Evaluating stress in natural populations of vertebrates: total CORT is not good enough


  • Creagh W. Breuner,

    Corresponding author
    • Wildlife Biology and Organismal Biology and Ecology, University of Montana, Missoula, Montana, USA
    Search for more papers by this author
    • All authors have contributed equally to the preparation of this document.
  • Brendan Delehanty,

    1. Department of Biological Sciences, Centre for the Neurobiology of Stress, University of Toronto Scarborough, Toronto, Ontario, Canada
    Search for more papers by this author
    • All authors have contributed equally to the preparation of this document.
  • Rudy Boonstra

    1. Department of Biological Sciences, Centre for the Neurobiology of Stress, University of Toronto Scarborough, Toronto, Ontario, Canada
    Search for more papers by this author
    • All authors have contributed equally to the preparation of this document.

Correspondence author. E-mail: creagh.breuner@umontana.edu


  1. Our goal in this review is to discuss how measures beyond simple quantification of total glucocorticoid levels are needed in comparative studies of stress. We need to measure corticosteroid binding globulin – CBG – and further downstream performance metrics to properly evaluate the significance and impact of stress in wild populations. We briefly cover the current literature, discuss methods that may enable detection of chronic stress and point to directions for future research to continue to clarify this field.
  2. CBG may regulate access of hormones to tissues, but disagreement remains as to the functional importance of total vs. free vs. bound hormone in the plasma. Here we focus on recent studies providing solid evidence supporting the free hormone hypothesis. These studies unequivocally indicate that the biologically active fraction consists only of the portion that is free (not total levels in the blood). We also present the ‘reservoir hormone hypothesis’, indicating the relevance of CBG-bound hormone in the plasma.
  3. The vast majority of physiological studies on stress in natural populations only measure blood corticosterone or its faecal metabolites. However, downstream metrics (e.g. immune function, oxidative stress and body mass changes) can assess the physiological impact of the changes in corticosterone and CBG and ultimately of the adaptiveness of these changes. Here we will discuss some of the most promising factors that can be measured downstream of the plasma and the rationale for their measurement, how to do it, and an introduction to the evidence.
  4. Although we draw on biomedical findings for some of this insight, we recognized that findings from a few mammalian laboratory species (primarily rats and mice) may not apply to the vast diversity of vertebrate species from fish to birds. In large measure we are still in the natural history phase of the accumulation of fundamental knowledge in terms of trying to extract general principles in the stress biology of natural populations. Our review outlines what is needed to lay the foundation for these principles.


The field of ecological stress physiology and its role in natural populations is now well over 60 years old. One of the first to link stress to population processes and health was John Christian (1950), and his ideas originated directly from the seminal work of Hans Selye (1936, 1937). However, it was difficult to translate the initial ideas on the role of stress into insightful natural processes because the tools were not available to quantify stress responses. The key innovation that moved the field along rapidly was the development in the late 1950s of radioimmunoassay (RIA) to reliably quantify glucocorticoids (CORT) and other hormones (reviewed in Abraham 1975). Simultaneously, it became apparent that measuring the absolute level of glucocorticoid in circulation may not be sufficient, as hormone effects are modified by binding proteins in the blood (see an early review by Sandberg et al. 1966; and a later one by Breuner & Orchinik 2002); thus we needed to know how much of the glucocorticoid was not bound to these proteins. Finally, even knowing these two components does not give us a comprehensive understanding of the impact of glucocorticoids on the physiology of the animal and ultimately of its fitness in the natural world. Glucocorticoids are known to influence the expression of approximately 10% of the genome and its targets include genes controlling metabolism, growth, repair, reproduction, and the management of resource allocation (Le et al. 2005). Thus we need to know how changes in glucocorticoid levels influence changes in energy mobilization and allocation, growth, reproduction, immune function, rates of aging, etc. Our goals in this manuscript are two: first to review and synthesize the evidence on the role of corticosteroid binding globulin and argue that free glucocorticoid levels must be measured as the biologically relevant component in blood; and second, to review the key downstream measures that can be quantified to assess the biological impact of these levels on the immediate and long-term health and fitness of individuals. There is no question that both sets of measurements will require a significant additional degree of sophistication for physiological ecologists. However, we believe that this will provide the foundation for a quantum leap in our understanding of the ecology of stress at the level of the individual, the population and the community.

Corticosteroid-binding globulin

Corticosteroid-binding globulin is a 50- to 60-kD protein found primarily in the plasma and may regulate CORT access to tissues. In the comparative literature, the role of plasma CBG has received relatively little critical attention, despite the fact that understanding the role of plasma CBG may be essential to the proper interpretation of CORT levels. Are the biological actions of CORT regulated by total or free hormone levels? At one extreme, CBG can be seen as a means of ‘locking up’ CORT in the circulatory system, rendering it biologically inactive. Under this view, only free CORT is biologically relevant, and total CORT concentrations become unreliable indicators of the stress response. This is the basis of the ‘free hormone hypothesis’ (defined and critiqued in Mendel 1989, 1992), stating that CORT regulates bioavailability and clearance rates of hormones. If CORT bound to CBG cannot enter tissues, it cannot bind to receptors nor be metabolized and cleared by the liver and kidneys. At the other extreme, CBG may play a very limited role in determining the biological relevance of CORT. Under this view, CBG is seen as a means for transporting a hydrophobic steroid; it is implicitly assumed that the hormone dissociates rapidly enough from CBG, and the tissue uptake rate is sufficiently rapid, that bound hormone is as biologically available as free hormone. Thus, total CORT concentrations are considered the biologically relevant measure of the stress response (the ‘total hormone hypothesis’). In between these two extremes is a third position recently elaborated by Malisch & Breuner (2010), who distinguish between biological activity and biological relevance. The authors suggest that only free hormone leaves circulation and is biologically active; the remaining bound hormone serves as a reservoir of CORT in the blood to be used as needed. In this study, we will refer to this as the ‘reservoir hormone hypothesis’ (complimentary to the free hormone hypothesis). Because both CBG and CORT concentrations can fluctuate independently of each other, trends in total hormone can be different from free (e.g. Boonstra & Singleton 1993; Breuner & Orchinik 2001; Delehanty & Boonstra 2011). Understanding the role of CBG in modulating the biological response to CORT is critical, but largely unexplored.

