OUP user menu

Training-dependent plasticity in patients with multiple sclerosis

Katrin Morgen, Nadja Kadom, Lumy Sawaki, Alessandro Tessitore, Joan Ohayon, Henry McFarland, Joseph Frank, Roland Martin, Leonardo G. Cohen
DOI: http://dx.doi.org/10.1093/brain/awh266 2506-2517 First published online: 29 September 2004

Summary

Cortical reorganization has been demonstrated in the motor network that mediates performance of a motor task in patients with multiple sclerosis. How this network responds to motor training is not known. This study examined functional MRI (fMRI) activation patterns associated with performance of a motor task, consisting of repetition of directionally specific voluntary thumb movements, before and after motor training in a group of multiple sclerosis patients with mild motor impairment of the right upper extremity. Patients and healthy subjects were scanned in one session before, during and after a 30 min training period. fMRI data obtained during rest, thumb flexion (trained movement) and thumb extension (untrained movement) were analysed using random effects analysis (SPM99). Motor kinematics of training motions and EMG from the resting hand were monitored with an accelerometer and surface EMG electrodes. Kinematics of thumb movements before, during and after training were comparable in the absence of mirror EMG activity in the resting hand. Before training, thumb movements elicited more prominent activation of the contralateral dorsal premotor cortex [PMd, Brodmann area (BA) 6] in multiple sclerosis patients than in controls. After training, unlike the control group, multiple sclerosis patients did not exhibit task-specific reductions in activation in the contralateral primary somatosensory (S1), motor (M1) and adjacent parietal association (BA 40) cortices. These results indicate that patients engage the contralateral PMd more than controls in order to perform directionally specific movements before training. The absence of training-dependent reductions in activation in S1, M1 and BA 40 is consistent with a decreased capacity to optimize recruitment of the motor network with practice.

  • multiple sclerosis
  • funcional MRI
  • plasticity
  • training
  • dorsal premotor cortex
  • BA = Brodmann area
  • EDSS = Expanded Disability Status Scale
  • IPL = inferior parietal lobule
  • M1 = primary motor cortex
  • NINDS = National Institute of Neurological Disorders
  • PMd = dorsal premotor cortex
  • ROI = region of interest
  • S1 = primary somatosensory cortex
  • S2 = secondary somatosensory cortex
  • SMA = supplementary motor area
  • TE = thumb extension
  • TF = thumb flexion

Introduction

Multiple sclerosis is characterized by a relapsing–remitting course of disease that eventually leads to cumulative disability, including sensorimotor deficits. Cortical plasticity, the capacity of the CNS to adapt to new environmental challenges or anatomical damage (Jacobs and Donoghue, 1991; Keller et al., 1992; Halter et al., 1995; Hess and Donoghue, 1996), contributes to functional recovery after brain injury (Schallert et al., 2000; Johansen-Berg et al., 2002a) and occurs in multiple sclerosis (Lee et al., 2000; Reddy et al., 2000; Pantano et al., 2002; Rocca et al., 2002a; Staffen et al., 2002). For example, performance of simple motor tasks is associated with more extensive activation of motor areas in multiple sclerosis patients than in healthy control subjects, possibly reflecting adaptation of the cortical networks to cope with increasing difficulties in generating appropriate motor output (Lee et al., 2000; Reddy et al., 2000; Pantano et al., 2002; Rocca et al., 2002a).

Motor training, required for skill acquisition and for relearning of lost functions (Indredavik et al., 1997; Stroke Unit Trialists Collaboration, 1997) and functional recovery (Johansen-Berg et al., 2002a) in patients with brain lesions, constitutes a key element of rehabilitative treatment in multiple sclerosis (Wiles et al., 2001). However, the neural substrates underlying motor performance after motor training in patients with multiple sclerosis are incompletely understood. To address this issue, we studied functional MRI (fMRI) activation patterns before and after motor training in a group of multiple sclerosis patients with mild motor impairment in the right upper extremity and in age-matched controls.

