Introduction

Physical activity (PA) is associated with reduced risk of acute myocardial infarction, stroke, hypertension, peripheral artery disease, erectile dysfunction, depression, dementia, malignancies, and improved status in diabetes, obesity, sarcopenia, and cognitive functioning and mental health. Thus, regularly exercising helps in slowdown aging consequences in association with an improvement of the quality of life, possibly promoting longevity [1]. Why exercising is so beneficial for well-being must be uncovered in human history. For hundreds-thousands years, food search and nomadic habits depended on daily endurance activity. Specific metabolic profiles were selected in order to satisfy these peculiar (among primates) energetic requirements. Only very recently, on evolutionary terms, lifestyle has turned to sedentary and the energy balance has become positive. This has led to the exponentially increased prevalence of several diseases whose pathogenesis is associated with suboptimal PA level [2–4]. Physical inactivity has been recognized, by the World Health Organization (WHO) as the fourth death cause (5.5 % of deaths globally) immediately following hypertension (12.8 %), smoke (8.7 %), and hyperglycaemia (5.8 %), and just before obesity and overweight (4.8 %) [5].

Exercising also profoundly affects bone turnover. Indeed, among the few modifiable factors, PA is a key determinant of bone mass and bone metabolism [6]. Exercise-induced gain in bone mass is prevalently due to mechanical load applied to the skeleton, but a critical role is covered by several endocrine axes modified by PA [7, 8]. This is not surprising since bone remodeling occurs constantly and simultaneously in several districts of the skeleton and thus the physiologic energy demands of the skeleton are great [9]. Several energy-associated hormones, myokines, adipokines, and neurotransmitters (i.e., insulin, leptin, adiponectin, epinephrine/norepinephrine) are involved in the fine regulation of bone turnover in response to energy availability/needs (Fig. 1). On the other side, bone regulates the energy metabolism by communicating its energetic needs based on loading thanks to osteocalcin (OC), in both its carboxylated (GlaOC) and undercarboxylated (GluOC) forms, which acts on (but not only) pancreatic β-cells and adipocytes [8].

Fig. 1
figure 1

Crosstalk mechanisms between skeletal muscle, adipose and bone tissues in response to exercise

Full size image

The strict relationship between bone and energy metabolisms emerged in pathology: while diabetic patients experience an increased fracture risk due to reduced bone mineral density (BMD), in osteoporosis metabolic dysfunctions (up to metabolic syndrome and diabetes) are common [9].

Based on this background, aim of this review is to give an overview about the most up-to-date knowledge in the field of exercise-induced changes in bone metabolism by contextualizing the role of novel hormones involved in this regulation.

Though we are aware that patented differences exist between the terms PA and exercise, we will use these two concepts as synonymous because several of the studies we will cite use the terms interchangeably.

Physical activity and bone metabolism

The effects of exercise on bone metabolism depend on type, intensity, and duration of the activity but, more importantly, on the load degree which is the main determinant for bone strength [6].

What is currently accepted is that adolescents and adults participating in endurance activities (i.e., running) as well as not weight-bearing (NWB) sports (i.e., cycling, swimming) have lower BMD than subjects participating in weight-bearing (WB) activities (i.e., ball and power sports) but also compared to their inactive peers [10–14]. Rector and colleagues demonstrated that men aged 20–59 years participating in road cycling had lower lumbar and whole-body BMD and lumbar bone mineral content (BMC) compared to age-matched runners. Osteopenia was diagnosed in 63 % of cyclists while in only 19 % of runners maybe as a consequence of the different load history which was halved in the cyclists [15]. We demonstrated that a NWB activity associated with huge energy consumption, as is a 3-week cycling stage race, determines osteoclast activation [16] sustained by osteocytes. These latter, stimulated by muscular traction in the absence of load, release Sost (sclerostin) and Fibroblast Growth Factor (FGF)-23, finally causing: (i) osteoblast activity inhibition [17]; (ii) increased renal excretion of phosphorous released by the bone resorption [18], despite constant dietary intakes and renal function [19].

The inability of muscular traction alone to sustain bone anabolism depends on Wnt (wingless-type MMTV integration site)-Sost antagonism. Spinal cord-injured rats displayed increased bone resorption, compared to sham rats. Electrically induced muscle contraction slightly attenuated bone resorption without, however, return to sham level. Sost expression was strongly induced in osteoblast treated with sera spinal cord-injured rats and reduced, of only one fourth, in culture treated with sera from electrically stimulated animals [20].

Importantly, high-level continuous exercising, both WB and NWB, affects bone metabolism both dependently and independently from canonical hormonal controls depending on vitamin D [18, 21], parathyroid hormone (PTH) [18, 22], testosterone, and cortisol [16, 17].

In pre-pubertal girls, gain in BMD per unitary increase in body mass was greater for gymnasts than for swimmers (synchronized and speed) and inactive controls together with a reduced amount in fat mass, percentage body fat, and lean mass [23]. In a very large cohort of Korean subjects (4299 males and 5226 women), lean body mass showed an independent association with bone mass and BMD, regardless of gender, age in men, and menopausal status in women [24].

Currently, there is no consensus about optimal intensity, duration, and frequency of PA able to maximize bone formation (or to reverse or slow down age-related bone loss). Heinonen and colleagues correlated intensity and stimulus specificity (sport discipline) and BMD measurement [25]. Recent findings suggest that prolonged multi-component training programs, e.g., WB/high impact (HI) plus resistance training, increase BMD at different sites in both premenopausal [26] and postmenopausal, osteoporotic or not, women [27] as well as in elderly, independently from gender [28, 29]. As a general rule, high strain rates and high peak forces are more effective in stimulating bone formation in young and adults and to limit bone loss in elderly [30, 31]. Indeed, while bone mass is influenced by the applied tension peak [32], bone formation rate depends on the frequency of stimulation [33]. Intermittent loading cycles determines a greater increase in bone formation compared to a single, even prolonged, stimulus [32, 33] and continuous stimulation desensitizes osteocytes [34].

