The effects of volume versus intensity of long-term voluntary exercise on physiology and behavior in C57/Bl6 mice

Cardiovascular exercise (CVE) is associated with healthy aging and reduced risk of disease in humans, with similar benefits seen in animals. Most rodent studies, however, have used shorter intervention periods of a few weeks to a few months, begging questions as to the effects of longer-term, or even life-long, exercise. Additionally, most animal studies have utilized a single exercise treatment group - usually unlimited running wheel access - resulting in large volumes of exercise that are not clinically relevant. It is therefore incumbent to determine the physiological and cognitive/behavioral effects of a range of exercise intensities and volumes over a long-term period that model a lifelong commitment to CVE. In the current study, C57/Bl6 mice remained sedentary or were allowed either 1, 3, or 12 h of access to a running wheel per day, 5 days/weeks, beginning at 3.5-4 months of age. Following an eight-month intervention period, animals underwent a battery of behavioral testing, then euthanized and blood and tissue were collected. Longer access to a running wheel resulted in greater volume and higher running speed, but more breaks in running. All exercise groups showed similarly reduced body weight, increased muscle mass, improved motor function on the rotarod, and reduced anxiety in the open field. While all exercise groups showed increased food intake, this was greatest in the 12 h group but did not differ between 1 h and 3 h mice. While exercise dose-dependently increased working memory performance in the y-maze, the 1 h and 12 h groups showed the largest changes in the mass of many organs, as well as alterations in several behaviors including social interaction, novel object recognition, and Barnes maze performance. These findings suggest that long-term exercise has widespread effects on physiology, behavior, and cognition, which vary by "dose" and measure, and that even relatively small amounts of daily exercise can provide benefits.

sectional studies found that individuals who exercised performed better than non-exercisers across a number of cognitive domains, including response time, reasoning, memory, vigilance, fluid intelligence and exhibited better performance on the Stroop and Trail-making tasks [10][11][12][13][14]. Results from longitudinal studies are also positive, linking mid-life physical activity to reduced risk of cognitive decline and dementia in later life [15,16]. In a study of healthy older adults, three days/week aerobic exercise increased memory performance compared to a stretching control group [8]. In a randomized controlled trial, 24 months of moderate intensity exercise had no effect on cognitive performance in previously sedentary older adults [17], suggesting that earlier, longer, or even life-long exercise interventions may be necessary to be effective. Importantly, these studies suggest that CVE may lead to physiological changes that alter the course of cognitive decline.
Studies in rodents have yielded similarly positive results, such that CVE is capable of improving learning and memory across various ages [18][19][20][21][22]. These studies typically employed either voluntary exercise interventions on running wheels or forced exercise on treadmills; however, a majority of these studies provided unlimited wheel access, resulting in large volumes of exercise not comparable to the human clinical scenario. While the methods outlined in these studies vary in the exercise-duration parameters, few systematically compared the effects of different amounts of exercise. In fact, exercise was typically presented as a treatment that was compared only with a sedentary control condition and did not assume that the intervention may have potential dose-response properties.
There are two important considerations then for both human health and the modeling of exercise in rodents. The first is to determine how different amounts of CVE affect physiological and behavioral measures. It may be that more is better, or alternatively, that an optimal level exists. Some evidence can be found in epidemiological studies showing that greater physical activity levels are linked to enhanced cognitive performance in healthy older women [23]. A meta-analysis found that exercise session durations of longer length (30-60 min) were more effective than shorter durations (15-30 min) in improving cognitive function [5]. In contrast, some studies suggest that only small doses of exercise are necessary to see significant health benefits [24,25], and there is evidence that shorter duration high-intensity regimens are equally beneficial to longer-duration/lower-intensity regimens in regard to cardiovascular and metabolic function and musculoskeletal benefits [26][27][28][29]. Assessment of exercise regimens of varying intensity in rodents has generally been performed using forced exercise (treadmill running) [30][31][32], which is a known stressor and argued to be more stressful than voluntary wheel running [33][34][35][36]. Therefore, it is of interest to create voluntary exercise regimens that naturally vary in volume and intensity in animal models.
The present study assessed the "dose-dependent" effects of exercise (one hour, three hours, or 12 h of daily running wheel access) compared to sedentary controls, providing the opportunity to assess how different exercise patterning affected behavior and physiological outcomes. We hypothesized that providing mice with differing lengths of access to a running wheel might result in differences in exercise volume (wheel rotations), as well as measures that could be indicative of higher running "quality"/"intensity". Animals with shorter access to a running wheel may run faster and/or take fewer breaks in order to maximize running output per available unit time; conversely, mice with longer access periods may achieve more cumulative running but take more frequent or longer rest breaks. These different patterns of exercise structure could presumably result in very different physiological adaptations and behavioral consequences. The second important consideration not represented in prior rodent studies is the effects of a lifetime of CVE. Virtually all previously mentioned exercise studies in rodents have had interventions lasting four months or less, while some clinical findings suggest that longer, or even life-long, exercise interventions may be necessary to be effective [17]. A meta-analysis also found that intervention length was a factor in the efficacy of exercise to exert beneficial effects [5]. Therefore, the current study utilized an eight-month intervention, assessing the impact of longer-term treatment. In sum, this study represents the first long-term intervention study examining different daily levels of exercise to model a life span of regular aerobic exercise in a human on physiological response measures and a comprehensive behavioral battery assessing motor function, temperament and cognition.

Animals
Forty C57/Bl6 mice were used in this experiment, split equally between males and females (Taconic Biosciences, Rensselear, NY) and housed individually for the duration of the experiment in a controlled room (22 ± 2°C and 40-60% humidity) with a 12 h reverse light-dark cycle (lights off 0800 h). C57/Bl6 mice were chosen as they are a common background strain of many transgenic mouse models of human disorders. Additionally, studies comparing running and its effects in several mouse strains found that C57/Bl6 mice run moderate amounts compared to other strains (~4 km/day) when given unlimited wheel access, which results in increased hippocampal neurogenesis and dentate gyrus volume, as well as improved cognitive performance [37].
Mice were allowed to habituate to a new housing facility and single housing for at least one week prior to the start of the experiment, referred to as the "pre-intervention" period. Baseline food intake and body weight measures were taken at the end of this period. Mice were then split into four experimental groups (n = 10/group): sedentary (Sed), one hour of wheel access/day (1 h), three hours of wheel access/ day (3 h), and twelve hours of wheel access/day (12 h). Purina Lab Diet chow was available ad libitum for the entire experiment, and body weight and food intake were recorded weekly throughout the entire experiment. All experiments were conducted in conformity with the National Academy of Sciences Guide for Care and Use of Laboratory Animals and approved by the Stony Brook University Institutional Animal Care and Use Committee.

