This study was part of a feasibility and proof-of-principle interventional study aiming at restoring walking function among patients with MS-related walking disability (51). The data on feasibility and restoration of walking have been reported elsewhere (51). This interventional study (10 weeks, 3×/week exercise training) with TMS assessment pre, post, and 3-month follow-up was approved by the local health ethics board prior to initiation (Health Research Ethics Board, #2019.0225, NCT04066972).

Ten participants were recruited via referral from neurologists and physiotherapists in the local MS clinic, as well as from an outpatient rehabilitation service discharge database. All participants signed informed consent prior to study inclusion. Recruitment and screening details have been described elsewhere (51). Participants were included if they (1) were diagnosed with progressive MS (SPMS or PPMS), (2) reported no relapses 3 months prior to inclusion, (3) presented with walking impairments (e.g., use of bilateral or unilateral gait aids), (4) had disability level ≥6.0 on the Expanded Disease Status Scale (EDSS), (5) were capable of participating in physical exercise [as per Physical Activity Readiness Questionnaire (PAR-Q) screening form (52)], and (6) were eligible to undergo TMS (53) and dual energy X-ray absorptiometry (DEXA) (54) as per screening procedures. Written informed consent was obtained from participants for the publication of any potentially identifiable images or data included in this article.

Two participants dropped out during the intervention (51), reporting not being able to commit to the proposed frequency of exercise sessions (3×/week). Eight participants (seven female) completed the intended exercise training, and pre–post data were collected. One participant (number 2) could not be reached during follow-up assessment. Participant demographics are presented in Table 1.

Participants underwent 10 weeks (3×/week) of vigorous treadmill walking exercise training in a temperature-controlled room (16°C) (51). The treadmill was equipped with a harness to prevent falls and to support ≤10% of participants' body weight. The dosage target of the exercise was 40 min (5 min warm-up and cool down) at a moderate-high intensity (40–65% heart rate reserve), which was adjusted throughout the training by increasing the speed and incline of the treadmill and/or reducing body weight support. Manual assistance to advance legs and resting breaks of ≤2 min were provided whenever necessary (51).

All outcome measures were assessed before the intervention (n = 8), after the 10-week period intervention (n = 8) and at 3-month follow-up after the exercise intervention had ended (n = 7).

A priori, we planned to use a one-way repeated measures analysis of variance and Friedman test when testing normal and non-normally distributed data, respectively. Because tests of normality (e.g., Shapiro–Wilk) typically require samples sizes of n ≥ 10 to generate reliable results (68), the more robust non-parametric alternative (i.e., Friedman test) (69) was preferred (70) to determine changes in TMS variables [RMT, AMT, and excitatory and inhibitory recruitment curves (MEP amplitudes_{105−155%AMT}, CSP time_{105−155%AMT}, slope, R², and area under the curve)], fitness (ml min⁻¹ kgLM-1, body fat %), and workload performed (kcal/session), at the different time points (pre, post, and follow-up). Analysis between time points (pre vs. post vs. follow-up) is reported as χ(degreesoffreedom)2 = test statistic, p-value. When statistically significant (p < 0.05), Bonferroni-corrected pairwise comparisons were performed to identify the difference across time points, and the adjusted p-value for multiple comparisons is reported. All data in the text are presented as median (Mdn).

Relationships between changes in cardiorespiratory fitness (ml min⁻¹ kg⁻¹_leanmass), lean mass (kg), body fat (%), levels of fatigue (FSS, MFIS), workload performed (kcal/session), and TMS changes were investigated with Spearman's coefficient (rho) at the unadjusted significance level of p < 0.05. Change scores were calculated as % changes = post – pre/pre.

Differences between TMS values of the stronger and weaker hand were investigated separately for each time point (pre, post, follow-up) with Wilcoxon non-parametric paired t-tests.

Friedman's test showed a significant difference for AMT between time points (pre, post, follow-up) in both stronger and weaker hands [χ²₍₂₎ ≥ 8.27, p ≤ 0.016]. Pairwise analysis revealed higher corticospinal excitability (i.e., lower AMT) in participants post- compared to pre-intervention in both stronger [MSO%; Mdn (pre vs. post) = 33 vs. 27, p = 0.033] and weaker hands [MSO%; Mdn (pre vs. post) = 41 vs. 37, p = 0.013), which returned to baseline at follow-up (Figures 1A,B). Higher variability was found for RMT; no change, increases, and decreases of RMT were noted across participants in both hemispheres (stronger and weaker hands), and no statistically significant changes were observed in either hemisphere (Figures 1C,D).