Direct endocrine effects of exercise on bone metabolism

From a cellular point of view, mechanosensation is the mechanism driving bone metabolism. The key actors in this context are osteocytes. Osteocytes representing the 95 % of bone cells are “buried-alive” in the mineralized matrix within lacunae, and a lacuna-canalicular system interconnects adjacent lacunae. Their countless dendrites form a complex network putting them in direct or indirect (by secretion) communication with other osteocytes and osteoblasts. Osteocytes sense mechanical pressure and load, although it is still under debate if this ability resides in the cell body or in dendrites [35, 36]. The importance of osteocytes in mechanosensation has been elegantly demonstrated in transgenic mice with specific ablation of osteocytes (70–80 %). These mice showed fragile fatty bones, due to intra-cortical and trabecular defects and fat infarction (aged skeleton), resistant to unloading-induced bone loss [37]. Mechanotransduction modulates the osteocyte expression of several molecules able to affect bone metabolism: Sost, insulin-like growth factor (IGF)-I and -II, OC, prostanoids, c-fos, and nitric oxide (NO) [38].

Sost is a secreted glycoprotein antagonist of Wnt through blockage of Wnt/β-catenin signaling which is responsible for osteoprogenitor expansion and reduced apoptosis rate in mature osteoblasts. Wnt and PPAR-γ (peroxisome proliferation-activating factor γ) are two factors involved in the key pathways determining the differentiation fate of the mesenchymal stem cell (MSC), osteogenesis, and adipogenesis, respectively [39].

Inactivating mutations of Sost gene result in sclerosteosis and Sost −/− mice have high bone mass [40]. Sost concentrations directly correlate with age, BMI, and BMC, and negatively with bone formation markers [41, 42]. Exercise increases the MSC pool and osteoblastogenesis, at expense of adipogenesis, by inducing Wnt [43–45]. In rodents, Sost expression is downregulated in high strain regions of bone while it is upregulated in unloaded regions [46]. A regulator of Sost expression has been identified in periostin (POSTN): mechanical loading-induced POSTN inhibits osteocytic Sost expression, while the load does not stimulate bone formation, in POSTN−/− mice, due to lacking Sost antagonism [47].

In general, highly active people have lower Sost than low-active people [42, 48]. Resting male athletes involved in WB activities (i.e., rugby, enduro, basketball) had higher serum Sost than NWB athletes (i.e., cyclists), although neither WB nor NWB differed from sedentary and HI athletes (tennis players). This apparently paradox could be read as (i) a feedback mechanism to avoid too much bone formation in WB, or (ii) an effect of resting [49]. Similarly, pre-pubertal rhythmic gymnast, despite their greater BMD compared to untrained controls, have higher sclerostin levels [50] as well as both young and old exercising mice [51]. In competing cyclists, increased Sost concentrations are associated with osteoclast activation [16, 17]. In ultramarathon, instead, Sost was unchanged as well as Dickkopf WNT signaling pathway inhibitor (Dkk)-1, another Wnt antagonist, despite the increase in bone resorption which is marked by the decrease in pro-collagen type I N-terminal propeptide (PINP) and the increase in C-terminal cross-linked collagen type I (CTx-I) [52]. During bed rest, contrarily to bone alkaline phosphatase (BAP) and CTx-I, Sost and Dkk-1 did not respond to either resistive or resistive-vibration exercise protocols, indicating that the impact of unloading on these hormones is greater than the effects of muscular traction [53]. Bone gain in post-menopausal women, following a 1-year long aerobic exercise protocol, was associated with osteoprotegerin (OPG) increase but not with receptor activator of nuclear factor κB ligand (RANKL) and Sost [54].

Muscle-bone axis

According to the “mechanostat theory,” muscle mass gain during growth induces a mild overload on the skeleton leading to an increase in bone mass and strength. Indeed, a lifelong linear association between BMC and lean body mass has been described in a large cohort of 1450 subjects from 2 to 87 years of age [55] and, compared to fat mass, lean mass is a better describer of whole-body BMD [56]. Furthermore, in young and adults, BMC and bone strength are linearly related to muscle size and force [57–59].

Muscle and bone constitute a functional unit. The first (and evident) fact supporting the existence of a “muscle-bone unit” is represented by their mechanical cooperation. From a biological point of view, this unit is characterized by common ontogenesis, genetic control, and hormonal control [60, 61]. Environmental (e.g., PA, nutrition, microgravity) and pathophysiological (e.g., aging, fracture, sarcopenia) conditions influence the whole locomotor system (or one apparatus as a consequence of dysfunctions in the other). This implies a two-way crosstalk [62, 63].

Muscle–bone crosstalk starts in fetal life. MyoD−/−/Myf5−/− mice, which congenitally lack skeletal muscles and die soon after birth, show low mineralization, morphological alterations of long bones, and a higher osteoclast number in new bone [64]. Bone shape and behavior are dependent on muscle contraction and mechanical load [65, 66]. In models of skeletal muscle inactivity, due to either congenital muscleless or induced paralysis, bone and joint show developmental alterations or increased resorption depending on the age at muscular damage [20, 67, 68]. In mice, limb muscles replacement with muscles implants (minced skeletal muscles, non-limb skeletal muscles, or cardiac muscle) allow the formation of cartilage and mineralized bone nodules, while liver tissue or sham implants do not [69]. Furthermore, muscle excision impairs bone fracture healing while, on the contrary, the process is enhanced by high molecular weight muscle-derived molecules [70, 71].

The myostatin system

Bone–muscle, if not associated to loading, fails in sustaining bone anabolism. This has been demonstrated, for example, by in vitro studies, carried out on bone marrow stromal cells (BMSCs) induced to differentiate into osteoblast. The lack of myostatin (growth-differentiation factor [GDF]-8), a member of the transforming growth factor (TGF)-β superfamily negatively regulating muscle growth and, possibly, bone mass [63], in BMSC-derived osteoblast enhanced differentiation but this effect was lost in unloading conditions [72]. In vivo, low-magnitude mechanical signals are sufficient to activate anabolism in both skeletal muscle and bone [72, 73]. In a diet-induced obesity murine model, the age-related loss in trabecular bone is reverted by application of low-magnitude mechanical signals (0.2 g at 90 Hz, 15 min/day, 5 days/week) [74].

The final effect of PA on bone and muscle metabolism depends on the kind of exercise. Short- and long-term recreational soccer training increases both BMD and muscle strength (with a positive strict correlation), compared to both sedentary controls and runners, in male adolescents [75] as well as in premenopausal women [76].