Classic comparative studies in mammals recognized the importance of measuring both free and bound (Seal & Doe 1966; Bradley, McDonald & Lee 1980a; Boonstra & Singleton 1993), and more recent work has elevated measures of CBG across vertebrate classes (Jennings et al. 2000; Breuner et al. 2006a; Delehanty & Boonstra 2009; Clinchy et al. 2011; for general CBG methods see Breuner et al. 2003 or Boonstra et al. 2001). While these studies describe intriguing patterns and characteristics of CBG and CORT, they do not differentiate between the hypotheses; hence, the conflict between free and total hormone relevance remains. Recent articles by Malisch & Breuner (2010), Perogamvros et al. (2011) and Qian et al. (2011) move the field significantly forward in resolving this debate. The first article provides a framework to consider the ‘reservoir hypothesis’, while the latter two provide solid experimental evidence that it is the free hormone that is biologically active.

Malisch & Breuner (2010) propose that the biologically active fraction of CORT will depend on the location in the body. There is clear evidence of binding sites for CBG in mammalian tissue culture studies, and even for second messenger signalling initiated by CORT-bound CBG (see Breuner & Orchinik 2002 for review). However, this evidence is limited to tissues; there is no evidence that the bound fraction of CORT in the plasma is biologically active. It is much more likely that only free hormone passes across capillaries (see discussion of Perogamvros and Qian below), and therefore, the free fraction in the plasma is the biologically active fraction. However, the total amount of hormone in the blood is still biologically relevant. In the majority of species, ~90% of hormone is bound to CBG in the plasma and can therefore act as a circulating reservoir of CORT (Desantis et al. 2013). This effect can be local (for example, CBG-bound CORT is released at sites of inflammation when CBG is cleaved by neutrophil elastase; Hammond et al. 1990), but it can also be systemic if CBG declines, or affinity for CORT decreases. In this case much of that 90% can be released from CBG and becomes available to tissues.

Perogamvros and Qian provide strong evidence for the free hormone hypothesis. Perogamvros et al. (2011) took an in vitro approach, using a bioassay in cell culture to measure CORT activity. Specifically, they expressed the human glucocorticoid receptor (GR) in a HeLa cell culture line. Once GR is activated in this system, it turns on transcription of the luciferase gene. Hence, colour change is equal to GR activation. Human serum was stripped of endogenous hormone and then diluted to one of four serum concentrations: 100%, 50%, 20% and 10%. Each dilution was spiked with the same amount of hormone and then added to the cell culture. Hence, each replicate of cells experiences the same amount of total hormone, but less and less CBG (less serum with a higher serum dilution). As serum concentrations decline, free hormone levels increase, and luciferase activity increases as well (Fig. 1a,b). That is, the 100% plasma has undiluted CBG in it, binding up much of the hormone present, resulting in less free CORT, lower GR activation and therefore lower luciferase activity. As the plasma is diluted, less CBG is present, less CORT is bound by CBG, and more is available to activate GR. This experiment elegantly demonstrates that less CBG leads to greater activity by the CORT present. It is possible that some factors in the plasma other than CBG are having this effect. However, there is no endogenous CORT left (it was stripped of hormone), and it is unlikely that any other factor in the plasma would bind and activate GR. Hence, this strongly suggests that it is free CORT which activates GR, not the bound.

Figure 1.

Serum dilution in vitro increases free CORT levels and bioactivity of CORT present. (a) As total CORT levels increase, free CORT levels are highly dependent on the amount of serum present. With serum diluted to 10% of media, most of the CORT present is free. (b) HELA cells with glucocorticoid receptor (GR)-alpha show greater luciferase activity when less serum is present, even though total CORT present is equivalent at each X-axis value across the four serum dilution levels. These data indicate that free hormone activates GR. Redrawn from Perogamvros et al. (2011).

Qian et al. (2011) directly measure free CORT levels in the blood and tissues of rats during and after various stressors. Rats were surgically implanted with microdialysis probes to measure free CORT levels in the jugular (plasma), the hypothalamus and a subcutaneous region in the neck. If free hormone is the biologically relevant fraction, then tissue levels of CORT will likely mimic plasma levels of free CORT, and the timing of CORT rise in the tissues will match the timing of free CORT in the plasma. If CBG does not limit access of hormones to tissues, then total CORT will predict the amount and timing of CORT increase in the tissues. The former prediction was supported. Forced swim stress elevated total plasma CORT to 2500 nM within 30 min of initiation of the stressor. Free plasma CORT also increased, but elevation was limited to 80 nM and did not peak until 60 min after swim stress (30 min later than total CORT peak). Tissue CORT levels match the level and timing of the free plasma CORT. In a separate study, Estrada-Y-Martin & Orlander (2011) demonstrate that tissue CORT levels mimic free CORT levels in the plasma and that both are approximately 10% of plasma total CORT. Together, these studies indicate that total hormone is not what is available to tissues but that free hormone, not bound by CBG, is the biologically active fraction.

