1 Introduction

Although physical activity has been proven to be substantially beneficial to human health (Ojiambo 2013), various negative effects including cardiovascular dysfunction (Webner et al. 2012), muscle damage (Howatson et al. 2010), increased propensity to upper respiratory tract infections (Robson-Ansley et al. 2012) and severe inflammation (Bonasia et al. 2015) have been associated with running extensive distances, such as that of endurance races. Even though these races have become increasingly popular, limited research is based on the elucidation of the effects of these races using a metabolomics approach. Metabolomics is defined as the identification and quantification of the small metabolites (< 1500 Da) present in a biological system in order to determine the physiological effects induced by a specific perturbation (Heaney et al. 2017). Since metabolites are typically the end-products of the genome, transcriptome and proteome, alterations in these are indicative of the overall physiological state of the investigated biological system (Heaney et al. 2017).

Previous metabolomics studies have indicated elevated concentrations of various carbohydrate/glycolysis intermediates, which suggest the utilization of free glucose as the preferred energy source during strenuous physical activity (Lewis et al. 2010; Salway 2012; Waśkiewicz et al. 2012). Furthermore, significant alterations to the tricarboxylic acid (TCA) cycle intermediates were induced by a marathon (Turer et al. 2014) and could be attributed to additional strain placed on the electron transport chain (ETC), causing an imbalanced NADH:NAD+ ratio (Esterhuizen et al. 2017). According to previous work (Stellingwerff 2012), free glucose and other carbohydrate stores can become depleted within approximately 90 min after the start of the race, which most likely lead to the utilization of alternative fuel substrates (lipids and amino acids) for energy production (Waśkiewicz et al. 2012). Increased lipolysis activity results in elevated serum glycerol and free fatty acids (Lewis et al. 2010; Waśkiewicz et al. 2012), the latter of which produce acetyl-CoA via β-oxidation and subsequent energy via the TCA cycle and ETC (Salway 2012). Furthermore, the increased synthesis of acetyl-CoA could also ascribe for the elevated ketone concentrations previously reported (Pechlivanis et al. 2010), as it is a key component of ketogenesis. In the event that the strenuous physical activity continues beyond the capacity of the athlete’s lipid stores, or if the traditional lipid oxidation pathways become saturated (Staron et al. 1989), the athlete’s metabolism shifts towards protein catabolism (resulting in reduced blood amino acid levels) in an attempt to synthesize the energy required to complete the marathon (Lewis et al. 2010). These amino acids are primarily oxidized to pyruvic acid and acetyl-CoA, both of which can serve as TCA cycle influx substrates for energy production (Salway 2012). Additionally, protein degradation has been shown to alter purine catabolism, resulting in elevated adenosine-monophosphate, inosine-monophosphate, hypoxanthine, xanthine, uric acid and allantoin (Turer et al. 2014), the latter of which is a uric acid derivative and a surrogate index of oxidative stress (Lewis et al. 2010).

Although these metabolomics studies provide some clues to the metabolic alterations that occur during strenuous physical activity such as endurance races, very few of these employed an untargeted metabolomics approach. Considering this, an untargeted two-dimensional gas chromatography time-of-flight mass spectrometry (GCxGC-TOFMS) metabolomics approach was used to holistically compare the serum metabolite profiles of 31 recreational marathon athletes before and after the completion of a marathon (42 km), in order to better characterize the acute metabolic changes induced by exercise stress.

2 Materials and methods

2.1 Participants

All participants completed a health and dietary questionnaire (including a menstrual cycle questionnaire for female participants) prior to the marathon in order to assess their eligibility. Individuals with food allergies, cardiovascular complications, musculoskeletal disorders/injuries, or those receiving anti-inflammatory treatment were excluded from the study. Athlete participation in this investigation was completely voluntary, and all the participants gave written and informed consent. A summary of the participant characteristics is presented in Table 1.

