Therapeutic potential of exerkines in neurodegenerative and mental disorders: a narrative review

Abstract

Neurodegenerative and mental disorders impose significant global disease burdens and pose serious social and economic challenges. Physical exercise (PE) exerts beneficial effects on brain health, contributing to a reduction in the risk of Alzheimer’s disease (AD), Parkinson’s disease (PD), depression, anxiety, and post-traumatic stress disorder (PTSD). To understand these effects of PE, a variety of molecules released from various tissues in response to PE have been discovered, which are collectively called ‘exerkines’. In particular, the skeletal muscle acts as an endocrine organ, secreting exerkines and is included in the category of myokines that facilitate direct or indirect crosstalk between the muscle and the brain. Although muscles actively interact with organs such as the liver, pancreas, and adipose tissue, the precise mechanisms of muscle–brain communication have yet to be fully elucidated. In the skeletal muscle, the types of exerkines secreted and their effects vary depending on the PE modality. Furthermore, these exerkines can cross the blood-brain barrier (BBB) to exert direct effects or act indirectly via molecular signaling pathways, contributing to the modulation of the brain microenvironment, attenuation of neuroinflammation, and neurodegeneration. Previous studies have indicated that brain-derived neurotrophic factor (BDNF), irisin, cathepsin B (CTSB), interleukin-6 (IL-6), and insulin-like growth factor 1 (IGF-1) are involved in enhancing cognitive performance and improving behavioral outcomes by promoting neurogenesis and synaptic plasticity. This review comprehensively discusses the effects of exerkines on the brain and the physiological responses manifested in neurodegenerative and mental disorders focusing primarily on findings from rodent models. Based on these insights, this review proposes future research directions to translate preclinical findings into therapeutic strategies.

1. Introduction

Neurodegenerative and psychiatric disorders represent some of the most significant global health challenges and, impose a substantial burdens on individuals, healthcare systems, and society (Mitchell et al., 2025). In particular, the increasing incidence of Alzheimer’s disease (AD), Parkinson’s disease (PD), and mental disorders such as depression, anxiety, and post-traumatic stress disorder (PTSD) is closely linked to disability and mortality, especially in the elderly population (Collaborators, 2021, Collaborators, 2024; Mao et al., 2025). Despite the urgency posed by an aging global population, effective treatments remain elusive as existing pharmacological interventions often present significant side effects and limited efficacy (Alzheimer’s Association, 2025; Bloem et al., 2021). Although the development of fundamental therapeutics is crucial, the most viable strategies for aging populations should focus on delaying disease progression and preventing disease onset.

Physical exercise (PE) is considered a potent intervention capable of mitigating many modifiable risk factors associated with neurodegenerative diseases and psychiatric disorders (Lee et al., 2023; Nagamatsu et al., 2014). Consistent with this perspective, accumulating clinical evidence demonstrates that regular PE improves cognitive function, reduces the risk of dementia, and alleviates the symptoms of psychiatric conditions such as depression and anxiety (Cámara-Calmaestra et al., 2022; Goodwin et al., 2008; López-Ortiz et al., 2023; Zhen et al., 2022). These beneficial effects are associated with enhanced neuroplasticity, increased hippocampal neurogenesis, and improved synaptic transmission (Campos et al., 2023; da Rocha et al., 2025; Guo et al., 2022). Despite these benefits, the precise molecular mechanisms through which PE induces structural and functional changes in the central nervous system (CNS) remain unclear.

To understand the physiological benefits of PE for brain health, several signaling molecules released from diverse tissues in response to PE have been identified. These secreted factors are collectively termed “exerkines” (Chow et al., 2022; Safdar et al., 2016). Among these factors, some exerkines are secreted from skeletal muscle in an activity- and context-dependent manner and play crucial roles in muscle–brain crosstalk (Chow et al., 2022; Pedersen, 2019). Their activity increases in a contraction dependent manner during PE (Fukada and Nakamura, 2021; Pedersen, 2023), while also exerting local autocrine and paracrine effects that support muscle metabolic homeostasis (Pedersen, 2019). Secretion profiles vary with PE modality (Severinsen and Pedersen, 2020) because different demands recruit different fiber types. Aerobic exercise (AE) engages oxidative type I fibers and induces metabolic stress, whereas resistance exercise (RE) recruits glycolytic type II fibers under mechanical tension (Bawa et al., 2014; Qaisar et al., 2016). Therefore, PE type influences the secretome, while PE intensity and duration further modulate its composition and overall magnitude (Hughes et al., 2018; Leuchtmann et al., 2021). However, integrated studies that systematically compare secretion patterns across PE types remain limited. Circulating signals can influence CNS function, making them promising biomarkers and targets (Pedersen, 2019). Fibronectin type III domain-containing protein 5 (FNDC5)/irisin, brain-derived neurotrophic factor (BDNF), cathepsin B (CTSB), interleukin-6 (IL-6), and insulin-like growth factor 1 (IGF-1) may affect AD and PD via neuroinflammation, energy metabolism, and hippocampal plasticity (Figure 1) (Lee et al., 2021), and may be related to depression, anxiety, and PTSD through BDNF-dependent neuroplasticity, oxidative stress, and hypothalamic–pituitary–adrenal (HPA) axis regulation (Montgomery and Grant, 2026).

Figure 1.

Schematic representation of physical exercise–induced muscle–brain crosstalk. Physical exercise activates intracellular signaling pathways in skeletal muscle (e.g., AMPK–PGC-1α, Ca²⁺–CaMKII, and MAPK), driving metabolic adaptations and the release of exercise-responsive exerkines and growth factors (e.g., irisin/FNDC5, CTSB, IL-6, and IGF-1, and exercise-associated BDNF). These circulating mediators reach the brain and signal across the blood–brain barrier (BBB) via humoral and neurovascular routes, promoting neurogenesis, synaptogenesis, and long-term potentiation (LTP) while attenuating neuroinflammation. Collectively, these mechanisms may improve cognition and mental health, highlighting the therapeutic potential of exercise in neurodegenerative diseases (AD, PD) and psychiatric disorders. AD, Alzheimer’s disease; PD, Parkinson’s disease; AMPK, AMP-activated protein kinase; PGC-1α, peroxisome proliferator-activated receptor gamma coactivator 1-alpha; CaMKII, Ca²⁺/calmodulin-dependent protein kinase II; MAPK, mitogen-activated protein kinase; BDNF, brain-derived neurotrophic factor; CTSB, cathepsin B; IL-6, interleukin-6; IGF-1, insulin-like growth factor 1; LTP, long-term potentiation; BBB, blood–brain barrier.

Previous reviews have primarily focused on the neuroprotective and therapeutic effects of exerkines within the CNS and the underlying mechanisms (Pedersen, 2019; Rai and Demontis, 2022; Wrann, 2015). In contrast, secretion mechanisms and regulatory functions in the skeletal muscles have received relatively limited attention. For leading candidates such as irisin and CTSB, it remains debated whether muscle contraction reliably drives their release into the circulation (Sugimoto et al., 2022). Moreover, integrated comparative studies that characterize secretion patterns across levels of PE intensity and duration remain scarce. Regarding central delivery, blood-brain barrier (BBB) permeability and transport pathways likely involve both direct and indirect routes, but key uncertainties and controversies remain (Maak et al., 2021; Rai and Demontis, 2022). Finally, research has focused relatively more on neurodegenerative disorders, whereas evidence in psychiatric conditions remains limited (Liu et al., 2024; Mojtabavi et al., 2020).

