Abstract

   Metabolomics and lipidomics techniques are capable of comprehensively
   measuring hundreds to thousands of small molecules in single analytical
   runs and have been used to characterize responses to exercise
   traditionally using venipuncture‐produced liquid samples. Advanced
   microsampling devices offer an alternative by circumventing the
   requirement to maintain frozen samples. This approach combines a
   microneedle puncture for blood draw with microfluidic sample collection
   onto a dried carrier and has thus far been employed for targeted
   measurements of a few analytes. To demonstrate the utility of advanced
   dried microsampling to characterize metabolomic and lipidomic changes
   during exercise, we obtained samples before and after a 2‐mile run from
   twelve (8 male, 4 female) healthy volunteers with various ranges in
   activity levels. Results highlighted significant changes in whole blood
   levels of several metabolites associated with energy (glycolysis and
   Tricarboxylic Acid cycle) and redox (Pentose Phosphate Pathway)
   metabolism. Lipid changes during this same period were individualized
   and less uniform. Sex‐based differences in response to running
   highlighted reliance on carbohydrate or fat substrate utilization in
   males or females, respectively. The results presented herein illustrate
   the ability of this approach to monitor circulating metabolome and
   lipidome profiles from field sampled blood in response to exercise.

   Keywords: bioenergetics, dried blood spot, exercise, fatty acid
   oxidation, glycolysis, lipidomics, metabolomics, running

1. INTRODUCTION

   Non‐communicable diseases including cardiovascular disease, as well as
   the metabolic factors that etiologically contribute to their onset
   (e.g., obesity and diabetes), currently represent the global leading
   cause of death.[28] ^1 Sedentary behavior is a conduit for the
   development of metabolic derangements that contribute to the
   development of cardiovascular diseases[29] ^2 and obesity,[30] ^3 the
   latter of which has nearly tripled since 1975 and accounts for one
   third of the 1.9 billion adults worldwide who are reported to be
   overweight.[31] ^4 As such, understanding how exercise rewires
   metabolism to prevent and mitigate obesity and associated
   cardiovascular disease is becoming an increasingly critical research
   endeavor.

   Mass spectrometry‐based metabolomics and lipidomics approaches have
   revived the field of sports metabolism,[32] ^5 by expanding classic
   concepts including the anaerobic or lactate threshold (equilibrium
   between synthesis and clearance[33] ^6 ) and substrate utilization,[34]
   ^7 with the ability to monitor thousands of small molecule analytes in
   a single comprehensive analysis. Metabolomics studies in the context of
   endurance, resistance, or high‐intensity bouts of exercise have been
   described in the literature.[35] ^8 , [36]^9 , [37]^10 , [38]^11 ,
   [39]^12 Most of these studies have been performed under well‐controlled
   laboratory settings, focusing on exercise and environmental
   characteristics (e.g., high‐intensity concentric‐eccentric exercise[40]
   ^13 under hypoxia[41] ^14 or circadian regulation of exercise in animal
   models).[42] ^15 However, limited literature is currently available on
   the impact of exercise in the field under real world conditions (i.e.,
   outside a controlled laboratory setting).

   A common limitation encountered during the design of these metabolomics
   studies pertains to the sample collection. Options are often limited to
   venipuncture, which requires trained phlebotomists and represents a
   logistical hurdle to sample collection in the field. Recently, we and
   others reported on the use of microcapillary‐based touch activated
   (TAP) devices for blood sampling[43] ^16 and its utility in
   characterizing whole blood metabolomes of elite professional
   cyclists.[44] ^10 TAP sampling, however, is limited by the requirement
   for freezing liquid whole blood samples to ‐80°C immediately after
   collection and throughout storage prior to mass spectrometry‐based
   characterization. As such, methods that provide dried immobilized blood
   samples offer an opportune mechanism to implement metabolomics and
   lipidomics for the characterization of outdoor sports training and
   performance monitoring (reviewed in[45] ^17 ). In addition to dried
   blood spot collection onto filter cards,[46] ^18 volumetric absorptive
   microsampling (VAMS, Neoteryx, Torrance, CA) that also utilizes a
   lancet for blood draw has been implemented for targeted mass
   spectrometry‐based quantification of small molecule metabolites[47] ^19
   and lipids.[48] ^20 Additional advanced dried sample devices that
   couple microcapillary‐based sampling to dried blood spot collection
   such as the TASSO‐M20 (Tasso, Inc., Seattle, WA) offers an alternative
   for ambient‐stable sample collection. As this device is capable of
   quantifying growth hormone[49] ^21 and other illicit doping
   substances,[50] ^22 , [51]^23 , [52]^24 , [53]^25 it has been welcomed
   by the World Antidoping Agency[54] ^26 and is on track for integration
   into the current anti‐doping testing system.[55] ^27 However, a
   metabolomics characterization is limited[56] ^28 to pilot explorations
   for clinical purposes and—to the best of our knowledge—never applied to
   assess the impact of exercise in field testing, which is the focus of
   the present study.

