The impact of exhaustive exercise on metabolic proles in young men: a metabolomics approach

Background: Exercise-induced fatigue leads to reduction in the ability exert physical performance. Prolonged and intensive exercise stimulates several metabolic pathways to produce energy. The purpose of the present study was to investigate the impact of exhaustive exercise on metabolomic pathways. Methods: Nine young recreationally active men were recruited to this study. Participants performed step incremental maximal exercise until maximum exhaustion. Saliva samples were collected pre- and post-exercise (immediately after exercise cessation) using a salimetric oral swab, and salivary metabolites were analyzed using capillary electrophoresis and time-of-ight mass spectrometry. Results: Two hundred ten metabolites were detected, representing different clustering principle component structures between pre- and post-exercise. Orthogonal partial least squares discriminant analysis identied 29 metabolites with highly related variable importance for projection score and 16 metabolites signicantly increased after exercise. Furthermore, increase in cyclohexylamine was positively correlated with an increase in fatigue on a visual analog scale. Conclusion: The present study demonstrated that exhaustive exercise changed the saliva metabolomic pathways related to energy production including glycolysis, lipolysis, amino acid metabolism, amines, and ketone bodies.


Introduction
Prolonged and strenuous exercise causes fatigue, which reduces force and power and then obstructs ability to continue physical performance. Multiple theories have been put forward to explain the complex and interactive nature of fatigue. [1] Fatigue causes internal environment changes that include a lack of oxygen and energy from glucose and fatty acids, an accumulation of carbon dioxide, hydrogen ions, lactate, and ammonia, and a heat loading. 2 In high-intensity exercise conditions (e.g., after onset of blood lactate accumulation (OBLA)), homeostasis of the internal environment is disrupted mainly by lactate, ammonia, and organic acid, which reduce the exercise workload or stop it completely. [2] It is plausible that maintaining metabolism of nutrients such as carbohydrate, fatty acid, and amino acid is important for reducing exhaustive fatigue.
The human metabolism could be analyzed by the metabolome, which can identify and quantify the speci c metabolites through a comprehension of biological compounds at the small molecule level. [3.4] The analysis of nontargeting approaches also provides a new insight into the biophysiological mechanism of response to disease or exercise. [5] Previous studies have demonstrated that fatigue symptom during competitive or intensive exercise were associated with changes in glucose and tricarboxylic acid (TCA) cycle metabolic pro le. [6] In addition, Monaf et al. (2018) [7] found changes in fatty acids, neurotransmitters, and indole metabolism following prolonged maximal exhaustive exercise at moderate intensity. However, little is known regarding the relationship between the human metabolome and exhaustive exercise-induced fatigue relatively short duration at high intensity. Therefore, the aim of the present study was to investigate the impact of exhaustive maximal exercise on fatigue and the metabolome. We performed a comprehensive comparison of saliva metabolites before and after exhaustion induced by gradually increased exercise.

Subjects
A total of 9 healthy and recreationally active young men participated in this study. All subjects in this study were medication-free and had no history of metabolic abnormalities such as diabetes mellitus, hypercholesterolemia, hypertension, and in ammatory disease. The present study was approved by the institutional review board at the Japan Institute of Sports Sciences. All subjects gave written informed consent prior to participating in the study. All study procedures were performed in accordance with relevant guidelines/regulations.

Study protocol
This study was a single-arm trial consisting of exhaustive cycling exercise using a metabolomics approach. The participants performed the exhaustive exercise test at least 3 h postprandially and had not consumed any caffeine and alcohol for 12 h prior to the exercise test. They completed a 5-min warm-up at 50-W, and then power output increased by 50 W every 8 min until reaching an exhaustive state modi ed on the basis of previous study. [8] The state of exhaustive fatigue is de ned as the time point where participant cannot maintain cadence, and the heart rate reaches 90% of the maximum heart rate. Heart rate was monitored using short-range radio telemetry (RS 800, Polor, Finland). We measured the average of oxygen uptake every 30 s during exercise test using online computer-assisted circuit spirometer (AE300S; Minato Medical Science, Osaka, Japan). Subjects were asked to assess fatigue using a visual analog scale (VAS) pre-and post-exercise.