The primary aim of this review is to provide a comprehensive overview of the biogenesis and secretion mechanisms of exerkines that may mediate muscle–brain crosstalk, and to evaluate their potential therapeutic roles in neurodegenerative and mental disorders. First, we systematically summarized prior evidence on the key exerkines implicated in brain function, focusing on their production pathways, secretion mechanisms, and physiological functions in target organs. To strengthen the molecular rationale for prescribing PE and to inform strategies for optimizing exerkine release, we compared exerkine secretion profiles by PE modality (AE vs. RE) and categorized them accordingly (Table 1). We then examined the current evidence and ongoing debates on whether, and through what direct or indirect routes, exerkines reach the brain, and discussed the possibility that they may help restore BBB–mediated crosstalk disrupted by brain disorders. Finally, drawing on recent studies, particularly those using well-controlled preclinical rodent models with clearly defined pathological phenotypes, we integrated evidence for exerkine-mediated neuroprotective mechanisms across a broad spectrum of diseases, including AD, PD, depression, anxiety, and PTSD (Table 2). This review proposes a conceptual framework to advance the clinical translation of exerkines as therapeutic mediators.

Table 1.

Exercise modality–specific exerkine response patterns.

Exerkine	Primary driver	Aerobic exercise (AE)	Resistance exercise (RE)
Irisin/FNDC5	intensity, sampling timing	► [Muscle] HIIT > MICT (overweight/obese adolescents) (Archundia-Herrera et al., 2017) ► [Blood] Isoenergetic high-intensity AE: ↑circulating irisin (Tsuchiya et al., 2014)	► [Blood] Peak ~1 h post-acute RE (Tsuchiya et al., 2015) ► [Blood vs Muscle] ↑circulating irisin despite unchanged muscle FNDC5 mRNA (Nygaard et al., 2015) ► [Blood] ↑basal irisin after 12-week RE in older adults (Kim et al., 2015)
BDNF	muscle mass, total work	► [Blood] High-intensity AE: ↑circulating BDNF (Cho et al., 2012) ► [Blood] Rapid decline toward baseline in early recovery (Heyman et al., 2012)	► [Blood] Hypertrophy-style (short rest/high volume)> strength-style (Marston et al., 2017) ► [Blood] No increase in plasma BDNF after acute RE (Correia et al., 2010)
CTSB	protocol, timing, training history	► [Blood] High-intensity AE (high %VO₂max): ↑CTSB (Mazo et al., 2022) ► [Blood] Chronic AE: ↑resting CTSB (Gaitan et al., 2021) ► [Blood] 6-week HIIT: no significant CTSB change (Nicolini et al., 2019) ► [Blood] Lower resting CTSB in endurance-trained (De la Rosa et al., 2019)	► [Blood] No early post-RE CTSB change after ~80% 1RM (~80% 1RM) (Johnson et al., 2020) ► [Blood] ↑resting serum CTSB after 12-week RE in obese women (Sung et al., 2017)
IL-6	duration, fuel stress	► [Blood] Marked increase during prolonged endurance AE (Steensberg et al., 2000) ► [Blood & Muscle] Attenuated increase with carbohydrate intake (Nieman et al., 2003) ► [Blood & Muscle] Amplified response under low glycogen (Keller et al., 2001) ► [Muscle] Blunted IL-6 mRNA response after endurance training (Fischer et al., 2004)	► [Blood & Muscle] Acute ↑muscle IL-6 mRNA, with ↑circulating IL-6 mainly under high metabolic stress (Annibalini et al., 2019) ► [Blood] Unchanged basal IL-6 after chronic resistance training (Libardi et al., 2012)
IGF-1	context (intensity/duration), age	► [Blood] Intensity-dependent acute ↑IGF-1 (Schwarz et al., 1996) ► [Blood] ↓circulating IGF-1 after prolonged endurance (Copeland and Verzosa, 2014) ► [Blood] ↑circulating IGF-1 (acute, low-volume sprint interval cycling) (Cui et al., 2015)	► [Blood] No IGF-I change; volume-dependent ↑IGFBP-1 (Nindl et al., 2009) ► [Muscle] ↑IGF-I mRNA ~48 h post high-load RE (eccentric > concentric) (Bamman et al., 2001)
CX3CL1 (fractalkine)	sampling timing, intensity	► [Blood & Muscle] Acute 1-h one-legged endurance exercise ↑CX3CL1 mRNA (Catoire et al., 2014) ► [Blood & Muscle] No change in CX3CL1 after 6-month aerobic training (Verheggen et al., 2016) ► [Muscle] Acute 1-h cycling ↑CX3CL1 mRNA and protein levels; vastus lateralis (Stromberg et al., 2016)	► [Muscle] Acute resistance exercise ↑CX3CL1 mRNA (Della Gatta et al., 2014) ► [Blood] Low-intensity resistance exercise program ↑serum CX3CL1 (Hashida et al., 2021)
MCP-1 (CCL2)	inflammatory, metabolic stress, duration	► [Blood & Muscle] Acute 1-h one-legged endurance exercise ↑MCP-1 mRNA and protein levels (Catoire et al., 2014) ► [Muscle & Blood] Acute 45-min cycling ↑MCP-1 mRNA (Balan et al., 2021) ► [Blood] 2-week moderate-intensity endurance training (MIT) ↑MCP-1 (Middelbeek et al., 2021)	► [Muscle] Acute isokinetic exercise ↑muscle MCP-1 expression (Della Gatta et al., 2014) ► [Blood] 12-week resistance training ↓circulating MCP-1 in older adults (Oliveira et al., 2020)

Table 2.

Summary of preclinical rodent studies on exerkine regulation and therapeutic outcomes in neurodegenerative and mental disorder.