The role of gravitational load could be even more impacting bone metabolism than muscle traction: low-magnitude vibrations (0.2 g at 90 Hz) in the absence of muscular activity, in hind limb unloaded mice (for 3 weeks), restore the bone marrow osteogenic potential [77]. Spaceflights and bedrest, also confirm the importance of load: resistive exercise during 90-day bed rest in young healthy males contrasted the loss in muscle cross-sectional area and BMC [78].

Wnt-independent bone resorption consequent upon a strong endurance effort, as is ultramarathon, has been linked to increased myostatin and decreased follistatin (FST) [52]. FST, follistatin-like (FSTL)1, and FSTL3, members of the TGF-β superfamily, are induced by exercise and act by inhibiting the binding of activins, inhibins, and myostatin to their receptors. The importance of FST-family proteins in bone and muscle development has been highlighted by the severe phenotypes caused by mutation in their genes. Fstl3 −/− mice fractured more often, lost mechanosensitivity and exercise-induced bone gain, and exercise-dependent Sost response. Moreover, aging is associated with decreased FSTL3 expression [79].

IL-6

IL-6, primarily synthesized by liver, is the responsible for acute phase reaction [80]. Low-grade chronic elevation of circulating IL-6 has pro-inflammatory effects [81]. However, IL-6 is also acutely produced by contracting muscles [82] and, in this case, it acts beneficially on energy metabolism (e.g., glucose sensitization) as well as an anti-inflammatory mediator. In order to avoid chronically elevated IL-6, for example, in continuously exercising athletes, the hepatosplanchnic system provides for its rapid removal from blood [83].

IL-6 is also essential for muscle trophism and regeneration [84] since IL-6-deficient mice were unable to recover from hind limb suspension-dependent muscular atrophy [85]. IL-6 responds to exercise depending on mode (muscle mass involved), intensity, and duration. For example, endurance and eccentric activities rise IL-6 up to 100-fold [86, 87]. Contrarily to other muscle markers (e.g., CK, LDH), which are released in blood following both muscle activity and damage [88], exercise-dependent IL-6 release is not associated with any sign of damage [87]. Noteworthy, exercise-dependent IL-6 release, which is not from macrophages, rushed at the muscle to repair damages [89, 90] and macrophage-derived IL-6 is unaffected by exercise [90–92]. The regulation of muscular IL-6 is independent from tumor necrosis factor (TNF)-α, suggesting a role, for IL-6, in metabolism rather than in inflammation [93]. Indeed, low intramuscular glycogen concentrations enhances contraction-induced IL-6 synthesis [94–96] while glucose intake during exercise inhibits it [82, 97].

While PA, mainly endurance, reduces resting IL-6 levels and chronic training attenuates exercise-dependent IL-6 rise, sedentary lifestyle and metabolic syndrome are associated with high basal IL-6 [87, 98]. In mice, 8-week training lowered IL-6 resting levels by 38 % [99]. However, this downregulation muscle sensitivity to IL-6 is increased through the upregulation of IL-6Rα [100]. Hence, in diabetes or metabolic syndrome, the behavior of IL-6 resembles that of glucose: muscle disuse is associated with IL-6 resistance, and high circulating IL-6 is associated with hyperinsulinemia, hyperleptinemia, insulin, and leptin resistance [101].

In vitro and in vivo treatment with rIL-6 increases glucose uptake by muscles, by inducing GLUT-4 translocation to the plasmalemma throughout AMP-activated protein kinase (AMPK) activation [102–105]; it enhances enzymatic activities in β-oxidation, and epinephrine-induced lipolysis in white adipose tissue (WAT) in order to improve fat utilization by muscles and by reducing their dependence on glucose [82, 106]. In vivo, the effects on whole-body glucose usage are potentiated when IL-6 is administrated during low-intensity exercise at doses mimicking those induced by high-intensity exercise [107].

Muscle-derived IL-6 acts as leptin-mimetic as it induces insulin signaling and AMPK activation, and IL-6−/− mice are obese and glucose intolerants [108–110]. Moreover, it mediates the anti-inflammatory and immunomodulatory effects of exercise since it modulates TNF-α and induces IL-1ra and IL-10 [111, 112].

In bone, IL-6 induces bone resorption by both directly through RANKL-dependently enhancing osteoclastogenesis/osteoclast differentiation [113, 114] and indirectly through osteoblast-derived prostaglandin E2 (PGE2)-dependent osteoclast activation [115–117]. In vivo, IL-6 transgenic mice have decreased osteoblasts and increased osteoclasts and develop osteoporosis, while IL-6 knock-out (KO), in a model of arthritis, showed mild-arthritis with impaired osteoclast recruitment at erosion sites [118, 119].

A low-impact high-intensity interval training (HIIT) protocol, applied to 23 recreationally active males, acutely increased serum BAP activity, OPG, and RANKL, as well as inflammatory cytokines (IL-1α, IL-1β, IL-6, and TNFα), while it decreased N-terminal cross-linked telopeptide of type I collagen (NTx). Bone formation was directly correlated with IL-10 and negatively with the inflammatory mediators (and IL-6) [120]. On the contrary, in professional rugby players, a session of intense training did not affect IL-6 [121]. After 4-month training in male and female military recruits, in this case, besides the unchanged inflammatory profile, bone turnover markers were all increased [122]. Ovariectomized (OVX) rats, as model of postmenopausal osteoporosis, after 3 months showed decreased trabecular bone volume (TBV), circulating estradiol and calcitonin, and higher bone-derived IL-1β, IL-6, and cyclooxygenase-2 (Cox-2). On the contrary, OVX rats submitted to treadmill exercise showed lower bony IL-1β, IL-6, and Cox-2 mRNA expression and higher TBV, estradiol, and calcitonin [123]. In 173 elderly obese subjects, the more severe obesity the higher IL-6 and adiponectin (but only in men), OC (in women), C-reactive protein (CRP), and leptin indicating that adiposity is associated with low BMD and pro-inflammatory status [124]. In a comparable population, 32-week combined loading protocol exercise increased muscle strength and BMD at various sites in association with decreased CRP, interferon (IFN-γ), IL-6 (only in men), while TNF-α and bone turnover markers were unchanged [29]. In postmenopausal women, baseline IL-6 concentrations were negatively correlated with BMD and muscle strength [125].

