Resilience in Aging Mice

Abstract

Recently discovered interventions that target fundamental aging mechanisms have been shown to increase life span in mice and other species, and in some cases, these same manipulations have been shown to enhance health span and alleviate multiple age-related diseases and conditions. Aging is generally associated with decreases in resilience, the capacity to respond to or recover from clinically relevant stresses such as surgery, infections, or vascular events. We hypothesize that the age-related increase in susceptibility to those diseases and conditions is driven by or associated with the decrease in resilience. Thus, a test for resilience at middle age or even earlier could represent a surrogate approach to test the hypothesis that an intervention delays the process of aging itself. For this, animal models to test resilience accurately and predictably are needed. In addition, interventions that increase resilience might lead to treatments aimed at enhancing recovery following acute illnesses, or preventing poor outcomes from medical interventions in older, prefrail subjects. At a meeting of basic researchers and clinicians engaged in research on mechanisms of aging and care of the elderly, the merits and drawbacks of investigating effects of interventions on resilience in mice were considered. Available and potential stressors for assessing physiological resilience as well as the notion of developing a limited battery of such stressors and how to rank them were discussed. Relevant ranking parameters included value in assessing general health (as opposed to focusing on a single physiological system), ease of use, cost, reproducibility, clinical relevance, and feasibility of being repeated in the same animal longitudinally. During the discussions it became clear that, while this is an important area, very little is known or established. Much more research is needed in the near future to develop appropriate tests of resilience in animal models within an aging context. The preliminary set of tests ranked by the participants is discussed here, recognizing that this is a first attempt.

Aging, Resilience, Health Span

Aging, Frailty, and Resilience

Aging is the leading risk factor for many of the chronic diseases that account for the bulk of morbidity, mortality, and health costs in most of the world. Indeed, aging is not merely “a risk factor” for chronic diseases; it often surpasses all other risk factors by one or more orders of magnitude. A common argument is that chronic age-related diseases occur late in life simply because of the time it takes for damage to accumulate. However, a more important reason might be that young organisms have robust defenses against homeostatic insults and challenges. With aging, these defense capabilities decline, contributing to the emergence of diseases that manifest clinically. A few examples of diseases primarily affecting the elderly adult are illustrative. Data from the Framingham study indicate that being 70 years old is, by itself, a higher risk factor for cardiovascular disease than high cholesterol, high blood pressure, and obesity combined ( 1 ), and this is because, as aging progresses, the ability of the organism to deal with insults of equal magnitude decreases: resilience is diminished.

A second example includes Alzheimer’s and other neurodegenerative diseases. In the case of familial Alzheimer’s disease—a small proportion of all cases, but an informative paradigm nevertheless—affected individuals have mutations known to cause Alzheimer’s disease inescapably, in genes such as amyloid precursor protein or presenilins ( 2 ). However, even though the mutant proteins are present from early in embryogenesis, the disease often does not appear until the fourth or, more often, the fifth decade of life ( 3 ), suggesting that age-related changes in inflammation and proteostasis may contribute to declines in resilience.

Finally, in the case of cancer, it is often argued that the disease occurs late in life simply because of the time needed for any single cell to acquire all the necessary tumorigenic mutations. However, it should also be noted that old mice of typical laboratory strains are very much prone to developing cancer. Most of them acquire it on average by only 2–3 years of age. Thus, it does not seem like the limiting factor is simply the passage of time, but in both mice and humans, cancer develops when affected individuals are roughly at two thirds of their life span. That is, cancer occurs in both species at the same relative physiological age, when decreases in resilience limit the ability of the organism to defend itself against the challenge. Interestingly, naked mole rats never develop cancer in spite of having the longest life span (more than 30 years) of any rodent ( 4 ). It seems likely that this phenotype reflects a defense system that is more robust in the longer lived species (naked mole rats), thus affording it increased resilience ( 5 ). Similarly, humans also have improved DNA repair in at least some of the pathways, compared with mice. This is to be expected because, in order to gain longevity, humans had to develop better resilience. Further evidence for this argument comes from evolutionary biology: In the case of dogs, whales, birds, and elephants, for example, cancer develops at different rates that scale with their respective life spans ( 6 ), again suggesting that in order to develop a longer life span, species need to increase their resilience to stress.