   Here we used an advanced microcapillary‐based blood sampling device to
   collect dried blood samples from 12 volunteers at baseline and after a
   2‐mile run in the field. We then applied a high‐throughput multi‐omics
   approach to analyze these samples, showing proof of concept for
   metabolomics and lipidomics analysis of field samples. Additional
   samples were collected 24 hours after completion of the trial to assess
   the feasibility of multi‐omics‐based characterization of inter‐subject
   heterogeneity in the recovery phase from the same exercise regimen.

2. MATERIALS AND METHODS

   Acute running test: Twelve recreational athletes (n = 8 males and n = 4
   females) voluntarily participated in this trial (demographic details
   provided in Table [57]1). After a 5‐minute warm‐up, subjects ran the
   same 2‐mile course guided at self‐pace, though all subjects started
   together. All study procedures were conducted in accordance with the
   Declaration of Helsinki.

TABLE 1.

   Subject information including means and standard deviations for each
   value
   Subject ID Gender Age (Years) Height (m) Weight (kg) BMI (kg/m2) Time
   (min:sec)
   S1 M 43 1.78 72.6 23.0 14:37
   S2 M 30 1.88 70.3 19.9 14:03
   S3 F 36 1.63 49.0 18.5 14:51
   S4 M 38 1.78 68.0 21.5 14:47
   S5 M 29 1.80 77.1 23.7 14:27
   S6 M 27 1.78 68.0 21.5 11:54
   S7 M 36 1.85 74.8 21.8 11:55
   S8 M 25 1.63 64.9 24.5 17:38
   S9 M 35 1.75 74.8 24.4 14:40
   S10 F 28 1.60 57.6 22.5 14:43
   S11 F 25 1.70 54.4 18.8 12:03
   S12 F 30 1.63 48.5 18.4 17:16
   Mean 32 1.73 65.00 21.5 14:24
   SD 5 0.09 9.74 2.1 1:45
   [58]Open in a new tab

   Dried blood sampling: Dried whole blood samples were obtained
   immediately before and after a 2‐mile run (n = 12), and 24 hours after
   the post‐run sample from a subset of male subjects (n = 7) using the
   TASSO‐M20 device (Tasso, Inc, Seattle, WA, USA) according to
   manufacturer's instructions. The device collects whole blood into 4
   individual dried blood samples (DBS) containing 17.5 µL each after
   which additional blood collects at the bottom of the sample pod thereby
   enabling visual confirmation that each of the spots are saturated.
   Furthermore, volumetric absorption ensures equivalent volumes of whole
   blood are uniformly collected in each spot, thus mitigating the
   “hematocrit effect”.[59] ^29 After collection, samples were dried for
   1 hour under ambient conditions, and upon confirmation that all spots
   were dry, were sealed in the provided airtight containment with a
   desiccant pouch. All samples were stored together at room temperature
   for 24‐48 hours depending on time of collection (pre‐test, immediately
   after or 24 h post‐exercise) before extraction for metabolomics or
   lipidomics analysis.

