A pilot study comparing the metabolic profiles of elite-level athletes from different sporting disciplines
Al-Khelaifi et al. Sports Medicine - Open
A pilot study comparing the metabolic profiles of elite-level athletes from different sporting disciplines
Fatima Al-Khelaifi 0 3
Ilhame Diboun 2
Francesco Donati 6
Francesco Botrè 6
Mohammed Alsayrafi 0
Costas Georgakopoulos 0
Karsten Suhre 5
Noha A. Yousri 1 4
Mohamed A. Elrayess 0
0 Anti Doping Laboratory Qatar , Sports City, P.O Box 27775, Doha , Qatar
1 Department of Genetic Medicine, Weill Cornell Medical College in Qatar , Education City, Qatar-Foundation, P.O. Box 24144, Doha , Qatar
2 Department of Economics , Mathematics and Statistics, Birkbeck , University of London , London WC1E 7HX , UK
3 University College London-Medical School, Royal Free Campus , London NW3 2PF , UK
4 Department of Computer and System Engineering, Alexandria University , Alexandria , Egypt
5 Department of Physiology and Biophysics, Weill Cornell Medical College in Qatar , Qatar-Foundation, P.O. Box 24144, Doha , Qatar
6 Laboratorio Antidoping, Federazione Medico Sportiva Italiana , Largo Giulio Onesti 1, 00197 Rome , Italy
Background: The outstanding performance of an elite athlete might be associated with changes in their blood metabolic profile. The aims of this study were to compare the blood metabolic profiles between moderate- and highpower and endurance elite athletes and to identify the potential metabolic pathways underlying these differences. Methods: Metabolic profiling of serum samples from 191 elite athletes from different sports disciplines (121 high- and 70 moderate-endurance athletes, including 44 high- and 144 moderate-power athletes), who participated in national or international sports events and tested negative for doping abuse at anti-doping laboratories, was performed using non-targeted metabolomics-based mass spectroscopy combined with ultrahigh-performance liquid chromatography. Multivariate analysis was conducted using orthogonal partial least squares discriminant analysis. Differences in metabolic levels between high- and moderate-power and endurance sports were assessed by univariate linear models. Results: Out of 743 analyzed metabolites, gamma-glutamyl amino acids were significantly reduced in both high-power and high-endurance athletes compared to moderate counterparts, indicating active glutathione cycle. High-endurance athletes exhibited significant increases in the levels of several sex hormone steroids involved in testosterone and progesterone synthesis, but decreases in diacylglycerols and ecosanoids. High-power athletes had increased levels of phospholipids and xanthine metabolites compared to moderate-power counterparts. Conclusions: This pilot data provides evidence that high-power and high-endurance athletes exhibit a distinct metabolic profile that reflects steroid biosynthesis, fatty acid metabolism, oxidative stress, and energy-related metabolites. Replication studies are warranted to confirm differences in the metabolic profiles associated with athletes' elite performance in independent data sets, aiming ultimately for deeper understanding of the underlying biochemical processes that could be utilized as biomarkers with potential therapeutic implications.
Metabolomics; Elite athletes; Power; Endurance; Steroids biosynthesis; Oxidative stress; Energy substrates
The emerging data provide a comprehensive
snapshot of athletes’ metabolism based on their
sports class as well as small molecule markers of
fitness, including changes in metabolites reflecting
sex steroid hormone biosynthesis and oxidative
The analysis confirmed previously reported changes
in the consumption of energy substrates in
glycolysis, lipolysis, adenine nucleotide catabolism,
and amino acid catabolism in response to exercise.
Once replicated and validated, these metabolic
signatures could be utilized as biomarkers for
excessive trainability associated with elite athletic
performance with potential therapeutic implications.
Excessive training of professional athletes causes
alterations in their blood metabolic profile that depends
largely on the type and duration of their training
]. Various behavioral, biochemical, hormonal,
and immunological markers are routinely used to assess
athletes’ physical status during a training program [
Previous studies, however, have demonstrated that
conventional tests could not detect the physiological
differences between endurance athletes and control subjects,
or differences before and after training sensitively [
Therefore, a more comprehensive metabolic profiling
has been considered in order to identify global
physiological changes in response to training.
Metabolomics offers a quantitative measurement of
the metabolic profiles associated with exercise in
professional athletes in order to identify biomarkers associated
with their performance, response to fatigue, and
potentially their respective sports-related disorders [
Non-targeted metabolomics allows the detection of
changes in response to various physiological states such
as pre-/post-exercise and offers identification of
metabolic signatures with potential translational impact for
both professional athletes and general public . These
changes include metabolites associated with glucose,
lipid, amino acid, and energy metabolism [
], such as
those involved in adenosine triphosphate (ATP)
synthesis, beta-oxidation of free fatty acids, and ketone bodies
. Previous studies in healthy volunteers have
demonstrated significantly reduced excretion of amino acids
with increased fitness levels and increased fat oxidation
rate during exercise [
]. Furthermore, metabolomics
profiling of athletes undergoing intensive exercise
revealed increase in plasma lactate [
] and adenine
breakdown products , indicating anaerobic
metabolism and ATP cycling, respectively. Further studies of the
effect of exercise showed elevated tricarboxylic acid
(TCA) cycle intermediates, markers of aerobic energy
production, in skeletal muscle biopsies [
exercise was also shown to trigger changes in the levels
of amino acids, including a moderate uptake of
glutamate in skeletal muscle leading to release of alanine to
promote ammonia metabolism [
11, 15, 16
corresponding changes in plasma concentrations of these
]. Elevation in serum levels of sex steroid
hormones was also reported in endurance athletes only
in response to high exercise intensities [
Athletes who have competed in national or
international sports events are considered elite athletes and
have been classified into two broad types according to
the kind and intensity of exercise: dynamic (isotonic)
and static (isometric) [
]. The dynamic exercise
represents changes in the muscle length due to regular
contractions producing a limited intramuscular power.
These changes are characteristic of high-endurance
sports such as marathon running, cycling, or
longdistance triathlons. Static exercise, on the other hand,
leads to a greater intramuscular power with little
changes in muscle length and is characteristic to power
sporting events such as sprinting, jumping, throwing,
and weightlifting. Some sports, however, require both
power and endurance such as boxing and rowing.
Dynamic exercise can also be further characterized
based on the maximal oxygen uptake percentage (VO2)
achieved with maximum cardiac output. Static exercise
can too be sub-categorized in relation to maximal
voluntary contraction percentage (MVC) gained with
increasing blood pressure .