Table 1 Summary of the participant demographical information
Full size table

2.2 Clinical samples

Blood samples were collected by antecubital fossa venesection of 31 marathon athletes (19 males and 12 females) 24 h before and immediately after completing the Druridge Bay Marathon (Northumberland, UK). Pre-marathon samples were collected the day preceding the race (between 10 am and 6 pm) as a means of reducing additional metabolic changes induced by the venesection stress as well as to limit interference to the athletes’ pre-marathon regimens. The individuals were required to be in a hydrated, yet fasted state (for a minimum of 2 h) at time of baseline sample venesection. The samples were collected in standard 10 mL vacutainer vials, placed on ice and transported to the laboratory (Faculty of Health and Life Sciences, Department of Sport, Exercise and Rehabilitation at the Northumbria University in Newcastle upon Tyne, UK) for immediate processing. Briefly, the blood was allowed to clot for 30 min and centrifuged at 3000×g for 10 min. The supernatant (serum) was then extracted and immediately frozen (− 80 °C) before being transported (on dry ice) to the North-West University, Human Metabolomics: Laboratory of Infectious and Acquired Diseases. These serum samples were stored at − 80 °C until metabolomics analyses commenced.

2.3 Chemicals and reagents

Methoxamine hydrochloride, 3-phenylbutyric acid and bis(trimethylsilyl)-trifluoroacetamide (BSTFA) with 1% trimethylchlorosilane (TMCS) were purchased from Sigma-Aldrich (St. Louis, Missouri, USA) and the pyridine was from Merck (Darmstadt, Germany). The acetonitrile used, was an ultra-pure Burdick and Jackson brand (Honeywell International Inc., Muskegon, Michigan, USA) and hence did not require any further purification.

2.4 Sample extraction and derivatization

A total metabolome extraction was performed on all the collected serum samples, along with appropriate quality control (QC) samples. During this process, 50 µL of internal standard (3-phenylbutyric acid dissolved in a chloroform:methanol:water [1:3:1] solution; 0.45 μg/mL) was added to 50 µL of each serum sample. Hereafter, the samples were placed on ice before 300 µL of acetonitrile was added as a protein precipitation agent. Samples were then subjected to the REAX D-91126 vortex (Heidolph Instruments GmbH & Co.KG, Schwabach, Germany) for 2 min and centrifuged at 3500×g for 10 min at 4 °C. The supernatant of the biphasic solution was then transferred to a clean GC–MS sample vile and dried at 40 °C under a light stream of nitrogen gas for approximately 45 min. Hereafter, 25 µL of methoxamine hydrochloride dissolved in pyridine (15 mg/mL) was added to each sample and incubated at 50 °C for 90 min. Thereafter, 40 µL BSTFA with 1% TMCS was added, followed by derivatization for 60 min at 60 °C. Each derivatized sample was transferred to a new GC–MS vile containing a vile insert, and capped.

2.5 GCxGC-TOFMS analysis and processing

The derivatized samples were placed on a multi-purpose auto-sampler tray (Gerstel GmbH and co. KG, Mülheim van der Ruhr, Germany) in a randomized order, and analyzed using a Pegasus 4D GCxGC-TOFMS system (LECO Africa (Pty) Ltd, Johannesburg, South Africa), fitted with an Agilent 7890A GC and TOFMS (LECO Africa). During analyses, 1 µL of each sample was injected using a 1:3 split ratio. Purified helium was used as a carrier gas and set at a constant flow rate of 1 mL/min, while the injector temperature was set to operate at a constant 270 °C throughout the entire sample analysis. The primary oven was fitted with a Restek Rxi-5MS capillary column (30 m; 0.25 µm diameter and 0.25 µm film thickness) and programmed to start the run at an initial temperature of 70 °C for 2 min, followed by an increase of 4 °C/min until a final temperature of 300 °C was reached and maintained for 2 min. The secondary oven, equipped with a Restek Rxi-17 capillary column (1 m; 0.25 µm diameter and 0.25 µm film thickness), was set at an initial temperature of 85 °C, which was increased by 4.5 °C/min until a final temperature of 300 °C was reached and maintained for 2 min. The thermal modulator was set to pulse streams of cold and hot nitrogen gas every 3 s, for 0.5 s. The detector was set to disregard all mass spectra information for the first 400 s of each run to exclude solvent detection; however, this was still included on the time axis of the primary column to reflect accurate retention times. Additionally, the transfer line and ion source were respectively held at a constant of 270 °C and 220 °C for the entire run, with a detector voltage of 1600 V and filament bias of − 70 eV. Mass spectra were acquired at a rate of 200 ms/s, over a range of 50–800 m/z. The total run time per sample was 111.28 min. Following GCxGC-TOFMS analysis, 838 peaks were identified, which was processed using LECO Corporation’s ChromaTOF software (version 4.32), as described by Luies and Loots (2016).