Corticospinal gain (excitatory recruitment curve slope) was statistically different between time points in both stronger and weaker hands [χ²₍₂₎ ≥ 8.40, p ≤ 0.015]. Pairwise analysis revealed increased capacity to recruit excitatory neurons with increased TMS stimulation intensities (i.e., higher slope) post- compared to pre-intervention [Mdn = (pre vs. post) = stronger: 1.33 vs. 2.20, p = 0.013; weaker: 0.67 vs. 2.08, p = 0.028], which returned to baseline at follow-up (Figure 2). Recruitment curve accuracy (R²) did not change in neither stronger or weaker hand [χ²₍₂₎ ≤ 4.00), p ≥ 0.135].

For MEP amplitudes, statistical significance between time points were noted at the intensities of 135% [χ²₍₂₎ = 7.00, p = 0.030] and 145% [χ²₍₂₎ = 9.33, p = 0.009] of AMT in the weaker hand and at 145% of AMT in the stronger hand [χ²₍₂₎ = 6.00, p = 0.050]. In all cases, pairwise analysis revealed increased corticospinal excitability (higher normalized MEP amplitudes) post- compared to pre-intervention with return to baseline at follow-up [% of largest baseline MEP; Mdn (pre vs. post): weaker hand: 135% of AMT: 85.49 vs. 111.39, p = 0.028; 145% of AMT: 85.78 vs. 151.66, p = 0.012; stronger hand: 145% of AMT: 88.73 vs. 127.05, p = 0.048; Figure 3].

In the stronger hand, differences between time points were noted for CSP investigated in all TMS intensities [105–155% of AMT; χ²₍₂₎ ≥ 6.00, p < 0.050]. Pairwise analysis revealed reductions in CSP time post- compared to pre-intervention across all intensities used (p ≤ 0.048), which returned to baseline level at follow-up (Figure 4A). In the hemisphere corresponding to the weaker hand, there was a statistically significant difference for CSP time at the different time points at lower TMS intensities {105–125% of AMT [χ²₍₂₎ = 6.33, p = 0.042]}; however, statistical significance was not reached during pairwise analysis (p ≥ 0.063; Figure 4B).

Lean body mass of the participants increased from pre- to post-intervention and from post-intervention to follow-up; however, only the change from pre to follow-up was statistically significant [χ²₍₂₎ = 7.00, p = 0.030; Mdn, lean mass (kg) (pre vs. follow-up): 41.74 vs. 48.57, p = 0.028] (Figure 5A). Body fat also decreased during follow-up, and a statistical significance was noted from post to follow-up [χ²₍₂₎ = 8.33, p = 0.016; Mdn, body fat % (post vs. follow-up): 40.00 vs. 37.35, p = 0.012; Figure 5B].

Although four out of eight participants improved their cardiorespiratory fitness (ml min⁻¹ kg⁻¹_leanmass), no overall statistical change was reached (p ≥ 0.368; Figure 5D). However, an increased capacity to perform exercise were noted as participants were able to perform a higher exercise workload (kcal/session) in the last compared to the first exercise session [χ²₍₂₎ = 7.14, p = 0.028; Mdn, kcal/session (pre vs. post) = 121.39 vs. 70.24, p = 0.023], and this capacity was maintained during follow-up (Figure 5C).

In the stronger hand, overall corticospinal excitation [area under the curve (AUC), normalized MEP amplitudes] differed between time points [χ²₍₂₎ = 11.14, p = 0.004]. Pairwise analysis revealed increased overall corticospinal excitation (higher AUC) post- compared to pre-intervention [Mdn, AUC_{105−155%ofAMT} (pre vs. post) = 3,237 vs. 3,947, p ≤ 0.016) with returned to baseline level at follow-up (Figure 6A). Relationship analysis demonstrated that greater increases in overall corticospinal excitation in the stronger hand were associated with greater reduction in fatigue severity levels measured with the FSS (rho = 0.762, p = 0.028; Figure 6B) and fatigue impact measured with the MFIS (rho = 0.962, p = 0.001; Figure 6C).