Despite multiple studies focusing on the impact of
exercise on athletes’ metabolomics profiling, the metabolic
differences between high- and moderate-power and
endurance athletes remain to be explored. This study aims
to identify the metabolic signature that differentiates
high- and moderate-power and endurance elite athletes
and to identify the potential metabolic pathways that
underlie these differences. Assessment of these changes
could provide valuable measures of the current physical
status of the athletes and their adaptation to training,
which may help directing future training programs,
preventing potential disorders associated with excessive
exercise as well as improving their overall performance.
Study participants included in this study were 191
consented elite athletes (171 males and 20 females) from
different sports disciplines who participated in national
or international sports events and tested negative for
doping substances at anti-doping laboratories in Qatar
and Italy. Spare serum samples collected for anti-doping
human growth hormone tests were used for
metabolomics studies. Briefly, samples were either collected IN or
OUT of competition. Once collected, samples were
delivered to the anti-doping labs within 36 h under cooling
conditions. Once received, samples were immediately
centrifuged to separate the serum and then stored at −
20 °C until analysis. Only information related to type of
sport and athlete’s gender were available to researches.
All other information was not available, including age,
ethnicity, or the time of recruitment (pre- or
postexercise), due to the strict anonymization process
undertaken by anti-doping laboratories and those dictated by
study’s ethics. This study was performed in line with the
World Medical Association Declaration of Helsinki. All
protocols were approved by the Institutional Research
Board of anti-doping lab Qatar (F2014000009). Sport
types can be dichotomized into low, moderate, and high
dynamic or static groups based on associated peak
dynamic (VO2) and peak static (MVC) components
achieved during competition, as suggested previously
]. In our study, few athletes belonged to low levels of
endurance and power, therefore were merged with the
corresponding moderate class of endurance and power,
respectively (Table 1A). For statistical analysis,
endurance and power athletes were each represented by a
categorical variable with two levels (high and moderate,
Table 1B). Table 1 further lists the number of
participants per sport type in each class and their genders.
Metabolomics profiling was performed using established
protocols at Metabolon, Durham, NC, USA. All methods
utilized a Waters ACQUITY ultra-performance liquid
chromatography (UPLC) and a Thermo Scientific
QExactive high resolution/accurate mass spectrometer
interfaced with a heated electrospray ionization (HESI-II)
source and Orbitrap mass analyzer operated at 35,000
mass resolution. The detailed description of the liquid
chromatography-mass spectrometry (LC-MS)
methodology was previously described [
] and is summarized in
the Additional file 1. Briefly, serum samples were
methanol extracted to remove the protein fraction. The resulting
extract was divided into five fractions: two for analysis by
two separate reverse phase (RP)/UPLC-MS/MS methods
with positive ion mode electrospray ionization (ESI), one
for analysis by RP/UPLC-MS/MS with negative ion mode
ESI, one for analysis by hydrophilic interaction
chromatography (HILIC)/UPLC-MS/MS with negative ion mode
ESI, and one sample was reserved for backup. Raw data
was extracted, peak-identified, and quality
controlprocessed using Metabolon’s hardware and software [
Compounds were identified by comparison to library
entries of purified standards or recurrent unknown entities
with more than 3300 commercially available purified
standard compounds. Library matches for each compound
were checked for each sample and corrected if necessary
]. Asterisks (*) indicated on IDs of some metabolites in
Tables 2 and 3, Additional file 2: Tables S2–S3 and S5–S8
refer to compounds that have not been officially
confirmed based on a standard, but their identities are
known with confidence.
Statistical analysis of metabolomics data
Metabolomics data were log-transformed to ensure
distribution normality. Batch correction was already
(A) Distribution of elite athletes in various categories based on sport type-associated peak dynamic (maximal oxygen uptake percentage; VO2) and peak static
(maximal voluntary muscle contraction percentage; MVC) components achieved during competition as described previously [
]. The number and gender (M for
males and F for females) of participants in each group are also indicated. (B) Categorization of sport types into classes based on power alone regardless of
endurance (left) and similarly for endurance alone ignoring power (right); these classes were used in the statistical analysis
performed by Metabolon by rescaling each metabolite so
that its median is equal to 1. Principle component
analysis (PCA) was initially undertaken using multivariate
techniques to achieve a global view of the data. PCA
components express a linear combination of the
metabolites levels weighted by the component’s loading
values. Orthogonal partial least square discriminant
analysis (OPLS-DA), a supervised multivariate regression
technique, was performed to identify components that best
differentiate between predefined classes of samples while
Sphingomyelin (d18:2/14:0, d18:1/14:1)*
Methionine, cysteine, SAM,
and taurine metabolism
Urea cycle; arginine and proline metabolism
Urea cycle; arginine and proline metabolism
Gamma-glutamyl amino acid
Phenylalanine and tyrosine metabolism
Methionine, cysteine, SAM,
and taurine metabolism
Primary bile acid metabolism
Asterisks (*) indicated on IDs of some metabolites refer to compounds that have not been officially confirmed based on a standard, but their identities are known
dissecting orthogonal components which do not
differentiate between these classes. In this study, OPLS-DA was
used to compare moderate versus high classes of
endurance and power separately. Both PCA and OPLS-DA were
run using SIMCA 14 with the default metabolite-wise
metabolite missingness threshold (percentage of missing
metabolite values across the samples) of 50%.
Univariate regression and enrichment analysis
Linear models for association analysis were run using the
R statistical package (version 2.14, www.r-project.org/). A
model incorporating power and endurance as a categorical
variable with two levels (moderate and high) was used.
Incorporating both endurance and power in the same
model made it possible to examine the effect of power
while correcting for endurance and vice versa. This is
sensible because the high-endurance class features a
mixture of high- and moderate-power sports while the
moderate-endurance class features only
moderatepower sports. An opposite pattern is observed with
power (Table 1B). With both analyses, covariates
including gender, hemolysis levels (determined visually
by Metabolon), and PCA components 1 and 2 were
included in the model. A stringent Bonferroni level of
significance of p ≤ 0.05/743 = 6.72 × 10 − 5 was used to infer
association. False discovery rate (FDR) multiple testing
correction was also performed. All p values included in
Tables 2 and 3, Additional file 2: Tables S2–S6 are
reported after performing the described multiple testing
correction. In order to identify metabolites that were
associated with endurance or power differently in males
versus females (endurance/power × gender), an
interaction term was added to the model. For simplicity, when
conducting the interaction analysis, both endurance and
power were used as continuous variables (since both come
in only two levels); hence, the analysis was reduced to
testing differences in the beta values between males and
females (where beta expressed the slope measuring the
effect of either power or endurance).