Further complexity in our understanding of binding of CORT in blood may occur because of changes in blood components such as free fatty acids (FFA) that affect the binding properties of either CBG or albumin (the most abundant blood-binding protein that normally binds CORT weakly). FFA are known to increase under the influence of adrenocorticotropic hormone (ACTH), and this increases in response to a stressor. Haourigui et al. (1993) found that CORT binding increases two- to three-fold in rats following an increase in FFA, with the effect lasting for up to 2 h. They argue that this is a result of increased binding by CBG because of temporary conformational changes in CBG under the influence of FFA. However, they were not able to exclude the role of other plasma proteins such as albumin. Boonstra et al. (1998) found a major temporary increase in the maximum corticosteroid-binding capacity of blood within 30 min following an ACTH injection in wild snowshoe hares. A similar increase was shown in laboratory rabbits (Boonstra & Tinnikov 1998). By selective degradation of CBG, Boonstra & Tinnikov (1998) showed that the increased CORT binding was entirely owing to changes in the binding properties of albumin, not to changes in CBG binding.

Fate of free CORT after an acute challenge

It is interesting to note, however, that much of the hormone secreted during an acute stress response may eventually make it into tissues. This is best illustrated by considering free and bound levels during each circulatory circuit of the body. Imagine a rapid, acute stressor increases CORT for 5 min. At the end of that 5 min, total plasma hormone has peaked, but only 10% of it is entering tissues. After a round through the capillaries, there will be less total hormone, but a similar fraction will still be free (with free hormone entering the tissues, more CORT will dissociate from CBG, so that there is still about 10% free). So with every round through the tissue, there is less total hormone, similar proportions of bound and free and so more hormone available to enter tissues. Of the hormones entering tissues, some will activate receptors and some will be degraded and cleared. However, the take home here is that much of the initial hormone secreted into the bloodstream during stress will make it to tissues to activate receptors. It will just arrive over an extended time frame, at lower levels than what would occur with no CBG present. The amount of hormone that eventually makes it into tissues will depend on the amount of CBG relative to CORT degradation rate in the liver and would need to be tested experimentally.

Comparative Studies

The biomedical literature has characterized the hypothalamic-pituitary-adrenal (HPA) axis of humans and laboratory species in great detail; however, laboratory data may be of limited significance in their application to wild species. Across different life-history strategies, HPA function can help animals adapt to the diverse challenges of the natural world. This is where comparative studies can provide particular insight. If patterns in CBG dynamics emerge across a range of species, life processes or in response to stress, then we can make inferences about CBG function and focus our inquiries on the most productive questions. Unfortunately, many researchers have not incorporated CBG levels into studies of stress in wildlife, so there are still relatively few studies to draw from when looking for patterns in CBG. Here, we briefly review some of the patterns that appear in the comparative literature and discuss how the reservoir hormone hypothesis affects the interpretation of existing studies.

CBG response to stress

In laboratory rodent models, studies have reported initial increases in binding capacity followed by longer-term declines. First over the short term (30 min to 2 h), there can be an increase in binding capacity (Haourigui et al. 1993; Qian et al. 2011; both discussed above). In contrast, over the long term, all evidence indicates that stressors cause a decline in binding capacity (Kattesh et al. 1980; Levin et al. 1987; Schlechte & Hamilton 1987; Frairia et al. 1988; Fleshner et al. 1995; Marti et al. 1997).

In studies on wild animals, a number of comparative studies on birds have also found rapid (within 1 h) or longer-term (≈24 h) changes, but – with a single exception – binding capacity decreases with stress. Five of 15 species show short-term (within an hour) declines in binding capacity (reviewed in Malisch & Breuner 2010). All three species examined 24 h after acute stress show a decline in CBG (Malisch, Crino & Breuner 2010; Malisch et al. 2010; Lynn et al. 2011). Two species (Japanese quail, Coturnix japonica and white-crowned sparrow, Zonotrichia leucophrys oriantha) have been examined for both short- and long-term changes: in both species, there is no short-term change in binding capacity, but a decline 24 h after a short-term stressor (Breuner et al. 2006a; Malisch, Crino & Breuner 2010; Malisch et al. 2010). Interestingly, the only wildlife studies to report an increase in CBG have measured CBG changes in response to simulated territorial intrusions. Charlier et al. (2009) and Landys et al. (2007) found that CBG levels in white-crowned sparrows (Zonotrichia leucophrys pugetensis) and male blue tits (Cyanistes caeruleus), respectively, increased after the stress of simulated territorial intrusion.

A similar mix of results is found in the few comparative mammalian studies to look at short-term changes in binding capacity. Within 1 h of capture, no change in CBG binding capacity was detected in either snowshoe hares (Lepus americanus, Boonstra & Singleton 1993) or arctic ground squirrels (Urocitellus parryii, Boonstra et al. 2001). However, within 4 h in red squirrels (Tamiasciurus hudsonicus, Boonstra & McColl 2000) and 5 h in Richardson's ground squirrels (Urocitellus richardsonii, Delehanty & Boonstra 2009), CBG levels started to decline. When snowshoe hares were injected with ACTH, there was a rapid increase in binding capacity by 30 min and lasting for at least 2 h (Boonstra & Singleton 1993; Boonstra et al. 1998). However, similar injections of ACTH had no such effect on arctic ground squirrels and red squirrels (Boonstra & McColl 2000). Thus, it seems that there may be species-specific mechanisms to alter binding capacity at least over the short term: directly by increasing CBG, indirectly in changing binding characteristics of other proteins or not at all. The near-uniformity of stress-induced long-term declines in CBG binding is compelling even with the limited number of existing studies and stands in stark contrast to the rapid and significant increase in CBG in rats (Qian et al. 2011). These declines fit easily into the reservoir hormone hypothesis. In species for which CBG declines rapidly during the course of the stressor, the decrease in CBG should strengthen the biological effects of the stress response by increasing the amount of hormone that reaches tissues. In species that show delayed CBG declines (24 h), the function of the decline seems more likely to prepare the animal for a more rapid free CORT response to subsequent stressors, although this would also depend on whether total CORT response to subsequent stressors changes. Species that show no short-term change in CBG in response to stress may be keeping CORT levels under tighter control, perhaps due to ecological pressures that favour reduced responsiveness to stressors (e.g. Breuner et al. 2003).