What is it about aging that makes it such a prominent risk factor? Aging is a very complex phenomenon in which genetics, epigenetics, the environment, and even chance play important roles. Aging also comprises the phenotypic effects of a lifetime of exposure to these variables. Here, two intimately connected properties that are inherent to the aging process and independent of disease will be considered: an increase in frailty and a decrease in resilience.

The separation between frailty and resilience is used here solely for operational purposes. Frailty and resilience might be considered as two sides of the same coin: an increase in frailty is related to a decrease in resilience and vice versa . For operational purposes, here we define resilience as the ability of the organism to respond to stress, whereas frailty is the decline in tissue and organism function that occurs with age. This allows a distinction between overt—or “frailty”—from covert—or “resilience”—measurements. Frailty can be observed and measured without a stress test, but resilience cannot. Thus, following this distinction, frailty can be evaluated by the typical clinical frailty measures such as grip strength, walking speed, and so on. Resilience represents the ability to respond to an added stress to the system, such as, for example, dobutamine echocardiographic stress tests in the cardiology clinic or other tests that can only be done sparingly in very old, frail people, but that can be conducted readily in animal models or younger subjects. Thus, frailty might be more closely related to the ability to meet the requirements of daily living, whereas resilience can be understood as the ability to deal with the unexpected or the unusual, including disease and clinical challenges.

Frailty and resilience might be related parts of the same process, and Figure 1 shows an example in which the loss of resilience occurs considerably earlier than frailty. For example, in adults aged 50–64 years, less than 10% of the population are frail ( 7 ), yet progressive declines in endurance capacity begins at approximately 35 years of age in nonelite and elite athletes ( 8 , 9 ). Thus it is possible that the loss of resilience might play a causative role in the development of age-related frailty. Much effort has been devoted to defining frailty in humans ( 10–12 ). Similar efforts have begun with respect to biomedicine’s animal model of choice: mice ( 13–15 ). In terms of using frailty and resilience for research purposes in animal models, scientists are still debating how to measure frailty in mice as a way to test possible interventions for potential use in humans ( 16 ). Care must be exercised in these efforts since, as mentioned, frailty occurs late in life and many researchers argue (correctly in our opinion) that using very old mice for aging research is not a good idea. Several attempts at describing measurements of health span or frailty in mice have been made, and several meetings have been devoted to that, including a joint NIA-Nathan Shock Center meeting held in Bandera, Texas in November 2012 ( 17 ) and an NIH-funded R24 Geroscience Network retreat in Santa Barbara, California in June 2014, which is reported in this issue. Despite these initiatives, the field has still not achieved consensus, with each investigator reporting different sets of outcome measurements that seem most appropriate for their particular model and study. There is however at least one point of general agreement: the most informative measures of frailty or health span are derived from assays in which the system under study is stressed. That is how resilience is operationally defined here: the ability to recover after stress.

Figure 1.

Age dependency of resilience and frailty. The broken line represents the speed (in miles/hr) of the world record holder for female marathon runners at different ages (data from http://arrs.net/SA_Mara.htm ). The solid line represents frailty in the InChianti study (personal communication, Ferrucci, L., June 2015)

There are already suggestions in the literature indicating that measurements of biological resilience in animal models might be informative. For example, aged mice and rats are more susceptible to acute challenges by cold, lipopolysaccharides, partial hepatectomy, oxidative stress, chemotherapy agents, heavy metals, and so on ( 18–20 ). Adequate response, both temporally and quantitatively, including the ability to survive—or recover from—such challenges is informative about the underlying health of the animal. In many cases, these experimental conditions are directly relevant to human health.

Approach

Purpose and Process of the Workshop

The purpose of the workshop was to consider the approaches and measures that most accurately reflect the physiological age of a mouse, in order to identify a relatively narrow set of optimal stressors that researchers can apply when assessing whether or not an intervention truly targets fundamental aging processes. Importantly, many tests have been developed that measure the effect of stress on the function of specific organs and systems in mice ( Table 1 ). There are very sophisticated methods to measure diverse aspects of cognition, cardiovascular fitness, muscle strength, and so forth. In mechanistic studies, “target purity” is important: the measurement should assess the function of the system under study with as little interference from other variables as possible. In contrast, when measuring overall frailty and resilience as an integrative measure of the individual’s health, the opposite is true: integrative responses involving multiple tissues, organs, and activities are desirable. Therefore, the focus of the discussion was on integrative measurements. The more central and integrative of various functions, the better the test will be at informing about the overall health status of the animal. Moreover, it is likely that a battery of tests, rather than a single all-encompassing one, will be most informative. An ideal battery of tests should have enough dynamic range in the response as to allow classification of individuals in easily distinguishable groups of robustly resilient, normal, and poorly or non-resilient. In contrast, a given individual will respond homogeneously across different components of the composite battery (ie if the tests are truly integrative, the individual should be either resilient or not in all domains, not in only some). There is at present no information as to whether a “truly integrative” response is an attainable goal, and it is entirely possible that only a battery of tests, challenging a set of semi-integrative responses, will provide such an overall assessment of health.