Methods

Subjects

Nine patients with definite multiple sclerosis (McDonald et al., 2001); five women and four men, (43.1 ± 10.4 years) and nine age- and gender-matched healthy volunteers (42.0 ± 8.6 years) participated in the study (Table 1). Criteria for inclusion in the study encompassed ability to perform the motor task appropriately, right-handedness according to the Edinburgh handedness inventory (Oldfield, 1971) and stability of clinical status in the preceding 3 months. Exclusion criteria included severe fatigue, severe spasticity/pain, severe uncontrolled medical problems and use of medications known to disrupt cortical plasticity, such as α-adrenergic antagonists or agonists, tranquillizers, clonidine, phenytoin, GABAergic agents (i.e. benzodiazepines), scopolamine, haloperidol, other neuroleptics and barbiturates (Feeney et al., 1993; Ikegaya et al., 1997; Goldstein, 2000; Sawaki et al., 2003a). Clinical evaluation of patients included complete neurological examination, determination of Expanded Disability Status Scale (EDSS; Kurtzke, 1983), score and the nine-hole peg test, a quantitative test of upper extremity fine motor coordination that reflects the time required to place and remove nine pegs in nine holes drilled on a board (Goodkin et al., 1988). The EDSS score evaluates functions mediated by activity in pyramidal, cerebellar and brainstem tracts as well as somatosensory, bowel, bladder and visual function. Most of the patients in our study had an EDSS score between 1.0 and 4.5, indicating that they were fully ambulatory. One of our patients had an EDSS score of 6.0. This patient required intermittent or unilateral constant assistance (cane, crutch or brace) to walk ∼100 m with or without resting. The study was approved by the Institutional Review Board of the National Institute of Neurological Disorders and Stroke (NINDS). All subjects gave written informed consent.

View this table:
Table 1

Patient characteristics

PatientAge (years)Years since diagnosisClinical courseEDSS scoreRight-handperformance in nine-hole peg test (in s)Left-hand performance in nine-hole peg test (in s)T2 lesion load (in cm3)T1 lesion load (in cm3)Brain parenchymal fraction
1*4810.0RR1.519.521.81.590.350.81
2*4929.5RR1.519.519.34.000.000.80
3230.3RR1.016.517.62.310.160.85
43615.0RR3.018.619.66.622.590.82
5§5413.5SP6.029.129.518.148.510.75
6*345.0RR1.019182.430.890.86
7532.0RR2.52319.62.330.190.84
8*503.5RR1.5262511.076.760.78
9*418.0RR1.523195.201.450.77
Mean ± SD43.1 ± 10.49.6 ± 9.02.2 ± 1.621.6 ± 4.021.0 ± 3.95.97 ± 5.442.32 ± 3.150.81 ± 0.04
  • RR = relapsing–remitting; SP = secondary progressive.

  • * Patients with a history of transient right-hand motor deficit during relapsing–remitting disease course;

  • § patient experienced recovery from upper extremity motor and sensory deficit. Performance in the nine-hole test reflects the time required to place and then remove a set of nine pegs in nine holes with the right and the left hand (s).

Structural image analysis

MRI examinations were performed on a 1.5 T Signa unit (General Electric, Milwaukee, WI) using a standard quadrature head coil. Imaging sequences included axial oblique proton density/T2-weighted (variable-echo, echo time 20/100 ms and repetition time 2000 ms) and gadolinium (Gd)-enhanced axial oblique T1-weighted images (echo time 20 ms, repetition time 600 ms). Axial oblique slices (either 5 mm × 27 slices or 3 mm × 42 slices) were prescribed from a mid-sagittal T1-weighted spin echo [400/14 (repetition time/echo time)] image. Data were acquired using contiguous interleaved slices with a 24 cm field of view and a matrix of 256 × 192. The following measures were calculated. (i) White matter lesion load was calculated from T2-weighted and proton density images using a semiautomated threshold technique on a Sun workstation [PV-Wave CL version 6.2; Sun4 Solaris Sparc, copyright 1995 (DeCarli et al., 1992)]. (ii) Lesion load on T1-weighted MRI (‘black holes’): using Medx software (Sensor Systems, Inc., Sterling, VA, USA), hypointense lesions on Gd-enhanced T1-weighted images were outlined manually on each slice after application of a semiautomated threshold technique. Lesion volume was then computed as the product of the pixel size [x, y = 0.9375 (mm)] and slice thickness. (iii) Brain parenchymal fraction, a measure of brain atrophy, was calculated as the ratio of cerebral brain matter to total cranial volume using a semiautomated threshold technique on a sun workstation [PV-Wave CL version 6.2; Sun4 Solaris Sparc, copyright 1995 (DeCarli et al., 1992)].

fMRI

Activation tasks

fMRI data were obtained in one session that consisted of an 8 min pre-training run, followed by a 30 min training period and an 8 min post-training run (total 46 min). During pre- and post-training runs, 30 s blocks of thumb flexion (TF; B) and thumb extension (TE; C) alternated with rest periods (A) (ABAC ABAC ABAC ABAC). During each 30 s block of TF (B) or TE (C), movements were visually cued at 1 Hz. Visual cues were also achieved at 1 Hz during 30 s rest periods in order to provide comparable visual stimulation. The 30 min training period consisted of flexion movements of the right thumb visually cued at 1 Hz. Subjects were instructed to flex or extend the right thumb briskly and let it return passively to the start position. Three 9 min periods of TF movements alternated with 1 min rest periods (total 30 min). The training right hand and forearm were stabilized in a cast. If necessary, subjects were encouraged via headphones to perform accurately and consistently between runs.

