You are on page 1of 12

Review Article

https://doi.org/10.1038/s42255-020-0258-x

Metabolic communication during exercise


Robyn M. Murphy1,2, Matthew J. Watt3 and Mark A. Febbraio   4 ✉

The coordination of nutrient sensing, delivery, uptake and utilization is essential for maintaining cellular, tissue and whole-body
homeostasis. Such synchronization can be achieved only if metabolic information is communicated between the cells and tissues
of the entire organism. During intense exercise, the metabolic demand of the body can increase approximately 100-fold. Thus,
exercise is a physiological state in which intertissue communication is of paramount importance. In this Review, we discuss the
physiological processes governing intertissue communication during exercise and the molecules mediating such cross-talk.

H
ippocrates, born in ~460 bc on the Greek island of Kos, is with endurance training, whereas the increase in VO2 max occurs
credited with being the first person to believe that diseases to a smaller extent7. Through a series of steps, glucose moieties
are a result of natural causes rather than superstitions or are converted to pyruvate, which has two fates. In the presence of
the will of gods. Acknowledged as the creator of Western medicine, O2, it is converted to acetyl-CoA and enters the Krebs cycle. In the
Hippocrates was also the first documented person to recognize absence of sufficient O2, to maintain at least some ATP production,
that an active lifestyle is beneficial to health, stating that “walking pyruvate undergoes anaerobic metabolism and is converted to lac-
is man’s best medicine,” and “if there is a deficiency in food and tate, in a much less efficient pathway that produces only ~6% the
exercise, the body will fall sick.” Inactivity is now well established to amount of ATP produced by aerobic metabolism.
lead to obesity, the accumulation of visceral fat, insulin resistance, Fatigue is of paramount concern to competing athletes and is
hyperlipidaemia and skeletal muscle atrophy1. Moreover, physical defined as an inability to maintain force production or a progres-
inactivity is now known to increase the risk of acquiring many life- sive decline in performance8. As their names suggest, central fatigue
style diseases including colon, endometrial and breast cancers, type occurs through the brain, whereas peripheral fatigue occurs in skel-
2 diabetes, cardiovascular disease, bone degenerative diseases, dis- etal muscle. The concept of central fatigue is discussed in further
eases involving cognitive impairment (such as dementia), and poly- detail below. Until the early 2000s, lactic acid (lactate and hydro-
cystic ovarian syndrome1–3. Much of the benefit of physical activity gen ions) had been blamed as the primary culprit determining
in preventing disease has been attributed to several mechanisms muscle fatigue, although this conclusion was based on correlative
including weight loss, increased cardiorespiratory fitness (maxi- rather than causative, data9. Research using a minimalist prepara-
mum rate of oxygen consumption, VO2 max) and the maintenance tion, mechanically skinned individual muscle fibres, has provided
of muscle mass2. However, over the past 20 years, interorgan com- clear evidence that lactate per se does not cause the muscle fibre to
munication during exercise has emerged as an alternative mecha- lose the ability to produce force via depolarization-induced calcium
nism through which physical activity can protect the body against release, also known as fatigue10. Indeed, the same minimalist prepa-
an array of lifestyle diseases. ration has also indicated that lactic acid, specifically the increases
in both hydrogen ions (or reduced pH)11 and lactate12, actually con-
Metabolic demand during exercise tribute to the maintenance of excitability of the muscle cell mem-
Metabolic demand during exercise varies greatly, in a manner brane and the ability to undergo repeated cycles of contraction via
dependent on exercise intensity and the requisite ATP supply to excitation–contraction coupling, thus completely debunking the
sustain skeletal muscle contraction4. In the most demanding exer- notion that lactic acid causes muscle fatigue.
cise, such as when an athlete competes in an ‘all-out’ capacity, the The precise cause of muscle fatigue is complex and dependent on
exercise duration is typically <60 s, and the metabolic rate in skel- several factors including the duration of exercise, the species studied
etal muscle can increase ~100-fold (ref. 5). Carbohydrate, fatty acids and the experimental model. During high-intensity exercise and/or
and protein can all be used for ATP production, and the efficiency in the single-fibre model, fatigue is thought to be linked to the accu-
is highest when sufficient oxygen (O2) is continually supplied to the mulation of inorganic phosphate as a result of the breakdown of cre-
muscle and is not limited to the mitochondria, thus ensuring oxida- atine phosphate, which has direct effects on sarcoplasmic reticulum
tive phosphorylation via the tricarboxylic acid cycle (also referred handling of Ca2+, and a decreased ability of the muscle to produce
to as the citric acid or Krebs cycle). force11,13
During periods of high energy demand, such as whole-body The functioning of excitable tissues (that is, the brain and the
exercise, O2 availability cannot meet local demands in the contract- skeletal and cardiac muscle) is energetically demanding. Each of
ing muscle6. The rate-limiting factor is the capacity for O2 delivery these tissues meets some demands with internal stores of glyco-
to skeletal muscle, because oxygen uptake (VO2) is able to meet gen, which can be enzymatically processed to produce glucose.
the delivery rate6. This aspect is perhaps best demonstrated by the Systemically, the liver plays an important role in supplying glucose
observations that mitochondrial oxidative enzymes increase rapidly to the circulation during periods of increased energy demand. The

Department of Biochemistry and Genetics, La Trobe Institute for Molecular Science, La Trobe University, Melbourne, Victoria, Australia. 2Department of
1

Physiology, Anatomy and Microbiology, School of Life Sciences, La Trobe University, Melbourne, Victoria, Australia. 3Department of Physiology, University
of Melbourne, Melbourne, Victoria, Australia. 4Monash Institute of Pharmaceutical Sciences, Monash University, Melbourne, Victoria, Australia.
✉e-mail: mark.febbraio@monash.edu

Nature Metabolism | www.nature.com/natmetab


Review Article NaTuRe MeTaBolISm

Box 1 | Lactate, the first identified muscle secretory factor

The relationship between muscle activity and the production of


IL-6
lactic acid from skeletal muscle was first observed in amphibians Lactate BAIBA
Metrnl
in 1907 (ref. 174). Skeletal muscle is now well known to release Cathepsin B
lactate during exercise, in a manner influenced by both exercise Irisin
intensity and muscle mass175. Generally, the release of lactate from BAIBA
contracting muscle is seen as a mere by-product of intense mus- SPARC
Irisin
Metrnl
cle contractions, and lactate must be rapidly cleared from the
contracting muscle for contractions to continue176. However,
lactate has long been known to be an important fuel source for
other tissues9. The finding that lactate can be oxidized by the brain
dates back more than 80 years (refs. 176,177). However, the notion
that lactate can be a fuel source for the brain was not adopted
until much later. With the advent of new in vivo markers, such as Fig. 1 | Muscle secretory factors during exercise. During muscle
the genetically encoded fluorescence resonance energy transfer contraction, an array of molecules are secreted that communicate with
sensor Laconic, in combination with two-photon microscopy, other organs such as the liver, BAT and WAT, colon and brain. These
the concept that lactate in the brain is differentially present in factors include those that signal to the brain and improve cognition and/or
various compartments and moves among these compartments memory (lactate, Metrnl, cathepsin B and irisin), those that signal to
via monocarboxylate transporters, with a lactate flux from astro- the liver and improve metabolism (IL-6 and BAIBA), those that signal
cytes to neurons, has been supported178. To trace lactate move- to adipose tissue and increase browning or energy expenditure (BAIBA,
ments in humans, van Hall and co-workers have analysed blood irisin and Metrnl) and those that provide protection against colon
samples from the arterial, jugular and femoral veins after the in- cancer (SPARC).
fusion of 13C-labelled lactate in participants cycling at 75% VO2
max179. The authors concluded that lactate contributes 20–25% of
the brain’s total energy demand, probably in preference to glucose
oxidation. The brain is clearly adaptable to substrate provision as (ref. 7). PGC-1α regulates gene expression in skeletal muscle and has
glucose or lactate, and once inside a cell, lactate is oxidized rather a major influence on the muscle’s metabolic response to exercise24.
than stored, thus maintaining a constant driving force for lactate The liquid chromatography–mass spectrometry metabolic screen
uptake by various brain regions179. Whether the increased deliv- identified four metabolites (β-aminoisobutyric acid (BAIBA),
ery of lactate to the brain with exercise is involved in processes GABA (γ-aminobutyric acid), cystine and 2´-deoxycytidine) that
other than substrate provision, such as through other signalling are released by PGC-1α-overexpressing myocytes, two of which
events, is unknown but warrants consideration. appear to have biological roles. BAIBA induces browning of
white fat and β-oxidation in the liver7,25, whereas GABA has been
implicated in muscle fibre-type switching8. The earlier observations
with respect to lactate, coupled with these recent findings, high-
extent to which liver glycogenolysis or gluconeogenesis contributes light the importance of skeletal muscle as a signalling organ during
to hepatic glucose production depends on factors including exercise exercise, in terms of metabolite cross-talk, a component of muscle
intensity, temperature and exogenous carbohydrate availability14,15. communication during exercise that is somewhat clear but often
Skeletal muscle is highly adaptable to enhancing glycogen stores, overlooked (Fig. 1).
and consequently, elite athletes are able to build large stores of gly-
cogen, a process sometimes referred to as glycogen supercompen- Muscle as an endocrine organ. IL-6 as the prototypical myokine. As
sation, as evidenced by the roughly spherical granular structures described above, the concept that skeletal muscle is an endocrine
seen under electron microscopy16. At least in rodent tissue, liver17, organ gained popularity in the early twenty-first century, when sev-
cardiac18 and brain19 glycogen also increases in response to exer- eral studies demonstrated that the cytokine interleukin-6 (IL-6) is
cise training. Glycogen stores are finite in all tissues, although the produced by, and subsequently released from, skeletal muscle dur-
amount of ATP obtainable is variable and dependent on oxygen ing exercise1,26. The term myokine, denoting a cytokine or peptide
availability, as described above. that is produced and released by skeletal muscle and subsequently
exerts paracrine or endocrine effects, was then coined1,26. The
Skeletal muscle as the primary metabolic communicator observation that muscle produces and releases IL-6 during exercise
during exercise and the recognition of myokines spawned this new field of research.
Muscle as a signalling organ. As early as the 1960s, skeletal muscle In a review published 4 years ago, we suggested that to satisfy the
was hypothesized to potentially be more than an organ that helps criteria for being considered a myokine, a protein or peptide must
maintain posture and provide locomotion, when Goldstein postu- (1) be derived from skeletal muscle cells (2) be secreted, either via
lated the existence of a humoral component in skeletal muscle that classical secretory pathways or via extracellular vesicles (EVs), and
may help regulate glucose homeostasis20. This concept that skeletal (3) induce a biological function in an endocrine and/or paracrine
muscle may be a signalling and/or endocrine organ was largely manner3.
ignored until the turn of this millennium, but the evidence that skel- The earliest and most extensively studied bona fide myokine is
etal muscle, is a signalling organ, particularly during contraction, IL-6, which was somewhat serendipitously discovered as the pro-
has been known for several decades21 (Box 1). totypical myokine. Plasma IL-6 was shown to increase in response
Although research has centred on the release of cytokines and/ to strenuous exercise27, and, given that IL-6 is a cytokine, the
or peptides from skeletal muscle, metabolites are also released from blood mononuclear cells (BMNCs) were assumed to be the source.
this organ. Roberts and co-workers22,23 have recently identified However, the messenger RNA27 and protein28,29 expression of IL-6
candidate metabolites released from skeletal muscle by using liq- does not increase in BMNCs during exercise in humans. Moreover,
uid chromatography–mass spectrometry metabolic profiling from the hepatosplanchnic bed clears, rather than releases, IL-6 during
muscle cells overexpressing the transcriptional coactivator PGC-1α exercise30.

