REvIEwS

Physical activity and muscle–brain

crosstalk
Bente Klarlund Pedersen   
Abstract | Neurological and mental illnesses account for a considerable proportion of the global
burden of disease. Exercise has many beneficial effects on brain health, contributing to decreased
risks of dementia, depression and stress, and it has a role in restoring and maintaining cognitive
function and metabolic control. The fact that exercise is sensed by the brain suggests that
muscle-induced peripheral factors enable direct crosstalk between muscle and brain function.
Muscle secretes myokines that contribute to the regulation of hippocampal function. Evidence
is accumulating that the myokine cathepsin B passes through the blood–brain barrier
to enhance brain-derived neurotrophic factor production and hence neurogenesis, memory
and learning. Exercise increases neuronal gene expression of FNDC5 (which encodes the
PGC1α-dependent myokine FNDC5), which can likewise contribute to increased brain-derived
neurotrophic factor levels. Serum levels of the prototype myokine, IL-6, increase with exercise
and might contribute to the suppression of central mechanisms of feeding. Exercise also
increases the PGC1α-dependent muscular expression of kynurenine aminotransferase enzymes,
which induces a beneficial shift in the balance between the neurotoxic kynurenine and the
neuroprotective kynurenic acid, thereby reducing depression-like symptoms. Myokine signalling,
other muscular factors and exercise-induced hepatokines and adipokines are implicated in
mediating the exercise-induced beneficial impact on neurogenesis, cognitive function, appetite
and metabolism, thus supporting the existence of a muscle–brain endocrine loop.

Physical activity
Over the centuries, several philosophers have expressed survival. Today’s humans often live in an environment
Any bodily movement ideas that are compatible with the existence of a muscle– where exercise is not an integral part of their daily life
produced by skeletal muscles brain endocrine loop: Friedrich Nietzsche said that “All and walking has become ‘a lost art’. Our life has become a
that requires energy truly great thoughts are conceived by walking”1; Søren mismatch with our evolutionary past, and our physically
expenditure.
Kierkegaard said “I have walked myself into my best inactive lifestyle puts us at risk of developing obesity,
thoughts, and I know of no thought so burdensome diabetes, depression and dementia5.
that one cannot walk away from it”2; and Jean-Jacques Both the historical and evolutionary perspectives sug-
Rousseau said that “my mind works only with my legs”3. gest a strong link between muscles and the brain. From a
Humans developed from tree-dwelling apes to Homo scientific perspective, evidence is also now accumulating
sapiens when they started to walk on two legs and that moderate-to-vigorous physical activity has many bene­
became bipedal. Early hominins underwent massive ficial effects on brain health and cognitive function6,7.
development of skeletal muscle and major changes Together with the well-described robust inverse associ-
occurred in their brains in parallel. Noakes and Spedding4 ation between exercise capacity and all-cause mortality8,
suggest that these changes rendered H. sapiens depend- physical activity decreases the risk of diseases such as
ent on physical exercise in order to maintain a healthy dementia9–11 and might also be useful in the treatment
Centre of Inflammation and brain. In other words, exercise does not just help to of this disease12. Exercise can lower the rate of cognitive
Metabolism (CIM) and Centre develop muscle and physical performance but is essen- decline in patients with neurodegenerative disorders
for Physical Activity Research tial to activating and increasing the number of neuronal and in healthy people of all ages13 and seems to have a
(CFAS), Rigshospitalet, connections. Our Palaeolithic ancestors lived in a world positive influence on stress, anxiety and depression12.
University of Copenhagen,
Copenhagen, Denmark.
where they had to exercise to chase down meat. Food Physical exercise has also been shown to have positive
intake fluctuated depending on the success of a hunt and effects on learning, memory and attention14; process-
e-mail: bente.klarlund.
pedersen@regionh.dk physical activity alternated between walking and gather- ing speed and executive functions15; reaction time and
https://doi.org/10.1038/ ing foods and periods of higher-intensity activities such language learning16; motor skills and learning17; verbal
s41574-019-0174-x as hunting, running from a predator or fighting for and visuospatial cognitive test results18; and academic

Nature Reviews | Endocrinology

Reviews

Key points and regeneration 34,35. Myokines mediate signalling

within the muscle and muscle–organ crosstalk to the
• Exercise can indirectly be sensed by the brain via adipose tissue (adiponectin) or the liver, gut, pancreas, adipose tissue, bone, vascular bed
liver (fibroblast growth factor 21 and insulin-like growth factor 1). and skin30,36. In addition, myokines with an anticancer
• Myokines mediate muscle–organ crosstalk to the liver, gut, pancreas, adipose tissue, effect have been identified37,38.
bone, vascular bed, skin and brain. The skeletal muscle secretome might mediate the
• Cathepsin B is an exercise-induced myokine required for exercise-induced effects of exercise on metabolic and cardiovascular
improvement in memory and adult neurogenesis. health36 and might be involved in the defence against
• Exercise enhances neuronal gene expression of FNDC5, the protein product of which malignancy37. Of particular interest to the present
might stimulate brain-derived neurotrophic factor in the hippocampus. Review is the idea that the muscle secretome might be
• Serum levels of the myokine IL-6 increase with exercise, and this myokine might involved in mediating the beneficial effects of exercise
regulate central mechanisms for food intake. on brain health.
• Exercise increases muscular expression of kynurenine aminotransferases, which Thus, the exercise-induced beneficial impact on
convert blood levels of neurotoxic kynurenine to the neuroprotective kynurenic acid, neurogenesis, cognitive function, appetite and meta­
thereby reducing depression-like symptoms. bolism might, at least in part, be mediated by myokine
signalling. Other possible mediators of the effects of
achievement in children and intelligence in adolescents18. exercise on the brain include exercise-induced alter­
Moderate-to-vigorous In addition, exercise improves several basic physiological ation in metabolites39, non-coding RNAs40 and hormo-
physical activity
functions such as sleep19, appetite20 and mood21. The ben- nal responses as well as changes in muscle enzymes that
Any activity with an energy
expenditure of ≥3 metabolic eficial effects of muscle activity on cognition have mostly influence the activity of circulating compounds such as
equivalents (for example, brisk been demonstrated with aerobic exercise; less is known kynurenine41.
walking); the WHO minimum regarding the effects of  resistance training. One study Some investigators have proposed that the sum total
recommendations are 150 min reported that resistance training for 8 weeks did not of all factors released in response to endurance exer-
of moderate-to-vigorous
physical activity each week (or
influence cognitive function22, whereas a meta-analysis cise (including peptides, nucleic acids and metabolites)
20 min or 10,000 steps on most reported that resistance exercise training was associated should be termed ‘exerkines’42. Many such exerkines
days of the week) for adults and with a significant reduction in depressive symptoms23. are released within extracellular vesicles known as
60 min of active playing on most A review of the literature concluded that a large net- exosomes, which represent subcategories of extracel-
days of the week for children
work of brain areas, equal to 82% of the total grey matter lular vesicles that are released into the circulation with
and adolescents.
volume, was modifiable by physical activity24. The hip- exercise43. Exosomes can contain peptides, nucleic
Aerobic exercise pocampus, which is involved in memory and learning, is acids, microRNA, mRNA and mitochondrial DNA. An
Exercise involving dynamic the brain region most affected by exercise. Exercise stud- acute bout of endurance exercise increases circulating
movements and large muscle ies in mice have documented that exercise leads to an exosomes and might mediate inter-tissue signalling44.
groups that predominantly rely
on aerobic metabolism for
increase in the volume of the hippocampus as well as an Together with circulating myokines and other mol-
fuelling muscle contractions; increase in blood flow to this part of the brain25. In par- ecules produced and released by skeletal muscles in
examples include jogging, ticular, exercise influences neurogenesis in the dentate response to exercise, direct feedback from skeletal mus-
running, swimming and rowing. gyrus in the hippocampus6,7. In various mouse models, cle through the peripheral nervous system and central
exercise has been shown to increase synapse plasticity as nervous system (CNS) to the brain is also involved in
Resistance training
Movement performed against well as to induce morphological changes in dendrites6,7. muscle-to-brain communication.
a specific external force that is A major question is which peripheral factor or fac-
regularly increased during tors enable direct crosstalk between working muscle The role of BDNF
training; examples include and brain function, including hippocampal-dependent The neurotrophin known as brain-derived neurotrophic
weightlifting and exercises
using resistance machines.
mood, learning, memory and appetite. factor (BDNF) seems to be a very important mediator of
the effects of exercise on the brain, especially cognition45.
Exercise training Myokines and other exercise mediators When rats have access to voluntary exercise, increased
A subset of physical activity that Robust findings regarding exercise and brain health BDNF levels are found in the hippocampus compared
is planned, structured and
suggest the existence of a muscle–brain endocrine loop, with mice without such access46,47. Wheel running for
repetitive and has a final or
intermediate objective of but it is not entirely clear which peripheral mechanisms 6 h was shown to increase Bdnf mRNA expression48 and
improving or maintaining elicit these positive effects of exercise. However, the past free access to activity wheels for 1–8 weeks was shown
physical fitness. The terms almost two decades have taught us that skeletal muscle to increase hippocampal Bdnf mRNA expression and
‘exercise’ and ‘exercise training’ is a secretory organ26. BDNF protein levels49–51. In addition to influencing BDNF
are used interchangeably to
refer to the cardiovascular
Muscle cells are highly metabolically active, and during concentrations directly, exercise has also been shown to
adaptations produced by this exercise skeletal muscles communicate with other organs modulate BDNF‐linked pathways52.
specific type of physical activity; by producing and releasing so-called myokines. The skel- Studies have demonstrated that BDNF is mechanis-
a single bout of exercise is etal muscle secretome in humans consists of hundreds of tically involved in mediating the exercise-dependent
referred to as ‘acute exercise’.
myokines, which are secreted from muscle cells during increase in proliferation of hippocampal dentate gyrus
Myokines proliferation and differentiation or in response to mus- cells and that BDNF is required for exercise-induced
Cytokines or peptides cle contractions. Myokines can exert autocrine, paracrine enhancement of cognitive functions such as memory
produced by skeletal muscle or endocrine effects27–33. Some myokines are involved in and learning53,54.
cells and subsequently energy supply during acute exercise, and repetitive acute The release of BDNF from the human brain has been
released into the circulation,
where they exert autocrine,
bouts are probably involved in mediating adaptation to shown to increase with an acute bout of exercise55,56,
paracrine or endocrine effects training in various organs. At rest, myokines are involved suggesting that exercise also mediates central BDNF
in other cells, tissues or organs. in the regulation of muscle proliferation, differentiation production in humans. BDNF promotes the growth

