Lack of adequate appreciation of physical exercise's complexities can pre-empt appropriate design and interpretation in scientific discovery

Transgenic/knockout animals

We will contend here that the physical activity level of genetically modified animals, in some cases, is critical to the interpretation of a gene's complete function. This is important because the gene often operates with fundamentally different functions depending on the physiological context of high or low physical activity (as described above). If the function of a gene is characterized only during low physical activity, but the function of this same gene differs in response to high physical activity, then the incomplete function of the gene is obtained and placed into the database. Our contention is demonstrated by two scenarios (Fig. 1).

Open in a separate window

Figure 1

Presence or absence of phenotype when comparing forced inactivity to physically active in Scenarios A and B and resultant conventional and correct interpretations

See the text for detailed explanations.

Scenario A of Fig. 1 is demonstrates how exercise can reveal phenotypes not present in resting transgenic mice. For example mice with a mutated myosin binding protein C appeared overtly healthy at rest, but they exhibited a phenotype of decreased exercise capacity and death in four out of eight of the mice within 24 h of ending treadmill running exercise in one study (Yang et al. 2001). A review wrote, ‘There are numerous examples of transgenic models in which the baseline cardiovascular phenotype is unchanged or minimally changed from the wild type, only to become manifest during the stress of exercise testing (Bernstein, 2003). Another example is hormone sensitive lipase (HSL) knockout mice that at rest have no differences in circulating fatty acids or liver glycogen content, but exercise reveals an inability to increase circulating fatty acids and a accelerated reduction in liver glycogen, thereby appropriately defining a physiological role for HSL (Fernandez et al. 2008). By exercising, disruption of physiological homeostasis necessitates attempts to restore homeostasis through the activation of the specifically targeted genes. Thus, physiological function of the genes can better be elucidated.

To ensure proper characterization of gene function, we believe that one must test the function of a gene in response to physiological stresses (i.e. high physical activity) that occurred during the evolution of our present genome. Throughout evolution, high levels of physical activity were obligatory for survival of animals and humans (Booth et al. 2000; Booth et al. 2002; Chakravarthy & Booth, 2004; Booth & Lees, 2007). Bennett and Ruben elegantly wrote, ‘The selective advantages of increased activity capacity are not subtle but rather are central to survival and reproduction. An animal with greater stamina has an advantage that is readily comprehensible in selective terms. It can sustain greater levels of pursuit or flight in gathering food or avoiding becoming food. It will be superior in territorial defense or invasion. It will be more successful in courtship and mating’ (Bennett & Ruben, 1979). Until only recently, physical activity was obligatory for survival. Today, mice and rats housed in cages unable to mimic evolution's physical activity levels may not completely elucidate the evolved gene's function. Thus, the failure to disturb homeostasis in a sedentary animal (without access to voluntary wheel running) will often fail to uncover the true evolutionary function of the gene. Humans also inherited evolutionary gene functions during a physically active lifestyle. For instance in scenario A, the appearance of a new phenotype following voluntary physical activity reveals the biological purpose for which the gene was selected.

