Characterization of the running mouse model
We have used a wheel running mouse model to evaluate the influence of spontaneous and voluntary exercise on physiological aspects of the brain. Within the first 5 days after getting access to the running wheels, mice adjusted to this new task and ran with increasing intensity, reaching the stable level of about 8-10 km per day from day 6 (Fig. 1A). Systemic adaptation of mice to exercise requires a three week running period [24]; therefore, our measurements were performed after 21 days of running. The animals were active almost exclusively during the dark phase of the 12:12 daily cycle (Supplementary Figure S1). As shown in Fig. 1A, the mice decreased spontaneous wheel-running starting at day 22; however, they still run a total distance in the range from ∼6 to 10 km, which has been reported as typical for C57BL/6J mice [24]. This drop of the running intensity was not observed in the mice predisposed to BrdU evaluation (i.e. the mice that were not been subjected to behavioral testing; data not shown), suggesting that the behavioral evaluation during the day affected the running activity at nights. No features of fatigue or changes in cognitive abilities (data not shown) were noted in the running mice, the findings that were consistent with the literature (Harrington, 2012; Zhang et al., 2018). We also did not detect any changes in plasma creatinine level, the marker of muscle breakdown, as the result of wheel running (Fig. 1B). All these observations are consistent with the fact that physical activity was fully voluntary and spontaneous. The running mice gained approximately the same weight as the inactive controls (Fig. 1C), despite increased food (p<0.0001; Fig. 1D) which was accompanied by higher water consumption (p<0.001; Fig. 1E). We also measured the level of triglycerides in the blood, and no differences were noted in the running group as compared to inactive mice (Fig. 1F).
To further validate the model, we evaluated neurogenesis which typically increases in exercised animals [27]. Specifically, we examined the levels of doublecortin (DCX), a marker of pro-neuronal differentiation, in the hippocampal dentate gyrus. DCX expression was significantly higher in the running mice than in controls (Fig. 1G, p=0.029), indicating enhanced differentiation into the neuronal lineage as a result of increased physical activity.
We also evaluated the number of proliferating cells in the dentate gyrus. Following 14 days of voluntary wheel running, mice received five injections with BrdU to label proliferating cells. To study a long-term effect of physical exercise, mice were allowed to maintain physical activity for the additional 20 days. The length of exercise was chosen based on the fact that immature newborn granule cells enter an integrated stage as early as 2 weeks after being born, and we extended this time frame by the factor of 2 to more precisely capture cell proliferation. Because of the high density of neuronal progenitor cells in the dentate gyrus, the method primarily reflects proliferation of this cell type. A significantly higher number of BrdU-positive cells was detected in the dentate gyrus from the running mice (Fig. 1H; p=0.029) as compared to inactive controls, suggesting enhanced cell proliferation and/or survival.
Bioenergetics metabolites are distinctly expressed in the FC and hippocampus in response to exercise
We analyzed the impact of physical activity on the metabolomic profile in two brain structures, namely in the hippocampus and the FC. Overall, 111 metabolites were identified (Figs. 2A-B), with 16 significantly changed in the hippocampus, and 15 significantly altered in the FC. All compounds that differed significantly between the running and inactive groups are presented in Supplementary Table S1.
The analysis revealed distinct metabolic responses to exercise in the hippocampus and the FC. Interestingly, the levels of the TCA cycle intermediates were not altered in the hippocampus, while a significant decrease was noted in the FC in the running mice as compared to their inactive controls. The levels of citric acid (p=0.034), alpha-ketoglutaric acid (p=0.048), fumaric acid (p=0.008), and malic acid (p=0.045) were all reduced in the FC of the running mice (Fig. 3). The concentration of niacinamide, the NAD+ donor in the TCA cycle was significantly lowered (p=0.031) in the running mice. The TCA results are different from those obtained in a rat model of forced and exhaustive exercise, likely because samples in our study were collected during a resting phase [28]. Our results based on a prolonged voluntary exercise indicated a decrease in the TCA metabolites and glycerol (an energetic substrate utilized by neurons [29], p=0.038) in the FC of the running mice, with a simultaneous increase in glycolysis metabolites in the hippocampus (Table 1). An increase in the glycolytic metabolites, such as 3-phosphoglycerate (p=0.0133) and fructose 6-phosphate (p=0.026) in the hippocampi of the running mice is consistent with previous reports on upregulation of the hippocampal enzymes involved in glucose utilization and metabolism in rodent models of physical activity [30–32]. Observed in our study was a decrease in the level of cortical glycerol (p=0.038), an energetic substrate utilized by neurons [29], may contribute to the insufficient TCA turnover (Table 1). A decrease in cortical GABA level (p=0.0023, Fig. 3) may also be related to the inefficient TCA cycle, because this neurotransmitter can be synthesized by glutamate/alpha-ketoglutarate/GABA pathway [33]. The median concentration of glutamic acid, the crucial neurotransmitter and substrate for the GABA synthesis in the pathway mediated by alpha-ketoglutaric acid [34,35] showed a tendency to be decreased in the FC of the running mice (Fig. 3). Despite the decreased cortical GABA level, the GABA/glutamate ratio was not changed (0.92 [0.83-1.16] vs 0.89 [0.11-1.15] for the running vs inactive mice, respectively). The right balance of GABA/glutamate is essential for normal brain functioning [36]; thus, the lowered cortical GABA probably does not affect behavioral phenotype in our running mice.
