensity, duration and frequency as these details are often critical in interpreting results and discerning consistencies (or inconsistencies) across experiments. Many of these experiments have employed restraint and immobilization as chronic intermittent stressors. Immobilization typically involves using adhesive tape to secure the four limbs of a test animal, usually a laboratory rat, to metal mounts attached to a supporting platform. Head movement is often restricted by concentric U-shaped metal loops mounted onto the platform at the rat’s neck region. The struggling that is typical of the first several minutes of an immobilization session places great pressure on the four limbs and animals are often physically exhausted at the end of an immobilization session. This model of stress has been used since the mid-1960s by Kvetnansky et al. at the Institute of Experimental Endocrinology in Bratislava, Slovakia (Kvetnansky & Mikulaj, 1970) and adopted by many other investigators.

In contrast, restraint stress involves placing a laboratory rat or mouse into a clear Plexiglas tube or a wire mesh container for a fixed period of time. The animal is restricted in its movements, but its limbs are not fastened and there is little risk of injury to the animal. There is general agreement that immobilization is a more intense stressor than restraint (Herman & Cullinan, 1997).

Chronic intermittent stress and the hypothalamic–pituitary–adrenocortical (HPA) axis

A critical response to homeostatic challenges, including stressful stimuli, is the synthesis and release of glucocorticoids from the adrenal cortex. In laboratory rats and mice, the principle glucocorticoid released from the adrenal cortex is corticosterone. Regulation of corticosterone synthesis and secretion involves a complex interplay of multiple brain regions, whose projections converge on the paraventricular nucleus (PVN) of the hypothalamus. The cell bodies of parvocellular neurosecretory neurons within the PVN release several peptide messengers into the hypophyseal portal blood supply at the level of the median eminence, including corticotropin-releasing hormone (CRH) and arginine vasopressin (AVP). CRH and AVP stimulate the synthesis and release of adrenocorticotropin (ACTH) from the anterior pituitary, which is mediated by CRH1 receptors. Increases in circulating ACTH then stimulate the synthesis and release of corticosterone from the adrenal cortex, which provides negative feedback at the level of the anterior pituitary, the hypothalamus, and higher brain centers to dampen further ACTH secretion (Figure 2). In addition, stress-induced release of corticosterone enhances glucose availability and suppresses immune system function and multiple intercellular mediators, which, if left unchecked, might cause damage to tissues and organs (Munck et al., 1984).

Figure 2. Influence of stress on the HPA axis, illustrating negative feedback effects of circulating glucocorticoids at the level of the pituitary, the paraventricular nucleus (PVN), and higher brain centers (e.g. hippocampus). CRH: corticotropin releasing hormone; AVP: arginine vasopressin; ACTH: adrenocorticotropin.

Recently, Frank et al. (2013) suggested that a different glucocorticoid-dependent process unfolds in brain following exposure to a stressor. Acute stress-induced increases in circulating glucocorticoids set in motion a delayed process of microglial sensitization in brain to prepare the organism for the possibility of a subsequent immunologic challenge. For example, if an animal previously exposed to an acute stressor or to a chronic intermittent stress regimen is then confronted with a peripheral inflammatory challenge (e.g. injection of lipopolysaccharide), the neuroinflammatory response is enhanced relative to unstressed controls. The authors referred to this process as “a neuroendocrine alarm signal of danger”.

CRH2 receptors are also involved in responses to stress, especially adaptation to and recovery from stress, and are broadly distributed in brain as well as peripheral tissues (Dautzenberg et al., 2001). CRH2 receptors selectively bind urocortins II and III, which are members of the CRH family of peptides, and they, too, are distributed widely in brain and peripheral tissues (Hsu & Hseuh, 2001; Lewis et al., 2001; Reyes et al., 2001).

The PVN receives stressor-specific input from several brain areas, including stimulatory input from brainstem catecholaminergic cell groups (A2, C1, C2, C3) and forebrain nuclei (central nucleus of the amygdala, posterior-cortical amygdaloid nucleus, medial amygdaloid nucleus and the lateral portion of the bed nucleus of the stria terminalis (BNST). Inhibitory input to the PVN arises from the hippocampus, lateral septum, ventral subiculum, medial portion of the BNST, the medial preoptic area, the dorsomedial and ventromedial hypothalamus, arcuate nucleus and the suprachiasmatic nucleus (Herman, 2013; Herman & Cullinan, 1997; Herman et al., 2003; Pacak & Palkovits, 2001).

Corticosterone exerts its central effects by binding to two classes of nuclear receptors, mineralocorticoid receptors (MRs) and glucocorticoid receptors (GRs) (de Kloet, 2000, 2013; de Kloet et al., 2005; McEwen, 2007). MRs are highly expressed in limbic brain areas, while GRs are found in highest concentrations in the PVN, the hippocampus, the amygdala and ascending catecholaminergic neurons. Early studies by Reul & de Kloet (1985) revealed that GRs have a 10-fold lower affinity for corticosterone compared to MRs. These differences in affinity mean that GRs are only fully activated when levels of corticosterone are elevated during exposure to a stressor or during peaks in the ultradian and circadian rhythms of corticosterone secretion. Activation of GRs is a crucial link in directing behavioral responses to stressful stimulation (Ter Horst et al., 2012).

Recent research has also documented occupation by corticosterone of low affinity, membrane-associated MRs and GRs (Joëls et al., 2008; Karst et al., 2005; Olijslagers et al., 2008). Membrane-associated MRs facilitate excitatory neurotransmission by enhancing the release of glutamate from presynaptic nerve terminals and post-synaptically by increasing neuronal excitability. In contrast, membrane associated GRs primarily exert inhibitory influences via GABA-ergic and endocannabinoid pathways (Tasker et al., 2006). These rapid nongenomic actions of membrane-associated MRs and GRs add an additional level of complexity to stress circuits in brain that regulate the HPA axis.

Several investigators have characterized stressors based upon their pattern of activation of central stress-sensitive pathways. For example, one may classify stressors as physical or psychological or as interoceptive versus exteroceptive (de Kloet, 2013; Grissom & Bhatnagar, 2009; Joëls & Baram, 2009). Alternatively, Herman & Cullinan (1997) proposed the existence of two distinct central pathways that regulate the HPA axis response to stress. The processive pathway involves input to the PVN from limbic structures and requires multisensory processing of the stressor, which is of low-to-moderate intensity and does not constitute an immediate threat to survival. Stressors in this category include restraint, exposure to a strange environment, noise stress and footshock, all of which constitute manageable behavioral challenges to the test animal. In contrast, the systemic pathway conveys information about closely monitored physiological variables that are essential for survival, including blood oxygenation (e.g. exposure to ether vapors as a stressor), blood glucose levels (insulin-induced hypoglycemia as a stressor) and blood volume (hemorrhage as a stressor). The systemic pathways arise from brainstem areas that project directly to the PVN, given their critical role in survival of the animal.

HPA responses to processive stressors are more likely to habituate following chronic intermittent exposure because animals gain experience with the nature of the stressor, its intensity, when it occurs, and how long and how frequently it is applied. With the first exposure to a processive stressor, the HPA response often overshoots the level required to respond to the physiological demands exacted by the stressor. However, as this information is processed, stored and retrieved over multiple stressor exposures, test subjects are able to mount HPA responses that are reduced compared to the initial exposure and that match more closely with the physiological demands of the stressor. One could argue that animals that have diminished HPA responses to chronic intermittent stressors would conserve valuable energy and enhance their survival over the longer term.

When HPA responses to a heterotypic stressor are enhanced in animals previously exposed repeatedly to a homotypic stressor, this, too, may convey a selective advantage. Just as living in a predictable yet modestly stressful environment often leads to habituation of HPA axis responses, it could also be advantageous for facilitation of HPA responses when there is a sudden departure from expectations. However, if the heterotypic stressor does not constitute a threat to survival, animals may actually habituate to the new stressor at a greater rate than normal, thus further conserving energy and enhancing survival.

Grissom & Bhatnagar (2009) provided an illuminating analysis of nonassociative properties of HPA responses to stress. Their detailed review of the literature on HPA responses to stress, coupled with the updated list of parameters of habituation described by Rankin et al. (2009), indicated that five of the 10 parameters have been satisfied (numbers 1, 4, 5, 8 and 10), one parameter has been largely confirmed (number 9), and four parameters have not been satisfied or have not been tested adequately (numbers 2, 3, 6 and 7) (Table 2).

Table 2. Do plasma corticosterone responses to chronic intermittent stress satisfy the ten criteria that characterize habituation? Characteristics of habituation included below are based upon the list originally presented by Thompson and Spencer (1966), and later revised by Rankin et al. (2009). This format was adopted from Grissom and Bhatnagar (2009).

Criterion	Met: Yes/No	Abbreviated description of parameter	Representative publications
1	Yes	Repeated stress sessions result in decreased response	Fernandes et al., 2002; Grissom et al., 2007; Natelson et al., 1988
2	No	Spontaneous recovery when stressor is withheld	No definitive test thus far; spontaneous recovery from restraint may take many weeks (Bhatnagar et al., 2002)
3	No	Potentiation of habituation	No relevant studies
4	Yes	The more rapid the frequency of stressor presentation, the more pronounced is habituation	De Boer et al., 1990; Ma and Lightman, 1998; Zhang et al., 2014
5	Yes	The weaker the stressor, the more pronounced is habituation	Prone restraint versus supine restraint (Natelson et al., 1988); restraint versus immobilization (Girotti et al., 2006; Weinberg et al., 2010); low versus high footshock levels (Rabasa et al., 2011a)
6	No	Beyond the zero or asymptotic response level	No relevant studies
7	No	Stimulus generalization in habituation	No relevant studies
8	Yes	Dishabituation upon exposure to a novel stressor	Bhatnagar and Dallman, 1998; Dallman and Jones, 1973; Daviu et al., 2014; Fernandes et al., 2002
9	?Yes?	Habituation of a dishabituated response	No direct tests; however, dishabituating stimuli that have been shown to habituate include restraint (Bhatnagar & Dallman, 1998) and placement on an elevated platform (Storey et al., 2006)
10	Yes	Long-term habituation	Bhatnagar et al., 2002

To amplify on the summary in Table 2, an impressive array of studies has reported significant habituation of HPA responses to chronic intermittent stress as measured by circulating levels of ACTH and/or corticosterone (Armario et al., 2012; Dallman, 1993; Martí & Amario, 1998). These include experiments that utilized restraint stress (Babb et al., 2014; Barnum et al., 2007; Bhatnagar et al., 2002; Bingham et al., 2011; Cole et al., 2000; Fernandes et al., 2002; Gray et al., 2012; Grissom & Bhatnagar, 2011; Grissom et al., 2007; Helm et al., 2004; Ma et al., 1999; Masini et al., 2012b; Natelson et al., 1988; Pace et al., 2001; Stamp & Herbert, 2001; Terrazzino et al., 1995; Uchida et al., 2008; Weinberg et al., 2010; Zelena et al., 2003), restraint stress plus partial immersion in 21 °C water (Mizoguchi et al., 2001), low-to-moderate intensity footshock stress (Pitman et al., 1990, 1995; Rabasa et al., 2011a), forced swimming in warm water (Rabasa et al., 2013), immobilization (Armario et al., 1988; Daviu et al., 2014; Garcia et al., 2000; Lachuer et al., 1994; Márquez et al., 2004; Rabasa et al., 2011b), handling (Dobrakovova et al., 1993), exposure to a novel environment (Bassett et al., 1973; File, 1982; Pfister, 1979), cold exposure (Kant et al., 1983), social defeat/social stress (Bhatnagar et al., 2006), and noise exposure (Day et al., 2009; De Boer et al., 1989; Masini et al., 2008, 2012a; Nyhuis et al., 2010a,b; Sasse et al., 2008). Taken together, these results provide strong support for HPA axis habituation to chronic intermittent stress. In addition, habituation has been consistently observed across a wide range of processive stressors of varying intensities.