Animal model			Exercise		Effects of exercise on exerkines			Related outcomes	Reference
Phenotype	Sex	Age	Type	Protocol	Blood	Muscle	Brain
5xFAD (AD)	Male	6 wk	Voluntary wheel running (AE)	6 mo 30 ± 1 km/week ~20 ± 1 h/week	—	—	[Hippocampus] ↑ BDNF	↓ Cognitive impairment ↑ Spatial learning & Memory	Belaya et al., 2020
3xTg (AD)	Male	9 mo	Resistance ladder climbing (RE)	4 wk (alternate days) 15 reps/session (2 min rest)	= IL-6	—	[Hippocampus] [F. cortex] ↑ IL-6 mRNA	↓ Neuroinflammation ↓ Cognitive impairment ↓ Aβ plaques & Tau ↑ Learning & Memory	Liu et al., 2020
3xTg (AD)	Female	3 mo	Treadmill running (AE)	9 wk 5 d/wk 15 m/min 30→90 min	—	[Gastrocnemius] = CTSB	[Hippocampus] ↑ IGF-1 = BDNF	↑ Motor performance	Pena et al., 2020
			Weighted ladder climbing (50→100% BW) (RE)	9 wk 3 sessions/wk 16→10 reps/session (1 min rest)	—			↓ Hippocampal Aβ ↑ Motor performance ↑ Grip strength
i.c.v. AβO model (AD)	Male	2.5-3 mo	swimming (AE)	5 wk 5 d/wk 1 h/day	—	[Gastrocnemius] = Fndc5	[Hippocampus] ↑ Irisin/FNDC5 (mRNA·protein) ↑ BDNF	↑ Cognitive & Memory function ↑ Synaptic plasticity	Lourenco et al., 2019
3xTg (AD)	Male	2.5 mo	Forced wheel running (AE)	12 wk 1-3 d/wk 60 min/d 8 m/min	[Serum] ↓ MCP-1	—	[Cortex] = MCP-1	—	Haskins et al., 2016
MPTP mouse (PD)	Male	6-8 wk	Treadmill running (AE)	10 wk 5 d/wk 30 min/session x 2 12 m/min	—	—	[Hippocampus] ↑ FNDC5 ↑ BDNF	↓ Depression-like behavior ↑ Learning & Memory ↑ Motor performance ↑ DA neuron integrity ↑ Synaptic structure	Shan et al., 2025
MPTP mouse (PD)	Male	8 wk	Treadmill running (AE)	10 wk 5 d/wk 10→60 min/d 8→12 m/min	—	—	[Hippocampus] ↑ Irisin/FNDC5	↓ α-synuclein ↓ Neuroinflammation ↑ Motor function ↑ TH+ cells & AHN ↑ Memory	Zhao et al., 2025
MPTP mouse (3-8 wk i.p.) (PD)	Male	7 wk	Treadmill running (AE)	8 wk 5 d/wk 60 min/d 12 m/min	[Serum] ↑ FNDC5	[Quadriceps] ↑ FNDC5	[Whole brain] ↑ FNDC5	↑ Cognitive & Memory function ↑ Motor function ↑ TH+ cells & DA	Tang et al., 2023
MPTP mouse (PD)	Male	2 mo	Treadmill running (AE)	10 wk 5 d/wk 10→60 min/d 8→12 m/min	[Serum] ↑ Irisin	—	[SN] ↑ Irisin	↓ Neuroinflammation ↑ TH+ cells ↑ Motor function	Wang et al., 2025
CUMS rat (depression)	Male	Adult	Treadmill running (AE)	8 wk 45 min 1.02 km/h	—	—	[Hippocampus] ↑ Bdnf mRNA ↑ IGF-1 ↑ FNDC5	↓ Hippocampal apoptosis ↑ Synaptic plasticity ↓ Depression	Kang et al., 2020
			Weight-bearing treadmill (interval) (RE)	8 wk 20 cycles/d 1.02 km/h (30% BW load)			↑ IGF-1 mRNA ↑ IGF-1 ↑ IGF-1R
CUS rat (depression)	Male	2 mo	Continuous treadmill running (AE)	6 wk 5 d/wk 22-42 min 23-27 m/min	—	—	[Hippocampus] ↑ BDNF ↑ FNDC5 (HIIT > Con)	↓ Anxiety ↓Depression	Babaei et al., 2021
			HIIT (RE)	6 wk 5 d/wk 38-42 m/min x (2-6) (2 min rest)
CUS rat (depression)	Male	6-8 wk	Treadmill running (AE)	6 wk 5 d/wk 20 min/d	—	—	[Hippocampus] ↓ IL-6	↓ Depression ↓ Neuroinflammation (IL-1β)	Xiao et al., 2021
SPS rat + OVX factorial (PTSD)	Female	Adult	Forced running wheel (FRW) (AE)	4 wk 5 d/wk 30 min/d 10 m/min	—	—	[Hippocampus] [PFC] ↑ BDNF ↑ IGF-1	↓ Anxiety ↑ Cognitive function	Amiri et al., 2024
SPS rat (PTSD)	Male and female	Adult	Treadmill running (pre-trauma) (AE)	4 wk 5 d/wk 30 min/d 10-15 m/min	[Serum] ↑ BDNF	—	[Hippocampus] ↑ BDNF	↓ Apoptosis ↓ Anxiety	Mirjalili et al., 2022
SPS rat + BMSCs (PTSD)	Male	Adult	Treadmill running (AE)	4 wk 5 d/wk 30 min/d 10 m/min	[Serum] ↑ BDNF ↑ IGF-1	—	[Hippocampus] ↑ IGF-1	↓ Anxiety	Eshaghi-Gorji et al., 2024
[PFC] ↑ IGF-1 ↑ BDNF

Level of protein/gene expression: ↑ increase; ↓ decrease; = no change. CUMS, chronic unpredictable mild stress; CUS, chronic unpredictable stress; SPS, single prolonged stress; OVX, ovariectomized; HIIT, high-intensity interval training; DA, dopamine; TH, tyrosine hydroxylase; F. cortex, frontal cortex; AβO, amyloid-β oligomers; BW, body weight; SN; substantia nigra; BMSCs, bone marrow mesenchymal stem cells.

2. Exerkines: secretion mechanisms and functions

2.1. FNDC5/Irisin

Irisin is a exerkine generated by proteolytic cleavage of FNDC5, a type I membrane protein induced in skeletal muscle through the AMP-activated protein kinase (AMPK)/peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC-1α) pathway (Boström et al., 2012; Shan et al., 2013). FNDC5 is processed by cleavage of its N-terminal extracellular domain (Boström et al., 2012), but the exact cleavage site and responsible protease(s), including possible a disintegrin and metalloproteinase (ADAM) family involvement (e.g., ADAM10), remain unresolved (Nie and Liu, 2017; Witmer et al., 2024). In humans, upstream ATG-dependent translation initiation, glycosylation, and downstream processing have been proposed, but their sequence and regulation are still unclear (Nie and Liu, 2017; Witmer et al., 2024). Irisin signals largely via αV integrins (e.g., αVβ5) (Kim et al., 2018) and can act endocrinologically as cargo within extracellular vesicles (EVs). PE increases the release of small EVs enriched in FNDC5/irisin. Likewise, muscle-specific FNDC5 rescue increases circulating FNDC5/irisin-rich EVs, supporting EV-mediated systemic delivery as a key endocrine route (Chi et al., 2022; Shi et al., 2024; Whitham et al., 2018). Systemic irisin promotes osteogenesis and reduces adiposity (Boström et al., 2012; Colaianni et al., 2015; Kim et al., 2018). It also enhances satellite cell function, mitochondrial biogenesis, and glucose uptake (Lee et al., 2015, Lee et al., 2019; Reza et al., 2017; Vaughan et al., 2014). In the brain, FNDC5/irisin induces hippocampal Bdnf mRNA and adult neurogenesis (Islam et al., 2021; Wrann et al., 2013), attenuates microglial inflammation (Wang et al., 2022), and reduces amyloid-β (Aβ) burden through astrocytic neprilysin release (Kim et al., 2023).