www.nature.com/nrendo
 Reviews

and proliferation of cells in the hippocampus, and aer- exercise but demonstrated depression-like symptoms
obic exercise training for 3 months has been shown to when they were forced to swim61. This latter finding is
increase hippocampal volume in healthy individuals in accordance with the well-known role of hippocampal
and in patients with schizophrenia by 12% and 16%, BDNF in protecting against anxiety and depression62.
respectively57. BDNF is involved in neuronal differenti- The elegant study by Moon et al.61 shows that the
ation, plasticity, cell survival, hippocampal function and myokine cathepsin B might be causally involved in medi-
learning58. Multiple studies support the idea that BDNF ating the exercise-induced improvement in hippocam-
has a dominant role in mediating the effects of physical pal neurogenesis, memory and learning. However, it
activity on cognitive changes59. However, the underlying remains to be shown to what extent this myokine is
mechanisms leading to increased BDNF levels during a determining factor in exercise-induced enhanced
exercise-induced cardiovascular and muscular activity cognitive function in humans61,63.
are unclear.
We showed that BDNF is a contraction-inducible PGC1α–FNDC5–BDNF pathway
protein in human skeletal muscle that is able to enhance FNDC5 is a membrane protein that is cleaved and
lipid oxidation in skeletal muscle via AMP-activated secreted into the circulation as the myokine irisin,
protein kinase (AMPK) activation60. However, we found which is known for its  browning effects in white adi-
little indication that muscle-derived BDNF was released pose tissue and might also be critical in mediating the
from muscle into the bloodstream and thus no evidence effects of physical activity on BDNF protein expression
that BDNF mediates a direct muscle–brain interaction60. in the brain (Fig. 1b). Exercise induces upregulation of
The question then remained as to whether exercise pro- PGC1α in skeletal muscle, and PGC1α has a central role
vokes muscle-derived circulating factors that can pass in mediating many of the metabolic effects of exercise
through the blood–brain barrier and stimulate BDNF locally within the muscle64. PGC1α is a transcriptional
production in the brain. co-activator of mitochondrial biogenesis and oxidative
metabolism in muscle and brown adipose tissue65. With
Cathepsin B and BDNF regulation exercise, PGC1α expression in muscle increases and
In 2016, Moon et al.61 identified a signalling pathway stimul­ates an increase in the expression of FNDC5 and is
linking exercising muscle with hippocampal function. involved in driving a shift of white adipose tissue into a
They found that in mice, exercise induces elevated brown-fat-like phenotype66.
systemic levels of a novel myokine, cathepsin B, which The group that showed that exercise-induced
promotes hippocampal expression of BDNF and stimu­ increased PGC1α expression leads to increased FNDC5
lates neurogenesis (Fig. 1a). First, L6 myotubes were expression66 later showed that endurance exercise leads to
treated with the AMPK agonist 5-aminoimidazole- an elevation in Fndc5 gene expression in the hippocampus
4-carboxamide ribonucleotide (AICAR) in an attempt to of mice58. Pgc1a−/− mice show reduced Fndc5 expression
model the effects of exercise in vitro. Subsequent screen- in the brain. FNDC5 overexpression in primary corti-
ing of the culture media for proteins using proteomic cal neurons stimulates an increase in BDNF expression.
and biochemical analyses led to the identification of By contrast, RNA interference-mediated knockdown of
elevated levels of cathepsin B in conditioned medium FNDC5 leads to a reduction in Bdnf gene expression66.
derived from skeletal muscle cell cultures treated Interestingly, increasing systemic levels of irisin by deliv-
with the AMPK agonist AICAR61. The researchers ery of FNDC5 to the liver via adenoviral vectors led to
then demonstrated that running resulted in increased increased expression of hippocampal Bdnf 66.
muscular expression of the Ctsb gene (which encodes The critical question is how exercise is sensed by
cathepsin B) in mice, which was accompanied by an the brain and how muscle can activate the neurologi-
increase in cathepsin B protein levels in plasma. The cal PGC1α–FNDC5–BDNF pathway 67. The finding
findings in rodents were supported by the finding of that expression of genes with potential neuroprotective
increased plasma levels of cathepsin B in rhesus maca­ functions, such as BDNF, is increased in the brain when
ques as well as in humans following 4 months of tread- systemic concentrations of FNDC5 are elevated suggests
mill training61. The authors also provided evidence that that the muscle-secreted form of FNDC5 might cross the
cathepsin B was able to cross the blood–brain barrier blood–brain barrier to exert its influence in the CNS —
in mice61. Although cathepsin B application did not for example, by inducing BDNF expression in the
affect hippocampal cell proliferation, it led to increased hippocampus — thereby playing a role in neurogenesis
Browning
expression of Bdnf mRNA and increased BDNF protein and reward-related learning and motivation58.
Browning can be defined as levels as well as increases in levels of doublecortin, a Whether exercise leads to increased levels of irisin
any substantial increase in the protein that has neuroprotective effects as it enhances in the bloodstream is controversial68. Some researchers
expression of uncoupling neuronal migration61. reported that neither acute nor chronic endurance or
protein 1 (UCP1) at the mRNA
To examine whether cathepsin B is directly involved resistance exercise led to increased FNDC5 expression
level occurring in what is
normally considered as a white in mediating exercise-induced hippocampal neuro- or circulating concentrations of irisin in healthy men69.
adipose tissue depot. The genesis and improvement in hippocampal function, Another study showed that a bout of acute exercise
resulting brown adipocytes Moon et al.61 performed studies in Ctsb-knockout mice. had no effect on muscle FNDC5 expression, whereas
that express UCP1 and appear They found that in wild-type control mice, exercise 20 days of high-intensity interval training led to acute
in white adipose tissue are
referred to as beige, brite,
induced the previously found enhancement of neurogen- increases in muscle expression of FNDC5 in healthy
convertible, ectopic, inducible esis and improvement of spatial memory. However, mice men70. Concerns have been expressed regarding the lack
or recruitable. lacking Ctsb did not show these effects with voluntary of specificity of anti-irisin antibodies68 used in the above

