Understanding the Neurophysiological and Molecular Mechanisms of Exercise-Induced Neuroplasticity in Cortical and Descending Motor Pathways: Where Do We Stand?

Exercise is a promising, cost-effective intervention to augment successful aging and neurorehabilitation. Decline of gray and white matter accompanies physiological aging and contributes to motor deficits in older adults. Exercise is believed to reduce atrophy within the motor system and induce neuroplasticity which in turn helps preserve motor function during aging and promote re-learning of motor skills, for example after stroke. To fully exploit the benefits of exercise, it is crucial to gain a greater understanding of the neurophysiological and molecular mechanisms underlying exercise-induced brain changes that prime neuroplasticity and thus contribute to postponing, slowing and ameliorating age- and disease-related impairments in motor function. This knowledge will allow us to develop more effective, personalized exercise protocols that meet individual needs, thereby increasing the utility of exercise strategies in clinical and non-clinical settings. Here, we review findings from studies that investigated neurophysiological and molecular changes associated with acute or long-term exercise in healthy, young adults and in healthy, postmenopausal women.

Exercise, physical activity, and fitness are often used interchangeably but describe different concepts (Caspersen et al., 1985;Hayes et al., 2013). Physical activity is defined as any movement of the body via skeletal muscles expanding energy and includes various activities encompassing leisure (e.g., walking, running, swimming, playing a sport), occupational (e.g., lifting heavy loads, climbing stairs, heavy construction, farming), domestic (e.g., gardening, cleaning, home repair), and transportation (e.g., walking or biking to and from work) (Caspersen et al., 1985). Physical activity is positively correlated to fitness, such that larger amounts of physical activity are associated with greater physical fitness (Caspersen et al., 1985). Exercise differs from physical activity in that the bodily movement generated by skeletal muscles and leading to energy expenditure is planned, structured, repetitive, and aimed to improve or maintain physical fitness (Caspersen et al., 1985). In other words, playing a sport for an hour or two, once every three weeks, is considered physical activity, while playing a sport for two hours, three times per week, for several months in order to improve fitness is considered exercise. Like physical activity, exercise is also positively correlated to fitness (Caspersen et al., 1985). Exercise protocols include moderate-intensity, continuous exercise (endurance-type) and more intense, intermittent exercise. The latter consists of brief bouts of intense exercise (for a total duration of less than 10 min) interspersed with intervals of low-intensity exercise or rest . Both exercise protocols enhance aerobic capacity, promote skeletal muscle adaptations, and reduce markers of disease risks, but intense, intermittent exercise involves a significantly shorter time commitment and lower total exercise volume (Donnelly et al., 2000;Gibala et al., 2006Gibala et al., , 2012Perry et al., 2008;Trapp et al., 2008;Gibala and McGee, 2008;Little et al., 2010Little et al., , 2011bLittle et al., , 2011aHood et al., 2011;Metcalfe et al., 2012;Hottenrott et al., 2012;Currie et al., 2013;Gillen et al., 2013;Iellamo et al., 2013;Cochran et al., 2014;Phillips et al., 2017). Contrary to physical activity and exercise, fitness does not refer to a movement of the body that expands energy, but to health-(i.e., cardiorespiratory endurance, muscular endurance and strength, body composition and flexibility) and skill-related (i.e., agility, balance, coordination, speed, power, reaction time) components, which people achieve and can be measured with specific tests (Caspersen et al., 1985). Cardiorespiratory fitness is of particular interest as it is associated with reduced risk of obesity, type-II diabetes, and cardiovascular diseases (Wei et al., 2000;Carnethon et al., 2003;LaMonte and Blair, 2006;Sui et al., 2008;Myers et al., 2015;Harber et al., 2017;Oktay et al., 2017) as well as increased brain health (Churchill et al., 2002;Barnes et al., 2003;Prakash et al., 2007Prakash et al., , 2011Burns et al., 2008;Honea et al., 2009;Johnson et al., 2012;Scheewe et al., 2013;Hayes et al., 2015). The gold standard for the assessment of cardiorespiratory fitness is graded maximal exercise tests on a treadmill or cycle ergometer which provide measures of oxygen utilization, such as the highest volume of oxygen consumed (VO 2peak ) or the value at which the consumed volume of oxygen plateaus or increases minimally (VO 2max ) (Day et al., 2003;Hayes et al., 2013;Beltz et al., 2016). Alternatively, estimates of the maximal oxygen consumption (VO 2max ) can be obtained using sub-maximal tests, such as the Astrand-Ryhming nomogram or Rockport one-mile walk test, which provide an indirect index of cardiorespiratory fitness ( Astrand and Ryhming, 1954;Cink and Thomas, 1981;Siconolfi et al., 1982;Kline et al., 1987;Fenstermaker et al., 1992;Dolgener et al., 1994;Macsween, 2001;Pober et al., 2002).
To date, it has been demonstrated that exercise creates an optimal environment for neuroplasticity in the primary motor cortex and other brain regions involved in motor control and that exercise-induced neuroplasticity promotes motor learning and function (Ziemann et al., 2006;Kleim and Jones, 2008;Dayan and Cohen, 2011;Rajji et al., 2011;Lehmann et al., 2020;Wanner et al., 2020). The evidence that exercise boosts neuroplasticity is corroborated by reports of enhanced neuroplasticity in response to non-invasive brain stimulation paradigms when the latter are preceded by exercise (Mellow et al., 2020). The ability of exercise-induced plasticity to facilitate motor learning could be valuable in maintaining motor function during aging and enhancing relearning of motor skills post-stroke or in Parkinson's patients who have deficits in motor skill acquisition and retention (Daley and Spinks, 2000;Seidler et al., 2010;Warraich and Kleim, 2010;Mang et al., 2013;Petzinger et al., 2013). However, knowledge gaps remain to be filled (Fig. 2).