Function enrichment analysis was performed using the
one-tailed Wilcoxon sum of the ranks test. For a given
biological function, the test assesses the probability of
observing the identified ranks of related metabolites
from the linear model analysis by chance. To gain
further insight into the biochemistry of identified
metabolites, the Kyoto Encyclopedia of Genes and Genomes
(KEGG) pathways were utilized. For heatmap analysis,
metabolites were z-scaled by subtracting their means
followed by division by standard deviations.
Multivariate analysis of athlete metabolomics data
Non-targeted metabolomics was applied to determine the
metabolic signatures of 191 elite athletes. PCA
components 1 and 2 (PC1 and PC2) captured together 25% of
the variance in the data. PC1 revealed two clusters of
samples, which were not explained by gender, sport types, or
classes (Fig. 1a). Examination of the loading plot in Fig. 1b
revealed a concentration of hemoglobin and heme
metabolites at the positive end of PC1. Furthermore, a t test
comparing the hemolysis measurement, between the two
clusters of samples revealed by PC1, was significant at the
0.01 significance level. These results led to the conclusion
that PC1 captured the extent of hemolysis in the samples.
Interestingly, there was also an enrichment of
arachidonate phospholipid metabolites at the positive end of PC1
as oppose to an enrichment of eicosanoids at the negative
end. While the biochemical link between the two sets of
metabolites is an obvious substrate/product relationship,
the link to hemolysis was rather obscure. There were no
clusters of samples according to PC2 (Fig. 1a). A closer
look at the loading plot revealed that TCA energy
metabolites and amino acids that feed into TCA cycle were
mostly located at the positive end of PC2 (Fig. 1c).
Moreover, a significant positive correlation between previously
identified changes in metabolites following 1 hour
postendurance exercise [
], also listed in Additional file 2:
Table S1, and our PC2 loading values for the same
metabolites (R = 0.6, p = 0.005) was identified. The enrichment
of dipeptides at the negative end of PC2 could indicate an
opposing anabolic effect. Although PCA did not explain
sport classes, it provided clues of possible confounders
(hemolysis and pre/post exercise) that we corrected for
Unlike PCA, OPLS-DA can identify sets of metabolites
that best distinguish between predefined classes of samples.
An OPLS-DA analysis comparing moderate versus high
classes of endurance revealed one class-discriminatory
component accounting for 66.7% of the variation in the
data due to endurance level (R-squared-Y = 0.66,
Qsquared = 0.45) (Fig. 2a). The corresponding loading score,
shown in Fig. 2b, suggests a reduction in diacyl glycerols
and gamma-glutamyl amino acids as oppose to an increase
in steroids, GABA derivatives, and monohydroxy fatty
acids with higher endurance levels.
OPLS-DA also revealed a clear separation between
moderate versus high power. One significant
predictive component explaining 88% of the variation in the
power (R-squared-Y = 0.88, Q-squared = 0.52) was
identified (Fig. 3a). The loading plot on Fig. 3b
suggests a decrease in gamma glutamyl amino acids as
oppose to an increase in sterols, phospholipids,
lysolipids, and xanthine metabolites with increased power.
OPLS results were confirmed by linear model in the
Univariate association tests and function enrichment analysis
A linear model was used to assess the significance of
metabolite-associations with the athletes’ class (moderate
versus high endurance) after correcting for gender,
hemolysis levels, PC1, PC2, and power. Thirty-eight
metabolites associated with endurance at a Bonferroni level
of significance (p ≤ 0.05/743 = 6.72 × 10−5) were
identified and their associated pathways listed (Table 2). More
metabolites associated with endurance at FDR and
nominal levels of significance are shown in Additional file 2:
Table S2. Similar results were obtained when analysis
was restricted to males only (Additional file 2: Table S3).
Enrichment analysis revealed an over-representation of
diacylglycerols, gamma-glutamyl amino acids, eicosanoids,
and monohydroxy fatty acids (FDR-corrected p-value
0.000122, 0.005, 0.017, and 0.04, respectively) among
metabolites most strongly associated with endurance,
irrespective of the direction of change. The steroid class
scored a nominal p-value of 0.05 but failed to remain
significant after FDR-based multiple testing. Interestingly,
these results are in considerable agreement with metabolic
effects identified through the OPLS-DA multivariate
approach previously discussed (Fig. 2b).
The results pertaining to steroids are certainly
remarkable if replicated and will be elaborated further in the
“Discussion” section. It is important to note that in
addition to the six Bonferroni significant steroids listed in
Table 2, seven more steroid species were FDR significant
at alpha = 0.05. These are etiocholanolone glucuronide
(FDR p value = 0.003);
5alpha-pregnane-3beta,20alphadiol disulfate (FDR p value = 0.01);
5alpha-pregnane3beta,20beta-diol monosulfate (FDR p value = 0.02);
androstenediol (3beta,17beta) disulfate (FDR p value =
0.025); 5alpha-pregnane-3beta,20alpha-diol monosulfate
(FDR p value = 0.029); pregnen-dioldisulfate (FDR p
value = 0.035); and androstenediol (3alpha, 17alpha)
monsulfate (FDR p value = 0.04). All Bonferroni and
FDR significant steroid metabolites were projected onto
KEGG Steroid Biosynthesis Pathway to highlight their
biochemical inter-relationships (Fig. 4). Significant
correlations among the identified steroid metabolites were
confirmed (Additional file 3: Fig. S1, Additional file 2:
Table S4), suggesting activation of sex steroid
biosynthesis pathway in high-endurance athletes.
A part of enrichments of functionally related sets of
metabolites, endurance association analysis also revealed
individual metabolic effects which are noteworthy.
Among these are derivatives of GABA cyclic lactam
2pyrrolidinone including succinimide (Bonferroni p value
= 0.00263), acisoga or
N-(3-acetamidopropyl)pyrrolidin2-one (FDR p value = 0.004), and 2-pyrrolidinone itself
(FDR p value = 0.03) as well as GABA derivative
4guanidinobutanoate (Bonferroni p value = 0.004). There
were significant correlations between 2-pyrrolidinone and
its derivatives including succinimide (R = 0.15, p = 0.04),
4-guanidinobutanoate (R = − 0.146, p = 0.04), and
guanidinosuccinate (R = − 0.186, p = 0.01), suggesting presence of
this drug and its derivatives in high-endurance athletes,
also seen in OPL-DA analysis (Fig. 2b).
