Low-fat diet, and medium-fat diets containing coconut oil and soybean oil exert different metabolic effects in untrained and treadmill-trained mice

Seven (7)-week old male C57BL/6N mice were purchased from Shimizu Laboratory Supplies (Hamamatsu, Shizuoka, Japan). Mice were housed in an animal room at 22 ± 0.5 °C and 50% humidity with a 12 h light-dark cycle (lights on and off at 6:00 and 18:00, respectively). Mice were acclimatized to this environment with ad libitum access to a standard chow diet (Oriental Yeast Co., Tokyo, Japan) and water for 7 to 10d before changing to assigned diets. Mice were randomly assigned to the following purified diets: low-fat diet (L; 20% kcal from casein, 70% kcal from cornstarch and 10% kcal from soybean oil); coconut oil diet (C; 20% kcal from casein, 50% kcal from cornstarch, 10% kcal from soybean oil and 20% kcal from coconut oil); and soybean oil diet (S; 20% kcal from casein, 50% kcal from cornstarch and 30% kcal from soybean oil). All diets contained vitamin and mineral mixes and were prepared by Research Diets (NJ, USA) based on the D12450K formulation. Training of mice was commenced following diet assignment. Dietary groups were divided into untrained (U) and trained (T) groups. Animal experiments were performed according to the Kyoto University Guidelines for the Ethical Treatment of Laboratory Animals as approved by the Kyoto University Animal Experimentation Committee with the number (29–39).

Training was conducted daily for 30d from 6:00 using a treadmill for rodents (MK-680; Muromachi, Tokyo, Japan). On the first 15d, mice ran for 1 h at 15 m/min at 3° incline. From the 16th day, intensity was increased to 18 m/min. Mice were forced to run by poking. All mice completed the training program.

Basal indirect calorimetry using the ARCO-2000 system (Tokyo, Japan) under ad libitum feeding and resting conditions as described in [24] was performed on a subset of mice. Mice were assigned to calorimetry chambers on the 28th day of training to facilitate acclimatization. Actual measurement was performed 24 h to 48 h post-training which represents a full light-dark cycle devoid of acute exercise effects. Mice were sacrificed after measurements. Sample collection was based on Manio, et al. [23].

Indirect calorimetry during exercise with treadmill endurance test was performed on a subset of mice. About 1.5-2 h before the run (6:00), mice were placed in sealed treadmill chambers (Mousebelt-200; Arco System, Tokyo, Japan) at 10° incline to acclimatize. After warming-up for 2 min, intensity was increased to 15 m/min. After 30 min, the intensity was increased to 18 m/min and maintained for 30 min. Then, the intensity was increased to 21 m/min and kept herein until exhaustion. Mice were stimulated with electrical stimulus (0.2 mA) and occasional noise and poking. Exhaustion was ruled if mice remained on electrodes or could not sustain running for 20s despite additional stimulation.

Fixed time run was performed on a subset of mice. Food was removed 2 h before running. Mice were individually placed on a moving treadmill set at an intensity of 10 m/min. After 2 min of warm-up, mice were transferred to a moving treadmill set at 15 m/min, 10° incline. After exactly 30 min, mice were removed from the treadmill and immediately sacrificed. Experiments were performed in the same conditions as exercise indirect calorimetry.

Serum was measured for glucose, triglycerides (TG), non-esterified fatty acids (NEFA) and beta-hydroxybutyrate (β-HB). Glycogen and TG were measured in organs [25‐27]. Measurements are detailed in [23].

