Physical activity and the endocannabinoid system: an overview

Abstract

Recognized as a “disease modifier”, physical activity (PA) is increasingly viewed as a more holistic, cost-saving method for prevention, treatment and management of human disease conditions. The traditional view that PA engages the monoaminergic and endorphinergic systems has been challenged by the discovery of the endocannabinoid system (ECS), composed of endogenous lipids, their target receptors, and metabolic enzymes. Indeed, direct and indirect evidence suggests that the ECS might mediate some of the PA-triggered effects throughout the body. Moreover, it is now emerging that PA itself is able to modulate ECS in different ways. Against this background, in the present review we shall discuss evidence of the cross-talk between PA and the ECS, ranging from brain to peripheral districts and highlighting how ECS must be tightly regulated during PA, in order to maintain its beneficial effects on cognition, mood, and nociception, while avoiding impaired energy metabolism, oxidative stress, and inflammatory processes.

Physical activity (PA) at a glance

Physical activity (PA) is widely defined as any movement produced by skeletal muscles that results in energy expenditure, while physical exercise is a subset of PA that is structured, planned, and repetitive [1]. In humans, PA can be done in multiple ways that differ for type, intensity, frequency, and duration. For instance, aerobic exercise involves dynamic activity of large groups of muscles, while resistance exercise specifically increases muscular strength, power, and endurance by using smaller groups of muscles [1]. Also, the intensity of the exercise protocol is important: fatty acids represent the main source of energy at low and moderate intensities, while at high intensities, glycogen and phosphocreatine become the main fuels [1, 2].

As discussed in more detail in the next sections, a correctly performed PA improves cognitive performance [3], and affects the brain reward system [4] by inducing a sense of satisfaction. By modulating several signaling mediators, exercise influences mood [3] and nociception [5], modulates immune system functions [6], as well as whole-body energy metabolism [7]. These features contribute to the concept that PA is not only important for a healthy everyday life, but may also help to prevent and even treat chronic diseases that afflict modern society because of a sedentary lifestyle [8]. In particular, the beneficial effect of physical exercise has been clearly documented under different pathophysiological conditions affecting the central nervous system (CNS) [8], as well as peripheral tissues [9–11].

It seems noteworthy that exercise can become harmful in the case of unaccustomed, vigorous, or eccentric muscle actions that overstretch and hence, disrupt sarcomeres, and induce oxidative stress and inflammatory responses [12]. Although the body has been shown to adapt to harmful exercises [12], damaging activities can clearly promote or aggravate pathological conditions [12, 13].

The endocannabinoid system (ECS)

Endocannabinoids (eCBs)

In the early 80s, several authors investigated the biological background of exercise-wide effects. Early hypotheses relied on the well-documented antidepressant and anxiolytic effect of PA [14], and pointed to a possible role of monoamines [15] and endorphines [16] in exercise neurochemistry. In the last decade, the evidence that PA induces some of the psychotropic effects elicited by the Cannabis sativa active ingredient Δ⁹-tetrahydrocannabinol (Δ⁹-THC, Fig. 1), like bliss, euphoria, and peacefulness, strengthened the hypothesis that endocannabinoids (eCBs) might mediate, at least in part, the central and peripheral effects of exercise [14].

Fig. 1.

Chemical structure of the main plant-derived and endogenous cannabinoids

The eCBs are lipid mediators including esters, amides, and ethers of arachidonic acid [the most active ω-6 polyunsaturated fatty acid (PUFA) in our body], which mimic the effects of Δ⁹-THC primarily by binding to and activating type-1 (CB₁) and type-2 (CB₂) cannabinoid receptors [17, 18]. The most studied eCBs are N-arachidonoylethanolamine (anandamide, AEA), an N-acylethanolamine (NAE) [19], and 2-arachidonoylglycerol (2-AG), a monoacylglycerol [20, 21] (Fig. 1). The NAE family also includes the appetite-suppressor N-oleoylethanolamine (OEA) and the anti-inflammatory and anti-proliferative N-palmitoylethanolamine (PEA) (Fig. 1) [22]. Both OEA and PEA are considered eCB-like compounds rather than authentic eCBs, because they usually do not bind to CB₁ or CB₂, but have some cannabimimetic or non-cannabimimetic actions through the activation of other molecular targets (see below) [22].

The eCB family also includes other bioactive members, like 2-arachidonoylglycerol ether (noladin ether) [23], O-arachidonoylethanolamine (virodhamine) [24] and N-arachidonoyldopamine (NADA) [25]. More recently, both N-docosahexaenoylethanolamine (DHEA) and N-eicosapentaenoylethanolamine (EPEA) (derived from the dietary ω-3 PUFA docosahexaenoic acid and eicosapentaenoic acid, respectively) (Fig. 1) have been shown to activate both CB₁ and CB₂ [26]. Although a clear understanding of the biological and pathophysiological relevance in vivo of these “ω-3 eCBs” is still missing, they are receiving growing attention as potential anti-proliferative agents [27].

eCB-binding proteins

AEA and 2-AG activate different signaling pathways depending on the specific receptor engaged [17, 18]. CB₁ and CB₂ are G-protein coupled receptors (GPCRs) that represent the main targets of eCBs [17, 18]. CB₁ is highly expressed in brain areas that control emotionality, cognition, memory, motor, and nociception, and that include cortex, limbic system, hippocampus, cerebellum, and several nuclei of the basal ganglia. CB₁ is also expressed in peripheral cells and tissues, including adipose tissue, liver, and skeletal muscle [17, 18]. CB₂ is mainly present in the immune system [28], but is also expressed within the CNS [29], where it might play a relevant role in coping with insults [30].

Both CB₁ and CB₂ are coupled to G_i/o proteins, thus inhibiting adenylyl cyclase and stimulating different members of the mitogen-activated protein kinase (MAPK) family [17, 18]. CB₁ can also activate A-type and inwardly rectifying K⁺ currents and inhibit N- and P/Q-type calcium currents [17, 18]. In addition, CB₁ has been recently shown to stimulate G_s proteins [18], and to be modulated by specific accessory proteins [17].