2.2. Brain-derived neurotrophic factor

BDNF is a member of the neurotrophin family (NGF), produced mainly by neurons and is also expressed in skeletal muscles (Barde, 2025). The BDNF locus uses multiple promoters to generate diverse 5′ untranslated region (5′-UTR) transcripts that encode the same protein precursor, preproBDNF (proBDNF), which can be stored/secreted as proBDNF or proteolytically processed into mature BDNF (Aid et al., 2007; Greenberg et al., 2009; Pruunsild et al., 2007). Mature BDNF binds to tropomyosin receptor kinase B (TrkB) and activates downstream mitogen-activated protein kinase (MAPK) and phosphoinositide 3-kinase (PI3K)/protein kinase B (Akt) signaling, thereby promoting neuronal survival, synaptic plasticity, and long-term potentiation (LTP), while also contributing to cognitive functions including learning, memory and mood regulation, in stress- and depression-related contexts (Barde, 2025; Duman and Monteggia, 2006; Lu et al., 2014). Impaired BDNF/TrkB signaling and dysregulated BDNF intracellular transport have been implicated in depression and AD, and the pharmacological enhancement of TrkB-dependent plasticity has emerged as a therapeutic strategy (Casarotto et al., 2022; Lu et al., 2014). In skeletal muscles, contraction induces BDNF expression, which mainly acts in an autocrine/paracrine manner to coordinate metabolic adaptation and muscle repair (Matthews et al., 2009). BDNF promotes AMPK-dependent lipid oxidation, mitochondrial quality control (remodeling/mitophagy) under lipid stress, and muscle regeneration via satellite cell differentiation (Ahuja et al., 2022; Clow and Jasmin, 2010; Matthews et al., 2009).

2.3. Cathepsin B

CTSB is a lysosomal cysteine protease widely expressed across tissues (Turk et al., 2012). PE and metabolic stress have suggested CTSB as a candidate exerkine, with skeletal muscle proposed as a source contributing to increased circulating CTSB levels (Moon et al., 2016). AMPK/PGC-1α activation induces Ctsb transcription (Moon et al., 2016). Newly synthesized CTSB undergoes endoplasmic reticulum (ER) entry/processing, Golgi glycosylation, and proteolytic maturation in acidic endolysosomal compartments (Xie et al., 2023). In muscle, CTSB has been shown to act in an autocrine manner to support myogenesis by promoting myoblast differentiation and fusion; during fusion, mature CTSB translocates to the cell surface and is released in an active form (Jane et al., 2006). Consistently, CTSB depletion or antisense-mediated knockdown in C2C12 cells impairs their survival and fusion, resulting in fewer and smaller multinucleated myotubes (Gogos et al., 1996). In neurons and glia, lysosomal CTSB supports proteostasis through lysosomal degradation pathways and has been associated with increased doublecortin (DCX) and BDNF (often assessed in the hippocampus) as well as improved cognitive performance (Moon et al., 2016; Ni et al., 2022). Conversely, under pathological conditions, lysosomal membrane permeabilization can lead to cytosolic release of CTSB, triggering inflammatory signaling and promoting cell death (Saudenova et al., 2022; Xie et al., 2023). In microglia, CTSB contributes to interleukin-1 beta (IL-1β) release and can amplify neuroinflammation (Bai et al., 2018; Nakanishi, 2020). PE-induced CTSB may also enter the circulation and influence remote tissues; however, its functions differ across the lysosomal, cytosolic, and extracellular pools, necessitating context-specific interpretation (Brix, 2018).

2.4. Interleukin-6

IL-6 is a representative myokine and the exerkine (Febbraio and Pedersen, 2005). Electrical pulse stimulation (EPS)-induced contractile activity activates MAPK signaling to promote IL-6 transcription (Whitham et al., 2012), after which IL-6 is processed and secreted via the conventional ER–Golgi apparatus secretory pathway (Verboogen et al., 2019). IL-6 acts via IL-6 receptor alpha (IL-6Rα) and the shared co-receptor gp130, producing autocrine/paracrine responses that depend on the magnitude and duration of exposure (Rose-John, 2012). Classic signaling requires membrane-bound IL-6Rα, whereas trans-signaling uses soluble IL-6R (sIL-6R) to activate gp130 on IL-6Rα negative cells (Rose-John, 2012). Acutely, IL-6 activates AMPK to enhance glucose and lipid metabolism, and may support hypertrophy by promoting satellite cell proliferation (Guo et al., 2024; Johnson et al., 2023). Systemically, PE-induced IL-6 mediates crosstalk with the liver, adipose tissue, and pancreas to maintain substrate availability and glycemic control (Ellingsgaard et al., 2020; Lin et al., 2023). In chronic conditions such as cancer cachexia, IL-6 has been implicated in muscle wasting (Bonetto et al., 2012). EV-displayed IL-6R may further modulate the target cell responsiveness, suggesting an additional endocrine-like route (Arnold et al., 2020).

In the CNS, IL-6 can exert neurotrophic and neuromodulatory actions that support synaptic plasticity and neurogenesis (Erta et al., 2012; Gruol, 2015). However, chronic IL-6 exposure may downregulate endothelial tight junction proteins (claudin-5 and occludin) and impair cognition (Takata et al., 2021). Thus, IL-6 effects are context-dependent and, differ between transient PE pulses and chronic inflammation (Pedersen and Febbraio, 2008).

2.5. Insulin-like growth factor 1

IGF-1 is a systemic endocrine effector of the growth hormone (GH)/IGF axis and a load-responsive exerkine expressed locally in skeletal muscles (Pedersen and Febbraio, 2012). IGF-1 is precisely regulated by the differential (often temporally patterned) induction of distinct isoforms and post-translational mechanisms (Philippou et al., 2014; Zanou and Gailly, 2013). Functionally, IGF-1 drives growth by enhancing protein synthesis via mechanistic target of rapamycin complex 1 (mTORC1) pathway (Schiaffino and Mammucari, 2011) and inhibiting FoxO-dependent ubiquitin–proteasome (UPS) catabolic programs (Sandri et al., 2004; Schiaffino and Mammucari, 2011). Collectively, these effects preserve regenerative capacity in atrophy and aging models, and induce hypertrophy in human myotubes (Musaro et al., 2001; Yoshida et al., 2010). Circulating IGF-1 is predominantly produced by the liver via GH-dependent, signal transducer and activator of transcription 5B (STAT5B)-mediated transcription and circulates largely in an insulin-like growth factor-binding protein 3 (IGFBP-3)/acid-labile subunit (ALS) complex, with target tissue bioavailability governed by IGFBP binding and proteolysis (Baxter, 2024; Rotwein, 2012). A recent study proposed that systemic IGF-1 can be packaged into EVs by choroid plexus epithelial cells, released into the cerebrospinal fluid, and taken up by hippocampal neurons in the immature brain (Ortenlof et al., 2024). In the CNS, IGF-1 activates the PI3K/Akt and MAPK pathways to support neuronal survival, synaptic plasticity, and glia-mediated metabolic adaptation (Fernandez and Torres-Aleman, 2012). PE-induced brain uptake of circulating IGF-1 supports hippocampal neurogenesis and neuroprotection, including reduced Aβ burden and protection of dopaminergic neurons (Carro et al., 2002; Trejo et al., 2001; Wang et al., 2020).

3. Exercise type-specific responses and regulation of exerkines

Exerkine responses to AE and RE depend on the intensity, duration, muscle recruitment, fatigue, and sampling timing immediately after PE compared with the early recovery phase, and circulating signals may diverge from intramuscular changes (Table 1) (Bettariga et al., 2024; Nygaard et al., 2015). Thus, identical protocols can yield different directions and magnitudes across serum versus plasma, and across sampling time points (Bettariga et al., 2024; Nygaard et al., 2015).