First, the neurobiological and motor behavioural effects of exercise have been predominantly investigated in healthy participants, thus it currently needs to be confirmed whether findings extend to older adults and clinical populations. Promisingly, some studies have shown that exercise can promote re- learning of motor skills after stroke (Forrester et al., 2008;Quaney et al., 2009;Mang et al., 2013) and improve motor learning and performance in Parkinson's patients (Fisher et al., 2008;Petzinger et al., 2013;Steib et al., 2018). Further, increases in peripheral brain-derived neurotrophic factor (BDNF) and corticospinal excitability have been found in 6-month poststroke patients following highintensity interval exercise along with a negative correlation between exercise-induced changes in BDNF and intracortical inhibition (Boyne et al., 2019). Taken together, these findings highlight the potential for exercise protocols as intervention strategies in rehabilitation and the importance of obtaining more evidence to confirm and maximize exercise benefits for aged and diseased populations. Second, the biological mediators of exercise effects and exercise-induced motor behavioural gains have been investigated in separate studies, thereby hindering the identification of the mechanistic links between exercise-driven motor benefits and exercise-driven systemic (i.e., structural and functional brain changes), cellular, and molecular changes. Indeed, there is evidence from neuroimaging, TMS, and molecular studies for exercise increasing functional connectivity in somatosensory areas (Rajab et al., 2014), positively impacting gray matter density and white matter microstructure Voss et al., 2013b;Herting et al., 2014;Schlaffke et al., 2014;Svatkova et al., 2015), modulating corticospinal excitability and intracortical motor circuits (Yamaguchi et al., 2012;McDonnell et al., 2013;Singh et al., 2014;Smith et al., 2014;Mooney et al., 2016;Lulic et al., 2017;Stavrinos and Coxon, 2017;Roeh et al., 2018;El-Sayes et al., 2019;MacDonald et al., 2019;Opie and Semmler, 2019;Nicolini et al., 2020), and upregulating molecular markers such as BDNF Knaepen et al., 2010;Huang et al., 2014;Skriver et al., 2014;Dinoff et al., 2017;Nicolini et al., 2020). There is also evidence for exercise improving formation and consolidation of motor memories as well as motor performance (Anshel and Novak, 1989;Roig et al., 2016Roig et al., , 2012Roig et al., , 2013Snigdha et al., 2014;Mang et al., 2014Mang et al., , 2016Statton et al., 2015;Taubert et al., 2015;Snow et al., 2016;Thomas et al., 2016aThomas et al., , 2016bOstadan et al., 2016;Baird et al., 2018;Hu¨bner et al., 2018;Opie and Semmler, 2019;Wanner et al., 2020;Lehmann et al., 2020). However, there is little evidence for how exercise is mechanistically linked to gains in motor learning and performance. It has been shown that greater improvements in motor skill retention are accompanied by larger reductions in neural activity within sensorimotor areas (Dal Maso et al., 2018) and by larger increases in corticospinal excitability  when exercise is performed immediately following motor task practice. Further, Lehmann et al. (2020) found a relationship between enhanced learning of a new motor task and changes in frontotemporal white matter microstructure following two weeks of exercise, while Skriver et al. (2014) reported an association between increases in BDNF and improvements in motor skill acquisition and retention after a single bout of exercise. However, Mang et al. (2014) failed to find a relationship between BDNF increases and both neurophysiological (i.e., enhanced experimentally-induced long-term potentiation-like plasticity) and behavioural (i.e., motor learning) gains observed following acute exercise. It follows that future studies need to obtain a greater understanding of the mechanistic link between exercise and motor improvements, especially in clinical settings. That is to say that exercise-induced, systemic, molecular, and cellular changes (e.g., increases in gray matter density, functional connectivity, corticospinal excitability, BDNF) need to be linked to exercise-induced, motor behavioural changes. Of note, as Stillman et al. (2016) pointed out, socioemotional mechanisms, such as exercise-induced changes in mood, stress levels, and sleep, might also contribute to modulating the beneficial effects of exercise on motor learning and performance, and the extent of their contribution should be further assessed in future research. Lastly, exercise parameters including intensity, type, duration, and frequency are believed to impact exercise efficiency and contribute to interindividual variability in exercise-related systemic, cellular, molecular, and behavioural outcomes (Herold et al., 2019). While optimal duration, type, and frequency of exercise to maximize motor benefits remain to be found, there is evidence supporting that high-intensity exercise leads to the greatest increases in corticospinal excitability, BDNF, and motor learning in both healthy individuals and stroke patients (Ferris et al., 2007;Winter et al., 2007;Schmolesky et al., 2013;Skriver et al., 2014;Thomas et al., 2016b;Nepveu et al., 2017;Boyne et al., 2019;Andrews et al., 2020;Nicolini et al., 2020). Identifying exercise mediators of motor behavioural improvements is key to developing personalized exercise interventions that more strongly and efficiently prime neuroplasticity and thus maximize motor benefits in both healthy and clinical populations.
Here, we review studies that investigated the effects of different exercise protocols on corticospinal excitability, intracortical motor circuits, and peripheral molecular markers in physically active or sedentary, healthy, young adults. We also summarize the findings from a study where we examined the relationships between cardiorespiratory fitness and measures of corticospinal excitability, intracortical motor circuits, primary motor cortex (M1) concentrations of c-aminobutyric acid (GABA) and glutamate, cortical thickness, and white matter microstructure integrity of sensorimotor and frontal areas in healthy, postmenopausal women.