Voluntary exercise intervention
A timeline of the experiment is shown in Fig. 1. Exercise intervention began at 3.5-4 months of age. The intervention lasted for approximately 8 months, followed by behavioral testing, and mice being euthanized at~12-13 months of age. Throughout the intervention period, mice were given access to a running wheel for five days per week during the dark cycle. Exercise mice were placed in 43 × 21 × 21 cm cages equipped with metal running wheels for the appropriate length of daily access (1 h, 3 h, or 12 h). The running wheel measured 6.5 in. in diameter with a 3 in. wide running platform. Number of rotations were recorded through a software system that responded to the closing of a switch produced by two magnets lining up one magnet located directly on the wheel, and another on the outside of the cage. Sedentary animals remained in their home cages during this time, as has been done in previous studies examining the effects of exercise on brain and behavior [38]. Voluntary wheel running was chosen over forced exercise, such as a treadmill running regimen, so that mice could choose when and how much to run. This minimized potential confounding variables introduced by forced exercise, such as stress [33][34][35][36] and allowed for the correlation between running and other measures.
Running behavior was recorded during each session to determine total daily exercise, as well as investigate patterns of activity throughout sessions. The number of revolutions performed in each minute of wheel access was recorded. Running behavior was investigated to assess differences in volume and intensity between exercise groups and determine their possible role in affecting physiology and behavior. Average number of total revolutions per session was calculated to determine running volume. Average rotations/min (for binned minute with > 5 rotations) and breaks/h (number of minute bins with < 5 rotations) were calculated as measures of running intensity.

Behavioral assessment battery
All mice underwent a battery of behavioral testing following the eight-month intervention period, including rotarod, open field, social interaction, marble burying, novel object recognition, Y-maze for spontaneous alternation, and Barnes maze. Exercise interventions were not continued throughout the behavioral testing period, with testing beginning four days after the last exercise session. This allowed for the investigation of the effects of chronic (rather than acute) exercise. All behavioral testing occurred during the animals' dark cycle. Tests were performed in the following order, with no mouse receiving more than one behavioral assay in the same day: Barnes Maze (Days 1-5), open field (Day 6 or 7), novel object recognition (Day 8, 9, or 10), social interaction (Day 11, 12, or 13), rotarod (Day 14 or 15), Y-maze (Day 16), marble burying (Day 17).

Rotarod
Rotarod was performed to assess balance, strength, and motor coordination. The apparatus (model ENV-575 M; MED Associates Inc.) is composed of a 30 cm long rod that is divided into five, equally sized 6 cm sections. Mice were placed on the rotarod, which spun at an increasing speed of up to 40 rpm over a five-minute period. The time spent on the rod until the mouse fell (maximum time of five minutes) was recorded. Mice were tested three times, with a minimum fiveminute inter-trial interval, and the average of the best two trials was used for analysis.

Open field
Mice were placed in a square 60 cm × 60 cm open field arena for 10 min. Behavior was recorded using ANY-maze software. Locomotor behavior was investigated with the measure of distance traveled, used to assess general activity levels and determine differences in motor function. Since mice tend to avoid open spaces and prefer exploring close to the walls of the open field apparatus (i.e. thigmotaxis), anxietylike behavior was assessed by time spent in the center of the open field as in previous studies [39]. Measures of general activity and anxietylike behavior are also useful in interpretation of any differences in performance observed in the Barnes maze.

Social interaction
Crawley's three chamber paradigm was adapted to assess sociability [40]. The apparatus used is a rectangular box measuring 46 cm × 21 cm, which consisted of three connecting compartments made of clear plexiglass walls. The side chambers measured 16 cm × 21 cm, and the middle chamber measured 14 × 21 cm. This test consisted of two five-minute trials that occurred consecutively, with the first trial serving as habituation to the chamber. In the second trial, the middle chamber was empty, one side chamber contained an empty cup, and the other side chamber housed another mouse in an identical cup (Stranger 1). Location of Stranger 1 (left or right side) was randomly assigned. Sociability was assessed by comparing time spent with Stranger 1 compared to the empty cup. Stranger mice were previously habituated to being confined in the cup within the arena to reduce distress during experimental trials.

Marble burying
Experimental procedures were adapted from a well-defined protocol [41]. Mice were placed in a rat-sized tub cage filled with 5 cm of corn cob bedding and 20 marbles in a 5 × 4 array for five minutes, during which time the latency to dig, number of digging episodes, and time spent digging were recorded. Digging was defined as coordinated movements of fore or hind limbs that displaced the bedding.

Novel object recognition task
A novel object recognition task was performed to assess object recognition memory. This task consisted of two trials, each lasting five minutes, with an inter-trial interval of 15 min. In the first trial, two of the same objects were placed in the open field arena, one in the front left quadrant, and one in the back right quadrant. In the second trial, one of the objects was replaced by a novel object, while the other object remained the same. Novel object recognition was assessed by the time spent with the novel object compared to the object used for the previous trial (familiar object).

Y-maze for spontaneous alternation
Mice were placed in a Y-shaped maze consisting of three arms. This maze was adapted from the unreinforced radial arm maze by blocking off five of the eight arms. Animals were allowed to freely explore the arms for three minutes, and number of arm entries was recorded. Order of arm entries were also manually assessed, and percent alternation was calculated [# alternations/(# arm entries -2) * 100]. Chance performance for continued alternation is 22.2%. In healthy exploration, mice should alternate traversing down arms in a circular pattern rather than repeatedly going down the same arms, based on the natural tendency of mice to explore novel environments. One alternation consisted of a mouse going down each of the three arms before returning to a previously-visited arm. Performance on this task is interpreted as a measure of spatial working memory [42].

Barnes circular maze
The Barnes maze was originally developed to test learning and memory in rats (Barnes, 1979). We used an adaptation of this maze, a circular wooden platform, 91 cm in diameter, elevated 75 cm off the ground. The platform has eight equally spaced escape holes along the periphery that are 24.5 cm apart. Under each hole, a shelf securely held an escape box, measuring 10 cm × 8.5 cm × 4 cm. There were visible distal cues placed around the room, which remained constant throughout the duration of testing. Testing was performed on five consecutive days, with two trials per day separated by a 15 min intertrial interval. Mice were placed onto the center of the maze at the beginning of each trial, then allowed to explore until the escape box was found and entered, or a maximum of five minutes. If the escape box was entered, the mouse remained there for one minute before being transferred back to its home cage. If escape box was not entered within five minutes, the mouse was placed in the escape box and left there for one minute. During each trial, the following measures were recorded: latency to find (amount of time taken to find escape box hole), hole entry rate (average time between subsequent hole entries as a measure of exploration speed), errors (number of nose-pokes into a hole that did not contain the escape box), and re-entry errors (number of nose-pokes into a previously visited hole that did not contain the escape box).