One area that needs further exploration before we draw firm conclusions about the biological significance of CBG changes is the role played by the standard capture–stress protocol. Almost all studies have looked at CBG changes in response to the sustained intense stress of capture and handling – a stressor that for most species has no natural equivalent. Qian et al. (2011) found that the CBG response to stress was affected by the intensity of the stressor. Most natural stressors are likely less severe than capture and handling; therefore, we need to determine whether more natural stressors (like the simulated territorial intrusion in Charlier et al. 2009) result in different CBG dynamics than what present studies suggest.

Semelparous marsupials

The most comprehensive wildlife studies of CBG and CORT to date concern a group of small Australian marsupials. The breadth of species and depth of life span covered offers a useful model to understand how life history has impacted changes in CBG and in CORT. Several of these dasyurid marsupials are semelparous – there is a complete die-off of adults after a single breeding season. Plasma CORT and CBG levels have been tracked through 5–9 months of the ~11-month life span in a number of semelparous species. In these species, a very clear trend is evident (Bradley, McDonald & Lee 1980a; McDonald et al. 1981; Bradley 1987): during pre-breeding, males maintain high CBG levels relative to total CORT levels. However, as the breeding season progresses, CBG levels decline (Fig. 2a) at the same time that total CORT levels increase. The result is that free CORT levels increase dramatically relative to pre-breeding levels (Fig. 2b). The same pattern is not seen in iteroparous dasyurids (e.g. Sminthopsis crassicaudata, McDonald et al. 1981) or other iteroparous mammals (Boonstra 2005).

Figure 2.

CBG capacity and estimated free CORT in semelparous vs. iteroparous dasyurid male marsupials. As semelparous males move through the breeding season, CBG levels decline in all four species tested, while estimated free CORT increases (Antechinus stuartii, Bradley, McDonald & Lee 1980a; A. swainsonii, A. flavipes, McDonald et al. 1981; Phascogale calura, Bradley 1987). In an iteroparous species (Sminthopsis crassicaudata), CBG increases through the breeding season, helping to keep free CORT levels low (McDonald et al. 1981).

Figure 3.

A model of free CORT levels under different temperature parameters. As temperature increases, affinity of CORT for CBG declines, increasing the amount of free CORT at each concentration of total CORT. The Y axis has been normalized to the amount free at 37 °C. Affinity of CORT for human CBG was assessed at each temperature; model was constructed assuming 600 nM CBG (a ‘normal’ amount in humans). Redrawn from Cameron et al., 2010.

This pattern also fits well with the reservoir hormone hypothesis. Under the reservoir hormone hypothesis, free hormone is a measure of immediate CORT action – it represents how much hormone is getting to tissues immediately. In contrast, what we currently know about the possible functions of the bound fraction (dealing with site-specific inflammation, release in response to fever, as a means to release more CORT via a drop in CBG) are all forward-looking functions as mechanisms to increase survival and ultimately fitness. As males approach the end of their life and must seek out and compete for mates, the value of these forward-looking functions is minimal, so the decline in CBG levels can be seen as a means of weighting the present over the future. In contrast, females still need to give birth to and raise their young to independence prior to dying. Thus, relative to males, they continue to maintain consistent CORT and CBG levels throughout their lifetime.

Variability as the dominant pattern

The final trend that emerges from the existing comparative literature is that variability is pervasive. In some species, CBG levels fluctuate seasonally, differ between the sexes or show high inter-individual variation (e.g. the section on 'Semelparous marsupials', Boonstra & Boag 1992; Boonstra et al. 2001; Boonstra, McColl & Karels 2001; Romero et al. 2008; Clinchy et al. 2011; Delehanty & Boonstra 2011; and several avian species reviewed in Malisch & Breuner 2010). This fact is critically important to the way in which comparative studies are carried out. Several recent reviews have attempted to detect broad trends in CORT levels and to interpret CORT data in terms of fitness or life-history strategies (Breuner, Patterson & Hahn 2008; Bokony et al. 2009; Bonier et al. 2009a; Hau et al. 2010). These reviews have relied on total CORT levels in their analyses. In the light of the variability in CBG levels within and among species, it is perhaps not surprising that Bonier et al. (2009a) found a mix of positive, negative and no relationships between CORT levels and fitness and that Bokony et al. (2009) found only weak relationships between brood value and CORT levels. Interestingly, Hau et al. (2010) found a stronger relationship between CORT levels and life-history traits than did Bokony et al. (2009) by using a more phylogenetically limited data set and controlling for body mass. However, under the reservoir hormone hypothesis, we predict that using free hormone levels instead of total will result in improved correlations between CORT levels and fitness or life-history traits.