In scenario B of Fig. 1 we discuss the situation where normal physiology rescues the disease phenotype that occurs at rest. As an example, we will use melanocortin-4 receptor knockout (MC4R^−/−) mice that develop a maturity-onset obesity syndrome associated with hyperphagia, hyperinsulinaemia, and hyperglycaemia (Huszar et al. 1997). In response to this phenotype, the Mouse Genome Informatics Website of Jackson Laboratories states, ‘Mutations in this gene (MC4R) result in hyperglycemia and weight gain’. However, this phenotypic information was obtained from laboratory mice with a low level of physical activity. When MC4R^−/− mice are allowed to increase physical activity levels with voluntary wheel running, this led Haskell-Luevano et al. (2004) to write, ‘Herein, we present the effects of voluntary exercise on the MC4R knockout mice in terms of bypassing the morbid obesity and hyperphagia phenotypes associated with this genetic obesity model’ (italics added); and also wrote (Haskell-Luevano et al. 2009), ‘voluntary exercise can prevent the genetic predisposition of MC4R-associated obesity and diabetes.’ Allowing mice to have normal physiology by placing wheels for running into a cage prevented obesity, making the interpretation of Jackson Laboratories only valid with sedentary mice. The opposite conclusion is reached when the mutation interacts with increased levels of physical activity, suggesting that MC4R plays no role in metabolic dysfunction. Studied in the context of a sedentary environment, gene interaction is more akin to a gene's role in pathology than in normal physiology. In this way the interaction between sedentary behavior and gene function may provide information about current sociological levels of physical activity in a pathological context, but not a physiological context. Indeed in the context of normal physical activity, the effect of MC4R on the lean phenotype is minimal. A revised entry into the functional protein database should read, ‘The response to the physiological stress of increased physical activity results in no phenotype in the MC4R mouse. However, in response to a sedentary pathological stress, the MC4R mouse develops obesity, hyperinsulaemia, and hyperleptaemia associated with type 2 diabetes.’ An amazing discovery is thus made for an interaction between the MC4R gene and low physical activity compared to evolutionary levels.

OLETF rats, have a spontaneous mutation to inactivate the cholecystokinin-1 receptor (CCKR1), causing hyperphagia. Similar to MC4R^−/− rats, above, no disease phenotype exists when OLETF rats are permitted voluntary running (Morris et al. 2008). However, a disease phenotype of obesity, type 2 diabetes and non-alcoholic fatty liver disease is revealed when voluntary running is abolished with CCKR1 mutation (Rector et al. 2008b), indicating that CCKR1, like MC4R^−/−, should be labelled as having no phenotype in physically active animals. If protein function databases wish to designate a disease phenotype, they need to indicate that the interaction of low physical activity is necessary to show disease. These examples provide evidence to study knockout transgenic models both during an acute exercise bout and at rest following chronic repeated exercise training better insight into ‘normal’ physiology is gained.

Exercise sciences

The first reservation is that human physiological adaptations to disruptions in homeostasis caused by exercise are so complex that no mono- or poly-pill will ever be able to mimic every exercise adaptation for six reasons.