Despite the positive metabolic effects of exercising muscle-derived IL-6, the benefits of exercise on bone are apparently dependent on decreased IL-6 (Fig. 2). Alternatively, a possible reading key resides in the differentiation between chronically high IL-6 and exercise-induced IL-6. In all the cases in which bone status (in terms of BMD/TBV or bone markers) is put in correlation with IL-6 values following a training protocol, the evaluations are performed on resting subjects and, thus, in a situation in which only systemic (an eventually chronically high) IL-6 is seen. Basal IL-6 concentrations (muscular contraction-independent) are always reduced by training, and this is correlated to an improved bone status.

Fig. 2
figure 2

Changes in IL-6 plasma concentration in response to low-to-moderate intensity physical activity, high-intensity physical activity, chronic inflammation, and sedentarism

Full size image

IL-7

Evidence about myokine-like action of IL-7 emerged recently. Myotubes, differentiated from satellite cells, express and secrete IL-7 depending on differentiation degree and inversely to IL-7 receptor (IL-7R). It negatively regulates differentiation by acting on satellite cells [126].

IL-7 is a key cytokine in T and B cells development. It mediates the estrogen deficiency-induced bone loss [127] through the indirect activation of osteoclasts [128] by increasing RANKL and TNF-α producing T cells [127]. In mice, systemic IL-7 administration enhances osteoclast differentiation [129] but it fails to increase bone resorption in nude mice [130]. Furthermore, it promotes survival and differentiation of B220+ cells, a subset of dendritic cells, in osteoclasts [130]. Eleven-week strength training program increased 3-to-fourfold the muscle biopsy expression of IL-7 [126]. As for IL-6, the biological meaning of muscle-derived and immune system-derived IL-7 and, possibly basal-to-peak differences may be the key to understand its role.

IL-8

IL-8 (CXCL-8) is accounted among chemokines (CXC family) as it is actively synthesized by monocytes/macrophages and endothelial cells (ECs) as a chemoattractant to leukocytes and pro-angiogenic factor [131]. However, it is virtually produced by all cell types, under various stimuli, as an early inflammatory mediator, in order to recruit and activate monocytes and neutrophils, at sites of inflammation [131, 132]. Being a chemoattractant, its biological activity depends on its gradient rather than on absolute concentrations [131, 132]. Due to its resistance to enzyme, temperature, and acid-dependent proteolysis, it persists (days/weeks) in the adverse inflamed environment [133].

CXC Chemokines are induced by IL-6: in mice, CXCL-1 (the murine homologue of human IL-8) is induced, in serum, muscle, and liver, after an exercise bout. This response is abolished in IL-6 KO mice, while it is enhanced in IL-6 transgenic mice. Furthermore, muscle IL-8 mRNA was increased after 2.5 h of cycling in well-trained cyclists, although no differences were found in plasma concentrations [134]. On the contrary, strenuous endurance activities, as marathons and ultramarathons, induce marked increases in IL-8, IL-6, and CRP [135, 136] together with increased bone resorption [52].

To date, there are no descriptions of the role of IL-8 in bone metabolism.

LIF

Leukemia inhibitory factor (LIF) is a myokine, belonging to the IL-6 superfamily, with pleiotropic effects ranging from hematopoietic stem cells (HSCs) proliferation, platelets and neural formation, and acute phase reactions. It also exerts effects on bone: enhanced bone turnover and bone formation in vivo, prostaglandin-dependent bone resorption, and osteoblast proliferation in vitro. This dual effect depends on differentiation stage: in osteoprogenitor it stimulates differentiation while in differentiated osteoblasts it inhibits mineralized nodule formation [137]. The effects of LIF also depend on the expression profile of the LIF receptor (LIFR) and its co-receptor gp130. Mild concentric exercise induced increased LIF mRNA expression in skeletal muscle biopsies and gradually recovered during the following 24, while LIF protein was unaffected. Myocytes treatment with ionomycin (a calcium ionophore) resulted in increased LIF mRNA and protein [138]. Exercise-derived LIF paracrinally acts on bone cells, particularly in the periosteum, by modulating their activity [139].

IGF-I

Insulin-like growth factor-I is primarily synthesized in liver, dependent on growth hormone (GH) stimulation. A negative feedback loop exists with circulating IGF-I acting on the pituitary gland to inhibit the GH synthesis and release. However, IGF-I is produced by many extrahepatic tissues, under the control of several hormones, and herein acting in a paracrine/autocrine fashion [140, 141]. GH stimulates the formation of a ternary complex between IGF-I, acid-labile soluble unit (ALS), and IGF binding protein (IGFBP)-3 which stabilizes the circulating IGF-I [140, 141].

The IGF system comprises two ligands (IGF-I, IGF-II), two tyrosine protein kinase receptor (IGF-IR, IGF-IIR), IGFBPs, and IGFBP proteases. They are all mainly involved in the cellular events involved in somatic growth (cell survival, proliferation, and differentiation). The receptors bind with high affinity both to IGF-I and IGF-II, but they are also able to bind, at lower affinity, insulin due to its 50 % homology with IGFs. However, the growth-promoting effects are mainly dependent on IGF-IR activation. This interaction leads, through the sequential activation of phosphatidyl-inositol 3-kinase (PI3K), mitogen-activated protein kinase (MAPK), extracellular signal-activated kinase (Erk)1/2, and the transcription factors elk-1/c-jun, to G1-to-S cell cycle progression, and apoptosis inhibition [142].

GH and IGF-I, and IGF-II are key regulators in skeletal homeostasis (e.g., longitudinal growth, maturation, bone mass acquisition/maintenance), although their relative importance varies during life [141].

IGF-I (more potently than IGF-II) exerts several GH-dependent and -independent actions on bone. In osteoblast, IGF-I expression is under control of PTH and cyclic AMP (cAMP)-inducers [143, 144] as for estrogens and triiodothyronine (T3), but contrarily to glucocorticoids [145–149]. Growth factors, produced locally, modulate IGF-I expression: platelet-derived growth factor (PDGF), and FGF inhibits IGF-I according to their inhibitory effects on osteoblast differentiation [150]. BMP-2, instead, increases IGF-I expression and, hence, it stimulates osteoblast differentiation [151].