Blood, organ, and muscle collection
Approximately one week following the completion of behavioral testing (~3-4 weeks after the completion of exercise), mice were euthanized between 1100 h and 1700 h under deep anesthesia with 2.5% avertin. Cardiac puncture was performed to collect blood, which was allowed to clot at room temperature for 30 min, spun at 2000 ×g for 10 min, and serum was collected and stored at -80°C until used in assays. Wet organ weights were taken of the brain, heart, lungs, kidneys, liver, spleen, pancreas, and adrenal glands, which were also normalized by body weight. Additionally, quadriceps, gastrocnemius, and soleus muscles were collected to assess exercise-induced differences in muscle mass.

Citrate synthase activity in muscle
Muscles collected during euthanasia were assayed for citrate synthase activity, which has been used as a marker of aerobic capacity and mitochondrial density in skeletal muscle [43]. Muscle samples were added to 20 mL of CelLytic MT mammalian tissue lysis/extraction reagent (Catalog Number C3228, Sigma Aldrich) per gram of muscle tissue and homogenized using Zirconium Oxide beads and a bullet blender at 4°C. Samples were spun at 16,000 ×g for 15 min at 4°C, and supernatant was collected and stored at -80°C until assayed for citrate synthase activity using a commercial kit (Catalog Number CS0720, Sigma Aldrich) according to the manufacturer's instructions. Citrate synthase activity was normalized by protein content, as done previously [44].

Enzyme linked immunosorbent assay (ELISA) for serum corticosterone
Serum samples were analyzed using a commercially available ELISA for corticosterone according to the manufacturer's instructions (Cayman Chemical). Absorbance was recorded using a plate reader (Spectramax).

Statistical analyses
One way ANOVAs were performed to determine differences between treatment groups on running parameters, body weight gain, overall food intake, organ and muscle mass, serum corticosterone concentration, and behavioral measures. Two-way repeated measures ANOVAs were performed to assess differences in treatment groups over time for running parameters, and food intake and body weight. Two way repeated measures ANOVAs were also performed to assess differences in exploration of objects (during the novel object recognition task) or cups (during social interaction) between treatment groups. When appropriate, Pearson correlations were performed to assess relationships between running parameters and behavioral and physiological measures. Analyses were performed using Statistica and SigmaPlot/Stat, and significance was set at alpha p < .05.

Results
Mice were subjected to 8 months of either sedentary or exercise conditions, with exercise groups being given either 1, 3, or 12 h of access to a running wheel (based on treatment group) 5 days per week. Mice then underwent an extensive battery of behavioral testing to characterize the effects of exercise on motor function, temperament, and cognition. Following behavior testing, mice were euthanized to collect blood, organs, and muscles for assessment of physiological outcomes. The current study included both males and females; however, separating analyses by sex results in an abundant loss of power. Because sex differences have been reported previously in voluntary running patterns and behavioral responses to running [e.g. [45,46]], symbols were added to figures to allow preliminary evaluation of these presently.

Rotations
The number of rotations performed by each mouse was recorded with computer software in one-minute bins during each exercise session. The total number of rotations performed was summed during each session as a measure of running volume. A two way repeated measures ANOVA was performed to determine the effect of treatment on average rotations performed per exercise session during each month of the intervention ( Fig. 2A). There was a significant main effect of treatment [F (2, 27) = 79.580, p < .001], with post-hoc analyses showing that mice with longer access to a running wheel performed a greater number of rotations (12 h > 3 h > 1 h; p < .001 for all). The main effect of time was significant [F(7,189) = 32.264, p < .001], with a trend of mice running less as time went on during the experiment. The treatment × time interaction was also significant [F(14,189) = 10.129, p < .001]. Post-hoc analyses found that as hypothesized, increased length of access to a running wheel results in greater running volume (12 h > 3 h > 1 h at all time points; p < .05 for all). In 1 h exercise mice, there was no significant variation in running throughout the course of the experiment; however, in 3 h and 12 h exercise mice, running dropped off in later months compared to early months of intervention.

Speed (rotations/min)
Running speed (rotations/min) was calculated by averaging the number of rotations performed in one-minute bins that were not counted as a "break", essentially averaging the number of rotations performed per minute when the number of rotations performed in that bin was greater than five. A two way repeated measures ANOVA was performed to determine the effect of treatment on running speed (rotations/min) during each month of the intervention (Fig. 2B). There was a significant main effect of treatment [F(2, 27) = 8.020, p = .002], with post-hoc analyses showing that 3 h and 12 h mice ran faster than 1 h mice (p < .001 for both). These results suggest that mice with longer access to a running wheel actually run faster, possibly due to increased fitness levels, rather than shorter access mice running faster to maximize exercise output. The main effect of time was also significant [F(7,189) = 41.747, p < .001], with a trend of mice running slower (reduced rotations/min) as time went on during the experiment.

Breaks/h
Breaks/h was calculated to determine the number of one-minute bins per hour during which the mouse was not running. A "break" was a one-minute bin in which the animal performed less than five wheel rotations. A two way repeated measures ANOVA was performed to determine the effect of treatment on breaks/h during each month of the intervention (Fig. 2C). There was a significant main effect of treatment [F(2, 27) = 29.114, p < .001], with post-hoc analyses showing that mice with shorter access to a running wheel took fewer breaks (12 h > 3 h > 1 h; p < .05 for all), supporting the hypothesis that mice with shorter access to a running wheel may take fewer breaks to maximize exercise output. The main effect of time was also significant [F(7,189) = 33.519, p < .001], with a trend of mice taking more breaks/h as time went on during the experiment. The treatment × time interaction was not significant [F(14,189) = 0.779, p = .691].