   Sample Preparation for LC‐MS: Prior to metabolomics or lipidomics
   analysis, individual DBS from each donor were placed in 1.5 ml tubes
   for either metabolite or lipid extraction. DBS were suspended in 180 µL
   of water/methanol (50:50 v/v) for metabolite extraction or 180 µL of
   isopropanol/methanol (50:50 v/v) for lipid extraction. Suspensions were
   sonicated for 20 minutes at room temperature, subsequently incubated at
   4°C for 30 minutes, and then centrifuged for 10 minutes, 18213 g, 4°C.
   Supernatants were isolated for LC‐MS.

   UHPLC‐MS data acquisition and processing: Analyses were performed as
   previously published.[60] ^30 , [61]^31 , [62]^32 Briefly, the
   analytical platform employs a Vanquish UHPLC system (Thermo Fisher
   Scientific, San Jose, CA, USA) coupled online to a Q Exactive mass
   spectrometer (Thermo Fisher Scientific, San Jose, CA, USA). Polar
   metabolite extracts were resolved in singlicate over a Kinetex C18
   column, 2.1 × 150 mm, 1.7 µm particle size (Phenomenex, Torrance, CA,
   USA) equipped with a guard column (SecurityGuard™ Ultracartridge—UHPLC
   C18 for 2.1 mm ID Columns, Phenomenex, Torrance, CA, USA) using an
   aqueous phase (A) of water and 0.1% formic acid and a mobile phase (B)
   of acetonitrile and 0.1% formic acid for positive ion polarity mode,
   and an aqueous phase (A) of water:acetonitrile (95:5) with 1 mM
   ammonium acetate and a mobile phase (B) of acetonitrile:water (95:5)
   with 1 mM ammonium acetate for negative ion polarity mode. The column
   was equilibrated at 5% B, and upon injection of 10 µl of extract,
   samples were eluted from the column using the solvent gradient:
   0.5‐1.1 min 5‐95% B at 0.45 mL/min; hold at 95% B for 1.65 min at
   0.45 mL/min, and then decrease to 5% over 0.25 min at 0.45 ml/min,
   followed by a re‐equilibration hold at 5% B for 2 minutes at
   0.45 ml/min. The Q Exactive mass spectrometer (Thermo Fisher
   Scientific, San Jose, CA, USA) was operated independently in positive
   or negative ion mode, scanning in Full MS mode (2 µscans) from 60 to
   900 m/z at 70000 resolution, with 4 kV spray voltage, 45 sheath gas, 15
   auxiliary gas, AGC target = 3e6, maximum IT = 200 ms. Non‐polar lipid
   extracts were resolved over an ACQUITY HSS T3 column (2.1 × 150 mm,
   1.8 µm particle size (Waters, MA, USA) using an aqueous phase (A) of
   25% acetonitrile and 5 mM ammonium acetate and a mobile phase (B) of
   90% isopropanol, 10% acetonitrile and 5 mM ammonium acetate. The column
   was equilibrated at 30% B, and upon injection of 10 µl of extract,
   samples were eluted from the column using the solvent gradient: 0‐9 min
   30‐100% B at 0.325 mL/min; hold at 100% B for 3 min at 0.3 mL/min, and
   then decrease to 30% over 0.5 min at 0.4 ml/min, followed by a
   re‐equilibration hold at 30% B for 2.5 minutes at 0.4 ml/min. The Q
   Exactive mass spectrometer (Thermo Fisher) was operated in positive ion
   mode, scanning in Full MS mode (2 µscans) from 150 to 1500 m/z at 70000
   resolution, with 4 kV spray voltage, 45 sheath gas, 15 auxiliary gas.
   When required, dd‐MS2 was performed at 17500 resolution, AGC
   target = 1e5, maximum IT = 50 ms, and stepped NCE of 25, 35 for
   positive mode, and 20, 24, and 28 for negative mode. Calibration was
   performed prior to analysis using the Pierce™ Positive and Negative Ion
   Calibration Solutions (Thermo Fisher Scientific).