1. Exercise adaptations are multi-organ. Adaptations to exercise and lack of exercise occur in almost every organ system, such as circulatory, neural, endocrine, skeletal muscle, connective tissue (bones, ligaments and tendons), gastrointestinal, immune and kidney (Fig. 2). Systems where exercise adaptations don't occur or are not well documented to occur are lungs and senses. Pharmaceuticals, on the other hand, must limit their effects to single organs. This is recognized in the Concluding Remarks and Future Perspectives in a review article by Puigserver and Spiegelman, writing, ‘Because of the role of PGC-1α in many important metabolic processes, it is worth asking whether and how the activities of PGC-1α might be modulated for therapeutic purposes …[of course, this wide range] wide range of physiological actions [by PGC-1α in multiple tissues] also points out a problem: activation or inhibition is going to have to be tissue selective or tissue specific to be useful. … This suggests that the usefulness of PGC-1α as a therapeutic target will all come down to an ability to achieve biological specificity’ (Puigserver & Spiegelman, 2003). Since exercise produces adaptations in multiple organs, exercise has a natural ‘built-in’ specificity that is unmatched by the therapeutic limitations of drugs.
2. Molecular biology has led to the idea that exercise is polygenetic. Already in 1996, 50 mRNAs had been reported to be altered by exercise/exercise training in skeletal muscle alone (Booth & Baldwin, 1996). More recently, 900 mRNAs in skeletal muscle had greater than 2-fold changes during the 24 h after either 3 or 12 h of voluntary running (Choi et al. 2005). Interestingly, lack of physical activity by hindlimb immobilization of rat (involving the soleus muscle) required an exhaustive search for an mRNA in skeletal muscle that did not change (Pattison et al. 2003). Both of these studies indicate that many more than just one or two mRNAs change after alterations in contractile activity. The cumulative data demonstrate that the complexity of tissue physiological responses to alterations in exercise levels extends to changes in thousands of mRNAs in skeletal muscle. In addition to skeletal muscle, data exist to support several hundred mRNAs changing in peripheral blood mononuclear cells (Radom-Aizik et al. 2009) and neurophils (Radom-Aizik et al. 2008) of humans, and in the liver (Colombo et al. 2005) and cardiac muscle of rats (Strom et al. 2005). Carey & Kingwell (2009) conclude with regard to health benefits of regular exercise. ‘Obviously no single pharmaceutical agent could mimic this response.’ Altogether, the combination of a number of different organs having individualized changes of hundreds of genes in response to exercise presents science that makes it very unlikely that one or two molecules could prove to be therapeutic targets for an ‘exercise pill’ in the near future.
3. Different exercise modalities result in different mRNA expressions, protein expressions, and physiological phenotypes. At the mRNA levels, differential gene expression occurs in the same tissue in response to resistance and endurance exercises (Holloszy & Booth, 1976; Stepto et al. 2009). For example, at the protein level, contractile protein content (per whole muscle) increases per whole skeletal muscle with resistance training, but remains unchanged per unit of mass. On the other hand in response to endurance training, contractile protein content and concentration remain largely unchanged. Conversely, both mitochondrial protein concentration and content (per whole muscle) increase in endurance-trained individuals, while for strength training only total content and not concentration increases. Lastly, an example of differing phenotypes in the same organ in response to two types of endurance exercise is provided by changes in bone mass. While bone mass increases in endurance running, it decreases in endurance cycling (Rector et al. 2008a). Thus even if an ‘exercise pill’ produced the same (or a very similar) effect as one type of exercise, it would have difficulty replicating other types of exercise that have their own unique health benefits.
4. Knockout animals show us that one key exercise gene is unlikely to exist. The observation that most mitochondrial genes were reduced in peroxisome proliferator-activated receptor γ coactivator 1 (PGC-1α) knockout mice led to the conclusion that PGC-1α was the ‘master regulator gene of mitochondrial biogenesis’ (Wu et al. 2002). However, these results were obtained with sedentary mice. When PGC-1α knockout mice were endurance-exercise trained, they still had some or all exercise adaptations in two different studies. Piltgaard and colleagues (Leick et al. 2008) found that endurance exercise training increased aminolevulinate synthase, cytochrome oxidase I, cytochrome c, and hexokinase II mRNA and protein levels in skeletal muscle equally in wild-type and PGC-1α whole-body knockout (KO) mice. They concluded that ‘PGC-1α is not mandatory for exercise training-induced adaptations in skeletal muscle mitochondrial proteins.’ (This is another example of an error in gene classification presented earlier in the first portion of this review.) Another putative master regulator of exercise induced phenotypes is AMPK, which is the target of the proposed ‘exercise mimetic’ AICAR. However, when Pilegaard and colleagues tested the response of whole-body α2- and α1-AMPK knockout mice to an endurance-exercise training protocol, a different conclusion was reached (Jorgensen et al. 2005). They stated, ‘In conclusion, KO of the α2- but not the α1-AMPK isoform markedly diminished AMPK activation during running. Nevertheless, exercise-induced activation of the investigated genes in mouse skeletal muscle was not impaired in α1- or α2-AMPK KO muscles.’ Pilegaard and colleagues also concluded, ‘other factors can take over for training-induced responses in muscles when either AMPK or PGC-1α is lacking’ (Leick et al. 2008).

Furthermore, AICAR does not mimic many biochemical adaptations to either an acute exercise bout or endurance exercise training. For example, (a) after a single injection of AICAR, plasma fatty acids fell with no change in plasma glycerol while an exercise bout increases both of them – in the same rats, AICAR had no effect on skeletal muscle glucose-6-phosphate or glycogen concentration, but both decreased after a single bout of exercise (Rantzau et al. 2008); (b) AICAR is less effective in attenuating ACCβ Ser²¹⁸ phosphorylation (inferring less AMPK activation) than exercise training (both 10 days of treatment) (McConell et al. 2008); (c) AICAR reduces IL-6 mRNA expression and protein release (Glund et al. 2009), while exercise increases both of them (Ostrowski et al. 1998); and (d) AICAR does not increase sarcolemmal fatty acid binding protein concentration, but muscle contraction does (Jeppesen et al. 2009). Therefore, AICAR does mimic biochemical adaptations to endurance exercise as claimed in a recent publication (Narkar et al. 2008).