Hourly analysis of running
Running parameters as discussed above were averaged over the entire daily access period (1 h, 3 h, or 12 h). It is possible that running patterns varied over the course of the day, particularly in the first hour or few hours of extended wheel access. Therefore, additional analyses The number of rotations performed by each mouse was recorded with computer software in oneminute bins during each exercise session. The total number of rotations performed was summed during each session as a measure of running volume. At all time points, there was a dose-dependent effect (12 h > 3 h > 1 h), and mice ran less as time went on during the experiment. (B) Running speed (rotations/min) was calculated by averaging the number of rotations performed in one-minute bins that were not counted as a "break", essentially averaging the number of rotations performed per minute when the number of rotations performed in that bin was greater than five. Overall, 3 h and 12 h mice ran faster than 1 h mice, and running speed declined in all groups over the course of the experiment. (C) Breaks/h was calculated to determine the number of one minute bins per hour during which the mouse was not running. A "break" was a one-minute bin in which the animal performed less than five wheel rotations. Overall, there was a dose-dependent effect (12 h > 3 h > 1 h), and mice took more breaks as time went on during the experiment. (D) Mean rotations performed throughout the daily wheel access period, averaged over the entire exercise intervention period. An analysis of only the first hour of running found that 12 h mice performed fewer rotations than 1 h and 3 h mice during that period (p < .05 for both). (E) Mean running speed throughout the daily wheel access period, averaged over entire exercise intervention period. An analysis of only the first hour of running found no significant difference in running speed during that time period, regardless of exercise group (p > .05). (F) Mean breaks/h taken throughout the daily wheel access period, averaged over entire exercise intervention period. An analysis of only the first hour of running found that mice with increased length of access to a running wheel took more breaks (12 h > 3 h > 1 h) during that time period. @p < .05 versus 1 h; %p < .05 versus 3 h; #p < .05 versus 12 h. In D-F: arrow = male average, plus sign = female average. were run to assess running patterns throughout the duration of daily wheel access. Hourly rotations (Fig. 2D), speed (Fig. 2E), and breaks/h ( Fig. 2F) were averaged across the 8 month exercise intervention period.
Mice in the 1 h group performed on average 1241 rotations in the hour of wheel access, while the 3 h and 12 h groups hourly ranges were 1226-1319 and 575-1100 rotations, respectively, suggesting consistent running in the 3 h group but variable running throughout the day in the 12 h group. Mice in the 1 h group ran at a speed of 25.3 rotations/min, while the 3 h and 12 h groups hourly ranges were 28.9-32.1 and 27.0-32.2 rotations/min, respectively, suggesting that 3 h and 12 h mice consistently ran faster than 1 h mice. Mice in the 1 h group took 14.0 breaks in that hour on average, while the 3 h and 12 h groups hourly ranges were 19.1-22.6 and 27.4-40.9 breaks/h, respectively, suggesting a consistent dose-dependent effect of exercise for breaks taken and considerable hourly variability in the 12 h group.
One way ANOVAs were performed to compare running parameters during the first hour of wheel access for all three exercise groups.

Body weight and food intake
A two way repeated measures ANOVA was performed to determine the effect of exercise on body weight over the course of the experiment (Fig. 3A). The main effect of treatment was not significant [F(3, 36) = 1.786, p = .167]. The main effect of time was significant [F (8,288) = 194.527, p < .001], and post-hoc analyses found that weight increased during each week of treatment (p < .05 for all), except between weeks 4 and 5 (p > .05), which likely reflects normal weight gain during the aging process. The treatment × time interaction was significant as well [F(24,288) = 3.956, p < .001]. Post-hoc analyses found no baseline differences between groups during the pre-intervention period, or in early treatment months (months 1-5; p > .05 for all). All exercise groups showed similarly reduced body weight despite sizable differences in running volume and quality. Additionally, these reductions in body weight took several months to become apparent (months 6-8; p < .05 for all). Similarly, a one way ANOVA found a significant main effect of treatment [F(3, 36) = 6.72, p < .001] for weight gain over the course of the experiment (Fig. 3B), such that exercise, regardless of length of access to the running wheel, similarly attenuated weight gain compared to sedentary mice (p < .05 for all).
Food intake was normalized to body weight, resulting in the measure of grams of food intake per kilogram of body weight (g/kg). A two way ANOVA was performed to assess the effect of exercise on daily food intake over the course of the experiment (Fig. 3C). The main effect of treatment was significant [F(3, 36) = 12.576, p < .001], and post-hoc analyses found that there was a dose-dependent increase in food intake with increasing access to a running wheel (p < .05 for all), except that 1 h and 3 h mice ate similar amounts (12 h > 3 h = 1 h > sedentary). The main effect of time was also significant [F(8,277) = 89.672, p < .001], and post-hoc analyses found several significant differences between time points during the experiment. Food intake increased during the first two months of intervention compared to the pre-intervention period (p < .001 for both). Food intake then decreased in month 3 (p < .05) and remained fairly stable between months 3-7 (p > .05 for all), then dropped again during month 8 of intervention (p < .05). The treatment × time interaction was significant as well [F (24,277) = 4.693, p < .001]. Post-hoc analyses found no baseline differences between groups during the pre-intervention period (p > .05 for all). Mice in all exercise groups ate more than sedentary mice during most months of intervention (p < .05 for all months except month 1 p = .092). Mice in the 12 h exercise group ate > 1 h mice in intervention months 1-3, 5-6, and > 3 h mice during all months of exercise intervention (p < .05 for all). Mice in the 1 h and 3 h groups ate similarly throughout the entire intervention period (p > .05 for all) Fig. 3. Mean (+SEM) body weight and food intake over the course of the exercise intervention period. (A) There were no baseline differences in body weight between groups during the pre-intervention period, or in treatment months 1-5. Mice in all exercise groups weighed less than sedentary mice in intervention months 6-8. (B) Exercise, regardless of length of access to the running wheel, similarly attenuated weight gain compared to sedentary mice. (C) Food intake was normalized to body weight, resulting in the measure of grams of food intake per kilogram of body weight (g/kg). Overall, there was a dose-dependent increase in food intake with increasing access to a running wheel, except that 1 h and 3 h mice ate similar amounts (12 h > 3 h = 1 h > sedentary). There were no baseline differences between groups during the preintervention period. Mice in all exercise groups ate more than sedentary mice in intervention months 2-8, and this trend was also apparent in month 1. Mice in the 12 h exercise group ate > 1 h mice in intervention months 1-3, 5-6, and > 3 h mice during all months of exercise intervention. Mice in the 1 h and 3 h groups ate similarly throughout the entire intervention period. (D) Food intake was averaged over the course of the intervention period, and normalized to body weight, resulting in the measure of grams of food intake per kilogram of body weight (g/kg). All exercise groups showed increased food intake, with 1 h and 3 h mice eating similarly, and 12 h mice eating the most. *p < . 05  Since there appeared to be dose-dependent effects of exercise on body weight and food intake, correlations were run between these measures and running parameters. While there was no relationship between body weight or food intake with running speed or breaks per hour, there were significant correlations for number of rotations performed with both % body weight change [r(38) = −0.446, p = .004] and food intake averaged over the experiment [r(38) = 0.686, p < .001]. These results suggest that greater running volume (but not intensity) was associated with reduced body weight, as well as increased food intake, likely in an attempt to compensate for energy deficits under exercise conditions.

