Increased physical activity and improved cardiorespiratory fitness (CRF) are associated with improvements in individual cardiometabolic risk factors and global cardiovascular disease risk (1–5); however, the molecular mechanisms underlying the salutary effects of exercise remain unclear. Substantial heterogeneity in the individual response to exercise training exists (6–8); thus, insight into the biochemical pathways involved in exercise offers promise toward early detection of cardiometabolic disease (9) and tailoring exercise “prescriptions” to the individual level (10).
Metabolites are substrates and by-products of metabolism with a diverse set of biochemical actions that serve as both biomarkers and effectors of disease states (11,12). The metabolome’s interface between gene–protein functional state and phenotype, and its dynamic response to environmental stimuli makes it well suited for the study of physiologic perturbations such as exercise. Advancements in high-throughput platforms have enabled the systematic assessment of large numbers of small-molecule metabolites in tissue and plasma samples that may participate in exercise-induced biochemical pathways (12,13).
Work by our group and by others has characterized changes in plasma metabolites after a single aerobic exercise session using advanced metabolite profiling technologies (13–15). Findings from these studies demonstrate exercise-induced changes in several key metabolites involved in glycolysis, lipolysis, glycogenolysis, the citric acid (tricarboxylic acid (TCA)) cycle, and amino acid metabolism, and identified differential substrate utilization among more and less fit individuals. By contrast, far less is known about the metabolic response to prolonged exercise training. Huffman et al. (16) examined the change in 69 plasma metabolites after 6 months of chronic exercise training in a small (n = 53) cohort of sedentary, middle-age, overweight, or mildly obese men and women, and found significant changes in leptin, monocyte chemoattractant protein, and arachidoyl carnitine compared with the inactive control group. Notably, this study focused on free fatty acids, acylcarnitines, and a subset of amino acids. More recently, the same group used a larger, targeted metabolomics platform in skeletal muscle biopsies in an expanded sample, and found that a high amount of vigorous-intensity aerobic exercise significantly increased concentrations of even-, medium-, and long-chain acylcarnitines, whereas resistance training and a low amount of vigorous-intensity aerobic exercise preferentially increased short- and medium-chain acylcarnitines (17).
The purpose of this study was to characterize the effect of a prolonged exercise intervention on a diverse set of circulating plasma metabolites. We used a targeted liquid chromatography tandem mass spectrometry (LC-MS/MS) platform in peripheral blood samples before and after participation in a 6-month randomized controlled exercise trial among 216 middle-age, abdominally obese men and women. We tested whether metabolite changes differed between groups and if metabolites were associated with changes in cardiometabolic traits.
The study was approved by the Queen’s University Health Sciences Research Ethics Board, and all participants provided written informed consent. Details of the study design and methods have been published previously (18). Briefly, the 24-wk exercise intervention randomized 300 inactive, abdominally obese (waist circumference (WC) >102 cm for men, 88 cm for women) participants into one of four exercise groups: control (no exercise), low amount low intensity (LALI; 180 and 300 kcal per session for women and men, respectively, at 50% V˙O2peak), high amount low intensity (HALI; 360 and 600 kcal per session, respectively, at 50% V˙O2peak), and high amount high intensity (HAHI; 360 and 600 kcal per session, respectively, at 75% V˙O2peak). CRF was assessed using standard open-circuit spirometry during a maximal graded exercise test. WC and blood pressure were measured at baseline and at 16 and 24 wk. Fasting insulin (pmol·L−1), homeostatic model assessment insulin resistance (HOMA-IR), 2-h glucose (mmol·L−1), and insulin area under the curve (AUC) were derived from a 2-h, 75-g oral glucose tolerance test at baseline and at 16 and 24 wk.
In the current analysis, participants from the original trial were excluded if they did not complete the 24-wk intervention (n = 84), which resulted in a study sample of 216 participants (control, n = 53; HAHI, n = 47; HALI, n = 60; LALI, n = 56).