Table 1.

Resilience Tests Currently Used in Mice

Organ specific
Heart	Cardiotoxin, isoproterenol, dobutamine, doxorubicin toxicity
Vascular	Carotid occlusion
Lungs	Smoking, bleomycin
Kidneys	Dehydration, renal artery ligation
Liver	Partial hepatectomy
Skin	Wounding, irritants, psoralins
Gastrointestinal	Chemical colitis
Joints	Ligament disruption
Immune	Vaccination, topical allergen, viral/bacterial infection
Cognitive	Anesthesia, barbiturates, circadian disruption, MPTP, iron
Integrative, acute
Exercise	Treadmill, wire dangling, swimming, RotaRod
Thermal	Heat, cold
Drugs	Paraquat/diquat, barbiturates, chemotherapy, carcinogens
Surgical	Wound closing, response to blood loss, anesthesia
Radiation	Total body, hind limb
Cognitive	Maze learning, anesthesia-induced delirium
Other	Cortisol, sleep deprivation, water deprivation
Integrative, chronic
Metabolic	High fat, cholesterol, high fructose, starvation, glucose tolerance test
Trauma	Fracture healing, muscle crush, muscle freezing
Diet restriction	Weight regain, normalization of blood chemistry

Other important issues that were identified include the number of tests administered and their technical difficulty. Using the current assessments of frailty in humans as a guide, it was felt important that the stresses and outcome measurements proposed to the aging research community be simple and limited in number. In humans, many approaches to measure frailty have been proposed, measuring anywhere between five and several dozen different parameters. Many of these parameters can be assessed rapidly and accurately in humans, but mice pose additional problems, primarily because questionnaires are not an option. Basically, there are two possible strategies: an extended battery of tests that will assess resilience and health in a comprehensive manner or a limited battery that could serve as a first approach to the issue. The group elected to focus discussion on the latter. It was felt to be important that any recommended tests be cheap, simple, and reliable, so that researchers in any institution might be able to perform them as a first approximation before engaging in further testing using more sophisticated equipment, knowledge, and expertise.

Another important issue discussed is relevance to human aging. It is clear that, in testing interventions in mice, it is the health of the mice that is the relevant outcome. In a rodent, impaired hair growth can be an important health indicator that is not often considered relevant to adult humans. However, because hair growth is important for the health of the mouse, an intervention affecting this parameter could be important in assessing the impact of the intervention on the animal’s overall health. Therefore, such a parameter could well serve as a valid outcome measurement, even though it has no direct relationship to human aging. A more extreme example, which is indeed used in research, is pharyngeal pumping by Caenorhabditis elegans ( 21 ). Most researchers agree that reduced pumping is in fact an age-related phenomenon in the worm. It is used routinely as a measure of health in this model, but most would agree that pharyngeal pumping is not a usual measure of “general health” in humans. Nevertheless, there is of course a clear added advantage to choosing health measurements in mice that are relevant and comparable to measurements in humans. As it turns out, with respect to stressors that can be used to test resilience, stress scenarios can often be chosen that have an analogous human clinical counterpart. Examples include exposure to chemotherapy agents, anesthesia, cold, high fat diets, infections, and so on that are relevant either as clinical or as “natural” stresses. Again the availability of these choices makes the focus on resilience more likely to be fruitful than measurements of the overt, static aspects of frailty.

Experimental Methods for Testing Resilience

Table 1 includes all the potential stressors that were discussed at the workshop. Using the ranking criteria considered earlier as a guide, and based on the experience of the participants and the very limited published data about these stressors, the group ranked them as shown in Table 2 . The general consensus was that these could serve as a very rough, initial guide for designing studies of resilience. These tests would need to be followed up by more detailed physical function analyses directed by this initial screen. The group also noted that few or no data are currently available about the validity, sensitivity, specificity, and reproducibility of most of these tests, particularly in the contexts of aging or animal models of age-related diseases. Therefore, the group felt strongly that considerably more work is needed before these tests can be accepted as valid. The most promising stressors are briefly outlined below.