3.1. Aerobic exercise

In AE, an intensity-driven cluster of acute responders is often observed first, particularly in terms of irisin and circulating BDNF levels (Archundia-Herrera et al., 2017; Lee and Ko, 2025; Tsuchiya et al., 2014). High-intensity interval training (HIIT) increases skeletal muscle irisin protein levels more than moderate-intensity continuous exercise (Archundia-Herrera et al., 2017), and high-intensity AE robustly elevates circulating irisin (Lee and Ko, 2025; Tsuchiya et al., 2014). In patients with PD, 12 weeks of regular AE increased serum irisin levels, and the magnitude of this increase was associated with improvements in balance function (Berg Balance Scale, BBS) (Zhang et al., 2023). Similarly, circulating BDNF levels exhibit an acute spike immediately after high-intensity or high-volume AE, followed by a rapid decline during early recovery (Cho et al., 2012; Heyman et al., 2012). Meta-analytic evidence indicates that exercise interventions tend to increase plasma BDNF levels in neurodegenerative disorders such as AD and PD (Ruiz-Gonzalez et al., 2021). In major depressive disorder (MDD), exercise training related increases in BDNF have been reported alongside symptom reduction, and changes in BDNF and IL-6 were associated with greater clinical improvement (da Cunha et al., 2023). In contrast, IL-6 levels during AE shows a duration dependent pattern, rising progressively during prolonged exercise with substantial energetic demand, typically after ~60–120 min (Keller et al., 2001; Steensberg et al., 2000). Accordingly, acute, muscle-derived IL-6 levels should be distinguished from chronically elevated basal IL-6 levels (Keller et al., 2001; Steensberg et al., 2000). In PTSD, both trauma-sensitive yoga and cognitive processing therapy reduced PTSD symptoms; however, IL-6 and C-reactive protein (CRP) differed across the arms, underscoring the sensitivity to intervention type and sampling timing (Zaccari et al., 2023). IGF-1 is also context-dependent; brief high-intensity bouts can transiently increase IGF-1 levels, whereas prolonged endurance exercise is often associated with decreases (Copeland and Verzosa, 2014; Cui et al., 2015). Overall, during AE, irisin and BDNF displayed intensity-sensitive, immediate post-exercise peaks, whereas IL-6 was driven primarily by exercise duration and metabolic load.

3.2. Resistance exercise

In RE, the primary determinants of exerkine responses are the amount of recruited muscle mass, total work performed (volume), and depth of fatigue (Kraemer and Ratamess, 2005; Marano et al., 2025; Marston et al., 2017). Accordingly, irisin and BDNF often present as transient, pulse-like increases (Marston et al., 2017; Tsuchiya et al., 2015). Irisin has been reported to peak within ~1 hour post-exercise (Tsuchiya et al., 2015), and resting levels may increase after ≥12 weeks of RE (Kim et al., 2015; Zhao et al., 2017). Likewise, BDNF levels tend to increase more prominently after hypertrophy- oriented, high-volume sessions that recruit large muscle groups (Correia et al., 2010; Marston et al., 2017). A similar acute pulse pattern has been observed in anxiety disorders. In panic disorder, a clinical study reported that a 30 min AE bout acutely increased reduced baseline circulating BDNF levels (Ströhle et al., 2010). In contrast, CTSB and IL-6 levels may show limited or inconsistent acute changes in circulating levels after a single RE session (Annibalini et al., 2019; Johnson et al., 2020). Nevertheless, repeated RE may be associated with a higher resting CTSB (Sung et al., 2017), and upregulation of IL-6 mRNA in skeletal muscle has also been reported (Annibalini et al., 2019). IGF-1 adaptations are often more evident locally (in the muscle) than systemically (in the blood), with volume-sensitive changes in IGFBP-1 and delayed, recovery-phase increases in intramuscular IGF-1/mechano growth factor (MGF) mRNA transcripts (Bamman et al., 2001; Nindl et al., 2009). Thus, reliance on circulating IGF-1 alone may underestimate RE-induced adaptations (Bamman et al., 2001; Nindl et al., 2009). Overall, RE responses are largely governed by volume and fatigue, and even when circulating changes are modest, intramuscular transcript-level adaptations may emerge more clearly, with a delay during the recovery period.

4. Exerkines: muscle–brain crosstalk and blood–brain barrier

4.1. Exerkine transport to the brain: mechanisms and controversies

Whether exerkines traverse the BBB is central to interpreting the muscle–brain axis. Mechanisms include indirect BBB mediated signaling without translocation and direct transport of selected exerkines into the CNS (Banks et al., 1994; Jones and Shusta, 2007; Kostka et al., 2024; Licinio and Wong, 1997). Even for exerkines proposed to enter the CNS directly, studies disagree on whether BBB transport occurs and, if so, on its magnitude, as well as whether the CNS-accessible form is an intact protein or a fragment (Banks et al., 1994; Pan et al., 1998; Pardridge, 2007). In contrast, IL-6 and IGF-1 have strong radiotracer evidence supporting saturable influx into the brain (Banks et al., 1994; Pan and Kastin, 2000; Threlkeld et al., 2010). For IL-6, the limited recovery of intact ligands in the brain or cerebrospinal fluid (CSF) suggests that the major effects may reflect endothelial IL-6R/STAT3/suppressor of cytokine signaling 3 (SOCS3) signaling or BBB modulation, rather than direct parenchymal action (Blecharz-Lang et al., 2018; Eskilsson et al., 2014). IGF-1 influx is supported by iodine-125-labeled IGF-1 (¹²⁵I–IGF-1), yet IGF-binding proteins (IGFBPs) can alter tracer bioavailability and confound influx estimates, necessitating models that incorporate serum protein transport interactions (Nieto-Estévez et al., 2016; Pan and Kastin, 2000).

FNDC5/irisin and BDNF are often discussed as direct-route candidates; however, their mechanisms remain controversial (Arosio et al., 2021; Kostka et al., 2024; Zhao, 2022). Peripheral irisin elevation following hepatic FNDC5 delivery or repeated recombinant administration is associated with hippocampal BDNF induction, supporting the possibility of brain access (Kim and Leem, 2019; Wrann et al., 2013). However, a specific receptor in the adult brain has not yet been established, and indirect anti-inflammatory, metabolic, or BBB-stabilizing effects may explain its neuroprotective effects (Islam et al., 2021; Kim and Leem, 2019). For BDNF, radiotracer studies report brain entry after peripheral administration, whereas other studies suggest minimal transport or limited passage of intact BDNF, implying dependence on the disease state, BBB integrity, and assay approach (Di Lazzaro et al., 2007; Pan et al., 1998; Wu and Pardridge, 1999). CTSB increases in the brain after intravenous injection, supporting peripheral entry (Moon et al., 2016). However, PE also elevates hippocampal Ctsb mRNA levels, indicating substantial local regulation (Moon et al., 2016). Overall, the BBB interactions were best described as graded exerkine-specific uncertainties. Interpreting these functions requires the separation of true CNS influx, molecular integrity, and indirect endothelial contributions (Banks et al., 1994; Jones and Shusta, 2007; Kostka et al., 2024).

