Understanding how genes, environment, and personal motivation operate to influence physical activity will require (i) inclusion of properly validated measures of putative mediators (e.g., cultural values, efficacy and control beliefs, goals, intentions, enjoyment, and self-management skills) and moderators (e.g., age or maturation, personality, race/ethnicity, fitness, fatness, skill, and competing behaviors) of physical activity, (ii) a search for candidate genes involved with motivational systems of energy expenditure in addition to energy intake pathways, (iii) assessment of specific features physical activity exposure (i.e., type, intensity, timing, and context), (iv) manipulation of physical activity or prospective observation of change in physical activity at multiple times, rather than cross-sectional association and linkage studies, and (v) use of statistical procedures that permit multilevel modeling (i.e., personal and group-level variables) of direct, indirect (i.e., mediated), and moderated (i.e., interactions of mediators with external factors) relations with physical activity within theoretical gene–environment networks.
The brain controls behavior in widely varying environmental circumstances. This control includes conscious decisions and less conscious drives or urges that are influenced by inherent and acquired traits, each of which are shaped by learning and memories of reinforcement history, by genes and by developmental, social, and environmental factors that alter gene expression. Brain neural circuits that regulate feeding, mood, pleasure, pain, memories about the reward and punishment of behavior, and cognitions such as liking, wanting, and decision making for appetitive behaviors have been elaborated (1,2,3,4). In contrast, little is known about how genes, the environment, and their interactions influence the brain's regulation of physical activity behavior.
Family and twin studies have reported that 30–70% of the variation in human physical activity is inherited (5,6,7,8), and a few association and linkage studies have implicated candidate genes that might explain small, but significant, portions of that variation (9). Features of this small literature suggest that four key issues be addressed during the search for gene–environment interactions that affect physical activity behavior: (i) the need to exploit and elaborate current understanding of putative mediators and moderators of voluntary physical activity, (ii) selection of candidate genes, (iii) measurement of physical activity, and (iv) multilevel statistical modeling of personal, environmental, and genetic influences on physical activity change across time.
Mediators and Moderators of Physical Activity
Although emerging animal and human studies suggest that genetic variation explains a substantial amount of physical activity, most of the variation in physical activity will be ultimately determined by personal choices and environmental opportunities or their interactions. Studies have not yet examined whether genetic variation in physical activity (i) occurs independently of putative mediators and moderators of physical activity, (ii) modifies (i.e., interacts with) the association of physical activity with those factors, or (iii) is explainable by common genetic variation in physical activity and select mediators or moderators (e.g., pleiotropic effects).
About 50 different behavioral correlates of physical activity have been reported among youths and adults (10,11). Putative moderators and mediators of interventions designed to increase physical activity include personal factors that influence decisions to be active, such as self-efficacy, perceived behavioral control, attitude, enjoyment, social norms, and perceptions of social support and access to physical activity settings. Personality appears to be weakly associated with physical activity (12,13), but it plausibly could influence spontaneous physical activity or moderate and help explain gene–environment influences on exercise behavior. In addition, real and perceived features of social and physical environments operate at the levels of families, schools, places of employment, and neighborhoods, which are all located within communities, to influence physical activity. Figure 1 illustrates these theorized relations and interactions.