Nature Metabolism | www.nature.com/natmetab


NaTuRe MeTaBolISm Review Article
Using muscle biopsy and arteriovenous (a-v) balance techniques, after endurance training43. Intriguingly, however, the authors have
the Pedersen and Febbraio laboratories have demonstrated that identified that the source of the increase in circulating IL-13 during
both the mRNA and protein expression of IL-6 increases in skeletal exercise is type 2 innate lymphoid cells, in a manner not muscle spe-
muscle cells during exercise31–33, and that IL-6 protein net release cific, and rather than studying animals with muscle-specific IL-13
from contacting limbs increases up to 100-fold during muscle con- deficiency, they used animals with whole-body IL-13 deficiency
traction34,35. Importantly, the production of IL-6 via skeletal muscle as their model. Therefore, whether IL-13 is a bona fide myokine
during exercise is somewhat unique. Whereas Il6 mRNA expres- remains to be determined.
sion in BMNCs during inflammation is under transcriptional con-
trol of the p65 subunit of the transcription factor NF-κB, exercise Irisin. Apart from the interleukins described above, other candidate
increases the phosphorylation of c-jun terminal kinase (JNK), the myokines have recently been identified. Similarly to the actions of
reporter activity of the downstream transcription factor AP-1 and the metabolite BAIBA22,25, myokines have been found to result in
the recruitment of AP-1 to the promoter region of the Il6 gene in adipose tissue browning. Using gene expression array analysis of
skeletal muscle36. Moreover, IL-6 production during exercise is skeletal muscle from PGC1α-transgenic mice, which exhibit a simi-
entirely dependent on this transcriptional pathway, because mice lar phenotype to that of endurance-trained animals, Böstrom et al.44
with a muscle-specific deletion of JNK do not show increased IL-6 have identified that irisin, a C-terminally cleaved portion of the pro-
transcription in response to exercise36. tein FNDC5, increases in plasma in response to exercise in both
Despite this definitive evidence that IL-6 is transcribed, trans- mice and humans. The protein is biologically relevant because its
lated and released from skeletal muscle during exercise, whether muscle-specific overexpression improves metabolic homeostasis via
it serves a biological function was unknown until recombinant browning of white adipose tissue (WAT)44. Although the Spiegelman
human IL-6 (rhIL-6) was infused in combination with exercise. One group has demonstrated the existence of irisin via tandem mass
hypothesis for why IL-6 is released during exercise is inhibition of spectrometry (MS)45 and has recently identified a receptor for this
the production of tumour necrosis factor (TNF), which is known ligand46, the data have been challenged, because human FNDC5
to cause pathogenesis in sepsis. Both exercise and rhIL-6 infusion has an unusual start codon of ATA rather than ATG, thus causing
to levels similar to the IL-6 levels during exercise completely sup- critics to suggest that irisin cannot be measured47. Moreover, avail-
pressed the TNF response to a bolus of Escherichia coli lipopoly- able antibodies to irisin have been reported to non-specifically bind
saccharide endotoxin, which induces low-grade inflammation serum proteins, thus prompting questions regarding the widely
in healthy humans37. This study has provided clear evidence that used method of measuring irisin by enzyme-linked immunosorbent
exercise is anti-inflammatory and that the underlying mechanism assys47–49 and highlighting the possible need to use the unequivocal
involves the production of IL-6. tandem MS method to obtain reliable irisin data.
As discussed above, the mechanisms that mediate the tightly con-
trolled production and clearance of glucose during exercise were not Meteorin-like protein. Another exercise-induced myokine that results
fully elucidated until relatively recently, and an unidentified ‘work in adipose tissue browning is meteorin-like protein (Metrnl), iden-
factor’ has been suggested to influence the contraction-induced tified by the Spiegelman group50. Metrnl is induced in muscle with
whole-body glucose production20. Experiments with rhIL-6 have exercise, but it also increases in adipose tissue with cold exposure,
demonstrated that IL-6 released during exercise contributes to the thus leading to increased Metrnl levels in the circulation under
increase in whole-body glucose production in humans38. These data these stimuli. Increasing circulating levels of Metrnl increase energy
provide potential new insights into factors that mediate glucose expenditure and enhance glucose tolerance50 .
production and disposal, and implicate IL-6 as the first myokine Apart from these important myokines, other muscle secretory
that results in muscle–liver cross-talk during contraction. factors, such as myostatin, fibroblast growth factor 21, myonectin,
myostatin, brain-derived neurotrophic factor (BDNF), IL-8, IL-15
Other interleukins. Beyond IL-6, several other cytokines have been and mitochondrially encoded peptide-c have all been suggested to
linked to the skeletal muscle signalling response during exercise. act as myokines. These myokine candidates have been discussed in
Soon after our work describing IL-6 as a myokine, we conducted previous reviews3,51
a cytokine gene expression screen of several immunoregulatory Although most of the above myokines were discovered seren-
cytokines. We detected skeletal muscle mRNA expression levels of dipitously, several groups have attempted to uncover the so-called
IL1Β, IL6, IL8 (official symbol CXCL8), IL18 and TNF from human ‘myokinome’. Using a computational approach, Schiaffino and col-
biopsy samples. Of note, however, the only cytokine other than IL-6, leagues have suggested that more than 300 proteins are potentially
influenced by exercise was IL-8 (ref. 39). Interestingly, as with IL-6, secreted by contracting skeletal muscle52. This number is, however,
the contraction-induced increase in the mRNA expression of IL8 likely to be markedly underestimated, because many proteins that
was exacerbated by low intramuscular glycogen. Critically, unlike lack a signal-sequence peptide and recognition of transmembrane
IL-6, plasma IL-8 did not increase during exercise39, thus suggesting regions can be secreted via EVs53. In fact, through a-v balance stud-
that it is not a contraction-induced myokine. ies across the contracting human limb, purification of EVs from
IL-15 is an interesting cytokine because it increases within skel- arterial and venous blood during exercise, and label-free quantita-
etal muscle during contraction and is linked to skeletal muscle tive proteomic techniques, we have identified more than 600 poten-
growth40. IL-15 has also been demonstrated to affect both adipose tial muscle secretory factors, most of which lack secretory-sequence
tissue mass41 and the ability of adipocytes to secrete adiponec- peptides53. The roles of EVs in tissue cross-talk during exercise are
tin42, thus suggesting muscle–adipose cross-talk during exercise. discussed in greater detail below.
Similarly to IL-8, however, IL-15 has not been found to be released
from the contracting limb during exercise, thereby suggesting that Muscle–brain cross-talk during exercise. Many studies have sug-
it is not a contraction-mediated myokine. gested that regular physical activity can modify both cognition
A recent study has identified IL-13 as a potential myokine. In a and brain biochemistry via circulating exercise-induced factors
comprehensive series of experiments, Lee and colleagues have shown or ‘exerkines’54,55, but identifying whether myokines or exerkines
that IL-13 increases in the plasma in exercise trained humans and in can affect brain function has been challenging. Apart from irisin’s
the skeletal muscle in exercise-trained mice43. Moreover, they have reported role in adipose tissue browning, a different role has recently
found that IL-13-deficient animals have defective fatty acid metab- been uncovered. Irisin levels are diminished in both the hippo-
olism after acute exercise and impaired mitochondrial biogenesis campus and cerebrospinal fluid of mouse models of Alzheimer’s