In vitro, IGF-I is slightly pro-differentiating in undifferentiated cells while it enhances function of mature osteoblasts [152–154]. It induces collagen type I (and reduces its degradation) and stabilizes β-catenin, making it available for the Wnt signaling, throughout PI3K and insulin-receptor substrate (IRS)-1 [154–156]. Osteoclasts express IGF-IR, and RANKL-induced IGF-I expression is counteracted by GH-mediated osteoprotegerin expression in osteoblasts [157–159]. These opposing actions may explain the modest in vivo effects of IGF-I [141].

IGFBPs act by either sequestering the circulating IGF or when embedded in the ECM, by making it bioavailable. Osteoblasts stage-specifically express different IGFBPs with the aim to reduce IGF-I activity in bone environment [141, 160]. In vivo, IGFBP-1 is involved in glucose homeostasis [161] while IGFBP-2 is linked to BMD and bone turnover rate in humans; although in mice this effect is gender-specific: igfbp-2 −/− increases cortical bone in females and decreases trabecular and cortical bone in males [162]. IGFBP-3 overexpression causes growth retardation and osteopenia [163] while IGFBP-4 and IGFBP-5, which are inhibitors of IGF-I, stimulate osteoblast activity in a IGF-independent manner [164]. IGFBP-6 has unknown effects on bone [141].

Myocytes express IGF-I and in mouse forelimb its expression is mainly localized at muscle–bone interface. IGF-IR is highly expressed in the periosteum adjacent to muscle attachment [165]. Muscle-derived IGF-I affects bone metabolism: in muscle atrophy due to disuse, overexpression of IGF-I and sonic hedgehog (Shh) within the gastrocnemius/soleus attenuates muscle atrophy and osteopenia [166]. IGF-I overexpression in normal mice muscles increases muscle mass and cortical bone [167] while the restriction of IGF-I expression to the liver causes a decrease in TBV [168].

Maximal exercise tests increase IGF-I, IGFBP-3, ALS (only in males), and IGFB-2 (only in females) in response to increased GH, and this was correlated with increased bone resorption marker C-terminal cross-linked telopeptide of type I collagen (ICTP) [169]. One-year WB and resistance training, in postmenopausal women, modestly enhanced BMD without any modification in the IGF-I/II-IGFBP-3 axis, regardless of the assumption of hormone-replacement therapy (HRT) [170]. In young eumenorrheic females and males, the association between IGF-I is associated with BMD in inactive subjects but not in physically active subjects [171, 172]. In adolescent girls, both femoral neck and lumbar spine BMD were not correlated with IGF-I and IGF-I-to-IGFBP-3 ratio in all the disciplines analyzed (track sprinting, swimming, cross-country skiing), but in rhythmic gymnastics [173]. Notably, the association of sarcopenia and osteoporosis, observed during aging, is linked to the physiological IGF-I decline [174]. Strong eccentric efforts rise IGF-I, bone resorption markers (NTx, tartrate-resistant acid phosphatase 5b (TRAP-5b)), and OC [175]. Obesity negatively affects IGF-I response to resistance exercise in both adult and young men [176, 177]. However, in T2D exercise-induced IGF-I partially explains positive effects of PA [178].

Many beneficial effects of exercise are dependent on the anabolic action of IGF-I (on muscle, bone, tendon, cognition, and metabolism). However, the precise role of systemic- versus locally produced IGF-I is not clear. Exercise may enhance local (muscle) IGF-I (possibly associated to a decrease in systemic IGF-I, and this explains the discrepancies evidenced above) and this stimulates bone anabolism [179, 180].

The PGC-1α-irisin system

A recent dispute questioned the existence (and the vaunted biological role) of irisin [181, 182]. However, more recently, the original discoverers of this hormone definitively demonstrated its existence and association with contractile activity [183].

Irisin induces the mitochondrial uncoupling protein (UCP)-1 in the WAT, together with a broad program of fat-like development, allowing it to turn toward brown adipose tissue (BAT). The resulting beige adipose tissue (WAT cells displaying the features of BAT) is more effective in thermogenesis and, hence, in dissipating energy [184, 185]. Irisin is mainly synthesized by muscles in response to endurance exercise, resulting in increased energy expenditure and improved glucose homeostasis and obesity. Irisin is the cleavage product of the membrane protein fibronectin III domain-containing protein (FNDC)5 and it is transcriptionally regulated by peroxisome proliferator-activated receptor (PPAR)-γ coactivator (PGC)-1α which could also be induced by cold [186]. PGC-1α expression, and hence circulating irisin levels, decreases age-dependently [183]. In browning WAT, p38-MAPK and ERK, activated by stress stimuli (i.e., exercise-induced radical oxygen species (ROS)), induce UCP1 expression. cAMP/PKA/MAPK signaling activates PGC-1α which induces FNDC5/irisin. On WAT, muscle irisin induces UCP1 expression and, consequently, of betatrophin. This last increases β-cell mass in mice and it is currently under consideration as a β-cell regenerative therapy [186].

As a myokine, irisin response to exercise depends on intensity, duration, kind of exercise, and training status [187–189]. Irisin peak production is also in function of the exercising muscle mass: women and elderly experience lower absolute serum irisin concentrations; however, the amplitude of exercise-induced increase is comparable to that observed in men [183].

Irisin effects on bone metabolism are still debated [190]. However, recently an association between low irisin levels and previous vertebral fractures in overweight postmenopausal women, but not with BMD and lean mass [191] has been suggested. This association is not affected by 3 month-long treatment with either teriparatide or denosumab (anti-RANKL) [192]. In vitro, irisin enhances osteoblast differentiation through the Wnt-β-catenin signaling and inhibits osteoclast differentiation by suppressing the RANK/RANKL pathway [193]. As described below, the female triad, a peculiar condition emerging in some female athletes and characterized by amenorrhea, anorexia, and osteopenia, is consequent to a chronic energy deficit state [194]. Amenorrheic athletes have low irisin and FGF-21 (also implicated in thermogenesis), compared to eumenorrheic counterparts, while only irisin is associated with BMD [195]. Fasting circulating irisin concentrations correlate with Sost in subjects with IGT (impaired glucose tolerance) or IFG (impaired fasting glucose) and, in men, they are negatively associated with BMI [196].

Recently, Garcia-Fontana and colleagues have demonstrated that irisin and myostatin serum concentrations are related in T2D patients and that they are modulated by triglyceride and glucose levels [197].