4.2. Exerkine modulation of blood–brain barrier function in brain disorders

AD, PD, PTSD, and depressive/anxiety disorders are frequently associated with systemic inflammation and, in some patients, metabolic dysfunction (including insulin resistance), which may activate the cerebral endothelium and compromise BBB integrity (Alkhalifa et al., 2023; Lau et al., 2024; Mehdi et al., 2023; Ritson et al., 2024; Sah and Singewald, 2025). Pro-inflammatory cytokines such as tumor necrosis factor-alpha (TNF-α) and IL-1β can increase oxidative stress, activate matrix metalloproteinases (MMPs), destabilize tight junctions, and increase BBB permeability, thereby promoting immune cell infiltration and microglia-driven synaptic dysfunction (Lau et al., 2024; Sweeney et al., 2018). In parallel, insulin resistance reduces insulin delivery to the brain, thereby impairing cerebrovascular function, metabolic homeostasis, and immune regulation (Heni et al., 2014; Rhea et al., 2018; Wang et al., 2022).

BDNF/TrkB signaling attenuates IL-1β- and TNF-α-induced endothelial hyperpermeability and prevents the cytokine-induced reduction in VE-cadherin protein levels (Matsuda et al., 2015). Endothelial TrkB activation can also stabilize vascular endothelial protein tyrosine phosphatase (VE-PTP)/VE-cadherin coupling and limit VE-cadherin cleavage, thereby maintaining junctional organization (Jiang et al., 2015). Consistent with this, VE-cadherin at adherens junctions helps preserve tight-junction integrity by promoting claudin-5 expression and tight-junction organization at the endothelial interface (Taddei et al., 2008). Beyond endothelial junctions, BDNF may reinforce the pericyte–endothelial axis by activating TrkB in mural cells (Anastasia et al., 2014). In addition, BDNF-mediated anti-inflammatory effects on microglia may indirectly favor barrier protection/recovery by reducing inflammatory stress (Charlton et al., 2023).

IGF-1 supports BBB integrity in a context-dependent manner via endothelial IGF-1R signaling (Gulej et al., 2024; Higashi et al., 2020). In ischemic models, IGF-I modulates BBB endothelial function and PI3K signaling (Bake et al., 2019, Bake et al., 2016). IGF-1 also promotes Akt-dependent endothelial nitric oxide synthase (eNOS) phosphorylation and is linked to reduced inflammation and oxidative stress (Higashi et al., 2010; Michell et al., 1999; Sukhanov et al., 2007). However, in neonatal lipopolysaccharide (LPS)-exposed rats, IGF-1 can exacerbate BBB leakage and intracerebral hemorrhage, highlighting age- and inflammatory context-dependent effects (Pang et al., 2010).

Meanwhile, exerkines such as IL-6 and CTSB display context-dependent actions on BBB integrity (Menard et al., 2017; Pedersen and Febbraio, 2008; Rose-John, 2012). IL-6 trans-signaling is pro-inflammatory and can activate endothelial cells (Rose-John, 2012), claudin-5 loss permits peripheral IL-6 brain entry and depression-like behavior (Menard et al., 2017). Conversely, the transient, PE-induced surge of muscle-derived IL-6 stimulates anti-inflammatory mediators and may indirectly support BBB stability by dampening systemic inflammation (Pedersen and Febbraio, 2008; Petersen and Pedersen, 2005). Pathological CTSB activity (e.g., after ischemic injury) may support a possible CTSB/calpain-associated contribution to neurovascular injury with MMP-9 upregulation after ischemia (Tsubokawa et al., 2006).

5. Therapeutic potential of exerkines in neurodegenerative and mental disorders

5.1. Alzheimer’s disease

AD is characterized by an imbalance in Aβ production, aggregation, and clearance, together with tau hyperphosphorylation and neurofibrillary tangle accumulation (Hampel et al., 2021; Long and Holtzman, 2019). These pathological axes reinforce one another through neuroinflammation and synaptic dysfunction (Hampel et al., 2021; Long and Holtzman, 2019). Accumulating evidence indicates that exerkines modulate glial programs involved in Aβ clearance and synaptic resilience (Kim et al., 2023; Lourenco et al., 2019). Through astrocytes and microglia, exerkines regulate Aβ-degrading enzymes, phagocytic clearance, and broader cytokine/chemokine signaling (Fakhoury, 2018; Kim et al., 2023). Irisin is reduced in the AD hippocampus and CSF, and restoring FNDC5/irisin improves synaptic plasticity and memory in AD models (Lourenco et al., 2019). Mechanistically, irisin binds astrocytic integrin αVβ5, downregulates extracellular signal-regulated kinase (ERK)/STAT3 signaling, and increases extracellular neprilysin release, thereby enhancing Aβ clearance (Kim et al., 2023). In humans, CSF irisin correlates positively with Aβ42, BDNF, and mini-mental state examination (MMSE) scores. This finding may be related to a muscle–brain endocrine link (Lourenco et al., 2020).

BDNF is a key neurotrophin implicated in synaptic function/plasticity in Alzheimer’s models and in PE-related neuroprotection (Belaya et al., 2020; Lourenco et al., 2019; Pena et al., 2020). In male 5xFAD mice, voluntary wheel running improves learning and memory and restores astrocyte-associated hippocampal BDNF signaling, suggesting trophic reprogramming may improve network output (Belaya et al., 2020). Recent research emphasizes neurotrophin receptor signaling, and the p75NTR modulator LM11A-31 showed safety with exploratory biomarker/imaging signals in mild-to-moderate AD (Shanks et al., 2024). CTSB appears context- and stage-dependent. It may worsen Aβ proteotoxicity under sustained burden, yet it may also promote microglial Aβ phagocytosis via PI3K/Akt signaling (Hook et al., 2020; Jiang et al., 2025; Siddiqui et al., 2024). Thus, CTSB activity, localization, and disease stage remain key interpretive variables. In female 3xTg-AD mice, treadmill exercise increased hippocampal IGF-1, whereas hippocampal BDNF and skeletal muscle mature cathepsin B (CTSB/Ctsb) were not significantly altered (Pena et al., 2020). RE reduced hippocampal Aβ without significant changes in hippocampal BDNF or skeletal muscle Ctsb, suggesting Aβ modulation can occur independent of CTSB/BDNF upregulation and may be context dependent (Pena et al., 2020). IL-6 effects depend on signaling mode and compartment, so transient PE-induced IL-6 responses may differ from chronic IL-6 activation in AD (Pedersen and Febbraio, 2008; Rose-John, 2012). In male 3xTg-AD mice, resistance ladder training increased hippocampal and frontal cortical IL-6 mRNA without changing serum IL-6, alongside reduced neuroinflammation, improved cognition, and lower Aβ and tau pathology (Liu et al., 2020). By contrast, persistent IL-6 pathway activation in AD-related contexts may be detrimental, as IL-6 deficiency suppresses STAT3-associated cGAS-STING signaling and reduces neuroinflammation and Aβ pathology (Liu et al., 2024). Collectively, these findings suggest transient PE-like activation may fundamentally diverge from chronic pathway engagement.