Fewer studies have investigated facilitation of a habituated HPA response. In an elegant series of experiments, Weinberg et al. (2009) reported habituation of HPA axis activity in rats exposed for 30 min per day for 14 consecutive days to the odor of a natural predator, the ferret (Mustela putorius). On the 15th day, rats were again exposed to ferret odor (the homotypic stressor) or were restrained for 30 min (the heterotypic stressor). Day 15 results indicated a significant habituation of plasma corticosterone responses in rats exposed to the homotypic stressor (ferret odor) for the 15th time compared to rats exposed to ferret odor for the first time. In contrast, rats exposed to ferret odor on days 1–14 had a significant elevation in plasma corticosterone when exposed to the heterotypic stressor of restraint compared to first-time restraint-stressed controls. Similar trends were apparent for plasma ACTH, although the differences between treatment groups did not attain statistical significance (ps > 0.05). Interestingly, habituation to the homotypic stressor developed much sooner than facilitation to the heterotypic stressor. This finding in the HPA axis of differential rates of habituation versus facilitation adds further support to the argument that these two forms of nonassociative learning are regulated by independent processes (Groves & Thompson, 1970; Thompson & Spencer, 1966; Rankin et al., 2009).

Other studies using laboratory mice or rats have also reported facilitation of the HPA axis to a heterotypic stressor following habituation to a homotypic stressor. They include: immobilization as the homotypic stressor and ether vapors as the heterotypic stressor (Hashimoto et al., 1988), immobilization as the homotypic stressor and tail shock as the heterotypic stressor (Armario et al., 1988), restraint as the homotypic stressor and lipopolysaccharide (LPS) as the heterotypic stressor (Fernandes et al., 2002), immobilization as the homotypic stressor and either insulin (Dronjak et al, 2004) or ether vapors as the heterotypic stressor (Hauger et al., 1990), 4–6 hours per day of cold exposure as the homotypic stressor and immobilization as the heterotypic stressor (Ma & Morilak, 2005; Pardon et al., 2003), chronic immobilization as the homotypic stressor and forced swimming as the heterotypic stressor (Garcia et al., 2000), chronic social defeat/social stress as the homotypic stressor and restraint as the heterotypic stressor (Bhatnagar & Vining, 2003), chronic intermittent cold stress (4–6 °C for 4 hours per day for 7 consecutive days) as the homotypic stressor and restraint as the heterotypic stressor (Bhatnagar & Dallman, 1998), chronic variable stress as the homotypic stressor and acute noise stress at 108 dBA as the heterotypic stressor (McGuire et al., 2010), and restraint as the homotypic stressor and noise stress and the heterotypic stressor (Spiga et al., 2009).

In an interesting methodological twist, Kearns & Spencer (2013) utilized an unexpected increase in the duration of a homotypic stressor to eliminate the habituation of the corticosterone response. Rats were exposed to restraint stress for 10 min per day for days 1–4 (homotypic stressor). On day 5, animals were again exposed to restraint stress for 10 min (expected duration) or for 30 min (unexpected increase in duration). Rats that were restrained for 10 min per day had significant reductions active struggling behavior and in plasma corticosterone on day 5 compared to controls that were restrained for the first time. In contrast, animals that were exposed to 30 min of restraint stress had levels of corticosterone that were similar to first-time stressed controls. In addition, rats exposed to an extended restraint stress session on day 5 had an elevation in active struggling behavior that started at 9 min into the restraint session. Although this was not an example of facilitation of corticosterone levels, it does point clearly to the fact that animals learn many aspects about the homotypic stressor, including its duration. A departure from the expected duration of the homotypic stressor over-rides the habituation process (Kearns & Spencer, 2013).

In summary, these experiments underscore the fact that facilitation of HPA responses is a robust response that occurs across a variety of homotypic-heterotypic stressor pairings. In addition, facilitation has been reported for heterotypic stressors that are processive (e.g. swim stress, restraint stress, noise stress) or systemic (e.g. ether vapors, LPS, insulin-induced hypoglycemia) in nature.

Sensitization of the HPA axis to chronic intermittent exposure to an intense stressor has also been reported. Examples of intense stressors include the use of inescapable electric shock to the tail (O’Connor et al., 2004; Servatius et al., 1994) or to the rump (Pitman et al., 1990) and a combination of immobilization, light flashes and noise (Vogel & Jensh, 1988).

The references included above are by no means an exhaustive listing of all relevant studies; rather, they provide a clear indication of the consistent findings in this area. Unfortunately, there are reports in the literature that have failed to detect habituation of HPA axis responses to chronic intermittent stress as assessed by measures of plasma ACTH and/or corticosterone. These include experiments with stressors such as immobilization (Armario et al., 1988; Clement et al., 1998; Dronjak et al., 2004), footshock (Bassett et al., 1973; Hajós-Korcsok et al., 2003; Kant et al., 1985), exposure to a predator (Blanchard et al., 1998), restraint stress in Fischer 344 rats (Dhabhar et al., 1997; Uchida et al., 2008), exposure to a hole board (Gagliano et al., 2008), daily handling and/or subcutaneous injections of saline (Hodges & Mitchley, 1970) and corticosterone responses to restraint (Garcia-Iglesias et al., 2013; Kant et al., 1985).

Studies that have failed to observe facilitation when previously stressed laboratory animals were exposed to a heterotypic stressor include reports by Armario et al. (1988); Babb et al. (2014); Chung et al. (2000); Masini et al. (2012b); Simpkiss & Devine (2003); and Terrazzino et al. (1995). Needless to say, these experiments are challenging to design and one can often second-guess choices for the homotypic or heterotypic stressor and their pairing, their intensities and the number of days for chronic intermittent stress prior to exposure to the heterotypic stressor.

Recently, Rabasa et al (2015) have offered a new perspective on HPA responses to chronic intermittent stress, arguing that in many instances, the defining parameters of habituation are not satisfied. Their major areas of concern, which were addressed in an extensive series of experiments, include (i) habituation developed more rapidly and was of greater strength in animals exposed to more intense stressors, (ii) even a single exposure to an intense stressor produced habituated responses that persisted for up to several weeks later, (iii) a single prolonged exposure to an intense stressor resulted in a greater habituation of the HPA response than multiple, brief exposures to the same stressor.

These investigators rejected nonassociative learning as an explanation for the pattern of HPA responses to chronic stress, as reported from their laboratory and from other laboratories. Instead, they proposed a new term, tolerance, to describe adaptation to repeated daily exposure to stressful stimulation (Rabasa et al., 2015). By taking this approach, they eliminated concerns about satisfying the ten parameters of habituation, as enumerated by Rankin et al. (2009) and summarized in Table 1. Unfortunately, they did not offer a description of the defining characteristics of tolerance and they were unable to replicate the basic findings of several well-established research groups, including De Boer et al. (1990) and Hill et al. (2010). In addition, they largely ignored the many studies of the HPA axis and the sympathetic-adrenal medullary system cited in this review that have produced findings that are consistent with nonassociative learning theory. Finally, Rabasa et al did not fully engage the literature on sensitization and facilitation of neuroendocrine responses to chronic stress and explain these responses within the context of tolerance.

With all due respect to the well-placed concerns of Rabasa et al. (2015), I conclude from the analysis of Grissom & Bhatnagar (2009) and additional findings included in the present review that adaptation of the HPA axis to chronic intermittent stress is largely consistent with a neuroendocrine form of habituation. I do not imply, however, that all 10 parameters of habituation have been satisfied or that all experimental results to date are consistent with nonassociative processes. Rather, I suggest that habituation provides a useful framework for the design of experiments and the interpretation of findings from studies of physiological adaptation to chronic intermittent stress. It is also important to keep in mind that the 10 parameters of habituation as presented by Rankin et al. (2009) are based in large measure upon experiments in models systems that lack the complexity of freely behaving laboratory rats and mice exposed to stressful stimulation. Thus, it may be that some of these parameters, which were formulated in model systems, may not be fully satisfied in studies of neuroendocrine responses to stress.

Chronic intermittent stress and the sympathetic-adrenal medullary system

The pioneering experiments of Cannon (1914, 1915) and Cannon & de la Paz (1911) on epinephrine secretion from the adrenal medulla depended upon bioassays for the quantification of circulating epinephrine. It was not until the mid-1970s that analytical methods were developed that made possible the routine measurement of picogram quantities of norepinephrine (NE) and epinephrine (EPI) in plasma samples of unrestrained laboratory rats (Peuler & Johnson, 1977). In the 1990s, newer analytical methods permitted measurement of plasma catecholamines and their metabolites by high-performance liquid chromatography (HPLC) with electrochemical detection (Holmes et al., 1994).

Because of the remarkable sensitivity of the sympathetic-adrenal medullary system of laboratory rats to handling, restraint and venipuncture necessary for the collection of blood samples, methods were also developed for the placement of indwelling catheters in arteries or veins of laboratory rats (Chiueh & Kopin, 1978; Paulose & Dakshinamurti, 1987; Steffens, 1969). Indwelling, chronic catheters provided researchers with the advantage of obtaining blood samples from freely behaving animals in the absence of handling or other disturbances. Remote sampling of blood via an indwelling arterial or venous catheter is an absolute requirement for obtaining accurate baseline measures of circulating NE and EPI in laboratory rats and tracking responses to stressful stimulation (Chiueh & Kopin, 1978; Kvetnansky et al, 1978). Sampling of blood from arterial or venous catheters is also important to capture the differing time courses of ACTH and corticosterone during and after stressful stimulation (Kearns & Spencer, 2013).

Circulating NE is derived from two principle sources. The primary neurotransmitter of postganglionic sympathetic nerves is NE. Sympathetic nerves innervate most of the tissues of the body (e.g. heart, blood vessels, spleen, intestines, etc.), where NE is released from sympathetic nerve endings and binds to adrenergic receptors on target cells. Some of the locally released NE (5–20% depending on the morphology of the neuroeffector junction) leaks away unmetabolized and enters the circulation. NE is also released from the adrenal medulla directly into the circulation upon stimulation of adrenal medullary chromaffin cells by sympathetic preganglionic nerves that release acetylcholine (Goldstein et al., 1983). Figure 3 shows an overview of the organization of the sympathetic-adrenal medullary system.

Figure 3. Neural pathways regulating the sympathetic-adrenal medullary system response to stress. Depicted in this illustration are sympathetic postganglionic projections to skin and a section of a blood vessel, where norepinephrine is released from nerve terminals (blue dots), and sympathetic preganglionic projections to the adrenal medulla, where epinephrine (EPI) is released into the circulation from chromaffin cells. PVN: paraventricular nucleus of the hypothalamus; LC: locus coeruleus; A5/C1: A5 noradrenergic and C1 adrenergic cell groups; RN: raphe nuclei; IML: intermediolateral cell column of the spinal cord; SG: sympathetic ganglion; EPI: epinephrine.

Yamaguchi & Kopin (1979) employed the pithed rat preparation to study the release of peripheral catecholamines when the spinal cord was stimulated electrically. During sustained electrical stimulation of the spinal cord, plasma levels of NE peaked later than levels of EPI (approximately a 3-minute delay) given that EPI is released directly into the circulation from the adrenal medulla whereas NE that is released from sympathetic nerve terminals must diffuse away from the neuroeffector junction before it enters the circulation. These investigators also reported that 25–40% of the circulating NE during direct stimulation of sympathetic outflow was derived from the adrenal medulla. Kvetnansky et al. (1979) found that approximately 30% of the NE in blood was derived from the adrenal medulla during the early stages of immobilization stress. After one hour of immobilization stress, there was a change-over, such that nearly 100% of the circulating NE was from sympathetic nerves.

Finally, McCarty & Kopin (1979) reported that the intensity of stressful stimulation affected the release of NE from the adrenal medulla. Under basal conditions, the adrenal medulla contributed a negligible amount of NE in the circulation; under mildly stressful conditions (handling and transfer to a strange environment), the percentage of NE from the adrenal medulla increased to 30%; and following an intense stressor (5 min of intermittent footshock), the percentage of NE from the adrenal medulla was 45%.