Nature Metabolism | www.nature.com/natmetab


Review Article NaTuRe MeTaBolISm

disease56. Moreover, decreasing irisin in the brain in these mouse of genes typically associated with energy depletion, whose products
models impairs cognition, but this impairment is rescued if irisin are involved in stress signalling pathways including MAPK, JNK,
levels are subsequently increased56. These recent studies highlight p53 and IL-6-like signalling64.
irisin as an important mediator of the beneficial effects of exercise Transcriptomic analyses of the liver have revealed
in Alzheimer’s disease preclinical models. exercise-induced changes in the mRNAs encoding secreted pro-
The notion that muscle secretory factors may affect cognition is teins, thus suggesting the likelihood of changes in liver protein
not, however, without precedent. Ruas and colleagues have identi- secretion65. Advances in sample preparation, liquid chromatogra-
fied that exercise training influences kynurenine metabolism and phy and MS technology have enabled the deepest coverage of the
protects against stress-induced depression57. They have shown that liver proteome, and 11,520 proteins have been identified in mouse
exercise increases the skeletal muscle expression of kynurenine liver66. Given that ~7% of these proteins enter the bloodstream
aminotransferases, thus enhancing the conversion of kynurenine through secretion, there is clearly a large scope for the involvement
into kynurenic acid, a metabolite unable to cross the blood–brain of hepatokine interorgan cross-talk. In this regard, our group has
barrier. Decreasing plasma kynurenine protects the brain from identified 564 hepatocyte-secreted proteins, 30% of which are clas-
stress-induced changes associated with depression57. This study has sically secreted67. Our recent unpublished work indicates that the
opened therapeutic avenues for the treatment of depression by tar- liver secretes ~2,500 proteins, 20% of which are classically secreted
geting skeletal muscle, without the need to cross the blood–brain and 80% of which are contained within EV’s.
barrier.
Shortly thereafter, Moon et al.58 demonstrated that cathepsin B, Hepatokines as exercise-regulated signals that affect metabo-
a muscle secretory factor, mechanistically underlies the finding that lism. HSP72. Although much of the research focus on hepato-
running is beneficial for neuroregeneration and cognition. Increased kines has been directed at elucidating their roles in the aetiology
cathepsin B levels have been observed in conditioned medium of metabolic diseases63, interest has grown in understanding the
derived from skeletal muscle cell cultures treated with the AMP roles of hepatokines in regulating systemic metabolism during both
kinase agonist AICAR, to mimic contraction. In mice, non-human exercise and recovery. We were one of the first groups to identify a
primates and humans, running increases plasma cathepsin B, which bona fide exercise-induced hepatokine, when we conducted human
importantly correlates with fitness and hippocampus-dependent hepatosplanchnic a-v balance studies almost two decades ago. We
memory function58. These recent findings56–58 suggest that myo- demonstrated that exercise induces the hepatosplanchnic release of
kines contribute to the beneficial effects of exercise on cognition. heat shock protein 72 (HSP72) in healthy humans during prolonged
bicycle exercise68.
Myokine effects on cancer. Multiple epidemiological studies have
indicated that physical activity can prevent several cancers includ- FGF21. Probably the best-studied and most physiologically impor-
ing colon, breast and endometrial cancer59. Circulating secreted tant hepatokine is fibroblast growth factor 21 (FGF-21). FGF21 is
protein acidic and rich in cysteine (SPARC) increases in the plasma expressed in many organs, most prominently the liver, and it has
in both mice and humans during exercise60. The effect of SPARC on been demonstrated to regulate an ever-expanding array of biologi-
colon tumorigenesis in SPARC-deficient mice has also been exam- cal functions69–71. The actions of FGF21 are mediated by binding to
ined. In addition, in an azoxymethane-induced colon cancer mouse FGF receptors and interaction with a coreceptor, β-Klotho, which is
model, regular low-intensity exercise educes the formation of aber- essential for effective FGF21 binding and signal transduction72. The
rant crypt foci in wild-type mice but not in SPARC-null mice60. target-organ selectivity of FGF21 is determined by the restricted tis-
These findings suggest that myokines are a molecular mechanism sue expression of β-Klotho, which is predominantly expressed in
through which exercise can decrease the incidence and/or progres- the liver, pancreas and adipose tissue. Hence, these tissues are con-
sion of certain cancers (Fig. 1). sidered the major targets of FGF21 (ref. 73).
FGF21 was initially shown to stimulate adipose tissue glucose
Role of the liver in intertissue communication and exercise uptake, and subsequent studies using transgenic overexpression of
metabolism FGF21 or FGF21 infusions in mice have shown that FGF21 improves
General considerations. The liver integrates complex signals whole-body insulin sensitivity and glycaemic control, increases adi-
derived from the systemic (that is, arterial blood) and portal (that pose tissue lipolysis and increases energy expenditure74. These effects
is, nutrient-rich blood from the intestines, pancreas or spleen) cir- are associated with diminished adiposity and hepatosteatosis in
culation, thereby regulating an array of metabolic functions includ- obese and diabetic animals74. More recent studies have demonstrated
ing the metabolism of carbohydrates, fatty acids and amino acids, roles of FGF21 in regulating adiponectin release from adipose tissue,
and the production and secretion of lipoproteins and hormones. As adipocyte browning, lifespan extension and protection against vas-
mentioned earlier, glucose derived from liver glycogen provides an cular inflammation and atherosclerotic-plaque formation75.
essential metabolic substrate to contracting skeletal muscle during FGF21 levels increase modestly during exercise in proportion
exercise14,15. The liver also produces very-low-density-lipoprotein to exercise intensity and more dramatically immediately after exer-
triacylglycerol (VLDL-TG), which can contribute as much as cise—effects that are sustained for ~4 hours (refs. 76–78). Although
10–15% of the total energy expenditure in the postabsorptive state61. controversial, there has been recent interest in exercise-induced
VLDL-TG secretion decreases during both exercise and recovery, ‘browning’ of WAT79,80, and FGF-21 is a candidate protein that
whereas VLDL-TG clearance remains unchanged; thus, VLDL-TG may drive this process and decrease adiposity, at least in mice.
oxidation makes a very small contribution to total energy expendi- Experiments in high-fat-fed FGF21-null mice have reported other
ture during exercise62.The extensive vascular network of the liver, impairments in the adaptation to exercise, and an absence of
particularly the ‘open-pore’ sinusoids located between hepatocyte improvements in glucose tolerance or decreases in hepatic triglyc-
planes, facilitates the free exchange of nutrients and hormones eride content81. Together, these studies clearly show that FGF21 is an
within the liver, and the hepatokines released in response to meta- exercise-regulated hepatokine whose action is necessary to achieve
bolic stresses are likely to be prominent in paracrine and autocrine the full metabolic benefits of exercise training.
regulation of hepatocyte function63. Extending the likelihood of sig-
nificant exercise-induced alterations in protein secretion, evidence Follistatin. Muscle growth is a hallmark of exercise training, particu-
from rodent studies has shown that the liver responds to an acute larly resistance exercise, and recent data indicate that a liver-secreted
bout of exercise, at least at the gene level, by upregulating an array factor participates in promoting muscle mass. Follistatin, a

Nature Metabolism | www.nature.com/natmetab


NaTuRe MeTaBolISm Review Article
member of the transforming growth factor (TGF)-β superfam-
ily, exerts pleiotropic actions including inhibition of myostatin to
regulate skeletal muscle growth. The liver is now understood to be FGF21
Follistatin
a primary source of exercise-induced follistatin, and studies using FGF21
hepatosplanchnic a-v balance have demonstrated substantial fol-
listatin release after exercise77,78,82,83. In support of these observations,
a striking increase has been observed in follistatin (Fst) mRNA in
the livers of mice after acute exercise, but not in other tissues includ- ANGPTL4
ing skeletal muscle82. Notably, follistatin does not increase during Follistatin
exercise, thus implicating an important role of follistatin in the HSP72
SeP
adaptation to exercise.
The likelihood that liver-derived follistatin can induce a
change in muscle mass is highlighted by an innovative study
that developed nanoparticles loaded with Fst mRNA to induce
follistatin translation and secretion from the liver84. With this Fig. 2 | Hepatokines during exercise. During exercise, hepatokines are
approach, serum follistatin levels increased to levels comparable secreted that improve brain function (FGF21), pancreatic function (FGF21
to those observed after acute exercise (for example, 2.5-fold above and follistatin), muscle function and metabolism (follistatin, HSP72 and
resting levels). Liver-derived follistatin was sufficient to decrease SeP), and adipocyte biology (ANGPTL4).
serum myostatin, and long-term repeated injections increased mus-
cle mass in mice by 10%. Follistatin exerts numerous other effects in
several target tissues, including modulation of glucagon and insulin during or after exercise in humans94. Nevertheless, these findings
secretion from the pancreas77,85, and improved whole-body glycae- in mice, coupled with a small trial in postmenopausal women, indi-
mic control85,86. cate that pretraining plasma SeP concentrations might be useful for
predicting the ineffectiveness of exercise training on aerobic exer-
ANGPTL4. Regular exercise training induces acute and long-term cise capacity.
adaptations that modulate lipid metabolism, particularly in skel-
etal muscle, adipose tissue and the liver. Angiopoietin-like protein Other hepatokines. Given the available evidence, most exercise-
4 (ANGPTL4) is a circulating protein that is highly expressed in responsive hepatokines appear to increase after rather than dur-
liver and that regulates lipid metabolism by inhibiting lipoprotein ing exercise, thus indicating roles in modulating metabolism in the
lipase activity, thereby decreasing uptake of triacylglycerol-derived recovery from exercise and/or regulation of the adaptation to exer-
fatty acids into tissues and stimulating adipose tissue lipolysis87. cise (Fig. 2). Other hepatokines including GDF15, inhibin βE, IGF1,
Plasma ANGPT4 levels increase only marginally during exercise IGFBP1 and LECT2 are likely to be identified as exercise responsive.
and more prominently in recovery from exercise. These effects However, compelling evidence is currently lacking. These hepato-
appear to be augmented with increasing intensity and duration kines have been reviewed extensively elsewhere95.
of exercise88,89. Examination of tissue ANGPTL4 mRNA levels has
indicated that liver and adipose tissue contribute more than muscle Adipose tissue as a metabolic communicator during
to the exercise-induced increase in circulating ANGPTL4 (ref. 89). exercise
These data were confirmed by Ingerslev et al.90, who have dem- General considerations. Lipolysis in WAT provides FFAs for ATP
onstrated direct secretion of ANGPTL4 from the liver in humans, production at rest and during exercise. During moderate-intensity
which appears to be mediated by a glucagon–cAMP–protein kinase exercise (~50% VO2 max), whole-body adipose tissue lipoly-
A pathway. sis increases approximately twofold by 30 min and progressively
Whole-body fatty acid oxidation is well known to be elevated increases to approximately fivefold higher than that at rest by 4 h.
for several hours after aerobic endurance exercise, partly because Rates of whole-body fatty acid oxidation closely match lipolysis
of increased plasma free fatty acid (FFA) availability secondary to during the first 90 min of moderate intensity exercise, and the FFA
increased adipose tissue lipolysis (~0.5–1.0 mM for 2–6 h)91. The supply from adipocytes is sufficient to provide all substrates for fatty
increase in lipolysis can be mediated by a residual increase in the acid–oxidation requirements at rest and during moderate-intensity
catecholamines, increased growth hormone and decreased insulin exercise91. Although the contribution of adipose-derived FFAs to
levels. ANGPTL4 might contribute to postexercise adipose tissue whole-body ATP production decreases during higher-intensity
lipolysis, particularly as the effects of the other lipolytic regulators exercise (>80% VO2 max), this effect remains substantial in
dissipate over time91. endurance-trained athletes96. The glycerol released from complete
lipolysis can be used by the liver to sustain gluconeogenesis and
Selenoprotein P. Selenoprotein P (SeP) was originally identified as a hepatic glucose output, which are most prominent during very pro-
hepatokine that functions as a selenium-transport protein, and later longed exercise97.
studies showed that SeP contributes to insulin resistance and hyper- Aside from these critical functions, adipose tissue is a major reg-
glycaemia in people with type 2 diabetes92. Intriguingly, recent work ulator of whole-body energy homeostasis through communication
suggests that SeP may explain the individual variations in respon- with the brain and other organs through the secretion of adipokines,
siveness to exercise training in people with metabolic diseases93. SeP which include small molecules, lipids and proteins. MS approaches
appears to signal via LDLR-related protein 1 (LRP1), which func- have identified a large diversity of secreted proteins from adipo-
tions as an uptake receptor of SeP, thus decreasing ROS production cytes98,99 and adipose tissue100. The distinction between adipocyte
in skeletal muscle and consequently blunting AMPK activity, as a and adipose tissue secretion is important because of the complex
result of exercise adaptations including mitochondrial biogenesis cellular composition of adipose tissue, which includes adipocytes
and exercise capacity93. and various populations of adipocyte progenitor cells, an array of
Although these studies using SeP-deficient or LRP1-deficient immune cells, endothelial cells and neurons.
mice provided the first evidence of an intrinsic inhibitory factor Moreover, adipose tissue is located at various anatomical sites,
that attenuates adaptation to exercise, the physiological relevance including the visceral and subcutaneous depots, and one study
remains unresolved, because circulating SeP levels are unchanged using comparison of isotope-labelled amino acid incorporation