Social-cognitive variables (i.e., beliefs that are formed by social learning and reinforcement history) are influences on self-initiated change in health behaviors such as physical activity (14). Self-efficacy is a belief in personal capabilities to organize and execute the courses of action required to attain a behavioral goal. Like self-efficacy, perceived behavioral control includes efficacy beliefs about internal factors (e.g., skills, abilities, and self-motivation or willpower) and external factors (e.g., time, opportunity, obstacles, and dependence on other people) that are imposed on behavior. Each construct is distinguishable from outcome expectancy, which is the perceived likelihood that performing a behavior will result in a specific outcome. Although people are more likely to form an intention to behave when they value an expected outcome of the behavior (i.e., they have a positive attitude), that likelihood is increased when a goal is set. People who set goals about being more active and who are dissatisfied with their current activity level will be more likely to adopt physical activity, especially if they possess high self-efficacy about their ability to be physically active (15). Like perceived behavioral control, self-efficacy affects behavior directly and also indirectly by influencing intentions. Efficacy beliefs can affect physical activity both directly by fostering self-management (16) and indirectly by influencing perceptions about sociocultural environments that provide assistance for physical activity which in turn directly influence physical activity (17). Self-efficacy can also moderate the relation between perceptions of social facilitators and physical activity change. Thus, beliefs in personal ability to overcome barriers to physical activity can sustain physical activity despite perceptions of declining social support of physical activity (18). Although beliefs are learned, genes might influence historical adaptations to exercise or success in physical activity settings so that contemporary relations between genes and physical activity might plausibly be explained by an interaction between genes, biological adaptability to exercise, and beliefs. Figure 2 shows a hypothetical gene–physical activity interaction whereby genes and people's exercise goals, or their interaction, might explain exercise adherence directly, or indirectly by influencing adaptability to exercise training.
Cross-sectional studies have found that self-reports and objective measures from geographic information systems of the social and built environments (e.g., neighborhood safety and facility accessibility) are weak correlates of physical activity and overweight among population-based samples. Whether perceived access and actual proximity to physical activity settings have direct relations with physical activity and/or indirect relations moderated or mediated by social-cognitive factors such as social support and efficacy beliefs about overcoming barriers to physical activity has not been determined. Although social networks may link physical activity habits in ways similar to those hypothesized for obesity (19), it is also possible that social affiliations have, to some extent, a genetic basis. Hence, genes might commonly influence features of both social affiliation and physical activity.
Like obesity, physical activity surely represents a complex, multifactorial trait influenced mainly by polygenes that each explain small portions (∼1%) of the variance in physical activity. However, in contrast to the literature on obesity in which at least 22 genes have been supported by at least five studies (20), less than 10 association studies (9) and two linkage studies (21,22) have examined candidate genes for physical activity, without much replication of results.
Most candidate genes for physical activity suggested by linkage or association studies have been selected for study based on understanding of energy intake pathways that influence energy balance (e.g., LEPR, AGRP, MC4R) (9,22) more so than models of otherwise motivated behavior (21,23). However, some genes that have been studied (e.g., DRD2, DRD4, SERT, 5-HT2A, 5-HT2C, Orexin A) might be involved in regulation of motivation systems for both feeding and physical activity. Evidence has been mixed as to whether alleles of SERT and DRD4 genes explain variations in personality (e.g., novelty seeking and extraversion) (e.g., 24,25,26), which plausibly could explain a portion of leisure or spontaneous physical activity.
Neurobiological regulation of physical activity by the central nervous system is poorly understood and has been mainly studied in order to model opioid regulation of anorexigenic running (27) and to understand central fatigue (i.e., a progressive decline in supraspinal motor drive to motor neurons) during exhaustive exercise (28,29) and prolonged, strenuous exercise under conditions of hypoxia, hyperthermia, and hypoglycemia (28,30,31,32). Putative brain mechanisms include increased brain serotonin activity, elevated ammonia levels, brain glycogen depletion, decreased striatal dopamine (DA), and inhibitory feedback from the exercising muscles.
In contrast to the study of factors that limit skeletal muscle control of locomotion under conditions of impairment, the neurobiological regulation of voluntary, nonstrenuous physical activity by healthy animals has received little study. The tight relationship between increased body mass and decreased spontaneous activity levels among rats and mice, irrespective of fat mass, has led to a hypothesis that a reduction in spontaneous activity is a symptom of obesity (33), but how body mass might be sensed to regulate voluntary physical activity is not known.
Reduced DA release or loss of DA receptors in brain appears related to the age-related decline in physical activity observed among many species (34). The meso-limbic (i.e., ventral tegmentum-nucleus accumbens) DA system is a critical component of the forebrain circuitry that regulates activational aspects of motivation (4) (Figure 3). Antagonists of DA and depletion of DA in the accumbens cause rats to reallocate their instrumental behavior away from food-reinforced tasks that have high work requirements and toward the selection of less-effortful types of food seeking (35), implicating DA brain circuitry in energy-related decision making.