Plasma sampling and metabolite profiling
Fasting plasma samples were acquired via the antecubital vein at baseline before the oral glucose tolerance test and again 48 h after the last exercise session at 16 and 24 wk. The plasma was immediately separated by centrifugation (10 min at 4250 rpm) and stored at −80°C.
We used two distinct, LC-MS/MS–based methods to profile 147 analytes including amino acids, organic acids, bile acids, indoles, nucleotides, and sugars. These platforms have been previously used to characterize a diverse set of biochemical pathways implicated in metabolic status (19,20). Briefly, fasting plasma samples (EDTA, 10 and 30 μL for positive and negative ion modes, respectively) were deproteinized using extraction solvent containing stable isotope–labeled internal standards. Samples were vortexed and centrifuged, and aliquots were transferred to 2-mL autosampler vials with glass inserts for LC-MS analysis. In a positive mode, normal-phase hydrophilic interaction chromatography using a 2.1 × 150-mm, 3-μm Atlantis column (Waters) was coupled to a 4000 QTrap triple-quadrupole mass spectrometer (Applied Biosystems/Sciex) equipped with an electrospray ionization source for targeted detection of 78 metabolites using a dynamic multiple-reaction monitoring mechanism. In a negative mode, hydrophilic interaction chromatography chromatography using a 2.1 × 100-mm, 3.5-μm Xbridge Amide column (Waters) was coupled to an Agilent 6490 triple-quadrupole mass spectrometer equipped with an electrospray ionization source for targeted detection of 69 metabolites using dynamic multiple-reaction monitoring. Metabolite peak areas were integrated using Sciex MultiQuant software (positive mode) or Agilent Masshunter Quantitative software (negative mode). All metabolite peaks were manually reviewed for peak quality in a blinded manner. In addition, pooled plasma samples were interspersed within each analytic run at standardized intervals, enabling the monitoring and correction for temporal drift in MS performance. The median coefficient of variation in metabolites was 5.88% in the positive mode and 5.23% in the negative mode.
All metabolites were log transformed to approximate a normal distribution, and analyses were adjusted for age and sex. Repeated-measures ANOVA was used to compare metabolite changes within and between the control, LALI, HALI, and HAHI groups. In secondary analyses, principal component analysis (PCA) was used to reduce dimensionality of the metabolite data set at baseline. Thirty-seven main principal factors were derived after orthogonal Varimax rotation (Table 1). Factors were retained if they had an eigenvalue of greater than 1 and individual metabolites with a factor load of greater than |0.4| for a given PCA-derived factor were included. Scores were calculated for each participant based on standardized scoring coefficients. Cross-sectional associations between PCA factor scores and cardiometabolic risk traits (body weight, body mass index (BMI), WC, 2-h glucose, fasting insulin, insulin AUC, HOMA-IR, systolic blood pressure (SBP), diastolic blood pressure (DBP)) were assessed with multiple linear regression adjusted for age and sex. In addition, multiple linear regression was used to determine cross-sectional associations between individual metabolites and cardiometabolic risk traits, baseline, and change in metabolite concentrations, and change in cardiometabolic traits. All analyses were performed using SPSS Statistics (Version 23), and statistical significance was determined at the Bonferroni-corrected P value < 3.0 × 10−4 (0.05/147). Nominal significance was set at P < 0.05.
Participant characteristics before and after treatment are summarized in Table 2. Briefly, participants were abdominally obese, inactive, middle-age adults with a relatively healthy metabolic profile. Significant reductions in body weight, WC, fasting insulin, HOMA-IR, and insulin AUC, and increases in CRF were observed in the exercise groups compared with the control group (P < 0.05).