Nature Metabolism | www.nature.com/natmetab


Review Article NaTuRe MeTaBolISm

rates MS has reported significant site-specific differences in Adipocyte-derived small molecules. LPA. Aside from proteins,
secretion from human adipose tissues100. A deeper evaluation and adipose tissue secretes several lipids that regulate metabolism.
comparative analysis are required across other adipose tissues and Lysophosphatidic acids (LPAs), consisting of a glycerol backbone, a
their constituent cells. Adding to this complexity, we have recently phosphate group and an ester-bound acyl chain, are potent bioactive
identified three human adipose progenitor cell subtypes that give lipids secreted by adipocytes. Among a multitude of biological roles,
rise to adipocytes with unique endocrine profiles101. The distribu- LPA impairs glucose homeostasis and inhibits insulin secretion in
tion of adipocyte progenitor cell subtypes varies between visceral obese mice121,122. Plasma LPA levels are associated with obesity and
and subcutaneous adipose tissues and is likely to underpin the dif- insulin resistance in mice, rhesus monkeys and humans122,123, and
ferences in protein secretion from these depots. this relationship appears to be dependent on dietary constituents
Adipokines have important roles in a plethora of biologi- and the activity of autotaxin, which catalyses LPA synthesis. The
cal process including but not limited to appetite and satiety, effects of exercise on adipocyte autotaxin secretion and LPA synthe-
insulin action, lipid metabolism, blood pressure, coagulation sis have not been described.
and reproduction102. The secretion of adipokines—including
leptin, adiponectin, retinol-binding protein 4 (RBP4), FGF21, Fatty acid esters of hydroxyl fatty acids. Fatty acid esters of hydroxyl
pigment-epithelium-derived factor, bone morphogenetic protein fatty acids are produced by WAT, and administration of certain spe-
(BMP)-4, BMP-7, vaspin, apelin and progranulin—is altered in obe- cies, namely palmitic acid esters of hydroxystearic acids (PAHSAs),
sity and contributes to a spectrum of obesity-associated diseases. increases glucagon-like peptide 1 and insulin secretion, and
This field has been reviewed in depth by others102. improves insulin action and glycaemic control124. Low circulating
PAHSA concentrations have been associated with insulin resis-
Adipokines and exercise metabolism. TGF-β2. Although exami- tance in humans124, and although the evidence is limited, one study
nation of the effects of adipokines on metabolic regulation, parti­ in older women has reported increased PAHSAs in adipose tissue
cularly in the context of obesity, is a burgeoning field, the effects and serum with exercise training125. Hence, there is a clear scope for
of acute exercise remain somewhat unclear, primarily because of further work examining the roles of acute exercise and prolonged
differences in the phenotypes of study participants and disparities training in regulating adipose-tissue-secreted lipids, to understand
in study designs. Overall, a single exercise bout appears to induce their roles in cellular signalling and their effects in resolving meta-
either no changes or extremely modest changes in circulating adi- bolic disease.
pokines, thus indicating that adipokines are unlikely to be impor-
tant regulators of metabolism in response to acute exercise103. A Effects of exercise on BAT-secreted factors
study by Goodyear and colleagues, however, has recently identi- Batokines. The concept that brown adipose tissue (BAT) is a meta-
fied TGF-β2 as a bona fide exercise-induced adipokine104. Exercise bolic communicator has gained much interest over the past decade
training increases TGF-β2 expression in subcutaneous WAT, serum and has led to the discovery of so-called ‘batokines’, which carry out
and fat explants. Most importantly, improved glucose tolerance this communication (Box 2). The breadth of batokines in regulating
has been observed after transplantation of subcutaneous WAT systemic physiology has been somewhat underappreciated. However,
from exercise-trained mice into donor mice, but critically not into it is not yet clear whether most these proteins are present in the cir-
TGF-β2 deficient donor mice104. Importantly, lactate, whose roles culation and thereby are bona fide endocrine regulators; whether
as a signalling molecule have been discussed at length (Box 1), such proteins circulate at effective concentrations, which is a con-
stimulates TGF-β2 production in adipocytes104, thereby suggesting cern given the relatively small mass of BATs in humans; and perhaps
a model through which muscle/lactate signals to adipose tissue/ most importantly, the identities of their target tissues and biological
TGF-β2, thereby modulating metabolism. actions. This will be an area of great interest in the coming years.

Adiponectin. The beneficial effects of regular physical exercise have BAT-derived lipids. The role of exercise in regulating BAT activa-
long been postulated to be partly mediated by changes in factors tion and protein secretion is not well documented. Although ther-
that are contained within adipose tissue and cause adipose tissue mogenic activation and norepinephrine are known to alter BAT
browning105 and/or adipokine secretion103, given that adipokine lev- protein secretion98, and exercise is highly likely to increase sympa-
els change dramatically in response to exercise training, particularly thetic drive and norepinephrine stimulation to BAT126, definitive
in obese individuals and people with type 2 diabetes103. Perhaps evidence is required to demonstrate exercise-mediated changes in
the most studied adipokine in this context is adiponectin, which protein secretion in vivo.
improves insulin sensitivity and stimulates FFA uptake and/or Perhaps the most exciting work in this domain relates to the
oxidation in muscle. Adiponectin is generally lower in obesity but effects of the so-called lipokines, which are lipids secreted from
importantly is increased after exercise training in overweight and adipose tissue. The lipokine 12,13-diHOME was identified by MS
obese individuals106,107. Similarly, long-term exercise training (≥2 lipidomics to be released from BAT in response to cold in mice and
weeks) is associated with a decrease in plasma leptin levels regard- humans, and to act in an autocrine–paracrine manner in increas-
less of age and sex, to an extent is largely dependent on the amount ing fatty acid uptake, lipolysis and thermogenesis in BAT; moreover,
of fat loss induced by training108. prolonged treatment with this lipokine decreases circulating triglyc-
eride levels127. Studies by Stanford and colleagues have shown that
Other adipokines. Other adipokines known to cause inflammation acute physical exercise in humans and mice increases the circulating
and insulin resistance, perturbed lipid metabolism, or fibrosis or vas- levels of 12,13-diHOME, and acute treatment with this lipokine in
cular dysfunction are consistently decreased with exercise training. mice increase skeletal muscle fatty acid uptake and oxidation, but
These include apelin109, RBP4 (refs. 110,111), visfatin112,113, chemerin114 has no effect on glucose uptake128. These are the first studies dem-
and lipocalin-2 (ref. 115). There are several exceptions; for exam- onstrating an endocrine role of BAT in response to exercise, thus
ple, the insulin-resistance-inducing adipokine PEDF116,117 and the indicating a novel mechanism for BAT–skeletal muscle cross-talk.
lipid-mobilizing factor zinc-α2-glycoprotein118 are unresponsive to
exercise training, whereas vaspin, which improves insulin sensitiv- Role of the brain as a metabolic communicator during
ity in mice, decreases with exercise training119,120. Nevertheless, the exercise
changes in adipokine secretion after exercise training are predicted Central fatigue. The ‘central fatigue hypothesis’ during exercise
to confer positive metabolic effects. postulates that an inability to maintain exercise capacity is associated

Nature Metabolism | www.nature.com/natmetab


NaTuRe MeTaBolISm Review Article
Box 2 | Batokines as metabolic communicators also been linked to cognition and memory, because its expression
in the hippocampus is diminished in people with depression and/
or Alzheimer’s disease141. In contrast, studies in both rodents and
BAT and beige adipose tissue are active secretory tissues releas-
humans have indicated that exercise training and/or increased fit-
ing batokines, which exert local and systemic actions. Brown
ness levels are associated with maintenance or improvements in
adipocytes have long been known to secrete proteins that con-
brain biology and function, and have been suggested to help pre-
tribute, in an autocrine or paracrine manner, to the control of
vent Alzheimer’s disease142–144. The protective role of exercise may be
BAT expansion and activity as well as the browning of WAT180.
due to several mechanisms. Exercise is able to enhance neurogenesis
Effector batokines include the positive regulators prostaglandins,
in the adult mouse brain145,146. It is also able to decrease amyloid-β
FGF-21 and CXCL14, and the negative regulators endothelin-1
pathology146, and animals that are more physically active have
and MSTN. More recent work has shown that BAT secretes pro-
diminished amyloid-β accumulation147. Together, such changes are
teins that signal to other cell types and assist in tissue remodelling
associated with improved memory in rodent behavioural testing146.
during BAT recruitment to thermogenic stimuli, primarily by
The peripheral factors that contribute to such exercise-induced
targeting innervation and vascularization. Prominent examples
changes in the brain remain largely unknown but are now begin-
include BMP8b, CXCL14 and NRG4 (ref. 180). Emerging evidence
ning to be identified and studied.
also indicates that brown adipose acts in an endocrine manner
Circulating BDNF is well known to be elevated during exer-
in influencing metabolism in distant tissues such as the heart181
cise148, and although its expression increases in skeletal muscle
and skeletal muscle182. Work by Kong et al.182 has described the
during exercise148, the increase in blood is likely to be derived from
first evidence of BAT–skeletal muscle cross-talk, demonstrating
platelets148,149. One potential mechanism through which exercise
that the BAT in mice lacking IRF4 secretes myostatin at thermo-
training may confer protection against neurodegeneration is the
neutrality, but activation of BAT lowers myostatin and improves
uptake of BDNF by the brain during exercise. Intriguingly, however,
exercise performance. These findings have potential implications
the brain releases rather than takes up BDNF during exercise150,151,
for muscle growth and exercise capacity; however, further work
thus suggesting that the central nervous system is also a source of
is needed to determine whether this signalling axis is relevant
the increased circulating levels of BDNF during exercise.
in humans and to ascertain the relevance of such cross-talk in
As discussed above, HSP72 was identified as one of the first
the context of regular exercise training. Proteomic analyses and
hepatokines nearly 20 years ago68. Of note, this protein is also
comparative mapping of proteins secreted from cultured hu-
released by the brain during exercise152, but whether this release
man brown and white adipocytes show that although most of
has a physiological role is unclear. Finally, the first identified myo-
the secreted proteins are detected in both adipocyte types, ~100
kine, IL-6, is released from the brain during prolonged exercise in
proteins are secreted exclusively from brown adipocytes98, thus
humans, although the release is neither rapid nor marked153. The
indicating an unappreciated breadth of batokines in regulating
physiological consequences of the release of these molecules from
systemic physiology.
the brain during exercise are unclear, but, as discussed above, some
muscle-derived factors appear to have roles in cognitive function
during exercise. Hence, the area of exercise and ‘tissue cross-talk’
with changes in the synaptic concentrations of neurotransmitters and their potential roles in neurodegenerative brain pathology is
during exercise. This increase in neurotransmitter concentration in of great interest and importance. Experiments are currently being
turn affects muscle function in the absence of hindered motor-unit undertaken to screen the flux of proteins across the human brain
recruitment within the skeletal muscle itself or substantial deple- with techniques developed by the Secher laboratory135. This work
tion of substrates for generation of ATP129–131. Thus, the brain during may shed light on the role of the brain as a metabolic communicator
exercise is acknowledged as a very important functional organ. during exercise.