Musclin

Osteocrin is expressed in embryonic bone, which disappears soon after birth concomitantly with the first steps of extracellular matrix mineralization. It is negatively regulated by 1,25-dihydroxyvitamin D, while in osteoblast cultures it inhibits mineralization, OC expression, and ALP activity [198]. Its protein sequence is identical to that of a muscle-derived factor, named musclin [199], belonging to the natriuretic-peptide family [200], which is, instead, expressed in adult mouse myocytes depending on the nutritional status. Insulin induced musclin muscular mRNA levels, while cAMP-inducers (i.e., forskolin, epinephrine, isoproterenol), counter-regulators of insulin, caused a reduction. Insulin-resistant obese mice showed a pathological overexpression of musclin [199]. FoxO1, a transcription factor controlling the GlaOC/GluOC balance in osteoblasts [201], inhibits musclin expression in muscles [202]. No published reports described the effects of exercise on musclin expression.

BDNF

Brain-derived neurotrophic factor (BDNF) belongs to the neurotrophins family and is involved in development, differentiation, survival, and functioning of brain cells [203]. Besides neurons, myocytes represent an important source of BDNF, mainly in response to exercise [204], although even in this case, the nervous system consistently contributes to this rise [205]. A meta-analysis on 29 studies comprising three exercise paradigms (single session, single session following a training program, rest after a training program), emphasized the exercise-dependent induction and highlighted a gender-specific response with women expressing lower BDNF [206]. Exercise-dependent rise in BDNF partially accounts for the neuroprotective effects of PA [207]. Eight-week training, however, induces adaptation to exercise, which lowers baseline BDNF concentrations, while detraining abolished exercise-induced BDNF rise [208]. Most interestingly, traffic-related air pollution during exercise may inhibit the BDNF rise [209].

In bone, BDNF and its receptor, TrkB, have been detected at sites of fracture healing being expressed by both osteoblasts and ECs [210], in epiphyseal cartilaginous growth plate by hypertrophic chondrocytes, at sites of intramembranous ossification, and in trabecular bone by active osteoblasts [211]. Importantly, BDNF brain-targeted conditional KO caused a metabolic phenotype characterized by hyperphagia, obesity, leptin resistance, and increased abdominal WAT, with no effects on UCP1. Bone showed increased femur length, high BMD, and BMC. Furthermore, females were most affected than males [205].

The skeletal muscle secretome consists of a huge set of (mainly unknown) molecules: other myokines (e.g., IL-15, osteoglycin), growth factors (FGF-1, FGF-2, BMP-1), effectors (MMP-2), transcription factors (MEF2C), and hormones (apelin, adropin) are affected by exercise [62, 101, 212, 213] and they reasonably have direct or indirect effects on bone but, however, too little is still known.

Table 1 summarizes the effects of exercise-induced release of molecules described above on bone tissue.

Table 1 Effects of exercise-induced myokines and adipokines release on bone tissue
Full size table

Indirect endocrine effects of exercise on bone metabolism: adipose tissue-to-bone axis

Osteoblasts, myofibers, and adipocytes are originated by a common precursor, the MSC [214]. The relative balance between different growth factors, in a specific microenvironment, drives the equilibrium toward a commitment rather than another. Osteoblastogenesis is driven, at expenses of adipocytogenesis, by the expression of core-binding factor alpha-1 (Cbfa-1) and runt-related transcription factor 2 (RUNX2). MyoD drives myocyte differentiation, while PPAR-γ induces adipogenesis. Although under normal conditions adipogenesis is slightly favored, osteoblastogenesis could be enhanced by local, systemic, and environmental (i.e., nutritional) factors [215, 216]. Adipogenesis is favored by chronic low-grade inflammation (ROS, TNF-α, PGE2, leukotriene B4 (LTB4)) which induces PPAR-γ [217]. PPAR-γ suppresses RUNX2 expression while TNF-α stimulates osteoclastogenesis determining bone loss. Chronic inflammation is also associated with muscle loss [218]. In a vicious cycle in which inflammation induces adipogenesis and increased adiposity induces inflammation, the net result is bone loss and, possibly, muscle loss. In this view, osteopenia, sarcopenia, and obesity, either combined or alone, are different presentations of the same pathological condition.

Terminal adipocyte differentiation is the addressing toward white or brown cells. BAT emerges earlier than WAT in fetal life [219]. Although the factors involved in adipocyte fate are still unclear, Wnt and retinoblastoma proteins (pRB, p107, p130) seem to inhibit brown differentiation [220]. Terminally, BAT expresses UCP-1, accounting for its thermogenic role, whereas WAT expresses leptin being deputed in fat storage [221, 222]. A positive energy balance (intake vs. expenditure) leads to excessive adiposity featured either by adipocyte hyperplasia in adults or adipocyte hyperplasia plus hypertrophy in young [223].

WATs from anatomically distinct locations are functionally different. Subcutaneous (sWAT) and visceral (vWAT) fat have, respectively, beneficial or detrimental effects in metabolism [224]: transplantation of sWAT to the vWAT reduces adiposity and improved glucose homeostasis, while the contrary has no effects [225, 226]. It is noteworthy that exercise remodels sWAT, making it even more metabolically active by inducing browning shift through mitochondrial biogenesis and thermogenesis regulators (e.g., UCP1), mainly mediated by FGF-21 [227].

PA shifts the energy usage from storage to consumption (i.e., thermogenesis): endurance training increases muscles reliance on lipids as an energy source [228].

Retired pro-athletes are commonly at increased risk of obesity due to decreased energy expenditure, not associated to reduced intakes, not balanced by a change in fat storage potential. Moreover, decreased muscle-derived myokines exercise-dependent pro-inflammatory mediators reduce lipolysis and appetite suppression [229].

Being BAT absent in adult humans (contrarily to rodents), thermogenesis takes place in skeletal muscle. Energy excess, as in obesity, induces the prevalence of type II muscular fibers, which display a low-efficiency energy usage profile [230].

During exercise, muscle contraction (but also bone remodeling) which requires energy, burns glucose and fatty acids. Muscle-derived IL-6 acts on WAT to induce lipolysis [231] in order to satisfy the energy needs, and pushes the browning transition [232].