New Directions to Consider in CBG Research

Recently, some of the most significant studies regarding the function of plasma CBG have been biomedical. However, they have important implications for comparative research. A recent article by Cameron et al. (2010) demonstrates a high temperature sensitivity of CBG affinity for CORT. That is, as temperatures increase within a biologically relevant range (from 35 to 41 °C), CBG affinity for CORT declines. These data support early studies by Westphal (1983), demonstrating large changes in CORT affinity for CBG in response to temperature. Affinity is a measure of how strongly a hormone binds its receptor, and small changes in affinity can have large effects on how much is bound. In the case of CBG, a high affinity means that more of the hormone present will be bound, whereas a lower affinity means less of the hormone present is bound. Hence, at low body temperatures (e.g. at night for a diurnal species), more of the CORT present in the plasma will be bound by CBG, and less entering tissues. At high body temperatures (e.g. fever), less of the CORT will be bound to CBG, and more entering tissues. Cameron et al. model the expected changes in free CORT given physiologically relevant changes in body temperature. Figure 3 suggests that free CORT levels could more than double at a high fever level than when body temperature is normal.

CORT is intimately tied to immune function, initially increasing inflammatory responses and then later inhibiting them. In cases where inflammation increases temperature locally, CBG may locally release CORT at that spot and further activate immune function. However, with longer-lasting fever, greater CORT access to tissues may inhibit immune function, putting a break on tissue damage and energy expenditure.

Many animals also enter torpor or hibernation, during which body temperatures can plummet. Does the temperature sensitivity of CBG protect tissues from any excessive CORT secretion while behavioural and metabolic options to respond to stress are limited? Westphal (1983) found that binding is 19 to 35 × greater at 4 °C than at 37 °C. Thus, if CBG is present during hibernation, then these animals must be incredibly well buffered from free CORT during the torpor phases, but much less so during the periodic arousal phases. Hence, it is possible that temperature swings in both directions could affect stress physiology in vertebrates that experience fluctuations in body temperature.

A second biomedical finding may also have implications for comparative studies. The main finding of Qian et al. (discussed above) demonstrates a significant increase in CBG capacity (total amount in the plasma) within 5 min of stress. It appears that upon initiation of the stressor, the liver dumps CBG into the plasma, increasing binding capacity by up to 50%. This increase in CBG buffers the tissues from the huge increase in total CORT and delays the increase in free CORT by up to 30 min. Rapid changes in CBG are not new to the comparative field. We have shown significant declines in CBG capacity with 30–60 min of capture and handling stress in five of the avian species examined discussed above. However, we have never detected an increase in CBG in response to the intense stress of capture and handling. The two avian studies that detected a rapid increase in CBG in response to stress (Landys et al. 2007; Charlier et al. 2009) used simulated territorial intrusions as the experimental stressor. This is arguably a more natural and less intense stressor than capture and handling, raising the possibility that birds have a more complicated relationship between CBG and stress than the rats in the Qian et al. (2011) study. The positive relationship between CBG levels and the intensity of stressors found by Qian may be limited to rats, or more broadly to mammals, or may happen in any of the avian species outside of the 15+ that have been tested. However, this effect would need to be tested across wild vertebrates before any general conclusions are drawn.

These two examples point to a role of bound CORT in the plasma, discussed in Malisch & Breuner (2010). Thus, there is now good evidence that only free CORT can enter tissues and have biological activity, but the bound fraction of hormone (often up to 95% in the plasma) is also biologically relevant. The bound fraction is available for use if the body degrades CBG or alters the binding affinity. A rapid decline in CBG (as reviewed above) or a decline in CBG–CORT affinity (as per Cameron et al. 2010) can rapidly increase the amount of CORT available to tissues, without increasing CORT secretion. In fact, Pemberton et al. (1988) initially hypothesized that CBG evolved as a steroid carrier so it could make CORT available to sites of inflammation when more CORT was needed. A common question in CBG research is the following: Why only have 5–10% of the hormone secreted available for cells? Why waste that production energy on hormone that is not used? The answer may lie in needing a reservoir for CORT in the plasma that can be manipulated through binding globulin affinity and capacity, allowing for local release of CORT to areas of the body that need it.

Thus, to summarize this section, we make four points. First, free, not total, CORT is the relevant measurement to make given that the free hormone hypothesis is so well supported. However, the reservoir of CORT left in the plasma has biological relevance, if not direct biological activity. Second, CBG levels generally decline in response to stress, except for three known examples, one in rats, one in Gambel's white-crowned sparrows, and one in blue tits. Third, CORT binding may change for reasons other than a change in CBG levels, such as changes in FFA that then affects biding properties of either CGB or of albumin. And fourth, there may be species differences in how CBG or other plasma proteins change under stress or for other biological reasons (e.g. reproduction), and we need to document and understand the reasons for these changes.

What happens next: downstream measures of chronic stress

The stress response sets off a downstream cascade of biochemical and physiological responses to deal with the stressor. Free glucocorticoid levels, as regulated by the CBG concentrations, are critical to this cascade. The magnitude and duration of these measures may give fundamental insight into the state of the animal and how different species have evolved to cope with the challenges of existence. However, why bother? There is utility in these measurements for two reasons. First, based on biomedical findings, chronic stressors are expected to have predictable impacts on a whole suite of physiological processes. So few animals in nature have been examined that we do not know the generality of the biomedical findings. Second, a series of articles have recently attempted to understand life-history patterns utilizing only the glucocorticoid (CORT) portion of the stress axis (Breuner, Patterson & Hahn 2008; Bokony et al. 2009; Bonier et al. 2009b; Hau et al. 2010, Fontaine et al., 2011). There is sufficient unexplained variation in their findings that a broader, multivariate analysis is warranted, although not presently possible given the few studies that have carried out the measurements.