Organ and muscle weights
For each organ and muscle collected, a separate one way ANOVA was performed to assess the effect of treatment on raw and normalized (% of body mass) weights of the brain, heart, lungs, liver, kidneys, spleen, pancreas, adrenal glands, soleus, gastrocnemius, and quadricep. Mean and SEM for raw and normalized organ and muscle weights can be seen in Table 1.
A one way ANOVA found a significant main effect of treatment for raw brain mass [F(3, 35) = 3.369, p = .029], such that 1 h mice had larger brains than sedentary and 3 h mice (p < .05 for both). A one way ANOVA found that treatment did not have a significant effect on normalized brain mass [F(3, 35) = 2.154, p = .111].
Although a one way ANOVA found no significant effect of treatment on raw heart mass [F(3, 35) = 1.213, p = .319], a one way ANOVA did find a significant main effect of treatment on normalized heart mass [F (3, 35) = 4.634, p = .008]. Post-hoc analyses found that 1 h and 12 h mice had bigger hearts (normalized to body weight) compared to sedentary and 3 h mice (p < .05 for all).
One way ANOVAs found that treatment did not have a significant main effect on raw [F (3,36)  Although a one way ANOVA found no significant effect of treatment on raw soleus mass [F(3, 33) = 1.128, p = .352], a one way ANOVA did find a significant main effect of treatment on normalized soleus mass [F (3, 33) = 4.082, p = .014]. Post-hoc analyses found that 3 h and 12 h mice had bigger soleus (normalized to body weight) compared to sedentary mice (p < .05 for all). Table 1 Mean and SEM of raw (g) and normalized (% of body weight) organ and muscle weights for each treatment group.  (Fig. 4A). Post-hoc analyses found that citrate synthase activity was higher in the gastrocnemius of 1 h mice compared to sedentary mice (p = .032). Additionally, the increase in citrate synthase activity in 3 h (p = .052) and 12 h (p = .077) versus sedentary mice approached significance. These results suggest that exercise similarly increased citrate synthase in gastrocnemius muscle across exercise conditions despite differences in running volume and intensity measures. Collapsing the three exercise groups into a single group and comparing it to the sedentary group using a t-test resulted in a significant effect, t(23) = 2.450, p = .022, with exercise mice having greater citrate synthase activity in the gastrocnemius compared to sedentary mice (Fig. 4B).

ELISA for serum corticosterone concentration
A one way ANOVA found that treatment had no significant effect on resting serum corticosterone levels [F(3, 36) = 1.125, p = .352] (Fig. 5A). Since there appeared to be a trend of all exercise groups showing reduced corticosterone levels compared to the sedentary group, the three exercise groups were collapsed and a t-test was used to compare this collapsed exercise group to the sedentary group (Fig. 5B). This difference approached significance [t(38) = 1.739, p = .09], such that there was a trend of exercise mice having reduced corticosterone compared to sedentary mice. Additionally, since this trend of exercise reducing corticosterone levels appeared that it could be dose-dependent, a correlation between corticosterone levels and running parameters were performed. Although the relationship between rotations performed and corticosterone was indeed negative, it was not significant [r(38) = −0.171, p = .292]. Corticosterone's relationships with running speed and breaks/h were also not significant (p > .05). Taken together, these findings suggest that exercise may marginally reduce serum corticosterone levels, with little variation due to differences in volume or intensity.

Rotarod
A one way ANOVA found a significant main effect of treatment on rotarod performance, as measured by amount of time on rod [F(3, 36) = 3.379, p = .029] (Fig. 6A). Post-hoc analyses found that all exercise groups spent a greater amount of time on the rotarod compared to the sedentary group (p < .05 for all). These findings suggest that exercise improves motor function, and that this effect does not appear to be dose-dependent.

Open field
One way ANOVAs found that treatment did not have a significant main effect on distance traveled [F(3, 36) = 1.248, p = .307] in the open field (Fig. 6B). Since there appeared to be a trend of all exercise groups showing increased distance traveled in the open field compared to the sedentary group, the three exercise groups were collapsed and a ttest was used to compare this collapsed exercise group to the sedentary group. This difference approached significance [t(38) = 1.731, p = .09], such that there was a trend of exercise traveling a greater distance compared to sedentary mice (data not shown).
A one way ANOVA found that treatment had a significant main effect on center time [F(3, 36) = 4.075, p = .014], and post-hoc analyses found that for both measures, all exercise groups displayed greater center activity compared to the sedentary group (p < .05 for all) (Fig. 7A). Taken together, these results suggest that the effects of exercise on general locomotion and anxiety-like behavior in the open field are significant but do not appear to be dose-dependent.  Since there appeared to be a trend of all exercise groups showing reduced corticosterone levels compared to the sedentary group, the three exercise groups were collapsed and a t-test was used to compare this collapsed exercise group to the sedentary group. This difference approached significance (p = .09), such that there was a trend of exercise mice having reduced corticosterone compared to sedentary mice. Arrow = male average, plus sign = female average.

Social interaction
A two way repeated measures ANOVA was performed to determine the effect of treatment on time spent sniffing the empty cup vs. conspecific cup during the social interaction task (Fig. 7B). According to the creator of this social interaction paradigm, groups can be assessed for exhibiting sociability (preference for conspecific cup vs. empty cup) using a repeated measures ANOVA when there are multiple groups, but degree of preference should not be compared as a graded parameter between groups [40]. There was a significant main effect of cup [F(1, 36) = 19.496, p < .001], such that mice spent more time with the conspecific cup than with the empty cup (p < .001). Although the treatment × cup interaction was not significant [F(3, 36) = 0.678, p = .571], only the 1 h and 12 h mice showed significant preference for the conspecific cup (p < .05 for both). Although the trend for sedentary (p = .135) and 3 h (p = .138) mice were in the same direction (more time spent with conspecific cup than empty cup), these differences did not reach significance, suggesting that exercise may dosedependently facilitate social behavior.

Marble burying
One way ANOVAs found that treatment did not have a significant main effect on number of digging episodes [F(3, 36) = 1.072, p = .373] (Fig. 7C), or digging time [F(3, 36) = 0.650, p = .588] (Fig. 7D). Since there appeared to be trends of exercise dose-dependently affecting digging time, correlations were run between this measure and running parameters. Digging time was correlated with rotations performed [r (38) = 0.301, p = .063]; however, this only approached significance (data not shown).