Using microdialysis of the adrenal medulla of laboratory rats, Kuzmin et al. (1995) studied the effects on catecholamine secretion of acute immobilization stress for 60 min. Adrenal medullary dialysate was collected before, during and after immobilization and later analyzed for content of NE and EPI. Their results indicated that under basal conditions, EPI/NE in dialysate was 3.7 ± 0.4 pmol/ml, approximately the ratio of the two catecholamines in adrenal tissue. There were dramatic increases in levels of both catecholamines at 15 min of immobilization, but the ratio of EPI/NE remained approximately 4.0. In contrast, 2-deoxy-d-glucose (500 mg/kg, i.v.) elicited a relatively selective increase in dialysate levels of EPI, while levels of NE remained constant. EPI/NE peaked at approximately 10.0 one hour after injection of 2-deoxy-d-glucose (Kuzmin et al., 1995). In an earlier study that combined adrenal medullary microdialysis and measurement of circulating catecholamines, Kuzmin et al. (1990) reported that plasma NE from the adrenal medulla of laboratory rats increased from 9% under basal conditions to 50% during hemorrhage.

In contrast to the complex origin of circulating NE, the origin of circulating EPI is much more direct. In laboratory rats, most if not all of the circulating EPI is derived from the adrenal medulla under basal conditions and during and following stressful stimulation (Kvetnansky et al., 1979; McCarty & Kopin, 1979; Yamaguchi & Kopin, 1979). Extra-adrenal sources of the EPI-synthesizing enzyme, phenylethanolamine N-methytransferase (PNMT) and measurable tissue levels of EPI have been reported by several laboratories, but these tissues do not appear to contribute in a significant manner to circulating EPI levels in laboratory rats. For example, EPI has been quantified in cardiac tissue (Caramona & Soares da Silva, 1985) and PNMT mRNA levels have been detected in several tissues, including skin (Pullar et al., 2006), spleen (Andreassi et al., 1998; Warthan et al., 2002), thymus (Andreassi et al., 1998; Warthan et al., 2002), cardiac tissue (Goncalvesova et al., 2004; Krizanova et al., 2001; Kvetnansky et al, 2006) and sympathetic ganglia (Kubovcakova et al., 2006; Kvetnansky et al., 2006).

The rate of removal of catecholamines from blood is relatively rapid, with a half-life of just over one minute (Yamaguchi & Kopin, 1979; Goldstein et al., 1983). Plasma levels of NE and EPI have a significant capacity to increase above basal levels, and the increments in plasma levels of NE and EPI increase monotonically with increasing intensities of acutely stressful stimulation. In contrast, plasma levels of corticosterone do not have as great of a dynamic range of response and do not track well with increasing intensities of acutely stressful stimulation (De Boer et al, 1990; Natelson et al., 1981, 1987). Finally, plasma levels of NE and EPI peak sooner and return toward baseline levels more rapidly than corticosterone following a brief period of stressful stimulation (De Boer et al., 1989). In addition, circulating corticosterone dampens the release of NE from sympathetic nerve endings during stressful stimulation, probably by acting at central sites that regulate sympathetic outflow (Kvetnansky et al., 1993).

One study has examined the genetic determinants of plasma levels of NE during various phases of an acute session of immobilization stress (Klimes et al., 2005). These investigators identified two genomic regions on chromosome 10 in rats that contain quantitative trait loci (QTL) that influence the sympathetic nervous system response to acute immobilization. Their findings indicate that temporal phases of the stress response to an acute stressor may have specific genetic determinants that can be identified and studied experimentally.

Several reports have described habituation of plasma NE and EPI responses to chronic intermittent stress. In a series of interesting studies, De Boer et al first used partial immersion in 35 °C water in a cage similar to the home cage for 15 min daily as the stressor. Plasma NE and EPI responses to this stressor decreased progressively and significantly over five daily sessions of water immersion. By comparison, if the five stress sessions occurred every 72 h instead of every 24 h, the rate of habituation of plasma NE and EPI responses was much slower (De Boer et al., 1990).

Using noise as a stressor (100 dBA for 10 min, intertrial interval of 30 min, three sessions in one day), De Boer et al. (1988) found that plasma levels of NE and EPI were reduced significantly over the three trials, consistent with a process of habituation. In a follow-up experiment (De Boer et al., 1989), shorter, more frequent noise exposures (100dBA, with 20 sessions in a single day) that were regularly spaced (4 min in duration every 16 min) or irregularly spaced (4 min in duration, mean intertrial interval of 16 min, range of 2–30 min) were employed. Plasma levels of EPI were reduced significantly on the 20th noise exposure compared to the first noise exposure in the regular and irregular groups compared to controls. Twenty-four hours later, the plasma EPI response to another noise exposure remained reduced in both groups compared to controls.

In contrast, the response patterns to noise stress for plasma NE were more complex. Specifically, plasma NE levels for rats in the irregularly spaced group were reduced significantly on the 20th session and the follow-up session the next day compared to controls. In contrast, in rats with regularly spaced noise stress sessions, basal plasma levels of NE were elevated prior to the twentieth noise stress session compared to controls or rats in the irregular group. With noise onset, plasma NE increased slightly at 0.5 min and then decreased progressively from 3.5 min after noise onset to 4 min after noise offset, then ending with a slight increase at 12 min after noise offset. On the following day, rats in the regularly spaced group had elevated NE levels at 0.5 min of noise onset, with levels declining progressively over the next 15 min.

These results indicate that plasma EPI habituated to repeated noise stress more rapidly than did plasma NE, and the habituated response was still evident 24 h after the 20th noise stress session. Surprisingly, there was no evidence that stressor predictability resulted in more rapid habituation of plasma catcholamine responses to repeated brief episodes of noise stress. In fact, in the case of plasma NE, the opposite was true. The irregularly spaced group actually displayed a more consistent pattern of habituation to repeated noise stress compared to the regularly spaced group (De Boer et al., 1989).

Kvetnansky et al. (1984), using laboratory rats, employed a 2.5-h period of immobilization stress each day for 1 or 40 consecutive days. Basal plasma levels of NE and EPI were elevated significantly in rats the morning following the 39th daily session of immobilization stress compared to unstressed controls. These elevated levels of both catecholamines may represent an anticipatory response to that day’s session of immobilization. In contrast, rats immobilized for the 40th time had significantly lower increments above baseline in levels of plasma NE and EPI compared to first-time immobilized animals, especially during the initial 20 min of immobilization.

In spite of the fact that immobilization is a relatively intense stressor (Herman & Cullinan, 1997), these findings are consistent with habituation of the plasma NE and EPI responses to chronic intermittent immobilization stress. Dronjak et al. (2004) replicated these findings by immobilizing rats for 2 h per day for 41 consecutive days. Plasma levels of NE, and especially EPI, were reduced significantly in rats immobilized for the 41st time compared to naïve controls immobilized for the first time.

Shaker stress is an interesting model to study habituation of neurohormonal responses to acute versus chronic stress. Laboratory rats in their home cages are placed on a shaker, no pain is involved and no handling required, and intensity of the stressor can be manipulated by changing shaker frequency (Nakata et al., 1993). Hashiguchi et al. (1997) compared the hormonal responses to 5 or 30 min of shaker stress (150 cycles per minute, 3.8 cm displacement) administered for one day or each day for 10 consecutive days. Plasma levels of EPI were reduced significantly after repeated shaker stress compared to first-time stressed controls in groups that were stressed for 5 min or 30 min per session. There was a trend for plasma NE levels to be reduced in repeatedly stressed rats, but the decreases were not statistically significant.

Other reports of habituation of plasma catecholamine responses to chronic intermittent stress include experiments that employed the following stressors: in four related studies, each encompassing 27 consecutive days, swim stress at 34 °C for 30 min each day (Konarska et al., 1990a), restraint stress for 30 min each day (Konarska et al., 1989a, 1990b), intermittent footshock for 10 min each day (Konarska et al, 1989a), or a variable stressor combination (footshock, restraint and swim stress at variable times each day) (Konarska et al., 1990b). In each study, plasma NE and EPI responses of chronically stressed rats were compared to those of handled, first-time stressed controls (McCarty et al., 1988).

There are fewer experiments on habituation of plasma catecholamines to chronic intermittent stress compared to the body of research on habituation of the HPA axis to chronic intermittent stress. In addition, the range of homotypic stressors employed to study habituation of plasma catecholamine responses is more limited. In spite of these limitations, it is clear that habituation of plasma catecholamine responses to chronic intermittent stress is a well-established phenomenon that comports well with the literature on habituation of the HPA axis.

Several studies of facilitation of plasma catecholamine responses to a heterotypic stressor following chronic intermittent exposure to a homotypic stressor have been described. The first study of this kind was reported by Kvetnansky et al. (1984), using immobilization stress (150 min daily for 40 consecutive days) as the homotypic stressor and acute exposure for 400 seconds to the Noble-Collip Drum (NCD, Noble & Collip, 1942) as the heterotypic stressor. The NCD represents an extreme form of traumatic stress, such that a ledge in a rapidly rotating drum carries a rat to the top of the drum, and then the rat drops to the bottom of the drum and the process begins again. Rats exposed to chronic intermittent immobilization stress for forty consecutive days exhibited facilitation of plasma NE and EPI responses during acute NCD stress on day 41 compared to control rats subjected to the NCD for the first time. The peak levels of NE and EPI that were attained when previously immobilized rats were exposed to the NCD were remarkably high; in fact, they may be the highest levels of plasma catecholamines ever recorded in freely behaving animals. They were comparable to plasma levels of both catecholamines in blood samples of immobilized rats collected by decapitation when all tonic supraspinal inhibitory influences are eliminated (refer to Table 3).

Table 3. Mean peak plasma levels of norepinephrine (NE) and epinephrine (EPI) in laboratory rats exposed to (i) immobilization stress 1 time or daily for 40 consecutive days, (ii) exposed to immobilization stress daily for 40 consecutive days and then placed in the Noble–Collip Drum (NCD) on day 41, (iii) exposed to the NCD for one time or daily for eight consecutive days or (iv) decapitated (Decap) during the first immobilization or the 30th consecutive daily immobilization. Data were estimated from results presented in Kvetnansky et al. (1984).

Stressor treatment	Plasma NE (ng/ml)	Plasma EPI (ng/ml)
Immobilized 1 time	0.95	2.13
Immobilized 40 times	0.71^a	1.12^a
Immobilized 40 times + NCD	5.49^b	29.90^b
NCD 1 time	2.70	7.31
NCD 8 times	4.82^a	12.79^a
Immobilized 1 time + Decap	0.99	2.75
Immobilized 30 times + Decap	6.76^a	20.13^a

^a p < 0.05 compared to first-time stressed matched controls.

^b p< 0.05 compared to first-time stressed NCD group.

Additional examples of facilitation of plasma catecholamine responses of chronically stressed laboratory rats come from a series of experiments by Konarska and colleagues (1989b). Specifically, rats were exposed for 26 consecutive days to one of three homotypic stressors for 30 min each day: restraint stress in a Plexiglas tube, swim stress in water maintained at 18 °C, or inescapable footshock (1.0 mA, 0.5-s duration, every 5 s). Rats were prepared with chronic tail artery catheters after the 26th stress session and were allowed to recover from surgery on day 27. On day 28, chronically stressed rats were exposed to a heterotypic stressor as follows: restraint-stressed animals were exposed to footshock stress, swim-stressed animals were exposed to restraint stress, and footshock-stressed animals were exposed to swim stress at 24 °C. For each of the three groups, there were matched groups of first-time stressed control animals. The results were consistent across the three pairings, with greater plasma NE and EPI responses to the heterotypic stressor compared to first-time stressed controls. These differences were statistically significant for the swim stress – restraint and the footshock – swim stress pairings (ps < 0.01) and approached but did not attain statistical significance for the restraint – footshock pairing (0.10 >p > 0.05). These enhanced plasma catecholamine responses were initially characterized as sensitization rather than facilitation (Konarska et al, 1989b).

Dronjak et al studied facilitation of plasma catecholamine responses by employing immobilization as the homotypic stressor (2 h per day for 41 consecutive days) and then, on day 42, one of three systemic stressors was employed as the heterotypic stressor: insulin (5 IU/kg, i.a.), 2-deoxy-d-glucose (2-DG, 500 mg/kg, i.a.) or cold stress (4 °C for 2 h). Administration of insulin resulted in significantly greater increases in plasma NE in previously immobilized rats at 15 min postinjection and in plasma EPI at 60 and 120 min postinjection. Administration of 2-DG resulted in enhanced plasma NE responses in previously immobilized rats at 30, 60 and 120 min postinjection and in plasma EPI at 15 and 30 min postinjection. Plasma levels of both catecholamines were similar in controls and previously immobilized rats exposed to cold stress continuously for 2 h (Dronjak et al., 2004).