Accumulated experimental evidence indicates that endogenous, plant-derived (like Δ⁹-THC and the non-psychoactive substance cannabidiol, CBD) (Fig. 1) and synthetic cannabinoids (synthocannabinoids) can activate the G protein-coupled receptor 55 (GPR55) [31, 32]. This orphan receptor shares low sequence homology with CB₁ and CB₂, is coupled to G_αq and G_α12/13, and has been proposed as a possible “CB₃”. It regulates several biological processes both in the CNS and peripherally, by activating small GTPases (like Rho) and distinct MAPK kinases [31, 32].

In addition to CB receptors, AEA but not 2-AG binds to transient receptor potential vanilloid type 1 (TRPV1) Ca²⁺ channels, highly expressed in a subset of primary sensory neurons, both centrally and peripherally, and in various non-neuronal cells [33]. By triggering signaling cascades associated with Ca²⁺ homeostasis, the AEA-TRPV1 interaction (that occurs at the inner side of the plasma membrane) leads to the control of CNS functions including nociception [34], and of basic biological processes like the induction of apoptosis [35, 36].

Finally, both AEA and 2-AG bind (directly or indirectly) to nuclear peroxisome proliferator-activated receptors (PPARs) α and γ, thus regulating lipid and glucose metabolism [37, 38], as well as inflammatory responses [39].

Concerning eCB-like compounds, both OEA and PEA seem to act through the activation of PPAR-α, thus either regulating feeding and body weight (OEA) or reducing inflammatory responses (PEA) [22, 40]. However, some effects of PEA might also depend on the activation of CB receptors, as well as of TRPV1 [22, 40], thus suggesting that eCB-like compounds deserve further investigation.

eCB metabolic enzymes

eCBs are produced in response to numerous stimuli (i.e., neuronal activity, glucocorticoids, insulin, and cytokines) [41–45] by a variety of cells throughout the body, from brain [41] to peripheral tissues like adipose tissue [38], muscles [45], heart, kidney [41], and immune cells [46, 47], just to name a few.

Indeed the biological activity of eCBs is tightly regulated by their metabolism [48]. The main biosynthetic enzyme of AEA is a Ca²⁺-dependent N-acylphosphatidylethanolamine-hydrolyzing phospholipase D (NAPE-PLD) [40], which is specific for NAEs and releases AEA from N-arachidonoylphosphatidylethanolamine (NArPE). Additional routes allow AEA biosynthesis [41]. For instance, a tyrosine phosphatase can release AEA from phospho-AEA, which in turn derives from NArPE through the action of a phospholipase C [49]. Moreover, the finding that mice lacking the NAPE-PLD gene have unaltered NAE levels in their brain allowed researchers to demonstrate that NAE biosynthesis also occurs through the sequential action of the α–β hydrolase enzyme ABH4, and of glycerophosphodiesterase 1 [50]. AEA is inactivated by a two-step process: cellular uptake and intracellular degradation. As yet, several hypotheses have been made to explain AEA transport across plasma membrane, including passive diffusion, carrier-mediated bidirectional transport through a specific eCB membrane transporter (EMT), and caveolae-related endocytosis [51, 52]. Moreover, it has recently been shown that AEA, once transported inside the cell, is shuttled to its final destinations by intracellular transporters, like fatty acid-binding proteins 5 and 7, heat-shock protein 70, albumin, and fatty acid amide hydrolase (FAAH)-like transporter [48, and references therein]. In this context, a relevant role is played by adiposomes that are intracellular lipid droplets representing a dynamic reservoir for AEA sequestration and a platform for accumulation, trafficking, metabolism, and signaling of this eCB [48, 53]. The most relevant degradative enzyme of AEA is the serine hydrolase FAAH (now called FAAH-1) that hydrolyzes it into arachidonic acid and ethanolamine [54, 55]. Additionally, AEA hydrolysis can be catalyzed by a newly discovered FAAH-2 [56], or by a lysosomal N-acylethanolamine-hydrolyzing acid amidase [41]. In this context, it should be recalled that both OEA and PEA share, at least in part, the catabolic routes of AEA [22, 40]. Therefore, they might potentiate the activity of true eCBs by inhibiting their degradation, through the so-called “entourage effect” [22].

Also the biosynthetic pathway of 2-AG begins with the hydrolysis of its phospolipid precursors, by the means of phospholipase C or phosphatidic acid phosphohydrolase that generate diacylglycerol [41]. The latter is then converted to 2-AG by sn-1-diacylglycerol lipases (DAGLs) α and β, which are abundant in adult and developing nervous tissues, respectively [57]. After its cellular uptake, most likely through the same EMT that transports AEA, 2-AG is hydrolyzed by monoacylglycerol lipase (MAGL) [58], although FAAH, and more interestingly, other lipases like αβ-hydrolases 6 and 12 [59, 60], can contribute to its degradation. A schematic representation of the ECS is reported in Fig. 2.

Fig. 2.

Schematic representation of the ECS. 2-AG 2-arachidonoylglycerol, AEA anandamide, AITs anandamide intracellular transporters, CBR cannabinoid receptors, DAGL diacylglycerol lipase, EMT endocannabinoid membrane transporter, FAAH-1/2 fatty acid amide hydrolase 1 and 2, MAGL monoacylglycerol lipase, NAPE-PLD N-acylphosphatidylethanolamine-hydrolyzing phospholipase D, PPAR nuclear peroxisome proliferator-activated receptor, TRPV1 transient receptor potential vanilloid type 1. See text for further details

Although the primary fate of AEA and 2-AG seems to be their inactivation via hydrolysis, they might also be oxidized by cyclooxygenase-2 and lipoxygenase isozymes, thus producing oxidized eCBs that are involved in regulating brain synaptic transmission and other biological processes [61, 62].