EFFECTS OF ACUTE EXERCISE Corticospinal Excitability
Several studies have investigated whether an acute bout of exercise modulates corticospinal excitability using single-pulse transcranial magnetic stimulation (TMS) to M1 (McDonnell et al., 2013;Singh et al., 2014;Smith et al., 2014;Neva et al., 2017;Stavrinos and Coxon, 2017;MacDonald et al., 2019;Opie and Semmler, 2019;Andrews et al., 2020). Findings are mixed. No changes in cortico-motor output have been repeatedly reported in moderately-to-low physically active participants after different exercise protocols, including low-(57% of age-predicted maximal heart rate, HR max (McDonnell et al., 2013); 40% of heart rate reserve, HRR ) and moderate-intensity (77% HR max (McDonnell et al., 2013); 65-70% HR max (Singh et al., 2014;Neva et al., 2017) ; 50% HRR (Andrews et al., 2020); 80% HRR ), continuous as well as high-intensity (90% HRR), interval exercise (Stavrinos and Coxon, 2017;Andrews et al., 2020). However, increases in cortico-motor excitability have been shown in moderately-to-highly physically active, healthy, young subjects after 20 min of moderate-intensity (40% and 50% HRR), but not low-intensity (30% HRR), continuous cycling (MacDonald et al., 2019). Also, similar results have been observed in individuals, whose fitness levels were not reported, after 30 min of both highintensity (77% HRR), interval and low-intensity (50% HRR), continuous exercise, although the increase was smaller after the latter (Opie and Semmler, 2019). Consistently, our findings demonstrate that a single bout of moderate-intensity (50-70% HR max ), continuous exercise enhances corticospinal output in highly, but not low, physically active individuals (Lulic et al., 2017) (Table 1). Interestingly, the increase in corticomotor excitability following acute exercise seen in highly fit, young adults appears to be independent of either biological sex or menstrual cycle phase (El-Sayes et al., 2019) (Table 1). Further, our work suggests that the relative intensity of the exercise stimulus is an important determinant for exercise-induced modulation of corticospinal pathway plasticity in a sedentary population. Namely, we have shown that a single session of intense, interval exercise potentiates cortico-motor excitability in sedentary individuals at a workload of 105-125% peak power (W peak ; determined during a peak oxygen uptake, VO 2peak , test) , but not 69% W peak (El-Sayes et al., 2020) (Table 1). To summarize, while acute, continuous exercise performed at low to moderate intensities (40-80% HR max or HRR) elicits no changes in cortico-motor output in moderately-to-low physically active participants, it increases cortico-motor output in more trained individuals with higher fitness. A strong exercise stimulus intensity (105-125% W peak ) is necessary to enhance corticospinal excitability acutely in a sedentary population.

Intracortical Circuits
Paired-pulse transcranial magnetic stimulation (TMS) paradigms have been used to noninvasively assess whether a single session of exercise induces neuroplasticity within M1 glutamatergic and GABAergic circuits including short-interval intracortical facilitation (SICF), short-interval intracortical inhibition (SICI), longinterval intracortical inhibition (LICI) and intracortical facilitation (ICF). Decreased GABA A -mediated inhibition (i.e., short-interval intracortical inhibition) has been found in individuals of unknown fitness levels after intense, interval (77% HRR), but not low-intensity (50% HRR), continuous exercise (Opie and Semmler, 2019). Similarly, reduced short-interval intracortical inhibition has been observed in moderately-to-low physically active participants 30 min after moderate-intensity (65-70% HR max ), continuous exercise, 10 min after intense, interval exercise (90% HRR) (Stavrinos and Coxon, 2017), immediately and 15 min after moderate-high-intensity (80% HRR), continuous exercise , and 15 min after low-moderate-intensity (40% HRR), continuous exercise . Despite these reports of an acute exercise-induced reduction in inhibition, Andrews et al. (2020) failed to observe a significant decrease in short-interval intracortical inhibition in subjects with moderate levels of physical activity 10 min following intense, interval (90% HRR) and moderateintensity (50% HRR), continuous exercise. Mooney et al. (2016) also reported no changes in short-interval intracortical inhibition between 10 and 50 min after 30 min of moderate-intensity (73% HRR), continuous cycling in highly-to-low physically active, young adults. These authors, however, showed a decrease in GABA Bmediated inhibition (i.e., long-interval intracortical inhibition) up to 20 min after exercise (Mooney et al., 2016). The reduction in long-interval intracortical inhibition has not been confirmed in moderately physically active individuals either after moderate-intensity (65-70 HR max (Singh et al., 2014); 50% HRR (Andrews et al., 2020)), continuous exercise, or intense, interval exercise (90% HRR) (Stavrinos and Coxon, 2017;Andrews et al., 2020). Findings from studies examining the effects of acute exercise on glutamatergic, intracortical circuits show increased intracortical facilitation after 20 min of moderate-intensity (65-70 HR max ) cycling in moderately physically active young adults (Singh et al., 2014;Neva et al., 2017) as well as after 30 min of light (48% HRR) cycling in similarly aged participants whose physical activity levels are not reported (Morris et al., 2020). However, no significant changes in intracortical facilitation have been observed in individuals aged 21-64 years and with moderate levels of physical activity following either intense, interval (90% HRR) or moderate-intensity (50% HRR), continuous exercise (Andrews et al., 2020). We found reduced intracortical facilitation and GABA Amediated inhibition as well as unchanged short-interval intracortical facilitation in both highly and low physically active subjects 10 min after a 20-min session of moderate-intensity (50-70% HR max ) exercise (Table1) (Lulic et al., 2017). A similar exercise protocol also resulted in decreased short-interval intracortical inhibition in both females and males with high cardiorespiratory fitness levels determined during a peak oxygen uptake (VO 2peak ) test (El-Sayes et al., 2019), while an acute bout of intense, interval exercise (105-125% W peak ) had no effect on short-interval intracortical inhibition and intracortical facilitation in sedentary males  (Table1). In summary, while a number of studies have reported that acute exercise, performed at moderate to high intensities (65-90% HR max or HRR), reduces GABA A -mediated inhibition in highly to low physically active participants, others have shown that similar exercise protocols do not elicit a change in short-interval intracortical inhibition in this population. Contrasting results have also been reported for GABA B -mediated inhibition (i.e., long-interval intracortical inhibition) and glutamatergic intracortical circuits (i.e., short-interval intracortical facilitation and intracortical facilitation). Thus, it is currently unclear whether acute exercise influences GABAergic and glutamatergic circuits within the primary motor cortex. Lastly, acute, intense, intermittent exercise does not elicit changes in either GABA A -mediated inhibition (i.e., short-interval intracortical inhibition) or intracortical facilitation in the sedentary, male population.