Using borderline hypertensive rats (BHRs) predisposed to develop hypertension (Lawler et al., 1981), Mansi & Drolet (1997) utilized restraint plus air jet puff stress (30 min per day, 5 days per week for eight consecutive weeks) as the homotypic stressor and inescapable footshock (0.8 mA, 0.5-s duration every 10 s for 15 min) as the heterotypic stressor. They reported that previously stressed rats had significantly greater plasma NE and EPI responses to the novel stress of footshock compared to first-time, footshock-stressed controls.

Finally, using a behavioral endpoint, Chung et al. (2000) examined the effects of repeated restraint stress (60 min per day for 10 consecutive days) on responses to a social stress test in laboratory rats. Compared to handle controls, previously restraint-stressed animals displayed decreased locomotion when placed into the home cage of an aggressive resident male, but with separation from the resident aggressive male by a partition. When the partition was removed, previously stressed intruder males displayed increased freezing in the presence of the resident male. These behavioral changes are consistent with increased fear and anxiety in previously stressed rats compared to handled control rats (Chung et al., 2000).

A limited number of studies has examined facilitation of plasma catecholamine responses to a heterotypic stressor. However, the results of these studies indicate that facilitation of plsma NE and EPI responses to a heterotypic stressor can occur independently or in concert. A range of heterotypic stressors has been employed, including processive and systemic stressors of varying intensities. Most surprisingly, facilitation of plasma NE and EPI responses was evident even when the heterotypic stressor was remarkably intense (e.g. NSD).

When laboratory rats are exposed to a highly intense stressor, the increments in plasma catecholamines grow with repeated exposure to that same stressor. Two clear examples of this sensitization process have been reported. As noted earlier, Kvetnansky et al. (1984) exposed laboratory rats to the NCD one time or once each day for eight consecutive days. Each exposure to the NCD was 400 s in duration. Plasma levels of NE and EPI were dramatically higher in rats exposed for the eighth time to NCD stress compared to controls exposed to NCD stress for the first time (Table 4). In a second example, Konarska et al. (1990a) compared the plasma catecholamine responses of laboratory rats to swim stress in water maintained at 18 °C or 24 °C for the first time or the 27th time. Compared to first-time stressed controls, rats exposed to swim stress for the 27th day at 18 °C or 24 °C had significantly enhanced plasma NE but not plasma EPI responses.

Table 4. Do plasma catecholamine responses to chronic intermittent stress satisfy the ten criteria that characterize habituation? Characteristics of habituation included below are based upon the list originally presented by Thompson and Spencer (1966), and later revised by Rankin et al. (2009). This format was adopted from Grissom and Bhatnagar (2009).

Criterion	Met: Yes/No	Abbreviated description of parameter	Representative publications
1	Yes	Repeated stress sessions result in decreased response	Konarska et al., 1989a, 1990b; Kvetnansky et al, 1984
2	No	Spontaneous recovery when stressor is withheld	No relevant studies
3	No	Potentiation of habituation	No relevant studies
4	Yes	The more rapid the frequency of stressor presentation, the more pronounced is habituation	De Boer et al., 1990
5	Yes	The weaker the stressor, the more pronounced is habituation	Konarska et al., 1990a for swim stress responses at different water temperatures
6	No	Beyond the zero or asymptotic response level	No relevant studies
7	No	Stimulus generalization in habituation	No relevant studies
8	Yes	Dishabituation upon exposure to a novel stressor	Kvetnansky et al., 1984; Konarska et al, 1989b
9	?Yes?	Habituation of a dishabituated response	No direct tests; however, dishabituating stimuli that have been shown to habituate include immobilization (Kvetnansky et al, 1984), restraint (Konarska et al., 1989a) and swim stress (Konarska et al, 1989a)
10	No	Long-term habituation	No relevant studies

Table 4 includes an analysis of nonassociative properties of plasma NE and EPI responses to chronic intermittent stress. Based upon the updated list of parameters of habituation described by Rankin et al. (2009), my analysis of the literature indicates that four of the 10 parameters have been satisfied (numbers 1, 4, 5 and 8), one parameter has been largely confirmed (number 9), and five parameters have not been satisfied or have not been tested adequately (numbers 2, 3, 6, 7 and 10).

This compilation is very similar to the results presented in Table 2 for the HPA axis. The one exception is that parameter 10 was satisfied for the HPA axis and has not been studied in the sympathetic-adrenal medullary system. Future experiments should tackle some of the remaining untested parameters, especially numbers 2 (spontaneous recovery), 3 (potentiation of habituation, and 7 (stimulus generalization). In addition, parameter 10 (long-term habituation) should be evaluated for the sympathetic-adrenal medullary system. If these experiments are conducted, the results would clearly indicate the extent to which nonassociative processes match with the regulation of HPA and sympathetic-adrenal medullary responses to stressful stimulation.

When considering the adaptive significance of plasma NE and EPI responses to chronic intermittent stress as described above, one must also factor in the stress-related changes in biosynthetic capacity of the adrenal medulla. We are fortunate that Kvetnansky and his collaborators over the years have conducted exquisitely detailed studies of catecholamine biosynthetic enzymes and catecholamine content in the adrenal glands of rats exposed to chronic intermittent immobilization stress. In a foundational study, Kvetnansky et al. (1970) reported that chronic intermittent immobilization stress of rats resulted in significant increases in the activities of the catecholamine biosynthetic enzymes, tyrosine hydroxylase (TH), dopamine-ß-hydroxylase (DBH) and PNMT, as well as levels of EPI in the adrenal medulla. Later studies revealed that mRNA levels for TH, DBH, and PNMT were also increased following chronic intermittent immobilization stress (Kvetnansky, 2004; Kvetnansky et al., 2002; McMahon et al., 1992). Sabban et al. (2004) extended these findings by reporting elevations in catecholamine biosynthetic enzymes in sympathetic ganglia. Finally, when chronically immobilized rats are exposed to novel stressors, levels of TH and PNMT in the adrenal medulla increase even further (Kvetnansky, 2004; Sabban & Serova, 2007; Sabban et al., 2006).

These studies clearly point to a dramatic upregulation of the biosynthetic and storage capacities of adrenal catecholamines during chronic immobilization stress. In contrast, chronically immobilized rats exhibit a significant decrease in circulating levels of NE and EPI during exposure to the last in a series of daily immobilizations (i.e. homotypic stressors). Why this paradox of increased tissue catecholamine biosynthetic capacity but decreased stress-induced catecholamine release?

To frame this paradox in evolutionary terms, animals exposed each day to a homotypic stressor have the benefit of living in a highly predictable environment. With each successive exposure to the homotypic stressor, animals have the capacity based upon experience to reduce responsiveness of physiological systems, including the sympathetic-adrenal medullary system, to the minimum extent necessary to maintain homeostatic balance. By conserving energy, especially glycogen stores in the liver, animals are enhancing their probability of surviving an unexpected future challenge. These hypotheses are in general agreement with the behavioral homeostasis theory advanced by Eisenstein et al (2001, 2006, 2012).

In spite of enjoying the advantage of living in a predictable if slightly stressful environment, these same animals hedge their bets by preparing for the unexpected novel stressor by enhancing the biosynthetic and storage capacities of the adrenal medulla and the sympathetic ganglia. When animals previously exposed to chronic intermittent stress are suddenly confronted with a novel (i.e. heterotypic) stressor, they activate the upregulated sympathetic-adrenal medullary system to a much greater degree than would be expected otherwise. We know this because the plasma NE and EPI responses to a novel stressor are significantly greater than the responses of a naïve animal exposed to the same stressor for the first time.

My colleagues and I have previously argued that this mismatch between plasma NE and EPI responses to homotypic and hterotypic stressors fits nicely within the theoretical framework of nonassociative learning (McCarty & Stone, 1984; McCarty et al., 1992; Stone & McCarty, 1983). These changes appear to result from centrally mediated alterations in sympathetic outflow as animals appraise the magnitude of the homeostatic challenge presented by a given stressor (Lazarus & Folkman, 1984). If the stressor is of low to moderate intensity and is familiar to and expected by the animal, the appraisal process leads to a dampening of sympathetic outflow. In contrast, if the stressor is novel, unexpected, or of high intensity, then the appraisal process activates sympathetic outflow to a much higher level.

Neural pathways regulating responses to chronic intermittent stress

In this section, I will not discuss in general terms the neural pathways regulating HPA and sympathetic-adrenal medullary responses to stressful stimulation. There are many excellent reviews of central stress pathways that are available (de Kloet et al., 2005; Herman, 2013; Herman & Cullinan, 1997; Pacak & Palkovits, 2001; Sawchenko et al., 1996, 2000). Rather, I will limit my consideration to studies that have addressed central mechanisms regulating habituation, facilitation or sensitization of HPA responses to chronic stress. Comparable studies of the sympathetic-adrenal medullary system have not been conducted and this is an appealing opportunity for future investigations.

HPA adaptation to stress: role of the posterior paraventricular thalamic nucleus

In a series of elegant experiments, Bhatnagar et al have investigated the involvement of the paraventricular nucleus of the thalamus (PVT) in the regulation of HPA responses to acute versus chronic intermittent stress in laboratory rats. A critical aspect of the success of these studies was the use of a consistent stress paradigm – chronic intermittent exposure over several days to a homotypic stressor, followed by acute exposure to a heterotypic stressor. Using this consistent stress paradigm, these investigators could probe brain areas and neurotransmitter systems involved in habituation and facilitation of HPA responses to stress.

In an initial set of studies, Bhatnagar & Dallman (1998) tackled the challenging issue of identifying a brain area(s) that regulates the facilitation of HPA responses to a heterotypic stressor (restraint) in laboratory rats previously exposed to a homotypic stressor (intermittent cold exposure at 4–6 °C for 4 h per day for 7 days). The experimental approach taken was to compare numbers of Fos-immunoreactive cells in 26 brain areas of control and previously cold-stressed rats at several time points following exposure to 30 min of restraint stress (i.e. the heterotypic stressor).

Thirty minutes after the termination of restraint stress, the density of Fos-stained cells was significantly greater in previously cold-stressed rats compared to restraint-stressed controls in six of the 26 brain areas examined. These brain areas included the parabrachial nucleus/Kölker-Fuse area; the posterior paraventricular thalamus (pPVT); the central (CNA), basolateral (BLA) and basomedial (BMA) nuclei of the amygdala; and the parvocellular paraventricular nucleus of the hypothalamus (PVN).

Based upon extensive supporting neuroanatomical evidence, the authors proposed that these 6 brain areas comprise a circuit that is critical to the facilitation response of the HPA axis, such that sensory information from the parabrachial nucleus → pPVT → amygdaloid nuclei → PVN, where a heightened ACTH response is marshaled. One hypothesis was that the pPVT could serve as a brain region that compares stored information regarding prior episodes of intermittent homotypic stress with novel incoming sensory information from the parabrachial nucleus during exposure to a heterotypic stressor, leading to the exaggerated HPA response to the heterotypic stressor (Bhatnagar & Dallman, 1998).

Several tests of this hypothesis have been conducted and summaries of six such tests are included below.

1. Employing the chronic intermittent cold acute restraint paradigm described above, Bhatnagar et al. (2000) reported findings consistent with release of cholecystokinin (CCK) into the pPVT to stimulate CCK-B receptors in response to a heterotypic stressor (restraint) in rats previously exposed to chronic intermittent cold stress. Locally released CCK dampens ACTH release in response to the heterotypic stressor in chronically stressed rats but not in rats without prior stress exposure. CCK-containing neurons that project to the pPVT arise from several areas, including the parabrachial nucleus, periaqueductal gray and dorsal raphe.