Changes in the ECS during PA

To our knowledge, the first experimental study aimed at investigating the influence of PA on ECS in humans was carried out in 2003 by Sparling and coworkers [63], who showed increased plasma AEA content after 45 min of moderate intensity exercise on a treadmill or cycle ergometer. Since then, other human studies have shown increased blood concentrations of AEA, but interestingly not of 2-AG, after 30–45 min and up to 5 h of aerobic exercise [64–67]. Of note, the eCB-like compounds PEA and OEA have been shown to parallel the changes of AEA during 60 min of moderate aerobic exercise, immediately followed by 30 min of intense exercise, reaching a peak during the subsequent 15 min of recovery [65].

A dependence of the increase of AEA concentration on exercise intensity has also been documented. Plasma levels of AEA significantly increased upon 30 min of moderate exercise (heart rate of 72 and 83 %), but not at lower and significantly higher exercise intensities, where the age-adjusted maximal heart rate was 44 and 92 %, respectively [67]. Finally, AEA levels were enhanced by combining exercise with mild hypobaric and hypoxic conditions due to high altitude [64].

Moreover, our group has recently shown that an active lifestyle upregulates lymphocyte FAAH activity through an interleukin (IL)-6-dependent stimulation of the FAAH gene promoter [42]. Finally, two additional studies have investigated the effects of human chronic exercise combined with dietary restriction on ECS. In the first study, a 1-year life intervention was able to lower AEA and 2-AG plasma levels in viscerally obese men, where an overstimulated ECS is known to promote weight gain and metabolic impairment [68]. The second study demonstrated that the expression of the FAAH-encoding gene was lowered in abdominal adipose tissue of overweight or obese women, who followed concomitant aerobic and diet programs [69].

Animal studies support these investigations, confirming that eCBs are also modulated in the CNS [70, 71]. For instance, 8 days of voluntary exercise is able to increase AEA levels, CB₁ density, and activity in rat hippocampus [70]. A very recent study has reported that in rats, acute aerobic exercise [a single bout on a rodent treadmill, until fatigue (≈50 min), after 3 days of training] triggers eCB-dependent analgesic effects at central and peripheral levels [71]. The latter is the first study to report increased 2-AG levels after a moderate acute aerobic exercise protocol, and is among the few articles indicating a positive peripheral effect of eCBs. Indeed, that eCBs increase during exercise nicely supports the belief in positive effects of PA on the CNS, but not on peripheral tissues (see below).

It seems of particular interest that short exercise protocols are accompanied by the upregulation of eCB signaling in murine striatum [72], whereas longer training is paralleled by reduced CB₁ expression in the striatum and hippocampus of adolescent rats [73], and adipose tissue of rats on a high-fat diet (HFD) [74]. Similarly, acute aerobic exercise affects plasma levels of AEA, but usually not of 2-AG [63–67], which is instead affected by chronic exercise only when combined with caloric restriction [68]. Altogether, these results seem to suggest that ECS not only activates different pathways depending on exercise length, but can even adapt to a physically active lifestyle [42]. This hypothesis is further supported by a transversal study conducted on humans and dogs (two cursorial species), and ferrets (a non-cursorial species). AEA, but not 2-AG, levels were found to increase in humans who ran for 30 min at moderate intensity, and in dogs that walked or ran for the same time and at the same intensity; in contrast, no changes were reported in ferrets under the same experimental conditions [66]. Humans and dogs have evolved a behavior suitable for running with a neurobiological motivational reward, and hence, a reduction in anxiety and pain, in order to improve their fitness condition; instead, ferrets are not adapted to long-distance running at high speeds, but rather use high accelerations over short distances [66]. In the light of these evolutionary differences, the available data appear supportive of a role for eCBs in PA of those mammals that are “made for” endurance exercise [66].

Implications of the ECS in motor activity: from brain to muscle

Every movement, whether simple or not, is based on a complex network of connections extending from different brain areas to muscles [75] and that, as schematically depicted in Fig. 3, have been shown to be influenced by the ECS. Solid experimental data, for instance, indicate that the ECS controls motor activity by interacting with the dopaminergic system in the basal ganglia network. Here, (1) CB₁ co-localizes with both type-1 and type-2 dopamine (D1 and D2) receptors [18]; (2) AEA production is enhanced by D2 receptor activation [43], thus inducing a process known as long-term depression (LTD) in high-frequency-stimulated striatal slices; and (3) a CB₁ blockade leads to long-term potentiation (LTP) expression [76].

Fig. 3.

Schematic representation of the main effects triggered by ECS in body areas involved in locomotor activity. ACh acetylcholine, LTD long-term depression, LTP long-term potentiation, SERCA sarcoplasmic reticulum Ca²⁺-ATPase. See text for further details

In the cerebellum, synaptic activation of Purkinje cells triggers the release of 2-AG that retrogradely mediates LTP and depolarization-induced suppression of excitation, thus contributing to the coordination, precision, and accurate timing of movement [77].

The ECS, and particularly CB₁, is also implicated in the regulation of contraction frequency. In the isolated lamprey spinal cord, eCBs released upon type-1 metabotropic glutamate receptor (mGluR1) activation induce LTD during rhythmic locomotion, in a CB₁-dependent manner [78]. More recently, it has been reported that 2-AG is mobilized from neurons of the spinal network during locomotor activity, and acts via CB₁ in synergy with nitric oxide (NO), to shift the balance between excitation and inhibition, and to potentiate locomotor frequency [79]. Interestingly, CB₁ has been shown to positively regulate transcription factors that control corticospinal motor neuron differentiation [80]. Accordingly, CB^−/−₁ mice show alterations in corticospinal motor neuron generation and subcerebral connectivity that cause defective skilled motor function [80].