2.6 Statistical analyses

Prior to statistical analysis, the dataset was normalized in relation to the internal standard and subjected to several “clean-up steps”, including a 50% zero filter, QC drift correction, QC coefficient of variation filter and zero value replacement (Fernandez et al. 2000; Luies and Loots 2016). Hereafter, a natural shifted log transformation was performed to correct for skewed variable distribution, as well as auto scaling to align all variables (Van den Berg et al. 2006) (exclusively during multivariate analysis).

Both multivariate and univariate statistical analyses were performed using MATLAB software (2012), adjusted with the PLS toolbox (2016), to identify the metabolite markers best describing the variation between the comparative groups. Multivariate approaches included principle component analysis (PCA) to determine whether a natural differentiation occurred between the comparative groups, and partial least square-discriminant analysis (PLS-DA) to characterize the group separation (Luies and Loots 2016). The univariate data analyses included a Wilcoxon signed rank test, corrected for multiple testing by the Benjamini–Hochberg procedure, to assess the statistical significance of each compound (Benjamini and Hochberg 1995; Rosner et al. 2006), while an independent effect size calculation (Wilcoxon signed effect size) was used to assess the practical significance of each detected metabolite (Tomczak and Tomczak 2014).

3 Results

The PCA differentiation between serum metabolite profiles (Fig. 1) of the marathon athletes before and after the completion of the marathon was clearly defined. The total amount of variance explained by the first three principle components (PCs) (R2X cum) was 53%, of which PC1 accounted for 24%, PC2 for 20% and PC3 for 9%. Additionally, the PLS-DA model (results not shown) showed a modelling parameter R2Y (cum) of 48.81% for the total variance of the response Y, and a Q2 (cum) of 83.37%, indicating the cross-validation variation due to the response Y.

Fig. 1
figure 1

Principle component analysis scores plot showing clear differentiation of the serum samples of marathon athletes before (denoted by circles) and after (denoted by squares) the completion of a marathon. The variance accounted for are indicated in parenthesis. PC principle component

Full size image

Since the aim of this study was to attain a holistic view of the altered human serum metabolome induced by a marathon, compounds with a PCA power value ≥ 0.5 or Wilcoxon p-value ≤ 0.017 (BH-critical value) or an effect size ω-value ≥ 0.3 were considered significant and interpreted based on their associated metabolic/biochemical pathways. This multi-statistical approach yielded an initial list of 78 metabolite markers, of which 70 metabolite markers could be annotated by comparison of their mass spectra and retention times to that of commercially available and in-house libraries developed from previously injected standards (see Table 2).

Table 2 The significant serum metabolite markers best describing the variation between the pre- and post-marathon groups, listed alphabetically
Full size table

4 Discussion

The altered metabolite markers listed in Table 2 are indicative of the major metabolic pathways affected by the marathon and are mainly associated with the macro-fuel substrate utilization pathways (carbohydrates, lipids and amino acids) and the regulation thereof (TCA, oxidative phosphorylation [OXPHOS] and gut microbiome). These metabolite pathways, along with the intermittent dietary-associated metabolite markers, are comprehensively discussed below and schematically presented in Fig. 2.

Fig. 2
figure 2

A schematic representation of the altered serum metabolome induced by a marathon. The altered metabolites are either donated as increased (↑) or decreased (↓) relative to the pre-marathon group. FAD flavin adenine dinucleotide, FADH flavin adenine dinucleotide + hydrogen, NAD nicotinamide adenine dinucleotide, NADH nicotinamide adenine dinucleotide + hydrogen, ATP adenosine triphosphate, ADP adenosine diphosphate