We divide the downstream response into two classes, based on the nature of the stressor – acute and chronic. Downstream responses to acute stressors (e.g. brief weather events, short-duration intra- or interspecific challenges, etc.) will vary as a function of age, sex, circadian cycles, breeding condition and seasonal cycles. We have so little information on the vast majority of wild vertebrates that at present it is hard to predict what the broadscale patterns of the downstream responses to an acute stressor will look like given species differences in life history, habitat and evolutionary relatedness. Thus, we will not discuss the literature relevant to acute stressors. However, the measurements we discuss can profitably be applied to these as well. In contrast, downstream responses to chronic stressors permit more universal predictions with respect to expected changes in energy mobilization, the immune system and the reproductive system. Thus, in reviewing the literature on downstream effects, we focus on those studies in which the animals experience chronic stressors. We further limit our focus in three ways. First, we only focus on externally imposed, rather than internally imposed, stressors. By the former, we mean unpredictable factors in either the abiotic or biotic environment that are of extended duration (e.g. drought, starvation, high predation risk, social conflict, etc.); by the latter, we mean predictable seasonal factors in the life cycle of the animal to which they have evolved species-specific responses. In the case of the latter, animals may have evolved to repress or accentuate certain portions of the stress axis (Wingfield & Sapolsky 2003; Boonstra 2005). Second, given that we often do not know what constitutes ‘normal’, our focus must always be comparative – we are less interested in the absolute levels of either the CORT or downstream measures than in the differences between classes of animals (i.e. comparisons are made between animals in a control state – the benchmark – and those in a chronically stressed state). Without a control, it is a tautology to conclude that the levels in any of these indices indicate a stressed state. Third, we make a distinction among three types of studies and focus only on the first two: (i) those documenting the impact of natural stressors in the wild; (ii) those using quasi-natural stressors in the laboratory; and (iii) those using artificial stressors in the field or laboratory (e.g. CORT supplements or implants). In the latter case, evolution has not shaped the animal's response to the stressor, and thus, there is the potential for artefact. The increase in artificial CORT through supplements or implants causes endogenous levels to be constant and thus removes both the pulsatile nature of CORT release and their circadian rhythms (Charmandari, Tsigos & Chrousos 2005). Supplements also remove the direct effects of CRF, AVP and ACTH on the brain and body over and above the effects of these on the adrenal cortex (Sapolsky, Romero & Munck 2000; Bornstein et al. 2008).

The vast majority of physiological studies on stress in natural populations only measure blood CORT or their faecal metabolites. For those collecting faeces, one can measure metabolites of CORT, androgens and some female steroid hormones such as progesterone, but virtually nothing else and thus fecal techniques are not amenable to measuring the downstream effects of CORT. We will not discuss these. For those collecting blood in a capture–bleed–release protocol, all have measured CORT, some have measured CBG, and surprisingly few have measured any other downstream components. There are two reasons for this. First, many of the changes are happening rapidly and thus are time dependent. Thus, it is critical to standardize the sampling protocol (see Sheriff et al. 2011 for solutions to this problem). Second, there is too little blood, and thus, one takes a triage approach: measure CORT first, then CBG and, if there is any blood/plasma left over, other components. However, some of the latter require so little blood; this may not be an impediment to their measurement. Here, we will discuss some of the most promising things that can be measured and why, how to do it and an introduction to the evidence. For those collecting both the blood and sacrificing the animals, a large suite of measures can be taken. We will focus primarily on animals from natural populations in which natural chronic stressors have provided selective pressures that have driven the evolution in these species, rather than on those from laboratory populations where we have imposed the selective pressures. Our list is not exhaustive, but it covers those that have proved most meaningful and are relatively easy to measure.


Prediction: Glucose stores in the form of liver glycogen should increase under chronically high CORT (Fujiwara et al. 1996) because of the enhanced capacity for hepatic gluconeogenesis (Miller & Tyrrell 1995). This comes at the expense of peripheral tissues by decreasing their glucose uptake and utilization, by the release of gluconeogenic substrate through an increase in protein breakdown in several tissues such as muscle, adipose and lymphoid, and by decreasing protein synthesis. When a challenge occurs, there should be a marked hyperglycaemia. To assess an animal's glucose stores and its capacity to mobilize glucose, we need to challenge the animal with a stressor (e.g. capture stress or ACTH challenge).

Measurement: Glucose levels in mammals are typically in the range of about 100 mg dL−1, whereas those in birds can reach 400+ mg dL−1 (Clinchy et al. 2004). Measurement of glucose levels is straightforward as glucometers designed for human diabetics can be used. We have found that the best devices are the FreeStyle glucometers (Abbott Diabetes Care, Alameda, CA, USA). These use a coulometric electrochemical detection method, have a linear range of 20–600 mg dL−1, require 0·3 μL of fresh blood and give a reading in 5 s.

Evidence: Increases in glucose have been seen in most studies of natural populations (Boonstra et al. 1998; Hik, McColl & Boonstra 2001; Ruiz et al. 2002; Clinchy et al. 2004; Sheriff et al. 2011). In experimental laboratory studies on wild species (primarily on fish), most stressors (e.g. variations in population density, temperature and handling) cause marked increases in glucose levels (e.g. Ackerman et al. 2000; Lankford et al. 2005; Costas et al. 2008).