Baseline associations between metabolites and cardiometabolic traits
PCA was used to reduce the large number of circulating metabolites into fewer factors containing highly correlated metabolites that are biologically related (Table 1). We detected strong associations between several principal factors, cardiometabolic risk factors, and CRF before the initiation of exercise (Table 3). Both PCA factor and individual metabolite associations revealed several well-established relationships between metabolic pathways and clinical traits, including body mass (tryptophan derivatives, aromatic amino acids, glutamate, uric acid, and adenine catabolism products) and glucose and insulin homeostasis (branched-chain amino acids, aromatic amino acids, tryptophan derivatives, and glutamate) (21–23). The hexosamine end-product, uridine diphosphate N-acetylglucosamine (UDP-GlcNAc), was also strongly and positively related to baseline levels of fasting and 2-h insulin, insulin AUC, and HOMA-IR. In addition, we identified novel inverse relationships between the aromatic amino acid, tyrosine (P = 2.9 × 10−4), tryptophan precursor and anthranilic acid (P = 1.5 × 10−4), and the adenine nucleotide catabolism product, inosine (P = 2.1 × 10−4), and directly measured CRF.
Change in metabolites at 24 wk
There were no significant metabolite changes between the exercise and control groups after 24 wk at the Bonferroni-adjusted threshold (P < 3.0 × 10−4). When exercise arms were collapsed to create a single-exercise group (n = 163), we found seven metabolites that changed at nominal significance (P < 0.05) compared with the control group. These included tryptophan metabolites ((indoxylsulfate (increased) and indole-3-lactic acid (increased)); metabolites derived from energy metabolism pathways ((aconitic acid (decreased), pyruvic acid (decreased), adenosine triphosphate (increased), malonic acid (increased)), and the purine degradation product xanthine (decreased). Changes in indole-3-lactic acid, indoxylsulfate, pyruvic acid, and xanthine seem to be driven by the high-amount exercise groups upon examination of within-group changes. We further identified several metabolites that changed from baseline levels within the exercise groups at the Bonferroni-adjusted level of significance. Products of adenine nucleotide metabolism (cyclic AMP, inosine, xanthine, and xanthosine); lipolysis (glycerol); the nucleotide sugar, UDP-GlcNAc; and nonpolar amino acids and intermediates (alanine and homogentisic acid) decreased, whereas serine and acetoacetic acid, a fatty-acid intermediate, both increased in HALI. Xanthine and indole-3-lactic acid also decreased and increased in LALI and HAHI, respectively.
Metabolite changes associated with changes in cardiometabolic traits
We detected strong associations between analyte changes and changes in cardiometabolic traits (P < 3 × 10−4; Table 4). Changes in the branched-chain amino acids leucine and isoleucine, and UDP-GlcNAc were inversely associated with changes in CRF (mL·kg−1·min−1), whereas change in the TCA cycle intermediate, isocitric acid, was positively correlated. These changes stayed significant even after adjustment for known relationships (BMI, fasting glucose). Changes in gluconeogenic amino acids (alanine, tyrosine, and proline) and UDP-GlcNAc were positively associated with change in BMI; UDP-GlcNAc was similarly associated with WC and insulin metabolism (insulin and insulin AUC). Xanthurenic acid was also associated with change in insulin AUC. These relationships remained significant after further adjusting for change in weight.
Changes in metabolites that were associated with changes in cardiometabolic traits at nominal significance are reported (see Table, Supplemental Digital Content 1, Associations between change in metabolites and change in cardiometabolic traits, http://links.lww.com/MSS/B221).
Baseline metabolites associated with change in cardiometabolic traits
We detected baseline metabolite levels measured before the initiation of exercise that predicted a favorable cardiometabolic trait response after 24 wk of chronic exercise, even after adjusting for the baseline clinical trait (Table 5). Baseline taurine and asymmetric dimethylarginine (ADMA)/symmetric dimethylarginine (SDMA) concentrations were inversely associated with changes in WC and 2-h glucose, respectively, whereas glutamate was positively associated with change in DBP (P < 3 × 10−4). Additional associations between baseline metabolites and change in cardiometabolic risk factors and CRF at nominal significance are reported (see Table, Supplemental Digital Content 2, Associations between baseline metabolites and change in cardiometabolic risk factors, http://links.lww.com/MSS/B222).