Full size image

Various carbohydrate metabolites were significantly elevated following the marathon and can be ascribed to gluconeogenic influx (MacLaren and Morton 2012) and a reduced insulin secretion (Richter et al. 1992), which is typically induced by an initial depletion of glucose and glycogen stores occurring approximately 90 min after the start of a marathon (exercise intensity dependent) at a VO2max > 75% (Stellingwerff 2012). The reduced insulin concentrations temporarily inactivate insulin-dependent glucose uptake systems (i.e. GLUT 4 transporters and glucokinase) (Salway 2012), preventing glucose absorption into cells. This could ascribe for the post-marathon elevations in serum glucose concentrations and various other associated metabolites, including glucaric acid (Żółtaszek et al. 2008) and mannose (Hu et al. 2016) as well as the elevated concentrations of gluconeogenesis-associated metabolites, i.e. myo-inositol (Eisenberg and Parthasarathy 1987), erythritol (synthesized via erythrose-4-phosphate in pentose phosphate pathway) (Hootman et al. 2017), glycerol and glyceric acid (Salway 2012; Wadman et al. 1976). Furthermore, the elevated concentrations of pyruvic acid observed in the post-marathon samples were anticipated since it is an end-product of the glycolysis pathway which feeds into the TCA cycle for further ATP production (Salway 2012). This is confirmed by the accumulation of various TCA cycle intermediates such as α-ketoglutaric acid, succinic acid, citric acid, fumaric acid and malic acid (Qiang 2015), which also indicate the accumulation of circulating NADH/FADH2 molecules (Esterhuizen et al. 2017) as a result of a saturated ETC activity. It is also important to mention that many of the aforementioned carbohydrate metabolites, along with elevated concentrations of mannose (Hu et al. 2016), sorbose (Guzik and Stachowicz 2016), mannitol (McNutt 2000), tagatofuranose (Kroger et al. 2006), and threonic acid (an ascorbic acid derivative) (Simpson and Ortwerth 2000), are well-known constituents of fruits/fruit juices, vegetables/vegetable juices, peanuts, energy bars, energy drinks and various other ergogenic aids consumed by athletes during the course of a marathon (Jeukendrup 2011; Pfeiffer et al. 2012).

The reduction in intracellular glucose due to the aforementioned cellular uptake inhibition is known to activate lipolysis of free and adipose tissue-bound triacylglycerol (TAG) (MacLaren and Morton 2012) as alternative fuel substrates. This is substantiated by the elevated concentrations of serum glycerol, monopalmitin and various free fatty acids (lauric acid, palmitic acid, palmitoleic acid, 11-eicosenoic acid, 11,14-eicosadienoic acid, myristoleic acid, α-linolenic acid, 5-dodecenoic acid, linoleic acid and oleic acid) (Kujala et al. 2013; Lewis et al. 2010; Peake et al. 2014; Waśkiewicz et al. 2012). Additionally, accumulated 3-hydroxy acids (β-hydroxyhexanoic acid) and 3-keto acids (β-hydroxy-α,β-didehydrosebacic acid) are indications of a saturated β-oxidation pathway, ascribed to the inhibition of the rate-limiting enzyme, β-hydroxyacyl dehydrogenase, which is pursued by 3-ketoacyl-CoA thiolase. This saturated β-oxidation pathway results in the catabolism of the accumulated fatty acids via α-oxidation (Roe and Ding 2001), thus substantiating the elevated concentrations of α-hydroxyoctanoic acid (C8; an α-oxidation intermediate) and the odd-chain fatty acids (OCFA; tridecanoic acid [C13], pentadecanoic acid [C15], heptadecanoic acid [C17] and 10-heptadecenoic acid [C17:1]) detected in the post-marathon serum. It should, however, be mentioned that these OCFA may also arise from elevated dietary intake (Jenkins et al. 2017) and/or autophagy of various cellular constituents during extensive energy-requiring states (Singh and Cuervo 2011). Nevertheless, irrespective of their origins, these OCFAs are ultimately catabolized to propionyl-CoA (Pfeuffer and Jaudszus 2016), hence the elevated β-hydroxypropionic acid observed in the post-marathon serum. The elevated concentrations of β-hydroxybutyric acid and acetoacetic acid are anticipated, as these are alternative fuel substrates for the brain (Cahill and Vech 2003) and skeletal muscles (Holloszy and Coyle 1984) in hypoglycemic states, and could also be an indication of an imbalanced redox state (Esterhuizen et al. 2017; Salway 2012). Furthermore, the post-marathon elevations of malonic acid typically indicate the accumulation of malonyl-CoA, which is a long-chain fatty acid (LCFA) transport inhibitor (Salway 2012) and could therefore be an additional reason for the increased cytosolic LCFAs (palmitic acid, palmitoleic acid, 11-eicosenoic acid, 11,14-eicosadienoic acid, myristoleic acid, α-linolenic acid, linoleic acid and oleic acid).