EMG activity was monitored on-line from the non-training hand with surface electrodes overlying the extensor pollicis brevis and flexor pollicis brevis muscles to screen for mirror activity.

fMRI acquisition

Functional imaging was conducted with a whole-body 1.5 T MR scanner (Horizon; General Electric Medical Systems, Milwaukee, WI) with a standard head coil. Blood oxygen level-dependent (BOLD)-sensitive, single-shot, blipped, gradient-echo, echo-planar images were acquired using the following parameters: repetition time 3 s, echo time 40 ms, flip angle 90° and voxel size 3.75 × 3.75 × 6.0 mm. Twenty-two contiguous axial slices were acquired to cover the whole brain in each subject. Sampling of data coincided with stimulus presentation. Timing of data sampling and stimulus presentation was based on the assumption of a steady-state response over each 30 s block to stimuli in cortical motor and sensory regions (Veltman et al., 2002).

Image pre-processing

Scans were transferred to a Sun workstation (Sun Microsystems, Palo Alto, CA). Data processing was performed using statistical parametric mapping (SPM99; London, UK) (Friston et al., 1995). The first four images of each time series were discarded to eliminate signal intensity variation caused by progressive saturation. To correct for head motion, images were spatially aligned to the first image in the time series with a six-parameter rigid-body transformation. Images were then normalized to the EPI brain template of the Montreal Neurological Institute (Cocosco et al., 1997) based on the reference system of Talairach and Tournoux (1988). The normalized images, interpolated to 2 × 2 × 2 mm voxels, were spatially smoothed with a 10 mm, full-width, half-maximum, isotropic Gaussian kernel. The anatomical location of functional group results was displayed on a normalized T1-weighted template.

Functional image analysis

Statistical parametric maps were created based on the general linear model used by SPM99 to identify areas of interest that co-varied with movement or change in movement (Friston et al., 1995). The different conditions (movement and rest) were modelled as boxcar functions convolved with the haemodynamic response function. Initially, differences between active and rest conditions were determined with a t test at each voxel for the pre- and post-training runs. Then, separate random effects group analyses were performed for the nine patients and the nine matched healthy volunteers as well as for a direct group comparison.

First, we compared activation for thumb movement and rest in statistical parametric maps: (flexion–rest)pre-training >0 and (extension–rest)pre-training >0, as well as (flexion–rest)post-training >0 and (extension–rest)post-training >0; P < 0.01 for peak height and cluster size >10 voxels. The activated areas in the four main effect maps were converted into a binary image for each group and used as a mask for the analysis of changes in activation as a result of training. The restriction to areas activated at baseline or after training facilitated the interpretation of training-induced changes in cerebral activity.

Secondly, we investigated the differential effects of training on activation induced by performance of the trained (TF) and untrained (TE) motor tasks using an analysis of interactions: (flexion–extension)pre-training—(flexion–extension)post-training and—[(flexion–extension)pre-training—(flexion–extension)post-training] (=task-specific effects). To clarify the results of these interaction analyses, we plotted the time course of signal intensities in voxels representing cluster maxima. We also performed a direct group comparison of task-related effects. For the group comparison, we limited statistical tests to a region of interest (ROI) based on a previous analysis of a larger group of healthy subjects. Our a priori hypothesis was that task-specific effects would differ between the groups in a region in left SM1 and parietal association cortex [Brodmann area (BA) 40] previously shown to exhibit training-related effects in healthy subjects (Morgen et al., 2004). The ROI encompassed the area in the left primary sensorimotor cortex (SM1) and adjacent parietal association cortex identified in the previous analysis at P < 0.01 voxel-level uncorrected and P < 0.05 cluster-level corrected. Voxels indicating group differences at P < 0.01 and k > 10 were considered significant if they occurred within the ROI. For the between-group analysis of task-specific effects in other regions of the brain and for individual group analyses of changes in activation pattern, we reported results as significant when P < 0.01 uncorrected for amplitude and P < 0.05 cluster-level corrected.

Monitoring of motor performance

Thumb movements were recorded with a three-dimensional accelerometer mounted on the proximal phalanx of the right training thumb (Kistler Instrument Corporation, Amherst, NY), and movement direction was calculated from the first-peak acceleration vectors. Acceleration signals were recorded in the vertical (extension and flexion) and horizontal (adduction and abduction) axes and digitized at 4000 Hz. Data were analysed using a data collection analysis program written in LabView (National Instruments, Austin, TX). To monitor the consistency of motor training kinematics across condition and groups, we measured the dispersion of training movement directions and the magnitude of the first-peak acceleration of these movements. Peak acceleration and angular dispersion of voluntary thumb movements before, during and after training were analysed separately using factorial ANOVA (analysis of variance) with factors subject (multiple sclerosis patients and controls) and time (TF before, during and after training and TE before and after training).