Muscle and liver were lysed in a 1% NP-40 buffer as detailed in [23]. Beta-hydroxyacyl-CoA dehydrogenase (β-HAD) activity was measured according to a procedure by Holloway, et al. [28]. Succinyl-CoA:3-oxoacid CoA-transferase (SCOT) activity was measured based on Williamson, et al. [29]. Citrate synthase (CS) activity was measured according to Srere [30]. Cytochrome c oxidase of the mitochondrial electron transport chain, also known and hereby referred to as Complex IV, was measured based on Mac Arthur, et al. and Spinazzi, et al. [31, 32]. Acetoacetyl-CoA thiolase (AACT) and deacylase (AACD) activities were measured based on Williamson, et al. [33]. All enzyme activity measurements were modified to adapt to a 96-well plate system as detailed in [23].

Total RNA was extracted with Tripure Isolation Reagent (Roche, Mannheim, Germany) and GenElute Mammalian Total RNA Miniprep Kit (Sigma-Aldrich, MO, USA) as detailed in [23]. Total RNA (1.8 μg and 1.5 μg for muscle and liver, respectively) was reverse transcribed with Transcriptor First Strand cDNA Synthesis Kit (Roche, Mannheim, Germany). Messenger RNA (mRNA) expression levels were quantified using intron-spanning primers and corresponding Universal Library Probes (Roche, Mannheim, Germany) listed in Table 1. Values were rationalized to Hprt expression [34].

Protein concentration of lysates was adjusted with lysis buffer and 4× Laemilli buffer containing 20% mercaptoethanol. Samples were loaded on 8% polyacrylamide gels. Semi-dry transfer to PVDF membranes was performed in a transfer buffer containing 20% methanol. After transfer, Ponceau S staining was performed. Excess stain was removed in distilled water. Membranes were visualized and digitized (LAS-3000; Fujifilm, Tokyo, Japan). Membranes were blocked in 5% skim milk powder in Tris-buffered saline with 0.1% Tween-20 buffer (TBST) containing 0.05% ProClin 300 (Sigma Aldrich, MO, USA). Membranes were cut at sections corresponding to regions of molecular weight previously identified to contain the protein of interest. Membranes were incubated in goat anti-fatty acid translocase/cluster of differentiation 36 (CD36) antibody (1:2000; AF2519, R&D Systems, MN, USA), goat anti-pyruvate dehydrogenase kinase 4 (PDK4) antibody (1:1000; C-16, Santa Cruz Biotechnology, CA, USA), rabbit anti-PGC1A antibody (1:1000; H-300, Santa Cruz Biotechnology, CA, USA) or rabbit anti-glucose transporter 4 (GLUT4) antibody (1:2000, AB1346, Chemicon International, CA, USA) at 4 °C for 20 h. Membranes were washed 3× in TBST to remove excess primary antibody. Horseradish peroxidase-labelled secondary antibody incubation was performed in anti-goat IgG (1:1000; P0449; Dako, Tokyo, Japan) or anti-rabbit IgG (1:1000; P0399; Dako, Tokyo, Japan) for 3 h at 4 °C. After washing 3× in TBST to remove excess secondary antibody, membranes were visualized by chemiluminescent detection (Western Lightning Plus ECL; Perkin Elmer, MA, USA). Signals corresponding to PGC1A (91 kDa), CD36 (88 kDa), GLUT4 (58 kDa), PDK4 (47 kDa) and Ponceau S signals were quantified using the software MultiGauge V3.2 (Fujifilm, Tokyo, Japan) with automatic background detection.

Statistical analyses were performed using the Prism 5.0 software (Graphpad Software, CA, USA). Points in exercise indirect calorimetry time-course data are presented as means ± SEM. Each time point was analyzed by one-way analysis of variance (ANOVA) to compare groups of different diets and by Student’s unpaired t-test to compare groups of different training states receiving the same diet. Other data are presented as means ± SEM and were analyzed accordingly using one-way ANOVA followed by Newman-Keuls post-hoc test and Student’s t-test. Significance level (α) was set at 0.05.