Most amino acids are catabolized into TCA cycle substrates via propionyl-CoA, succinyl-CoA, pyruvic acid or acetyl-CoA, depending on the specific amino acid (Salway 2012). Reduced concentrations of amino acids (serine, glycine, alanine, aspartic acid, phenylalanine, tyrosine, threonine and methionine) and altered amino acid-associated metabolite concentrations [dimethylglycine (Holm et al. 2003), pyroglutamic acid (Kumar and Bachhawat 2012), indole-3-acetic acid (Salway 2012) and glutaric acid (Peake 2016)] were detected in the post-marathon serum, which indicate amino acid catabolism during the marathon. Furthermore, elevated concentrations of α-hydroxybutyric acid (an intermediate in the threonine/methionine pathway), further supports the aforementioned NADH:NAD+ imbalance, ketoacidosis, reduced insulin secretion and impaired glucose absorption (Gall et al. 2010). Additionally, elevated p-hydroxyphenylacetic acid and p-hydroxyphenyllactic acid could be indicative of mild liver injury/damage (Ghoraba et al. 2014; Liebich and Pickert 1985), which concur with previous findings (Jastrzebski et al. 2015; Lippi et al. 2011).

In accordance with previous findings (Pechlivanis et al. 2010), reduced serum concentrations of the branched-chain amino acids (BCAAs) isoleucine (0.05 vs 0.02 µg/mL, p = 0.018), valine and leucine were detected in the post-marathon samples, while their catabolism intermediates i.e. α-ketoisovaleric acid, β-hydroxyisobutyric acid (valine metabolites) and α-ethylhyrdracrylic acid (allo-isoleucine metabolism) (Wendel et al. 1989), were significantly elevated. Allo-isoleucine catabolism intermediates may serve as alternative substrates in the valine catabolism pathway (Ryan 2015), however, since the valine catabolism intermediates remained elevated in the post-marathon serum it can be deduced that allo-isoleucine is instead catabolized to α-ethylhydracrylic acid (Korman et al. 2005) via α-keto-β-methylvaleric acid (explaining its reduced concentrations post-marathon) with the subsequent production of FADH2 (Korman et al. 2005). The elevated concentrations of β-hydroxyisovaleric acid detected post-marathon is typically associated with ketone rich environments (as observed in the current investigation), resulting from increased leucine catabolism and the subsequent production of isovaleryl-CoA`23 (Mock et al. 2011). This metabolite is also a well-known constituent of athlete supplementation (Brioche et al. 2016) and could be elevated due to dietary ingestion during the marathon. Furthermore, reduced concentrations of these BCAAs, in particular leucine, results in mammalian target of rapamycin complex 1 (mTOR1) inhibition (Laplante and Sabatini 2009), which in turn activates various catabolic processes such as autophagy (of organelle and plasma membrane constituents) to release additional embedded fuel substrates (Singh and Cuervo 2011). mTOR1 inhibition can also be induced by other factors including elevated 5′-AMP-activated protein kinase during energy deprivation, reduced oxygen levels, reduced essential amino acids and inflammation (Laplante and Sabatini 2009), all of which are associated with endurance races and evidently occur in the marathon athletes investigated in this study. Considering that autophagy contributes to elevated amounts of cellular debris, the aforementioned accumulation of fatty acids (especially the OCFAs) in the post-marathon serum may also be explained by the autophagosomal degradation of phospholipids, sphingolipids (Kishimoto et al. 1973; Maes et al. 1996) and phytosphingosines (Kitamura et al. 2017; Kondo et al. 2014) found in cell and organelle membranes. Additionally, the elevated concentrations of squalene and 5-pregnene-3β,20α-diol in the post-marathon serum suggest cholesterol degradation (Charlton-Menys and Durrington 2007; Salway 2012), further supporting the activation of autophagy as the latter is a common constituent in cell membranes (Salway 2012). Elevated 5-pregnene-3β,20α-diol (also known as 20α-dihydropregnenolone) post-marathon also indicates steroid metabolism activation via cholesterol catabolism as it is produced by the reduction of pregnenolone via 20-hydroxysteroid dehydrogenase (Ebner et al. 2006). Pregnenolone is a precursor for the production of aldosterone and cortisol, the latter of which stimulates lipolysis, gluconeogenesis and protein catabolism (Salway 2012).