Free Fatty Acids

Prediction: Lipid stores and their breakdown products – FFA – are expected to decline because of continuous lipolysis under chronic stress (lipid mobilization and catabolization, McClelland 2004). FFA provide primarily metabolic fuel, as only the glycerol portion of the triglyceride molecule can be converted to glucose (Vander, Sherman & Luciano 1990). To assess an animal's lipid stores and its capacity to mobilize them, we need to challenge the animal with a stressor. Trapping and handling may be a sufficient stressor, although a more rigorous and repeatable stressor can be achieved hormonally with ACTH. However, there may be variation among species in the response to a stressor. Thus, for example, ACTH is sufficient to mobilize FFA in lagomorphs (Desbals, Desbals & Agid 1970; Boonstra & Tinnikov 1998), whereas in rodents, both ACTH and CORT are needed (Baggen et al. 1987).

Measurement: Non-esterified fatty acids can rapidly be measured with a colorimetric kit on 96-well microtitre plates (NEFA-C; Wako Chemicals USA Inc., Richmond VA, USA). The assay requires 2·5–5 μL of plasma depending on endogenous concentrations and has a linear range between 0·01 and 4·00 mM L−1 (these units are relative to palmitic acid).

Evidence: Free fatty acids have been found to decline with chronic stress in some studies on natural populations of birds and mammals (Boonstra et al. 1998; Clinchy et al. 2004), but more comprehensive measurement of FFA is warranted.

Packed Red Blood Cell Volume (Haematocrit)

Prediction: Haematocrit levels are expected to decline under conditions of chronic stress. Stress can directly suppress erythropoiesis (Leung & Gidari 1981; Northrop-Clewes 2008).

Measurement: Haematocrit is an index of the relative amount of red blood cells in the total blood volume and is measured after whole blood has been drawn into small capillary tubes and spun on a dedicated centrifuge (e.g. IEC Micro-Hematocrit Centrifuge, International Equipment Company, Chattanooga, TN, USA).

Evidence: Lower haematocrit values have been linked to increased predation risk (Boonstra et al. 1998; Hik, McColl & Boonstra 2001; Clinchy et al. 2004) and to poorer nutritional and health status (Lochmiller et al. 1986; Hellgren, Rogers & Seal 1993; Moreno et al. 1998).

Reproductive Hormones

Prediction: Reproductive hormones are expected to decline under chronic stress (Sapolsky 2000).

Measurement: Although it is theoretically possible to measure these declines in all the hormones, the concentrations of many of them are so low that too much blood would be required. Thus, the focus is often on the more abundant ones or on a specific one. A challenge protocol (either using a capture protocol or with an ACTH injection, followed by serial bleeds) can help to discriminate control from chronically stressed animals. Either radioimmunoassay or enzyme immunoassay methods can be used.

Evidence: Declines in testosterone have been seen with high predation risk (Boonstra et al. 1998; Hik, McColl & Boonstra 2001) and social conflict (Sapolsky 1985; Arnold & Dittami 1997), in progesterone with social conflict (Hackländer, Möstl & Arnold 2003) and in prolactin with increases in environmental severity (Delehanty et al. 1997).


Prediction: Chronic stress suppresses both the adaptive and constitutive (innate) immune responses through mechanisms that involve suppression of leucocyte numbers, trafficking, and function and changes in cytokine balance (Sapolsky, Romero & Munck 2000; Dhabhar 2008; Martin 2009). A special feature in Functional Ecology has reviewed ecological immunology, focusing primarily on disease and parasite processes, particularly as these relate to evolutionary trade-offs and fitness (i.e. see Martin, Hawley & Ardia 2011), and less attention being paid to impact of chronic environmental stressors on immunosuppression.

Measurement: In that special feature, Boughton, Joop & Armitage (2011) in their Table 1 provide an excellent overview of what can be measured, how, the problems and an introduction to the literature. We will highlight a number of studies not covered in Boughton et al.'s review or in the recent review by Davis, Maney & Maerz (2008), examining only the change in the haematology with environmental stressors.

Evidence: Environmental stress increases polychromasia – the proportion of immature red blood cell (Clinchy et al. 2004; Travers et al. 2010), declines in the levels of most lymphocytes (Baker, Gemmell & Gemmell 1998; Boonstra et al. 1998; Hanssen, Folstad & Erikstad 2003; Davis 2005; Lobato et al. 2005; Sheriff, Krebs & Boonstra 2011) and changes in the ratio of heterophils (or neutrophils in mammals) to lymphocytes in blood (cf. Clinchy et al. 2004; Mueller, Jenni-Eiermann & Jenni 2011).

Oxidative Stress

Prediction: Chronic stress affects energy mobilization and utilization and thus should increase the rate of oxidative stress. This occurs because of an imbalance between damaging reactive oxygen species (free radicals, produced as a by-product of aerobic metabolism) and endogenous and exogenous antioxidants that neutralize their effects. A special feature in Functional Ecology has reviewed oxidative stress, and we direct readers there for an overview of the primary research directions (e.g., see Costantini et al. 2010) and for reference to methods (Horak & Cohen 2010). However, the central research thrust of these articles is on evolutionary questions related to ageing, life history and reproduction (e.g. Costantini et al. 2010), not on the impacts of chronic stress.

Evidence: Chronic effects have been seen in damselflies experiencing high predation risk (Slos & Stoks 2008), in stingrays experiencing high tourist disturbance (Semeniuk et al. 2009) and in song sparrows experiencing simulated predation pressure on eggs (Travers et al. 2010).