Studies on vertebrate striated neuromuscular junctions have shown that activation of CB₁ in nerve terminals (where the receptor is highly expressed) inhibits acetylcholine release [81]; similarly, in isolated fast and slow muscles of frog, CB₁ has been reported to reduce tension evoked by caffeine, which is known to act on sarcoplasmic calcium release [82]. In this context, it should be stressed that calcium handling and energy supply are important events for movement. Indeed, ATP is necessary for muscle contraction, and also for key calcium-handling proteins like sarcoplasmic reticulum Ca²⁺-ATPase (SERCA), whose uncoupling causes muscle fatigue [83, 84]. NADA stimulates SERCA uncoupling with subsequent cytoplasmic ATP depletion [85], while both AEA and AM404 (an EMT inhibitor) directly inhibit Ca²⁺ flux in transverse tubule membrane vesicles derived from rabbit skeletal muscle cells [86, 87].

Concerning the general effects of systemic administration of (e)CBs, it is known that the locomotor activity of mice and rats is enhanced by AEA and other CB₁ agonists in some studies [88, 89], but is decreased in others [90–92]. Moreover, the finding that both the activation and the blockade of CB₁ decreased running activity in mice selectively bred for high voluntary wheel running, as well as in their wild-type littermates [93, 94], further confirms that the effects of ECS on movement not only depend on variables in the experimental paradigm (e.g., strain, sex, age, and locomotor measurements), but also on the intrinsic complexity of the system per se.

Interplay between PA and ECS in the CNS

Several experimental data support the hypothesis that ECS might, at least in part, explain PA effects on brain functions, because: (1) CB₁ is the most abundant GPCR in the brain participating in neuronal plasticity [18]; (2) eCBs are involved in several brain responses that greatly overlap with the positive effects of exercise; (3) eCBs are able to cross the blood–brain barrier [95]; and (4) exercise increases eCB plasma levels [64–67].

Cognition

Physical activity improves brain plasticity and cognition by increasing hippocampal neurogenesis and cell proliferation, as well as dendritic length and complexity [3, 96]. Moreover, exercise critically influences the pleiotropic substance and mood controller brain-derived neurotrophic factor (BDNF), responsible for the development, regeneration, survival, and maintenance of neurons [3, 97]. In addition, ECS, and particularly CB₁, is involved in the modulation of neurogenesis (Fig. 4). Indeed, neural progenitor cells express high levels of CB₁, able to promote in vitro neuroprogenitor differentiation and maturation, via inhibition of the extracellular-signal-regulated kinase 1/2 (ERK1/2) signaling pathway [98]. In vivo, CB₁ contributes to embryonic and adult hippocampal neurogenesis [99], which coherently decreases in CB^−/−₁ mice [100], along with impaired progenitor cell proliferation and astrogliogenesis [101]. It seems noteworthy that increased hippocampal neurogenesis correlates with both high AEA levels and increased CB₁ activity in the same brain area [70]. Finally, eCBs positively regulate BDNF activity in the brain: reduced levels of this neurotrophin are detected in the brain of CB^−/−₁ mice [102], whereas increased BDNF content is found upon CB₁ activation [103] (Fig. 4).

Fig. 4.

Schematic representation of the interplay between PA and ECS in the CNS. Red arrows indicate negative effects, green arrows indicate positive effects. Specific references are included for each item. References in red PA/ECS interplay confirmed in the specific brain area, references in green suggested PA/ECS interplay for the indicated function, but not necessarily related to the particular area, references in black ECS effects not yet proven to be related to exercise. See text for further details

An interplay between ECS and PA in neurogenesis has been recently suggested [70]. Daily administration of the CB₁ agonist AM251 to rats with free access to a running wheel for 8 days abrogates a voluntary exercise-induced increase in cell proliferation [70] (Fig. 4). Yet, these data seem to be at variance with another study that failed to show any difference in hippocampal neurogenesis between CB^−/−₁ mice allowed free access to a running wheel for 6 weeks and their wild-type littermates [91]. Nonetheless, these discrepancies might depend on differences in species (rats vs mice), experimental approach (pharmacological vs genetic), and/or duration of running (8 days vs 6 weeks).

Pharmacological elevations of AEA (via inhibition of FAAH activity) enhance learning and memory in rats, in a CB₁- and PPAR-α-dependent manner [104, 105] (Fig. 4). In this context, it should be recalled that OEA and PEA are also substrates of FAAH, and that these eCBs-like compounds, though inactive at CB₁, bind to and activate PPAR-α [105]. Therefore, it is tempting to speculate that both OEA and PEA might contribute to AEA-induced improvement of hippocampal functions via PPAR-α signaling. On the other hand, other studies failed to confirm such a positive effect of eCBs. Indeed, FAAH and EMT inhibitors (URB597 and AM404, respectively), as well as CB₁ agonists (Δ⁹-THC, HU-210 or R-methanandamide) were found to negatively influence working memory and/or short-term memory of rats [106]. These apparently conflicting data might be explained by recalling that pharmacological approaches might increase the eCBs content above physiological levels, thus affecting other brain processes that, in turn, might reduce performance in cognition tests [107, 108].

Mood

Due to the increased demand for energy, wheel running activates the hypothalamic–pituitary–adrenal (HPA) axis, thus raising circulating levels of glucocorticoids, endocrine signals that change in adaptation to stress [109, 110]. Moreover, acute and chronic exercise influences the activity of BDNF [3, 65], and reduction of the latter substance in the hippocampus (observed, for example, following stress-dependent activation of the HPA axis) critically contributes to stress vulnerability and depression [3]. Accordingly, voluntary running has been found to positively correlate with hippocampal BDNF levels in a rodent model of depression and anxiety [3].

(e)CBs and BDNF might interact in mood control (Fig. 4). For example, Δ⁹-THC treatment of intact and olfactory bulbectomized rats (an animal model of depression) increased BDNF expression in hippocampus and frontal cortex, thus exerting anti-depressant properties [111] (Fig. 4). Pharmacological inhibition of FAAH elicited anxiolytic-like and antidepressant-like effects in chronically and mildly stressed rats (a behavioral model with high isomorphism to human depression) [112]. Instead, a blockade of CB₁ precipitated depression-like symptoms in clinical trial subjects [113].

eCBs have been also shown to regulate the HPA axis (Fig. 4). Glucocorticoids, indeed, quickly increase brain levels of eCBs, which suppress excitatory inputs in the paraventricular nucleus, rapidly shutting down the HPA axis in a CB₁-dependent manner [44]. Involvement of the ECS in stress response is also supported by a study performed on a rat model of post-traumatic stress disorder, showing that WIN55,212-2 (injected systemically or in basolateral amygdale) prevents stress-induced effects [108]. Accordingly, CB^−/−₁ mice exhibit an increased susceptibility and neuroendocrine response to chronic stress [114, 115].