   Data analysis: Acquired data was converted from raw to mzXML file
   format using Mass Matrix (Cleveland, OH, USA). Samples were analyzed in
   randomized order with a technical mixture injected after every 10
   samples to qualify instrument performance. Metabolite assignments were
   performed using accurate intact mass (sub‐10 ppm), isotopologue
   distributions, and retention time/spectral comparison to an in‐house
   standard compound library (MSMLS, IROA Technologies, NJ, USA) using
   MAVEN (Princeton, NJ, USA).[63] ^33 Lipidomics data were analyzed using
   LipidSearch 4.0 (Thermo Scientific), which provides lipid
   identification on the basis of accurate intact mass, isotopic pattern,
   and fragmentation pattern to determine lipid class and acyl chain
   composition. Graphs, heat maps and statistical analyses (either T‐Test
   or ANOVA), multivariate analyses including Principal Component Analysis
   (PCA), Partial Least Squares‐Discriminant Analysis (PLS‐DA),
   hierarchical clustering analysis (HCA), and metabolite pathway
   enrichment analysis were performed using MetaboAnalyst 5.0.[64] ^34
   Individual metabolite graphs were plotted with GraphPad Prism 9
   (GraphPad Software Inc., La Jolla, CA, USA). Pathway graphs were
   prepared on BioRender.com.

3. RESULTS AND DISCUSSION

3.1. Dried blood spot metabolomics reveals metabolic changes in response to
acute running

   Whole blood samples were obtained from twelve subjects (8 males and 4
   females) before and after a 2‐mile run and dried at ambient conditions
   before subsequent analysis by mass spectrometry‐based metabolomics and
   lipidomics 3 days later (Figure [65]1A), providing relative
   quantitative values for 225 metabolites and 1118 lipids after data
   curation. Partial Least Squares‐Discriminant Analysis (PLS‐DA) of
   metabolomics data discerned samples taken before and after the run
   across the Component 1 axis, which described 22.8% of the variability
   (Figure [66]1B). The top 20 metabolites that contributed to this
   clustering pattern involved intermediates of energy and redox
   metabolism including glycolysis (bisphosphoglycerate, fructose
   1,6‐bisphosphate, glyceraldehyde 3‐phosphate, phosphoenolpyruvate,
   pyruvate, and lactate), the Tricarboxylic Acid (TCA) Cycle (succinate,
   fumarate, malate), guanylates and adenylates (GDP, GTP, adenosine, AMP,
   ADP, ATP), and the Pentose Phosphate Pathway (PPP, 6‐phosphogluconate,
   erythrose 4‐phosphate) (Figure [67]1C). Hierarchical clustering of
   subject samples using the top 50 significant metabolites that changed
   during the run demonstrated distinct circulating metabolite signatures
   and was able to distinguish all samples according to time point except
   for one Pre timepoint sample that clustered together with the Post time
   point samples, and closest to the respective Pre timepoint from this
   subject (Figure [68]1D). Notably, this person was the fastest to
   complete the trial (sub 6 min/mile pace) and the most trained in this
   cohort as the only subject to have completed multiple marathons. Thus,
   the closest relative proximity of these two samples suggests a smaller
   magnitude of metabolic changes within this subject during the run and
   might be related to the relatively high training status of this
   individual with respect to the entire cohort.

FIGURE 1.

   FIGURE 1
   [69]Open in a new tab

   Dried blood spot metabolomics reveals metabolic changes in response to
   acute running. (A) Dried whole blood was sampled from 12 moderately
   active individuals (8 male, 4 female) before and after a 2‐mile run and
   analyzed by mass spectrometry‐based metabolomics and lipidomics. (B)
   Partial Least Squares‐ Discriminant Analysis (PLS‐DA) distinguishes the
   samples on the basis of relative metabolite levels with the top 20
   metabolites contributing to pattern (Variable Importance in Projection,
   VIP) plotted in (C). (D) The top 50 significant metabolites by ANOVA
   were used to hierarchically cluster metabolite patterns and runners