Nature Reviews | Endocrinology

Reviews

a Blood–brain
barrier
Muscle

Exercise

↑ BDNF
↑ Doublecortin
Cathepsin B
Hippocampal
neurogenesis

Hippocampal-dependent
learning, memory and mood

↑ PGC1α
Exercise

FNDC5
Irisin Irisin ↑ BDNF

Hippocampal
neurogenesis

Hippocampal-dependent
learning, memory and mood

c Tryptophan

Liver KYN

↑ KAT

Exercise ↑ PGC1α
KYNA KYN Depression

Fig. 1 | Ways in which exercise might beneficially affect neurogenesis, learning, memory , mood and depression-like
symptoms. a | The myokine cathepsin B is released by skeletal muscle during exercise and might influence neurogenesis,
learning, memory and mood. AMP-activated protein kinase (AMPK) activation elicits cathepsin B secretion in skeletal muscle
cells. Running increases cathepsin B levels in mouse gastrocnemius muscle, and exercise leads to an increase in levels of the
myokine cathepsin B in the plasma of mice, rhesus macaques and humans. In vivo studies provided evidence that, in mice,
peripheral cathepsin B crosses the blood–brain barrier. In vitro studies on hippocampal progenitor cells showed that
cathepsin B increased both mRNA and protein levels of brain-derived neurotrophic factor (BDNF) as well as doublecortin.
These factors are known to be important for neuronal migration and neurogenesis and thereby affect learning, memory and
mood. b | FNDC5, a membrane protein that is cleaved and secreted into the circulation as the myokine irisin during exercise,
might influence BDNF and influence neurogenesis, learning, memory and mood. Exercise in mice and humans induces an
upregulation of PGC1α in skeletal muscle. In mice, PGC1α expression in muscle stimulates an increase in the expression
of FNDC5. Exercise also leads to a PGC1α-dependent elevation of FNDC5 in the hippocampus of mice. Peripheral delivery of
FNDC5 to the liver via adenoviral vectors, resulting in elevated blood levels of irisin, induces expression of BDNF in the
hippocampus, suggesting that irisin passes through the blood–brain barrier and induces BDNF expression in the brain, which
will lead to enhanced learning, memory and mood. c | High levels of the neurotoxic kynurenine (KYN) are associated with
depression. Exercise enhances the PGC1α-dependent muscular expression of the enzyme kynurenine aminotransferase (KAT),
which converts neurotoxic KYN into neuroprotective kynurenic acid (KYNA), thereby reducing depression-like symptoms. In
contrast to KYN, KYNA is not able to pass through the blood–brain barrier. The imbalance between the neuroprotective KYNA
and the neurotoxic KYN metabolites has been proposed to be critical for the development of depression.

www.nature.com/nrendo
 Reviews

studies, but a mass spectrometry state-of-the-art of IL-6 has been shown to lead to an improvement in
method showed that human irisin exists, circulates and insulin action in inflammatory murine models86.
is ­regulated by exercise71. In contrast to the above findings, a large number
Some researchers found that exercise provokes of studies have identified IL-6 to have a beneficial
an increase in irisin levels in human plasma72, and role in metabolism regulation. IL-6-deficient mice are
confirmation of these findings would support the characterized by whole-body insulin resistance and
­possible existence of a muscle–brain endocrine loop. late-onset obesity87,88. Studies in rodents have shown
An exercise-induced release of muscle-derived irisin that IL-6 stimulates expansion of pancreatic α-cells
into the blood might provide a link between PGC1α, in the obese state 89 and promotes enhancement of
FNDC5 and hippocampal expression of BDNF. New glucagon-like peptide 1 (GLP1) production and con-
findings in 2019 suggest a role for FNDC5 and irisin sequently increased insulin secretion89. Moreover, IL-6
in regulating synaptic function and memory in mouse signalling in murine macrophages and hepatocytes
models of Alzheimer disease73. induces anti-inflammatory effects, thereby improving
glucose homeostasis90,91.
β-Hydroxybutyrate and BDNF Studies in humans clearly show that physiological
Exercise might also induce increases in BDNF levels concentrations of IL-6 have multiple metabolic effects.
by altering the epigenetic markers of the BDNF pro- When IL-6 is infused into healthy humans, it enhances
moters74. Physical activity induces multiple metabolic insulin-stimulated glucose uptake92, stimulates lipolysis
changes, and it is conceivable that an exercise-induced and fat oxidation93 and delays gastric emptying with
endogenous molecule can serve as a regulator of BDNF beneficial effects on postprandial glucose control94.
transcription. Moreover, acute infusion of IL-6 to humans inhibited
Ketone bodies are markedly increased in the circula- endotoxin-induced elevation in circulating TNF levels,
tion and brain after fasting, dieting and intense exercise75. suggesting an anti-inflammatory effect of IL-6 (ref.95).
Under conditions of reduced glucose levels, ketone bod- IL-6 infusion into humans also stimulates the produc-
ies, in the form of β-hydroxybutyrate and acetoacetate, tion of the anti-inflammatory molecules IL-1 receptor
serve as an energy source75. Prolonged exercise has been antagonist and IL-10 (ref.96).
shown to stimulate an increase in β-hydroxybutyrate Interestingly, although IL-6 seems to be released from
levels in the blood75. β-Hydroxybutyrate crosses the fat tissue in the obese state, during a bout of exercise
blood–brain barrier, accumulates in the hippocampus IL-6 is released by myocytes, leading to an exponential
and increases the expression of BDNF76. Direct applica- increase in plasma levels of IL-6; levels are up to 100-fold
tion of β-hydroxybutyrate into the brain of rats has been higher than basal levels immediately after exercise and
shown to increase the expression of BDNF76. Previous then return to basal levels within a couple of hours after
work with β-hydroxybutyrate showed that it is an effec- exercise. Muscle-derived IL-6 is associated with improved
tive neuroprotective agent in experimental models of insulin sensitivity and fat oxidation30,97. The exercise-
Huntington disease77 and Parkinson disease78, protecting induced increase in IL-6 is an acute effect mediated by
striatal and dopaminergic neurons, respectively. working muscles. The beneficial effects of regular exer-
Taken together, the available evidence suggests that cise might result from the accumulation of repeated
β-hydroxybutyrate provides a link between aerobic acute bouts of exercise-induced alterations in homeo-
exercise and BDNF gene expression in the brain. stasis. The signalling pathways for IL-6 differ between
muscle cells and macrophages. In macrophages, IL-6
The sympathetic nervous system and BDNF transcription requires activation of the nuclear factor-κB
Physical exercise is accompanied by the activation of the (NF-κB) signalling pathway. In myocytes, IL-6 is reg-
sympathetic nervous system, which can have multiple ulated by the Ca2+–NFAT (nuclear factor of activated
effects on peripheral organs, as well as on the brain79. T cells) and glycogen–p38 MAPK (mitogen-activated
β2-Adrenergic stimulation in rats has been shown to protein kinase) pathways. Therefore, although IL-6 cre-
reduce expression of inflammatory cytokines and ates a pro-inflammatory response in macrophages, the
to increase expression of BDNF in the hippocampus80. contraction-induced activation of IL-6 in muscle cells is
In addition, the noradrenaline reuptake inhibitor, venla- independent of NF-κB activation97.
faxine, has been shown to increase expression of BDNF The role of IL-6 in muscular adaptations to exercise
protein in the frontal cortex, which might represent the training has been studied extensively. Contracting skel-
mechanism whereby venlafaxine improves depression81. etal muscles produce IL-6 (ref.98) in a TNF-independent
fashion99, suggesting that muscle-derived IL-6 has a role
IL-6 and appetite regulation in metabolism rather than in inflammation.
Epidemiological studies have shown that IL-6 is impli- Several studies have demonstrated that the release
cated in the chronic inflammation that accompanies of IL-6 from muscle is regulated by muscle glycogen
conditions such as obesity and type 2 diabetes melli- content and glucose ingestion, as reviewed elsewhere97.
tus30. The origin of elevated plasma IL-6 levels in these The exercise-induced increase in plasma IL-6 level is
conditions seems to be immune cells located in adipose diminished by glucose intake during exercise. In addi-
tissue82. Increased systemic IL-6 levels in humans are tion, muscular IL6 mRNA expression100 and IL-6 protein
associated with obesity83 and the metabolic syndrome84. release from contracting muscle101 are increased when
In support of this idea, IL-6 infusion into mice has been intramuscular glycogen is low, indicating that IL-6 works
shown to impair insulin sensitivity85, and neutralization as an energy sensor and that muscular IL-6 production