Results

Subjects

All subjects performed the motor task and training appropriately and completed the experimental protocol (Tables 1 and 2).

View this table:
Table 2

Kinematics of thumb movements in fMRI experiments

Peak acceleration for flexionPeak acceleration for extensionAngular dispersion forflexionAngular dispersion for extension
Pre-training
    Patients5.17 ± 0.285.15 ± 0.240.91 ± 0.030.90 ± 0.02
    Controls5.35 ± 0.345.52 ± 0.300.95 ± 0.030.96 ± 0.03
During training
    Patients4.80 ± 0.37n/a0.92 ± 0.03n/a
    Controls5.24 ± 0.45n/a0.94 ± 0.03n/a
Post-training
    Patients5.03 ± 0.324.89 ± 0.300.91 ± 0.020.92 ± 0.02
    Controls5.31 ± 0.365.67 ± 0.310.92 ± 0.030.94 ± 0.03
  • Peak acceleration is expressed in m/s2. Angular dispersion is expressed as length of unit vector. n/a refers to the fact that thumb movements during training were in flexion.

Disability among the patients ranged from an EDSS score of 1.0 to 6.0 (mean 2.2). All patients were able to perform this simple motor task appropriately throughout the experiment, despite a high EDSS score in patient 5 and slowed nine-hole peg performance and high T1 and T2 lesion load in patients 5 and 8 (Table 1). None of the patients had suffered a relapse within the 3 months preceding the study.

Motor kinematics

Evaluation of first-peak acceleration of voluntary thumb movements did not show significant effects of factors subject (F = 3.83, NS) or time (F = 0.28, NS) or the interaction subject × time (F = 0.24, NS). Similarly, evaluation of angular dispersion of voluntary thumb movements did not show significant effects of factors subject (F = 3.14, NS) or time (F = 0.16, NS) or the interaction subject × time (F = 0.30, NS) (Table 2). In addition, all subjects were able to perform the training task in the absence of overt mirror movements or mirror EMG activity from extensor pollicis brevis or flexor pollicis brevis in the resting hand. Thus performance was consistent throughout the experiment in both groups.

fMRI

Activation before training

In healthy volunteers and patients, TF and TE movements activated the contralateral primary somatosensory cortex (S1), primary motor cortex (M1), dorsal premotor complex (PMd) and inferior parietal lobule (IPL; BA 40); bilateral secondary somatosensory cortex [S2; located in the upper bank of the lateral fissure and extending into the operculum (Disbrow et al., 2000)]; bilateral supplementary motor area (SMA) and pre-SMA; anterior cingulate cortex (BA 32); and ipsilateral cerebellum relative to rest. In addition to these areas, patients activated the contralateral cerebellum (Table 3, TF versus rest before training; Table 4, TE versus rest before training; and Fig. 1) and the left thalamus with TE; control subjects showed activation in the left thalamus with both movements.

Fig. 1

Activation patterns associated with performance of voluntary thumb flexion and extension movements visually paced at 1 Hz preceding motor training in healthy volunteers and patients (P < 0.01 voxel-level uncorrected, P < 0.05 cluster-level corrected; see also Tables 3 and 4).

View this table:
Table 3

Thumb flexion versus rest before training

Gyral locationBrodmann areaTalairach coordinatesZ-score*P cluster-level (corrected)
xyz
Healthy controls (n = 9)
    Left precentral gyrus, M14−34−17584.230.0001
    Left postcentral gyrus, S11, 2−51−28573.900.0001
    Left postcentral gyrus and operculum, S2−65−18252.960.04
    Right postcentral gyrus and operculum, S265−24252.980.05
    Left dorsal premotor cortex, PMd6−46−3554.530.0001
    Left superior frontal gyrus, SMA6−44504.230.001
    Right superior fontal gyrus, SMA6, 32210514.050.001
    Right superior frontal gyrus, pre-SMA81022433.590.001
    Left inferior parietal lobule, IPL40−42−38533.980.0001
    Right cerebellum, anterior lobe18−51−163.580.0001
    Right cerebellum, posterior lobe26−55−173.560.0001
    Left thalamus−10−13123.260.08 (trend)
Multiple sclerosis patients (n = 9)
    Left precentral gyrus, M14−46−15525.860.0001
    Left postcentral gyrus, S11, 2−44−17565.360.0001
    Left postcentral gyrus and operculum, S2−53−16275.170.0001
    Right postcentral gyrus and operculum, S267−18302.860.05
    Left dorsal premotor cortex, PMd6−40−3555.000.0001
    Left superior frontal gyrus, SMA6−6−1593.760.0001
    Right superior frontal gyrus, pre-SMA8416532.840.0001
    Left inferior parietal lobule, IPL40−34−38543.130.0001
    Left cerebellum, posterior lobe−30−57−164.680.05
    Right cerebellum, anterior lobe14−53−113.570.05
    Right cerebellum, posterior lobe24−57−163.460.05
    More activation in healthy controls than MS patients
    –
More activation in multiple sclerosis patients than healthy controls
    Left dorsal premotor cortex, PMd6–40–3553.500.07§
  • * Z-score of peak activation;

  • § In the same location as the difference in PMd activation for thumb extension (Table 4).