Telomere Length

Prediction: High levels of CORT and increased energy demands associated with chronic stress may cause an increase rate of telomere shortening (reviewed in Epel 2009). Telomeres are the end caps of chromosomes and have a variety of functions, all which promote chromosome integrity and stability (Monaghan & Haussmann 2006). Reactive oxygen species play a role in their shortening, and this is associated with ageing, but chronic stress may also accelerate this process. Thus, it may be a good molecular biomarker of chronic stress.

Measurement: Telomere length can be measured with a telomere restriction fragment assay. It has been measured in a variety of tissues, but those that are rapidly dividing such as the leucocytes in blood are likely to be most indicative and easy to obtain (e.g. Haussmann et al. 2012).

Evidence: Telomere shortening associated with chronic life stress has been seen in natural populations in humans (Epel et al. 2004) and in male lizards associated with predation induced tail loss (Olsson et al. 2010) and in experimental populations in house mice (Kotrschal, Ilmonen & Penn 2007). For additional references see Monaghan & Haussmann (2006).

Acute Phase Proteins

Prediction: Stress, infection, trauma and inflammation can result in a non-specific response by the body to increase liver production of acute phase proteins (also called stress proteins) that enhance resistance to the infection and promote repair of damaged tissue (Northrop-Clewes 2008). Although they are involved in the normal housekeeping functions in cells, their presence under stressful conditions increases because of accelerated cellular metabolism, resulting in a need to prevent damage (Sørensen, Kristensen & Loeschcke 2003). In general, the response is transitory, although the changes in concentration of some of these proteins may be long term and indicate a chronically stressed state, Hsp70 being a good candidate (Slos & Stoks 2008).

Measurement: At present, quantification requires antibodies to the protein and immunoblotting, with the samples coming either from blood or from body tissues (e.g. Slos & Stoks 2008; Wiseman et al. 2011).

Evidence: The proteins can be up-regulated under predator stress in some vertebrates (Kagawa & Mugiya 2000; Fleshner et al. 2004) and invertebrates (Pauwels, Stoks & De Meester 2005; Slos & Stoks 2008; Slos, Meester & Stoks 2009), but not in others (Pijanowska & Kloc 2004; Pauwels et al. 2007).

Body Mass Changes

Prediction: Body mass should decline as a consequence of both gluconeogenesis and lipolysis as discussed above. Body mass could also decline because chronic stress may inhibit growth hormone, although there are species-specific differences (Pickering 1993; Giustina & Veldhuis 1998).

Measurement: There are at least three approaches. The first approach is to rely on ‘natural’ experiments, make longitudinal measures of the same animal when it is and is not stressed (Hodges, Boonstra & Krebs 2006). The second approach is to obtain an index based on both skeletal size and body mass and determine how these diverge in samples of animals that are stressed from those that are not (Romero & Wikelski 2001; Scheuerlein, Hof & Gwinner 2001; Sheriff, Krebs & Boonstra 2011). For long-lived species under natural stressors, this may be the best approach. The third approach is to experimentally impose a stressor and compare control and treatment groups (Zanette et al. 2011).

Evidence: Chronic stressors during adult life result in loss of body mass (e.g. Scheuerlein, Hof & Gwinner 2001; Pérez-Tris, Díaz & Tellería 2004; Hodges, Boonstra & Krebs 2006). Chronic stress during egg laying and offspring rearing or pregnancy and lactation may directly stress the mothers causing maternal transfer of glucocorticoids to the offspring (e.g. mammals: Sheriff, Krebs & Boonstra 2009, 2010; birds: Coslovsky & Richner 2011), and/or it may affect parental behaviour, causing them to feed their offspring less (Zanette et al. 2011).

Downstream conclusions

Most, but not all, of the studies examining natural stressors show the expected changes in downstream effects. Critical in terms of interpreting these findings is an in-depth understanding of the environmental stress, the reproductive state of the animals and the demography of the population, if possible before and after the measurements. The experimental studies are more variable as the stressors are more constrained by the experimenter. Two key downstream effects having organizational impact on subsequent animal function and fitness should ultimately be considered for measurement. Both require a high degree of technical skill. Hippocampal-HPA axis down-regulation in regulatory CORT receptors (MRs – mineralocorticoid receptors – and GRs – glucocorticoid receptors) occurs with chronic stress having potentially long-term effects on both the individual and its offspring (e.g. de Kloet, Joels & Holsboer 2005; see Dickens et al. 2009 for an application to starlings). Epigenetic changes of both the adults and programming of offspring may affect the animal's ability to cope with subsequent environmental challenges (e.g. methylation of the DNA: Bossdorf, Richards & Pigliucci 2008).

Future directions

In the light of the data and ideas presented within this review, we have several suggestions for future work in this field. First, it is critical to measure CBG and thus calculate free CORT levels if we are going to try to make sense of the patterns and processes in the ecology of stress in vertebrates and the vast majority of studies to date have measured only total CORT. We recommend that CBG be measured in all field studies across a broad array of vertebrate species and over the complete spectrum of their life history to assess sex-specific, reproductive state-specific and ecological situation-specific changes in CBG. It will only be then that we can assess what general principles occur. Second, we recommend detailed studies of the dynamics of free CORT and CBG under conditions of stress (over the short term and long term), to assess the generality of the laboratory findings. Third, even fewer field studies measure any metrics related to CORT (Total and/or Free). We recommend the extension of studies in wild animals to include downstream metrics, to assess the physiological impact of the changes in CGs and CBG and ultimately of the adaptiveness of these changes.


We thank the National Science Foundation (PSI-0747361to CWB) and the Natural Sciences and Engineering Research Council of Canada for funding.