Nature Reviews | Endocrinology

Reviews

is regulated not only by muscle contractions but also by in mood, anxiety or cognition113; however, up to 95% of
the energy status of the muscle100,101. the bioavailable tryptophan is metabolized to kynure-
Several studies have demonstrated that global dele- nine, which upon accumulation in the CNS can lead to
tion of Il6 in mice results in accumulation of adipose depression and stress114.
tissue87,88, whereas adeno-associated viral delivery of Defects in kynurenine signalling have been noted
IL-6 into rat hypothalamus102 and central overexpres- in mouse models of Alzheimer and Huntington dis-
sion of IL-6 in mice103,104 are associated with decreased eases115,116. High levels of kynurenine are also found
fat content and suppression of body weight gain. These in people with depression117, and kynurenine might
results indicate that IL-6 has an important role in body be directly involved in the pathogenesis of depression
weight control. IL-6 might cause these effects through by inducing death of neuronal cells and neuroinflam-
regulation of hypothalamic neuropeptides involved in mation 118. Kynurenine aminotransferases convert
energy homeostasis104–106. kynurenine to kynurenic acid119. Kynurenic acid is
The absence of muscular IL-6 in male mice results neuroprotective and, unlike kynurenine, is not able to
in decreased body weight and reductions in food con- cross the blood–brain barrier119. An imbalance between
sumption in response to leptin, suggesting that muscle the neuroprotective kynurenic acid and the neuro-
IL-6 has an impact on the CNS and has a role in mouse toxic kynurenine metabolites has been proposed to be
metabolism not only during exercise but also in the involved in the development of depression41.
basal state and in situations in which energy balance is As mentioned above, exercise leads to upregulation
altered107. When IL-6 is centrally administered by intra­ of PGC1α in skeletal muscle cells64. Activation of PGC1α
cerebroventricular injections in mice, it suppresses feed- leads to an enhancement of processes such as mitochon-
ing and improves glucose tolerance108. Of note, although drial biogenesis, fatty acid oxidation and resistance to
obese mice demonstrate leptin resistance, the ability of muscle atrophy120,121. Transgenic murine models with
IL-6 to inhibit feeding is more pronounced in obese mice specific PGC1α overexpression in the skeletal muscle
than in lean mice108. To determine whether the ability of demonstrate many of the adaptations to aerobic exer-
IL-6 to suppress feeding was dependent on the actions cise training such as mitochondrial biogenesis, oxi-
of IL-6 centrally in the brain, the researchers injected the dative metabolism and the formation of slow-twitch
same dose of IL-6 intraperitoneally and found no effect muscle fibres122.
on food intake, clearly demonstrating that the observed Overexpression of PGC1α in muscle has been shown
effect was mediated via the CNS108. However, when mice to have an antidepressant-like effect by reducing
were injected peripherally with a fourfold higher IL-6 entry of neurotoxic kynurenine into the brain41 (Fig. 1c).
concentration than the dose injected centrally, the high Agudelo et al.41 showed that overexpression of PGC1α
dose of IL-6 was associated with a significant reduction in muscle induces a change in kynurenine metabolism
in food intake during refeeding. One conclusion from leading to protection from stress-induced depression.
this study is that peripherally derived IL-6 at high con- The authors demonstrated that exercise led to activation
centrations can pass through the blood–brain barrier of the PGC1α–PPARα–PPARδ pathway, which stimu-
and suppress feeding in mice. This finding leaves open lates the expression of kynurenine aminotransferase
the possibility that the increases in muscle-derived IL-6 within skeletal muscles. High expression of kynure-
that occur during exercise of high intensity and long nine aminotransferase led to increased conversion
duration might inhibit appetite. of kynurenine into kynurenic acid. Reduction in plasma
Healthy elderly people (aged >70 years) maintain kynurenine level protected the animals from stress-
the capacity to produce and release IL-6 in response induced damage in the brain. Wild-type mice were sen-
to dynamic exercise, with no difference compared sitive to stress-induced depression, whereas transgenic
with younger individuals109. Satellite cells derived from mice with muscle-specific overexpression of PGC1α
humans with insulin resistance demonstrate IL-6 resist- were resistant to stress-induced depression. Exposure
ance with regard to AMPK activation, but it remains to be to chronic, mild stress led to a decrease in hippocampal
shown whether such functional impairments exist with levels of BDNF and increased neuro­inflammation in
other biological actions of IL-6 or in other subpopulations wild-type mice but not in PGC1α transgenic mice41.
such as older individuals110. In general, lifelong physi- However, although the transgenic mice with muscle-
cal activity prevents age-associated insulin resistance specific PGC1α overexpression were protected from
in human skeletal muscle myotubes111. the brain damage, neuroinflammation and depression
induced by chronic stress, they did not show overall
PGC1α–kynurenine axis protection from stress-induced inflammation. In fact,
Worldwide, depression and stress are leading causes of transgenic mice exposed to chronic stress displayed
disability and negatively influence the quality of life weight loss and increased inflammation in skeletal
of millions of people. Physical exercise is sometimes muscle compared with nonstressed transgenic mice41.
prescribed to people suffering from poor mental health112. Exercise training has been shown to increase pla­
Tryptophan is an essential amino acid critical for sma kynurenic acid levels in rodents41, and extensive
protein synthesis, but it also serves as a substrate for the endurance exercise has been shown to increase plasma
generation of important compounds that have various kynurenic acid levels in humans123. Diabetic and obese
physiological roles. The best-known fate of tryptophan mice as well as humans with type 2 diabetes mel-
is its conversion to serotonin (5-hydroxytryptamine), litus demonstrate reduced muscular expression of
an important neurotransmitter linked with alterations PGC1α124. The fact that PGC1α shows links with both

www.nature.com/nrendo
 Reviews

insulin resistance and inflammation, and also with the insulin-like growth factor 1 (IGF1) seem to be involved
kynurenine pathway, might be of clinical importance. in crosstalk between the liver and brain tissues, and
Dysregulation in the PGC1α–kynurenine pathway such crosstalk can be modulated by exercise. FGF21 is a
might explain not only the impaired insulin signalling hepatokine134 capable of inducing insulin sensitivity and
and inflammation in this disease but also the twofold to weight loss135. Exercise induces an increase in FGF21
threefold increased risk of dementia and depression125,126. secretion from the human liver, which contributes to
Increased peripheral FNDC5 levels have been shown an increase in systemic FGF21 levels via a mechanism
to mediate neuroprotective effects by increasing BDNF involving glucagon136.
expression in the brain58; however, BDNF levels were Studies in rodents have shown that prolonged fasting
unchanged in skeletal muscle-specific PGC1α trans- stimulates hepatic production of FGF21 via a mecha-
genic mice, which suggests that the FNDC5–BDNF nism that includes activation of the transcription factor
pathway might not be involved in the defence against PPARα. After production in the liver, FGF21 enters the
stress-induced depression. brain and activates the hypothalamic–pituitary–adrenal
In the absence of a psychological stress challenge, axis for release of corticosterone, thereby stimulating
endurance exercise training is sufficient to activate skel- hepatic gluconeogenesis137.
etal muscle kynurenine to kynurenic acid conversion, Studies in mice, monkeys and humans show that
resulting in considerably elevated circulating kynurenic FGF21 regulates simple sugar intake and preferences
acid levels41. Peripheral kynurenic acid has been shown for sweet foods via signalling through FGF21 receptors
to possess anti-inflammatory properties127. Moreover, in the paraventricular nucleus of the hypothalamus
kynurenic acid increases energy utilization by activat- and correlates with reduced dopamine neurotransmis-
ing G protein-coupled receptor GPR35, which stim- sion within the nucleus accumbens138–140. FGF21 has
ulates lipid metabolism and anti-inflammatory gene also been shown to prevent cognitive decline in obese
expression in adipose tissue127. Kynurenic acid thereby insulin-resistant rats by improving hippocampal syn-
suppresses weight gain in animals fed a high-fat diet, aptic plasticity and brain FGF21 signalling141. Taken
improves glucose tolerance and reduces adipose tissue together, the available data on FGF21 indicate that
inflammation128. exercise causes an increase in circulating FGF21, which
In conclusion, exercised skeletal muscle might pos- crosses the blood–brain barrier and affects metabolic as
itively influence the balance between kynurenine and well as cognitive functions.
kynurenic acid by increasing the conversion of kynure- The main source of IGF1 is believed to be the liver,
nine to kynurenic acid. The fact that kynurenic acid is but IGF1 has also been shown to be released by contract-
not able to cross the blood–brain barrier protects the ing muscles, and circulating levels of IGF1 are upregu-
brain from stress-induced kynurenine accumulation, lated by exercise142,143. The role of IGF1 in muscle–brain
neuroinflammation and changes in synaptic plasticity communication is supported by the finding that the
associated with depression, and kynurenic acid also effect of physical exercise on upregulation of hippocam-
contributes to an overall improvement in metabolism. pal BDNF expression and adult neurogenesis is blocked
by neutralizing IGF1 antibodies144.
Adipocyte–brain link
Adiponectin is a protein secreted by adipocytes and is Exercise and ageing
primarily produced and released into the circulation Physical exercise offers protection against cognitive
from adipose tissue129. Studies in rodents have shown decline in ageing145, but it is not clear whether regular
that adiponectin is also expressed and released from exercise training slows the trajectory of normal ageing
muscle in association with exercise129. by influencing metabolic and vascular risk factors or
Adiponectin has insulin-sensitizing, antidiabetic, whether it repetitively boosts brain function, for exam-
anti-inflammatory and antiatherogenic properties129 as ple, by inducing neurochemical and structural changes
well as neuroprotective effects130,131. Adiponectin can pass in the hippocampus, thereby protecting memory and
through the blood–brain barrier and induce increased learning145.
cell proliferation and decreased depression-like behav- To what extent the ageing muscle can elicit physiolog-
iour. Exercise-induced reduction in depression-like ical stimuli capable of inducing sufficient responses in
behaviour was abrogated in adiponectin-deficient mice the brain is also unclear. Time course studies in healthy
via a mechanism that included impairment of AMPK rodents have shown, however, that the age-related loss
in the hippocampus. These findings suggest that adi- of skeletal muscle mass and function (sarcopenia) is
ponectin is involved in mediating the effect of exercise strongly associated with denervation of myofibres and
on hippocampal neurogenesis and depression130,131. The that exercise can prevent this sarcopenia121,146–148.
discovery of adipokines, such as adiponectin, suggests One-third of the hippocampal neurons are subject to
a direct crosstalk between adipose and brain tissues, exchange149, with 700 new neurons added in each hip-
Sarcopenia which can be modulated by exercise. pocampus per day in adult humans. This figure corre-
Derived from the Greek sarx sponds to an annual turnover of 1.75% of the neurons
(‘flesh’) and penia (‘loss’), Liver–brain link with only a modest decline in this turnover during age-
sarcopenia is the age-induced Similar to muscle and adipose tissue, the liver releases ing. Neurons are generated throughout adulthood in
loss of muscle mass and
function that typically
secretory proteins, known in this case as hepatokines, both mice and humans149, suggesting that exercise can
manifests as reduced gait which are involved in metabolism132,133. Hepatokines positively influence brain hippocampal neurogenesis
speed. such as fibroblast growth factor 21 (FGF21) and throughout our lifespan.