Food intake was measured every third day. Total food and calorie intake for 30 days were calculated (Table 2). Body weight and weight of some organs were measured. Total food intake decreased in C and S relative to L in either training state (p < 0.05). Training did not increase food intake in C and S but significant increase was observed in L (p < 0.05). Total calorie intake showed no difference among diet groups in either training state. Training increased calorie intake in L (p < 0.05). Despite this, diets nor training did not cause significant body weight change at the end of the experimental period relative to day 1. While significant differences were not observed in muscle and fat weight among diet groups of either training state, training significantly increased liver weight in L and S (p < 0.05) but not in C. Therefore, the type of diet combined with training affected calorie intake and liver weight but not body weight.

Indirect calorimetry at rest 24 h after the last training session was performed to determine the effects of diet and training on whole-body metabolism. Basal RER significantly increased in L relative to C and S in either training state (p < 0.05) (Fig. 1a). Furthermore, training increased RER in all diet groups (p < 0.05). Basal VO₂ and energy expenditure (not shown) were not significantly affected by diet nor training (Fig. 1b). L in either training state increased carbohydrate oxidation (CHO) than C and S (p < 0.05) (Fig. 1c). Furthermore, training increased CHO in all diet groups (p < 0.05). Fat oxidation (FAT) in L was significantly decreased relative to C and S in either training state (p < 0.05) (Fig. 1d). Training decreased FAT in all diet groups but significance (p < 0.05) was observed only in L and S but not in C (p = 0.0726). Therefore, at rest in ad libitum fed state, training shifted substrate utilization favoring CHO beyond the effects of diet. Furthermore, coconut oil and soybean oil slightly but differently affected FAT with training.

Energy metabolism was measured during the first hour of endurance test. Diet groups had similar RER in either training state (Fig. 2a, b). Training decreased RER in several time points during the run especially at the onset of running (p < 0.05). Diet groups had similar VO₂ in either training state (Fig. 2c, d). Training lowered VO₂ at several points during the run (p < 0.05). However, cumulative data showed that S with training was not significantly different to its untrained counterpart (p = 0.0845). Diet groups had similar CHO in either training state (Fig. 2e, f). However, trained groups had consistently lower CHO during the run (p < 0.05). Furthermore, drastic elevation in CHO at the onset of running in untrained groups was absent with training. Cumulative CHO was lower with training but significant only in L (p < 0.05, p = 0.0746 and p = 0.0571 in L, C and S, respectively). Diet groups had similar FAT in either training state (Fig. 2g, h). FAT was higher in the trained groups particularly in the first few time points of the run (p < 0.05). Except in S, a reversal was observed as the run progressed with untrained groups having higher FAT than trained groups. In all parameters, no significant differences at any time point were observed among diet groups in either training state. Therefore, diet composition did not affect whole-body metabolism among groups of the same training status during exercise. On the other hand, training strongly affected exercise metabolism causing a slight shift to FAT if relative contribution to energy expenditure was considered as reflected in the mild but observable lowering of RER.

All mice were able to run past 60 min and exceed the intensity of 18 m/min (Fig. 3a). Diet groups had similar time-to-exhaustion and distance in either training state (Fig. 3b, c). Training increased to nearly double the treadmill endurance of untrained counterparts (p < 0.05) regardless of diet. Calculating for work revealed that C increased treadmill endurance than L and S in the untrained but not in the trained state (p < 0.05) (Fig. 3d). Therefore, coconut oil slightly but significantly improved endurance in the untrained state but did not potentiate the endurance improvement observed with training.