Molecular Markers
The molecular mechanisms mediating exercise-induced M1 neuroplasticity largely remain to be elucidated. The neurotrophin brain-derived neurotrophic factor (BDNF), known to promote the positive effects of exercise on learning and memory (Go´mez-Pinilla et al., 2002;Cotman et al., 2007;Erickson et al., 2011;Bechara and Kelly, 2013), is a likely mediator of exercise-induced M1 plasticity. In humans, increases in peripheral BDNF have been reported after various acute exercise protocols including graded maximal exercise, continuous exercise performed at moderate to high intensities (60-80% HR max or HRR; 55-75% of maximal power output determined during a maximal oxygen uptake test) and intermittent exercise at higher intensities (90% of maximal power output) Ferris et al., 2007;Tang et al., 2008;Goekint et al., 2008;Bos et al., 2011;Cho et al., 2012;Heyman et al., 2012;Schmolesky et al., 2013;Mang et al., 2014;Skriver et al., 2014;. At present, however, a relationship between acute exercise-induced increases in BDNF and changes in TMS-probed intracortical and corticospinal motor networks has yet to be established. Mang et al. (2014) found no correlation between increases in serum BDNF and paired-associative-stimulation-induced corticospinal excitability after acute, intense, interval exercise. In line with these findings, we showed that increases in circulating BDNF were not statistically related with increases in corticospinal excitability in sedentary, healthy, young males following a single bout of intense, interval exercise (Table 1) . A potential confounding factor in these studies is that changes in peripheral BDNF levels might not reflect those occurring in the brain, although some evidence supports that they do (Pan et al., 1998;Karege et al., 2002;Rasmussen et al., 2009;Seifert et al., 2010;Klein et al., 2011). Thus, at present, it remains to be determined whether BDNF contributes to mediating exercise-induced plasticity in M1. Further investigation of the proteins that are believed to facilitate BDNF increases in response to exercise is needed and includes the growth factor insulin-like growth factor 1 (IGF-1) (Ding et al., 2006), the bonederived hormone osteocalcin (OCN) (Khrimian et al., 2017;Kosmidis et al., 2018). the myokines cathepsin B (CTSB) (Moon et al., 2016) and irisin (Wrann et al., 2013;Belviranli et al., 2016;Lourenco et al., 2019). These molecules are all elevated both peripherally and endoge-nously in the brain following exercise and all cross the blood brain barrier, but whether their peripheral or central versions are responsible for exercise-induced increases in brain BDNF is unclear. Animal studies have shown that IGF-1 promotes exercise-driven neurogenesis (Trejo et al., 2001) and that blocking IGF-1 receptors abolishes exercise-induced increases in hippocampal BDNF expression (Ding et al., 2006). Peripheral delivery of the irisin precursor skeletal muscle membrane protein fibronectin type II domain-containing protein 5 (FNDC5) increases circulating irisin and hippocampal expression of BDNF in mice (Wrann et al., 2013), while recombinant cathepsin B enhances BDNF expression in cultured, hippocampal progenitor cells. (Moon et al., 2016). We reported no changes in systemic levels of IGF-1, cathepsin B or irisin in sedentary males 30 min after acute, intense, intermittent exercise (Table1) . However, there was a positive relationship between the percent change in BDNF and the percent change in irisin after intense, interval exercise (Table 1) , indicating that greater increases in BDNF are associated with larger increases in irisin. This suggests that circulating irisin might contribute to promoting exercise-induced BDNF increases which would, in turn, facilitate neuroplasticity. The lack of significant changes in the relative levels of IGF-1, cathepsin B and irisin could be due to the 30-min time point used in our study, as a rise in these markers might be evident immediately after acute exercise and return to baseline by 30 min after the end of the exercise bout (Cappon et al., 1994;Schwarz et al., 1996;Oliff et al., 1998;Skriver et al., 2014;Lo¨ffler et al., 2015;Ma´derova´et al., 2019). A time-course study could rule out whether these markers are elevated at earlier or later times following exercise. Lastly, osteocalcin has increasingly drawn attention for its potential role in exercise-induced BDNF release. It is secreted by osteoblasts and found in the blood in carboxylated as well as partially or fully decarboxylated forms (Zoch et al., 2016). Uncarboxylated OCN is the active form, which crosses the blood brain barrier and enhances BDNF mRNA and protein expression as well as trafficking of BDNF-containing vesicles to synapses (Khrimian et al., 2017). Mera et al. (2016) have proposed that exercise promotes the production of active osteocalcin (i.e., uncarboxylated osteocalcin) by increasing skeletal muscle secretion of interleukin-6, which, in turn, facilitates osteoblast-driven bone resorption occurring at an acidic pH, ideal for osteocalcin decarboxylation (Ferron et al., 2010). We showed increased uncarboxylated osteocalcin (expressed as a ratio to total intact osteocalcin) and decreased carboxylated osteocalcin in sedentary, healthy males following intense, intermittent exercise (Table 1) , suggesting that this exercise protocol increases osteocalcin decarboxylation and activation (i.e., formation of uncarboxylated osteocalcin). We also found that larger increases in BDNF were associated with greater increases in uncarboxylated osteocalcin and decreases in carboxylated osteocalcin (Table 1)  , suggesting that uncarboxylated osteocalcin, like irisin, might contribute to BDNF release in response to exercise, and ultimately to changes in neuroplasticity.