3.2. Energy metabolite changes in response to acute running

   Interrogation of individual metabolites that distinguished pre‐ and
   post‐run profiles highlighted significant effects that running elicits
   on circulating metabolic profiles (Figure [70]2A). While the
   mobilization of glucose via glycogenolysis varied amongst subjects,
   significant glycolytic flux was indicated from steady state
   measurements by decreased early stage and increased late‐stage
   glycolytic intermediates, resulting in the accumulation of both
   pyruvate and lactate (Figure [71]2B). In similar fashion, intermediates
   of the PPP were significantly lower after running, suggesting a
   re‐routing of glucose‐derived carbon towards glycolysis
   (Figure [72]2C). This down‐regulation of the PPP likely occurs in red
   blood cells to fuel late glycolysis in a pH‐dependent process.[73] ^13
   Trends in TCA Cycle intermediates were also significant, with lower
   citrate and higher levels of succinate, fumarate, and malate
   (Figure [74]2D). Indeed, release of succinate into the extracellular
   compartment is also a pH‐dependent process and is coupled with impaired
   mitochondrial bioenergetics.[75] ^35 , [76]^36 In contrast,
   α‐ketoglutarate variably changed along with transamination couple
   glutamate (Figure [77]2E). Significant accumulation of alanine,
   however, suggests elevated alanine aminotransferase (ALT) activity to
   cope with rapidly accumulating pyruvate and prevention of lactate
   generation (Figure [78]2E). In addition to utilization of
   carbohydrate‐derived carbon to fuel the TCA cycle via pyruvate, fatty
   acids also provide a significant source of acetyl‐CoA for TCA cycle
   function. During this run, fatty acid oxidation was highly variable
   amongst subjects with wide ranges of observed liberated fatty acids
   (Figure [79]2F) and acylcarnitines (Figure [80]2G) in circulation.
   These results recapitulate previous metabolomics studies on the impact
   of running[81] ^37 (or other short bouts of high intensity exercise[82]
   ^14 , [83]^38 ) on the circulating metabolome, including activation of
   glycolysis and concomitant accumulation of lactate, in addition to
   elevated levels of TCA cycle intermediates.

FIGURE 2.

   FIGURE 2
   [84]Open in a new tab

   Energy Metabolism. (A) A pathway map for energy generation is shown,
   along with plots for metabolites in (B) Glycolysis, (C) the Pentose
   Phosphate Pathway, (D) the Tricarboxylic Acid (TCA) Cycle, (E)
   corresponding transamination products, (F) free fatty acids, and (G)
   acylcarnitines. y‐axis values given as peak areas (AU) unless otherwise
   noted. P‐values for Post/Pre paired comparisons are indicated as
   * < 0.05, ** < 0.01, *** < 0.001, **** < 0.0001

   In addition, exposome agents such as those obtained by diet (e.g.,
   caffeine), substance usage (cotinine, oxyresveratrol), or medication
   (e.g., ibuprofen) were detected (Supplemental Figure [85]S1). While
   this study was not powered enough to assess any associated impact
   between performance and the presence or abundance of these compounds,
   it would be interesting to expand on the molecular effects of these
   molecules in an exercise setting considering the established impact
   they have on red blood cell function,[86] ^39 especially in the case of
   caffeine[87] ^40 , [88]^41 and the nicotine‐catabolite, cotinine.[89]
   ^42