2. Bhatnagar et al. (2002) returned to the issue of habituation of HPA responses to chronic intermittent stress and the role of the pPVT, using ibotenic acid injections to lesion the pPVT prior to stressor exposure and restraint stress was employed instead of cold stress. Ibotenic acid is a highly specific neurotoxin that destroys neuronal perikarya but spares axons of passage and terminals in the vicinity of the injection site (Olney, 1986). In sham-lesioned rats, plasma levels of ACTH and corticosterone were significantly reduced in rats exposed to restraint stress for the eighth time compared to the first time. Ibotenic acid lesions of the pPVT completely blocked habituation of ACTH and corticosterone responses to chronic intermittent restraint stress but did not affect ACTH and corticosterone responses to the initial exposure to restraint stress.

3. The role of MR and GR receptors in the pPVT in the habituation of HPA responses to chronic intermittent restraint stress was examined in a comprehensive series of experiments by Jaferi & Bhatnagar (2006). Initially, the investigators described the distribution of GR receptors in the PVT (anterior, medial and posterior divisions) to supplement a previously published study of the distribution of MR receptors in the PVT (Ahima et al., 1991). Next, MR and/or GR antagonists (spironolactone and RU-38486, respectively) or vehicle were injected into the pPVT of rats exposed for the first time or the eighth time to restraint stress. The results indicated that vehicle-injected rats restrained for the eighth time had significantly lower plasma ACTH levels than vehicle-injected rats restrained for the first time. Neither GR nor MR antagonists affected integrated plasma ACTH responses to the first or eighth session of restraint. However, the combination of MR + GR antagonists into the pPVT resulted in a significantly higher integrated ACTH response to the first but not to the eighth session of restraint. If MR + GR antagonists were administered before the first through the seventh restraint stress sessions, then integrated plasma ACTH levels were significantly greater during the eighth restraint stress session compared to vehicle-treated rats restrained for the eighth time. In fact, the combination of antagonists normalized the integrated plasma ACTH level to that of first-time restrained rats.

   The chronic restraint-stress paradigm was repeated in adrenalectomized (ADX) rats provided with corticosterone replacement that fixed circulating corticosterone levels at 3.1–5.8 μg/dl for all groups. Rats received implants of corticosterone or cholesterol into the pPVT and were then assigned to either a single 30-min session of restraint stress or a 30-min restraint stress session each day for eight consecutive days. In rats with cholesterol implants in the pPVT, plasma levels of ACTH were significantly lower in repeatedly stressed rats compared to first-time stressed controls. Corticosterone implants into the pPVT significantly reduced plasma ACTH responses in chronically restrained rats compared to cholesterol-implanted controls. In contrast, corticosterone implants did not affect plasma ACTH responses in first-time, restraint-stressed controls (Jaferi & Bhatnagar, 2006). These findings built upon an earlier study that demonstrated that the pPVT is a site for negative feedback of circulating corticosterone in chronically stressed but not first-time stressed rats (Jaferi et al., 2003).

4. In their initial report, Bhatnagar & Dallman (1998) lesioned the pPVT using a Halász-type knife positioned stereotaxically. Plasma ACTH responses to restraint stress were significantly greater in pPVT-lesioned rats previously exposed to cold stress compared to sham-operated cold-stressed controls. In contrast, pPVT lesions did not affect plasma ACTH responses to acute restraint stress in unstressed controls. Plasma corticosterone responses to restraint stress were unaffected by pPVT lesions.

5. In a subsequent study, Bhatnagar & Dallman (1999) examined the influence of the pPVT on circadian rhythms in metabolic functions, including core body temperature and energy balance. Compared to unstressed controls, chronically cold-stressed laboratory rats (4–6 °C for 4 h per day for seven consecutive days) had reductions in the amplitude of core body temperature rhythms. Ibotenic acid lesions of the pPVT prior to stressor exposure blocked this effect, but lesions of the pPVT did not alter body temperature rhythms of unstressed control rats. Food intake (g/100 g body weight) was significantly greater in cold-stressed rats compared to controls over the course of the experiment. Lesions of the pPVT resulted in significant increases in food intake in unstressed control rats but did not affect food intake in cold-stressed rats. Finally, chronic cold stress resulted in significant decreases in subcutaneous white adipose tissue weight in sham-lesioned, but not in pPVT-lesioned rats. Ibotenic acid lesions of the pPVT were without effect on subcutaneous white adipose tissue weights of unstressed controls. Taken together, results from these experiments point to a role for the pPVT in inhibiting the amplitude of body temperature rhythms and weights of white adipose tissue in chronically stressed rats but not in unstressed controls. While the pPVT is not a primary brain area involved in temperature and energy balance, it does appear to play an important role in fine-tuning these systems through its extensive ascending connections to hypothalamic and limbic areas and descending connections to brainstem nuclei. However, this fine-tuning role is not evident under basal conditions and appears to be activated only under conditions of chronic intermittent exposure to a homotypic stressor.

6. Recently, Heydendael et al. (2011) examined the role of orexins, or hypocretins, in the pPVT on HPA adaptations to chronic intermittent stress. Orexins (ORXs) are neuropeptides that come in two forms, ORX-A and ORX-B and are produced from a common precursor, preproorexin. They serve as endogenous ligands for two G-protein-coupled receptors, ORX-1 and ORX-2 (for a recent review, Johnson et al., 2012). A striking feature of the neuroanatomy of the brain ORX system is that virtually all ORX-containing cell bodies are found in a subregion of the hypothalamus that contains the perifornical nucleus, the dorsomedial hypothalamic nucleus, and the dorsal and lateral hypothalamic areas. A few cells were also detected in the posterior hypothalamic area and the subincertal nucleus. The projections from these cell bodies are widely distributed in brain, with especially dense projections to other hypothalamic nuclei, the locus coeruleus, the nucleus of the solitary tract, the PVT, the BNST, the periaqueductal gray, the nucleus raphe dorsalis and the nucleus raphe magnus (Peyron et al., 1998). In a series of experiments, Heydendael et al. (2011) used a paradigm of swim stress (15 min per day at 25 °C) each day for four consecutive days (homotypic stressor) followed by acute exposure to 30 min of restraint stress (heterotypic stressor). Blockade of ORX receptors with localized injections of SB334867 into the pPVT before each of the four daily swim stress bouts blocked the enhanced plasma ACTH, hypothalamic c-Fos and PVN CRH mRNA responses to acute restraint (heterotypic stressor). SB334867 is an antagonist of ORX receptors and has a 50-fold greater selectivity for ORX-1 versus ORX-2 receptors (Smart et al., 2001). By the fourth swim stress bout, ORX-1 receptors were translocated from membrane to cytosolic fractions of the pPVT. Finally, pPVT cells of control rats and rats exposed to four bouts of swim stress were studied using a whole cell, patch-clamp recording technique. Results pointed to an increased responsiveness of pPVT cells to ORX-A in repeatedly swim-stressed rats compared to unstressed controls. These and other findings from this report provide strong evidence that ORXs released in the pPVT and acting on ORX-1 receptors play an important role in habituation and facilitation of HPA responses to chronic intermittent stress (Heyendael et al, 2011).

The results of these multiple lines of investigation point consistently to the pPVT as a brain area that is critical in regulating habituation and facilitation of HPA responses to stress. Given that three different homotypic stressors were employed in these experiments (e.g. cold exposure, restraint and swim stress), it appears that pPVT involvement is not linked to a specific stressor, but rather, that the pPVT monitors ongoing stressful stimulation of diverse types. When a heterotypic stressor is introduced to an animal previously exposed intermittently to a homotypic stressor, the pPVT notes the discrepancy between the expected stressor and the novel stressor and directs an enhanced HPA response. In virtually all neuroanatomical and neuropharmacological manipulations involving the pPVT, the HPA effects were specific to chronically stressed animals and were not observed in unstressed or first-time stressed controls. Taken together, these findings point to a highly specific role for the pPVT in habituation and facilitation of HPA responses (Figure 4).

Figure 4. Critical role of the basolateral amygdala (BLA) and the posterior paraventricular nucleus of the thalamus (pPVT) in regulating habituation (coded in green) and facilitation (coded in blue), respectively, of HPA responses to familiar and novel stressors. This diagram summarizes a series of experiments from the laboratory of Professor Seema Bhatnagar (see text and references for a summary of experiments and relevant citations) that positions the pPVT as a brain area where discriminations are made between homotypic stressors and heterotypic stressors, but only in animals that have been exposed previously to chronic intermittent stress. Afferent inputs to the pPVT include those originating from the parabrachial nucleus (PB), locus coeruleus (LC), dorsal raphe (DR) and the periaqueductal gray (PAG). Efferent projections from the pPVT include those to the medial prefrontal cortex (mPFC), ventral subiculum (vSUB), medial nucleus of the amygdala (MeA) and central nucleus of the amygdala (CeA), all of which project to the bed nucleus of the stria terminalis (BNST). Note that the pPVT does not project directly to the PVN. Also depicted is a projection from the basolateral nucleus of the amygdala (BLA) to the CeA that plays an important role in habituation of HPA responses to repeated daily exposure to a homotypic stressor.

The pPVT is well-positioned anatomically to play an important role in habituation and facilitation of HPA, behavioral, and possibly autonomic responses to chronic intermittent stress (Hsu et al., 2014). The neural inputs to the rat pPVT have been studied using tract-tracing techniques, and include prelimbic, infralimbic and insular cortical areas; the hippocampal subiculum; the dorsomedial nucleus, suprachiasmatic nucleus, arcuate nucleus and the lateral hypothalamic area of the hypothalamus; periaqueductal gray; parabrachial nucleus; entorhinal cortex; zona incerta; amygdala; and the BNST (Krout & Loewy, 2000; Li & Kirouac, 2012). Brain areas receiving projections from the pPVT include the infralimbic cortex, dorsal agranular insular cortex, entorhinal cortex, shell of the nucleus accumbens, central nucleus of the amygdala, basolateral nucleus of the amygdala, BNST and the suprachiasmatic nucleus (Li & Kirouac, 2008; Parsons et al., 2007; Shin et al., 2008). As summarized by Hsu et al. (2014), the pPVT, by virtue of its afferent and efferent projections, receives information related to stressful stimuli, circadian rhythms, and memory processes while influencing HPA responses to stress, mood, motivation, reward, and drug-seeking behavior.

In addition, the PVT plays an important role in learned fear paradigms involving a conditioned stimulus (CS) paired with an unconditioned stimulus (US). In one recent study (Furlong et al., 2016), fear extinction occurred following habituation to a US (loud sound), when the US only was presented after conditioning. Habituation to the US also reduced the level of fear elicited by a CS (tone) previously paired with the aversive US. Nonassociative mechanisms have been advanced to explain the decrease in learned fear following CS habituation (Rescorla, 1973). Thus, the PVT is involved in several stress-associated processes (neuroendocrine habituation and facilitation and fear extinction) as revealed by increases in c-fos expression in PVT neurons following chronic intermittent stress or CS-US pairings, respectively.

Another aspect of the brain circuitry involved in facilitation of HPA axis and possibly sympathetic-adrenal medullary responses to a heterotypic stressor is the importance of the BNST. As shown in Figure 4, stress-related information from the pPVT is directed to the prefrontal cortex, subiculum, and amygdaloid nuclei, and then conveyed to the BNST, which influences the HPA axis through efferent projections to the PVN. In addition, the BNST projects to brainstem nuclei involved in regulation of sympathetic outflow (Crestani et al., 2013). It is possible, therefore, that the pPVT and the BNST participate in the regulation of HPA axis and sympathetic-adrenal medullary responses to heterotypic stressors. Future experiments should investigate these possibilities and include simultaneous measures of ACTH, corticosterone, and plasma levels of NE and EPI in the same animals.

Finally, it is not surprising that several key brain areas involved in habituation and facilitation of neuroendocrine responses to stress also play critical roles in modulating memories for aversive events. These brain areas include the prefrontal cortex, amygdala and BNST (Gold & Korol, 2014; McGaugh, 2004, 2015; Reul, 2014; Roozendaal et al., 2009). Corticosterone, which readily crosses the blood–brain barrier, has access to these stress-sensitive brain circuits. In contrast, EPI, which does not readily penetrate the blood–brain barrier, has been demonstrated to modulate emotional memories by its effects on ascending vagal fibers, which in turn stimulate neurons in the ventrolateral medulla, locus coeruleus and nucleus of the solitary tract, leading to release of NE from nerve terminals in the amygdala (Chen & Williams, 2012). In addition, EPI stimulates release of glucose from glycogen stores in the liver and stress-induced elevations in circulating glucose have also been shown to modulate memory storage processes (Gold & Korol, 2014).