Novel object recognition task
A one way ANOVA found no significant main effect of treatment on the total time of exploration of the two objects during the novel object recognition task [F(3, 36) = 1.301, p = .289] (Fig. 8A). A two way repeated measures ANOVA was performed to determine the effects of treatment on time spent with the familiar vs. novel objects (Fig. 8B). This ANOVA found a significant main effect of object [F(1, 36) = 89.246, p < .001], and post-hoc testing found that mice spent more time interacting with the novel object than the familiar object (p < .001), suggesting intact object recognition memory. The main effect of treatment [F(3, 36) = 1.301, p = .289] and the object × treatment interaction [F(3, 36) = 0.929, p = .437] were not significant. However, although there were no differences in exploration of the familiar objects between groups (p > .05 for both), 1 h and 12 h exercise mice spent more time interacting with the novel object compared to sedentary mice (p < .05 for both), suggesting that exercise may dose-dependently facilitate performance on the novel object recognition task.

Y-maze
One way ANOVAs found no significant main effect of treatment on   (Fig. 9C), suggesting that a greater running volume was associated with enhanced performance on this task assessing spatial working memory.
3.4.3. Barnes maze 3.4.3.1. Latency to find. A two way repeated measures ANOVA was performed to assess the effect of treatment across time (days) on latency to find the escape hole in the Barnes maze task (Fig. 10A) A one way ANOVA was performed to determine the effects of treatment on latency to find the escape hole in the Barnes maze, averaged across days 2-5 (Fig. 10B). This ANOVA found a marginally significant main effect of exercise [F(3, 36) = 2.872, p = .050], and post-hoc analyses found that 3 h mice took longer to find the escape box compared to sedentary mice (p = .010).

Hole entry rate.
Since there appeared to be group differences in latency to find the escape box, hole entry rate was assessed, which is the average amount of time (s) between subsequent hole entries, to determine if this difference could be due to slower rates of exploration. A two way repeated measures ANOVA was performed to assess the effect of treatment across time (days) on hole entry rate in the Barnes maze task (Fig. 10C). This ANOVA found a significant main A one way ANOVA was performed to determine the effects of treatment on hole entry rate in the Barnes maze, averaged across days 2-5 (Fig. 10D). This ANOVA found a significant main effect of treatment [F(3, 36) = 4.312, p = .011], with trends of exercise mice traversing more slowly from hole to hole than sedentary mice. Post-hoc analyses found that 3 h mice were slower traveling between holes compared to all other groups (p < .05 for all except 12 h p = .073). The 12 h mice also moved from hole to hole more slowly than sedentary mice, though this only approached significance (p = .092). These findings suggest that longer access to a running wheel may slow exploration speed.

3.4.3.3.
Errors and re-entry errors. "Errors" were counted as the number of hole visits before finding the escape box hole. A two way repeated measures ANOVA was performed to assess the effect of treatment across time (days) on errors in the Barnes maze task (Fig. 10E). This ANOVA found a significant main effect of time [F(4,144) = 7.290, p < .001], such that mice made more errors on day 1 compared to all other days A one way ANOVA was performed to determine the effects of treatment on errors in the Barnes maze, averaged across days 2-5 (Fig. 10F). This ANOVA found no significant main effect of exercise [F (3, 36) = 1.726, p = .179], though post-hoc analyses found that 1 h mice made fewer errors than sedentary mice, though this difference Fig. 8. Mean (+SEM) measures of performance on the novel object recognition (NOR) task. (A) There was no significant treatment effect on total time spent exploring objects (p > .05). (B) There was a main effect of object, such that mice spent more time interacting with the novel object than the familiar object, and this was significant for all groups (^p < .05); however, although there were no differences in exploration of the familiar objects between groups (p > .05 for both), 1 h and 12 h exercise mice spent more time interacting with the novel object compared to sedentary mice (*p < .05 for both). Arrow = male average, plus sign = female average. 038], suggesting that a greater running volume was associated with enhanced performance on this task assessing spatial working memory (*p < .05). Arrow = male average, plus sign = female average. only approached significance (p = .089).
"Re-entry errors" were counted as the number of hole visits repeated to a previously visited hole before finding the escape box hole. A two way repeated measures ANOVA was performed to assess the effect of treatment across time (days) on re-entry errors in the Barnes maze task (Fig. 10G). This ANOVA found a significant main effect of time [F (4,144) = 8.932, p < .001], such that mice made more errors on day 1 compared to all other days (p < .05 for all). The main effect of treatment [F(3, 36) = 1.460, p = .242] and the treatment × time interaction [F(12,144) = 1.470, p = .142] were not significant.
A one way ANOVA was performed to determine the effects of treatment on re-entry errors in the Barnes maze, averaged across days 2-5 (Fig. 10H). This ANOVA found that the main effect of exercise approached significance [F(3, 36) = 2.625, p = .065], and post-hoc analyses found that 12 h mice made fewer re-entry errors than sedentary mice, though this difference only approached significance (p = .086). Taken together, these findings suggest that exercise may facilitate spatial learning and memory, as measured by decreases in errors and re-entry errors.

Discussion
Although the benefits of cardiovascular exercise (CVE) on healthy aging have been extensively investigated in rodents, most studies have tested a single exercise treatment group (usually unlimited running wheel access), resulting in large volumes of exercise that may not be clinically relevant. In agreement with this, the results of the current study indicate that mice given access to a running wheel for the entirety of their active cycle ran for about 50% of the time, or six hours. Therefore, it was of interest to determine whether smaller "doses" of CVE could exert similar benefits. Additionally, we hypothesized that giving mice different lengths of access to an exercise wheel may result in exercise regimens that varied not only by volume, but by intensity (as defined by several measures, discussed below) as well. Moreover, previous studies have generally utilized a shorter intervention period (ranging from a few weeks to a few months), while there is evidence suggesting that earlier, longer, or even life-long exercise interventions may have different effects [5,17,47]. This is the first long-term intervention study examining different daily levels of CVE to model a human life span of varying volume and intensity exercise regimens on physiological measures and a comprehensive behavioral battery assessing motor function, temperament, and cognition.