In mice 7-month high-fat diet (HFD), compared to regular feeding, doubled total adiposity, halved trabecular bone (−61 %), and reduced B cell number in both marrow and blood. Low-magnitude mechanical signals (0.2 g at 90 Hz, 15 min/day, 5 days/week), in the second half of HFD, restored both bone structure (osteoclast inhibition) and B cells [74]. In obese elderly humans, 12-month diet restriction reduced body weight, femur cross-sectional area, cortical thickness, and BMD, and increased Sost. PA intervention reverted these negative changes [233].

Insulin resistance is also associated with visceral adiposity. Therapeutic strategies aimed to reduce adiposity result in insulin sensitivity improvements.

Besides decreasing WAT mass (decrease in cell size and lipid content), increasing lipolysis, and WAT browning, regular PA modifies the hormonal profile of adipocytes. Adipose tissue synthesizes and secretes many proteins (i.e., TNF-α, IL-6, resistin, visfatin, adiponectin), collectively termed adipokines. Their function ranges from satiety regulation to carbohydrate and lipid metabolism and insulin sensitivity. Exercise-dependent effects on fat storing and insulin resistance are dependent upon a modulation in the adipokines expression profile [234].

These cytokines also profoundly affect bone remodeling and BMD (Table 1). One example of fat-bone connection derives from a pathogenic hypothesis for osteoporosis according to which increased bone resorption depends on increased bone marrow adiposity [235–237].

Leptin

This hormone represents the link between bone and adipose tissue metabolisms at the central nervous system (CNS) level [238]. Leptin is an adiposity signal that suppresses appetite. Both leptin and leptin receptor deficiency (ob−/− and db−/− mice) causes obesity but also higher bone mass [239–241]. Intracerebroventricular leptin infusions reverse the high bone mass phenotype and improve the metabolic phenotype [240–243]. Bone formation is modulated by leptin through sympathetic nervous system (SNS) and the inhibition adrenergic signaling results in a high bone mass uncorrected by intracerebroventricular leptin infusions [244]. Thus leptin/adrenergic circuit-dependent modulation of bone mass is independent on the adipose tissue-derived leptin [238].

Paradoxically, leptin administration to women who have become hypogonadal due to strenuous exercise slightly increases bone mass. However, this improvement depends on leptin-induced estrogen restoration [245].

In lean professional male cyclists, 3-week racing decreased leptin, increased bone resorption markers and GluOC-to-GlaOC ratio [16]. Likewise, in obese young males, 8-week exercise training decreased fat mass and leptin, improved insulin sensitivity, and increased both total OC and GluOC. The authors suggested that leptin and OC were independent predictors of changes in insulin sensitivity throughout a training program [246]. Interestingly, bed rest induced rises in leptin, insulin, and OC independently from resistive vibration exercises [247]. In young obese, metabolic status, BMD, and adiponectin improvements were greater in those who underwent to combined aerobic-resistance training than in those who underwent to either aerobic or resistance protocols [248, 249]. In elderly obese, weight loss intervention accelerated bone turnover. Interventional PA prevents this effect, attenuates hip BMD loss, and greatly decreases leptin levels [250]. Leptin levels are age-dependent but this association is lost when leptin concentrations are adjusted for fat mass [251].

Leptin is detrimental for bone also dependently on its pro-inflammatory action: stimulates neutrophil chemotaxis and phagocytic function, induces pro-inflammatory cytokines in monocytes, and induces T helper (Th)1-type cytokines [252].

Taken together the aforementioned, it seems that exercise benefits on bone metabolism are also exerted through the reduction of circulating leptin.

Adiponectin

According to a recent systematic review, adiponectin is the most relevant adipokine associated (negatively) to BMD and fracture risk, independently from gender and menopausal status [235]. Adiponectin, is inversely associated with body fat, promotes insulin sensitivity, and β-oxidation but it exerts negative effects on bone mass. On the other side, adiponectin is induced by OC which improves glucose tolerance [253, 254]. Anti-inflammatory/immunomodulatory effects of adiponectin are mediated by the inhibition of TNF-α release from macrophage [252].

Actually, conflicting evidences have emerged about the role of adiponectin on bone metabolism [255]. It acts both directly and indirectly on osteoblasts and osteoclasts which express the adiponectin receptor (but also adiponectin itself) [256, 257]. In vitro and in vivo studies suggest that autocrine/paracrine action of adiponectin on osteoblast stimulates their function, while systemic adiponectin inhibits osteoblasts activity and increases osteoclastogenesis [258, 259].

Cycling races increase adiponectin, in association to energy consumption, increased osteoclasts activity, and increased GluOC-to-GlaOC ratio [16]. Moderate to high-intensity regular aerobic exercise enhances adiponectin concentrations [260–262]. However, a single bout of exercise does not affect adiponectin levels in IGT subjects suggesting that a modification in body composition is needed to increase its concentration [263]. Adiponectin levels are associated with BMD in untrained subjects (pre-pubertal girls, elderly woman) but in trained subjects [264, 265]. This suggests that others factors induced by regular PA have a greater impact on bone metabolism (i.e., load) than adiponectin changes.

Acute negative energy balance (i.e., single or multiple bouts of endurance exercise) increases adiponectin [266]. The load, even in this case, plays a key role with NWB activities pushing bone resorption (decrease in total OC and relative increase in GluOC) [16], while WB activity stimulates bone formation (increased total OC) [266].

Visfatin

Visfatin/Nampt, also named PBEF (pre-B cell colony-enhancing factor), is primarily synthesized by vWAT [267]. It exists in an intracellular form, acting as nicotinamide phosphoribosyltransferase (Nampt [EC2.4.2.12]) in NAD biosynthesis, and an extracellular form which is an insulin-mimetic, pro-inflammatory/immuno-modulating adipokine. Its circulating levels, although with some concerns [268], are associated with obesity, insulin resistance, and energy-bone crosstalk [269].

It activates insulin-receptor (IR) signaling in primary osteoblasts and, like insulin, it stimulates their proliferation [270].

However, in vitro, visfatin induces osteoblastic differentiation but inhibits OC expression [271]. It inhibits osteoclastogenesis by suppressing RANK and nuclear factor of activated T cells (NF-AT) pathways [272]. In MSCs it is induced by PPAR-γ activations and drives adipogenesis [273].