Activation of neuronal activity in hypothalamic reward regions during spontaneous locomotion by rats was established 40 years ago (36,37), and electrical self-stimulation of the ventral tegmental area (VTA) has been used to artificially motivate treadmill running (38) and weight lifting (39) in rats. However, little is known about the role of the meso-limbic DA system in the motivation of voluntary physical activity such as wheel running or other types of spontaneous physical activity in home cages or novel open fields. Treadmill running acutely increases DA release (40) and turnover (41) and chronically upregulates D2 receptors (42) in the striatum of rats, but forced treadmill running by rats and mice likely confounds exertion with emotional stress and is thus a poor model of voluntary physical activity.
Rats and mice have been selectively bred for running (43,44), but genes that might help explain motivated running or other spontaneous physical activity in rodents have yet to be identified (45). Nonetheless, early gene expression has been implicated in voluntary wheel running. A recent study found that c-fos and delta fosB in the nucleus accumbens were activated during wheel running in rats, and mice that overexpress delta fosB selectively in striatal dynorphin-containing neurons increased their daily running compared with control litter mates (46). Delta fosB could plausibly facilitate wheel running by inhibiting the release of dynorphin by GABA neurons, which otherwise binds with kappa opioid receptors to inhibit DA release in the VTA or accumbens (46). The hypothalamic neuropeptide orexin A stimulates both feeding and spontaneous physical activity in rats when it is injected into the lateral hypothalamus (47,48). Neurons that contain orexin A project from caudal hypothalamic areas throughout the neuroaxis and appear to enhance spontaneous physical activity, in part by inhibiting hypothalamic GABA release and stimulating or synergizing with DA signaling in the striatum (49) or by possibly modulating GABAA inhibition of the accumbens (50). Orexin-containing neurons in the VTA appear directly involved in opioid-dependent appetitive behavior and increased locomotion by activation of the meso-limbic dopamine pathway between the VTA and the accumbens (51), possibly by potentiation of glutamate-N-methyl-D-aspartate (NMDA) receptor–mediated neurotransmission (52). Chronic activity wheel running appears to downregulate the striatal GABAA response after open-field locomotion (53) and has a neurotrophic effect in the VTA (54). Wheel and treadmill running (55,56) also increase gene expression for brain galanin which helps regulate feeding and may also play a role in modulation of the dopaminergic meso-limbic system. Galanin expression is not sensitive to other forms of stress, which can permanently upregulate expression of some HPA axis genes (57), but possible epigenetic effects of physical activity have not received much study.
Measurement of Physical Activity
Physical activity is bodily movement produced by contraction of skeletal muscles that results in varying amounts and rates of energy expenditure that are positively related to physical fitness and health depending upon the stimulus features of physical activity such as its type, intensity, regularity and timing. Among humans, physical activity can be spontaneous and sporadic (e.g., leisure play or fidgeting) or purposeful and repetitive (e.g., jobs, chores, organized sports, or exercise performed with the goal of improving or maintaining physical fitness, function or health).
Most studies have used self-reports of physical activity. Fewer, more recent studies have used standardized observational systems or objective monitoring by accelerometry (e.g., 21). Although some self-report measures have been validated by demonstrating correlations with objective measures, those correlations are typically modest, accounting for <30% shared variance. Self-report measures of physical activity used in observational, population-based association studies generally were designed to categorize or rank people according to frequency of physical activity or total energy expenditure estimated crudely by metabolic equivalents (i.e., MET·h) rather than to assess specific features of physical activity such as type, intensity, and timing that are important for understanding the direction, effort, and persistence of human motivation. Self-report measures of physical activity used in studies of human gene cohorts have been too crude to distinguish spontaneous physical activity (e.g., fidgeting or pacing about) from other leisure or purposeful physical activities. Likewise, commonly used objective measures of physical activity or energy expenditure do not assess the type or setting of physical activity (e.g., accelerometry) or any specific features of physical activity (e.g., energy expenditure, using doubly labeled water technique). Studies of physical activity should compare results using both subjective and objective measures to enhance convergence of methods and exploit their unique assessment features.