Finally, a recent study enrolling healthy, trained male cyclists showed that intense aerobic PA behaves as a physiological stressor, able to increase plasma AEA levels, that in turn might mediate the neuroplastic and anti-depressant effects of exercise through a BDNF-dependent mechanism [65] (Table 1; Fig. 4).

Table 1.

Effect of PA on the ECS in humans

Subjects	Exercise protocol	Main finding	Reference
Normal weight male runners (n = 8), cyclists (n = 8), controls (n = 8)	Acute exercise. Running on a treadmill/cycling on a stationary bike for 45 min (HRmax = 70–80 %)	AEA content increases in plasma	[63]
Asymptomatic obese men (n = 49)	Chronic exercise. 1-year life intervention (diet plus exercise program, but with no indication on intensity)	Decrease of plasma AEA and 2-AG levels correlates with reduced visceral adipose tissue	[68]
Trained healthy volunteers (n = 12)	Acute exercise. A: strenuous hiking below an altitude of 2,100 m; B: strenuous hiking and descent the next day; C: passive ascent with helicopter (no indication on intensity parameter in any protocols)	Plasma AEA content increases with altitude Plasma 2-AG content does not change	[64]
Overweight/obese women: caloric restriction alone (n = 9) or plus moderate (n = 13) or high intensity (n = 8) exercise	Chronic exercise. 20-week intervention: caloric restriction alone (VO2max or in combination with 3 day/week of moderate (HRmax = 45–50 % for 15–55 min) or vigorous (HRmax = 70–75 % for 15–30 min) intensity exercise	FAAH gene expression decreases in abdominal adipose tissue, if exercise is combined to caloric restriction	[69]
Sedentary or minimally active (<60 min/week) females (n = 8) and males (n = 4)	Chronic exercise. 10 interventions during 2 weeks with a 30 min treadmill exercise (HR = 60 %)	Chronic exercise decreases craving for marijuana	[119]
Humans (n = 10), mixed-breed dogs (n = 8) and ferrets (n = 8)	Acute exercise. 30 min treadmill running (HRmax = 72 %) or walking (HRmax = 45 %)	Plasma AEA content increases only in cursorial animals	[66]
Females (n = 4), males (n = 6)	Acute exercise. 30 min treadmill running at low (HRmax = 44 %), moderate (HRmax = 72 and 83 %), or high (HRmax = 92 %) intensities	Plasma AEA content increases only at moderate intensities	[67]
Healthy, trained male cyclists (n = 11)	Acute exercise. Ergometric bicycle: 60 min moderate exercise (55 % of maximal trial power output), followed by 30 min intense exercise (75 % of maximal trial power output) and 15 min recovery	Increased plasma AEA, PEA, and OEA levels correlate with those of BDNF and cortisol	[65]
16 healthy males: sedentary (n = 8) and active, regularly practicing running, swimming, cycling for 8.1 ± 1.2 h/week (n = 8)	Chronic exercise. Regular aerobic activity. Samples collected in resting conditions (i.e., no exercise 12 h prior to testing)	FAAH activity is higher in lymphocytes from active subjects	[42]

HR _max maximum heart rate, VO _2max maximal oxygen uptake

Rewarding

Cognitive processes are also influenced by reward circuitry and by dopamine signaling [107], which are crucial in processing environmental rewarding stimuli, in drug addiction, and in voluntary exercise [4, 72, 116]. Expanding on the reward system, it should be recalled that dopamine and eCBs are intimately connected and cooperate in a fine-tuned manner, thus influencing reward due to food, drugs of abuse, and drugs of electrical brain stimulation [117]. Moreover, much like addictive drugs, exercise activates dopamine transmission [4]; consequently, it might reduce drug abuse-dependence by counteracting positive-reinforcing effects produced by cocaine and marijuana [118, 119] (Table 1).

It seems apparent that the reinforcing properties of exercise are mediated by the ECS in mouse striatum, where voluntary exercise alters sensitivity of CB₁ in controlling γ-aminobutyric acid (GABA) transmission (Fig. 4) [72]. Indeed, the synaptic response to the selective CB₁ agonist HU210 was normal after a single day of exposure to a running wheel or to sucrose consumption, but it was potentiated after 7 days of treatment. Remarkably, GABA functions slowly returned to normal after treatment discontinuation [72], thus confirming that the GABA receptor adapts to CB₁ signaling in response to running activity. In keeping with these data, conditional deletion of CB₁ selectively associated with GABA neurons leads to decreased wheel-running performance in mice [116]. Moreover, negative effects of intra-ventral tegmental area (VTA) administration of CB₁ antagonists on voluntary exercise were lost in CB^−/−_1GABA mice, thus indicating that VTA dopaminergic activity is also controlled by eCBs during voluntary exercise [116] (Fig. 4).

Finally, in obese Zucker rats that were allowed to press a lever in order to reach a running wheel, 2-AG was seen to decrease the rewarding effect of exercise in obese rats in a dose-dependent manner, by modulating both lever press and revolution on the running wheel [120] (Fig. 4).

Pain

Pain reduction during exercise is crucial to improving performance and allowing individuals to continue exercise. The mechanism of analgesia associated with exercise is not yet fully understood [5]; nonetheless, several pieces of data suggest that ECS might be involved in exercise-mediated analgesic effects (Fig. 4). eCBs and their molecular targets are, indeed, present in pain-related regions [34], where they have been shown to exert analgesic properties by the retrograde activation of CB₁ [121, 122]. Such an effect is potentiated by the pharmacological or genetic inactivation of FAAH [123, 124]. Coherently, levels of the two eCB congeners, PEA and N-stearoylethanolamine (both substrates of FAAH), have been found to increase after low-force exercise in women with chronic widespread or neck/shoulder pain, in association with reduced pain intensity [125].