Habituation of the HPA axis to stress: role of the basolateral amygdala

Grissom & Bhatnagar (2011) hypothesized that the basolateral amgydala was involved in the habituation of the HPA axis to chronic intermittent stress through activation of ß-adrenergic receptors. This hypothesis was based in large measure on the pioneering studies of McGaugh and his colleagues on memories for emotionally arousing or stressful experiences (for reviews, see McGaugh, 2004, 2015; Roozendaal et al., 2009). In an initial test of this hypothesis, the ß-adrenergic antagonist, propranolol, was microinjected (0.3 μg/0.2 μl into the basolateral amygdala on each side of the brain) following each of four consecutive daily 30-min session of restraint stress. This treatment prevented the development of HPA axis (plasma levels of ACTH and corticosterone) and behavioral (amount of time spent struggling) habituation to restraint stress on day 5.

Microinjections of propranolol into the basolateral amygdala also had a significant impact on AVP but not CRH gene expression in the PVN of rats exposed to chronic restraint stress. Propranolol prevented the increase in AVP mRNA in the PVN that typically occurs following repeated restraint stress. The basolateral amygdala does not project directly to the PVN, so this effect of chronic stress on AVP mRNA in the PVN is probably mediated by other intermediate limbic nuclei. Next, the ß-adrenergic agonist, clenbuterol, was microinjected (1, 3 or 10 ng/0.2 μl into the basolateral amygdala on each side of the brain) following each of four consecutive daily 30-minute sessions of restraint stress. The effects of this drug followed an inverted-U dose-response curve, with enhancement of HPA axis habituation at the lowest dose but greatly attenuated behavioral habituation at the higher doses.

Based on this interesting series of experiments, Grissom & Bhatnagar (2011) concluded that habituation of the HPA axis to chronic restraint stress requires ß-adrenergic neurotransmission in the basolateral amygdala during the period immediately following each daily bout of stress (refer to Figure 4). In addition, poststress ß-adrenergic activity in the basolateral amygdala is critical for the increased expression of AVP mRNA in the PVN, an effect that appears to be associated with sustained HPA responses to chronic intermittent stress (Aguilera et al., 2008).

HPA adaptation to stress: role of the endocannabinoid system

The endocannabinoid (ECB) system plays an important role in regulation of the HPA axis and behavioral responses to stress (Akirav, 2013; Gorzalka et al., 2008; Haring et al., 2012; McLaughlin et al., 2014). ECBs, including anandamide and 2-arachidonoyl glycerol (2-AG), are neither stored in nor released from vesicles; rather, they are synthesized and released in response to alterations in neuronal activity. Not surprisingly, the ECB system is localized in brain areas known to be involved in the regulation of stress responses and emotional behavior, including cortical areas, hippocampus, striatum, substantia nigra, cerebellum, amygdala, and nucleus accumbens (Herkenham et al., 1990, 1991). Early studies suggested that type-1 cannabinoid (CB1) receptors were solely responsible for central effects of ECBs, but more recent studies indicate that some actions of ECBs in brain may be mediated by type-2 cannabinoid (CB2) receptors (Marco et al., 2011).

Several laboratories have examined the role of ECB signaling on the HPA axis by comparing wild-type mice with mice lacking CB1 receptors. For example, mice lacking CB1 receptors (using a CD-1 background) were reported to have elevated basal and novelty stress-induced levels of ACTH and corticosterone when compared to wild-type mice. However, in vitro release of ACTH from anterior pituitary fragments under basal conditions and following stimulation with CRH were similar between mice of the two groups. The investigators concluded that brain and not pituitary CB1 receptors influence HPA axis function under basal conditions and during acute exposure to stress (Barna et al., 2004).

Fride et al. (2005) evaluated HPA axis function in CB1-deficient mice (using a C57BL/6J background) by measuring plasma levels of ACTH and corticosterone under basal conditions and following noise stress (a bell ringing at 95–103 dBA). Plasma ACTH levels were similar in the two groups of mice. However, basal plasma corticosterone was elevated in CB1 receptor-deficient mice and the increment in response to stress was blunted compared to wild-type controls. It should be noted that blood samples were collected 6 min after a 4-min period of noise stress and this brief delay may not have provided sufficient time for the stress-related peak in corticosterone secretion. CB1-deficient mice displayed greater behavioral inhibition during exposure to stressful stimulation, suggesting greater responsiveness to the stressor.

Cota et al. (2007) reported that mice lacking CB1 receptors (using a predominant C57BL/6N background) displayed elevated levels of plasma corticosterone at the onset of the dark phase of the light-dark cycle compared to wild-type controls. In addition, CB1-deficient mice had elevated levels of CRH mRNA in the PVN but not in other brain areas where CB1 receptors and CRH mRNA are co-localized (e.g. amygdala and piriform cortex). Finally, mice lacking CB1 receptors exhibited downregulation of GR mRNA levels in the CA1 region of the hippocampus but not in the dentate gyrus or the PVN. MR mRNA levels were unchanged in wild-type versus CB1-deficient mice.

A key finding linking ECB activity with HPA axis fast-feedback responses to stress (Keller-Wood & Dallman, 1984) was described by Tasker et al in an elegant series of experiments (Di et al., 2003). Employing whole-cell, patch-clamp recordings in an in vitro rat hypothalamic slice preparation, they demonstrated a rapid suppression of excitatory glutamatergic input to parvocellular neurosecretory neurons of the PVN by corticosterone as well as dexamethasone. Further studies pointed to actions of the glucocorticoids at a postsynaptic membrane receptor and not an intracellular receptor, with subsequent release of a retrograde messenger. The actions of the retrograde messenger were blocked by two CB1 receptor antagonists, and its effects were mimicked by a cannabinoid receptor agonist, demonstrating actions of an endogenously released ECB (Di et al., 2003).

Building upon the findings of Tasker’s laboratory, Patel et al. (2004) investigated the effects of ECBs on restraint stress-induced activation of the HPA axis in laboratory mice. They noted that administration of a CB1 receptor antagonist/inverse agonist (SR141716) was accompanied by a modest dose-dependent increase in serum corticosterone. However, the highest dose of SR141716 did not increase neuronal activity in the PVN, as assessed by induction of Fos protein. Acute exposure of mice to 30 minutes of restraint stress also increased serum levels of corticosterone, but did not produce a consistent increase in Fos expression in the PVN. Pretreatment of mice with SR141716 prior to 30 min of restraint stress resulted in a significant enhancement of serum corticosterone levels and Fos expression in the PVN. Restraint stress-induced corticosterone release was decreased significantly in mice pretreated with a CB1 receptor agonist, an ECB transport inhibitor, or a fatty acid amide hydrolase inhibitor. Hypothalamic levels of 2-AG were reduced in mice following acute restraint stress compared to controls, but were increased significantly in mice following the fifth daily bout of restraint stress. Chronic intermittent restraint stress also resulted in reduced levels of serum corticosterone, evidence of a habituated response (Patel et al., 2004). In a subsequent series of experiments, it was reported that enhanced stimulation of CB1 receptors resulted in habituated neural and behavioral responses to chronic intermittent restraint stress (Patel et al., 2005).

The central sites of action of ECBs in regulating HPA responses of laboratory rats to acute versus chronic intermittent stress were explored in experiments by Hill et al (2009, 2010). These investigators found that 30 min of restraint stress each day for nine consecutive days resulted in an elevation in basal corticosterone levels, but a decrease in the corticosterone response to the final episode of restraint compared to first-time stressed controls. Further experiments addressed the central sites of action of ECBs in regulating the two distinct HPA responses to chronic intermittent exposure to a homotypic stressor – increased basal levels of corticosterone but decreased corticosterone responses to the ninth daily bout of restraint stress. Anandamide levels within the prefrontal cortex were strongly negatively correlated with elevated basal levels of corticosterone in chronically stressed rats. In contrast, there was a significant negative correlation between 2-AG content in the amygdala and the corticosterone response to the ninth bout of restraint stress.

These and other findings demonstrated a clear dissociation between the effects and sites of action of anandamide and 2-AG on HPA responses to acute versus chronic intermittent stress. In rats exposed acutely to restraint stress, anandamide levels in the amygdala were inversely correlated with serum corticosterone levels, suggesting that acute exposure to stress reduces anadamide signaling in the amygdala and this reduction, in turn, enhances HPA activation (Hill et al., 2009). Reductions in anandamide levels in the prefrontal cortex following chronic intermittent stress contributed to the elevated basal levels of corticosterone, while increased 2-AG signaling within the amygdala following chronic intermittent stress contributed to the habituation of corticosterone responses (Hill et al., 2010).

Neural adaptations to stress: role of transcriptional changes

Studies of gene transcriptional changes in brain in response to acute versus chronic intermittent stress have progressed from studies of single genes to approaches involving gene microarrays and RNA sequencing (Rubin et al., 2014). Much of what has been learned about stress-sensitive brain circuits over the past two decades has depended upon studies of the immediate early gene, c-fos, and the nuclear phosphoprotein it encodes, Fos.

Fos or c-fos mRNA serve as ideal indices of neural activity in that they are found in very low levels under basal conditions and increase rapidly and transiently in response to synaptic stimulation (Kovács, 1998; Sagar et al., 1988). Importantly, c-fos mRNA and Fos protein also increase in specific neuronal populations in response to a variety of stressful stimuli and have been employed to map stress circuits in brain (Ceccatelli et al, 1989; Cullinan et al, 1995; Kovács & Sawchenko, 1996; Pacak & Palkovits, 2001).

It is important to note that other immediate early genes and transcription factors may also respond to acute versus chronic stress and influence neural, endocrine, and behavioral responses. Examples include zif-268, Erg-1, arc, c-jun, jun-B, jun-D, ΔFosB and ERK1/2 (Flak et al., 2012; Gutièrrez-Mecinas et al., 2011; Melia et al., 1994; Ons et al., 2010; Reul, 2014; Schreiber et al., 1991), and some of these will be discussed below.

Included as a small part of a larger set of experiments, Sasse et al. (2013) exposed laboratory rats to noise stress (98 dB for 30 min) for 1 day or 11 consecutive days. Unstressed control rats were exposed to ambient sound created by exhaust fans (60 dB) in the same noise chambers for the same amount of time. Acute exposure to noise stress was followed by significant increases in c-fos mRNA in the infra-limbic cortex, prelimbic cortex, ventrolateral BNST, ventrolateral septum and PVN above levels in unstressed controls. In rats exposed to noise stress for 11 days, levels of c-fos mRNA were reduced significantly compared to first-time stressed controls in each of the brain areas noted above, indicating habituation to repeated noise stress.

To examine habituation and facilitation of immediate early gene expression, Melia et al. (1994) exposed laboratory rats to one or nine daily sessions of restraint stress (2 h per day). A single exposure to restraint stress increased c-fos mRNA in several brain regions previously associated with regulation of the HPA axis (cortex, hippocampus, hypothalamus, septum and brainstem). In contrast, c-fos mRNA levels were dramatically reduced in these same brain regions in rats exposed to restraint stress each day for nine consecutive days. Similar results were reported for the immediate early gene, jun-B, but not for zif-268, c-jun or jun-D mRNA expression. If rats exposed repeatedly to restraint stress were then exposed to a heterotypic stressor (swim stress in 23 °C water for 20 min), levels of c-fos mRNA in all brain regions studied were comparable to levels seen in first-time swim-stressed controls (Melia et al., 1994).

In a similar study, Watanabe et al. (1994) exposed rats to acute versus repeated restraint stress (1 hour per day for 1 or 14 consecutive days) and quantified levels of mRNA for the immediate early genes c-fos and zif-268 in several brain regions. Compared to unstressed controls, acutely stressed rats had significant increases in c-fos mRNA in the PVN, locus coeruleus, midbrain raphe, and central gray. No significant changes were detected in midbrain A9 and A10 dopaminergic neurons. Acute restraint stress resulted in significant increases in zif-268 mRNA only in the PVN.