Cerebral blood flow. Through the pioneering work of the Secher Role of extracellular vesicles in metabolic communication
laboratory, cerebral blood flow is now known to increase by approx- during exercise
imately 25% during exercise, and a parallel increase in metabolism General considerations. As briefly discussed above, in recent
is observed132–134. During activation, an increase in cerebral O2 sup- years, exercise has been increasingly appreciated to transiently
ply is required, because there is no capillary recruitment within stimulate the release of a host of systemic factors into circula-
the brain, and increased metabolism becomes dependent on an tion via EVs. EVs were first described in detail in the early 1980s
enhanced O2 diffusion gradient135. Moreover, the increase in cere- by two research groups154,155, which reported that transferrin
bral oxygenation during exercise136 contrasts with the contracting receptors located on small reticulocytes are secreted into the extra-
skeletal muscle, where oxygenation progressively decreases135. This cellular medium. These studies were largely ignored for many
difference between cerebral and muscle oxygenation is likely to years, because EVs were dismissed as unimportant cellular waste
arise because skeletal muscle can sustain activity even if O2 satura- organelles used only to dispose of molecules unwanted by cells.
tion is decreased by 90% (ref. 137), whereas brain function deterio- However, emerging research has revealed that EVs play a far more
rates when its average O2 saturation is decreased by 10% (refs. 138,139). complex role in the body as a facilitator of cell-to-cell communica-
Therefore, maintaining cerebral blood flow and brain oxygenation tion156,157. EVs are now well accepted to be actively secreted from
is clearly paramount for sustaining exercise performance. However, virtually all cells in the body158, and growing evidence indicates that
the extent to which the brain plays a role as a metabolic communi- EVs have important roles in tissue-to-tissue communication during
cator during exercise is less clear. exercise3,159,160.
EVs transport and deliver bioactive materials such as enzymes,
Myokines and hepatokines in peripheral-organ and central-organ proteins and microRNAs (miRNAs) to target cells in a paracrine
cross-talk. In the brain, lactate is produced in astrocytes and pro- fashion or via the circulation161,162, which contains a heterogeneous
vided as a substrate to nearby neurons. Many of the identified myo- mixture of EVs comprising apoptotic bodies, microvesicles shed
kines and hepatokines discussed above have also been found to from membranes and exosomes. In the absence of markers and
exhibit flux across the brain during exercise. One such peptide is purification strategies to clearly distinguish these subtypes, EVs
BDNF, a member of the neurotrophic family of proteins that facili- of unclear subcellular origin can be grouped according to size,
tates neurogenesis, synaptic plasticity and cell survival140. BDNF has into large EVs and small EVs161,162. Shotgun label-free proteomic

Nature Metabolism | www.nature.com/natmetab


Review Article NaTuRe MeTaBolISm

Transmembrane
Effect of exercise on EV cargo. Exercise-induced EVs secreted into
Phospholipid
proteins bilayer
the circulation are enriched in glycolytic enzymes matching the high
energy demands of exercise53. Of note, glycolytic enzymes delivered
DNAs
by EVs can influence the glycolytic rate in recipient cells170,171, but the
RNAs functional relevance of the enrichment of glycolytic enzymes imme-
Proteins diately after exercise in vivo remains to be elucidated. Many proteins
have been observed in EVs after exercise, including CD36, Flotillin1,
Exosomes
α-sarcoglycan, HSP72, and amyloid-β (reviewed in ref. 105), but the
biological importance of these observations remains unclear.
Through a proteomic screening approach, Whitham et al.53 have
observed 322 proteins whose abundance differs between rest and
exercise, although again, the biological meaning of these changes
is unknown. EVs are also enriched in miRNAs, and several studies
have demonstrated that exercise results in an increase in an array
of miRNAs in EVs (reviewed in ref. 105). Similarly to the observa-
tions with proteins, the biological meaning of such increases is also
Fig. 3 | EVs traffic biological signals during exercise. Proteins, miRNAs, unknown.
DNAs and enzymes are carried in EVs from contracting muscle to the One method to decipher the roles of miRNAs in a biological con-
liver and possibly other organs, such as the brain, pancreas and adipose text has been elegantly demonstrated by Kahn and co-workers. To
tissue, where they release these molecules, thereby modifying biology and process miRNAs, cells must express the miRNA-processing enzyme
possibly providing a mechanism for the protective effects of exercise in Dicer. Briefly, these researchers have demonstrated that mice with
disease states. a fat-specific deletion of Dicer, as well as humans with lipodystro-
phy, have significantly less circulating EV miRNA than controls172.
Moreover, the authors identified several miRNAs in EVs that partic-
analyses of plasma samples have increased in precision in the past ipate in intracellular communication and tissue cross-talk173. During
decade, thus overcoming the challenge of identifying high dynamic exercise, EVs are likely to be liberated by the contracting skeletal
ranges of protein abundance, without limiting the identification muscle, because studies of a-v balance across the contracting limb in
of lower-abundance proteins99. Unbiased, hypothesis-free and humans suggest that many bona fide myokines are contained in EVs
high-coverage proteomic experiments have recently enabled char- in the contracting muscles of the limb53. Therefore, exercising skel-
acterization of the secretome during exercise3,159,160. etal muscle–specific Dicer-deficient animals and adoptively trans-
ferring the purified EVs from these and littermate-control animals
Role of exercise in extracellular-vesicle mobilization. During into resting donor mice may provide insight.
exercise, the number of EVs in the circulation increases approxi- In summary, EVs during exercise are clearly an important mecha-
mately twofold (ref. 53) to fourfold (ref. 163), and the number returns nism through which tissue cross-talk can occur (Fig. 3). Physical
to pre-exercise levels within 4 h of recovery53,163. This exercise- activity is well known to modulate many biological processes in organs
induced increase appears to be highly enriched in small EVs, distant from the contracting skeletal muscle. Whether EVs play roles
because significant increases in the small-EV markers ACTN4 in these processes as well as in the protective effects of exercise in
(ref. 164), ADAM10 (ref. 162), ALIX165, ANAX11 (ref. 162) and CD81 many diseases will be a fertile area of research in the coming years.
(refs. 162,165,166) are elevated in EVs in postexercise samples53. Recently,
platelets, endothelial cells and leukocytes have also been found Conclusions and future directions
to contribute to the exercise-induced release of EVs167, and these Research interest in cell-to-cell and organ-to-organ communication
cellular subtypes may contribute to the increased number of EVs during exercise has grown extensively since the turn of the millen-
in the circulation observed during exercise. As described above, nium. We now know that interorgan communication during exer-
EVs contain a mixture of enzymes, proteins and miRNAs, and cise contributes to the mechanism through which physical activity
recent studies have examined all three of these bioactive materials can protect the body against an array of lifestyle diseases. Although
to some extent. the myokine field has been well studied, the roles of the liver, brain,
The regions to which exercise-induced EVs relocate are crucial and both WAT and BAT beds in mediating metabolism during exer-
to the understanding of the biological importance of their cargo. For cise requires further study.
many years, researchers hypothesized a ‘muscle–liver axis’ during Clearly the most intriguing and emerging field is that of EVs.
exercise168, and early research on the role of IL-6 as a myokine sug- Although several groups have shown that exercise modulates the
gested that this axis is important for maintaining glucose homeosta- number of EVs secreted and the tissues to which EVs are trans-
sis38. To investigate the biodistribution of EVs after their liberation ported after an exercise bout, very little else is known. In fact, to our
into circulation during exercise, Whitham et al.53 have isolated EVs knowledge, no studies have pinpointed the specific cargo sent from
from exercised or non-exercised donor mice, labelled them with one tissue to another during exercise, or the roles, if any, of EVs
lipophilic carbocyanine, intravenously injected them into recipient and their cargo in regulating metabolism and/or contributing to the
mice and analysed the tissue distribution of labelled EVs by intra- protective effect of exercise in the progression of diseases, such as
vital imaging. EVs from exercised, compared with non-exercised, cognitive impairments and prevalent metabolic diseases including
donors revealed marked localization of labelled EVs in the liver type 2 diabetes and non-alcoholic fatty liver disease. This field is
and to a lesser extent, the spleen and lung53. Experiments are cur- likely to be a productive area of research in the near future.
rently underway using stable-isotope labelling with amino acids in
cell culture (SILAC) mice, originally described by the Mann labo- Received: 28 April 2020; Accepted: 2 July 2020;
ratory169. Through donation of purified EVs from SILAC mice, we Published: xx xx xxxx
are currently constructing an exercise/EV mouse atlas in which
we can determine not only the tissue distribution of exercise EVs References
in the donor mice but also the identity of the proteins delivered by 1. Pedersen, B. K. & Febbraio, M. A. Muscles, exercise and obesity: skeletal
these EVs. muscle as a secretory organ. Nat. Rev. Endocrinol. 8, 457–465 (2012).