Nerve conduction speed (MEP latency) did not change in either side [χ²₍₂₎ ≤ 1.14, p ≥ 0.565; Mdn, milliseconds (pre vs. post vs. follow-up): stronger hand, 24.17 vs. 24.51 vs. 22.12; weaker hand, 26.26 vs. 25.94 vs. 25.97].

All the TMS values (median and range), differences between stronger and weaker hands across time points, and reasons for missing values across time points are reported in Table 2.

Motor thresholds and MEP amplitudes are considered indicators of corticospinal excitation, mediated by glutamate and its activity on N-methyl-d-aspartate (NMDA) and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors (24, 30). In fact, higher glutamatergic receptor activity is associated with greater capacity for synaptic plasticity (72, 73), and disruption of this excitatory circuitry is responsible for diminished neuroplasticity and lower capacity to learn new tasks and recover from neurological damage (e.g., aging, stroke, MS) (4, 7, 71). Therefore, there are important initiatives underway to develop new treatments (e.g., exercise, pharmacological, non-invasive brain stimulation) aimed at increasing glutamatergic-mediated brain excitation in the injured brain to enhance neuroplasticity and recover function (9, 45, 71, 74–76). For instance, studies using TMS have confirmed that, in comparison to those who are less physically active, individuals with higher fitness have lower motor thresholds and higher MEP amplitudes (29) (i.e., higher corticospinal excitability) and demonstrate superior increases in MEP amplitudes (i.e., greater neuroplastic response) following paired associative stimulation to induce neuroplasticity (28, 77).

We have previously shown that acute exercise increases corticospinal excitation (i.e., higher MEP amplitude) and reduces intracortical inhibition (i.e., shorter CSP) among people with walking disability due to progressive MS (47). Importantly, this effect was noted only in the stronger hand (47), likely due to a more intact (i.e., less affected) contralateral corticospinal representation (33). Here, we showed bilateral reductions in AMT, increases in MEP amplitudes, and superior motor neuronal recruitment (higher recruitment curve slope) after 10 weeks of aerobic exercise training. This suggests that the stimulus from regular exercise training may have led to the chronic enhancements in excitatory synaptic transmission noted in these participants. Moreover, even though the hemisphere corresponding to the weaker hand, which was likely more affected by MS (33, 78), was unresponsive after one exercise session (47), in this longer term exercise training, it demonstrated capacity to improve in synaptic excitatory transmission. It is interesting to observe that Nicoletti et al. (9) recently reported enhanced corticospinal excitation in people with progressive MS after 4 weeks of d-aspartate treatment, which aimed to enhance NMDA receptor activity (9). They also showed increases in MEP amplitudes following intermittent theta burst stimulation (i.e., enhanced neuroplasticity) (9). It appears that exercise training has comparable benefits in terms of enhancing capacity for neuroplasticity in progressive MS. It is important to note that the corticospinal excitability enhancements reported here and those by Nicoletti et al. (9) were short term and disappeared 3 months after cessation of the intervention. Therefore, we suggest that treatments that enhance neuroplasticity, such as physical exercise training, should be prescribed continuously in progressive MS to protect the brain, improve brain function, and likely to potentiate the effects of treatments (e.g., drugs) and other neuroplasticity-inducing protocols (e.g., non-invasive brain stimulation).

When applying suprathreshold TMS stimulations to the primary motor cortex with participants performing a tonic muscle contraction of the contralateral target muscle, the length of the period of cessation of muscle activity (CSP) is an indicator of intracortical inhibition mediated by the activity of the inhibitory neurotransmitter GABA on its ionotropic and metabotropic receptors (GABA_A and GABA_B, respectively) (24, 32). Although the cortical and spinal contribution to the CSP length is still unclear (24, 79), it is generally accepted that the cortex is the main modulator of CSP change (32). Because excessive GABAergic-mediated intracortical inhibition is considered pathological (80, 81), detrimental to neuroplasticity (39, 40, 81, 82), and is associated with disease progression in MS (36) and stroke (83), decreasing its activity is an attractive treatment strategy to boost neuroplasticity (40, 81).