Salivary metabolome
Saliva samples were collected using an oral swab cotton swab and a storage tube (Salimetrics oral swab; Salimetrics, USA) before and immediately after the exhaustive exercise. Participants were asked to abstain from any food and drink before the saliva collection. Before and after exercise, they sat and rinsed their mouth with distilled water 3 times, then rested for at least 5 min, both of pre-and postexercise. Saliva production was stimulated by chewing on cotton for 1 min at a rate of 1 chew/s. [9] The obtained saliva samples were separated from the cotton by centrifugation at 1500 g, and the samples were frozen at -80 ˚C until analysis.
Metabolite levels were determined using capillary electrophoresis and time-of-ight mass spectrometry (CE-TOFMS) analysis. A saliva sample of 25 µL and Mili-Q water of 25 µL were combined with 400 µmol/L of commercial standard solution (H3304-1002; Human Metabolome Technologies, Japan) and passed through a 5 kDa cut-off lter to remove proteins and macromolecules. The ltrate was analysed using CE-TOFMS performed using an Agilent capillary electrophoresis system (Agilent Technologies, Waldbronn, Germany) as described previously. [10] Brie y, cationic metabolites were analyzed through a fused silica capillary (50 µm internal diameter, 80 cm length) with a commercial buffer as the electrolyte. The sample was injected at a pressure of 5 mbar for 10 s. The applied voltage was set at 28 kV. Electrospray ionization-mass spectrometry was conducted in the positive ion mode. The spectrometer was scanned with mass-to-charge ratio (m/z) ranging from 50 to 1000. Anionic metabolites were analyzed through a fused silica capillary (50 µm internal diameter, 80 cm length) with a commercial buffer as the electrolyte. The sample was injected at a pressure of 50 mbar for 25 s. The applied voltage was set at 30 kV. Electrospray ionization-mass spectrometry was conducted in the negative ion mode.
The spectrometer was scanned using m/z ratios ranging from 50 to 1000. The obtained data were analyzed using proprietary autonomic integration software (MasterHabds; Human Metabolome Technologies, Tsuruoka, Japan). Each metabolite was identi ed and quanti ed by the peak information including m/z ratio, migration time, and peak area.

Statistical analyses
Data are expressed as the means ± SD. Statistical analysis was performed using R software (version 3.4.2, https://www.r-project.org/). Principle component analysis (PCA) was carried out using multivariate techniques of metabolic global view by the R packages FactoMineR and factoextra. PCA components express a linear combination of the metabolite levels weighted by the component's leading values. Orthogonal partial least squares discriminant analysis (OPLS-DA) was also used to identify the metabolites factors that change before and after exercise using the R package ropls. A variable importance on projection (VIP) scores greater than 1.5 was used as the cut-off to select the most important metabolites responding to exercise for further analysis. To identify metabolites that were signi cantly changed by exercise, we used nonparametric Wilcoxson's signed-rank test. An estimate of the false discovery rate (FDR) was calculated to consider the multiple comparison that normally occurs in metabolomics-based studies, with FDR < 0.5 used as the cutoff for signi cance. For heatmap analysis, metabolites standardization was z-scaled by subtracting their means and followed by division by standard deviations. Relationship between saliva metabolites and fatigue VAS was analyzed by using a Pearson correlation coe cient, p < 0.05 for statistically signi cant. Table 1 shows subject characteristics and exhaustive exercise data. All subjects completed the exercise reaching an exhaustive fatigue state for an average duration of 1713 ± 191 s. Non-targeted metabolomics were applied to determine the 210 metabolites. PCA (Fig. 1) showed that principle components 1 and 2 captured together 40% of variance of the data, as well as the differential clustering before and after exercise. OPLS-DA modeling revealed a difference between pre-and post-exercise samples (R 2 Y = 0.929, Q 2 Y = 0.596) and identi ed 29 metabolites with VIP scores of 1.5 and higher (Table 2). These metabolites were attributed to carbohydrate metabolism, glycolysis, ketoacidosis, fatty acid metabolism, protein metabolism, and TCA cycle metabolism. These metabolites were differently clustered in the heatmap before and after exercise (Fig. 2). Sixteen metabolites were signi cantly increased post-exercise compared to pre-exercise after adjusted FDR (Table 2). Furthermore, the fatigue VAS was signi cantly correlated only with an increase in cyclohexylamine (Fig. 3).