Habituation of the c-fos mRNA response to chronic intermittent restraint stress was evident in the PVN, locus coeruleus, midbrain raphe and central gray. For zif-268, there was habituation to chronic intermittent restraint in the PVN only. If rats exposed to chronic intermittent restraint stress were then exposed to a heterotypic stressor (shaking stress at 100 cycles per minutes for 30 min), levels of mRNA for c-fos and zif-268 were generally similar across brain regions to levels observed in control rats exposed to shaking stress for the first time. Thus, there was no evidence of facilitation of the immediate early gene response to a heterotypic stressor in either of these studies (Malia et al., 1994; Watanabe et al., 1994). In contrast, Weinberg et al. (2009) focused on the PVN and prefrontal cortex and reported habituation of the c-fos response to chronic daily exposure of laboratory rats to ferret odor and facilitation of the c-fos response to a heterotypic stressor (restraint) following chronic daily exposure to ferret odor.

Recently, several laboratories have utilized gene chip technologies to study transcriptional activity of several hundred to more than 25,000 genes simultaneously in brain tissue samples from laboratory animals exposed to various chronic stress paradigms (e.g. Andrus et al., 2012; Datson et al., 2012, 2013; Feldker et al., 2006; Kim & Han, 2006; Liu et al., 2010; Polman et al., 2012). Such an approach represents a powerful tool to explore adaptive transcriptional regulation in brain areas following chronic intermittent stress (Rubin et al., 2014). I will limit my summary of relevant experimental findings on stress-induced transcriptional changes to two superb studies that were designed within the framework of habituation, facilitation, or sensitization processes using chronic, intermittent stress paradigms.

The extensive experimental results connecting the pPVT to habituation and facilitation of HPA responses (described above) presented an ideal opportunity to explore how transcriptional changes might be crucial in this brain nucleus. In an important study by Heydendael et al. (2012), the genes that mediate the effects of ORXs in the pPVT of swim-stressed (15 min at 25 °C for four consecutive days) and unstressed control rats were analyzed using a custom PCR array of 186 mRNAs. The ORX-1 antagonist, SB334867 or vehicle was infused directly into the pPVT prior to each of the four swim stress sessions and brain tissue was sampled 24 h later. Surprisingly, the expression levels of only nine genes were altered significantly by chronic intermittent swim stress. Of the 47 growth factor related-genes, vascular endothelial growth factor A (Vegfa) and Bcl-2-associated × protein (Bax) were increased following swim stress, and the increases were blocked by the ORX-1 antagonist. In contrast, metallothionein-3 (Mt3) was increased following ORX-1 blockade in non-stressed rats but decreased by ORX-1 blockade in swim-stressed rats.

Of 46 G-protein coupled receptor-related mRNAs studied, only three were altered significantly. SB334867 micro-injected into the pPVT resulted in significant decreases in expression of adenosine A_2A receptor mRNA in control and swim-stress rats. In unstressed control rats, there was an increased expression of group II metabotropic glutamate receptor mRNA (Grm2) following administration of SB334867 and this effect was not observed in swim-stressed rats. Finally, for CRH receptor1 (Crhr1) mRNA, there was an increase in expression in vehicle-treated, swim stressed rats compared to vehicle-treated control rats. However, expression of Crhr1 mRNA in SB334867-treated rats exposed to repeated swim stress was not different from levels in vehicle-treated rats exposed to repeated swim stress.

Only two of 20 immune-related mRNAs were altered significantly in this study. For each of these two mRNAs, there were significant but opposite effects of SB334768 microinfusion into the pPVT but no effects of stress conditions. In the first, prostaglandin endoperoxide synthase 2 (Ptgs2) expression levels were decreased significantly in rats receiving SB334768 while CXC3 chemokine receptor 1 (Cx3cr1) expression levels were increased significantly in rats receiving SB334768.

Expression of 31 epigenetic-related mRNAs was quantified and only histone deacetylase 5 (Hdac5) was altered significantly. In unstressed rats that received micro-injections of SB334768 into the pPVT, Hdac5 expression was increased. SB334768 was without effect on Hdac5 in swim-stressed rats. There were no significant alterations observed in 26 intracellular signaling-related genes, 11 common ion channel-related genes, or 6 neuropeptide-related genes.

A careful review of the patterns of change of the nine genes described above reveals that only 3 of the genes (Vegfa, Bax and Crhr1) had expression profiles that matched with the facilitated HPA response to a heterotypic stressor that was blocked by daily treatment with SB334867 prior to swim stress sessions, as described in an earlier study (Heydendael et al., 2011). These three genes, and possibly others not included in this study, appear to be crucial in mediating the effects of ORX released within the pPVT on facilitation of HPA responses to a heterotypic stressor in rats previously exposed to a homotypic stressor (Heydendael et al., 2012).

In an ambitious recent study, Gray et al. (2014) examined the effects of a homotypic stressor, restraint, and a heterotypic stressor, swim stress, on gene expression patterns in the whole hippocampus of C57CL/6J mice. Restraint stress was for 2 h per day for 21 consecutive days. The forced swim stress was in room temperature water (approximately 22 °C) for 6 min. The experimental groups included (a) unstressed controls, (b) chronic intermittent restraint for 21 days, (c) chronic intermittent restraint for 21 days + swim stress, (d) chronic intermittent restraint for 21 days followed by no treatment for 21 days, (e) chronic intermittent restraint for 21 days followed by no treatment for 20 days followed by swim stress on day 42, (f) acute swim stress. RNA was extracted from hippocampal samples and subjected to microarray analysis of approximately 19,000 probes.

The results indicated that hippocampal gene transcription patterns differed dramatically between the homotypic stressor, the heterotypic stressor, and the recovery from homotypic stress conditions. After 21 consecutive days of restraint stress, expression levels of 773 genes (42% increased, 58% decreased) were altered significantly compared to controls. In contrast, transcription levels of 3,999 genes (28% increased, 72% decreased) were altered significantly in chronically stressed mice exposed to swim stress compared to unstressed controls. Only 77 genes were altered significantly by both swim stress and chronic restraint, indicating distinct patterns of transcriptional responses to these two distinctly different stressors. After a 21-day recovery from chronic restraint stress, 689 genes (53% increased, 47% decreased) were significantly different from unstressed controls. Restraint-stress recovered mice exposed to acute swim stress had 1251 genes (54% increased, 46% decreased) whose transcription levels differed from unstressed controls.

Two interesting bundles could be created from this expansive data set. In the first bundle, the transcription levels of 95 genes were altered significantly in animals following exposure to the heterotypic stressor, regardless of prior stress history (i.e. acute swim stress, chronic restraint + acute swim stress, chronic restraint + recovery + acute swim stress). Many of these genes have been identified in previous studies of acute stress and some are associated with increased neural activity (i.e c-fos). The second bundle comprises 327 genes that overlap between the chronic restraint + acute swim stress and the chronic restraint + recovery + acute swim stress groups, and many of these genes have been associated with chronic stress conditions (i.e. brain-derived neurotropic factor, BDNF).

Among the points emphasized by the authors in their concluding comments were:

• Three weeks after the end of 21 days of chronic intermittent restraint stress, gene transcription patterns remained distinctly different from unstressed controls. Thus, the transcriptional effects of chronic intermittent restraint stress persist for at least 3 weeks and may alter the response patterns of these mice to other stressors.

• A greater number of genes was altered by the chronic restraint + acute swim stress condition compared to acute swim stress alone or chronic restraint + recovery + acute swim stress. This heightened level of transcriptional reactivity may be analogous to the enhanced response of the HPA axis in facilitation.

• These results provide a valuable baseline of transcriptional changes in response to chronic stress, recovery from chronic stress and exposure to a heterotypic stressor following chronic stress. Future studies may focus on transcriptional responses in animal models characterized by increased susceptibility to stress-related behavioral changes.

• The homotypic stressor + heterotypic stressor paradigm may allow for the identification of new targets for drug development for such conditions as depression and anxiety (Gray et al., 2014).

A complementary level of analysis involves studies of epigenetic mechanisms associated with chromatin remodeling, especially in the hippocampus (Carter et al., 2015; Mifsud et al., 2011). The hippocampus has traditionally been a brain area of intense focus because it contains high levels of glucocorticoid receptors, making it susceptible to stressors, and it contains many of the enzymes responsible for chromatin remodeling (McEwen, 2007; Trotter & Archer, 2007).

Chromatin is made up of nucleosomal units, a complex of 8 histone octomers, containing two copies each of H2A, H2B, H3 and H4, around which the DNA is super-coiled. Given the highly condensed structure of chromatin, transcriptional activators can only gain access to individual genes through posttranslational changes in histones and in DNA. Histones contain N-terminal tails (25–30% of the mass of the individual histone) that extend from the core of the nucleosome, and the tails can be reversibly modified by enzymes at a number of the amino acids by acetylation, methylation, ubiquitinylation, and phosphorylation, creating a histone code. For example, histone acetylation of H3 results in transcriptional activation by canceling the positive charges of lysine residues 9, 14, 18 and 23, with an increase in the spacing between nucleosomes. This process is stimulated by histone acetyltransferase and reversed by histone deacetylase (for a review, see Strahl & Allis, 2000).

There is extensive documentation that acute and chronic stressors can stimulate posttranslational modification of histones. For example, exposure to acute psychological stressors (e.g. forced swimming, exposure to a predator) but not to physical stressors (ether vapors, cold environment) resulted in histone H3 phosphorylation in the dentate gyrus (Bilang-Bleuel et al., 2005). In addition, exposure of laboratory rats to acute novelty stress (exposure to a clean cage) was associated with enhanced phosphorylation and phospho-acetylation of histone H3 in the dentate gyrus, followed by induction of c-fos (Chandramohan et al., 2007).

Finally, Hunter et al. (2009) explored the effects of acute versus chronic restraint stress on histone H3 methylation at lysines 4, 9 and 27 in areas of rat hippocampus. Their results point to rapid, regionally specific changes in chromatin remodeling in the hippocampus of rats exposed to acute versus chronic restraint stress. In particular, for some of these changes, acute restraint stress resulted in much larger changes in histone trimethylation at lysines 9 and 27 compared to chronic intermittent restraint stress. These differences in chromatin remodeling in rats exposed to acute versus chronic intermittent restraint may be related in part to habituation of the HPA axis to chronic intermittent restraint stress. In an extension of their work, Hunter et al. (2015) considered the ways in which exposure to stress could produce persistent alterations in neural and endocrine responses and behavior. They suggested that expression of retrotransposons may play a critical role in stabilizing genomic activity in stress-sensitive brain regions such as the hippocampus

The studies by Bhatnagar & Dallman (1998) and Grissom & Bhatnagar (2011) marked a critical point of departure for the field of stress research. By exploiting the neural and endocrine differences between animals exposed chronically to a homotypic stressor versus animals exposed chronically to a homotypic stressor and then exposed acutely to a heterotypic stressor, they were able to focus in on brain regions that are critical for habituation (basolateral amygdala) versus facilitation (pPVT) of HPA responses. As reviewed above, other experiments have delineated some of the neural pathways and neurotransmitter systems that participate in these non-associative influences on regulation of the HPA axis. Other investigators have employed similar experimental paradigms to reveal transcriptional changes that are associated with habituation and facilitation of HPA responses to stress. These advances in the neurobiology of stress have the potential to reveal new targets for the development of drug therapies that could interrupt stress-sensitive processes that contribute to the development of pathological changes.

Chronic intermittent stress and psychopathology: clinical implications

Stressful stimulation has been viewed by many investigators as an etiologic factor in the onset of symptoms associated with mental illnesses and the development of dependance on drugs of abuse (Bangasser & Valentino, 2014; Bisagno & Cadet, 2014; Checkley, 1996; de Kloet et al., 2005; Gold et al., 2015; Grippo & Johnson, 2009; Solomon & Herman, 2009; Van Bockstaele et al., 2010). Results of experiments with animal models point to disruptions of physiological responses to acute versus chronic intermittent stressors that appear to play a role in disease processes, including those associated with depression and post-traumatic stress disorder (PTSD). To connect the previous discussion of neural and endocrine responses to homotypic and heterotypic stressors with psychopathology, I will focus on two mental illnesses, depression and post-traumatic stress disorder, where the connections between stress and the disorders have received considerable attention.