Pre- (basal) and post-exercise (30 min run) glucose-related metabolites were measured and presented in Table 3. Pre-exercise serum glucose was similar among diet groups in either training state. Training tended to decrease serum glucose. At post-exercise, all groups had significantly increased serum glucose (p < 0.05). L tended to decrease serum glucose than C and S and this was highlighted in trained groups (p < 0.05). Pre-exercise muscle glycogen was similar among diet groups in either training state. However, training increased this in L and S but not in C (p < 0.05). At post-exercise in the untrained state, it decreased (p < 0.05) relative to pre-exercise values in L and C while a non-significant decrease (p = 0.0672) was observed in S. With training, muscle glycogen also significantly decreased in L and S (p < 0.05) but not in C. Pre-exercise liver glycogen was similar among diet groups in the untrained state. It was significantly lower in C than L in the trained state (p < 0.05). Also, a non-significant decrease (p = 0.0735) was observed in C with training relative to its untrained counterpart. However, a longer run might be required to observe utilization of glycogen in the liver. Therefore, coconut oil prevented glycogen accumulation in the muscle and liver with training but preserved it in the muscle during exercise.

Lipid- and ketone body-related metabolites at pre- and post-exercise are also presented in Table 3. Pre-exercise serum TG and NEFA was similar among diet groups in either training state. However, training caused significant decrease in NEFA in L (p < 0.05). At post-exercise, TG significantly decreased in S in both training states relative to pre-exercise values (p < 0.05). Post-exercise NEFA also decreased significantly in S without training (p < 0.05) and non-significantly in other C and S groups regardless of training (p < 0.1). Pre-exercise serum β-HB was similar among diet groups in either training state. However, it decreased with training and was significant in L (p < 0.05). At post-exercise, drastic elevation was observed in C in both training states and was emphasized by training relative to L and S (p < 0.05). Similarly, training relatively lowered β-HB levels post-exercise in all diet groups. Pre-exercise muscle TG was not significantly affected by diets nor training. Significant decrease was not observed in any untrained group at post-exercise. With training, muscle TG non-significantly decreased (p = 0.0522) in L and significantly decreased (p < 0.05) in C and S relative to pre-exercise values. Therefore, coconut oil caused ketosis during exercise but not at rest. On the other hand, training promoted intramuscular TG utilization during exercise.

The effects of diets and exercise on mitochondrial enzyme activities of the muscle were assessed. CS catalyzes the formation of citrate from oxaloacetate and acetyl-CoA and is commonly used as a measure of mitochondrial density in muscle [35]. In either training state, CS activity increased in C and S relative to L (p < 0.05) (Fig. 4a). Training significantly increased CS activity in all diet groups (p < 0.05). β-HAD activity reflects the capacity of cells to oxidize fatty acid [28]. In the untrained state, C had a non-significant increase in β-HAD relative to L and S. β-HAD activity was similar among diet groups in the trained state (Fig. 4b). Also, training significantly increased β-HAD activity in L and S (p < 0.05). An increase in β-HAD activity was also observed in C but this was not statistically significant. SCOT is an enzyme involved in ketolysis in the muscle [36]. In the untrained state, SCOT activity increased in C and S relative to L (p < 0.05) (Fig. 4c). Significant difference among diet groups in SCOT activity was not observed with training. Training increased SCOT activity in all groups but significant only in L and C (p < 0.05). Complex IV catalyzes the last step of the electron transport chain in the mitochondria and also regulates oxidative phosphorylation [37]. Similar complex IV activity was observed among diet groups in either training state (Fig. 4d) and training increased its activity (p < 0.05). Overall, coconut oil and soybean oil increased some muscle mitochondrial enzyme activities in the untrained state. However, training increased and normalized these muscle mitochondrial enzyme activities across the diet groups.

Changes in mitochondrial enzyme activities and whole-body metabolism are influenced by transcriptional adaptations in the muscle. We measured known transcription factors affecting the said processes. Diet and training failed to significantly influence Pgc1a (Fig. 5a). Ppara non-significantly increased in C and S relative to L (Fig. 5b). While training failed to increase Ppara in C and S, 40% increase albeit not significant (p = 0.0702) was observed in L. Pparb/d expression was not affected in untrained diet groups (Fig. 5c). Training increased Pparb/d expression in L and S but not in C (p < 0.05). ERRs showed relative inhibition of expression in C and S in the untrained state (Fig. 5d-f). Significant decrease in expression was observed only in Errg (p < 0.05). Training increased the expression of Erra and Errb in all groups but significant only in S (p < 0.05). On the other hand, Errg decreased with training in L while significant increase was not observed in C and S (p < 0.05).