Table 2.

Proposed Resilience Measurements in Animal Models

Starvation (including weight regain after dietary restriction)

Water deprivation

Anesthesia, chemotherapy

Cold stress

Trauma

Paraquat

Others: Cortisol, circadian rhythm, infection, barbiturates

Starvation

Exposing mice to periodic cycles of prolonged fasting, including water-only fasting, increases health span and life span as well as the resilience to exogenous stressors such as chemotherapeutic agents (see subsequently). However, prolonged water-only fasting represents a major challenge that involves a remarkable metabolic reprogramming associated with a reduction in the number of cells, ranging from hepatocytes to immune and muscle cells, leading to reduced size of various organs. By Days 4 and 5, depending on the genetic background, mice given only water die from causes that are poorly understood ( 22 , 23 ). Although humans can withstand periods of water-only fasting of 30 or more days, the complexity and extreme challenge imposed by water-only fasting in mice represents an ideal and inexpensive way to evaluate resilience, thereby avoiding killing of the mice. In line with the regulations of most Institutional Animal Care and Use Committees, loss of body weight should not exceed 20% during the fasting interval, thereby making routine inspections necessary at least every 12 hours. The 20% weight loss limit ensures the full recovery of animals and effectively standardizes the fasting protocol, thereby allowing evaluation of resilience. Following the onset of starvation, mice display coprophagy and increased activity related to foraging and should therefore be placed in clean housing prior to starvation. Signs of severe discomfort including hunched body posture, lethargy, and coldness to touch, which may occur independently of duration of starvation or weight loss, are strong indicators that food should be resupplied. This increase or decrease in activity can also be considered for experiments such as behavior tests. Frail or old mice are less likely to be resilient to prolonged starvation, making shortened periods of food deprivation and close monitoring in experimental settings necessary. For example, whereas multiple cycles of a 4-day fasting-mimicking diet were tolerated in middle-aged C57Bl/6 mice, old mice struggled to regain weight that resulted in increased mortality ( 24 ). Resilience was reestablished by shortening the fasting-mimicking diet to 3 days.

Body weight measurements during the water-only fast and in the refeeding period can be utilized to assess resilience. For example, in young mice, refeeding normalizes body weight to pre-fasting weight within 2–3 days ( 25 ). Old mice may require more time to recover from weight loss, potentially due to an age-dependent decrease in intestinal nutrient transport ( 26 ). The following are examples of measurements that could be easily carried out to determine resilience in mice of different ages:

1. Monitoring of the rate of weight loss during water-only fasting.

2. Monitoring of the rate of weight regain and ability to return to pre-fast weight after mice reach the 20% weight loss point. Monitor glucose and ketone bodies.

3. Ability of mice to perform behavior tests after 2 days of water-only fasting (RotaRod, Barnes maze, novel object recognition).

4. Measuring muscle strength.

Water deprivation

Water deprivation has been used in mice as a stress of the endocrine, renal, and central nervous system components that control fluid and electrolyte balance, a means to evoke water-seeking behavior, or a way to enable provision of water to be used as a reward during cognitive function testing ( 27 ). The thirst control system, including the posterior pituitary, adrenal, and renal systems that control water balance, become dysfunctional in many humans and experimental animals in old age ( 28–30 ). Older humans can have impaired thirst and osmotic control, altered activity of the arginine vasopressin and renin–aldosterone systems in response to water deprivation, and renal focal glomerulosclerosis with decreased urine concentrating ability ( 31 ). The impact of water deprivation in older mice has not been studied extensively. However, it does appear that overnight fluid deprivation or provision of food as dry pellets results in less fluid consumption over a 1-hour period in 25 than 3- or 12-month-old mice ( 29 ), paralleling the decline in dehydration-induced thirst that is observed in some older humans. However, thirst induced by hypertonic saline administration elicits the same drinking response in old as younger mice. The impact of water deprivation on cognitive and other domains of function may be greater in older than in younger animals ( 32 ). On balance, water deprivation appears to be a potentially good test of resilience in multiple systems in older mice, a point that merits further investigation. Since there are many age-related changes in the systems affected by water deprivation, the most interpretable experimental paradigms may be those in which animals of the same age are given a test intervention or placebo and then are subjected to water deprivation. Practically, 24 hours of water deprivation is often used in mice, but this appears to be quite severe. A starting point in older mice might be 6 hours. Effects of water deprivation in the test intervention-treated and placebo groups on physical or cognitive function, as well as tests of the renin–aldosterone system, blood urea nitrogen:creatinine ratio, blood and urine osmolality, and blood and urine sodium, could follow the period of water deprivation. If the test intervention actually alleviates fundamental aging processes, it would be anticipated to result in less severe effects of water deprivation on these outcome measures than placebo.