Depression

As noted below, an inability to habituate to repeated episodes of stressful stimulation may place susceptible humans at greater risk of developing depression (Kendler et al., 1999). A particular focus of studies of stress and depression has been on the brain circuitry that regulates the HPA axis, the autonomic nervous system and appropriate behavioral responses. One possible link between the HPA axis, the autonomic nervous system and symptoms of depression is a lack of habituation of CRH-containing neurons to repeated episodes of stressful stimulation (Gold & Chrousos, 2002). CRH-containing neurons in the PVN regulate the release of ACTH from the anterior pituitary and stimulate autonomic responses to stressful stimulation via projections to the brainstem. In addition, CRH released within limbic areas serves as a modulator of behavioral responses to fearful or threatening situations (Gold & Chrousos, 2002).

Elevated circulating levels of cortisol have been reported frequently for patients with major depression. However, Gold et al. (1984) first reported that depressed patients with elevated cortisol levels exhibited blunted plasma ACTH responses to ovine CRH, but had a significant cortisol response. The higher the circulating cortisol level, the lower the plasma ACTH response. Apparently, in these depressed patients, the adrenal cortex had experienced hypertrophy and was now more sensitive to the stimulatory effects of ACTH. In addition, CRH-containing neurons in the PVN were less sensitive to the negative feedback effects of elevated circulating cortisol. Finally, projections from PVN CRH neurons to the locus coeruleus stimulate NE-containing neurons, which project to the hypothalamus and limbic areas. Thus, brain CRH and NE neurons cooperate to upregulate the HPA axis and this cooperation may be reflected in enhanced physiological and behavioral responses to future stressors (Gold & Chrousos, 2002).

In a compelling review of studies linking stress and depression, Gold (2015) noted that evolution has favored the development of neural, endocrine, and behavioral systems that are activated during the onset of a controllable stressor. These systems tend to remain in an activated state for the minimum amount of time necessary to reestablish homeostatic balance. Unfortunately, these same feedback systems that regulate stress circuits in the brain may go awry when susceptible individuals are exposed to uncontrollable stressors. Therefore, central stress pathways may represent the weak links in the onset of depression and explain the connections between stressful life experiences and the development of major depressive disorder (Gold et al., 2015).

Animal models afford opportunities for delineating the mechanisms underlying the connection between stress and depression. In particular, the hippocampus is a stress-sensitive brain area that participates in the regulation of the HPA axis. For example, exposure of laboratory rats daily to a chronic unpredictable stress paradigm for 1 month resulted in a significant reduction in total dendritic length and a reduction in density of distal dendritic branching in hippocampal CA3 neurons. In addition, the unpredictable stress paradigm also led to dendritic regression in both granule cells and CA1 pyramidal cells. These dramatic structural changes could affect hippocampal inhibition of the HPA axis during repeated exposure to a homotypic stressor (Sousa et al., 2000).

In an intriguing study, Wood et al. (2012) exposed laboratory rats to the stress of resident-intruder aggression for 30 min each day for seven consecutive days to study the effects of stress on the appearance of behaviors and physiological changes that are similar to co-occurring depression and cardiovascular disease in humans. In addition, some groups of stressed rats received either a CRH-R1 receptor antagonist (NBI 30775), desipramine (DMI), or an appropriate vehicle injection prior to resident-intruder aggression tests. The CRH-R1 receptor antagonist but not DMI increased in intruder rats the time spent in an upright, defensive posture and the latency to assume a fully submissive posture toward the resident. The CRH-R1 antagonist also reduced ACTH and corticosterone responses to social defeat and reduced adrenal hypertrophy and decreases in heart rate variability (i.e. increased sympathetic drive to the heart). Both drugs decreased immobility in a forced swim test. These findings suggest that central CRH systems play an important role in regulating HPA axis and sympathetic nervous system responses and behaviors associated with the stress of resident-intruder testing.

Post-traumatic stress disorder (PTSD)

It was not until 1980 that post-traumatic stress disorder (PTSD) was included as a diagnostic category in the Third Edition of the Diagnostic and Statistical Manual of Mental Disorders of the American Psychiatric Association (1980). Although the impetus for recognition of PTSD as a mental disorder reflected intense lobbying from veteran’s groups, it goes without saying that PTSD may also develop in individuals across a wide age range and from traumatic experiences unrelated to combat (Beck & Sloan, 2012).

Contrary to other mental disorders, the causative agent for PTSD is generally accepted to be acute or chronic intermittent exposure to a psychologically traumatic event(s). But not all individuals exposed to traumatic events develop symptoms of PTSD, and those who do vary from 1% to 50%, depending upon the intensity of the traumatic event and the biological vulnerability of the victims, as well as the systems of social support available to them (Barlow & Durand, 2015). I will limit my discussion of PTSD to select areas where nonassociative principles relating to stress may come into play in developing animal models, in formulating theories regarding the etiology of PTSD, or in developing therapeutic interventions.

Currently, there is no widely accepted animal model of PTSD. Several experimental approaches have been taken to develop such model systems, each with an eye toward psychological and biological features of PTSD as gleaned from studies of patient populations. To guide these efforts, Yehuda & Antelman (1993) advanced five criteria by which animal models should be evaluated to determine their validity for informing research on patients with PTSD. They include the following:

1. Exposure to very brief stressors should be capable of inducing biological and behavioral changes characteristic of PTSD.

2. The stressor should be capable of producing changes characteristic of PTSD in a dose-dependent manner.

3. The stressor should produce biological changes that persist over time or become more pronounced over time.

4. The stressor should produce biological and behavioral changes, some of which may represent increases over prestress levels, while others may represent decreases over pre-stress levels.

5. Variability between individuals in responses to a stressor should be present based upon differences in prior experience, genetic differences or an interaction of the two.

These five criteria continue to guide investigators who work on animal models of PTSD. Symptoms of PTSD in animal models and in patients may reflect a sensitization of stress-related brain circuitry following exposure to the initial trauma. Conversely, sympthoms of PTSD may reflect a failure to habituate to experimental conditions that later serve as a reminder of the initial exposure to trauma.

An early series of studies from Natelson’s laboratory focused on developing an animal model of PTSD. Their experimental approach involved exposing laboratory rats to a 2-h session of 40 inescapable tailshocks (2.0 mA) for 1 day or for three consecutive days. They found that acutely (1X) and chronically (3X) stressed rats exhibited enhanced startle responses for 7–10 days after termination of the stressor. In addition, 3X rats had enhanced levels of HPA axis activity and reduced levels of open field activity. When exposed to a single tail shock ten days after the final stress session, 3X rats exhibited higher levels of corticosterone that were slower to return toward basal levels and elevated basal corticosterone levels the following day (Ottenweller et al., 1992, 1994; Servatius et al., 1994, 1995). This model approximated some but not all of the key characteristics of PTSD described by Yehuda & Antelman (1993).

More recently, other experimental approaches have been taken to develop animal models that more closely approximate key features of PTSD in humans. One approach that has been employed widely is the single prolonged stress (SPS) paradigm, whereby laboratory rats are exposed to a series of stressors in a single day (e.g. restraint stress + swim stress + ether anesthesia) and then left undisturbed for at least one week (sensitization period). This experimental design is reminiscent of time dependent sensitization (Antelman et al., 2000) and produces physiological and behavioral changes that match well with symptoms of PTSD in humans (Whitaker et al., 2014; Yamamoto et al., 2009).

Rats exposed to SPS exhibit many features consistent with PTSD, including an enhanced startle response, decreased extinction retention, enhanced expression of GR mRNA in the hippocampus up to 2 weeks following SPS, and evidence of greater negative feedback regulation of the HPA axis to restrain basal plasma corticosterone levels up to 2 weeks post-SPS (Kohda et al., 2007; Liberzon et al., 1999). The model fails to replicate two of the five key features of PTSD listed above, including production of PTSD symptoms in an intensity-dependent manner (i.e. only one intensity level has been employed to date for the SPS paradigm) and inter-individual variability in production of PTSD-like symptoms (Whitaker et al., 2014).

Another promising model of PTSD involves exposure of laboratory animals, usually laboratory rats, to a predator or its odor. Predators that have been employed include cats and ferrets, and predator odors have included those of cats, ferrets and foxes. Typically, laboratory rats are exposed to a predator or its odor in a test chamber from which there is no escape. As an example, Adamec et al. (1993) placed rats individually into a small room with a cat for 5 min. The rats were followed for up to 21 days after predator exposure, and behavioral measures were taken of each animal in an elevated plus maze. The results indicated that the occurrence of anxiety-like behaviors in the plus maze was strongly correlated with the proximity of the cat to the rat during initial testing. Thus, there was interindividual variability in the behavioral responses to the threatening experience, and the behaviors persisted for an extended period of time. Both of these features are important for the development of an animal model of PTSD.

In an important methodological advance that recognized the heterogeneity of responses to predator stress, Cohen et al. (2003) exposed individual rats to a cat or its odor for a 10-minute period. Seven days later, predator-exposed rats were tested in an elevated plus maze and their behaviors were compared to handled control rats. In analyzing the data from their study, the investigators confirmed that predator exposure increased anxiety-like behaviors in laboratory rats. They then employed cutoff behavioral scores in the elevated plus maze to distinguish between “mal-adapted” and “well-adapted” rats. Approximately one-quarter of the predator-exposed rats fell into the mal-adapted group and an equal proportion fell into the well-adapted group. Mal-adapted rats exhibited elevations in basal plasma levels of ACTH and corticosterone, increased sympathetic activity, and reductions in vagal tone compared to well-adapted and control rats.

Another methodological twist that has been included in these studies of predator stress is a situational reminder condition, where predator odor-exposed rats are presented with clean cat litter as a reminder of the earlier exposure to cat odor 7–10 days before. The situational reminder results in anxiety-like behaviors that are similar to those observed during the original exposure to cat odor (e.g. Cohen et al., 2006). The efforts of Cohen et al represent an important advance in attempts to develop a viable and widely accepted animal model of PTSD that satisfies the five criteria established by Yahuda & Antelman (1993).

Major refinements have occurred in the development of animal models of PTSD, in moving from repetitive exposure to high-intensity stressors on one or more occasions, (e.g. inescapable tail shocks), to use of an ethologically relevant stressor such as exposure of laboratory rats to a predator or its odor. In particular, studies using predator stress as an animal model of PTSD have identified new approaches to the development of next-generation therapies, including use of protein synthesis inhibitors to prevent behavioral changes consistent with PTSD and the involvement of proinflammatory responses in hippocampus, frontal cortex and hypothalamus in the development of PTSD symptoms (Cohen et al., 2006; Levkovitz et al., 2015). These new approaches hold the promise of advances in the treatment of this debilitating condition.

Summary and conclusions

Over the past two decades, major advances have been made in the study of how an organism adapts to acute versus chronic intermittent exposure to various stressors. For many years, experiments merely achieved incremental gains by demonstrating that a given stressor of intensity X and duration Y produced a measurable change in a particular neural, endocrine, or behavioral parameter. With refinements in experimental design that permitted studies of habituation and sensitization of neuroendocrine pathways, it was possible to explore the role of specific brain nuclei in adaptive processes during chronic intermittent exposure to stress. These studies also extended into the realm of genomic adaptations, which hold great promise for the future.

Finally, these same stress pathways in brain are now recognized as critical players in the onset and maintenance of mental disorders. Experiments with animal models of depression and PTSD have unmasked novel targets for the development of therapeutic interventions to treat these devastating conditions. These basic experimental findings have informed clinical studies in patients with mental disorders and provide insights into neural circuits and neuroendocrine response patterns that link stressful stimulation with adaptive and mal-adaptive behavioral changes. Ongoing research efforts will hopefully lead to better pharmacological and behavioral treatment options for patients.

Acknowledgements

I thank Vanderbilt University for a sabbatical leave for the 2014–2015 academic year that made possible the preparation of this manuscript. I also thank Professors John Gore and Sandy Rosenthal for support and encouragement and Ms. Megan Rojas for preparation of figures. Three anonymous reviewers provided many helpful suggestions for improving the manuscript and I am grateful for their input.

Disclosure statement

The author reports no conflicts of interest. The author alone is responsible for the content and writing of this article.