Varying access time to running wheels produces differences in running volume and intensity
There has been debate in the literature regarding the efficacy of exercise to produce health benefits regarding volume and intensity. Some studies suggest that longer-duration exercise is more beneficial compared to regimens shorter in duration [5]. Others suggest that only small amounts of exercise are necessary to see significant health benefits [24,25], and there is evidence that shorter duration high-intensity regimens are equally beneficial to longer-duration/lower-intensity regimens in regard to cardiovascular and metabolic function and musculoskeletal benefits [26][27][28][29]. Therefore, creating exercise regimens of different doses with varying duration/volume and intensity was of interest to test their effects on physiology, behavior, and neurobiology. Although studies have previously utilized varying duration and intensity of exercise using forced exercise (e.g., treadmill running), fewer studies have attempted to create different "doses" of exercise using voluntary wheel running. While treadmill running is a known stressor, some studies have shown that voluntary wheel running is significantly less stressful [33][34][35][36]. Additionally, wheel running has the benefit of allowing for correlational analyses of volume and intensity measures with outcome measures. Most studies that utilize wheel running as the form of exercise provide unlimited wheel access. As nocturnal animals, mice will not usually run during their light (sleep) cycle [48]. Therefore, our 12 h group is likely comparable to the unlimited wheel access groups in other studies. Compared to 12 h mice, 1 h mice performed 88% fewer wheel rotations, ran 19% slower, and took 60% fewer breaks; whereas 3 h mice performed 62% fewer wheel rotations, ran at the same speed (1% slower), and took 40% fewer breaks. There was a dose-dependent increase in running volume, such that, as hypothesized, animals that had longer access to a running wheel performed a greater number of wheel rotations. It was also hypothesized that animals with shorter access to a running wheel may attempt to compensate by increasing running speed or taking fewer breaks to maximize running volume. In agreement with this, we found that 1 h and 3 h mice took about half as many breaks compared to 12 h animals. However, mice with shorter access to a running wheel did not run faster to compensate; instead, 3 h and 12 h mice ran faster than 1 h mice, which may be due to increased fitness levels in these groups or simply represent the physiological limits of mice to sustain nearly continuous exercise for an hour. These findings suggest that providing mice with different lengths of access to a running wheel produces exercise regimens that not only differ in exercise volume, but also measures that may be indicative of some aspect of running intensity. The 1 h group ran the least and at the lowest speed but took the fewest number of breaks. The 3 h group ran an intermediate volume at a higher speed and took and took an intermediate number of breaks. The 12 h group ran the most and at a higher speed and took the most number of breaks.
These findings raise a perhaps more important question about the nature of aerobic exercise "intensity" as a construct, with regard to its equivalency between human studies and those in rodents. The construct of exercise volume (defined as the accumulated amount of exercise per some unit of time such as a day, week or month for humans, and total rotations per session for our mice; Fig. 2A) is unambiguously defined presently and dose-dependent as arrayed across our running wheel access periods. Intensity is a more difficult conceptual challenge, presently. It should first be noted that intensity can be defined differently in various human settings. Clinical studies and from public service sources such as the Centers for Disease Control often refer to exercise intensity for humans in somewhat imprecise terms (such as "light", "moderate" or "vigorous") [49]. At the other extreme, "gold standard" laboratory methods establish intensity referenced to metabolic capacity (e.g. VO 2 maximum, heart rate maximum or lactate threshold) using physiological measurement apparatus in controlled conditions that do not transfer readily out of the lab, though reasonably accurate estimations can be made using various formulae [50]. More convenient methods used to define intensity in humans for conditioning purposes are heart rate ranges indexed to age, power (sustained force) and perceived exertion on a numerical rating scale [51][52][53]. The latter measure is clearly not obtainable from rodents and getting heart rates or lactate thresholds, while possible in rodent engaging in exercise, were not practical measurements to take in a chronic running wheel format [54]. Therefore, we must look to the exercise parameters themselves in the present study to define the construct.
Classic heart rate zone-based training, such as high intensity interval training (HIIT) regimens, define workout intensity usually in terms of periods of exertion at fixed heart rate targets followed by precise, defined rest periods [55]. The running parameters that we measured in the mice do have distinct on/off periods that model this. However, our analyses include several potential measures of intensity based on the periods of time within which one integrates "on" periods with "off" periods. "Speed" (rotations per minute; Fig. 2B), which may mimic short intense bouts of exercise, such as Tabata intervals, is identical and highest for 3 and 12 h groups, and may estimate the maximum exertion they are willing to produce [56]. In contrast, the exercise period frequency or "density" (defined as the number of breaks per hour) reflects a somewhat different definition of intensity, not as clearly translatable to human HIIT routines (Fig. 2C). Furthermore, within-session analyses reveal yet other relevant patterns, in which both 1 h and 3 h groups defend the same high density of running, but in different ways: the 1 h group took fewer breaks whereas the 3 h group maintained a higher speed when they did run. The 12 h animals ran at higher densities earlier in the dark phase, producing a range of intensities. This discussion clearly raises more questions than it can answer, but nonetheless illustrates that future preclinical work that looks at rodent disease models will need to carefully consider these conceptual challenges and may want to anchor these observed patterns acute with laboratory tests in rodents that estimate lactate threshold and other metabolic measures that serve as the basis for human studies.

Exercise volume-dependently alters physiology
All doses of exercise resulted in similarly reduced body weight gain, which only became apparent during the last 3 months of exercise. Generally, there was a volume-dependent increase in food intake; however, no differences were seen between the 1 h and 3 h groups, despite 3 h mice running nearly three times as much as 1 h mice and exhibiting similar weight gain. There was also a significant positive correlation between running volume and food intake, and a significant negative correlation between running volume and body weight gain. Taken together, these findings suggest that mice will increase food intake when exercising more, likely to compensate for increased energy expenditure to maintain a healthy body weight. Additionally, there was no relationship between food intake or weight gain and running speed or breaks/h, suggesting that these physiological effects may be more likely mediated by exercise volume rather than intensity.
Interestingly, although there were trends in each exercise group, only 1 h mice showed a significant increase in raw brain mass compared to sedentary mice, which cannot be attributed to differences in body mass (since they weighed less than sedentary mice). This increased brain mass may be due, in part, to synaptogenesis, and cell and synapse survival during aging [57,58]. Brain volume is correlated with cardiovascular fitness in the elderly [59], and an experimental study found that aerobic exercise increased the volume of both gray and white matter regions in the aging population, suggesting a sparing of tissue [60]. Mice in the 1 h group also had larger kidneys (normalized to body mass) compared to all other groups. This is likely not fully attributable to a decrease in body mass with exercise, as exercise similarly attenuated body mass in all exercise groups, but only the kidney mass of 1 h mice appears to be affected. Exercise has been shown to alter renal function [61] and improve renal function in patients with chronic kidney disease [62]. Additionally, exercise (1 h & 12 h groups) also increased the normalized mass of the heart compared to sedentary animals. Previous studies have noted cardiac hypertrophy as an adaptation in response to aerobic exercise [63][64][65]. Lastly, exercise (1 h & 12 h) reduced the size of the pancreas, and this was true for both raw and normalized masses, in contrast to a previous exercise study [66].
Exercise was also shown to increase relative muscle mass, in agreement with prior findings [66], and this effect was similar across groups. This finding, however, may be attributable to a reduction in fat mass (and overall body weight), rather than an actual increase in muscle, as raw muscle mass was unaffected by exercise. Mitochondrial function in these muscles was assessed using citrate synthase activity assays. Citrate synthase activity has been used as an objective fitness measure, signifying aerobic capacity. Exercise increased citrate synthase activity (relative to protein content) in the gastrocnemius muscle only. The gastrocnemius is composed of red muscle fibers, and is referred to as a "fast-twitch" muscle since it is involved in fast movement like running and jumping. The soleus (a "slow-twitch" muscle composed of white muscle fibers, involved in standing and walking) and quadricep (a muscle composed of a mix of red and white fibers) were less affected by exercise intervention in the current study. Interestingly, CS activity was similarly increased across all exercise groups, but the increase seen was < 10% above sedentary levels. Previous studies have shown that citrate synthase and other mitochondrial enzyme levels in muscle increase following exercise training in both rodents and humans, with decreases seen following weeks to months of exercise cessation [67][68][69][70][71]. Since exercise training ended about a month before tissue collection, this could explain why increases in citrate synthase were small in the gastrocnemius and non-significant in other muscles, in the current study.
There was a trend of exercise reducing resting serum corticosterone levels, and this effect did not appear to be exercise pattern-dependent. Corticosterone levels were also not significantly correlated with measures of running volume or intensity. In previous studies, intense aerobic exercise has been shown to elevate glucocorticoid release [72], and reductions in resting corticosterone levels may represent adaptations of the hypothalamic-pituitary-adrenal axis in response to chronic exercise.