View this table:
Table 4

Thumb extension versus rest before training

Gyral locationBrodmann areaTalairach coordinatesZ-score*P cluster-level (corrected)
xyz
Healthy controls (n = 9)
    Left precentral gyrus, M14−34−21584.960.0001
    Left postcentral gyrus, S12−44−38573.330.0001
    Left postcentral gyrus and operculum, S2−63−20274.250.0001
    Right postcentral gyrus and operculum, S263−24203.290.03
    Left dorsal premotor cortex, PMd6−34−5545.220.0001
    Left medial frontal gyrus, SMA6−61534.780.0001
    Right medial frontal gyrus, SMA683624.410.0001
    Left superior frontal gyrus8−420512.570.0001
    Right medial frontal gyrus3288462.790.0001
    Left inferior parietal lobule, IPL40−53−24194.060.0001
    Right cerebellum, anterior lobe20−51−143.380.01
    Right cerebellum, posterior lobe28−55−193.010.01
    Left thalamus−6−1784.160.001
    Left putamen−201473.660.01
Multiple sclerosis patients (n = 9)
    Left precentral gyrus, M14−36−28594.440.0001
    Left postcentral gyrus, S11, 2−40−25534.690.0001
    Left postcentral gyrus and operculum, S2−48−18214.180.0001
    Right postcentral gyrus and operculum, S267−18253.250.03
    Left dorsal premotor cortex, PMd6−42−3504.820.0001
    Left superior frontal gyrus, SMA6−41644.040.0001
    Right medial frontal gyrus, SMA641573.870.0001
    Right superior frontal gyrus, pre-SMA8626482.870.0001
    Left inferior parietal lobule, IPL40−40−28503.970.0001
    Left temporal lobe, fusiform gyrus37−36−57−114.260.0001
    Left cerebellum, posterior lobe−40−63−194.550.0001
    Right cerebellum, anterior lobe14−51−93.710.0001
    Right cerebellum, posterior lobe30−59−143.650.0001
    Left thalamus−12−1143.330.03
More activation in healthy controls than multiple sclerosis patients
    –
More activation in multiple sclerosis patients than healthy controls
    Left dorsal premotor cortex, PMd6–42–5463.500.01§
  • * Z-score of peak activation.

  • § In the same location as the difference in activation for thumb flexion (Table 3).

In direct comparison, patients showed greater activation than controls in the contralateral PMd (BA 6; Tables 3 and 4, Fig. 2) during TF and TE. PMd activation in the patient group did not correlate with performance, as measured by the magnitude of the first-peak acceleration of thumb movements and by the nine-hole peg test, or with disease severity, measured by EDSS and lesion load on MRI.

Fig. 2

Direct comparison of activation patterns in patients and controls preceding training (P < 0.01 voxel-level uncorrected, P < 0.05 cluster-level corrected; see also Tables 3 and 4). Note more activation in the left PMd (BA 6) in MS patients than in healthy volunteers during thumb flexion (A) and extension (B).

Task-specific changes in activation after training

We investigated changes in activation patterns for the trained task (TF) relative to the untrained task (TE) with an analysis of interactions. Training led to task-specific decreases in activation in the contralateral S1, M1 and IPL (BA 40) (Table 5, Fig. 3) in healthy volunteers but not in patients (Table 5). There were no significant task-specific increases in activation in either group. A direct comparison of task-related effects documented group differences in contralateral S1 and IPL (BA 40).

Fig. 3

Three-dimensional rendering (A and B) of task-specific reduction in activation in contralateral IPL (BA 40; A and C), M1 (B and D, and S1 (B and E) in healthy volunteers (P < 0.01 voxel-level; see also Table 5). C, D and E illustrate average signal intensities (across subjects, adjusted for low-frequency effects) in voxels representing cluster maxima in IPL (BA 40), M1 and S1 for 30 s blocks of thumb flexion (trained movement, grey bars), extension (control movement, blue bars) and rest. Note that training led to a decrease in signal intensity predominantly for the trained movement (arrows).