Amino acids can be converted to metabolic intermediates and utilized as the TCA substrates. The level of phenylalanine was decreased (p=0.025) in the FC and the level of L-valine dropped (p=0.022) in the hippocampi of the running mice. D-Pinitol, a marker of consumption of soy products [37], was elevated in the hippocampi of the running mice compared to the controls (p<0.0001), suggesting a possible accumulation of this metabolite as the result of increased consumption of standard rodent’s chow that contains 14% of Hi-Pro soybean meal. This suggests another possible mechanism through which voluntary exercise may contribute to the modulation of the metabolome, namely, by increasing consumption of food and deposition of foodborne compounds in the brain.
Physical activity alters FA profile in the brain
Analysis of the content of the hippocampal and FC samples showed remarkable differences in the FA profile due to physical activity (Fig. 4). Saturated FA (Fig. 4A), such as heptadecanoic (17:0), stearic (18:0), and palmitic acids (16:0) were increased in the running mice as compared to inactive controls in both the hippocampi and the FC. Myristic acid (14:0), another common saturated FA, was increased in the hippocampal samples, and showed a high tendency to be elevated in the FC. In contrast, unsaturated FA were generally decreased in the running mice (Fig. 4B). Specifically, the levels of arachidonic acid decreased in both examined brain regions of the running mice (p=0.032 for FC and p=0.021 for the hippocampus). In addition, the concentration of docosahexaenoic acid was lowered in the hippocampal samples (p=0.015) but it tended to be increased in the FC (Fig. 4B).
Because palmitate, and to lesser extent, myristate and stearate are the primary end products of de novo lipogenesis [38], we hypothesized that this process may be boosted in the brain of the running mice. Therefore, we assessed the expression of FAS, which utilizes acetyl CoA and malonyl CoA to elongate FAs by two carbons catalyzing the biosynthesis of palmitate [39]. We observed that the levels of this maker of lipogenesis were elevated in the hippocampi but not in the FC of the running mice (Fig. 3C; ANOVA: F=12.43, p=0.001; results of post hoc testing are marked on the graph).
We also considered that the observed changes in FA composition could affect lipid enriched myelin sheath as a consequence of physical activity. To explore this possibility, we measured the expression of myelin basic protein (MBP), a marker of myelin, by immunoblotting. No changes were observed on the impact of exercise on the cortical and hippocampal level of this protein as compared to inactive controls (data not shown), suggesting that myelin levels were not changed in the running mice.
Cortical and hippocampal metabolites correlate with activity level
We next evaluated a possible correlation between alterations of brain metabolites and mouse activity levels. Based on Pearson’s correlation, and we identified several metabolites which concentrations was positively or negative correlated with the level of activity calculated as the average daily distance run by an animal (Table 2).
The brain concentration of eight metabolites closely correlated with the activity level of the mice. In the hippocampus, tyrosine (p=0.049), threonine (p=0.048), phenylalanine (p=0.048) and galactose-6-phosphate (p=0.02) showed positive correlation with the distance run by mice. Phenylalanine, as a tyrosine precursor, is a substrate for tyrosine hydroxylase in the rate-limiting step in catecholamine synthesis [40]. Threonine regulates mTOR signaling [41], which is stimulated in the hippocampus and other brain structures of exercised rodents [42]. Galactose-6-phosphate is in constant equilibrium with fructose-6-phosphate, and the positive correlation between these two compounds was significant in the hippocampi of the running mice (r=0.82, p<0.05; data not shown). Among these metabolites, galactose-6-phosphate levels were increased in the running mice by 192.4% (p=0.013) as compared to controls (Supplementary Table 1). In the FC, citric acid (p=0.038), L-arabitol (p=0.00014), aspartic acid (p=0.035) and phosphoethanolamine (p=0.022) negatively correlated with running activity level. Relatively to the inactive control mice, the cortical levels of citric acid, phosphoethanoloamine, and L-arabitol were 82% (p=0.04), 76% (p=0.04), and 68% (p=0.0023), respectively. Cortical fumaric acid showed a tendency for negative correlation with activity level; however, the changes were not significant (r=-0.71, p=0.07). A very strong negative correlation (r=-0.98, p<0.0001) was observed for arabitol, a sugar alcohol, that can be found in most biofluids, although its role in living organisms is unclear. Information about its role in the brain is scare; however, in the light of the present findings, its functions should be reconsidered.
Physical activity enhances anxiolytic behavior
Accumulation of even-chain saturated brain FA affect anxiety responses [43]; therefore, we also performed several tests focusing on anxiety evaluation in the running and inactive mice. In the EPM test, the running mice spent significantly more time in the open arms of the EPM area (p=0.024, Fig. 5A, left panel) and less in the closed arms (p=0.024, Fig. 5A, middle panel) than inactive animals, indicating enhanced anxiolytic responses. In addition, the cumulative time of nose exposing beyond the area (so called, dipping time) showed tendency to increase in the running mice (p=0.065, Fig. 5A, right panel).
In the D/L test, the running mice spent significantly more time in the light area than the inactive controls (p=0.048, Fig. 5B), which confirms that the active animals were less anxious. These results were consistent with the OF test, in which the running mice, but not the inactive controls, explored more intensively the center compartment of the cage (p=0.011, Fig. 5C). In addition, the running groups spent approximately the same time in the center as compared to cage margins. In contrast, the inactive mice spent less time in the center area and more in the margin (p=0.002, Fig. 5D). Additional functional parameters of the OF test are provided in Table 3.