Other interesting effects include a Bonferroni
significant increase in citrate together with an FDR significant
increase in 2-methylcitrate (FDR = 0.012). Other
associations include acyl carnitines, phospholipids, and
sphingolipids among others (Table 2).
Power associated metabolites
When considering power, the categorical variable
“power” becomes the explanatory variable of interest in
the previous model and “endurance” becomes a
confounder that is corrected for. Thirty-three metabolites
were significantly associated with power according to
this model; these are listed in (Table 3). Enrichment
analysis revealed an over-representation of phospholipids
(p = 0.00042), lysolipids (p = 0.00042), gamma-glutamyl
amino acids (p = 0.000846), and sterols (p = 0.005)
amongst metabolites most strongly associated with
power. Other significantly changed metabolites in
moderate- versus high-power classes included
guanidinoacetate, N-acetylcarnosine, cholate, imidazole lactate,
indolelactate, and 3-methylxanthine (Table 3).
Among FDR significant changes, an increase in
creatine (estimate = 0.6, p = 0.001) and a decrease in
creatinine (estimate = − 0.1, p = 0.002) were also
detected in the high-power group although did not
reach Bonferroni significance. More metabolites
associated with power at FDR level of significance are
shown in Additional file 2: Table S5. Similar results
were obtained when analysis was restricted to males
only (Additional file 2: Table S6).
Metabolites with FDR corrected p values of less than
0.01 from the endurance and power models were
projected on the heatmap in Figs. 5 and 6, respectively.
The heatmaps give a snapshot summary of the actual
intensities of these metabolites after correcting for
confounders in the linear model described earlier. Samples
were ordered by sports type within their respective sport
groups (moderate power/moderate endurance, moderate
power/high endurance and high power/high endurance).
Gender-sports class interaction
Gender-endurance interaction analysis identified 60
significant metabolites with a nominal p value (less than
0.05) amongst which none remained significant after
FDR correction (Additional file 2: Table S7). As for
power, 144 metabolites were differently associated with
power between males and females, among which 35
metabolites remained significant after FDR correction
(Additional file 2: Table S8).
Metabolic profiling of athletes’ blood in response to
exercise has recently revealed unique metabolic signatures
associated with various types and durations of exercise
]. However, metabolomics of elite athletes from
different sport disciplines remains to be investigated. In
particular, the metabolic pathways of endurance and
power athletes should shed light on the molecular
mechanisms underlying variations with functional relevance
or those that can be used as potential biomarkers for
their respective sport class. In this study, metabolomics
analysis was utilized to characterize the unique serum
metabolic signature of elite athletes who participated in
national or international sports events following the
successful completion of anti-doping tests. Despite limited
information about the participants and possible
confounding factors influencing their metabolic profiling,
the emerging data revealed significant differences in
metabolite levels between high- versus moderate-power and
endurance sport types. Inclusion of PC1 and PC2 in the
linear model has likely corrected for expected confounders
including hemolysis and pre-post exercise effects to reveal
common as well as distinctive metabolic mechanisms
underlying endurance and power. These include a clear
signature of oxidative stress common to both high-power
and high-endurance sports alike, yet steroids and
polyamine pathways appeared more prominent in endurance,
while sterols, adenine-containing purines, and energy
metabolites were most evident with power.
Metabolites associated with endurance
Exercise can cause changes in sex steroid hormone
concentrations in the serum of non-athletes as well as
], including levels of testosterone and cortisol
]. One interesting finding in this study is the
elevated levels of various metabolites involved in sex steroid
hormone biosynthesis in the high-endurance athletes.
Some of these metabolites were conjugated with one or
more sulfate group(s) which renders them inactive.
However, these can be reactivated through the activity of
enzyme steroid sulfatase [
]. The list of elevated steroids
included pregnenolone that mediates biosynthesis of
corticosteroids and progesterone and 21-hydroxypregnenolone
disulfate that mediates biosynthesis of corticosteroids,
corticoids (cortisol and cortisone), various metabolites
of progesterone (pregnanediol,
testosterone precursor (androstenediol (3beta,17beta)),
and testosterone metabolites
(etiocholanoloneglucuronide, androstenediol (3alpha, 17alpha)) (Fig. 4).
Elevated cortisol-related metabolites in response to
sustained aerobic exercise were shown to correlate
positively with intensity of exercise as measured by
oxygen uptake [
]. However, exercise-induced alterations
in sex steroid hormone levels are usually short lived
(1–3 h) [
]. The habitual exercise regiments of the
elite endurance athletes may have accounted for this
maintained systemic increase. Sex steroid hormones
play a crucial role in glucose metabolism and protein
synthesis in the muscle as well as in the regulation of
redox homeostasis [
]. Some act as neurosteroids
that alter neuronal excitability such as
pregnendioldisulfate that works as a potent negative allosteric
modulator of the GABAA receptor [
pregnenolone sulfate that acts as a potent negative allosteric
modulator of the GABAA receptor and a weak positive
allosteric modulator of the NMDA receptor [
stimulatory effects of steroids on muscle mass, energy
generation, and neuronal excitability may have
accounted for the higher endurance ability of the
highendurance group compared to their lower endurance
counterparts. Given that athletes included in this study
have successfully passed anti-doping tests, changes in
steroids levels may reflect either enhancement in
endogenous anabolic steroids biosynthesis, physiological
adaptation to exercise, and/or increased dietary intake.
A genetic association study is needed to reveal the
potential genetic variants underlying increased activity of
enzymes involved in steroid biosynthesis. Interestingly,
in addition to elevation in a number of neurosteroids,
our data suggested increased elevated levels of a number
of GABA derivatives including 2-pyrrolidinone, the cyclic
lactate form of GABA [
], its derivatives succinimide,
acisoga (N-(3-acetamidopropyl)pyrrolidin-2-one), and
4guanidinobutanoate, perhaps contributing to
GABAmediated muscle growth in response to exercise [
Other metabolic changes associated with high
endurance included reduced diacylglycerols (DAGs) and fatty
acid (FA)-carnitine and increased acylated carnitine.