View this table:
Table 5

Task-specific changes in activation patterns with the trained task (TF) relative to the untrained task (TE)

Gyral locationBrodmann areaTalairach coordinatesZ-score*P cluster-level (corrected)
xyz
Reductions in activation in healthy controls (n = 9)
    Left precentral gyrus, M14−34−18622.870.02
    Left postcentral gyrus, S11, 2−48−30603.350.02
    Left inferior parietal lobule, IPL40−38−38553.920.02
Reductions in activation in multiple sclerosis patients (n = 9)
Increases in activation in healthy controls (n = 9)
Increases in activation in multiple sclerosis patients (n = 9)
More pronounced reductions in activation in healthy controls than in multiple sclerosis patients
    Left postcentral gyrus, S12−46−28552.35k = 12
    Left inferior parietal gyrus, IPL40−44−32552.31k = 12
Other between-group differences in task-specific effects
*
  • * Aside from the group differences identified in BA 2 and 40, listed above, there were no voxels that passed the threshold for amplitude at P ≤ 0.01.

Discussion

The main findings of this study, designed to gain information on training-dependent changes in neural substrates underlying performance of simple, directionally specific thumb movements in patients with multiple sclerosis, were (i) preceding training, a greater activation in contralateral PMd (BA 6) in multiple sclerosis patients than in controls; and (ii) following training, less pronounced task-specific reductions in activation in contralateral S1 and IPL (BA 40) with a similar trend in M1 in patients than in controls, areas involved in monitoring movement kinematics and cortical motor output.

Motor performance and kinematics

Motor training consisted of the stereotyped repetition of simple thumb movements in a specific direction. This type of training leads to storage of a memory trace in the primary motor cortex that encodes the kinematic details of the practised movements (Classen et al., 1998; Butefisch et al., 2000, 2002; Sawaki et al., 2002a,b, 2003a), a process that may be regarded as a short-term memory for movement and may be the first step in skill acquisition (Classen et al., 1998). Such training does not lead to measurable performance improvements in the ability to move the thumb, as measured by the magnitude of the first-peak acceleration or the angular dispersion of voluntary thumb movements, probably because of the simplicity of the task (Classen et al., 1998; Butefisch et al., 2000, 2002; Sawaki et al., 2002a,b, 2003a). Mechanisms operating in this form of plasticity include NMDA and muscarinic and α-adrenergic receptor function, as well as GABAergic neurotransmission (Classen et al., 1998; Butefisch et al., 2000; Sawaki et al., 2002a). These mechanisms decay with age (Sawaki et al., 2003b) and can be enhanced by pharmacological agents (Butefisch et al., 2002; Sawaki et al., 2002b, 2003a) and cortical stimulation strategies (Butefisch et al., 2004).

Motor performance, measured by the magnitude of the first-peak acceleration of thumb motions and the angular dispersion of these movements, was comparable across the three experimental periods (before, during and after training) in both patients and controls. Therefore, fatigue or inconsistency in motor performance did not affect motor function in either group and there was no evidence of overt motor learning. We scanned participants during performance of a trained (TF) and an untrained (TE) task involving the same finger to (i) control for non-specific changes dependent on the experimental context and possible attentional shifts; and (ii) characterize activation differences related to the specific direction of the trained thumb movement (TF). In this way, contrast between the two tasks reflects changes in the neural substrates underlying performance of the specifically trained movement.

All patients were able to perform the task, did not have mirror movements as measured with EMG recorded from the non-training hand, and exhibited comparable performance with controls, despite differences in clinical status and disease burden on MRI. It is of note that even the patient with a high EDSS score (patient 5, Table 1) and the two patients with high disease burden on MRI (patients 5 and 8, Table 1) performed appropriately.

Activation patterns preceding motor training

In our study, consistent with previous reports, both patients and controls activated contralateral S1, M1, PMd and IPL (BA 40); bilateral S2 [located in the upper bank of the lateral fissure and extending into the operculum (Disbrow et al., 2000)]; bilateral SMA and pre-SMA; anterior cingulate cortex (BA 32); and ipsilateral cerebellum (Jenkins et al., 1994; Fink et al., 1997).

The main group difference preceding training was a more prominent activation of the left PMd contralateral to the moving hand in the patient group. The reproducibility of this result with two different thumb motions emphasizes the strength of the effect, which cannot be explained by changes in motor performance, because EMG and movement kinematics were carefully monitored. The more prominent activation of the left PMd in patients relative to controls in our study could be related, in part, to the nature of the motor task. Subjects performed repetitive thumb movements that required close attention to movement kinematics, particularly movement direction and speed. The emphasis on accuracy of movement kinematics may have led to a more prominent involvement of the left PMd, because this area is specialized in planning and processing spatial patterns or trajectories of intended movements (Schubotz et al., 2003). The left PMd may have been more strongly involved in the current patient group because of its close interaction with the left parietal association cortex (Cavada and Goldman-Rakic, 1989) and their mutual contribution to body scheme representation in motor programming, essential to the correct execution of directionally specific thumb movements (Kosslyn et al., 1993; Suchan et al., 2002; Schubotz et al., 2003).