Nature Metabolism | www.nature.com/natmetab


NaTuRe MeTaBolISm Review Article
2. Booth, F. W., Roberts, C. K. & Laye, M. J. Lack of exercise is a major cause 32. Starkie, R. L., Arkinstall, M. J., Koukoulas, I., Hawley, J. A. & Febbraio, M. A.
of chronic diseases. Compr. Physiol. 2, 1143–1211 (2012). Carbohydrate ingestion attenuates the increase in plasma interleukin-6, but
3. Whitham, M. & Febbraio, M. A. The ever-expanding myokinome: not skeletal muscle interleukin-6 mRNA, during exercise in humans.
discovery challenges and therapeutic implications. Nat. Rev. Drug Discov. J. Physiol. (Lond.) 533, 585–591 (2001).
15, 719–729 (2016). 33. Hiscock, N., Chan, M. H., Bisucci, T., Darby, I. A. & Febbraio, M. A.
4. Gaitanos, G. C., Williams, C., Boobis, L. H. & Brooks, S. Human muscle Skeletal myocytes are a source of interleukin-6 mRNA expression and
metabolism during intermittent maximal exercise. J. Appl. Physiol. 75, protein release during contraction: evidence of fiber type specificity.
712–719 (1993). FASEB J. 18, 992–994 (2004).
5. Hawley, J. A., Maughan, R. J. & Hargreaves, M. Exercise metabolism: 34. Steensberg, A. et al. Production of interleukin-6 in contracting human
historical perspective. Cell Metab. 22, 12–17 (2015). skeletal muscles can account for the exercise-induced increase in plasma
6. Boushel, R. et al. Muscle mitochondrial capacity exceeds maximal oxygen interleukin-6. J. Physiol. (Lond.) 529, 237–242 (2000).
delivery in humans. Mitochondrion 11, 303–307 (2011). 35. Steensberg, A. et al. Interleukin-6 production in contracting human skeletal
7. Saltin, B. & Rowell, L. B. Functional adaptations to physical activity and muscle is influenced by pre-exercise muscle glycogen content. J. Physiol.
inactivity. Fed. Proc. 39, 1506–1513 (1980). (Lond.) 537, 633–639 (2001).
8. Allen, D. G., Lamb, G. D. & Westerblad, H. Skeletal muscle fatigue: cellular 36. Whitham, M. et al. Contraction-induced interleukin-6 gene transcription in
mechanisms. Physiol. Rev. 88, 287–332 (2008). skeletal muscle is regulated by c-Jun terminal kinase/activator protein-1.
9. Brooks, G. A. Lactate doesn’t necessarily cause fatigue: why are we J. Biol. Chem. 287, 10771–10779 (2012).
surprised? J. Physiol. (Lond.) 536, 1 (2001). 37. Starkie, R., Ostrowski, S. R., Jauffred, S., Febbraio, M. & Pedersen, B. K.
10. Dutka, T. L. & Lamb, G. D. Effect of lactate on depolarization-induced Ca2+ Exercise and IL-6 infusion inhibit endotoxin-induced TNF-alpha
release in mechanically skinned skeletal muscle fibers. Am. J. Physiol. Cell production in humans. FASEB J. 17, 884–886 (2003).
Physiol. 278, C517–C525 (2000). 38. Febbraio, M. A., Hiscock, N., Sacchetti, M., Fischer, C. P. & Pedersen, B. K.
11. Pedersen, T. H., Nielsen, O. B., Lamb, G. D. & Stephenson, D. G. Interleukin-6 is a novel factor mediating glucose homeostasis during
Intracellular acidosis enhances the excitability of working muscle. Science skeletal muscle contraction. Diabetes 53, 1643–1648 (2004).
305, 1144–1147 (2004). 39. Chan, M. H., Carey, A. L., Watt, M. J. & Febbraio, M. A. Cytokine gene
12. de Paoli, F. V., Ørtenblad, N., Pedersen, T. H., Jørgensen, R. & Nielsen, O. B. expression in human skeletal muscle during concentric contraction:
Lactate per se improves the excitability of depolarized rat skeletal muscle by evidence that IL-8, like IL-6, is influenced by glycogen availability. Am. J.
reducing the Cl- conductance. J. Physiol. (Lond.) 588, 4785–4794 (2010). Physiol. Regul. Integr. Comp. Physiol. 287, R322–R327 (2004).
13. Westerblad, H., Allen, D. G. & Lännergren, J. Muscle fatigue: lactic acid or 40. Nielsen, A. R. et al. Expression of interleukin-15 in human skeletal muscle
inorganic phosphate the major cause? News Physiol. Sci. 17, 17–21 (2002). effect of exercise and muscle fibre type composition. J. Physiol. (Lond.) 584,
14. Angus, D. J., Febbraio, M. A., Lasini, D. & Hargreaves, M. Effect of 305–312 (2007).
carbohydrate ingestion on glucose kinetics during exercise in the heat. 41. Carbó, N. et al. Interleukin-15 mediates reciprocal regulation of adipose
J. Appl. Physiol. 90, 601–605 (2001). and muscle mass: a potential role in body weight control. Biochim. Biophys.
15. Coggan, A. R. & Coyle, E. F. Reversal of fatigue during prolonged exercise Acta 1526, 17–24 (2001).
by carbohydrate infusion or ingestion. J. Appl. Physiol. 63, 2388–2395 (1987). 42. Quinn, L. S., Strait-Bodey, L., Anderson, B. G., Argilés, J. M. & Havel, P. J.
16. Nielsen, J. N. et al. Glycogen synthase localization and activity in rat Interleukin-15 stimulates adiponectin secretion by 3T3-L1 adipocytes:
skeletal muscle is strongly dependent on glycogen content. J. Physiol. evidence for a skeletal muscle-to-fat signaling pathway. Cell Biol. Int. 29,
(Lond.) 531, 757–769 (2001). 449–457 (2005).
17. Baldwin, K. M., Fitts, R. H., Booth, F. W., Winder, W. W. & Holloszy, J. O. 43. Knudsen, N. H. et al. Interleukin-13 drives metabolic conditioning of
Depletion of muscle and liver glycogen during exercise: protective effect of muscle to endurance exercise. Science 368, eaat3987 (2020).
training. Pflugers Arch. 354, 203–212 (1975). 44. Bostrom, P. et al. A PGC1-α-dependent myokine that drives brown-fat-like
18. Saitoh, S., Shimomura, Y., Tasaki, Y. & Suzuki, M. Effect of short-term development of white fat and thermogenesis. Nature 481, 463–468 (2012).
exercise training on muscle glycogen in resting conditions in rats fed a high 45. Jedrychowski, M. P. et al. Detection and quantitation of circulating human
fat diet. Eur. J. Appl. Physiol. Occup. Physiol. 64, 62–67 (1992). irisin by tandem mass spectrometry. Cell Metab. 22, 734–740 (2015).
19. Matsui, T., Soya, M. & Soya, H. Endurance and brain glycogen: a clue 46. Kim, H. et al. Irisin mediates effects on bone and fat via αV integrin
toward understanding central fatigue. Adv. Neurobiol. 23, 331–346 (2019). receptors. Cell 175, 1756–1768.e17 (2018).
20. Goldstein, M. S. Humoral nature of the hypoglycemic factor of muscular 47. Albrecht, E. et al. Irisin: a myth rather than an exercise-inducible myokine.
work. Diabetes 10, 232–234 (1961). Sci. Rep. 5, 8889 (2015).
21. Owles, W. H. Alterations in the lactic acid content of the blood as a result of 48. Erickson, H. P. Irisin and FNDC5 in retrospect: an exercise hormone or a
light exercise, and associated changes in the CO2-combining power of the transmembrane receptor? Adipocyte 2, 289–293 (2013).
blood and in the alveolar CO2 pressure. J. Physiol. (Lond.) 69, 214–237 (1930). 49. Albrecht, E. et al. Irisin: still chasing shadows. Mol. Metab. 34,
22. Roberts, L. D. et al. β-Aminoisobutyric acid induces browning of white fat 124–135 (2020).
and hepatic β-oxidation and is inversely correlated with cardiometabolic 50. Rao, R. R. et al. Meteorin-like is a hormone that regulates immune-adipose
risk factors. Cell Metab. 19, 96–108 (2014). interactions to increase beige fat thermogenesis. Cell 157, 1279–1291 (2014).
23. Roberts, L. D. et al. Inorganic nitrate mimics exercise-stimulated muscular 51. Priest, C. & Tontonoz, P. Inter-organ cross-talk in metabolic syndrome.
fiber-type switching and myokine and γ-aminobutyric acid release. Diabetes Nat. Metab. 1, 1177–1188 (2019).
66, 674–688 (2017). 52. Bortoluzzi, S., Scannapieco, P., Cestaro, A., Danieli, G. A. & Schiaffino, S.
24. Handschin, C. & Spiegelman, B. M. The role of exercise and PGC1α in Computational reconstruction of the human skeletal muscle secretome.
inflammation and chronic disease. Nature 454, 463–469 (2008). Proteins 62, 776–792 (2006).
25. Kammoun, H. L. & Febbraio, M. A. Come on BAIBA light my fire. Cell 53. Whitham, M. et al. Extracellular vesicles provide a means for tissue
Metab. 19, 1–2 (2014). crosstalk during exercise. Cell Metab. 27, 237–251.e4 (2018).
26. Febbraio, M. A. & Pedersen, B. K. Muscle-derived interleukin-6: 54. Tsai, C. L., Pai, M. C., Ukropec, J. & Ukropcová, B. Distinctive effects of
mechanisms for activation and possible biological roles. FASEB J. 16, aerobic and resistance exercise modes on neurocognitive and biochemical
1335–1347 (2002). changes in individuals with mild cognitive impairment. Curr. Alzheimer
27. Ullum, H. et al. Bicycle exercise enhances plasma IL-6 but does not change Res. 16, 316–332 (2019).
IL-1 alpha, IL-1 beta, IL-6, or TNF-alpha pre-mRNA in BMNC. J. Appl. 55. Tsai, C. L., Ukropec, J., Ukropcová, B. & Pai, M. C. An acute bout of
Physiol. 77, 93–97 (1994). aerobic or strength exercise specifically modifies circulating exerkine levels
28. Starkie, R. L., Angus, D. J., Rolland, J., Hargreaves, M. & Febbraio, M. A. and neurocognitive functions in elderly individuals with mild cognitive
Effect of prolonged, submaximal exercise and carbohydrate ingestion on impairment. Neuroimage Clin. 17, 272–284 (2017).
monocyte intracellular cytokine production in humans. J. Physiol. (Lond.) 56. Lourenco, M. V. et al. Exercise-linked FNDC5/irisin rescues synaptic
528, 647–655 (2000). plasticity and memory defects in Alzheimer’s models. Nat. Med. 25,
29. Starkie, R. L., Rolland, J., Angus, D. J., Anderson, M. J. & Febbraio, M. A. 165–175 (2019).
Circulating monocytes are not the source of elevations in plasma IL-6 and 57. Agudelo, L. Z. et al. Skeletal muscle PGC-1α1 modulates kynurenine
TNF-alpha levels after prolonged running. Am. J. Physiol. Cell Physiol. 280, metabolism and mediates resilience to stress-induced depression. Cell 159,
C769–C774 (2001). 33–45 (2014).
30. Febbraio, M. A. et al. Hepatosplanchnic clearance of interleukin-6 in humans 58. Moon, H. Y. et al. Running-induced systemic cathepsin B secretion is
during exercise. Am. J. Physiol. Endocrinol. Metab. 285, E397–E402 (2003). associated with memory function. Cell Metab. 24, 332–340 (2016).
31. Ostrowski, K., Rohde, T., Zacho, M., Asp, S. & Pedersen, B. K. Evidence 59. Moore, S. C. et al. Association of leisure-time physical activity with
that interleukin-6 is produced in human skeletal muscle during prolonged risk of 26 types of cancer in 1.44 million adults. JAMA Intern. Med. 176,
running. J. Physiol. (Lond.) 508, 949–953 (1998). 816–825 (2016).