These examples clearly show that no one putative master regulator of exercise phenotype can completely prevent exercise-induced adaptations from occurring. Indeed, from an evolutionary viewpoint, the ability to adapt to increased levels of physical activity was necessary for survival, likely containing multiple overlapping pathways responsible for single adaptations. Thus, both the science and evolutionary logic predict that a single molecule will be unable to ‘mimic’ exercise-induced adaptations.

(5) Multiple signalling transduction. It is beyond the permitted word length here to review the tremendous number of signalling pathways activated by exercise, but others have just commented on the complexity of metabolic homeostasis during exercise and the inappropriateness of the term ‘exercise mimetic’ (Carey & Kingwell, 2009). In the above examples, two such pathways are examined. What will be mentioned is that some of these pathways are likely to be differentially modified by resistance- and endurance-type exercise. It is unlikely that a single molecular target, like AMPK, would activate the same, or similar, pathways for increasing mitochondrial concentration and skeletal muscle mass to meet the definition of mimetic as claimed (Narkar et al. 2008). It would be dangerous to claim, as is being done on Nova's Marathon Mouse web site (http://www.pbs.org/wgbh/nova/sciencenow/0403/03.html), that AICAR is an exercise pill for the elderly because AICAR does not counter loss of muscle strength, a major cause of mortality in elderly (Metter et al. 2002; Newman et al. 2006; Cesari et al. 2009), but decreases signalling by inhibiting signalling proteins that regulate the initiation of protein synthesis in skeletal muscle (Deshmukh et al. 2008).

(6) Claims of ‘exercise pill’ are flawed. Methodological concerns for the paper claiming that AMP kinase is an exercise mimetic have been presented in our recent Viewpoint article (Booth et al. 2009).

Pharmacological sciences

ED₅₀ and LD₅₀. The dose of drug required to produce a specified effect in 50% of the population is the median effective dose (ED), abbreviated as ED₅₀. In preclinical studies of drugs, the median lethal dose (LD), as determined in experimental animals, is abbreviated LD₅₀. The ratio of LD₅₀ to ED₅₀ is an indication of the therapeutic index. An exercise pill would be required by the US Food and Drug Administration to obtain a therapeutic index in pre-clinical testing. Comparisons of therapeutic indices would be demanded by us in order to prove that an ‘exercise pill’ was as effective as natural physical activity in healthy populations.

No known LD₅₀ exists for exercise. It is unlikely that an exercise LD₅₀ exists. In order to determine whether LD₅₀ exists for exercise or not, an American Heart Association Statement on Exercise and Acute Cardiovascular Events – Placing the Risks into Perspective was used as a reference source by Thompson et al. (2007) and some of their cited studies were used to obtain the following death rates. The absolute rate of exercise-related death among US high-school and college athletes was 1 per 133 000 men and 1 per 769 000 women (numbers include all sports-related non-traumatic deaths, so other causes than cardiovascular events were included) (Van Camp et al. 1995). An incidence of 1 sudden death per 33 000 was reported in Italy in older individuals participating in more intense sports. In less intense sports, death rates in adults have been reported in two articles. The first indicated that 1 death occurred per 396 000 person-hours of jogging, or at a rate of 1 death per year for every 7620 joggers; these numbers include known or readily diagnosed coronary artery disease (Thompson et al. 1982). However, when only previously healthy individuals were examined in the same study, death rates were about half – 1 death per 792 000 hours or 15 260 subjects. The second article reported 1 death per 82 000 club members at a large commercial fitness chain, a rate of 1 death per 2.57 million workouts (Thompson et al. 2007). Club members who exercised infrequently or less than once a week composed ∼50% of the exercise-related deaths. In agreement, individuals who exercise regularly have both a lower base-line risk of myocardial infarction and lower relative risk during heavy physical exertion (Mittleman et al. 1993).