Nature Reviews | Endocrinology

Reviews

Conclusions central mechanisms it suppresses feeding and improves

Our understanding of how exercise is sensed by the glucose tolerance. In addition, exercise enhances the
brain and how hippocampal neurogenesis, cognition, PGC1α-dependent muscular expression of the enzyme
mood and appetite become activated or regulated by kynurenine aminotransferase, which induces a benefi-
muscle has been limited. Findings now suggest the exist- cial shift in the balance between the neurotoxic kynure-
ence of a muscle–brain endocrine loop. Working skele- nine and the neuroprotective kynurenic acid, thereby
tal muscle secretes myokines or expresses muscle factors reducing depression-like symptoms. Direct muscle-
that can alter hippocampal function either directly or to-brain crosstalk is mediated by myokines and meta­
via an effect on BDNF protein levels. Evidence is accu- bolites released by muscle, but exercise is also sensed
mulating that the myokine cathepsin B, when increased by the brain indirectly via adipose tissue and the liver.
peripherally by exercise, can pass through the blood– These organs secrete adipokines and hepatokines, which
brain barrier and enhance BDNF production and hence can pass through the blood–brain barrier. The identifi-
neurogenesis, memory and learning. Moreover, the cation of exercise-related factors that have a direct or
PGC1α-dependent myokine irisin, which is released indirect effect on brain function has the potential to
into the circulation by cleavage of FNDC5, might reach highlight novel therapeutic targets for neurodegenera-
the brain and activate the FNDC5–BDNF pathway. The tive diseases and cognitive enhancers for people of all
ketone body β-hydroxybutyrate might also provide a ages. These factors might also be useful as markers to
link between aerobic exercise and BDNF gene expres- be used for moni­toring the amount, intensity and mode
sion in the brain. The myokine IL-6 is produced in mus- of exercise that is required in order to prescribe exercise
cle and released into the blood with exercise, and plasma as a booster of neurological and mental health.
levels increase exponentially with exercise duration.
IL-6 can pass through the blood–brain barrier, and via Published online xx xx xxxx