Furthermore, a significant decrease in α-aminomalonic acid concentrations was observed in the post-marathon serum samples. Although the precise origin of this metabolite is unknown, it has been associated with abnormal protein oxidation, macrophage accumulation, non-essential amino acid oxidation via reactive oxygen species and pulmonary arterial hypertension (Bujak et al. 2016).

The microbiome plays a crucial role in many essential metabolic processes required during strenuous exercise, including the regulation of energy metabolism, oxidative stress and inflammatory response pathways (Mach and Fuster-Botella 2017). To this end, various microbial-associated metabolites (i.e. tagatofuranose, talofuranose, ethyl-α-d-glucopyranoside, arabitol, indole-3-propionic acid and d-rhamnose) were significantly altered in the post-marathon samples. Most of these metabolites are synthesized from host carbohydrate intermediates, either via catabolism or fermentation processes [arabitol: a sugar alcohol produced from arabinose/glucose (Kumdam et al. 2014); and ethyl-α-d-glucopyranoside: an ethnolysis product from glucose (Hu et al. 2013)]. Indole-3-propionic acid is the deamination product of tryptophan, produced by gut microbes (e.g. Clostridium sporogenes) (Wikoff et al. 2009) and further supports the notion of reduced amino acids. The reduced concentrations of d-rhamnose, a component of most Gram-positive bacterial cell walls (Mistou et al. 2016), contradict the above-mentioned microbial product elevations. Since the particular nature of this reduction following completion of the marathon is unclear, it warrants future investigation.

In addition to these endogenously and microbially produced metabolites, elevated concentrations of ibuprofen were also observed post-marathon. Ibuprofen is a well-known non-steroidal, anti-inflammatory drug (Nieman et al. 2006) commonly used by athletes for preventing muscle damage/soreness and inflammation and was most likely consumed by some athletes prior to/during the marathon (Mcanulty et al. 2007).

5 Conclusion

The results of this metabolomics investigation suggest that the body utilizes various fuel substrate pathways to comply with the high energy demands required during the marathon, including catabolism of carbohydrates, lipids (β-oxidation and α-oxidation) and amino acids, as well as the activation of ketogenesis and autophagy via mTOR1 inhibition. Considering the results of the current investigation and previous literature, the possible cascade of events contributing to this metabolic “snapshot” could be summarized as follows: (a) A proposed initial reduction in carbohydrate catabolism and glucose uptake via insulin-dependent transporters lead to glycolysis dysregulation, ketogenesis activation and increased serum glucose. (b) A metabolic shift towards fatty acid utilization (from either endogenous or dietary TAGs) is induced, which (c) overwhelms/saturates the β-oxidation pathway, resulting in the α-oxidation of fatty acids. (d) Amino acids (from either endogenous or dietary protein catabolism) are also used as alternative fuel substrates, resulting in (e) mTOR1 inhibition and autophagy as the body desperately tries to generate the necessary fuel substrates to comply with the energy demand. (f) Lastly, various metabolic processes are activated to reduce oxidative stress and regulate/correct the redox imbalance.

Possible limitations of this study include human genotype/phenotype variation (an inevitable confounder) and the uncontrolled dietary intake of the athletes during the marathon. However, convincing athletes to deviate from their individualized supplementation protocols would be extremely difficult, if not impossible. Validation using a larger sample cohort could further substantiate the current findings. Nonetheless, these findings indicate the extensive metabolic changes induced by the marathon perturbation. Possible future prospects could be to investigate the effects of supplementing with amino acids, pre- and/or pro-biotics, and β-hydroxyisovaleric acid as a means of improving aerobic exercise performance, reduce skeletal muscle/liver damage, and enhance recovery.