Therapeutic index (LD₅₀/ED₅₀). Drugs with wide therapeutic indices are better than ones with narrow indices. No level of exercise is known to be so toxic as to produce 50% mortality. Therefore, no LD₅₀ exists for exercise. Theoretically then the dosage (duration × intensity) of exercise to cause a LD₅₀ must be very high. Therefore, the LD₅₀/ED₅₀ ratio approaches infinity. Any ‘exercise mimetic’ drug must also have an extremely high LD₅₀ and a therapeutic index of near infinity to perform as well as natural exercise. Without an exercise mimetic as therapeutically effective as exercise, it is questionable whether it would be ethical to give an exercise pill to someone capable of being physically active.

Health sciences

In addition to meeting scientific criteria to qualify as an exercise mimetic, any such pharmaceutical should ultimately improve health with a similar efficacy as exercise. Simply put, an exercise mimetic must have, at least, similar ED₅₀ levels to physical activity guidelines to justify cost effectiveness. While the health benefits of exercise are undisputed based on numerous epidemiological and clinical studies, it will be difficult to obtain a single ED₅₀. For each patient, the ED₅₀ of exercise for a specific chronic disease will vary depending on exercise type, age, sex and exercise frequency, intensity and duration (Church & Blair, 2009). Nevertheless in attempting to determine the level of physical activity that an exercise mimetic must meet for substantial health benefits to be achieved, we defer to the Physical Activity Guidelines Advisory Committee Report (Physical Activity Guidelines Advisory Committee, 2008). While the report astutely states that there are no simple guidelines for daily physical activity, they make the following statement as a general overall recommendation for aerobic exercise: ‘However, as a starting place for overall public health benefit, data from a large number of studies evaluating a wide variety of benefits in diverse populations generally support 30 to 60 minutes per day of moderate- to vigorous-intensity physical activity on 5 or more days of the week. For a number of benefits, such as lower risk for all-cause mortality, coronary heart disease, stroke, hypertension, and type 2 diabetes in adults and older adults, lower risk is consistently observed at 2.5 hours per week (equivalent to 30 minutes per day, 5 days per week) of moderate- to vigorous intensity activity’ (Physical Activity Guidelines Advisory Committee, 2008). The Physical Activity Guidelines Advisory Committee Report presents data showing that this amount of aerobic physical activity reduces mortality risk by 30%, cardiovascular diseases by 20–35%, and metabolic diseases by 30–40% in healthy populations. In addition the Committee recommends that progressive muscle strengthening exercises that target all major muscle groups be performed on two or more days per week consisting of 8–12 repetitions of each exercise performed to volitional fatigue in order to improve muscle strength.

Known negative health consequences of drugs that enhance physical performance

Although there are many current drugs that mimic some adaptations to exercise, none have been considered exercise pills. For example, while never being called exercise pills, anabolic steroids and recombinant erythropoietin both mimic and enhance a minimal number of exercise adaptations. However, these drugs contain major side effects. For example, Sjoqvist et al. (2008) list some side effects of anabolic steroids as endocrinological, blood lipid dysfunction, cardiac arrest, premature mortality and neuropsychiatrical. Side effects of erythropoietin are myocardial infarction, cerebrovascular disease and serious thromoembolic events. A true exercise mimetic would not have any side effects that do not occur with exercise (normal side effects of exercise can be muscle soreness, fatigue, etc.), and certainly no major side effects. Carey & Kingwell (2009) contend that an exercise polypill would ‘likely be associated with unwanted effects.’

In summary, an exercise mimetic would not be able to have a side effect that exercise does not have. It is simple logic, with reference to the definition of ‘mimetic’, that an exercise pill cannot increase the risk of any unhealthy condition that exercise itself doesn't cause. Again the ethics of harming the patient must be considered.