1. Nietzsche, F. W. Beyond Good and Evil Aphorism 34 18. Aberg, M. A. et al. Cardiovascular fitness is associated family of chemokines by muscle cells. Mol. Biosyst. 7,
(Cambridge Univ. Press, 2002). with cognition in young adulthood. Proc. Natl Acad. 311–321 (2011).
2. Kierkegaard, S. Kierkegaard’s Writings, XXV, Volume Sci. USA 106, 20906–20911 (2009). 35. Henningsen, J., Rigbolt, K. T., Blagoev, B., Pedersen, B. K.
25: Letters and Documents (ed. Rosenmeier, H.) letter 19. Kelley, G. A. & Kelley, K. S. Exercise and sleep: & Kratchmarova, I. Dynamics of the skeletal muscle
150, 1847 (Princeton Univ. Press, 2009). a systematic review of previous meta-analyses. secretome during myoblast differentiation. Mol. Cell.
3. Gros, F. A Philosophy of Walking (Verso Books, 2014). J. Evid. Based Med. 10, 26–36 (2017). Proteomics 9, 2482–2496 (2010).
4. Noakes, T. & Spedding, M. Olympics: run for your life. 20. Blundell, J. E., Gibbons, C., Caudwell, P., Finlayson, G. 36. Benatti, F. B. & Pedersen, B. K. Exercise as
Nature 487, 295–296 (2012). & Hopkins, M. Appetite control and energy balance: an anti-inflammatory therapy for rheumatic
5. Booth, F. W., Roberts, C. K., Thyfault, J. P., impact of exercise. Obes. Rev. 16 (Suppl. 1), 67–76 diseases-myokine regulation. Nat. Rev. Rheumatol.
Ruegsegger, G. N. & Toedebusch, R. G. Role of (2015). 11, 86–97 (2015).
inactivity in chronic diseases: evolutionary insight 21. Crush, E. A., Frith, E. & Loprinzi, P. D. Experimental 37. Hojman, P., Gehl, J., Christensen, J. F. & Pedersen, B. K.
and pathophysiological mechanisms. Physiol. Rev. 97, effects of acute exercise duration and exercise Molecular mechanisms linking exercise to cancer
1351–1402 (2017). recovery on mood state. J. Affect. Disord. 229, prevention and treatment. Cell Metab. 27, 10–21
6. Cotman, C. W., Berchtold, N. C. & Christie, L. A. 282–287 (2018). (2018).
Exercise builds brain health: key roles of growth factor 22. Ruiz, J. R. et al. Resistance training does not have 38. Pedersen, L. et al. Voluntary running suppresses
cascades and inflammation. Trends Neurosci. 30, an effect on cognition or related serum biomarkers tumor growth through epinephrine- and IL-6-
464–472 (2007). in nonagenarians: a randomized controlled trial. dependent NK cell mobilization and redistribution.
7. Mattson, M. P. Energy intake and exercise as Int. J. Sports Med. 36, 54–60 (2015). Cell Metab. 23, 554–562 (2016).
determinants of brain health and vulnerability to 23. Gordon, B. R. et al. Association of efficacy of resistance 39. Rai, M. & Demontis, F. Systemic nutrient and
injury and disease. Cell Metab. 16, 706–722 (2012). exercise training with depressive symptoms: meta- stress signaling via myokines and myometabolites.
8. Arem, H. et al. Leisure time physical activity and analysis and meta-regression analysis of randomized Annu. Rev. Physiol. 78, 85–107 (2016).
mortality: a detailed pooled analysis of the clinical trials. JAMA Psychiatry 75, 566–576 (2018). 40. Makarova, J. A. et al. Exercise immunology meets
dose-response relationship. JAMA Intern. Med. 175, 24. Batouli, S. A. H. & Saba, V. At least eighty percent MiRNAs. Exerc. Immunol. Rev. 20, 135–164
959–967 (2015). of brain grey matter is modifiable by physical activity: (2014).
9. Santos-Lozano, A. et al. Physical activity and Alzheimer a review study. Behav. Brain Res. 332, 204–217 41. Agudelo, L. Z. et al. Skeletal muscle PGC-1α1
disease: a protective association. Mayo Clin. Proc. 91, (2017). modulates kynurenine metabolism and mediates
999–1020 (2016). 25. Kobilo, T. et al. Running is the neurogenic and resilience to stress-induced depression. Cell 159,
10. Aarsland, D., Sardahaee, F. S., Anderssen, S. & neurotrophic stimulus in environmental enrichment. 33–45 (2014).
Ballard, C. Is physical activity a potential preventive Learn. Mem. 18, 605–609 (2011). 42. Safdar, A., Saleem, A. & Tarnopolsky, M. A.
factor for vascular dementia? A systematic review. 26. Febbraio, M. A. & Pedersen, B. K. Contraction-induced The potential of endurance exercise-derived exosomes
Aging Ment. Health 14, 386–395 (2010). myokine production and release: is skeletal muscle an to treat metabolic diseases. Nat. Rev. Endocrinol. 12,
11. Williams, J. W., Plassman, B. L., Burke, J. & Benjamin, S. endocrine organ? Exerc. Sport Sci. Rev. 33, 114–119 504–517 (2016).
Preventing Alzheimer’s disease and cognitive decline. (2005). 43. Whitham, M. et al. Extracellular vesicles provide a
Evid. Rep. Technol. Assess. (Full Rep.) 193, 1–727 27. Giudice, J. & Taylor, J. M. Muscle as a paracrine and means for tissue crosstalk during exercise. Cell Metab.
(2010). endocrine organ. Curr. Opin. Pharmacol. 34, 49–55 27, 237–251 (2018).
12. Pedersen, B. K. & Saltin, B. Exercise as medicine - (2017). 44. Safdar, A. & Tarnopolsky, M. A. Exosomes as
evidence for prescribing exercise as therapy in 26 28. Karstoft, K. & Pedersen, B. K. Skeletal muscle as a mediators of the systemic adaptations to endurance
different chronic diseases. Scand. J. Med. Sci. Sports gene regulatory endocrine organ. Curr. Opin. Clin. exercise. Cold Spring Harb. Perspect. Med. 8,
25 (Suppl. 3), 1–72 (2015). Nutr. Metab. Care 19, 270–275 (2016). a029827 (2018).
13. Voss, M. W., Nagamatsu, L. S., Liu-Ambrose, T. & 29. Pedersen, B. K. Muscle as a secretory organ. 45. Loprinzi, P. D. & Frith, E. A brief primer on the
Kramer, A. F. Exercise, brain, and cognition across the Compr. Physiol. 3, 1337–1362 (2013). mediational role of BDNF in the exercise-memory
life span. J. Appl. Physiol. 111, 1505–1513 (2011). 30. Pedersen, B. K. & Febbraio, M. A. Muscles, exercise link. Clin. Physiol. Funct. Imaging 39, 9–14 (2019).
14. Cotman, C. W. & Berchtold, N. C. Exercise: a and obesity: skeletal muscle as a secretory organ. 46. Neeper, S. A., Gomez-Pinilla, F., Choi, J. & Cotman, C.
behavioral intervention to enhance brain health and Nat. Rev. Endocrinol. 8, 457–465 (2012). Exercise and brain neurotrophins. Nature 373, 109
plasticity. Trends Neurosci. 25, 295–301 (2002). 31. Hoffmann, C. & Weigert, C. Skeletal muscle as an (1995).
15. Smith, P. J. et al. Aerobic exercise and neurocognitive endocrine organ: the role of myokines in exercise 47. Liu, P. Z. & Nusslock, R. Exercise-mediated neurogenesis
performance: a meta-analytic review of randomized adaptations. Cold Spring Harb. Perspect. Med. 7, in the hippocampus via BDNF. Front. Neurosci. 12, 52
controlled trials. Psychosom. Med. 72, 239–252 a029793 (2017). (2018).
(2010). 32. Schnyder, S. & Handschin, C. Skeletal muscle as an 48. Oliff, H. S., Berchtold, N. C., Isackson, P. &
16. Snowden, M. et al. Effect of exercise on cognitive endocrine organ: PGC-1α, myokines and exercise. Cotman, C. W. Exercise-induced regulation of
performance in community-dwelling older adults: Bone 80, 115–125 (2015). brain-derived neurotrophic factor (BDNF) transcripts
review of intervention trials and recommendations for 33. Raschke, S. & Eckel, J. Adipo-myokines: two sides of in the rat hippocampus. Brain Res. Mol. Brain Res.
public health practice and research. J. Am. Geriatr. Soc. the same coin—mediators of inflammation and 61, 147–153 (1998).
59, 704–716 (2011). mediators of exercise. Mediators. Inflamm. 2013, 49. Van Hoomissen, J. D., Chambliss, H. O., Holmes, P. V.
17. Gaser, C. & Schlaug, G. Brain structures differ 320724 (2013). & Dishman, R. K. Effects of chronic exercise and
between musicians and non-musicians. J. Neurosci. 34. Henningsen, J., Pedersen, B. K. & Kratchmarova, I. imipramine on mRNA for BDNF after olfactory
23, 9240–9245 (2003). Quantitative analysis of the secretion of the MCP bulbectomy in rat. Brain Res. 974, 228–235 (2003).