Considerations and Gaps
To summarize, moderate-intensity, continuous exercise can increase corticospinal excitability only in moderately-to-highly physically active (Lulic et al., 2017;MacDonald et al., 2019) or fit (El-Sayes et al., 2019), young adults but not in those with moderate-to-low levels of physical activity (McDonnell et al., 2013;Singh et al., 2014;Smith et al., 2014;Andrews et al., 2020) or fitness (El-Sayes et al., 2020). Promisingly, more intense, intermittent exercise appears to enhance cortico-motor output in the latter population , although not across all studies employing a similar, but nevertheless less intense, exercise protocol (Stavrinos and Coxon, 2017;Andrews et al., 2020;El-Sayes et al., 2020). Indeed, the intensity of the exercise stimulus might have contributed to the discrepancy in outcomes among these studies and might represent a key determinant of the changes in neurophysiological and molecular markers which are thought to drive exercise-induced motor neuroplasticity. Reports of high-, rather than moderateintensity, exercise increasing BDNF release (Ferris et al., 2007;Winter et al., 2007;Schmolesky et al., 2013;, motor learning (Thomas et al., 2016b), and M1 long-term potentiationlike plasticity in response to a type of non-invasive brain stimulation (intermittent theta burst stimulation) (Andrews et al., 2020) provide support for this hypothesis. Thus, it might be important to consider exercise intensity in attempts to prime plasticity within descending and intracortical motor pathways in sedentary, healthy males. Whether this extends to sedentary, healthy females, older adults or individuals with clinical conditions needs to be assessed. Other factors that might contribute to the differences in study results include the range of ages of recruited participants and of TMS pulse intensities delivered to acquire motor-evoked potential (MEP) recruitment curves for the assessment of cortico-motor excitability. Specifically, studies that have failed to observe an increase in corticomotor output recruited individuals with a wide range of ages, i.e., 18-60 years (McDonnell et al., 2013;Smith et al., 2014;Stavrinos and Coxon, 2017;Andrews et al., 2020). It is possible that age impacts motor plasticity responses to exercise in that, for example, a stronger exercise stimulus or multiple sessions might be required to observe enhanced corticospinal excitability in older adults and thus a wide age spread in a cohort of participants might impede the detection of exercise-induced changes. Further, studies that reported no changes in corticomotor output following acute exercise in sedentary and moderately active subjects used TMS pulse intensities ranging from 100% to 140% of resting motor threshold (RMT) to build stimulus-response curves (Singh et al., 2014;Smith et al., 2014;Neva et al., 2017;Stavrinos and Coxon, 2017;El-Sayes et al., 2020), while we delivered TMS pulses at intensities between 90% and 200% RMT and observed an increase in the excitability of descending motor pathways in a similar cohort . This suggests that, to detect exercise-related changes in MEP amplitude in a sedentary population, it might be necessary to extend the range of TMS pulse intensities delivered to generate MEP recruitment curves so not just to capture the ascending portion, but also the plateau of the curve at higher intensities (i.e., 140-200% RMT). Lastly, questions have been raised on interindividual and intraindividual response variability impacting the validity of TMS findings. With regard to this matter, Chipchase et al. (2012), and more recently Pellegrini et al. (2020), have highlighted the importance of increasing the homogeneity of technical and methodological factors (e.g., TMS device characteristics, coil size and orientation, TMS pulse parameters) as well as participant-specific ones. The latter includes selection criteria (e.g., sex and menstrual cycle phase, age, medical and medication history, handedness, specialized hand use such as playing an instrument (Ziemann et al., 1996;Priori et al., 1999;Smith et al., 2002;Nordstrom and Butler, 2002;Ziemann, 2004;Hammond et al., 2004;Inghilleri et al., 2004;Rosenkranz et al., 2007;Nitsche et al., 2008;Rossi et al., 2009;Fujiyama et al., 2014;Heise et al., 2014;Zoghi et al., 2015;Goodwill et al., 2015;Ansdell et al., 2019)), preparation prior to the experimental session (e.g., alcohol consumption, hours of sleep (Civardi et al., 2001;Scalise et al., 2006;Conte et al., 2008;Kreuzer et al., 2011;Nardone et al., 2012;Placidi et al., 2013)), scheduling (e.g., time of the day, days between sessions (Nitsche and Paulus, 2001;Nitsche et al., 2003Nitsche et al., , 2008Sale et al., 2007Sale et al., , 2008), and instructing of participants throughout the testing session (e.g., head and neck posture, muscle activity during the testing session, attention level (Fujiwara et al., 2009); (Lazarski et al., 2002;Andersen et al., 2003;Stefan et al., 2004;Fujiwara et al., 2009;Kotan et al., 2015;Kuhn et al., 2018)). Controlling for these factors is crucial to minimize response variability and thus to use TMS to reliably evaluate changes in corticospinal excitability. Consistency and rigor in reporting methodological and participant-specific factors will also allow a more reliable interpretation and comparison of TMS findings across studies.
The molecular mechanisms mediating motor plasticity after acute exercise are still largely unknown. We have recently demonstrated that, aside from enhancing serum levels of BDNF, intense, intermittent exercise alters the carboxylation state of osteocalcin in sedentary, healthy, young males . Further, greater increases in BDNF are accompanied by larger increases in uncarboxylated osteocalcin and irisin after intense, intermittent exercise in the same cohort . These results suggest that active osteocalcin and irisin might be mediators of BDNF release induced by acute exercise which, in turn, facilitates neuroplasticity. It remains to be tested whether these findings can be extended to sedentary, healthy, female or normally aging or clinical populations. In addition, relationships between motor performance and exercise-induced neurophysiological and molecular changes remain to be established. Indeed, one of the questions that remains unanswered is whether changes in molecular markers and cortical and descending motor pathways in response to exercise are associated with improved motor performance in the aforementioned populations.
Genetic factors are believed to contribute to betweensubject variability, impact individual responses to exercise interventions, and modulate their efficacy in inducing M1 plasticity and motor performance improvements in older adults and during rehabilitation. A single nucleotide polymorphism (Val66Met; rs6265) at codon 66 of the BDNF gene, causing a valine-to-methionine substitution and resulting in reduced activity-dependent BDNF secretion (Egan et al., 2003), has drawn increasing attention. The BDNF Val66Met polymorphism has been shown to attenuate M1 plasticity responses to a facilitatory repetitive TMS paradigm (i.e., intermittent theta burst stimulation) following intense interval exercise (Andrews et al., 2020), that is, Val66Met appears to reduce the priming effects of intense, interval exercise on motor plasticity. Conversely, McDonnell et al. (2013) reported that, although Met carriers had lower baseline BDNF levels than Val/Val carriers, the BDNF genotype did not influence M1 plasticity responses to continuous theta burst stimulation, an inhibitory repetitive TMS paradigm. Of note, these authors found that moderate-intensity exercise had no effect on MEP amplitudes before and following continuous theta burst stimulation (McDonnell et al., 2013). Further, Singh et al. (2014) showed that the response of corticospinal excitability and intracortical motor circuits (i.e., intracortical facilitation, short-interval intracortical inhibition and long-interval intracortical inhibition) to exercise was not affected by this BDNF polymorphism. However, the sample size was small (n = 6, Met carriers; n = 6, Val/Val), and two trends were evident (Singh et al., 2014). The reduction in short-interval intracortical inhibition (i.e., GABA A -mediated inhibition) was stronger and there was no change in long-interval intracortical inhibition (i.e., GABA A -and GABA B -mediated inhibition) in Met carriers compared to Val homozygotes after exercise (Singh et al., 2014). This evidence suggests that exercise might modulate GABA A -and GABA B -mediated inhibition differently in Val/Val individuals versus Met carriers. Nonetheless, it remains unclear at present whether plasticity responses to exercise within the motor system in young, aging, or clinical populations are influenced by the Val66Met polymorphism. This should be examined in future studies, as it might hinder the assessment of exercise priming effects on motor plasticity by masking exercise-induced changes in neurophysiological, molecular, and behavioural measures. Findings from this research will indicate whether genetic variants, particularly BDNF genotype, should be considered when designing exercise strategies aimed to maximize the priming of neuroplasticity, e.g., during normal aging or rehabilitation.