3.3. Metabolic profiles return to baseline within 24 hours

   To demonstrate the ability of this approach to provide longitudinal
   resolution of metabolic responses to running, we also obtained a sample
   24 hours after the run in a subset (n = 7) of subjects. PLS‐DA
   distinctly clustered the post‐run timepoint away from the overlapping
   samples isolated before the run and 24 hours afterward (Figure [90]3A).
   While 47 metabolites or 17 metabolites significantly decreased or
   increased, respectively, during the run, they predominantly returned to
   baseline levels within 24 hours (Figure [91]3B and C). Only 7
   metabolites remained significantly changed at this timepoint. Compounds
   related to redox management (lactoylglutathione, methionine S‐oxide,
   γ‐glutamyl‐taurine), nitrogen homeostasis (5‐methylthioadenosine,
   argininosuccinate), and fatty acid oxidation (AC(18:4)) remained lower
   after 24 hours, while tryptophan catabolite indole‐5,6‐quionine was
   elevated (Figure [92]3D). Responses to oxidative stress are a
   well‐established impact of exercise.[93] ^43 Thus, these compounds may
   reflect a persistent response to oxidative signaling elicited by the
   run, either in tissues or in red blood cells which have been reported
   to accumulate oxidative lesions after exercise.[94] ^44 In this view,
   persistently lower levels of argininosuccinate and
   5‐methylthioadenosine may reflect alterations to nitrogen homeostasis
   resulting from nitric oxide production to stimulate vasodilation,[95]
   ^45 , [96]^46 as well as mediate ongoing purine salvage in red blood
   cells, which is also modulated by oxidative stress.[97] ^47 , [98]^48
   Finally, higher levels of indole‐5,6‐quinone may indicate an
   involvement of gut microbiota, which has been increasingly appreciated
   to contribute to sport performance[99] ^49 as well as motivation to
   exercise.[100] ^50 While speculative, future studies will be required
   to disentangle the mechanisms underlying these persistent metabolic
   changes after running.

FIGURE 3.

   FIGURE 3
   [101]Open in a new tab

   Recovery of metabolic profiles after 24 hours. (A) A PLS‐DA of samples
   for which a 24‐hour recovery timepoint had also been collected (n = 7).
   (B) Hierarchical clustering of the top 50 significant metabolites by
   ANOVA. (C) Volcano plot of Pre versus Post run for this cohort, and (D)
   Pre versus 24‐hour recovery

3.4. Lipidomic measurements reveal two distinct sub‐groups

   In addition to metabolite profiling, dried blood spot sampling is
   amenable for lipidomics analysis as well. Samples could not be
   distinguished by either unsupervised PCA or partially supervised PLS‐DA
   (Supplementary Figure [102]S2). However, analysis of delta values
   (intra‐subject normalized to Pre), provided distinct patterns. A linear
   model adjusting for the covariate of sex was capable of discerning
   lipids with exercise‐induced changes in a sex‐ and BMI‐dependent
   fashion despite the small cohort studied here (Supplementary Figure
   [103]S3). In addition, PCA of delta lipid values identified two
   distinct, sex‐independent groups (Figure [104]4A). These groups were
   further distinguished by PLS‐DA (Figure [105]4B and Supplementary
   Figure [106]S4) and hierarchical clustering of the top 50 significant
   lipids demonstrated a clear pattern of lipid depletion in Cluster A in
   contrast to accumulation of these same lipids in Cluster B
   (Figure [107]4C). Subsequent re‐analysis of metabolomics delta values
   categorized by these two clusters also revealed distinct patterns by
   PLS‐DA (Figure [108]4D and Supplementary Figure [109]S2). Hierarchical
   clustering of the top 20 significant metabolites between these two
   clusters highlighted that Cluster A blood samples were enriched for
   short chain free fatty acids after the run, while samples from Cluster
   B had elevated levels of long chain free fatty acids (FA(18:2, 18:3,
   20:3, 20:4) which were decreased in Cluster A (Figure [110]4E). In
   light of the fact that three of the five members of Cluster A were the
   first 3 to finish the run, and four of the five members reported being
   highly active, it is interesting to note that this cluster had lower
   levels of lipids and long chain free fatty acids after the run, a
   profile of which has been reported as a marker of fitness and
   associates with performance capacity.[111] ^51 However, steady state
   measurements cannot discern whether fat is utilized more or mobilized
   less, and while this study was not designed prospectively nor powered
   to test this hypothesis, future studies can interrogate fatty acid
   oxidation using isotope tracers[112] ^52 in combination with dried
   blood field sampling coupled with respirometry measurements taken in
   laboratory settings.[113] ^51

FIGURE 4.