Exercise improves motor function without significantly affecting general locomotor activity or exploration
All exercise patterns equally improved motor coordination and balance, as measured by performance on the rotarod task. No pattern of exercise had a significant effect on general activity levels or exploratory behavior as measured by activity in the open field and arm entries in the Y-maze. Previous studies in mice have also found no effects of exercise on general activity levels in the open field [73,74]. While these findings suggest that exercise improves motor function without significantly affecting general locomotor activity or exploration, future additional tests such as grip strength and balance beam walking could further elucidate specific aspects of motor function that benefit from long-term wheel running.

Exercise reduces anxiety-like behavior and facilitates social behavior and marble burying
All exercise groups produced similar reductions in anxiety-like behavior in the open field (as displayed by more center activity). These findings are in agreement with previous studies that have found that exercise reduces anxiety-like behavior in animals [73][74][75][76] and anxiety measures in humans [2,77,78]. Furthermore, these findings indicate that even small doses of exercise are potentially anxiolytic. Group differences were also observed in social behavior on the three-chamber social interaction paradigm. Although there were trends of all mice spending more time with the conspecific cup compared to the empty cup, this was only significant for 1 h and 12 h exercise mice. These findings suggest that exercise may facilitate social behavior, which may be linked to a reduction in anxiety. However, while anxiolytic-like effects of 3 h exercise were seen in the open field task, this benefit was not measurable using the social interaction paradigm, which may be a better indicator of social rather than more general anxiety. Previously, exercise animals that showed reduced anxiety-like behavior in the open field and other tasks also showed increased social behavior [76]. Lastly, although there were no group differences, there were trends of exercise dose-dependently facilitating digging behavior in the marble burying task, in addition to correlations of digging time with both running volume and speed. Burying behavior has also been used as a test of anxiety [79,80], though performance on burying tasks has been shown not to correlate with performance on other tests of anxiety [79,81], as seen in the current study. It is also argue that marble burying is more reflective of a repetitive and perseverative behavior [81], and others have shown trends that exercised rodents show increased marble burying compared to sedentary controls [82]. Future additional tests including elevated plus maze or light/dark box for anxiety, and tail suspension or forced swim test for depression, could help further elucidate the effects of various exercise patterns on emotionality.

Exercise enhances cognitive-measure performance depending on dose
All groups of mice showed a significant preference for interacting with the novel object compared to the familiar object during the test phase of the novel object recognition task, a test of non-spatial memory. Exercise did not significantly affect total object exploration time or time with the familiar object, with exercising mice spending more time with the novel object than sedentary mice. Although this trend was apparent for all exercise groups, it was only significant for the 1 h and 12 h groups. Additionally, although there were no group differences, there were trends of exercise dose-dependently enhancing performance on the y-maze for spontaneous alternation task, a test of spatial working memory. In this task, there was also a significant positive correlation of the % alternation with running volume measures. Moreover, there were trends of 1 h and 12 h mice committing fewer errors and re-entry errors on the Barnes maze, a test of spatial memory, while speed of exploration of the Barnes maze was dose-dependently reduced by exercise, possibly as a result of reduced anxiety (as seen in the open field center time). Taken together, these results suggest that exercise dose-dependently enhances cognitive performance on a number of tasks, in agreement with several previous studies in animals and humans [37,[83][84][85][86][87][88]. While the lowest (1 h) and highest (12 h) doses appeared to be similarly beneficial for novel object recognition (a non-spatial memory task) and Barnes maze task (a hippocampal-dependent/spatial memory task [89]), Y-maze performance (a prefrontal cortex-and hippocampaldependent/spatial working memory task [90,91]) seemed to benefit most from higher volumes of exercise.

Conclusions
This study represents the first long-term (8 months) intervention study examining different daily levels of exercise to model a life span of regular aerobic exercise in a human on physiological response measures and a comprehensive behavioral battery assessing motor function, temperament and cognition. Varying length of access to a running wheel resulted in exercise regimens that differed by not only volume, but by measures of intensity as well. Exercise reduced body weight and increased relative muscle mass similarly across groups. While all exercise groups showed increased food intake, this was greatest in the 12 h group but did not differ between 1 h and 3 h mice. Additionally, all exercise groups showed improved motor function on the rotarod, and reduced anxiety in the open field. While exercise dose-dependently increased working memory performance in the y-maze, the 1 h and 12 h groups showed the largest changes in the mass of many organs, as well as alterations in several behaviors including social interaction, novel object recognition, and Barnes maze performance. It is likely that some of the effects of exercise were variable and/or diminished, particularly later behavioral assays and end-point physiological measures, as exercise was discontinued prior to the start of behavior testing. Therefore, authors have included results of analyses that trend towards significance, which may point to areas for future study using a shorter interval between the end of exercise and testing. Additionally, the current study included both males and females; however, separating analyses by sex results in an abundant loss of power. Because sex differences have been reported previously in voluntary running patterns and behavioral responses to running [e.g. [45,46]], symbols were added to figures to allow evaluation of these presently. However, future studies powered to confidently include sex as a biological variable are necessary. Overall, our findings suggest that there are widespread effects of long-term exercise on physiology, behavior, and cognition, and that even relatively small amounts of exercise can provide lasting benefits in mice.