In this context, IL-6 intersects chemokine networks that govern glial state transitions and immune-cell recruitment. In vivo IL-1β challenge amplifies astrocyte-associated C-C motif chemokine ligand 2 (Ccl2) and IL-6 responses, whereas C-X3-C motif chemokine receptor 1 (Cx3cr1) deficiency in 5xFAD microglia drives a Ccl2-high degenerative program (Lopez-Rodriguez et al., 2021; Puntambekar et al., 2022). In humans, rising circulating monocyte chemoattractant protein-1 (MCP-1; also known as CCL2) tracks poorer memory and, in genotype-stratified analyses, higher AD risk and neuropathology (Bettcher et al., 2019; Huang et al., 2025). High endurance-training status is associated with lower age-related plasma MCP-1, and PE normalizes elevated serum and brain MCP-1 in 3xTg-AD mice (Balan et al., 2021; Haskins et al., 2016). Acute endurance PE increases C-X3-C motif chemokine ligand 1 (CX3CL1), and peripheral CX3CL1 enhances hippocampal BDNF and recognition memory in aged mice (Catoire et al., 2014; Takei et al., 2022). Microglial IGF-1 signaling can enhance astrocytic phagocytosis, whereas long R3 IGF-1 remodels plaques without cognitive rescue in 5xFAD mice (Engel et al., 2025; Zhang et al., 2026). Collectively, these findings support stage- and mechanism-tailored combination strategies that integrate amyloid clearance, inflammatory control, and synaptic resilience.

5.2. Parkinson’s disease

PD is a progressive neurodegenerative disorder marked by degeneration of nigrostriatal dopaminergic neurons and Lewy pathology enriched in misfolded α-synuclein (α-syn) (Leak et al., 2024), producing motor impairment and prominent cognitive and affective symptoms (Alfaidi et al., 2024; Hoglinger et al., 2024). Recent studies on PD and α-syn increasingly converge on interconnected hubs involving transcription factor EB (TFEB)-regulated autophagy–lysosomal function, mitochondrial homeostasis, and glia-driven neuroimmune crosstalk (Giamogante et al., 2024; Ma et al., 2025). These networks also encompass oligodendrocytic prosaposin (PSAP)/G protein-coupled receptor 37 (GPR37)/IL-6 signaling and impaired neurotrophin receptor signaling (Giamogante et al., 2024; Ma et al., 2025). Mechanistically, α-syn can bind the TrkB kinase domain and inhibit BDNF/TrkB signaling by disrupting TrkB trafficking and ubiquitination (Kang et al., 2017). In addition, pS129/α-syn promotes membrane retention of TrkB, collapsing the BDNF/ERK/CREB/mTOR loop and further decreasing BDNF expression (Ma et al., 2025). Within this framework, α-syn overexpression reduces mitochondria-lysosome apposition and perturbs local Ca²⁺ transfer, alters Ca²⁺-dependent TFEB nuclear translocation and lysosomal homeostatic transcription (Giamogante et al., 2024). Lysosomal proteases, particularly CTSB, sit at this convergence point. CTSB inhibition impairs autophagy, increases lysosomal cargo accumulation, reduces glucocerebrosidase (GCase/GBA1) activity, and diminishes clearance of pre-formed α-syn fibrils (Jones-Tabah et al., 2024). Conversely, CTSB activation promotes fibril clearance in cellular models and human iPSC-derived dopaminergic lineages, although some cleavage products may show increased aggregation propensity under specific conditions (Jones-Tabah et al., 2024). Thus, CTSB biology appears beneficial but context dependent.

Against this backdrop, exerkines can function as hub modulators, and FNDC5/irisin is emerging as a multi-node regulator extending beyond mitochondrial protection toward immune-autophagy reprogramming (Cai et al., 2026; Zhang et al., 2023). Irisin enhances autophagic flux and restores TFEB-driven lysosomal function by regulating receptor for advanced glycation end products (RAGE) ubiquitination, while promoting NOD-like receptor family pyrin domain containing 3 (NLRP3) inflammasome degradation to suppress microglial inflammation (Cai et al., 2026). Irisin also attenuates oxidative stress and improves mitochondrial complex I activity through integrin-dependent Akt and ERK1/2 signaling, supporting mitochondrial biogenesis and normalized dynamics (Zhang et al., 2023). These mechanisms align with findings from PE studies in the MPTP model of Parkinson’s disease, a neurotoxin model that damages nigrostriatal dopaminergic neurons. Treadmill training increased irisin/FNDC5 in serum and brain regions, reduced α-syn and neuroinflammation, and improved motor outcomes, tyrosine hydroxylase (TH)-positive neuron preservation, neurogenesis, memory, and dopamine-related indices (Tang et al., 2023; Wang et al., 2025; Zhao et al., 2025). In MPTP models, treadmill exercise consistently enhances FNDC5/irisin signaling and downstream neuroprotection—improving BDNF-linked synaptic/cognitive function (Tang et al., 2023), suppressing microglia-driven inflammation via AMPK/silent mating type information regulation 2 homolog 1 (SIRT1) while preserving TH-positive neurons and motor function (Wang et al., 2025), and reducing α-syn/NLRP3 pathology while restoring neurogenesis and memory (Zhao et al., 2025). Neurotrophic signaling is tightly interwoven with this response: aerobic treadmill exercise increased hippocampal FNDC5 together with BDNF and improved depression-like behavior and synaptic structure (Shan et al., 2025).

In PD, neuroimmune convergence includes oligodendroglial PSAP/GPR37 signaling driving neuroinflammation, while IL-6 effects remain source-, sex-, and context-dependent (Chen et al., 2025; Ma et al., 2025). IL-6, MCP-1, and CX3CL1 support a central-peripheral inflammatory axis (Qu et al., 2023). MCP-1 recruits monocytes, amplifies neuroinflammation (Singh et al., 2021), and predicts memory decline, whereas CX3CL1/CX3CR1 counter-regulates microglia and preserves dopaminergic neurons, although context remains dependent (Pabon et al., 2011; Szymura et al., 2020). CX3CL1 is PE-responsive (Della Gatta et al., 2014; Swalsingh et al., 2025). Finally, IGF-1 associates with PD risk, while IGF1R transcytosis supports α-syn immunotherapy delivery (An et al., 2025; Gao et al., 2025; Shin et al., 2022).

5.3. Mental disorders

Depression, anxiety disorders, and PTSD converge on dysfunction within prefrontal–limbic circuitry, particularly the prefrontal cortex–hippocampus–amygdala network, supporting shared circuit-level mechanisms across diagnostic boundaries (Kredlow et al., 2022; Maren and Holmes, 2016; Tozzi et al., 2024). Within this network, maladaptive stress responsivity, including HPA-axis dysregulation, interacts with neuroimmune disruption to undermine synaptic plasticity and compromise emotional learning and updating (Feng et al., 2025; Lawrence and Scofield, 2024; Maren and Holmes, 2016).

Evidence increasingly supports inflammation-linked subtypes across mood and stress-related disorders, motivating biomarker-guided stratification (Miller, 2025; Sah and Singewald, 2025; Zeng et al., 2024), and meta-analytic evidence suggests that immunomodulatory interventions yield larger benefits in high-inflammation subgroups (Giollabhui et al., 2026). Within this subtype-oriented framework, IL-6 and chemokine signaling appear central. Meta-analyses consistently report elevated IL-6 in major depressive disorder, and IL-6 is repeatedly implicated in PTSD-associated inflammatory signatures (Dowlati et al., 2010; Haapakoski et al., 2015; Passos et al., 2015). CCL2 has also emerged as a candidate chemokine altered in depression (Leighton et al., 2018). Mechanistically, IL-6 trans-signaling can induce endothelial CCL2 through JAK/STAT3 and PI3K/Akt pathways, providing a plausible route from peripheral inflammatory tone to vascular/BBB remodeling and central immune-cell mobilization (Zegeye et al., 2018). By contrast, neuron-derived CX3CL1 tunes microglial CX3CR1 signaling and synaptic homeostasis (Sheridan and Murphy, 2013), and a cohort study suggests a functional dissociation in PTSD, with CCL2 linked to onset risk and CX3CL1 to resilience (Zhang et al., 2020).