Nature Metabolism | www.nature.com/natmetab


Review Article NaTuRe MeTaBolISm
60. Aoi, W. et al. A novel myokine, secreted protein acidic and rich in cysteine 92. Misu, H. et al. A liver-derived secretory protein, selenoprotein P, causes
(SPARC), suppresses colon tumorigenesis via regular exercise. Gut 62, insulin resistance. Cell Metab. 12, 483–495 (2010).
882–889 (2013). 93. Misu, H. et al. Deficiency of the hepatokine selenoprotein P increases
61. Gormsen, L. C. et al. Impact of body composition on very-low-density responsiveness to exercise in mice through upregulation of reactive
lipoprotein-triglycerides kinetics. Am. J. Physiol. Endocrinol. Metab. 296, oxygen species and AMP-activated protein kinase in muscle. Nat. Med. 23,
E165–E173 (2009). 508–516 (2017).
62. Sondergaard, E. et al. Effects of exercise on VLDL-triglyceride oxidation 94. Sargeant, J. A. et al. The influence of adiposity and acute exercise on
and turnover. Am. J. Physiol. Endocrinol. Metab. 300, E939–E944 (2011). circulating hepatokines in normal-weight and overweight/obese men.
63. Watt, M. J., Miotto, P. M., De Nardo, W. & Montgomery, M. K. The liver as Appl. Physiol. Nutr. Metab. 43, 482–490 (2018).
an endocrine organ-linking NAFLD and insulin resistance. Endocr. Rev. 40, 95. Weigert, C., Hoene, M. & Plomgaard, P. Hepatokines: a novel group of
1367–1393 (2019). exercise factors. Pflugers Arch. 471, 383–396 (2019).
64. Hoene, M. & Weigert, C. The stress response of the liver to physical 96. Tsiloulis, T. & Watt, M. J. Exercise and the regulation of adipose tissue
exercise. Exerc. Immunol. Rev. 16, 163–183 (2010). metabolism. Prog. Mol. Biol. Transl. Sci. 135, 175–201 (2015).
65. Hoene, M. et al. Activation of the mitogen-activated protein kinase 97. Ahlborg, G., Felig, P., Hagenfeldt, L., Hendler, R. & Wahren, J. Substrate
(MAPK) signalling pathway in the liver of mice is related to plasma glucose turnover during prolonged exercise in man: splanchnic and leg metabolism
levels after acute exercise. Diabetologia 53, 1131–1141 (2010). of glucose, free fatty acids, and amino acids. J. Clin. Invest. 53,
66. Azimifar, S. B., Nagaraj, N., Cox, J. & Mann, M. Cell-type-resolved 1080–1090 (1974).
quantitative proteomics of murine liver. Cell Metab. 20, 1076–1087 (2014). 98. Deshmukh, A. S. et al. Proteomics-based comparative mapping of the
67. Meex, R. C. et al. Fetuin B is a secreted hepatocyte factor linking steatosis secretomes of human brown and white adipocytes reveals EPDR1 as a novel
to impaired glucose metabolism. Cell Metab. 22, 1078–1089 (2015). batokine. Cell Metab. 30, 963–975.e967 (2019).
68. Febbraio, M. A. et al. Exercise induces hepatosplanchnic release of heat 99. Crowe, S. et al. Pigment epithelium-derived factor contributes to insulin
shock protein 72 in humans. J. Physiol. (Lond.) 544, 957–962 (2002). resistance in obesity. Cell Metab. 10, 40–47 (2009).
69. von Holstein-Rathlou, S. et al. FGF21 mediates endocrine control of 100. Roca-Rivada, A. et al. CILAIR-based secretome analysis of obese visceral
simple sugar intake and sweet taste preference by the liver. Cell Metab. 23, and subcutaneous adipose tissues reveals distinctive ECM remodeling and
335–343 (2016). inflammation mediators. Sci. Rep. 5, 12214 (2015).
70. Talukdar, S. et al. FGF21 regulates sweet and alcohol preference. Cell Metab. 101. Raajendiran, A. et al. Identification of metabolically distinct adipocyte
23, 344–349 (2016). progenitor cells in human adipose tissues. Cell Rep. 27,
71. Soberg, S. et al. FGF21 is a sugar-induced hormone associated with sweet 1528–1540.e1527 (2019).
intake and preference in humans. Cell Metab. 25, 1045–1053.e1046 (2017). 102. Funcke, J. B. & Scherer, P. E. Beyond adiponectin and leptin: adipose
72. Kurosu, H. et al. Tissue-specific expression of βKlotho and fibroblast tissue-derived mediators of inter-organ communication. J. Lipid Res. 60,
growth factor (FGF) receptor isoforms determines metabolic activity of 1648–1684 (2019).
FGF19 and FGF21. J. Biol. Chem. 282, 26687–26695 (2007). 103. Stanford, K. I. & Goodyear, L. J. Muscle-adipose tissue cross talk. Cold
73. Ding, X. et al. βKlotho is required for fibroblast growth factor 21 effects on Spring Harb. Perspect. Med. 8, a029801 (2018).
growth and metabolism. Cell Metab. 16, 387–393 (2012). 104. Takahashi, H. et al. TGF-β2 is an exercise-induced adipokine that regulates
74. Fisher, F. M. & Maratos-Flier, E. Understanding the physiology of FGF21. glucose and fatty acid metabolism. Nat. Metab. 1, 291–303 (2019).
Annu. Rev. Physiol. 78, 223–241 (2016). 105. Otero-Díaz, B. et al. Exercise induces white adipose tissue browning across
75. Babaknejad, N., Nayeri, H., Hemmati, R., Bahrami, S. & Esmaillzadeh, A. the weight spectrum in humans. Front. Physiol. 9, 1781 (2018).
An overview of FGF19 and FGF21: the therapeutic role in the treatment of 106. Becic, T., Studenik, C. & Hoffmann, G. Exercise increases adiponectin and
the metabolic disorders and obesity. Horm. Metab. Res. 50, 441–452 (2018). reduces leptin levels in prediabetic and diabetic individuals: systematic
76. Kim, K. H. et al. Acute exercise induces FGF21 expression in mice and in review and meta-analysis of randomized controlled trials. Med. Sci. (Basel)
healthy humans. PLoS ONE 8, e63517 (2013). 6, 97 (2018).
77. Hansen, J. S. et al. Exercise-induced secretion of FGF21 and follistatin are 107. Yu, N., Ruan, Y., Gao, X. & Sun, J. Systematic review and meta-analysis of
blocked by pancreatic clamp and impaired in type 2 diabetes. J. Clin. randomized, controlled trials on the effect of exercise on serum leptin and
Endocrinol. Metab. 101, 2816–2825 (2016). adiponectin in overweight and obese individuals. Horm. Metab. Res. 49,
78. Willis, S. A. et al. Effect of exercise intensity on circulating hepatokine 164–173 (2017).
concentrations in healthy men. Appl. Physiol. Nutr. Metab. 44, 108. Fedewa, M. V., Hathaway, E. D., Ward-Ritacco, C. L., Williams, T. D. &
1065–1072 (2019). Dobbs, W. C. The effect of chronic exercise training on leptin: a systematic
79. Tsiloulis, T. et al. No evidence of white adipocyte browning after endurance review and meta-analysis of randomized controlled trials. Sports Med. 48,
exercise training in obese men. Int. J. Obes. 42, 721–727 (2018). 1437–1450 (2018).
80. Aldiss, P. et al. Exercise-induced ‘browning’ of adipose tissues. Metabolism 109. Krist, J. et al. Effects of weight loss and exercise on apelin serum
81, 63–70 (2018). concentrations and adipose tissue expression in human obesity.
81. Loyd, C. et al. Fibroblast growth factor 21 is required for beneficial effects Obes. Facts 6, 57–69 (2013).
of exercise during chronic high-fat feeding. J. Appl. Physiol. 121, 110. Numao, S. et al. Effects of exercise training on circulating retinol-binding
687–698 (2016). protein 4 and cardiovascular disease risk factors in obese men. Obes. Facts
82. Hansen, J. et al. Exercise induces a marked increase in plasma follistatin: 5, 845–855 (2012).
evidence that follistatin is a contraction-induced hepatokine. Endocrinology 111. Choi, H. Y. et al. Effects of a combined aerobic and resistance exercise
152, 164–171 (2011). program on C1q/TNF-related protein-3 (CTRP-3) and CTRP-5 levels.
83. Hansen, J. S. et al. Glucagon-to-insulin ratio is pivotal for splanchnic Diabetes Care 36, 3321–3327 (2013).
regulation of FGF-21 in humans. Mol. Metab. 4, 551–560 (2015). 112. Haus, J. M. et al. Decreased visfatin after exercise training correlates
84. Schumann, C. et al. Increasing lean muscle mass in mice via with improved glucose tolerance. Med. Sci. Sports Exerc. 41,
nanoparticle-mediated hepatic delivery of follistatin mRNA. Theranostics 8, 1255–1260 (2009).
5276–5288 (2018). 113. Seo, D. I. et al. Effects of 12 weeks of combined exercise training on visfatin
85. Davey, J. R. et al. Intravascular follistatin gene delivery improves and metabolic syndrome factors in obese middle-aged women. J. Sports Sci.
glycemic control in a mouse model of type 2 diabetes. FASEB J. 34, Med. 10, 222–226 (2011).
5697–5714 (2020). 114. Chakaroun, R. et al. Effects of weight loss and exercise on chemerin serum
86. Tao, R. et al. Inactivating hepatic follistatin alleviates hyperglycemia. concentrations and adipose tissue expression in human obesity. Metabolism
Nat. Med. 24, 1058–1069 (2018). 61, 706–714 (2012).
87. Aryal, B., Price, N. L., Suarez, Y. & Fernández-Hernando, C. ANGPTL4 in 115. Moghadasi, M. & Mohammadi Domieh, A. Effects of resistance versus
metabolic and cardiovascular disease. Trends Mol. Med. 25, 723–734 (2019). endurance training on plasma lipocalin-2 in young men. Asian J. Sports
88. Górecka, M. et al. Effect of mountain ultra-marathon running on plasma Med. 5, 108–114 (2014).
angiopoietin-like protein 4 and lipid profile in healthy trained men. Eur. J. 116. Joham, A. E. et al. Pigment epithelium-derived factor, insulin sensitivity,
Appl. Physiol. 120, 117–125 (2020). and adiposity in polycystic ovary syndrome: impact of exercise training.
89. Norheim, F. et al. Regulation of angiopoietin-like protein 4 production Obesity (Silver Spring) 20, 2390–2396 (2012).
during and after exercise. Physiol. Rep. 2, e12109 (2014). 117. Duggan, C., Tapsoba, J. D., Wang, C. Y., Schubert, K. E. F. & McTiernan, A.
90. Ingerslev, B. et al. Angiopoietin-like protein 4 is an exercise-induced Long-term effects of weight loss and exercise on biomarkers associated with
hepatokine in humans, regulated by glucagon and cAMP. Mol. Metab. 6, angiogenesis. Cancer Epidemiol. Biomarkers Prev. 26, 1788–1794 (2017).
1286–1295 (2017). 118. Ge, S. & Ryan, A. S. Zinc-α2-glycoprotein expression in adipose tissue of
91. Lundsgaard, A. M., Fritzen, A. M. & Kiens, B. The importance of fatty acids obese postmenopausal women before and after weight loss and exercise +
as nutrients during post-exercise recovery. Nutrients 12, 280 (2020). weight loss. Metabolism 63, 995–999 (2014).