In contrast, the right PMd, involved in skill acquisition and storage of motor sequences (Sadato et al., 1996; Kawashima et al., 1998), is more prominently active when subjects perform rapid finger movements (Reddy et al., 2002) or finger opposition sequences (Pantano et al., 2002), possibly because of greater task complexity. It is also possible that the left PMd was more involved because the testing strategy involved contralateral right-hand movements, an interpretation supported by previous neuroimaging and physiological findings in patients with stroke (Weiller et al., 1992, 1993; Mima et al., 2001; Carey et al., 2002).

The PMd, with strong projections to α-motoneuron pools in the spinal cord (Dum and Strick, 1991), is probably involved in functional recovery after brain lesions, such as stroke (Liu and Rouiller, 1999; Johansen-Berg et al., 2002b; Rushworth et al., 2003; Fridman et al., 2004). A similar mechanism could operate in patients with multiple sclerosis, despite the more extensively distributed nature of the lesions relative to those present in stroke. Support for this explanation comes from primate (Liu and Rouiller, 1999) and human (Fridman et al., 2004) studies. Liu and Rouiller (1999) demonstrated that functional recovery of hand movements following a focal infarct of the primary sensorimotor cortex is reversed by inactivation of PMd contralateral to the paretic moving hand. Of note, Frost et al. (2003) documented reorganization in the hand representation of ventral premotor cortex contralateral to a paretic hand in the process of functional recovery from focal M1 lesions.

In humans, recovery of motor function after chronic stroke is associated with enhanced activation of the contralateral PMd (Weiller et al., 1992, 1993; Mima et al., 2001; Carey et al., 2002). More importantly, disruption of activity of the PMd contralateral to the paretic hand in chronic stroke patients with good motor recovery and lesions of the corticospinal tract originating in M1 results in abnormal motor behaviour in the paretic hand (Fridman et al., 2004). Therefore, our results, in combination with these reports, are consistent with the view of a contributory role for PMd in recovery of motor function in patients with multiple sclerosis (Pantano et al., 2002). The lack of correlation of PMd activation with T2 lesion load, EDSS score or pegboard performance could be a consequence of the purposeful exclusion of patients with more severe motor disability (those unable to perform the motor task accurately).

Performance of a motor task results in fMRI activation differences in patients with multiple sclerosis (Lee et al., 1999; Filippi et al., 2002; Pantano et al., 2002; Reddy et al., 2002; Rocca et al., 2002a, 2003c) or myelitis (Rocca et al., 2003b) relative to healthy controls. In general, multiple sclerosis patients appear to recruit a more extensive neuronal network, often bilateral, to maintain proper motor output (Pantano et al., 2002; Reddy et al., 2002; Rocca et al., 2002a). For example, fMRI obtained from one multiple sclerosis patient in association with hand movements early in the disease showed a large volume of activation in both hemispheres that, over time, returned to a normal pattern of activation contralateral to the moved hand. These activation differences in the setting of abnormal N-acetylaspartate (NAA) levels were interpreted as suggestive of cortical reorganization supporting motor output in the presence of persistent axonal injury (Reddy et al., 2000).

Additionally, an index of laterality, measured by the ratio of activation in the two hemispheres, shifted toward increased activation in the hemisphere ipsilateral to the moving hand in patients with higher disease burden (T2 lesion volume; Lee et al., 2000), lower NAA levels (Cifelli and Matthews, 2002) and increased disability (Reddy et al., 2002). It has been argued that the similar ipsilateral activation patterns elicited by passively induced and voluntary movements in patients with multiple sclerosis indicate ‘true’ cortical reorganization in the network processing proprioceptive information, because they cannot be explained by a simple change in the patient's strategy to move a weak hand voluntarily (Robertson and Murre, 1999; Reddy et al., 2002). Overall, these findings are consistent with the view that ipsilateral activation, a feature of cortical reorganization in multiple sclerosis patients, is more prominent in more impaired individuals (Lee et al., 2000; Cifelli and Matthews, 2002) and is influenced by such factors as disease stage (Pantano et al., 2002) or subtype (Rocca et al., 2002b, 2003a).