Potential negative health consequences from altered behaviour by the public using ‘exercise mimetics’

If individuals believe exercise is a treatment that can be replaced by drugs, harm to the individual will occur in two ways. First, the patient will fail to achieve optimal health. As outlined above, the beneficial effects of exercise are so widespread and diverse (Fig. 2) that current drug approaches directed at mimicking a limited number of exercise pathways only primarily prevent a limited number of chronic diseases. Due to the limited effectiveness such a lifestyle supplement, a false sense of pharmacological health would be created, leading many people to stop exercising altogether. The fact that such a drug would be called an ‘exercise mimetic’ rather than a more appropriate title such as an anti-diabetic drug or anti-obesity drug is purely sensationalizing the particular drug, again leading to a false sense of health security in the general public. Obviously, the goal of biomedical research is to make people healthier; however, continued use of the term ‘exercise pill’ or ‘exercise mimetic’ for drugs incapable of truly mimicking exercise according to the definition of mimetic in the Oxford English Dictionary online may actually lead to a reduction in health of our general population. If this were to occur, we would consider it as unethical.

When drug targets of exercise pathways make sense

We agree that it is a worthwhile goal to find drug targets to a small number of the total exercise adaptations that will keep patients from losing functional capacity in a small number of the total organ systems that decondition when they are unable to exercise. Such pharmaceuticals could be clinically advantageous during recovery from illness or surgery or in patients too frail to engage in exercise. Additionally, pharmaceuticals that attenuate losses of organ or exercise capacities for those incapable of sufficient of amounts of exercise would be beneficial. The current concept of an ‘exercise pill’ illustrates a lack of knowledge about the totality of adaptations to exercise. In summary, implying that a single ‘exercise pill’ will provide all adaptations to exercise when the ‘exercise pill’ only provides a clinically significant gain in physiological capacity for one specific exercise target (out of the multitude of adaptations from actual exercise) is deemed by us as medical malpractice, because the single target pill would not mimic the overwhelming majority of all healthy adaptations to exercise. The ethically correct terminology is ‘partial exercise pill’ to reflect it will improve the clinical condition of patients unable to exercise but the ‘term also reflects that the partial exercise pill’ will not replace all the health benefits obtained with natural exercise.

Summary

Our analysis of projected economic costs, in concert with the diseases whose prevalence is increased by the lack of exercise (Fig. 2), if correct, does not justify the cost-effectiveness of replacing natural exercise in those physically capable of exercising for primary prevention of all chronic diseases. If pills only mimicking a small percentage of all health benefits of a physically active lifestyle, but labelled incorrectly as ‘exercise pills’, were to cause individuals to believe they could get all the benefits of exercise without daily exercise, then the following would likely ensue. The prevalence of chronic health disorders shown in Fig. 2 that the ‘exercise pill’ does not prevent would increase, potentially decreasing quality of life, increasing sick days, increasing disabilities, resulting in more chronic illness, and more mental health issues in those who give up a physically active lifestyle. If all US individuals were physically active to the level recommended to the Secretary of Health and Humans Services (240 minutes of moderate to vigorous exercise per week), then a significant portion of the annual $150 billion in direct health care costs as a result of inactivity could be saved.

Criteria for an ‘exercise pill’ or ‘exercise mimetic’ must:

1. Usage of “mimetic” must be consonant with its definition which is ‘Same or similar’ adaptive effects of exercise training in preventing chronic disorders (Fig. 2) (Oxford English Dictionary, 1989, online).
2. Not produce unhealthy side effects that are not produced by moderate exercise training itself.
3. Be cost effective, thus lowering US health care costs in populations unable to exercise.
4. Decrease mortality rate and have a therapeutic index more favourable than exercise training.
5. Be ethical.

The entirety of molecular, physiological and epidemiological data on the benefits exercise strongly supports the criteria listed above.

We thank Espen Spangenburg, Eva Chin and Grant Simmons for reading parts or all of manuscript and offering important suggestions; Ronald Terjung for offering the title. Grant was written while authors were supported by University of Missouri Research Board and NIH AG18780 (F.W.B.) and Danish National Research Foundation (M.J.L.). We thank Donald Connor and Howard Wilson for assistance in labelling Fig. 2.