Lastly, it should be acknowledged that, across studies, the intensity of the exercise stimulus is differently controlled and tailored to each participant based on age-predicted maximal heart rate (HR max ; i.e., 220-age), heart rate reserve (HRR; i.e., HR max -resting heart rate) or peak workload (W peak ; determined during a peak oxygen uptake test). This hinders comparison of results, and future studies should assess how controlling exercise intensity by H max , HRR or W peak affects exercise-induced motor plasticity.

EFFECTS OF EXERCISE TRAINING Corticospinal Excitability and Intracortical Circuits
While several studies (reviewed above) have examined the priming effects of acute exercise on motor plasticity using single-and paired-pulse TMS paradigms, only two investigations have, at present, considered traininginduced responses, i.e., the impact of more long-term exercise interventions. One study assessed whether six weeks of high-intensity, interval training (HIIT) induced changes in corticospinal excitability, intracortical GABA A -mediated inhibition, and glutamatergic facilitation in young, healthy, sedentary males . The other investigated the relationships between cardiorespiratory fitness (determined by estimating maximal oxygen consumption, VO 2max , using a submaximal test) and neurophysiological measures (i.e., MEP recruitment curves, short-and long-latency afferent inhibition (SAI, LAI), interhemispheric (SIHI, LIHI) inhibition, short-interval intracortical inhibition, and intracortical facilitation) in healthy, postmenopausal women   (Table 2). Specifically, Nicolini et al. (2019) showed that eighteen bouts of intense, intermittent exercise over six weeks did not alter cortico-motor excitability or short-interval intracortical inhibition, but reduced intracortical facilitation ( Table 2). The reduction in intracortical facilitation induced by long-term, intense, interval exercise might enhance the propensity for acute plasticity within M1 by facilitating the reduction in GABA A -mediated inhibition in response to an acute exercise bout. Interestingly, a relationship between percent change in cardiorespiratory fitness and percent change in short-interval intracortical inhibition was seen in Val homozygotes after 6-week HIIT, such that greater increases in cardiorespiratory fitness were associated with larger increases in GABA Amediated inhibition (Table 2)  . This finding is consistent with previous reports of changes in peripheral BDNF, hippocampal and temporal lobe volumes, and episodic memory only in Val/Val individuals in response to a 16-week multimodal exercise program or habitual physical activity (Brown et al., 2014;Canivet et al., 2015;Nascimento et al., 2015). It also suggests that BDNF genotype might modulate the effects of long-term exercise on GABA A -mediated inhibition in sedentary males. Lastly, fitness does not impact motor plasticity in healthy postmenopausal women, as no relationships between cardiorespiratory fitness and TMS-probed corticospinal excitability or intracortical motor circuits have been found (Table 2)  . It is, however, possible, and remains to be tested, that high levels of cardiorespiratory fitness prime plasticity responses to acute exercise within the motor system.

Molecular Markers
Acute exercise increases peripheral BDNF levels Ferris et al., 2007;Goekint et al., 2008;Tang et al., 2008;Cho et al., 2012;Heyman et al., 2012;Mang et al., 2013;Skriver et al., 2014;Nicolini et al., 2020). In contrast, most studies have found no effect of exercise training on circulating BDNF (Schiffer et al., 2009;Erickson et al., 2011;Ruscheweyh et al., 2011;Voss et al., 2013a;Maass et al., 2016;Goldfield et al., , 2019Gourgouvelis et al., 2018), although there are exceptions. Zoladz et al. (2008) reported increased plasma BDNF following five weeks of endurance training in physically active males, whereas Leckie et al. (2014) found elevated serum BDNF only in subjects older than 65 years following one year of moderate-intensity walking. Additionally, Heisz et al. (2017), while observing no group differences in serum BDNF between participants who underwent sixweeks of HIIT and those who did not, showed that individuals with greater increases in cardiorespiratory fitness had larger increases in BDNF. Consistent with the majority of reports, we found that six weeks of HIIT had no effect on serum BDNF in sedentary males (Table 2) . We also showed that this exercise protocol did not affect serum levels of IGF-1 or cathepsin B in the same cohort (Table 2)  . Two other studies have assessed the effects of long-term exercise on peripheral cathepsin B in low-active, young adults. One reported an increase after four months of supervised treadmill training (Moon et al., 2016), while the other found no changes following eight weeks of a combination of resistance and moderate-to-vigorous-intensity aerobic exercise (Gourgouvelis et al., 2018). Our findings together with Gourgouvelis et al. (2018) seem to indicate that cathepsin B might not be involved in mediating the effects of long-term exercise. Nonetheless, we found that increases in cardiorespiratory fitness were accompanied by decreases in total and precursor cathepsin B (Table 2) , suggesting that greater aerobic capacity might require more mature cathepsin B, which is the enzymatically active form produced from cleavage of the inactive precursor (Mach et al., 1994;Mort and Buttle, 1997;Hook et al., 2015). Therefore, despite no changes in total and precursor cathepsin B having been detected after long-term exercise in low-active, young adults, it cannot be ruled out at present that the mature, active form of cathepsin B might contribute to mediating neuroplasticity responses to long-term exercise.