Alterations in these lipids may suggest enhanced hydrolysis
of DAGs, shuttling of FA intracellularly, followed by
fatty acid oxidation and energy generation [
acids and lipids are preferred substrates for exercising
the muscle, and the emerging data suggest a greater beta
oxidation of fatty acids in athletes belonging to higher
endurance sports. Hence, those athletes are perhaps more
capable of activating lipolysis during physical activity than
moderate-endurance athletes. Furthermore, accumulation
of acylated carnitine may provide a favorable effect on the
recovery from exercise stress since carnitine can reduce
post-exercise plasma lactate and prevent cellular damage
]. Citrate and isocitrate were also significantly increased
in high-endurance elite athletes, indicating enhanced
aerobic energy generation through TCA.
Metabolites associated with power athletes
Changes in creatine, creatinine, and guanidinoacetate
were significant between high- and moderate-power
athletes. Whereas creatine increased in the high-power
group, its breakdown product (creatinine) and precursor
(guanidinoacetate) were both significantly reduced, thus
maintaining the previously reported balance of creatine
]. Creatine (Cr) and creatine phosphate
(CrP) play essential roles in the storage and transmission
of phosphate-bound energy. Changes in creatine
homeostasis in high-power athletes may suggest more
adaptable muscular storage of CrP that during exercise can
constitute an essential source for high energy to
replenish ATP in the first few seconds of intense activity.
Other energy-related metabolites elevated in high-power
athletes were 3-methylxanthine and 7-methylxanthine
(adenine breakdown products), perhaps reflecting
heightened utilization of fuel substrates in several
metabolic pathways [
]. Xanthine supplementation allows
athletes to exercise at a greater power output for longer
]. Additionally, N-acetylcarnosine was
significantly reduced in high-power athletes. This metabolite
acts as oxidative stress scavenger in muscles especially
against lipid peroxidation through its imidazolium group
that stabilizes adducts formed at the primary amino
]. Various derivatives of phosphatidates were
increased with increased power, perhaps reflecting
changes in cellular membrane dynamics in response to
oxidative stress [
]. Among those, inositol
phospholipids were previously shown to accumulate in response
to muscle contraction during hypoxia [
metabolite likely to be a result from stress-induced
membrane dynamics is 12,13-DHOME. This long-chain fatty
acid enhances adipogenesis and inhibits asteogenesis due
to its role as a proliferator-activated receptor (PPAR)
gamma 2 ligand [
Global stress response in both high-power and high-endurance athletes
Intensive exercise has been implicated in the
promotion of free radical generation in active skeletal
muscle resulting in the formation of oxidized lipids
]. Overall in both power and endurance athletes,
there was a clear stress metabolic response. Changes
in gamma-gultamyl amino acids, associated with
elevated cysteine-glutathione disulfide (change 0.24,
nominal p value of 0.03), between high- and
moderateperformance athletes may indicate active gamma-glutamyl
cycle that plays an important role in the
glutathionemediated radical detoxification during oxidative stress
]. The cycle involves synthesis and degradation of
glutathione by transferring gamma-glutamyl functional
groups from glutathione to an amino acid, leaving the
cysteine products intact, which leads to the preservation
of intracellular homeostasis in case of oxidative stress
]. Reduction in serum levels of
gamma-glutamylamino acids in high-performance athletes (both high
power and high endurance) may indicate increased
glutathione synthesis. The accumulation of glutathione
in the blood stream marks increased oxidative stress
and reactive oxygen species scavenging activity.
Despite lack of FDR significant differences in metabolites
associated with endurance in males versus females,
differences in a number of metabolites were nominally
significant, including a number of gamma-glutamyl
amino acids and steroid metabolites among others.
Differences in these metabolites between high and
moderate levels of endurance were mostly going in the same
direction in males and females but were more
pronounced in females. As per power-associated
metabolites, there were FDR significant differences between
males and females in a number of metabolites including
TCA-mediators such as malate, fumarate, succinate, and
alpha keto glutarate as well as lactate where in females
there was increase with higher power with no FDR
significant effects in males. These gender-related
differences need to be further investigated, especially in light
of low number of studied females (n = 20).
One main limitation of this study is the relatively low
number of participants, especially the females; therefore, a
replication study is essential for confirmation of these
findings. Furthermore, since athletes’ blood samples were
collected at multiple sites, a batch effect was inevitable,
likely attenuating correlations between metabolite
concentrations and sports class. This batch effect may have
included various crucial pre-analytical features that can
significantly influence the metabolic profiling of samples
such as the blood collection process and time (IN or OUT
of competition) and transportation conditions, including
time to reach anti-doping laboratories, sample processing,
and sample storage [
]. Despite these factors, clear
signatures were identified after correcting for potential
confounders. Additionally, the lack of information about
participants including their age, ethnicity, and body mass
index was another major limitation of this study. However,
the young age of elite athletes in general and the wide
range of sports included in this study may have diluted
out other potential confounders. Ambiguity in the exact
description of the subcategories of athletes’ sports was an
additional issue this study has faced due to the limited
information provided by the anti-doping laboratories
following the strict anonymization process. This has prompted
the adoption of the general sports class grouping based on
previously published work [
] despite the differences
among different members of the same team such as such
as breast-stroke and freestyle swimming or football
midfielders and goal keepers. Another limitation of this study
is the group number bias as some sports were
overrepresented and others underrepresented. Finally, differences in
dietary intake between high- and moderate-power and
endurance elite athletes, including various supplements,
medications, and other ergogenics, may have influenced
their metabolic profile [
]. Such differences are difficult to
account for as they vary among different sports and
athletes and are not usually publicized. Taken all these
limitations into account, it is critical to stress that this is a pilot
study that needs further replication and validation as
finding biomarkers from the identified differentiating
significant compounds still requires optimization of
targetspecific analytical methods and validation of these methods
with their reference materials and proficiency tests [
The emerging data provide a comprehensive snapshot of
athletes metabolism based on their sports class as well as
small molecule markers of fitness, which requires further
validation. Metabolomics of elite athletes classified according
to their sports class into endurance or power revealed for
the first time changes in metabolites reflecting sex steroid
hormones biosynthesis and oxidative stress substrates
(glutathione metabolism). The analysis confirmed previously
reported changes in the consumption of energy substrates
in glycolysis [
], lipolysis [
], adenine nucleotide
catabolism , and amino acid catabolism [
] in response to
1, 55, 56
]. These metabolic signatures could be
utilized as pilot indicators of excessive trainability associated
with elite athletic performance with potential applications in
directing future training programs, preventing potential
disorders associated with excessive exercise as well as
improving their overall performance. Changes in these metabolic
signatures may also provide valuable clues for anti-doping
research related to Athlete Biological Passport.