Since eCBs are lipids that are able to readily cross the blood–brain barrier [95], it is conceivable that increased eCB concentrations in blood, observed during exercise, might lead to central reduction in pain perception, thus possibly contributing to an immediate feeling of well-being. Indeed, acute aerobic exercise has been demonstrated to increase plasma levels of AEA, 2-AG, PEA, and OEA in rats. Systemic and intrathecal injections of specific antagonists of CB₁ and CB₂, as well inhibitors of eCB transport and hydrolysis, showed that eCBs and their congeners contribute to exercise analgesic effects, either centrally or peripherally. Of note, the wide pharmacological approach used in this study was able to indicate a likely CB₁ and CB₂ involvement in the eCB anti-nociceptive effects during this exercise protocol [71], a fact that should be further explored in the future in order to deepen our knowledge on exercise and ECS.

Appetite

Exercise can raise energy expenditure several fold, so that compensatory mechanisms are activated at both central (i.e., in hypothalamus) and peripheral levels, with an impact on food intake [126].

High-intensity training increases the orexigenic neuropeptide Y (NPY) [127, 128]; similarly, acute exercise protocol increases agouti-related protein (AGRP) plasma levels [128, 129], while the anorexigenic pro-opiomelanocortin and cocaine- and amphetamine-regulated transcript (CART) are reduced [128].

Physical activity also influences release of peripheral orexigenic/anorexigenic mediators that enable communication among different energy-related tissues and the brain. In particular, exercise decreases the release of anorexigenic leptin [130, 131], while raising circulating levels of orexigenic adiponectin (both deriving from adipocytes) [131]. The orexigenic hormone ghrelin (primarily secreted from stomach) seems less responsive, although some studies reported an exercise-dependent decrease [130]. Nevertheless, most of these energy signals are strongly dependent on exercise type, duration, and intensity [132]. Leptin release, indeed, is affected by long-term exercise, while adiponectin levels are influenced by short-term exercise [132]. Likewise, acute exercise elevates post-prandial concentrations of peptide YY (a gut anorexigenic hormone), but not of ghrelin [133], while resistance exercise decreases ghrelin levels, but not peptide YY content [134].

eCBs influence feeding behavior by mainly acting on the hypothalamus, thus contributing to energy homeostasis modulation [for complete reviews, see Refs 45, 135]. 2-AG and AEA, indeed, increase feeding behavior in rodents [136] (Fig. 4). Their levels, reduced by leptin [137] and enhanced by ghrelin and glucocorticoids [44, 138], significantly increase in the hypothalamus during fasting, thus returning to basal levels after feeding [45]. Although levels of AEA are less sensitive to food restriction [45], this eCB increases NPY release in vitro [139]. Finally, FAAH^−/− mice have reduced levels of CART in appetite-related hypothalamus areas [140]. To date, it is clear that (e)CB effects on feeding are CB₁-dependent. Indeed, CB^−/−₁ (much like FAAH^−/−) mice are lean, hypophagic, and resistant to diet-induced obesity [141, 142], while CART-deficient mice are resistant to inhibition of food intake dependent on rimonabant (a CB₁ antagonist) [140]. Genetically and diet-induced obese mice have recently been reported to have more CB₁-expressing inhibitory inputs and 2-AG biosynthetic enzyme DAGL in the lateral hypothalamus, compared with lean littermates, and such alterations are reverted to lean mice conditions upon leptin administration [143].

Interestingly, the influence of wheel running (over a 6-week period) and genotype (CB^−/− vs CB^+/+ mice) on food intake has been documented [91]; accordingly, voluntary exercise on a running wheel enhances the CB₁ inverse agonist-dependent reduction of appetite in mice after 1 day of treatment [144].

Interplay between PA and the ECS in peripheral tissues

Peripheral energy homeostasis

Due to increased energy demand during exercise, PA induces the metabolic remodeling of several tissues that contain lipid and carbohydrate stores, and thus it exerts a general positive effect on peripheral energy homeostasis. Indeed, PA improves insulin sensitivity and metabolic flexibility in muscles [145, 146], diminishes accumulation of lipids in liver [147, 148], and increases lipolysis in adipose tissue by: (1) activating the sympathetic nervous system; (2) increasing adipose tissue blood flow; and (3) stimulating fatty acid oxidation [149] (Fig. 5).

Fig. 5.

Schematic representation of effects of PA and the ECS at the peripheral level. ROS reactive oxygen species, RNS reactive nitrogen species. ↑: increase; ↓:decrease; ↑↓: increase or decrease depending on exercise intensity, type, and duration (when referred to PA), and on activated pathways (when referred to ECS). See text for further details

Conversely, eCB tone mainly exerts a negative influence on peripheral energy homeostasis [45, 135, 150]. For instance, AEA-treated soleus muscles from lean rats showed decreased basal and insulin-induced glucose uptake [151]. Accordingly, a CB₁ blockade in L6 rat myotubes increased glucose uptake through phosphatidylinositol 3-kinase/protein kinase B (Akt) [152] and ERK1/2 kinases [153], without affecting the expression of GLUT1 or GLUT4 glucose transporters [152] (Fig. 5).

To date, no effect has been found on glucose uptake using CB₂ or TRPV1 antagonists [152], though it has been reported that CB^−/−₂ mice, which did not develop diet-induced obesity and insulin resistance, increased insulin-mediated glucose uptake in skeletal muscles [154]. Treatment of primary skeletal muscle cells with AEA or with a pre-adipocyte-conditioned medium inhibited insulin-dependent glucose uptake, decreased Akt phosphorylation, and activated ERK1/2 and p38-MAPKs [155]. Moreover, AEA has been found to further disturb the insulin pathway, by increasing insulin receptor substrate 1 phosphorylation at Ser307, and thus triggering an inhibitory action [155].