PE may recalibrate this circuit-immune coupling through exerkines that connect peripheral physiology to central plasticity (Severinsen and Pedersen, 2020). By modulating NF-κB–linked inflammatory signaling and microglial states, PE may shape downstream trophic pathways, including BDNF and IGF-1 (Mee-Inta et al., 2019; Strohm and Majewska, 2024). FNDC5/irisin exemplifies this integration: in a single prolonged stress (SPS) PTSD model, exogenous irisin reduced anxiety-like behavior and rescued fear-extinction deficits through an AMPK-dependent mechanism, increasing p-AMPK in the hippocampus, frontal cortex, and amygdala while suppressing NF-κB-linked inflammatory mediators (Xie et al., 2025). Complementary work further suggests microglial integrin αVβ5/AMPK signaling and autophagy engagement as part of this anti-inflammatory irisin response (Zhang et al., 2024).

PE studies in rodent depression- and PTSD-related stress models reinforce a coupled trophic–inflammatory response to physical training. In chronic stress paradigms, treadmill exercise preconditioning upregulates hippocampal FNDC5/BDNF signaling and improves stress-related behavior (Babaei et al., 2021), increases hippocampal BDNF and IGF-1 alongside synaptic plasticity readouts and behavioral recovery (Kang et al., 2020), and attenuates hippocampal inflammatory cytokines including IL-6 and IL-1β (Xiao et al., 2021). In PTSD paradigms, treadmill exercise improves anxiety-like behavior and is accompanied by increased neurotrophic support (BDNF, and in some designs IGF-1) in stress-relevant regions, with effect magnitude shaped by hormonal status and the timing of training relative to the stressor (Amiri et al., 2024; Eshaghi-Gorji et al., 2024; Mirjalili et al., 2022). Collectively, these findings support a subtype-sensitive cascade in which elevated inflammatory tone may constrain neuroimmune signaling, whereas PE-responsive FNDC5/irisin-linked trophic pathways (BDNF/IGF-1) promote plasticity relevant to affect regulation and stress resilience.

6. Conclusion and future perspectives

This review argues that exerkines mediate muscle–brain crosstalk across AD, PD, and stress-related mental disorders. FNDC5/irisin, BDNF, CTSB, IL-6, and IGF-1 show modality-, intensity-, and time-dependent kinetics and may reprogram glial pathways governing proteostasis, mitochondria–lysosome function, inflammatory tone, and synaptic plasticity, potentially via BBB trafficking and endothelial signaling. Human studies remain difficult to compare due to heterogeneous biospecimens, inconsistent sampling timepoints and fasting status, variable preprocessing, and limited normalization to PE dose, muscle mass, and fitness. Preclinical work rarely establishes molecule-specific causality because paradigms are nonstandardized, full-length proteins versus cleavage products are not consistently distinguished, BBB protein/EV tracing is incomplete, muscle origin is often unconfirmed, and secretion kinetics are poorly aligned with human datasets. Under characterized candidates such as CX3CL1 and MCP-1 may be relevant to inflammatory subtypes of depression, anxiety, and PTSD, yet their regulation and causal roles remain unclear. Priority next steps include standardized workflows, harmonized longitudinal assays, origin-confirming tagging with protein/EV tracing, and multi-omics discovery with causal validation.

Glossary

¹²⁵I–IGF-1

iodine-125–labeled insulin-like growth factor-1

5′-UTR

5′ untranslated region

Aβ

amyloid-β

AD

Alzheimer’s disease

ADAM

a disintegrin and metalloproteinase

AE

aerobic exercise

Akt

protein kinase B (PKB)

ALS

amyotrophic lateral sclerosis

AMPK

AMP-activated protein kinase

α-syn

alpha-synuclein

BBB

blood–brain barrier

BBS

Berg Balance Scale

BDNF

brain-derived neurotrophic factor

CNS

central nervous system

CRP

C-reactive protein

CSF

cerebrospinal fluid

CTSB

cathepsin B

CX3CL1

C-X3-C motif chemokine ligand 1 (fractalkine)

CX3CR1

C-X3-C motif chemokine receptor 1 (fractalkine receptor)

DCX

doublecortin

eNOS

endothelial nitric oxide synthase

EPS

extrapyramidal symptoms

ER

endoplasmic reticulum

ERK

extracellular signal-regulated kinase

EVs

extracellular vesicles

GH

growth hormone

GPR37

G protein-coupled receptor 37 (prosaposin receptor)

HIIT

high-intensity interval training

HPA

hypothalamic–pituitary–adrenal (axis)

IGF-1

insulin-like growth factor-1

IGFBP-3

insulin-like growth factor binding protein-3

IL-1β

interleukin-1 beta

IL-6

interleukin-6

IL-6Rα

interleukin-6 receptor alpha

LPS

lipopolysaccharide

LTP

long-term potentiation

MAPK

mitogen-activated protein kinase

MCP-1

monocyte chemoattractant protein-1 (CCL2)

MDD

major depressive disorder

MGF

mechano growth factor

MMPs

matrix metalloproteinases

mTORC1

mechanistic target of rapamycin complex 1

MPTP

1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine

NGF

nerve growth factor

NLRP3

NLR family pyrin domain containing 3

PD

Parkinson’s disease

PE

physical exercise

PGC-1α

peroxisome proliferator-activated receptor gamma coactivator 1-alpha

PI3K

phosphoinositide 3-kinase

PSAP

prosaposin

PTSD

post-traumatic stress disorder

RAGE

receptor for advanced glycation end products (AGER)

RE

resistance exercise

sIL-6R

soluble interleukin-6 receptor

SIRT1

sirtuin 1

SOCS3

suppressor of cytokine signaling 3

SPS

single prolonged stress

STAT3

signal transducer and activator of transcription 3

STAT5B

signal transducer and activator of transcription 5B

TFEB

transcription factor EB

TH

tyrosine hydroxylase

TNF-α

tumor necrosis factor-alpha

TrkB

tropomyosin receptor kinase B

UPS

ubiquitin–proteasome system

VE-PTP

vascular endothelial protein tyrosine phosphatase

Funding Statement

The author(s) declared financial support was received for this work and/or its publication. This work was supported by the Ministry of Education of the Republic of Korea and the National Research Foundation of Korea (2025S1A5A2A03014466).

Author contributions

SY: Visualization, Writing – original draft, Investigation, Writing – review & editing. H-WS: Writing – review & editing. SK: Writing – review & editing. E-JC: Writing – review & editing. YC: Writing – review & editing. DK: Writing – review & editing. YJ: Writing – review & editing. D-HP: Writing – review & editing. H-BK: Conceptualization, Writing – original draft, Project administration, Supervision, Writing – review & editing. JY: Supervision, Conceptualization, Funding acquisition, Writing – original draft, Writing – review & editing.

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.