Additional file 1: Materials and Methods. (DOCX 16 kb)
Additional file 2: Table S1. Comparison of previously published
metabolite changes in plasma at 60 min after completion of exercise [
and their corresponding PC2 loading values obtained in this study.
Table S2. Metabolites differentiating between moderate- and
high-endurance athletes (p ≤ 0.05). Table S3. Metabolites differentiating
between moderate- and high-endurance athletes (p ≤ 0.05) in males only.
Table S4. Pearson’s Correlations between various sex steroid metabolites.
Significant p values are highlighted (* < 0.05, ** < 0.01, *** < 0.001).
Table S5. Metabolites differentiating between moderate- and
highpower athletes (p ≤ 0.05). Table S6. Metabolites differentiating between
moderate- and high-power athletes (p ≤ 0.05) in males only. Table S7.
Gender-endurance interaction metabolites. Columns A–F show the effect
of endurance on gender-interaction metabolites in males only. Columns
H to L show the different effect in females. Table S8. Gender-power
interaction metabolites. Columns A–F show the effect of power on
gender-interaction metabolites in males only. Columns H to L show the
different effect in females. (XLSX 1377 kb)
Additional file 3: Figure S1. Heatmap (left) and hierarchical clustering
(right) of steroid metabolites featured in this study. The significant
metabolites from the linear model associated with endurance are
highlighted in red (right). (PPTX 73 kb)
ATP: Adenosine triphosphate; DAGs: Diacylglycerols; ESI: Electrospray
ionization; FA: Fatty acid; HESI-II: Heated electrospray ionization;
HILIC: Hydrophilic interaction chromatography; LC-MS: Liquid
chromatography-mass spectrometry; MVC: Maximal voluntary contraction
percentage; OPLS-DA: Orthogonal partial least square discriminant analysis;
PCA: Principle component analysis; RP: Reverse phase; TCA: Tricarboxylic acid;
UPLC: Ultra-performance liquid chromatography; VO2: Maximal oxygen uptake
We would like to thank Qatar National Research Fund (QNRF) for funding this
project (Grant number NPRP7-272-1-041). We would like to thank Dr. Edward D.
Karoly from Metabolon, Inc. for his detailed description of the metabolomics
analysis adopted by Metabolon used in the methods section.
This research was sponsored by Qatar National Research Fund (QNRF), Grant
number NPRP7-272-1-041 (MAE, KS, CG, and FB).
Availability of data and materials
All datasets on which the conclusions of the manuscript rely are presented
in the additional supporting file in excel format and will be made available
at Metabolomics Workbench.
FK, ID, FD, FB, MA, CG, KS, NY, and MAE collected samples, carried out
analysis, wrote the paper, and reviewed and accepted its final version. MAE
(corresponding) is responsible for the integrity of the work as a whole. All
authors read and approved the final manuscript.
CG and FB are directors of anti-doping labs in Qatar and Italy, respectively.
ID, KS, and NY are the bioinformatics/biostatistics team. MAE is the lead PI.
Ethics approval and consent to participate
This study was performed in line with the World Medical Association
Declaration of Helsinki. Only consented participants were included in the
study. All protocols were approved by the Institutional Research Board of
anti-doping lab Qatar (F2014000009).
Consent for publication
Fatima Al-Khelaifi, Ilhame Diboun, Francesco Donati, Francesco Botrè,
Mohammed Alsayrafi, Costas Georgakopoulos, Karsten Suhre, Noha A. Yousri,
and Mohamed A Elrayess declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
1. Lewis GD , Farrell L , Wood MJ , Martinovic M , Arany Z , Rowe GC , et al. Metabolic signatures of exercise in human plasma . Sci Transl Med . 2010 ; 2 ( 33 ):33ra7. https://doi.org/10.1126/scitranslmed.3001006.
2. Pitsiladis YP , Maughan RJ . The effects of exercise and diet manipulation on the capacity to perform prolonged exercise in the heat and in the cold in trained humans . J Physiol . 1999 ; 517 (Pt 3): 919 - 30 .
3. Lac G , Maso F . Biological markers for the follow-up of athletes throughout the training season . Pathologie-biologie. 2004 ; 52 ( 1 ): 43 - 9 . https://doi.org/10.1016/S0369- 8114 ( 03 ) 00049 - X .
4. Rietjens GJ , Kuipers H , Adam JJ , Saris WH , van Breda E , van Hamont D , et al. Physiological, biochemical and psychological markers of strenuous traininginduced fatigue . Int J Sports Med . 2005 ; 26 ( 1 ): 16 - 26 . https://doi.org/10.1055/s-2004-817914.
5. Yan B , A J , Wang G , Lu H , Huang X , Liu Y et al. Metabolomic investigation into variation of endogenous metabolites in professional athletes subject to strength-endurance training . J Appl Physiol 2009 ; 106 ( 2 ): 531 - 538 . doi: https://doi.org/10.1152/japplphysiol.90816. 2008 .
6. Bogdanis GC . Effects of physical activity and inactivity on muscle fatigue . Front Physiol . 2012 ; 3 : 142 . https://doi.org/10.3389/fphys. 2012 . 00142 .
7. Nicholson JK , Connelly J , Lindon JC , Holmes E. Metabonomics : a platform for studying drug toxicity and gene function . Nat Rev Drug Discov . 2002 ; 1 ( 2 ): 153 - 61 . https://doi.org/10.1038/nrd728.
8. Heaney LM , Deighton K , Suzuki T . Non-targeted metabolomics in sport and exercise science . J Sports Sci . 2017 : 1 - 9 . https://doi.org/10.1080/02640414. 2017 . 1305122 .
9. Morris C , Grada CO , Ryan M , Roche HM , De Vito G , Gibney MJ , et al. The relationship between aerobic fitness level and metabolic profiles in healthy adults . Mol Nutr Food Res . 2013 ; 57 ( 7 ): 1246 - 54 . https://doi.org/10.1002/mnfr. 201200629.
10. Goodwin ML , Harris JE , Hernandez A , Gladden LB . Blood lactate measurements and analysis during exercise: a guide for clinicians . J Diabetes Sci Technol . 2007 ; 1 ( 4 ): 558 - 69 .