In healthy people and people with stroke, studies have confirmed that even a single bout of aerobic exercise is able to acutely reduce short intracortical inhibition (59, 83–86) assessed with TMS paired pulse, a TMS biomarker of GABA_A-receptor activity (24). We recently reported a similar effect after acute aerobic exercise in people with progressive MS (47). Interestingly, here, we showed that after 10 weeks of exercise training, CSP duration was reduced at all TMS intensities, indicating reductions in both GABA_A and GABA_B-mediated intracortical inhibition. This result aligns with findings in healthy individuals demonstrating that 4–12 weeks of strength exercise training reduced both GABA_A- and GABA_B-receptor activity, as decreasing in short-intracortical inhibition and duration of the CSP elicited at higher TMS intensities, respectively (26). We have previously shown that among people with MS, superior cardiorespiratory fitness was related to shorter CSP (44). In our present findings, although there were no significant improvements in cardiorespiratory fitness measured as the peak of oxygen consumption (VO_2peak), there were other indicators of improved physical health (48) such as higher capacity to perform exercise (i.e., kcal/session), greater lean mass, and lower body fat percentage, and increases in other parameters of cardiorespiratory fitness such as the oxygen uptake efficiency slope [for details, see (51)]. The fact that the beneficial reduction (acute and long term) in intracortical inhibition was only observed in the brain hemisphere corresponding to the stronger hand may suggest a greater neuroplastic potential of inhibitory mechanisms in the hemisphere thought to be less affected by MS. Furthermore, our walking training provided a high degree of task-specific training (18, 87, 88). Ziemann et al. has shown that less GABAergic-mediated intracortical inhibition, assessed with TMS, was essential for motor learning processes from task-specific training to occur (89). Decreasing GABAergic-mediated intracortical inhibition has also been proposed to be an important factor initiating increases in muscular strength (26, 27, 31). Although we did not measure muscular strength (e.g., MVC pre–posttraining), we did note increases in lean mass at post and follow-up as well as improvements in walking function [e.g., walking speed; see (51)]. Altogether, this indicates that long-term physical exercise that utilizes task-specific training in highly disabled people with progressive MS reduces intracortical inhibition and possibly improves and restores physical function through enhanced neuroplasticity. Although, because no correlation between changes in intracortical inhibition, body composition, and walking function was noted, it remains to be answered whether decreasing intracortical inhibition would lead to improvements in learning and restoration of function in people with MS. Future research should examine whether such effects would take place in a larger sample with different walking abilities using a randomized controlled design. As well, because we measured overall gains in walking function (51) and body composition, future research should examine whether the enhanced plasticity (reduced inhibition) measured in the hemisphere corresponding to the stronger side of the body indeed translates into global brain function improvement (60) (e.g., bilateral and cognitive function) or whether it is restricted to the contralateral representation. This would be an important discovery for interventions aiming at improving function of the most affected side.

It is interesting that, when compared to healthy controls, some studies have shown reduced intracortical inhibition (shorter CSP) in MS patients (90, 91). Nantes et al. reported that shorter CSP correlated with lower whole brain cortical volume (MRI, magnetic transfer ratio) in progressive MS and that, interestingly, longer CSP was a predictor of upper extremity motor dysfunction (92). Therefore, when compared to the healthy central nervous system (CNS), the CNS affected by MS may display decreased activity of inhibitory mechanisms that, curiously, may work as a compensatory mechanism during brain disease. The concept that there are compensatory mechanisms that increase brain excitation and decrease brain inhibition in order to preserve brain function in CNS disease has been recently proposed by other authors (7, 33, 44, 93–95). However, these processes are certainly not uniform across CNS disorders. For instance, in Parkinson's disease, Fisher et al. (96) showed that high-intensity treadmill exercise program improved walking performance and lengthened CSP time (96), which is typically shortened in people with Parkinson's disease (97). Thomas et al. (98) also showed lengthening of CSP in people with incomplete spinal cord injury after a regimen of treadmill training. Although the mechanisms are not entirely clear, our work and the work of others suggests that rehabilitation and exercise prime the CNS as measured by shifting of the CSP.