In vivo, exercise induces (rats, humans) intracellular Nampt expression within the skeletal muscle, which sustains NAD+ production [274], while it decreases circulating visfatin in healthy [275], obese [276], and T2D patients [277, 278]. However, recently, Lubkowska and colleagues found that, in obese young men, a six-month-long exercise program associated to whole-body cryotherapy (WBC, for review [279]) increased visfatin concentration, while it had no effects on adiponectin, resistin, and leptin [280]. This apparently contrasting finding could depend on the exercise training protocol as well as on the specific anthropometric features of the study population. In children and adults, circulating visfatin levels behave as an inflammatory marker, being associated with fat mass and inflammatory status. Moreover, visfatin concentrations change depending on lifestyle interventions (diet, exercise) [281–284], energy expenditure [285], and insulin sensitivity improvements [286].

No study associates circulating visfatin levels, PA, and bone status, and also indirect associations are often controversial. In master male rowers, despite higher BMD and better metabolic profile, serum visfatin and resistin did not differ from sedentary [287]. Visfatin is an independent BMD predictor in a cohort of Chinese women [288] but not in Chinese men [289] and postmenopausal Iranian women [290]. In 13-to-15 year-aged female athletes, participating in different sports, visfatin was not associated with BMC and BMD [291]. Metabolic syndrome patients had higher plasma visfatin (and low adiponectin) compared to age-matched controls, and it was correlated with lumbar spine BMD only in men [292].

Resistin

Resistin is an adipose tissue-derived hormone early marker of insulin resistance, linking obesity and inflammation [293]. Osteoclasts, osteoblasts, and BMSCs express resistin, and in vitro studies suggest that it may stimulate the overall bone remodeling by stimulating osteoclastogenesis as well as osteoblast proliferation [270, 271]. As for visfatin, in MSCs resistin is induced by PPAR-γ activations and drives adipogenesis [273]. Its (negative) correlation with BMD, in vivo, is debated since there are data suggesting an association in women [288] and in men [294] and other excluding it [289, 295, 296]. In menopause, circulating resistin levels are doubled compared to premenopause [297]. Interestingly, resistin concentrations were positively associated with previous osteoporotic fracture and the correlation was strengthened by the coexistence of diabetes [298]. Resistin levels are not associated with GluOC [299], but are decreased by exercise (aerobic, endurance), depending on type, duration, and intensity [300], and also by weight loss and insulin sensitivity improvements [300, 301].

No studies associate the exercise-dependent changes in resistin with bone metabolism. However, due to the pro-inflammatory/pro-osteoclastogenic effects of resistin, its exercise-induced decrease may be beneficial for bone.

The female triad: an example of homeostatic failure in the female athlete

The female triad is a peculiar pathological condition, frequently observed in endurance athletes, characterized by low energy availability with or without eating disorders, menstrual dysfunctions, and low BMD. An early recognition and intervention is needed, in this situation, to avoid the evolution toward anorexia, amenorrhea, and osteoporosis [302].

The etiological factor is the energy deficiency due to the unbalanced dietary intake compared to the expenditure to support homeostasis (health, daily activities, and growth). Thus, this condition is not a triad of three separate entities but rather a syndrome characterized by different components [194]. Although the literature has focused on females, the energy deficiency and the health consequences also occur in men [303].

The disordered eating (<30 kcal/Kg FFM/day) is pathogenically multifactorial: cultural, familial, individual, and genetic/biochemical factors are involved [304] with a prevalence of 13–20 % in elite female athletes, greatly varying among sports [305]. Furthermore, practicing sports may directly affect diet: performance enhancing diets, need to lose weight, overtraining, injuries, inappropriate coaching [303].

Low energy availability is the main cause of menstrual dysfunctions (primary or secondary amenorrhea) since it alters the equilibrium of several metabolic hormones and substrate like insulin, cortisol, GH, IGF-I, 3,3,5-triiodothyronine, ghrelin, leptin, glucose, fatty acids, and ketones [306].

The consequences of this metabolic unbalance are, on one hand the compromised performance level, and on the other hand health problems such as nutrient deficiency (anemia), chronic fatigue, increased risk of infection and illness, and cardiovascular, gastrointestinal, endocrine, reproductive, renal, and nervous complications [307]. Moreover, muscle protein synthesis is decreased and muscle strength is strongly depressed [308].

This situation affects bone metabolism. Estrogen and progesterone are key regulators of peak bone mass acquisition during growth, and an imbalance, even in subclinical ovulatory dysfunctions, has negative effects on bone maturation [309]. In addition to energy deficiency and PA-dependent increased stress hormones levels (cortisol, catecholamines), amenorrhea further limits the pro-osteogenic effects of exercise [310].

Hypoestrogenism was firstly recognized as the key factor in the low BMD female triad; however, it has been demonstrated that low energy availability is the leading cause of bone metabolism deregulation due to decreased IGF-I and, consequently, the reduced bone formation rate [311]. Bone loss in athletes also seems irreversible [312].

Stress fracture risk is associated with low BMD, dietary insufficiencies, menstrual dysfunction, compulsive exercising, low BMI, and previous fractures [194], conditions that are often encountered in the triad syndrome.

Elevated bone turnover and low BMD, for example, has been associated with low body mass, menstrual dysfunctions, hypoestrogenism, and hypovitaminosis D in female adolescent runners [313].

At a molecular level, the events underlying this condition are still poorly understood, although they are a clear evidence about the existence of an energy-to-bone crosstalk, acting independently from the canonical hormonal pathways, which governs the whole-body homeostasis in the athlete.

Conclusions

Myokines and adipokines are key factors in keeping the homeostasis (Fig. 1). About myokines, the emerging picture is that they are mainly produced by other tissues than skeletal muscle, acting as inflammatory mediators when chronically and constantly elevated. Exercise interventions induce peaks in their production and, hence, provoke a “variation-on-theme” on the chronically inflamed phenotype. Probably, the myokines-mediated beneficial effect of exercise depends on the intermittent exposure to these mediators. This is particularly true in exercise effects on bone. Bone is negatively affected by several chronically elevated inflammatory mediators while, as broadly evidenced in this review, oscillations in their level have a beneficial impact on bone metabolism.

Adipokines, instead, have a still debated role in bone metabolism, and the association between exercise-dependent adipokine changes and bone status has not been properly developed. It seems that adipose tissue-derived inflammatory mediators have detrimental effects on bone, and the beneficial effects of exercise are indirectly due to reduction in their circulating levels rather than a direct action on osteoblasts and/or osteoclasts.