11. Berton R , Conceicao MS , Libardi CA , Canevarolo RR , Gaspari AF , ChaconMikahil MP , et al. Metabolic time-course response after resistance exercise: a metabolomics approach . J Sports Sci . 2017 ; 35 ( 12 ): 1211 - 8 . https://doi.org/10. 1080/02640414. 2016 . 1218035 .
12. Dudzinska W , Lubkowska A , Dolegowska B , Safranow K , Jakubowska K . Adenine, guanine and pyridine nucleotides in blood during physical exercise and restitution in healthy subjects . Eur J Appl Physiol . 2010 ; 110 ( 6 ): 1155 - 62 . https://doi.org/10.1007/s00421-010-1611-7.
13. Howarth KR , LeBlanc PJ , Heigenhauser GJ , Gibala MJ . Effect of endurance training on muscle TCA cycle metabolism during exercise in humans . J Appl Physiol . 2004 ; 97 ( 2 ): 579 - 84 . https://doi.org/10.1152/japplphysiol.01344. 2003 .
14. Peake JM , Tan SJ , Markworth JF , Broadbent JA , Skinner TL , Cameron-Smith D . Metabolic and hormonal responses to isoenergetic high-intensity interval exercise and continuous moderate-intensity exercise . Am J Physiol Endocrinol Metab . 2014 ; 307 ( 7 ): E539 - 52 . https://doi.org/10.1152/ajpendo. 00276. 2014 .
15. Henriksson J . Effect of exercise on amino acid concentrations in skeletal muscle and plasma . J Exp Biol . 1991 ; 160 : 149 - 65 .
16. Leibowitz A , Klin Y , Gruenbaum BF , Gruenbaum SE , Kuts R , Dubilet M , et al. Effects of strong physical exercise on blood glutamate and its metabolite 2- ketoglutarate levels in healthy volunteers . Acta Neurobiol Exp . 2012 ; 72 ( 4 ): 385 - 96 .
17. Sahlin K , Katz A , Broberg S. Tricarboxylic acid cycle intermediates in human muscle during prolonged exercise . Am J Phys . 1990 ; 259 ( 5 Pt 1 ): C834 - 41 .
18. Eriksson LS , Broberg S , Bjorkman O , Wahren J . Ammonia metabolism during exercise in man . Clin Physiol . 1985 ; 5 ( 4 ): 325 - 36 .
19. Sato K , Iemitsu M , Katayama K , Ishida K , Kanao Y , Saito M . Responses of sex steroid hormones to different intensities of exercise in endurance athletes . Exp Physiol . 2016 ; 101 ( 1 ): 168 - 75 . https://doi.org/10.1113/EP085361.
20. Asmussen E. Similarities and dissimilarities between static and dynamic exercise . Circ Res . 1981 ; 48 ( 6 Pt 2 ): I3 - 10 .
21. Mitchell JH , Haskell W , Snell P , Van Camp SP. Task Force 8: classification of sports . J Am Coll Cardiol . 2005 ; 45 ( 8 ): 1364 - 7 . https://doi.org/10.1016/j.jacc. 2005 . 02 .015.
22. Evans AM , Bridgewater BR , Liu Q , Mitchell MW , Robinson RJ , Dai H , et al. High resolution mass spectrometry improves data quantity and quality as compared to unit mass resolution mass spectrometry in high-throughput profiling metabolomics . Metabolomics . 2014 ; 4 ( 132 ) https://doi.org/10.4172/ 2153 - 0769 . 1000132 .
23. DeHaven CD , Evans JM , Dai H , Lawton KA . Software techniques for enabling high-throughput analysis of metabolomic datasets . Metabolomics, Dr Ute Roessner (Ed), InTech . 2012 ; Chapter 7 . doi:https://doi.org/10.5772/31277.
24. Smith AA , Toone R , Peacock O , Drawer S , Stokes KA , Cook CJ . Dihydrotestosterone is elevated following sprint exercise in healthy young men . J Appl Physiol . 2013 ; 114 ( 10 ): 1435 - 40 . https://doi.org/10.1152/ japplphysiol.01419. 2012 .
25. Budde H , Machado S , Ribeiro P , Wegner M. The cortisol response to exercise in young adults . Front Behav Neurosci . 2015 ; 9 : 13 . https://doi.org/10. 3389/fnbeh. 2015 . 00013 .
26. Budde H , Pietrassyk-Kendziorra S , Bohm S , Voelcker-Rehage C . Hormonal responses to physical and cognitive stress in a school setting . Neurosci Lett . 2010 ; 474 ( 3 ): 131 - 4 . https://doi.org/10.1016/j.neulet. 2010 . 03 .015.
27. Fietz D , Bakhaus K , Wapelhorst B , Grosser G , Gunther S , Alber J , et al. Membrane transporters for sulfated steroids in the human testis-cellular localization, expression pattern and functional analysis . PLoS One . 2013 ; 8 ( 5 ): e62638. https://doi.org/10.1371/journal.pone. 0062638 .
28. Fuqua JS , Rogol AD . Neuroendocrine alterations in the exercising human: implications for energy homeostasis . Metab Clin Exp . 2013 ; 62 ( 7 ): 911 - 21 . https://doi.org/10.1016/j.metabol. 2013 . 01 .016.
29. Jacob MH , RJD d, Jahn MP , Kucharski LC , Bello-Klein A , Ribeiro MF . Agerelated effects of DHEA on peripheral markers of oxidative stress . Cell Biochem Funct . 2010 ; 28 ( 1 ): 52 - 7 . https://doi.org/10.1002/cbf.1619.
30. Ferrando AA , Sheffield-Moore M , Yeckel CW , Gilkison C , Jiang J , Achacosa A , et al. Testosterone administration to older men improves muscle function: molecular and physiological mechanisms . Am J Physiol Endocrinol Metab . 2002 ; 282 ( 3 ): E601 - 7 . https://doi.org/10.1152/ajpendo.00362. 2001 .
31. Sato K , Iemitsu M , Aizawa K , Ajisaka R . Testosterone and DHEA activate the glucose metabolism-related signaling pathway in skeletal muscle . Am J Physiol Endocrinol Metab . 2008 ; 294 ( 5 ): E961 - 8 . https://doi.org/10.1152/ ajpendo.00678. 2007 .