Anesthesia

The majority of people who have surgery requiring anesthesia are elderly patients ( 33 ). An almost daily occurrence in health care facilities is the need to ascertain the risk of complications of anesthesia and surgery in frail, older patients for whom surgery is indicated. Such subjects have increased risks of prolonged time in the postoperative recovery room, increased frequency of delirium, delay in time to recover mobility, increased risk of pneumonia, prolonged hospitalization, and increased frequency of discharge to rehabilitation facilities or nursing homes instead of home ( 34 ). Anesthesia itself may confer risks of these complications in older subjects above and beyond those of the associated surgical procedure ( 35 , 36 ). The effects of anesthesia have previously been modeled to some degree in mice ( 37–39 ). More work is needed on the potential value of anesthetic stress as a test of effectiveness of interventions that target fundamental aging processes.

Chemotherapy

Chemotherapy and radiation treatment are interventions that around a third of people will experience in their lifetime. These interventions frequently induce short-term side effects including weakness, fatigue, anorexia, and confusion that tend to be more severe in older than younger subjects, especially in frail subjects, to the point that doses sometimes need to be reduced or treatment foregone in elderly subjects. Intermediate and longer term side effects include the same symptoms as well as hematological dysfunction and many agent-specific side effects, such as heart failure in the case of doxorubicin or pulmonary fibrosis in the case of bleomycin, which may also be more severe in older than younger subjects ( 40 , 41 ). In addition, chemotherapy and radiation treatment have been associated with development of an accelerated aging-like state years later, suggesting a link between these stresses and fundamental aging mechanisms ( 23 , 42–45 ). Indeed, both chemotherapy and radiation can induce cellular senescence in vitro and in vivo ( 46 ).

Experimental animals exposed to chemotherapy or radiation develop similar short- and long-term side effects ( 47 ). As in humans, these side effects may be more severe in older than younger mice (personal communication, Longo, V., February 2016). Brief caloric restriction appears to reduce the severity of short-term side effects of chemotherapy in both mice and humans ( 48 , 49 ). The clinical relevance of these interventions, the more frequent adverse effects in older than younger mice, their relation to fundamental aging mechanisms, and their alleviation by at least brief caloric restriction makes chemotherapy and radiation attractive as perturbations for resilience testing in experimental animals to measure the effectiveness of interventions that target basic aging processes. Among the agents that have been used in mouse aging, studies are doxorubicin, bleomycin, cyclophosphamide, as well as from 6 to 60 gray radiation, either to the whole body or locally (above 7 gray of whole body radiation is frequently lethal in mice, but higher levels can be tolerated when restricted areas are radiated). It must be noted that different chemotherapy treatment regimens adversely affect different organ systems. Thus, the effects of these compounds on overall resilience may be secondary to the primary organ-specific effects. For example, doxorubicin is particularly toxic to cardiomyocytes and could be used to assess the effects of cardiac function on overall resilience. Conversely, cyclophosphamide primarily targets bone marrow cells and could be used to assess the role of the hematopoietic system in resilience. A great deal of work remains to be done to establish reasonable ranges of doses for resilience studies in mice of different ages, strains, and functional levels. Furthermore, work needs to be done to ascertain which outcome measures, for example, treadmill endurance, novel object recognition, or other tests, are more sensitive for detecting whether interventions that target fundamental aging mechanisms enhance resilience in response to chemotherapy or radiation stress.