Importantly, AEA was also able to induce basal glucose uptake on its own at high concentrations (10 μM), perhaps due to 5′ AMP-activated protein kinase (AMPK) activation [156] (Fig. 5). Indeed, two distinct pools of GLUT4 exist in muscle cells [157] and insulin regulates GLUT4 trafficking principally via the PI3K/Akt pathway; instead, other stimuli like exercise appear to do so (at least in part) via activation of the stress kinase AMPK2 [158]. In line with this, AEA treatment of human primary skeletal muscle myotubes increased AMPK-α mRNA expression [61]. In the same study, AEA (5 μM) was also able to increase the expression of PPAR-γ co-activator-1α, a transcription factor engaged to regulate mitochondrial biogenesis, glucose uptake, and fat oxidation [61]. The latter data might suggest a possible eCB positive effect on muscles, through alternative pathways that could also be targeted by an eCB-like compound, like OEA and PEA. Should this be the case, further studies on eCB-triggered signaling might provide further support to the beneficial effect of an exercise-dependent rise of plasma eCBs.

eCB molecular targets, which are present in several adipose depots, such as visceral and subcutaneous adipose tissues, are linked to energy homeostasis in different ways (Fig. 5). While CB^−/−₂ mice fed HFD maintain insulin sensitivity and do not show signs of obesity-induced inflammation [154], GPR55 expression is elevated in obese subjects and correlates with visceral adipose tissue; moreover, its activation increases Ca²⁺ levels and expression of lipogenic enzymes in differentiated adipocytes [159]. CB₁ activation stimulates glucose uptake in adipocytes [38], and enhances their endocrine function, by increasing visfatin expression and decreasing adiponectin expression [160]. Once again, these effects appear in contrast with those generally elicited by PA [131, 132]. Finally, rimonabant treatment in HFD rats produces profound metabolic remodeling of adipocytes, and hence it counteracts fat accumulation by: (1) affecting glycolysis, via induction of glyceraldehyde-3-phosphate dehydrogenase expression [161]; (2) influencing the tricarboxylic acid cycle and β-oxidation pathway, via increase of carnitine acetyltransferase and carnitine palmitoyltransferase II expression [162]; and (3) impairing lipogenesis, via inhibition of stearoyl-Coenzyme A desaturase 1 expression [163] and pre-adipocyte proliferation [161]. It has also been postulated that ECS might modulate adipose tissue biology by interacting with PPARδ, shown to play a role in adipose remodeling during endurance exercise [164]. In agreement with this hypothesis, HFD rats showed all signs of metabolic syndrome, like adipocyte hypertrophy, elevated CB₁ expression, and reduced PPARδ, which disappeared after a chronic exercise consisting of swimming for 1 h per day thrice a week, for 6 months [74].

Moreover, a study enrolling viscerally obese subjects, who followed a lifestyle program combining a healthy diet and physical activity for 1 year, showed an improved profile of conventional metabolic risk factors, body weight loss, waist circumference reduction, and a 7 % decrease in plasma AEA and a 62 % decrease in plasma 2-AG levels [68] (Table 1). Coherently, lean and agouti mice, given AM251 (a CB₁ antagonist) combined with voluntary exercise, lost more body weight than mice supplemented with the CB₁ antagonist alone [144]. Since obesity and physical inactivity affect insulin and leptin, which both negatively regulate eCB levels [137, 150, 165], it is conceivable that in obese and sedentary subjects, increased eCB tone might depend on the lack of control exerted by the two hormones. Moreover, given that the upregulation of eCB tone affects peripheral energy homeostasis in favor of fat accumulation, a compensatory mechanism aimed at limiting the excessive production of eCBs upon exercise might exist in order to prevent anabolic processes and to favor catabolism and energy supply. In this context, it should be stressed that the in vivo biological actions of eCBs are tightly regulated by a metabolic control, where FAAH has been recognized as a major player [166]. Incidentally, we found that physically active subjects (practicing regular aerobic exercise for about 8 h per week) have increased lymphocyte FAAH activity when compared to sedentary individuals [42].

The liver is a key organ for the storage and disposal of carbohydrates, proteins, and fats, and it is known to be influenced by both PA [146] and ECS [45, 135] (Fig. 5). HFD mice, indeed, showed increased CB₁ expression and AEA levels in hepatic cells [167, 168]. Consistently, a CB₁ blockade decreased liver steatosis and dyslipidemia [167, 168]. On the other hand, at least in healthy subjects it seems that previous chronic exercise (8-week training) does not affect liver ECS response to fasting-refeeding cycles; indeed, in rats fed with fructose (an animal model of liver steatosis), diet-dependent increases of CB₁, CB₂ and FAAH expression were not affected by 8 weeks of treadmill exercise, and instead lowered plasma levels of free fatty acids and triacylglycerols [169].

Collectively, only a few reports confirm [71], suggest, or hypothesize [61, 156] a positive peripheral eCB effect during exercise. Thus, it would be very important to further investigate eCB-dependent mechanisms and/or their tissue specificity, in order to better understand the relationship between PA and ECS. At any rate, it has been convincingly demonstrated that the major role of eCBs at the periphery is opposite to that of PA and, if dysregulated, might also lead to pathological conditions [45, 150]. Therefore, it is conceivable that eCB metabolic routes are modulated to control eCB tone, a hypothesis that has been already demonstrated for FAAH in lymphocytes of subjects practicing regular physical activity [42]. Proving this concept also for other ECS elements might be an asset in the future.