Advancements in metabolomics technologies have enabled efforts to systematically profile blood-based metabolites to characterize changes predictive (13,24) or reflective (9) of physiologic states. Here, we used a targeted LC-MS/MS platform to describe the changes in a large and varied sets of plasma metabolites in a cohort of 216 abdominally obese subjects randomized to 24 wk of inactivity or one of three aerobic exercise interventions. We found no significant differences between metabolites in the control and exercise arms at our a priori level of statistical significance. Products of lipolysis, adenine nucleotide metabolism, and both polar and nonpolar amino acids changed within individual exercise groups, albeit not differently from control. We confirmed cross-sectional relationships between metabolites and cardiometabolic risk traits. In addition, we demonstrated associations between changes in metabolite level and changes in these traits.
Prior metabolomics investigations into the effects of chronic exercise training were either limited by a small sample size and less expansive metabolite platform (16), or characterized changes in skeletal muscle rather than blood (17). Huffman et al. (16) randomized participants into exercise groups varying in amount and intensity and a control group and observed only three nominally significant metabolite changes between groups. The same group also studied changes in skeletal muscle metabolite concentrations after exercise training and found nominally (P < 0.05) significant increases in even-chain acylcarnitines and TCA cycle intermediates, whereas skeletal muscle concentrations of amino acids, glycolytic metabolites (lactate and pyruvate), and all other TCA cycle intermediates did not change (17).
Similarly, our group did not find significant differences in metabolites between the exercise and control groups at the Bonferroni-adjusted level of significance (P < 4 × 10−4) but did detect significant metabolite changes within groups. Several potential explanations exist for these findings. First, the use of a strict statistical threshold may have masked biologically relevant associations, particularly when several of the metabolites under scrutiny cluster together and participate in known biologic pathways (11). Second, the presence of significant interindividual variability may have blunted group comparisons. Within-group analysis using each individual as his/her own biologic control revealed significant changes within the exercise intervention arms, an experimental design that has been previously used in small metabolomics (9,13,16) and proteomic (25) pertubational studies. Although we cannot attribute exercise effects to the changes seen within individual exercise arms, our findings underscore the importance of performing much larger trials that are better powered to characterize the molecular response to physical activity (26).
The cross-sectional associations between baseline metabolites and clinical risk factors demonstrated here extend previous findings. Elevations in aromatic (tyrosine) and branched-chain amino acids, and tryptophan breakdown products (anthranilic acid, kynurenines) have been strongly correlated with elevated body mass and insulin resistance phenotypes (23,27,28). In addition, we observed novel inverse associations between tyrosine, anthranilic acid, and inosine, and peak oxygen consumption (V˙O2peak), which is an established predictor of morbidity and mortality.
We further explored whether improvements in cardiometabolic traits were related to changes in circulating metabolite levels. For instance, decreased UDP-GlcNAc concentrations were associated with improvements in anthropometric and insulin resistance traits, whereas increased levels were associated with worse CRF. UDP-GlcNAc, a nucleotide sugar, is the end-product of the hexosamine biosynthetic pathway and acts as a substrate for the O-linked N-acetylglucosamine transferase enzymes (OGT). Chronically elevated glucose concentrations are associated with increased UDP-GlcNAc concentrations via flux through the hexosamine biosynthetic pathway, augmenting O-GlcNAcylation of numerous proteins across the body and altering their function (29). Aberrant OGT activity has been implicated in insulin resistance, dyslipidemia, and diabetic cardiomyopathy (30–32), and increases in a graded fashion with increased UDP-GlcNAc concentrations. In addition, increased O-GlcNAc mitochondrial protein expression within myocardium was seen among rats selected for low capacity running in comparison to high-capacity runners, suggesting a role for OGT activity in aerobic capacity (33). Investigation into the effects of chronic exercise on OGT activity has been limited to myocardial tissue in different mouse models undergoing heterogeneous exercise interventions, and its role is not well understood (29). Our findings validate UDP-GlcNAc’s relationship to metabolic health traits and further suggest its potential role as either a marker or an effector of cardiorespiratory status.