Neurotransmitter Concentrations, Cortical Thickness, and White Matter Microstructure M1 neurotransmission can be probed using magnetic resonance spectroscopy, which allows the quantification of neurotransmitter concentrations, such as the most prevalent inhibitory (GABA) and excitatory (glutamate) neurotransmitters (Levy et al., 2002;Novotny et al., 2003;Floyer-Lea et al., 2006;Singh et al., 2009;Stagg et al., 2011;Stagg, 2014). Further, cortical thickness and integrity of white matter microstructure within the motor system can be assessed from anatomical magnetic resonance images (Fischl and Dale, 2000) and via diffusion tensor imaging (Alexander et al., 2007), respectively. To date, few studies have used these tools to assess how long-term exercise modulates motor plasticity in young, aging, or clinical populations. Recently, it has been shown that myelination within the M1 area containing the motor outputs to the legs is increased in older adults (>65 years) after twelve weeks of cycling at 64% VO 2max (Rowley et al., 2020). Further, in older adults, higher cardiorespiratory fitness has been associated with greater white matter integrity in the cingulum (Marks et al., 2011;Oberlin et al., 2016;Chen et al., 2020), cerebral peduncle (Chen et al., 2020), anterior corona radiata (Oberlin et al., 2016), anterior internal capsule (Oberlin et al., 2016), fornix (Oberlin et al., 2016), and corpus callosum (Johnson et al., 2012;Oberlin et al., 2016). Interestingly, Johnson et al. (2012) found that the portions of the corpus callosum with greater fitness-related integrity are those connecting premotor cortex and frontal regions, which are involved in high-level motor planning. Tseng et al. (2013) further supported the notion that physical activity attenuates age-related decline in white matter integrity by showing that life-long, aerobic exercise training in older Masters athletes was associated with greater white matter microstructural integrity in the cingulum, posterior thalamic radiation, inferior and superior longitudinal fasciculi as well as in the non-dominant superior corona radiata, superior longitudinal fasciculus, and inferior fronto-occipital fasciculus. In line with these findings, we found that higher cardiorespiratory fitness levels were linked to greater white matter microstructural integrity in dorsal pre-motor cortex, primary somatosensory cortex, and cingulum (Table 2)  in postmenopausal women, a population that is at a greater risk of age-related injuries than men (Peel et al., 2002;Stevens and Sogolow, 2005;Schiller et al., 2007;Hartholt et al., 2011;Canada Public Health Agency of Canada, 2014;Do et al., 2015;Johansson et al., 2016). These results suggest that high levels of cardiorespiratory fitness preserve white matter microstructure in these areas and thus, ultimately, maintain motor function (e.g., faster reaction times, better motor coordination, greater tactile acuity) (Kantak et al., 2012;Borich et al., 2015) and memory (Kantarci et al., 2011) in this population. In addition to being effective in counteracting white matter degradation and motor performance decline associated with aging, high cardiorespiratory fitness seems to mitigate age-related reductions in cortical thickness. Lee et al. (2016) showed a positive relationship between self-reported physical activity and cortical thickness in healthy, older adults, where longer exercise duration (1 h/day) was linked to greater cortical thickness, both overall and in frontal regions. Of note, subjectively measured physical activity levels do not reflect objectively quantified cardiorespiratory fitness (i.e., determined using a VO 2peak test) (Tager et al., 1998). Williams et al. (2017) reported a significant positive relationship between cardiorespiratory fitness and cortical thickness in older, but not in young, adults, such that higher cardiorespiratory fitness was linked to thicker cortex, with the strongest effects in the precentral (M1), paracentral, parahippocampal, and supramarginal gyri as well as in the nondominant orbitofrontal and middle temporal cortices. These authors also found that highly fit, older adults had greater cortical thickness than low fit, older adults in the same brain areas, which are also the most vulnerable to age-related atrophy. Additionally, cortical thickness of some brain regions did not differ in highly fit, older versus young adults, suggesting that cardiorespiratory fitness attenuates age-related cortical decline in older adults, particularly in the precentral gyrus (M1), pars triangularis, and non-dominant precuneus (Williams et al., 2017). Further, Jonasson et al. (2017) showed that cardiorespiratory fitness was linked to greater cortical thickness in the dorsolateral prefrontal cortex in older adults, while Wood et al. (2016) reported that old Masters athletes, who started competitive aerobic training early in life, sustained it for over 30 years, and had high levels of cardiorespiratory fitness, exhibited greater cortical thickness throughout a wide range of cortical areas, especially in the preand postcentral gyri, medial prefrontal cortex, and insula, compared to age-matched, healthy individuals (Wood et al., 2016). Contrary to these studies, we found no significant relationship between cardiorespiratory fitness and cortical thickness in sensorimotor cortex (i.e., primary somatosensory, primary motor, pre-motor cortices, and the supplementary motor area) or frontal regions (i.e., superior, middle, inferior, and orbitofrontal gyri) in postmenopausal women (Table 2) . It should, however, be investigated further whether longterm exercise training is protective against age-related cortical thinning in aging women. Lastly, while there is one report of increased GABA but not glutamateglutamine-glutathione concentration in sensorimotor cortex after a single bout of intense, interval exercise (Coxon et al., 2018), no study has assessed the effects of cardiorespiratory fitness, or long-term exercise training, on GABA and glutamate levels within M1. We were the first to report that cardiorespiratory fitness appears to have no effect on M1 GABA and glutamate levels in aging women (Table 2) . In summary, long-term exercise training counteracts white matter degradation, reductions in cortical thickness, and decline in motor function associated with aging. Fitness, however, does not seem to impact M1 GABA and glutamate concentrations or to preserve cortical thickness within sensorimotor and frontal cortices in postmenopausal women.

Considerations and Gaps
No baseline changes in neurophysiological and molecular measures (aside from a reduction in intracortical facilitation) were evident at the end of a 6-week HIIT protocol. It is, however, possible that this exercise intervention might prime the motor system for acute plasticity, that is sedentary males who perform six weeks of HIIT might show changes in the tested measures immediately after a single bout of exercise as opposed to those who do not. It is also possible that longer exercise interventions are required to facilitate M1 plasticity in a sedentary population. Further, given the paucity of studies that have assessed the priming effects of long-term exercise on intracortical and descending motor networks, it is unclear and should be established in future work whether the findings collected so far can be extended to sedentary female, moderately-to-highly fit, normally aging, or clinical populations. Additionally, despite failing to observe a relationship between cardiorespiratory fitness and M1 GABA or glutamate levels in postmenopausal women, high levels of cardiorespiratory fitness might result in greater GABA and/or glutamate within M1 following acute exercise. Lastly, it should be assessed whether a long-term exercise intervention can increase M1 neurotransmission, cortical thickness, and white matter microstructural integrity in sedentary, aging women. Evidence from such studies is crucial to optimize individualized exercise programs for maximal priming of plasticity.