In our patients, we have not found significant increments in ipsilateral activity relative to controls, most probably because of the stated exclusion of patients with poor motor function. This interpretation is consistent with findings in patients with stroke that document the association of (i) therapy-related improvements in hand function (Johansen-Berg et al., 2002a) and better recovery (Johansen-Berg et al., 2002b; Fridman et al., 2004) with predominantly contralateral activation (see, for example, Figure 4C in Johansen-Berg et al., 2002b); and (ii) beneficial effects of rehabilitative treatments with a shift of activation patterns toward the contralateral affected hemisphere (Rocca et al., 2003a). Therefore, motor function in the paretic hand of stroke patients with good recovery appears to rely predominantly on reorganized activity in the affected hemisphere (Liu and Rouiller, 1999; Frost et al., 2003; Werhahn et al., 2003; Murase et al., 2004; see also Figure 7B in Rushworth et al., 2003) and the same result is postulated in multiple sclerosis.

Activation patterns following motor training

To our knowledge, these results represent the first description of training-dependent changes in the neural substrates underlying performance of a motor task in patients with multiple sclerosis. The main findings of the study were that such training resulted in task-specific decreases in activation in S1, M1 and IPL (BA 40) in controls but not in patients. Direct comparison documented less pronounced task-specific decreases in activation in S1 and IPL (BA 40) in patients than in controls.

Reductions in activation in M1 in healthy volunteers may reflect acquisition of a task-relevant set or routine, a form of training-dependent plasticity (Karni et al., 1995, 1998; Morgen et al., 2004) consistent with previous transcranial magnetic stimulation reports (Classen et al., 1998; Butefisch et al., 2000, 2002; Sawaki et al., 2002a,b, 2003a). Given the close attention to movement kinematics required, particularly movement direction and speed, it is not surprising that this process involved S1 and IPL. Activity in S1 may contribute to storage of sensory information required for monitoring movement kinematics (Harris et al., 2002). The IPL is involved in processing spatial information (Corbetta et al., 1991; Pardo et al., 1991; Jenkins et al., 1994; Clower et al., 1996; Seitz et al., 1997) and may contribute to monitoring direction accuracy of voluntary thumb movements. This attenuation may reflect a shift from effortful to more automated performance (Gabrieli, 1998) or, alternatively, a decrease in the monitoring effort required to maintain proper output, a form of perceptual learning (Ahissar et al., 2001; Harris et al., 2002).

Patients failed to show this attenuation. It is possible that diffuse neuronal damage led to a reduced ability to process spatial information in parietal regions (Corbetta et al., 1991; Pardo et al., 1991; Jenkins et al., 1994; Clower et al., 1996; Seitz et al., 1997) and enhanced efforts to monitor directional accuracy and generate thumb movements (Ahissar et al., 2001; Harris et al., 2002), resulting in persistent activity in IPL and S1. These results are reminiscent of the recruitment of larger cortical areas in patients with stroke relative to controls at the beginning of rehabilitative therapies (Marshall et al., 2000; Calautti et al., 2001; Ward et al., 2003a,b). It is possible that more prolonged training than that implemented in this study would lead to an attenuation of activity in S1 and IPL similar to that seen in controls, a prediction supported by the finding of a progressive decrease of initially widespread activation of brain regions over several sessions in patients with stroke (Ward et al., 2003a).

This proposal would be consistent with the documented decrease in parietal activity with skill acquisition in healthy subjects (Gabrieli, 1998) and may reflect a less efficient use of neuronal resources in patients, a form of delayed perceptual learning (Ahissar et al., 2001; Harris et al., 2002). If so, the lack of training-dependent attenuation in multiple sclerosis patients may indicate a non-specific strategy of the CNS to respond to neural damage independently of the aetiology. We do not know if these findings in multiple sclerosis patients correlate with functional recovery, because our study was cross-sectional and the inclusion criteria excluded patients with poorer function.

On the other hand, our findings may reflect a disease-specific form of plasticity, given that, contrary to stroke, immunopathological mechanisms leading to withdrawal of trophic signals and trans-synaptic degeneration in chronic multiple sclerosis result in diffuse reduction of axonal density, even in brain regions that appear unaffected on MRI (Bjartmar et al., 2003; Bruck and Stadelmann, 2003). It is possible that this diffuse damage underlies the lack of training-dependent attenuation reported in this study. Another difference from stroke is the evolution to a secondary progressive stage characterized by steadily increasing motor disability and lack of remissions, possibly reflecting decreased effectiveness of compensatory CNS resources to cope with increasing deficits (Trojano et al., 2003). Studies of training-dependent changes in neural substrates underlying motor function over time could provide useful information on the ability of the CNS to cope with evolving damage and, possibly, predict disease progression in multiple sclerosis patients.

The absence of task-specific training effects on fMRI activation patterns in the current group of multiple sclerosis patients indicates differences in training-dependent plasticity between patients and healthy subjects that may influence the results of rehabilitative treatments. A more thorough understanding of training-dependent plasticity in multiple sclerosis may help optimize therapies and, thereby, prolong preservation of motor function.

Footnotes

  • * These authors contributed equally to the work

References

View Abstract