Immune system and redox state

A growing body of evidence points to the ability of exercise to induce, in a type-, duration-, and intensity-dependent manner, physiological changes in the innate and adaptive immune system (Fig. 5). Acute exercise (1) increases the number of circulating monocytes, neutrophils, and lymphocytes [170]; (2) reduces the CD4⁺/CD8⁺ T cell ratio [171]; (3) increases mobilization of memory (but not naïve) lymphocytes [171]; (4) enhances the number and activity of natural killer cells [172]; and (5) markedly increases pro-inflammatory cytokines, such as IL-1β, IL-6, IL-8, and tumor necrosis factor (TNF)-α [173]. Nonetheless, all parameters returned to basal levels within 24 h from the end of exercise [13]. Instead, prolonged exercise or frequent bouts of vigorous exercise temporarily depress immune system, thus resulting in higher susceptibility to infections, particularly among athletes [13]. Against this background, it is well-recognized that regular PA exerts general positive effects in preventing and reducing systemic inflammation, by: (1) downregulating the release of pro-inflammatory cytokines from monocytes and macrophages; (2) decreasing the number of circulating activated immune cells [174]; and (3) increasing the amount of blood T regulatory (Treg) cells [175]. By reducing the accumulation of visceral adipose tissue (a condition often occurring in the case of a sedentary lifestyle), exercise also decreases macrophage infiltration in this organ, thus contributing to lower levels of pro-inflammatory cytokines in the blood [176]. In this context, a key role is played by skeletal muscle that, during contraction, releases myokines like IL-6, IL-8, and IL-15 [177]. Among them, IL-6 plays a metabolic role by increasing insulin sensitivity, hepatic glucose production, lipolysis, and fat oxidation [177, 178], and promotes an anti-inflammatory environment that suppresses the long-term effects of elevated pro-inflammatory cytokines [179].

Regular PA also promotes the generation of reactive oxygen and nitrogen species (ROS and RNS, respectively), which positively regulate redox-sensitive transcription factors (NF-κB and AP-1) [180], force production by contracting muscles [181], exert a beneficial impact on vascular functions [182] and induce satellite cell proliferation and differentiation [183]. The downside is that, when exercise is inappropriate, ROS and RSN levels dramatically increase, thus triggering oxidative stress that contributes to muscle fatigue and damage, as well as to pro-inflammatory responses [181, 184–186] (Fig. 5).

All ECS components are highly expressed in the immune system, where they are differentially modulated depending on leukocyte activation state: CB₂ (undetectable in naïve T cells) and FAAH increase and decrease, respectively, in activated lymphocytes [46, 187–189], while AEA content augments (due to upregulated NAPE-PLD activity) in LPS-stimulated macrophages [46, 47].

The crucial role of ECS in immune responses is well-recognized [28] (Fig. 5): for instance, AEA/CB₂ interaction strongly suppresses CD4⁺/CD8⁺ T lymphocyte proliferation, impairs the pro-inflammatory responses of both T helper (Th) 17 and Th1 cells [189], and stimulates anti-inflammatory activity of Treg and Th2 cells [190] (Fig. 5).

The ECS (particularly CB₁) also modulates the redox state, by regulating inducible and endothelial nitric oxide synthase expression, mitochondrial biogenesis, respiration rate, and ROS/RNS generation [79, 191–194]. Furthermore, CBD has been shown to increase the production of ROS that, in turn, trigger CD4⁺ and CD8⁺ T cell apoptosis [195] (Fig. 5).

Overall, based on its ability to modulate immune responses, ECS might play a role in exercise-related changes of immune functions and cellular redox state. The interplay among eCB tone, exercise, and the immune system is, indeed, evident in subjects habitually practicing moderate exercise, who show increased FAAH activity in their T lymphocytes (but not in other blood cells), when compared to sedentary individuals. Such an upregulation, which is mediated by factors released during exercise, including IL-6 [42], IL-10 (whose release is enhanced by IL-6), and IL-4 [6, 179], represents a compensatory mechanism occurring in active individuals and aimed at avoiding an excessive increment of eCBs upon exercise. Indeed, high eCB levels exert the same effects as prolonged exercise, i.e., depression of the immune system; therefore, both PA and eCB tone should be maintained in an optimal range, in order to favor an anti-inflammatory environment able to counteract side effects like muscle damage and immune suppression [6, 179].

Concluding remarks and future directions

A growing body of evidences strongly indicates an interplay between PA and eCB tone, both centrally and peripherally. Indeed, eCBs are crucial in controlling locomotor activity, and much like exercise, they positively affect cognitive functions, nociception, and other brain processes. Thus, it is conceivable that some exercise-dependent effects in the CNS might depend (at least in part) on eCB signaling. This hypothesis is strongly supported by several human studies showing that exercise is able to increase the circulating levels of eCBs in a type-, intensity- and time-dependent manner. Yet, in the majority of reports, unlike exercise, eCB tone negatively affects peripheral energy homeostasis towards fat accumulation, which better describes an obesity-directed condition or a sedentary lifestyle. It seems necessary, therefore, that eCBs are tightly regulated during PA, in order to maintain their beneficial effects while avoiding impaired energy metabolism, oxidative stress, and inflammatory processes.

Otherwise, future research should probably aim at clarifying these aspects, for example by exploring the influence of different exercise protocols on distinct proteins of the ECS, as well as on different eCB-dependent signaling cascades. As yet, data on the effect of combining both exercise and diet on human eCB levels are still very few, and are all based on chronic exercise protocols. Nonetheless, it is well-known that dietary fat composition influences tissue levels of eCBs [196]: a standard Western diet rich in ω-6 PUFA (i.e., linoleic and arachidonic acids) enhances AEA and 2-AG levels, while reducing the levels of ω-3 PUFAs and their derivatives (e.g., DHEA and EPEA) [197].

Collectively, exercise is increasingly seen as a “disease-modifier”, i.e., a powerful tool to help prevent and treat several diseases, from metabolic disorders to neurodegenerative pathologies. A better understanding of the role of eCB in exercise could further increase the therapeutic efficacy of exercise protocols. This could also broaden exercise-based treatments to diseases with motor deficit, cognitive impairment, as well as oxidative or inflammatory disturbances. Indeed, eCBs might also be an alternative target for drug design, keeping in mind that conventional medicines imply tolerance, drug resistance, or unwanted side-effects.

In conclusion, the study of the combined effects of exercise and eCBs can lead to a more holistic, cost-saving clinical approach in related (and rather common) behaviors, like nicotine and alcohol dependence.