Exercise is a relatively low-cost intervention for maintaining or enhancing motor function during normal aging (Rikli and Edwards, 1991;Buckwalter, 1997;Visser et al., 2002;Brach et al., 2003;Means et al., 2005;Buchman et al., 2007) as well as for ameliorating motor performance deficits associated with clinical conditions (Bergen et al., 2002;Crizzle and Newhouse, 2006;Herman et al., 2007;Forrester et al., 2008;Gobbi et al., 2009;Cooke et al., 2010;Alberts et al., 2011;Vreugdenhil et al., 2012;Pitka¨la¨et al., 2013;Coelho et al., 2013;De Andrade et al., 2013;Sobol et al., 2016;Linder et al., 2017;Rosenfeldt et al., 2019). Mechanisms at different levels (i.e., molecular, cellular, systemic, and socioemotional) modulate exercise benefits on motor learning and function (Stillman et al., 2016). Particularly important mediators of exercise positive effects on motor behaviour are increases in corticospinal excitability and BDNF, both established modulators of neuroplasticity (Ziemann et al., 2001;Siebner and Rothwell, 2003;Siebner et al., 2004;Bramham and Messaoudi, 2005;Thickbroom et al., 2006;Kleim et al., 2006;Bekinschtein et al., 2008;Yoshii and Constantine-Paton, 2010;Fritsch et al., 2010;Carson and Kennedy, 2013;Tunovic et al., 2014;Lu et al., 2014;Leal et al., 2017;Kowian´ski et al., 2018), a mechanism by which our brain learns motor behaviour (Rioult-Pedotti et al., 2000;Muellbacher et al., 2002;Doyon and Benali, 2005;Monfils et al., 2005;McConnell et al., 2009;Dayan and Cohen, 2011;Cantarero et al., 2013). Attention should also be paid to osteocalcin, a molecular marker that is likely to contribute to modulating the positive behavioural transfer effects of acute exercise by enhancing BDNF release (Mera et al., 2016;Khrimian et al., 2017;Nicolini et al., 2020). Further, exercise prescription (e.g., intensity, duration, type, frequency) is thought to influence exercise-related systemic, cellular, molecular, and behavioural outcomes. To date, there is only evidence supporting high-intensity exercise being optimal to maximize motor benefits, at least acutely. Optimal duration, type, and frequency of exercise to prime maximal neuroplasticity within intracortical and corticospinal motor networks remain to be identified. Another important knowledge gap that needs to be filled concerns the mechanistic link between exercise-induced systemic, cellular, and molecular changes and exercise-induced improvements in motor learning and function. As of now, exercise-induced changes in systemic, cellular, molecular, and behavioural outcomes have been examined in separate studies. For example, TMS findings have shown that exercise increases corticospinal excitability (Lulic et al., 2017;El-Sayes et al., 2019;MacDonald et al., 2019;Opie and Semmler, 2019;Nicolini et al., 2020), while neuroimaging ones have demonstrated that exercise positively influences functional connectivity, gray matter density, and white matter microstructure in motor areas Voss et al., 2013b;Herting et al., 2014;Rajab et al., 2014;Schlaffke et al., 2014;Svatkova et al., 2015;Rowley et al., 2020). Further, molecular studies have reported BDNF upregulation after exercise (Rasmussen et  behavioural ones have shown that exercise results in improved motor learning (Roig et al., 2012;Statton et al., 2015;Thomas et al., 2016aThomas et al., , 2016bWanner et al., 2020). Based on these findings, it is likely that increases in cortical excitability, functional connectivity, and BDNF might contribute to mediating improvements in motor behaviour following exercise as these systemic, cellular, and molecular changes have been related to motor learning independently of exercise (Rioult-Pedotti et al., 2000;Sanes and Donoghue, 2000;Donchin et al., 2002;Muellbacher et al., 2002;Doyon and Benali, 2005;Monfils et al., 2005;Kleim et al., 2006;Boyke et al., 2008;Taubert et al., 2010;Teixeira et al., 2010;McHughen et al., 2010;Dayan and Cohen, 2011;He et al., 2013;Sehm et al., 2014;Tunovic et al., 2014;Lee et al., 2014). Five studies, mentioned in the Introduction and all conducted with healthy and young adults, have assessed exercise-induced changes in neurobiological and behavioural outcomes, with four reporting a relationship (Skriver et al., 2014;Ostadan et al., 2016;Dal Maso et al., 2018;Lehmann et al., 2020) and one failing to do so (Mang et al., 2014). It is important that more studies, particularly involving aging and diseased individuals, examine exercise-linked systemic, cellular, molecular, and behavioural outcomes in the same study, in order to mechanistically establish whether exercise-induced increases in TMS-probed corticospinal excitability, in neuroimaging-assessed functional connectivity, gray matter, and white matter as well as in molecular measures, such as BDNF, are related to improvements in motor behaviour. Indeed, in order to be able to successfully utilize exercise protocols in rehabilitation and to preventively off-set motor deficits associated with physiological aging, we need to determine whether exercise-induced systemic, cellular, molecular, and behavioural changes seen in healthy adults are also found in older individuals and clinical populations. Lastly, the effects of long-term exercise on motor learning and performance need to be investigated in future research. Although some promising findings suggest that exercise training has positive transfer effects on motor learning and performance for older adults, stroke survivors, and Parkinson's patients (Bakken et al., 2001;Miyai et al., 2002;Herman et al., 2007;Quaney et al., 2009;Wang et al., 2020), the motor benefits of long-term exercise remain largely speculative due to the paucity of month-to-year-long exercise intervention studies. Filling the remaining knowledge gaps will allow the design of exercise interventions that optimally enhance motor plasticity.

CONFLICTS OF INTEREST
All authors have no potential sources of conflict of interest to declare.