Paraquat exposure

Oxidative stress has been postulated to contribute to aging phenotypes since the 1950s ( 50 ) and has been studied extensively in the decades since ( 51 , 52 ). Several lines of evidence suggest that oxidative damage accumulates with advancing age ( 51 ). Moreover, young animals and animals with increased longevity due to life span-extending interventions are often protected from chemically induced oxidative stress ( 53–58 ). As such, the administration of chemical stressors (eg paraquat) represents another potential avenue by which resilience can be tested in animal models. Given that calorie restriction can reduce the toxic effects of oxidative stress ( 55 , 56 ), it seems logical that other pro-longevity interventions could be tested in this manner. Historically, a large bolus of paraquat has been administered followed by the plotting of survival curves. Future studies in animal models could certainly be performed in this fashion, but longer term administration of sublethal doses may also have a role since age-related changes in short term stress responses might differ from age-related changes in longer term adaptations to stress.

Cold stress

Humans can develop cold-intolerance and higher susceptibility to hypothermia with advancing age ( 59 ). The mechanisms underlying these declines are multifactorial and may include a reduction in skin sympathetic nervous system activity, mild hypothyroidism, and the onset of age-related lipodystrophy and sarcopenia ( 60 , 61 ). Studies in mice and rats have also indicated that rodents can have similar age-related declines in thermoregulation and resistance to cold stress ( 62–64 ). Cold stress is potentially a cheap and easily reproducible challenge for the evaluation of resilience in rodents. Most rodent vivariums are maintained at temperatures conducive to human comfort (19°C–22° C) and not rodent thermoneutrality (30°C–32°C) ( 65 ). As such, laboratory rodents are often already subjected to mild cold stress that has been shown to affect heart rate, metabolic rate, and immune function ( 66–69 ). Detailed studies are needed to refine protocols for applying cold stress and to determine its effects in old mice, including studies of such variables as the ambient temperature at which mice are raised before the cold stress is applied, the duration of the cold exposure stress, and the temperature applied during the stress.

Trauma

Older adults currently account for more than 25% of all trauma-related hospital admissions in the United States, and this number is projected to increase as the population grows older ( 70 , 71 ). Preexisting comorbidities often render older adults who have succumbed to traumatic injuries much less resilient than younger adults during the recovery phase, thus leading to longer hospital stays, higher incidence of intensive care admissions, and greater mortality rates ( 70 , 72 ). Several types of experimental trauma models have been developed and therefore provide opportunities to evaluate resilience in the context of these models. The models that appear most relevant to the aging population include long bone fracture ( 73 , 74 ), traumatic brain injury ( 75–77 ), and burn injury models ( 78 , 79 ). The vast majority of these models are easily reproducible, although an initial development and implementation phase is required for new users of the models. In addition, detailed experimentation is currently needed to refine these protocols for use in old or frail mice and to optimize the dynamic range in younger animals. Once the appropriate magnitude of traumatic stress and appropriate measures of recovery have been established, initial studies in long-lived mouse models should provide baseline data for evaluation of resilience for comparison purposes.

Responses to Stresses

In addition to more research on determining the value of particular stresses for studies of resilience, more work is needed to devise, validate, and optimize measurements of the responses relevant to each stress. Of course the optimal outcome would be a fast and full return to homeostasis after cessation of the stress. This might not always occur in aged animals, and close attention must be paid to the kinetics of response. In young individuals, response to a challenge is usually fast, and return to basal levels occurs after a variable time, depending on the time it takes to neutralize the specific insult. In contrast, in older individuals, many additional elements in the kinetics of the response (lag time, amplitude, duration, and so on) can play a role and might be informative of the individual’s response capacity. For example, and for reasons that are poorly understood, basal levels of several cytokines are already elevated in older individuals. In response to an inflammatory stimulus such as lipopolysaccharide, young and old animals likely differ in the lag time to response, the slope, the maximal response, the duration of the response, and even the residual (that might or might not be related to the fact they have increased cytokine levels to start with). Of particular importance therefore is the fact that, due to these differences in kinetics, single time point measurements can lead to erroneous conclusions, since the time of maximal response is not necessarily the same in the young as in the old. In Figure 2A , several of these parameters have been changed in a hypothetical situation (dashed response line), where the basal level is higher, the lag longer, and so on. In this hypothetical example, measurement at the time of maximal response in the continuous line would lead the experimentalist to conclude that the response is blunted in individuals represented by the dashed line, which is not the case when the entire area under the curve is examined instead. Which of these variables are likely to be the most relevant and informative? Although some challenge/response paradigms are suitable for continuous measurement (eg insulin/glucose), not all of them are. A related parameter is the refractory period or the time before the organism can effectively respond to a repetition of the same or a different stress. This is illustrated in Figure 2B , where—absent hormesis—shortening the time between stresses will lead to a decreased response as illustrated, or eventually to inability to respond, resulting in death, if the stimuli are too closely spaced. In that sense, it is also important to consider whether survival to a lethal stress is more or less informative than measurement of the kinetics of response or ability to recover and to regain ability to sustain a new insult.