We also determined the expression of some PPAR targets. In the untrained state, Lpl expression was significantly increased in C and S relative to L (p < 0.05) (Fig. 5g). With training, the intergroup differences were abolished which could be attributed to the non-significant increase (20%) in L. In the untrained state, Cpt1b was significantly increased in C relative to L but not to S (p < 0.05) (Fig. 5h). With training, Cpt1b did not increase in C in contrast to considerable and significant (> 50%; p < 0.05) increases in L and S. Glut4 was significantly decreased in C and S relative to L (p < 0.05) (Fig. 5i). Training did not change its expression in L and C. In the S however, significant increase similar to the level of L was observed (p < 0.05). Fatp1 expression was similar among diet groups in either training state. Training caused a non-significant increase (p = 0.0567) in S but not in L and C (Fig. 5j). Myh2 non-significantly decreased in S in the untrained state (Fig. 5k). Training significantly increased Myh2 expression and normalized its expression with L and C (p < 0.05). Lastly, no significant differences in Cd36 and Pdk4 were observed because of diet nor training (not shown).

In summary, the amount of fat but not the type significantly affected the expression of PPARs and ERRs in the untrained state. With training, however, the responses of these genes and their targets varied. PPARs were influenced by the type of fat while the ERRs with the exception of ERRγ were unaffected by diet. In particular, coconut oil inhibited increases in PPARβ/δ mRNA expression as well as some of its targets.

Protein expression of glucose and fatty acid metabolism-related genes in the muscle were measured by immunoblotting. GLUT4 protein expression was non-significantly decreased in S in the untrained state (Fig. 6a, b). Training increased GLUT4 protein in all diet groups but significant only in L and S (p < 0.05) but not in C (p > 0.1). Protein expression of PGC1A, CD36, and PDK4 were not significantly influenced by diet nor training (Fig. 6a; quantification not shown). Therefore, training increased GLUT4 protein but coconut oil prevented the same degree of induction as in other diets.

The effects of diets and training on mitochondrial functions of the liver were assessed. AACT and AACD catalyze the reversible conversion of acetyl-CoA to acetoacetyl-CoA and the non-reversible conversion of acetoacetyl-CoA to acetoacetate, respectively [33]. AACT and AACD activities were significantly different among diet groups with training but not in the untrained state (Fig. 7a, b). AACT significantly increased in S than L and C (p < 0.05). AACD, on the other hand, was significantly higher in S than C and was significantly higher in C than L (p < 0.05). Similar to AACT, training significantly increased AACD activity in C and S but not in L (p < 0.05). Complex IV activity was not significantly affected by diet nor training but a trend towards increased activity could be observed in C and S in both training states (Fig. 7c). β-HAD was similar among diet groups in either training state (Fig. 7d). However, there was a robust but non-significant (p = 0.0579) increase in β-HAD activity with training in C. Therefore, in the liver, coconut oil tended to improve β-oxidation with moderate ketogenic enzyme activation while soybean oil strongly activated ketogenic enzymes but not β-oxidation with training. However, increased mitochondrial density was not observed as universal changes in mitochondrial functions were absent.

PPARα influences hepatic ketogenesis and β-oxidation [10]. Ppara in the liver was similar among diet groups in the untrained state (Fig. 8a). Training significantly increased Ppara in C relative to L and S, and to its untrained counterpart (p < 0.05). Cpt1a non-significantly increased with training in C (Fig. 8b) while Hmgcs2 was not significantly affected by diet nor training (not shown). Therefore, in the untrained state, the type of fat did not affect PPARα gene expression in the liver while coconut oil but not soybean oil induced its expression with training.