32. Belelli D , Gee KW . 5 alpha-pregnan-3 alpha,20 alpha-diol behaves like a partial agonist in the modulation of GABA-stimulated chloride ion uptake by synaptoneurosomes . Eur J Pharmacol . 1989 ; 167 ( 1 ): 173 - 6 .
33. Harteneck C . Pregnenolone sulfate: from steroid metabolite to TRP channel ligand . Molecules . 2013 ; 18 ( 10 ): 12012 - 28 . https://doi.org/10.3390/ molecules181012012.
34. Fasolato C , Bertazzon A , Previero A , Galzigna L . Effect of 2-pyrrolidone on the concentration of GABA in rat tissues . Pharmacology . 1988 ; 36 ( 4 ): 258 - 64 .
35. Powers ME , Yarrow JF , McCoy SC , Borst SE . Growth hormone isoform responses to GABA ingestion at rest and after exercise . Med Sci Sports Exerc . 2008 ; 40 ( 1 ): 104 - 10 . https://doi.org/10.1249/mss.0b013e318158b518.
36. Hoppel CL . Carnitine and carnitine palmitoyltransferase in fatty acid oxidation and ketosis . Fed Proc . 1982 ; 41 ( 12 ): 2853 - 7 .
37. Karlic H , Lohninger A . Supplementation of L-carnitine in athletes: does it make sense? Nutrition . 2004 ; 20 ( 7-8 ): 709 - 15 . https://doi.org/10.1016/j.nut. 2004 . 04 .003.
38. Walker JB . Creatine: biosynthesis, regulation, and function . Adv Enzymol Relat Areas Mol Biol . 1979 ; 50 : 177 - 242 .
39. Tullson PC , Terjung RL . Adenine nucleotide metabolism in contracting skeletal muscle . Exerc Sport Sci Rev . 1991 ; 19 : 507 - 37 .
40. Graham TE . Caffeine and exercise: metabolism, endurance and performance . Sports Med . 2001 ; 31 ( 11 ): 785 - 807 .
41. Boldyrev A , Abe H . Metabolic transformation of neuropeptide carnosine modifies its biological activity . Cell Mol Neurobiol . 1999 ; 19 ( 1 ): 163 - 75 .
42. Powers SK , Jackson MJ . Exercise-induced oxidative stress: cellular mechanisms and impact on muscle force production . Physiol Rev . 2008 ; 88 ( 4 ): 1243 - 76 . https://doi.org/10.1152/physrev.00031. 2007 .
43. Coburn RF , Baron C , Papadopoulos MT . Phosphoinositide metabolism and metabolism-contraction coupling in rabbit aorta . Am J Phys . 1988 ; 255 ( 6 Pt 2 ): H1476 - 83 .
44. Lecka-Czernik B , Moerman EJ , Grant DF , Lehmann JM , Manolagas SC , Jilka RL . Divergent effects of selective peroxisome proliferator-activated receptorgamma 2 ligands on adipocyte versus osteoblast differentiation . Endocrinology . 2002 ; 143 ( 6 ): 2376 - 84 . https://doi.org/10.1210/endo.143.6.8834.
45. Circu ML , Aw TY . Glutathione and modulation of cell apoptosis . Biochim Biophys Acta . 2012 ; 1823 (10): 1767 - 77 . https://doi.org/10.1016/j.bbamcr. 2012 . 06 .019.
46. Lu SC . Regulation of glutathione synthesis . Mol Asp Med . 2009 ; 30 ( 1-2 ): 42 - 59 . https://doi.org/10.1016/j.mam. 2008 . 05 .005.
47. Yokoyama H . Gamma glutamyl transpeptidase (gammaGTP) in the era of metabolic syndrome . Nihon Arukoru Yakubutsu Igakkai zasshi = Jpn J Alcohol Studies Drug Depend . 2007 ; 42 ( 3 ): 110 - 24 .
48. Yin P , Lehmann R , Xu G. Effects of pre-analytical processes on blood samples used in metabolomics studies . Anal Bioanal Chem . 2015 ; 407 ( 17 ): 4879 - 92 . https://doi.org/10.1007/s00216-015-8565-x.
49. Williams MH . Dietary supplements and sports performance: introduction and vitamins . J Int Soc Sports Nutr . 2004 ; 1 : 1 - 6 . https://doi.org/10.1186/ 1550 -2783-1-2-1.
50. Naz S , Vallejo M , Garcia A , Barbas C . Method validation strategies involved in non-targeted metabolomics . J Chromatogr A . 2014 ; 1353 : 99 - 105 . https://doi. org/10.1016/j.chroma. 2014 . 04 .071.
51. Van Hall G , Jensen-Urstad M , Rosdahl H , Holmberg HC , Saltin B , Calbet JA . Leg and arm lactate and substrate kinetics during exercise . Am J Physiol Endocrinol Metab . 2003 ; 284 ( 1 ): E193 - 205 . https://doi.org/10.1152/ajpendo. 00273. 2002 .
52. Ormsbee MJ , Thyfault JP , Johnson EA , Kraus RM , Choi MD , Hickner RC . Fat metabolism and acute resistance exercise in trained men . J Appl Physiol . 2007 ; 102 ( 5 ): 1767 - 72 . https://doi.org/10.1152/japplphysiol.00704. 2006 .
53. Goto K , Ishii N , Sugihara S , Yoshioka T , Takamatsu K. Effects of resistance exercise on lipolysis during subsequent submaximal exercise . Med Sci Sports Exerc . 2007 ; 39 ( 2 ): 308 - 15 . https://doi.org/10.1249/01.mss. 0000246992 . 33482.cb.
54. Sahlin K , Gorski J , Edstrom L . Influence of ATP turnover and metabolite changes on IMP formation and glycolysis in rat skeletal muscle . Am J Phys . 1990 ; 259 ( 3 Pt 1 ): C409 - 12 .
55. Chorell E , Moritz T , Branth S , Antti H , Svensson MB . Predictive metabolomics evaluation of nutrition-modulated metabolic stress responses in human blood serum during the early recovery phase of strenuous physical exercise . J Proteome Res . 2009 ; 8 ( 6 ): 2966 - 77 . https://doi.org/10.1021/pr900081q.
56. Pohjanen E , Thysell E , Jonsson P , Eklund C , Silfver A , Carlsson IB , et al. A multivariate screening strategy for investigating metabolic effects of strenuous physical exercise in human serum . J Proteome Res . 2007 ; 6 ( 6 ): 2113 - 20 . https://doi.org/10.1021/pr070007g.