Lastly, we explored whether baseline metabolite levels could predict exercise-induced response in a given cardiometabolic trait and identified three significant associations. Baseline glutamate levels were positively associated with change in DBP. This extends previous cross-sectional findings, wherein glutamate was positively associated with a range of cardiometabolic risk factors (insulin, HOMA-IR, SBP, and DBP) (23). ADMA/SDMA levels were inversely related to changes in 2-h glucose, a finding previously supported (34). In addition, baseline taurine (a sulfated organic compound) was negatively associated with change in WC. These findings are consistent with previous observations investigating the relationship between taurine and obesity. In animal studies, a decrease in blood taurine concentration was associated with obesity (35) and taurine treatment reduced inflammatory processes in adipose tissue via reduction in macrophage infiltration. Our observations extend prior literature by suggesting that higher plasma taurine concentrations may predict a more favorable exercise response for abdominal obesity.
A limitation of our study is that the sample was relatively homogeneous, which may have attenuated the strength of observed relationships. However, given that nearly 40% of US and Canadian adults are abdominally obese (36,37), our findings are relevant to this population. Furthermore, although we measured a diverse panel of ~150 metabolites in plasma, our analysis was targeted and not comprehensive of the entire plasma metabolome.
In summary, our study provides additional details on the metabolic changes induced by chronic exercise training, in addition to potential predictors of changes in metabolic traits. These data motivate additional studies in larger, heterogeneous cohorts.
This study was supported by the Canadian Institutes of Health Research (Grant OHN-63277). R. E. G. received support from the National Institutes of Health Molecular Transducers of Physical Activity Consortium.
The authors have no conflicts of interest to disclose. The results of the present study do not constitute endorsement by the American College of Sports Medicine. The results of the study are presented clearly, honestly, and without fabrication, falsification, or inappropriate data manipulation.
1. Leon AS, Connett J, Jacobs DR, Rauramaa R. Leisure-time physical activity levels and risk of coronary heart disease and death. The Multiple Risk Factor Intervention Trial. JAMA
2. Myers J, Prakash M, Froelicher V, Do D, Partington S, Atwood JE. Exercise
capacity and mortality among men referred for exercise
testing. N Engl J Med
3. Kraus WE, Houmard JA, Duscha BD, et al. Effects of the amount and intensity of exercise
on plasma lipoproteins. N Engl J Med
4. Janiszewski PM, Ross R. The utility of physical activity in the management of global cardiometabolic risk. Obesity (Silver Spring)
. 2009;17(3 Suppl):S3–14.
5. Ross R, Blair SN, Arena R, et al. Importance of assessing cardiorespiratory fitness in clinical practice: a case for fitness as a clinical vital sign: a Scientific Statement From the American Heart Association. Circulation
6. Bouchard C, Rankinen T. Individual differences in response to regular physical activity. Med Sci Sports Exerc
. 2001;33(6 Suppl):S446–51.
7. Bouchard C, Blair SN, Church TS, et al. Adverse metabolic response to regular exercise
: is it a rare or common occurrence? PLoS One
8. de Lannoy L, Clarke J, Stotz PJ, Ross R. Effects of intensity and amount of exercise
on measures of insulin and glucose: analysis of inter-individual variability. PLoS One
9. Sabatine MS, Liu E, Morrow DA, et al. Metabolomic identification of novel biomarkers of myocardial ischemia. Circulation
10. Neufer PD, Bamman MM, Muoio DM, et al. Understanding the cellular and molecular mechanisms of physical activity–induced health benefits. Cell Metab
11. Lewis GD, Asnani A, Gerszten RE. Application of metabolomics
to cardiovascular biomarker and pathway discovery. J Am Coll Cardiol
12. Shah SH, Kraus WE, Newgard CB. Metabolomic profiling for the identification of novel biomarkers and mechanisms related to common cardiovascular diseases: form and function. Circulation
13. Lewis GD, Farrell L, Wood MJ, et al. Metabolic signatures of exercise
in human plasma. Sci Transl Med
14. Muhsen Ali A, Burleigh M, Daskalaki E, Zhang T, Easton C, Watson DG. Metabolomic profiling of submaximal exercise
at a standardised relative intensity in healthy adults. Metabolites
15. Enea C, Seguin F, Petitpas-Mulliez J, et al. (1)H NMR-based metabolomics
approach for exploring urinary metabolome modifications after acute and chronic physical exercise
. Anal Bioanal Chem
16. Huffman KM, Slentz CA, Bateman LA, et al. Exercise
-induced changes in metabolic intermediates, hormones, and inflammatory markers associated with improvements in insulin sensitivity. Diabetes Care
17. Huffman KM, Koves TR, Hubal MJ, et al. Metabolite signatures of exercise
training in human skeletal muscle relate to mitochondrial remodelling and cardiometabolic fitness. Diabetologia
18. Ross R, Hudson R, Stotz PJ, Lam M. Effects of exercise
amount and intensity on abdominal obesity and glucose tolerance in obese adults: a randomized trial. Ann Intern Med
19. Kimberly WT, O’Sullivan JF, Nath AK, et al. Metabolite profiling identifies anandamide as a biomarker of nonalcoholic steatohepatitis. JCI Insight
. 2017;2(9): e92989.