Nature Metabolism | www.nature.com/natmetab


NaTuRe MeTaBolISm Review Article
119. Youn, B. S. et al. Serum vaspin concentrations in human obesity and type 2 149. Fujimura, H. et al. Brain-derived neurotrophic factor is stored in human
diabetes. Diabetes 57, 372–377 (2008). platelets and released by agonist stimulation. Thromb. Haemost. 87,
120. Kadoglou, N. P., Vrabas, I. S., Kapelouzou, A. & Angelopoulou, N. The 728–734 (2002).
association of physical activity with novel adipokines in patients with type 2 150. Seifert, T. et al. Endurance training enhances BDNF release from the
diabetes. Eur. J. Intern. Med. 23, 137–142 (2012). human brain. Am. J. Physiol. Regul. Integr. Comp. Physiol. 298,
121. Rancoule, C. et al. Lysophosphatidic acid impairs glucose homeostasis and R372–R377 (2010).
inhibits insulin secretion in high-fat diet obese mice. Diabetologia 56, 151. Rasmussen, P. et al. Evidence for a release of brain-derived neurotrophic
1394–1402 (2013). factor from the brain during exercise. Exp. Physiol. 94, 1062–1069 (2009).
122. D’Souza, K., Paramel, G. V. & Kienesberger, P. C. Lysophosphatidic acid 152. Lancaster, G. I. et al. Exercise induces the release of heat shock
signaling in obesity and insulin resistance. Nutrients 10, 399 (2018). protein 72 from the human brain in vivo. Cell Stress Chaperones 9,
123. Michalczyk, A., Budkowska, M., Dołęgowska, B., Chlubek, D. & Safranow, 276–280 (2004).
K. Lysophosphatidic acid plasma concentrations in healthy subjects: 153. Nybo, L., Nielsen, B., Pedersen, B. K., Møller, K. & Secher, N. H.
circadian rhythm and associations with demographic, anthropometric and Interleukin-6 release from the human brain during prolonged exercise.
biochemical parameters. Lipids Health Dis. 16, 140 (2017). J. Physiol. (Lond.) 542, 991–995 (2002).
124. Yore, M. M. et al. Discovery of a class of endogenous mammalian lipids 154. Harding, C. & Stahl, P. Transferrin recycling in reticulocytes: pH and iron
with anti-diabetic and anti-inflammatory effects. Cell 159, 318–332 (2014). are important determinants of ligand binding and processing. Biochem.
125. Brezinova, M. et al. Exercise training induces insulin-sensitizing PAHSAs in Biophys. Res. Commun. 113, 650–658 (1983).
adipose tissue of elderly women. Biochim Biophys Acta Mol Cell Biol Lipids 155. Pan, B. T. & Johnstone, R. M. Fate of the transferrin receptor during
1865, 158576 (2020). maturation of sheep reticulocytes in vitro: selective externalization of the
126. Sanchez-Delgado, G. et al. Role of exercise in the activation of brown receptor. Cell 33, 967–978 (1983).
adipose tissue. Ann. Nutr. Metab. 67, 21–32 (2015). 156. Kalra, H., Drummen, G. P. & Mathivanan, S. Focus on extracellular vesicles:
127. Lynes, M. D. et al. The cold-induced lipokine 12,13-diHOME introducing the next small big thing. Int. J. Mol. Sci. 17, 170 (2016).
promotes fatty acid transport into brown adipose tissue. Nat. Med. 23, 157. Mathivanan, S., Ji, H. & Simpson, R. J. Exosomes: extracellular
631–637 (2017). organelles important in intercellular communication. J. Proteomics 73,
128. Stanford, K. I. et al. 12,13-diHOME: an exercise-induced lipokine that 1907–1920 (2010).
increases skeletal muscle fatty acid uptake. Cell Metab. 27, 158. Deatherage, B. L. & Cookson, B. T. Membrane vesicle release in bacteria,
1111–1120.e1113 (2018). eukaryotes, and archaea: a conserved yet underappreciated aspect of
129. Davis, J. M. & Bailey, S. P. Possible mechanisms of central nervous system microbial life. Infect. Immun. 80, 1948–1957 (2012).
fatigue during exercise. Med. Sci. Sports Exerc. 29, 45–57 (1997). 159. Vechetti, I. J., Valentino, T., Mobley, C. B. & McCarthy, J. J. The role of
130. Roelands, B., de Koning, J., Foster, C., Hettinga, F. & Meeusen, R. exosomes in skeletal muscle and systematic adaption to exercise. J. Physiol.
Neurophysiological determinants of theoretical concepts and mechanisms https://doi.org/10.1113/JP278929 (2020).
involved in pacing. Sports Med. 43, 301–311 (2013). 160. Whitham, M. & Febbraio, M. A. Redefining tissue crosstalk via shotgun
131. Roelands, B., De Pauw, K. & Meeusen, R. Neurophysiological effects of proteomic analyses of plasma extracellular vesicles. Proteomics 19,
exercise in the heat. Scand. J. Med. Sci. Sports 25, 65–78 (2015). (Suppl. 1). e1800154 (2019).
132. Jørgensen, L. G., Perko, M., Perko, G. & Secher, N. H. Middle cerebral 161. Lötvall, J. et al. Minimal experimental requirements for definition of
artery velocity during head-up tilt induced hypovolaemic shock in humans. extracellular vesicles and their functions: a position statement from the
Clin. Physiol. 13, 323–336 (1993). International Society for Extracellular Vesicles. J. Extracell. Vesicles 3,
133. Jorgensen, L. G., Perko, G. & Secher, N. H. Regional cerebral artery mean 26913 (2014).
flow velocity and blood flow during dynamic exercise in humans. J. Appl. 162. Kowal, J. et al. Proteomic comparison defines novel markers to characterize
Physiol. 73, 1825–1830 (1992). heterogeneous populations of extracellular vesicle subtypes. Proc. Nat. Acad.
134. Jorgensen, L. G., Perko, M., Hanel, B., Schroeder, T. V. & Secher, N. H. Sci. USA 113, E968–E977 (2016).
Middle cerebral artery flow velocity and blood flow during exercise and 163. Frühbeis, C., Helmig, S., Tug, S., Simon, P. & Krämer-Albers, E. M. Physical
muscle ischemia in humans. J. Appl. Physiol. 72, 1123–1132 (1992). exercise induces rapid release of small extracellular vesicles into the
135. Secher, N. H., Seifert, T. & Van Lieshout, J. J. Cerebral blood flow and circulation. J. Extracell. Vesicles 4, 28239 (2015).
metabolism during exercise: implications for fatigue. J. Appl. Physiol. 104, 164. Willms, E. et al. Cells release subpopulations of exosomes with distinct
306–314 (2008). molecular and biological properties. Sci. Rep. 6, 22519 (2016).
136. Ide, K., Horn, A. & Secher, N. H. Cerebral metabolic response to 165. Haraszti, R. A. et al. High-resolution proteomic and lipidomic analysis of
submaximal exercise. J. Appl. Physiol. 87, 1604–1608 (1999). exosomes and microvesicles from different cell sources. J. Extracell. Vesicles
137. Astrand, P. O., Cuddy, T. E., Saltin, B. & Stenberg, J. Cardiac output during 5, 32570 (2016).
submaximal and maximal work. J. Appl. Physiol. 19, 268–274 (1964). 166. Keerthikumar, S. et al. Proteogenomic analysis reveals exosomes are more
138. González-Alonso, J. et al. Brain and central haemodynamics and oncogenic than ectosomes. Oncotarget 6, 15375–15396 (2015).
oxygenation during maximal exercise in humans. J. Physiol. (Lond.) 557, 167. Brahmer, A. et al. Platelets, endothelial cells and leukocytes contribute to
331–342 (2004). the exercise-triggered release of extracellular vesicles into the circulation.
139. Madsen, P. L. & Secher, N. H. Near-infrared oximetry of the brain. J. Extracell. Vesicles 8, 1615820 (2019).
Prog. Neurobiol. 58, 541–560 (1999). 168. Gleeson, M. Interleukins and exercise. J. Physiol. (Lond.) 529, 1 (2000).
140. Mattson, M. P., Maudsley, S. & Martin, B. BDNF and 5-HT: a dynamic 169. Krüger, M. et al. SILAC mouse for quantitative proteomics uncovers
duo in age-related neuronal plasticity and neurodegenerative disorders. kindlin-3 as an essential factor for red blood cell function. Cell 134,
Trends Neurosci. 27, 589–594 (2004). 353–364 (2008).
141. Tsai, S. J. Brain-derived neurotrophic factor: a bridge between major 170. Garcia, N. A., Moncayo-Arlandi, J., Sepulveda, P. & Diez-Juan, A.
depression and Alzheimer’s disease? Med. Hypotheses 61, 110–113 (2003). Cardiomyocyte exosomes regulate glycolytic flux in endothelium by direct
142. Radak, Z. et al. Exercise plays a preventive role against Alzheimer’s disease. transfer of GLUT transporters and glycolytic enzymes. Cardiovasc. Res. 109,
J. Alzheimers Dis. 20, 777–783 (2010). 397–408 (2016).
143. Voss, M. W., Vivar, C., Kramer, A. F. & van Praag, H. Bridging animal and 171. Zhao, H. et al. Tumor microenvironment derived exosomes pleiotropically
human models of exercise-induced brain plasticity. Trends Cogn. Sci. 17, modulate cancer cell metabolism. eLife 5, e10250 (2016).
525–544 (2013). 172. Thomou, T. et al. Adipose-derived circulating miRNAs regulate gene
144. Duzel, E., van Praag, H. & Sendtner, M. Can physical exercise in old age expression in other tissues. Nature 542, 450–455 (2017).
improve memory and hippocampal function? Brain 139, 662–673 (2016). 173. Mori, M. A., Ludwig, R. G., Garcia-Martin, R., Brandão, B. B. & Kahn, C. R.
145. van Praag, H., Kempermann, G. & Gage, F. H. Running increases cell Extracellular miRNAs: from biomarkers to mediators of physiology and
proliferation and neurogenesis in the adult mouse dentate gyrus. disease. Cell Metab. 30, 656–673 (2019).
Nat. Neurosci. 2, 266–270 (1999). 174. Fletcher, W. M. Lactic acid in amphibian muscle. J. Physiol. (Lond.) 35,
146. Choi, S. H. et al. Combined adult neurogenesis and BDNF mimic 247–309 (1907).
exercise effects on cognition in an Alzheimer’s mouse model. Science 361, 175. Van Hall, G. et al. Leg and arm lactate and substrate kinetics during
eaan8821 (2018). exercise. Am. J. Physiol. Endocrinol. Metab. 284, E193–E205 (2003).
147. Lazarov, O. et al. Environmental enrichment reduces Aβ levels and amyloid 176. Schurr, A. Cerebral glycolysis: a century of persistent misunderstanding and
deposition in transgenic mice. Cell 120, 701–713 (2005). misconception. Front. Neurosci. 8, 360 (2014).
148. Matthews, V. B. et al. Brain-derived neurotrophic factor is produced by 177. Quastel, J. H. & Wheatley, A. H. Oxidations by the brain. Biochem. J. 26,
skeletal muscle cells in response to contraction and enhances fat oxidation 725–744 (1932).
via activation of AMP-activated protein kinase. Diabetologia 52, 178. Machler, P. et al. In vivo evidence for a lactate gradient from astrocytes to
1409–1418 (2009). neurons. Cell Metab. 23, 94–102 (2016).

Nature Metabolism | www.nature.com/natmetab


Review Article NaTuRe MeTaBolISm
179. van Hall, G. et al. Blood lactate is an important energy source for the Author contributions
human brain. J. Cereb. Blood Flow Metab. 29, 1121–1129 (2009). R.M.M., M.J.W. and M.A.F. wrote sections of the manuscript. M.A.F. prepared the figures.
180. Villarroya, F., Cereijo, R., Villarroya, J. & Giralt, M. Brown adipose tissue as
a secretory organ. Nat. Rev. Endocrinol. 13, 26–35 (2017).
181. Ruan, C. C. et al. A2A receptor activation attenuates hypertensive cardiac Competing interests
remodeling via promoting brown adipose tissue-derived FGF21. Cell Metab. The authors declare no competing interests.
28, 476–489.e475 (2018).
182. Kong, X. et al. Brown adipose tissue controls skeletal muscle function via Additional information
the secretion of myostatin. Cell Metab. 28, 631–643.e633 (2018). Correspondence should be addressed to M.A.F.
Peer review information Primary Handling Editor: Christoph Schmitt.
Reprints and permissions information is available at www.nature.com/reprints.
Acknowledgements
M.A.F. is supported by a Senior Principal Research Fellowship (APP1116936) and an Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
Investigator Award (APP1194141) from the National Health & Medical Research Council published maps and institutional affiliations.
of Australia. © Springer Nature Limited 2020

Nature Metabolism | www.nature.com/natmetab

You might also like