Figure 2.

Responses to a stress. (A) A response to a stress can have several stages and elements, including basal level (B), lag time (L), rate of increase (Δ), amplitude (A), length of sustainability (Δ + RB), and rate of return to basal (RB). There might also be a residual if the parameter does not return to basal level (Rʹ).

Summary

The notion of testing the impact of interventions based on the fundamental biology of aging on health span is relatively new, with numbers of studies increasing over the past 5 to 6 years. Unlike in other areas of basic or applied biomedical research, such as in endocrine or cardiovascular biology, very little formal work examining the value, utility, or relevant technical details involved in measuring resilience in aging experimental animals is available. Pockets of expertise in particular types of dynamic studies in aging mice are beginning to appear, but a great deal more research is needed to develop resilience tests suitable for aging biology studies and to create an evidence base. Among the many questions that need to be addressed are: (a) is the change in resilience with aging general or segmental, affecting only a limited subset of physiological domains?, (b) how much value will these tests add to available standard static measures of frailty, health, or pathology in aging mice?, (c) will a screening battery, standardized across laboratories, be useful in comparing interventions that affect fundamental aging processes?, (d) how well will these tests correlate with gold standard measures such as life span or other health span measures?, and (e) will these tests actually be more sensitive to effects of interventions than static tests of health span? Although much more research is still needed in each domain, the workshop aimed at beginning to address the issues and hurdles still ahead, as well as to identify those stressors that, in light of current knowledge, might be the most informative.

It is important to emphasize that a battery of tests may be optimal. No one test will likely suffice, and just as for health span, it is not expected that any one intervention will improve all aspects of the battery. In addition to tests of resilience proper, measures of regular “health span” will need to be included that can be used to quantify, for example, recovery of voluntary ambulation after a given, measured stress. Since loss of resilience occurs earlier than overt signs of frailty, it is possible that some resilience tests might also predict future health. That would be “icing on the cake,” but the focus at this early stage should remain on “current health status of the animal.” Furthermore, it remains to be established whether the different measurements relate to each other. If they are truly measures of overall health, at least some overlap and cross-effects should be expected. Careful selection of tests will always be necessary, rather than a fixed “panel” of measurements. Some life span interventions might affect the parameter under study. For example, we already know that some of the manipulations that extend life span also improve drug metabolism. This is particularly likely in the context of chemical stresses (chemotherapy, anesthesia, diquat, etc.), where the “stressful” molecule is not the one provided to the animal but a metabolite produced by the liver. If the liver is more effective at producing the active compound, then there could be an unintended effect on the concentration of the stressor. One possible solution to this problem would be to temporarily stop the treatment before testing, to allow enough time for the drug to clear. Indeed, a “proof-of-principle” for this approach was published after the workshop ( 80 ). Of course, this is not so easy for nonpharmacological interventions. We also need to be aware of interventions affecting the parameter under study: a stress to muscle might not be a good choice for animals on dietary restriction. In longitudinal studies, it will be very difficult to avoid “carry over” effects, including hormesis. Finally, attention must be given to possible interference by disease states. Therefore additional studies including pathological assessment also need to be conducted.

Acknowledgments

Proceedings of a workshop at the National Institute on Aging in Bethesda, MD on August 27, 2014. Additional meeting participants included Rafael de Cabo, PhD, Luigi Ferrucci, MD, PhD, Tamara Harris, MD, MS, Donald K. Ingram, PhD, Arnold J. Kahn, MS, PhD, Sean X. Leng, MD, PhD, Valter D. Longo, PhD, Richard A. Miller, MD, PhD, Janko Nikolich-Žugich, MD, PhD, Charlotte A. Peterson, PhD, Arlan G. Richardson, PhD, George E. Taffet, MD, Mark I. Talan, PhD, and Jeremy D. Walston, MD The authors are indebted to them for their intellectual contributions to both the workshop proceedings and this manuscript.

Author notes

Decision Editor: Rafael de Cabo, PhD