   FIGURE 4
   [114]Open in a new tab

   Dried blood spot lipidomics. (A) PCA of delta lipid values
   (intra‐subject fold change of post/pre) categorized by males and
   females identifies two distinct subclusters, referred to as Cluster A
   (red) and Cluster B (blue). (B) PLS‐DA of delta lipid values in Cluster
   A and B. (C) Hierarchical clustering of the top 50 T‐Test significant
   lipids between Cluster A and B. (D) PLS‐DA of delta metabolite values
   in Cluster A and B. (E) Hierarchical clustering of the top 20 T‐Test
   significant lipids between Cluster A and B

3.5. Sex specific metabolic responses to running

   Hierarchical clustering of metabolite fold‐changes (intrasubject
   post/pre ratios) was able to separate male from female runners
   (Figure [115]5A). Pathway enrichment analysis of these fold‐changes
   highlighted that sex‐specific patterns pertinent to glycolysis, the TCA
   Cycle, PPP, and transamination reactions (Alanine, Aspartate, Glutamate
   Metabolism) (Figure [116]5B). Analyzing metabolites central to these
   pathways as a function of sex revealed several significant differences.
   For instance, while early glycolytic and PPP intermediates
   significantly decreased during the run, the decrease was smaller in
   magnitude for female runners who finished with significantly higher
   levels of these compounds than males (Figure [117]5C and D). The clear
   dimorphism observed here may result from the fact that glucose
   6‐phosphate dehydrogenase (G6PD), which is the rate‐limiting enzyme of
   the PPP, is coded by a gene on chromosome X and is expressed higher in
   women.[118] ^53 In contrast, the accumulation of pyruvate and lactate
   was higher in males (Figure [119]5D). These results support reported
   findings in swimmers, which showed that males generated higher
   submaximal blood lactate levels than females in a 30‐minute swim
   test..[120] ^54 In addition, it has been proposed that smaller
   reductions in glycogen content in type I fibers observed in women after
   sprinting may also contribute to lower accumulation of blood
   lactate.[121] ^55

FIGURE 5.

   FIGURE 5
   [122]Open in a new tab

   Comparison of male and female responses to acute running. (A)
   Hierarchical clustering of the top 50 T‐test significant metabolite
   fold changes (intraindividual Post/Pre normalized). (B) A pathway
   enrichment map prepared from intraindividual normalized values
   (Post/Pre) stratified by sex. Metabolites from (C) the PPP, (D)
   glycolysis, (E) TCA Cycle, (F) acylcarnitines, and (G) free fatty acids
   plotted at pre and post run values, with males (circles) on the left
   and females (squares) on the right. y‐axis values given as peak areas
   (AU). P‐values for Post/Pre comparisons are indicated as * < 0.05,
   ** < 0.01, *** < 0.001, **** < 0.0001

   Similar sex‐dependent trends were observed for TCA cycle intermediates
   citrate, succinate, and fumarate (Figure [123]5E). Sex dimorphism in
   carboxylate accumulation is consistent with previous reports describing
   a lower respiratory exchange ratio in women compared to men, indicating
   lesser reliance on carbohydrate oxidation to support fuel requirements
   for exercise.[124] ^56 Tentative explanations for this phenomenon could
   depend on hematological, muscular or cardiovascular differences.
   Hematocrit percentages and hemoglobin levels[125] ^57 are sex‐dependent
   and could contribute to differential reliance on glycolytic versus
   mitochondrial metabolic pathways via altered oxygen delivery. In
   addition, dimorphisms in skeletal muscle,[126] ^58 respiratory muscle
   morphology (females typically have smaller lungs than males even when
   matched for height[127] ^59 , [128]^60 ), and lung geometry,[129] ^61
   ultimately impact VO[2] max (lower in females) and subsequent fatigue
   resistance.