www.nature.com/nrendo
 Reviews

50. Farmer, J. et al. Effects of voluntary exercise on 76. Sleiman, S. F. et al. Exercise promotes the expression of muscle glycogen content. FASEB J. 15, 2748–2750
synaptic plasticity and gene expression in the dentate of brain derived neurotrophic factor (BDNF) through (2001).
gyrus of adult male Sprague-Dawley rats in vivo. the action of the ketone body beta-hydroxybutyrate. 101. Steensberg, A. et al. Interleukin-6 production in
Neuroscience 124, 71–79 (2004). eLife 5, e15092 (2016). contracting human skeletal muscle is influenced by
51. Adlard, P. A., Perreau, V. M. & Cotman, C. W. 77. Lim, S. et al. D-β-hydroxybutyrate is protective in pre-exercise muscle glycogen content. J. Physiol. 537,
The exercise-induced expression of BDNF within the mouse models of Huntington’s disease. PLOS ONE 6, 633–639 (2001).
hippocampus varies across life-span. Neurobiol. Aging e24620 (2011). 102. Li, G. et al. Induction of uncoupling protein 1 by
26, 511–520 (2005). 78. Kashiwaya, Y. et al. D-β-hydroxybutyrate protects central interleukin-6 gene delivery is dependent on
52. Li, X. et al. Blocking of BDNF-TrkB signaling inhibits neurons in models of Alzheimer’s and Parkinson’s sympathetic innervation of brown adipose tissue and
the promotion effect of neurological function recovery disease. Proc. Natl Acad. Sci. USA 97, 5440–5444 underlies one mechanism of body weight reduction in
after treadmill training in rats with spinal cord injury. (2000). rats. Neuroscience 115, 879–889 (2002).
Spinal Cord 57, 65–74 (2019). 79. Sothmann, M. S., Hart, B. A. & Horn, T. S. Sympathetic 103. Hidalgo, J. et al. Transgenic mice with
53. Vaynman, S., Ying, Z. & Gomez-Pinilla, F. Hippocampal nervous system and behavioral responses to stress astrocyte-targeted production of interleukin-6 are
BDNF mediates the efficacy of exercise on synaptic following exercise training. Physiol. Behav. 51, resistant to high-fat diet-induced increases in body
plasticity and cognition. Eur. J. Neurosci. 20, 1097–1103 (1992). weight and body fat. Brain Behav. Immun. 24,
2580–2590 (2004). 80. Gleeson, L. C., Ryan, K. J., Griffin, E. W., Connor, T. J. 119–126 (2010).
54. Vaynman, S., Ying, Z. & Gomez-Pinilla, F. Exercise & Harkin, A. The β2-adrenoceptor agonist clenbuterol 104. Senaris, R. M. et al. Interleukin-6 regulates the
induces BDNF and synapsin I to specific hippocampal elicits neuroprotective, anti-inflammatory and expression of hypothalamic neuropeptides involved
subfields. J. Neurosci. Res. 76, 356–362 (2004). neurotrophic actions in the kainic acid model of in body weight in a gender-dependent way.
55. Rasmussen, P. et al. Evidence for a release of excitotoxicity. Brain Behav. Immun. 24, 1354–1361 J. Neuroendocrinol. 23, 675–686 (2011).
brain-derived neurotrophic factor from the brain (2010). 105. Schele, E. et al. Inter-relation between interleukin
during exercise. Exp. Physiol. 94, 1062–1069 81. Cooke, J. D., Grover, L. M. & Spangler, P. R. Venlafaxine (IL)-1, IL-6 and body fat regulating circuits of the
(2009). treatment stimulates expression of brain-derived hypothalamic arcuate nucleus. J. Neuroendocrinol.
56. Seifert, T. et al. Endurance training enhances BDNF neurotrophic factor protein in frontal cortex and inhibits 25, 580–589 (2013).
release from the human brain. Am. J. Physiol. Regul. long-term potentiation in hippocampus. Neuroscience 106. Benrick, A. et al. Interleukin-6 gene knockout
Integr. Comp. Physiol. 298, R372–R377 (2010). 162, 1411–1419 (2009). influences energy balance regulating peptides in the
57. Pajonk, F. G. et al. Hippocampal plasticity in response 82. Mohamed-Ali, V. et al. Subcutaneous adipose tissue hypothalamic paraventricular and supraoptic nuclei.
to exercise in schizophrenia. Arch. Gen. Psychiatry 67, releases interleukin-6, but not tumor necrosis J. Neuroendocrinol. 21, 620–628 (2009).
133–143 (2010). factor-alpha, in vivo. J. Clin. Endocrinol. Metab. 82, 107. Molinero, A. et al. Role of muscle IL-6 in
58. Wrann, C. D. et al. Exercise induces hippocampal 4196–4200 (1997). gender-specific metabolism in mice. PLOS ONE 12,
BDNF through a PGC-1α/FNDC5 pathway. Cell Metab. 83. Carey, A. L. et al. Interleukin-6 and tumor necrosis e0173675 (2017).
18, 649–659 (2013). factor-alpha are not increased in patients with type 2 108. Timper, K. et al. IL-6 improves energy and glucose
59. de Assis, G. G. & de Almondes, K. M. Exercise- diabetes: evidence that plasma IL-6 is related to fat homeostasis in obesity via enhanced central IL-6
dependent BDNF as a modulatory factor for the mass and not insulin responsiveness. Diabetologia 47, trans-signaling. Cell Rep. 19, 267–280 (2017).
executive processing of individuals in course of 1029–1037 (2004). 109. Pedersen, M. et al. Does the aging skeletal muscle
cognitive decline. a systematic review. Front. Psychol. 84. Wedell-Neergaard, A. S. et al. Cardiorespiratory maintain its endocrine function? Exerc. Immunol. Rev.
8, 584 (2017). fitness and the metabolic syndrome: roles of 10, 42–55 (2004).
60. Matthews, V. B. et al. Brain-derived neurotrophic factor inflammation and abdominal obesity. PLOS ONE 13, 110. Scheele, C. et al. Satellite cells derived from obese
is produced by skeletal muscle cells in response to e0194991 (2018). humans with type 2 diabetes and differentiated into
contraction and enhances fat oxidation via activation 85. Kim, H. J. et al. Differential effects of interleukin-6 myocytes in vitro exhibit abnormal response to IL-6.
of AMP-activated protein kinase. Diabetologia 52, and -10 on skeletal muscle and liver insulin action PLOS ONE 7, e39657 (2012).
1409–1418 (2009). in vivo. Diabetes 53, 1060–1067 (2004). 111. Bunprajun, T., Henriksen, T. I., Scheele, C., Pedersen, B. K.
61. Moon, H. Y. et al. Running-induced systemic cathepsin 86. Cai, D. et al. Local and systemic insulin resistance & Green, C. J. Lifelong physical activity prevents
B secretion is associated with memory function. resulting from hepatic activation of IKK-β and NF-κB. aging-associated insulin resistance in human skeletal
Cell Metab. 24, 332–340 (2016). Nat. Med. 11, 183–190 (2005). muscle myotubes via increased glucose transporter
62. Martinowich, K., Manji, H. & Lu, B. New insights into 87. Matthews, V. B. et al. Interleukin-6-deficient mice expression. PLOS ONE 8, e66628 (2013).
BDNF function in depression and anxiety. Nat. Neurosci. develop hepatic inflammation and systemic insulin 112. Global Burden of Disease Study 2013 Collaborators.
10, 1089–1093 (2007). resistance. Diabetologia 53, 2431–2441 (2010). Global, regional, and national incidence, prevalence,
63. Suzuki, W. A. How body affects brain. Cell Metab. 24, 88. Wallenius, V. et al. Interleukin-6-deficient mice develop and years lived with disability for 301 acute and chronic
192–193 (2016). mature-onset obesity. Nat. Med. 8, 75–79 (2002). diseases and injuries in 188 countries, 1990-2013:
64. Olesen, J., Kiilerich, K. & Pilegaard, H. PGC-1α- 89. Ellingsgaard, H. et al. Interleukin-6 enhances insulin a systematic analysis for the Global Burden of Disease
mediated adaptations in skeletal muscle. Pflugers secretion by increasing glucagon-like peptide-1 Study 2013. Lancet 386, 743–800 (2015).
Arch. 460, 153–162 (2010). secretion from L cells and alpha cells. Nat. Med. 17, 113. Canli, T. & Lesch, K. P. Long story short: the serotonin
65. Spiegelman, B. M. Transcriptional control of 1481–1489 (2011). transporter in emotion regulation and social cognition.
mitochondrial energy metabolism through the PGC1 90. Mauer, J. et al. Signaling by IL-6 promotes alternative Nat. Neurosci. 10, (1103–1109 (2007).
coactivators. Novartis Found. Symp. 287, 60–63 activation of macrophages to limit endotoxemia and 114. Muller, N. & Schwarz, M. J. The immune-mediated
(2007). obesity-associated resistance to insulin. Nat. Immunol. alteration of serotonin and glutamate: towards an
66. Bostrom, P. et al. A PGC1α-dependent myokine that 15, 423–430 (2014). integrated view of depression. Mol. Psychiatry 12,
drives brown-fat-like development of white fat and 91. Mauer, J., Denson, J. L. & Bruning, J. C. Versatile 988–1000 (2007).
thermogenesis. Nature 481, 463–468 (2012). functions for IL-6 in metabolism and cancer. Trends 115. Beal, M. F. et al. Replication of the neurochemical
67. Ding, Q., Vaynman, S., Akhavan, M., Ying, Z. & Immunol. 36, 92–101 (2015). characteristics of Huntington’s disease by quinolinic
Gomez-Pinilla, F. Insulin-like growth factor I interfaces 92. Carey, A. L. et al. Interleukin-6 increases acid. Nature 321, 168–171 (1986).
with brain-derived neurotrophic factor-mediated insulin-stimulated glucose disposal in humans and 116. Wu, W. et al. Expression of tryptophan
synaptic plasticity to modulate aspects of exercise- glucose uptake and fatty acid oxidation in vitro via 2,3-dioxygenase and production of kynurenine
induced cognitive function. Neuroscience 140, AMP-activated protein kinase. Diabetes 55, pathway metabolites in triple transgenic mice and
823–833 (2006). 2688–2697 (2006). human Alzheimer’s disease brain. PLOS ONE 8,
68. Albrecht, E. et al. Irisin - a myth rather than an 93. van Hall, G. et al. Interleukin-6 stimulates lipolysis e59749 (2013).
exercise-inducible myokine. Sci. Rep. 5, 8889 (2015). and fat oxidation in humans. J. Clin. Endocrinol. 117. Claes, S. et al. The kynurenine pathway in major
69. Pekkala, S. et al. Are skeletal muscle FNDC5 gene Metab. 88, 3005–3010 (2003). depression: haplotype analysis of three related
expression and irisin release regulated by exercise 94. Lang Lehrskov, L. et al. Interleukin-6 delays gastric functional candidate genes. Psychiatry Res. 188,
and related to health? J. Physiol. 591, 5393–5400 emptying in humans with direct effects on glycemic 355–360 (2011).
(2013). control. Cell Metab. 27, 1201–1211 (2018). 118. Myint, A. M. & Kim, Y. K. Network beyond IDO in
70. Eaton, M. et al. Impact of a single bout of high- 95. Starkie, R., Ostrowski, S. R., Jauffred, S., Febbraio, M. psychiatric disorders: revisiting neurodegeneration
intensity interval exercise and short-term interval & Pedersen, B. K. Exercise and IL-6 infusion inhibit hypothesis. Prog. Neuropsychopharmacol. Biol.
training on interleukin-6, FNDC5, and METRNL mRNA endotoxin-induced TNF-alpha production in humans. Psychiatry 48, 304–313 (2014).
expression in human skeletal muscle. J. Sport Health FASEB. J. 17, 884–886 (2003). 119. Schwarcz, R., Bruno, J. P., Muchowski, P. J. & Wu, H. Q.
Sci. 7, 191–196 (2018). 96. Steensberg, A., Fischer, C. P., Keller, C., Moller, K. Kynurenines in the mammalian brain: when physiology
71. Jedrychowski, M. P. et al. Detection and quantitation & Pedersen, B. K. IL-6 enhances plasma IL-1ra, IL-10, meets pathology. Nat. Rev. Neurosci. 13, 465–477
of circulating human irisin by tandem mass and cortisol in humans. Am. J. Physiol. Endocrinol. (2012).
spectrometry. Cell Metab. 22, 734–740 (2015). Metab. 285, E433–E437 (2003). 120. Arany, Z. PGC-1 coactivators and skeletal muscle
72. Wrann, C. D. FNDC5/irisin - their role in the nervous 97. Pedersen, B. K. & Febbraio, M. A. Muscle as an adaptations in health and disease. Curr. Opin. Genet.
system and as a mediator for beneficial effects of endocrine organ: focus on muscle-derived Dev. 18, 426–434 (2008).
exercise on the brain. Brain Plast. 1, 55–61 (2015). interleukin-6. Physiol. Rev. 88, 1379–1406 (2008). 121. White, Z. et al. Voluntary resistance wheel exercise
73. Lourenco, M. V. et al. Exercise-linked FNDC5/irisin 98. Febbraio, M. A. & Pedersen, B. K. Muscle-derived from mid-life prevents sarcopenia and increases
rescues synaptic plasticity and memory defects in interleukin-6: mechanisms for activation and possible markers of mitochondrial function and autophagy in
Alzheimer’s models. Nat. Med. 5, 165–175 (2019). biological roles. FASEB J. 16, 1335–1347 (2002). muscles of old male and female C57BL/6J mice.
74. Koppel, I. & Timmusk, T. Differential regulation of Bdnf 99. Keller, C., Hellsten, Y., Steensberg, A. & Pedersen, B. K. Skelet. Muscle 6, 45 (2016).
expression in cortical neurons by class-selective Differential regulation of IL-6 and TNF-α via calcineurin 122. Lin, J. et al. Transcriptional co-activator PGC-1α drives
histone deacetylase inhibitors. Neuropharmacology in human skeletal muscle cells. Cytokine 36, 141–147 the formation of slow-twitch muscle fibres. Nature
75, 106–115 (2013). (2006). 418, 797–801 (2002).
75. Mitchell, G. A. et al. Medical aspects of ketone body 100. Keller, C. et al. Transcriptional activation of the IL-6 123. Lewis, G. D. et al. Metabolic signatures of exercise in
metabolism. Clin. Invest. Med. 18, 193–216 (1995). gene in human contracting skeletal muscle: influence human plasma. Sci. Transl Med. 2, 33ra37 (2010).