Fatigue is one of the most disabling symptoms in MS (44–46). Although the etiology of MS-related fatigue is not completely understood, neuroimaging studies [e.g., MRI, functional MRI (fMRI)] have proposed that its development and progression is due to structural and functional abnormalities in both cortical and subcortical areas (45). Previous studies have shown that 10–12 weeks of physical exercise training can lessen subjective fatigue in people with MS (99), including progressive MS (51, 100). Based on previous findings showing an association between shorter CSP and lowered levels of subjective fatigue in a cohort of people with MS (44), we proposed that improving fitness through exercise training could mitigate fatigue by decreasing GABAergic-mediated intracortical inhibition (i.e., shortening CSP). In this current pilot study, we reported a strong association between increases in corticospinal excitation (recruitment curve; AUC) and reductions in subjective fatigue (FSS and MFIS). Nicoletti et al. (9) also demonstrated reductions in subjective fatigue (FSS) and increases in corticospinal excitation (intracortical facilitation) after d-aspartate treatment in people with progressive MS (9). Furthermore, Créange et al. (101) have also shown increases in corticospinal excitation (e.g., RMT reduction) and reduction in levels of fatigue after erythropoietin treatment to improve walking in people with progressive MS. Our results and the results of others support that there is a link between corticospinal excitation/inhibition and fatigue, which should be examined in larger trials. In fact, non-invasive brain stimulation methods (repetitive TMS, transcranial direct current stimulation), which aim to increase cortical excitation and treat MS fatigue, have been recently proposed (45). It is important to note that the abovementioned experiments, and the present study, measured perceived (i.e., subjective) fatigue and not fatigability (i.e., muscle/performance fatigability measured during contraction). Nonetheless, because perceived fatigue and fatigability closely associate (102), our results showing reduced levels of perceived fatigue and improved fitness suggests that following training, subjects required less physical effort to perform activities of daily living, suggesting superior energy availability and reduced fatigability (102). Therefore, we propose that exercise training might be able to mitigate symptoms of fatigue possibly by acting through increases in excitatory circuitry.

There are some important limitations to consider when interpreting the results of the present study. First, this was a small pilot study, and no statistical sample size calculation was conducted for the outcomes presented in this manuscript, which limits the statistical power to obtain conclusive results. Second, no control group was included, which limits the conclusion on the true effect of the intervention. Third, as only patients with progressive MS and severe MS-related walking disabilities (EDSS, 6.0–6.5) were included, the findings may not be applicable for relapsing–remitting and/or less disabled MS patients. Despite these limitations, the novel insights from this study may serve as a rationale for larger studies and continued efforts in investigating the effects of exercise and physical rehabilitation on neuroplasticity and functional recovery in MS.

As for considerations for future studies, although the aim of this study was to investigate changes in corticospinal excitability in a non-exercised hand muscle to demonstrate widespread effects of exercise training on global brain plasticity (59, 60), investigating muscles that were more involved in the walking training (e.g., lower limb muscles) could provide more insight regarding the link between the trained muscle and cortical function (TMS) (27). Moreover, having participants' neuroimaging data (e.g., magnetic resonance imaging) could help to better understand the role of lesion volume and location on exercise-induced corticospinal excitability changes. We determined averaged MEP amplitudes and CSP times from a small number of trials (three to six) as done previously by others (28, 65, 66), and with participants performing tonic contraction, in order to reduce intrasubject variability (27). Future studies should examine the optimal number of stimulation trials (103) in order to produce reliable MEP/CSP data. With respect to the TMS recruitment curve parameters, we used linear regression (TMS intensities by MEP amplitudes), as done by others (25, 34), in an attempt to assess the corticospinal tract recruitment gain (slope) and accuracy (R²); biomarkers were previously proposed by Potter-Baker et al. (67) to reflect morpho-physiological integrity of the corticospinal tract in stroke. However, more studies are necessary in order to understand what the best model is [e.g., sigmoidal (67) or linear (25, 34)] when calculating these parameters while taking into consideration the different TMS methodologies (e.g., range of TMS intensities employed), the clinical population (e.g., stroke, MS), and lesion profile (e.g., lesion volume, location).