Similarly, physical activity studies of mice and rats have used various types of physical activity that are uncorrelated and have widely different behavioral interpretations depending upon the setting. For example, voluntary wheel running seems to represent a unique motivational system in rodents (58). Mice selectively bred for higher velocity wheel running (42), and rats selectively bred for higher treadmill running endurance (59) each have high aerobic capacities, but the treadmill running trait is unrelated to voluntary wheel running (60) and neither running trait is associated with other types of spontaneous physical activity such as locomotion in the home-cage or in a novel open field (61,62,63). Moreover, commonly used measures of spontaneous physical activity by rodents (e.g., home-cage locomotion) used to mimic human nonexercise energy expenditure (e.g., reference (64)) do not seem well suited for understanding human motivation for either leisure play or purposeful physical activities of daily living. Spontaneous physical activity in cages or in an open field can be selectively bred as a trait (65) and can model fear, anxiety, agitation, depression, aggression, social dominance, or exploration, depending upon the social and environmental conditions of observation.
Multilevel Models of Physical Activity Change
Conceptual models of physical activity should include variables measured at the level of the person, including family and home environment, but also measured at the level of community catchment areas (e.g., neighborhoods, churches, work sites, schools) (Figure 4). Complex models are needed to describe the independent and interactive contributions of key variables at each level to change physical activity across multiple points in time. Such models require analytical methods that address inherently complex features, such as (i) the multilevel nature of the data, (ii) the analysis of change across time, (iii) hypothesis tests of independent (i.e., direct), mediated (i.e., indirect), and moderated (i.e., interactive) relations, (iv) the use of different data forms including self-ratings, as well as objective measures of the physical and social environment, and (v) the need to demonstrate equivalence of the measurement properties of several person-level variables between age, race, and gender groups and within each of those groups across time. Common statistical techniques such as ANOVA and ordinary regression cannot fully address these complexities. Advanced techniques such as structural equation modeling and latent growth modeling provide precision for multilevel, theoretically derived analysis of mediated change (66) in physical activity, as illustrated by Figure 5.
Understanding how genes, the environment, and personal motivation operate to influence physical activity will be hindered by a search for candidate genes predominantly restricted to energy intake pathways, continued use of cross-sectional, univariate association studies, failure to include measures of personality or putative social-cognitive mediators and moderators of physical activity, use of physical activity measures that are poorly validated or improperly applied according to context, and by scarce use of statistical procedures that permit simultaneous estimates of direct and indirect relations/effects between genes, motivation, the environment, and physical activity. Future research should:
•Search for candidate genes involved with brain modulation of motivated energy expenditure in addition to those involved with energy intake pathways of energy balance.
•Use prospective cohort studies and randomized controlled trials to examine gene interactions with motivational traits or social-cognitive mediators of physical activity.
•Select or develop measures of spontaneous physical activity in animals that, in addition to assessing energy expenditure, are appropriate for modeling human motivation for exercise or leisure physical activity. Select or develop measures of physical activity for use in studies of human gene cohorts that distinguish motivated leisure and purposeful physical activity from other types of spontaneous physical activity that are not motivated by instrumental goals.
•Use multilevel (i.e., personal and group-level variables) modeling of direct, indirect (i.e., mediated), and moderated (i.e., interactions of mediators with external factors) relations of physical activity within theoretical gene–environment networks that include putative mediators and moderators of change in physical activity across multiple time periods. Discrete categories of physical activity can be clinically meaningful (e.g., program adherence/dropout or meeting/not meeting participation guidelines for public health). However, three or more measurement periods are needed to examine change, while assessing interindividual variation in initial status (i.e., baseline) and inter- and intraindividual variation in change (67).
The author has declared no financial interest.
This publication was sponsored by the National Cancer Institute (NCI) to present the talks from the “Gene-Nutrition and Gene-Physical Activity Interactions in the Etiology of Obesity” workshop held on September 24–25, 2007. The opinions or assertions contained herein are the views of the authors and are not to be considered as official or reflecting the views of the National Institutes of Health.