Nature Reviews | Endocrinology

Reviews

124. Eckardt, K., Taube, A. & Eckel, J. Obesity-associated refeeding and overfeeding. Diabetes 63, 4057–4063 145. Duzel, E., van Praag, H. & Sendtner, M. Can physical
insulin resistance in skeletal muscle: role of lipid (2014). exercise in old age improve memory and hippocampal
accumulation and physical inactivity. Rev. Endocr. 135. Gaich, G. et al. The effects of LY2405319, an FGF21 function? Brain 139, 662–673 (2016).
Metab. Disord. 12, 163–172 (2011). analog, in obese human subjects with type 2 diabetes. 146. Krishnan, V. S. et al. A neurogenic perspective of
125. Roy, T. & Lloyd, C. E. Epidemiology of depression and Cell Metab. 18, 333–340 (2013). sarcopenia: time course study of sciatic nerves from
diabetes: a systematic review. J. Affect. Disord. 142, 136. Hansen, J. S. et al. Exercise-induced secretion of aging mice. J. Neuropathol. Exp. Neurol. 75,
S8–S21 (2012). FGF21 and follistatin are blocked by pancreatic clamp 464–478 (2016).
126. Pouwer, F. Should we screen for emotional distress in and impaired in type 2 diabetes. J. Clin. Endocrinol. 147. Krishnan, V. S., Shavlakadze, T., Grounds, M. D.,
type 2 diabetes mellitus? Nat. Rev. Endocrinol. 5, Metab. 101, 2816–2825 (2016). Hodgetts, S. I. & Harvey, A. R. Age-related loss of
665–671 (2009). 137. Liang, Q. et al. FGF21 maintains glucose homeostasis VGLUT1 excitatory, but not VGAT inhibitory,
127. Fallarini, S., Magliulo, L., Paoletti, T., de Lalla, C. by mediating the cross talk between liver and brain immunoreactive terminals on motor neurons in spinal
& Lombardi, G. Expression of functional GPR35 in during prolonged fasting. Diabetes 63, 4064–4075 cords of old sarcopenic male mice. Biogerontology 19,
human iNKT cells. Biochem. Biophys. Res. Commun. (2014). 385–399 (2018).
398, 420–425 (2010). 138. von Holstein-Rathlou, S. et al. FGF21 mediates 148. Snyder-Warwick, A. K., Satoh, A., Santosa, K. B.,
128. Agudelo, L. Z. et al. Kynurenic acid and Gpr35 endocrine control of simple sugar intake and sweet Imai, S. I. & Jablonka-Shariff, A. Hypothalamic Sirt1
regulate adipose tissue energy homeostasis and taste preference by the liver. Cell Metab. 23, protects terminal Schwann cells and neuromuscular
inflammation. Cell Metab. 27, 378–392 (2018). 335–343 (2016). junctions from age-related morphological changes.
129. Dai, Y., Pang, J., Gong, H., Fan, W. & Zhang, T. M. 139. Talukdar, S. et al. FGF21 regulates sweet and alcohol Aging Cell 17, e12776 (2018).
Roles and tissue source of adiponectin involved in preference. Cell Metab. 23, 344–349 (2016). 149. Spalding, K. L. et al. Dynamics of hippocampal
lifestyle modifications. J. Gerontol. A Biol. Sci. Med. 140. Soberg, S. et al. FGF21 is a sugar-induced hormone neurogenesis in adult humans. Cell 153, 1219–1227
Sci. 68, 117–128 (2013). associated with sweet intake and preference in (2013).
130. Yau, S. Y. et al. Physical exercise-induced hippocampal humans. Cell Metab. 25, 1045–1053 (2017).
neurogenesis and antidepressant effects are mediated 141. Sa-Nguanmoo, P. et al. FGF21 improves cognition Acknowledgements
by the adipocyte hormone adiponectin. Proc. Natl by restored synaptic plasticity, dendritic spine density, The Centre for Physical Activity Research (CFAS) is supported
Acad. Sci. USA 111, 15810–15815 (2014). brain mitochondrial function and cell apoptosis in by a grant from TrygFonden. The author also thanks
131. Li, A., Yau, S. Y., Machado, S., Yuan, T. F. & So, K. F. obese-insulin resistant male rats. Horm. Behav. 85, Alzheimer-forskningsfonden for support.
Adult neurogenic and antidepressant effects of 86–95 (2016).
adiponectin: a potential replacement for exercise? 142. Hamrick, M. W. A role for myokines in muscle-bone Competing interests
CNS Neurol. Disord. Drug Targets 14, 1129–1144 interactions. Exerc. Sport Sci. Rev. 39, 43–47 The author declares no competing interests.
(2015). (2011).
132. Cooper, C., Moon, H. Y. & van Praag, H. On the run for 143. Berg, U. & Bang, P. Exercise and circulating insulin-like Publisher’s note
hippocampal plasticity. Cold Spring Harb. Perspect. growth factor I. Horm. Res. 62, (Suppl. 1), 50–58 Springer Nature remains neutral with regard to jurisdictional
Med. 8, a029736 (2018). (2004). claims in published maps and institutional affiliations.
133. Stefan, N. & Haring, H. U. The role of hepatokines 144. Chen, M. J. & Russo-Neustadt, A. A. Running
in metabolism. Nat. Rev. Endocrinol. 9, 144–152 exercise- and antidepressant-induced increases in Reviewer information
(2013). growth and survival-associated signaling molecules Nature Reviews Endocrinology thanks M. Grounds, S. Schiaffino
134. Markan, K. R. et al. Circulating FGF21 is liver are IGF-dependent. Growth Factors 25, 118–131 and other anonymous reviewer(s) for their contribution to the
derived and enhances glucose uptake during (2007). peer review of this work.

www.nature.com/nrendo