   In addition to carboxylates, males finished the run with higher
   circulating levels of abundant long chain acylcarnitines including
   AC(18:0) and AC(18:1) (Figure [130]5F). In contrast, females finished
   with higher levels of abundant free fatty acids including FA(18:1) and
   FA(18:3), which significantly decreased only in males (Figure [131]5G).
   These results highlight an increased reliance on fatty acid, rather
   than glucose oxidation in females, consistent with reported
   findings.[132] ^62 Female‐predominant elevation in the levels of
   very‐long chain poly‐ and highly unsaturated fatty acids observed after
   exercise are notable considering that fatty acid carbon chain length is
   inversely proportional to the efficiency of fatty acid oxidation in
   mitochondria,[133] ^63 a phenomenon that is at least in part
   counterbalanced by the degree of unsaturation.[134] ^64 Activation of
   endogenous fatty acid desaturases, including those present in red blood
   cells,[135] ^65 is promoted by oxidant stress. In this view, it is
   worth noting that fatty acid synthesis and oxidation are processes that
   rely on the redox equilibrium of the cofactor nicotinamide adenine
   dinucleotide phosphate (NADP), the production of which is also
   predominantly dependent on G6PD activity and level. Thus, these
   observations have clear implications not just in fatty acid
   homeostasis, but also in redox homeostasis. Since exercise‐induced
   oxidant stress is a well‐established phenomenon,[136] ^43 and
   polymorphisms to G6PD are extremely common in humans (∼500 million
   people suffer to some degree of G6PD deficiency that impacts blood cell
   metabolism and susceptibility to stress‐induced hemolysis[137] ^66 ),
   it is interesting to speculate that sex dimorphism and, perhaps, G6PD
   deficiency may play an as of yet unappreciated role in exercise
   performance.

4. CONCLUDING REMARKS

   Successful optimization of field sampling may help pave the way towards
   the “democratization” of metabolomics‐guided approaches for exercise
   interventions, as determined by availability, distribution, and
   economic considerations. This potential is salient for sports training
   and exercise performance applications—a global market for blood lactate
   measurement[138] ^67 alone was valued at $128 M in 2021.[139] ^68
   Indeed, in the Olympic spirit of equity and fairness in sports, all
   athletes, not just elite professionals with economically solid
   sponsors, should have access to latest generation omics for exercise
   and training guidance. Similar considerations hold true for the
   emerging practices of exercise‐based metabolic interventions in fields
   as diverse as cardiopulmonary diseases (e.g., chronic obstructive
   pulmonary disease[140] ^69 or pulmonary hypertension[141] ^70 ) and
   cancer.[142] ^71 , [143]^72 , [144]^73

   One main roadblock to the feasibility of remote testing of at home
   collected blood samples for these emerging medical applications is not
   only tied to collection and shipment protocols, but also to the
   intrinsically elevated costs of metabolomics studies. To tackle this
   issue, we recently developed an ultra‐high‐throughput metabolomics
   method that enables the analysis of ∼1000 samples/day, making it
   economically and logistically feasible to scale metabolomics testing
   for large volumes of samples.[145] ^74 Coupling of dried blood field
   sampling and ultra‐high‐throughput omics approaches can pave the way
   for the dissemination of metabolomics from a laboratory science to a
   more widely applicable discipline; a revolution foretold by early
   application of metabolomics approaches in the fields of clinical
   biochemistry[146] ^75 and medical diagnostics and prognostics,
   especially when coupled to machine learning approaches.[147] ^76 Based
   on our results, we foresee the feasibility of omics applications that
   leverage at‐home sample collection in combination with remote training
   by expert physicians and nutritionists or even simply paired to
   increasingly common at home electronic training work‐out stations
   (e.g., for cycling, running or bodybuilding) to enhance physiological
   responses to these activities and potentially augment or innovate
   capabilities of real‐time health monitoring wearables.

AUTHOR CONTRIBUTIONS

   FC, TN, and AD designed the study. FC performed exercise tests and
   collected the samples in the field setting. TN, FC, and AD performed
   metabolomics analyses; TN and FC performed data analyses and TN
   prepared figures. All authors contributed to writing the manuscript.

CONFLICT OF INTEREST

   The authors declare that TN and AD are co‐founders of Omix
   Technologies, Inc. AD is a SAB member for Hemanext Inc, Macopharma, and
   Forma Therapeutics Inc.

Supporting information

   Supporting Information
   [148]ANSA-4-37-s002.pdf^ (498.2KB, pdf)

   Supporting Information
   [149]ANSA-4-37-s001.xlsx^ (674.1KB, xlsx)

ACKNOWLEDGEMENTS