Discussion
In this study, we investigated the impact of exhaustive exercise on salivary metabolites and the relationship between fatigue and metabolites. The main ndings of this study were as follows: rst, exhaustive exercise led to changes in metabolic pro le. Second, bioinformatic techniques including OPLS-DA analysis found that the 29 metabolites related to carbohydrate, fatty acid, protein, and TCA cycle metabolism were associated with the changes in metabolites after exercise. Furthermore, the changes in cyclohexylamine levels were signi cantly correlated with subjective fatigue. Therefore, these metabolites could play an important role in fatigue during exhaustive exercise.
The metabolomics approaches have revealed that acute exercise changed the metabolic pro le associated with adenosine triphosphate production, including glycolysis (e.g., pyruvate and lactate), fatty acid oxidation (e.g., palmitate), protein metabolism (e.g., alanine), and ketone bodies (e.g., hydroxybutyrate). [11.12] Consistent with previous studies, the present study found that exhaustive exercise increased these energy substrates related to carbohydrate, fatty acid, and amino acid metabolism ( Table 2). These metabolic responses were depicted in principal components of the PCA scores plot (Fig. 2) accounting for about 40% of the total variance. Therefore, these results suggest that exhaustive exercise acutely changes the metabolic pro le related to the energy production pathway.
Prolonged exhaustive exercise was associated with an increase in long chain fatty acids, medium chain fatty acids, fatty acid oxidation products, and ketone bodies. [13] Nieman et al. (2017) [14] have showed that these fatty acid metabolites changed after exhaustive running at 70% maximum oxygen uptake for a means duration of 2.26 h. Moreover, Manaf et al. (2018) [7] have found changes in metabolites related to fatty acid, indole, neurotransmitter substances, and amino acid pathways after exhaustive cycling at an intensity of 3 mmol/L of lactate for a means of 1.2 h. The present study investigates the metabolomic pro le response to a step-load cycling exercise until exhaustion (below 0.5 hours), a protocol that would be able to elicit central fatigue in a relatively short duration by high intensity exercise. [8] The results of the present study demonstrated that our exhaustive exercise changes the metabolomic pro le mainly associated with glycolysis (e.g., lactate, pyruvate, dihydroxyacetone phosphate, and 2-hydroxyvaleric acid) and ketoacidosis (e.g., 2-oxousovaleric acid and 4-methyl-2oxovaleric acid). Therefore, metabolite response may be related to the nature of exercise factors such as duration and intensity.
Metabolite pro le clustering was different before and after exhaustive exercise, and carbohydrate glucose metabolites showed increased VIP scores and fold changes compared to others (Table 2). In addition, TCA cycle intermediates, ketoacidosis, and some amino acid metabolites were associated with postexercise metabolic pro le changes, although most TCA cycle intermediates did not increase signi cantly. TCA cycle is the nal pathway of carbohydrate, lipid, and some amino acid to produce energy via oxidative phosphorylation mainly during aerobic exercise. Taken together, the exhaustive exercise in the present study would re ects mostly anaerobic glycolysis metabolism, rather than the aerobic oxidation pathway, although both pathways were included in the exercise protocol. During relatively high intensity exercise, glucose metabolism is facilitated and pyruvate and lactate were increased. Under these conditions, pyruvate, lactate, and amino acid (alanine) are involved in restoration of glucose metabolism in the liver. [15] Indeed, we observed that pyruvate, lactate, and alanine were increased post-exercise in this study, suggesting that the glucose metabolism pathway was accelerated at the end of the exercise.
In this study, the changes in fatigue VAS were correlated with an increase in cyclohexylamine within the VIP metabolites (Fig. 3). Cyclohexylamine is an acyclic aliphatic amine, and known as a metabolite produced by the arti cial sweetener cyclamate. [16] An early study has reported that ingestion of cyclohexylamine increased blood pressure and plasma free fatty acids, and that cyclohexylamine have a sympathomimetic capability. [17] Animal studies demonstrated that cyclohexylamine was deaminated to the corresponding ketones by microsomes in the rabbit liver. [18] Interestingly, metabolomics analysis in the present study could not detect cyclohexylamine at the baseline; however, cyclohexylamine increase was detected in several subjects after exercise. These ndings may imply that exhaustive exerciseinduced fatigue is in part mediated by cyclohexylamine. In this regard, further studies are necessary to elucidate a relationship between exercise and cyclohexylamine.
In conclusion, this study investigated the salivary metabolomic pro le following exhaustive exerciseinduced fatigue. We have identi ed energy-related metabolites that were signi cantly increased after exhaustive exercise. Our ndings show an increase in metabolites related to glycolysis, lipolysis, amino acid metabolism, and ketone bodies. Furthermore, changes in cyclohexylamine levels were associated with an increase in fatigue. This metabolomic pro le signature would serve as a pilot understanding of exercise induced fatigue.  The principal component analyses (PCA) scores plot pre-and post-exercise.

Figure 2
The dendrogram heatmap of variable importance on projection metabolites pre-and post-exercise.