20. Wang TJ, Larson MG, Vasan RS, et al. Metabolite profiles and the risk of developing diabetes. Nat Med
21. Newgard CB, An J, Bain JR, et al. A branched-chain amino acid–related metabolic signature that differentiates obese and lean humans and contributes to insulin resistance. Cell Metab
22. Pedersen ER, Tuseth N, Eussen SJ, et al. Associations of plasma kynurenines with risk of acute myocardial infarction in patients with stable angina pectoris. Arterioscler Thromb Vasc Biol
23. Cheng S, Rhee EP, Larson MG, et al. Metabolite profiling identifies pathways associated with metabolic risk in humans. Circulation
24. Shah SH, Bain JR, Muehlbauer MJ, et al. Association of a peripheral blood metabolic profile with coronary artery disease and risk of subsequent cardiovascular events. Circ Cardiovasc Genet
25. Ngo D, Sinha S, Shen D, et al. Aptamer-based proteomic profiling reveals novel candidate biomarkers and pathways in cardiovascular disease. Circulation
26. National Institutes of Health. Molecular transducers of physical activity in humans. 2017 [cited 2017 Aug 15]. Available from: https://commonfund.nih.gov/moleculartransducers/overview
27. Ho JE, Larson MG, Ghorbani A, et al. Metabolomic profiles of body mass index in the Framingham Heart Study reveal distinct cardiometabolic phenotypes. PLoS One
28. Guasch-Ferré M, Hruby A, Toledo E, et al. Metabolomics
in prediabetes and diabetes: a systematic review and meta-analysis. Diabetes Care
29. Myslicki JP, Belke DD, Shearer J. Role of O
-GlcNAcylation in nutritional sensing, insulin resistance and in mediating the benefits of exercise
. Appl Physiol Nutr Metab
30. Yang X, Ongusaha PP, Miles PD, et al. Phosphoinositide signalling links O
-GlcNAc transferase to insulin resistance. Nature
31. Hu Y, Belke D, Suarez J, et al. Adenovirus-mediated overexpression of O
-GlcNAcase improves contractile function in the diabetic heart. Circ Res
32. Clark RJ, McDonough PM, Swanson E, et al. Diabetes and the accompanying hyperglycemia impairs cardiomyocyte calcium cycling through increased nuclear O
-GlcNAcylation. J Biol Chem
33. Johnsen VL, Belke DD, Hughey CC, et al. Enhanced cardiac protein glycosylation (O-GlcNAc) of selected mitochondrial proteins in rats artificially selected for low running capacity. Physiol Genomics
34. Stühlinger MC, Abbasi F, Chu JW, et al. Relationship between insulin resistance and an endogenous nitric oxide synthase inhibitor. JAMA
35. Tsuboyama-Kasaoka N, Shozawa C, Sano K, et al. Taurine (2-aminoethanesulfonic acid) deficiency creates a vicious circle promoting obesity. Endocrinology
36. Janssen I. The public health burden of obesity in Canada. Can J Diabetes
37. National Center for Health Statistics. Health, United States, 2016: With Chartbook on Long-term Trends in Health
. Hyattsville (MD); 2017. pp. 2–466.