Sage Journals: Discover world-class research

Abstract

Background

Motor recovery varies across post-stroke individuals, some of whom require a better rehabilitation strategy. We hypothesized that macrostructural neuroplasticity of the motor control network including the cerebellum might underlie individual differences in motor recovery. Objectives. To gain insight into the macrostructural neuroplasticity after stroke, we examined 52 post-stroke individuals using both the Fugl-Meyer assessment and structural magnetic resonance imaging.

Methods

We performed voxel-based lesion symptom mapping and cross-sectional voxel-based morphometry to correlate the motor scores with the lesion location and the gray matter volume (GMV), respectively. Longitudinal data were available at ~8 and/or 15 weeks after admission from 43 individuals with supratentorial lesions. We performed a longitudinal VBM analysis followed by a multiple regression analysis to correlate between the changes of the motor assessment scores and those of GMV overtime.

Results

We found a cross-sectional correlation of residual motor functioning with GMV in the ipsilesional cerebellum and contralesional parietal cortex. Longitudinally, we found increases in GMV in the ipsilesional supplementary motor area, and the ipsilesional superior and inferior cerebellar zones, along with a GMV decrease in the ipsilesional thalamus. The motor recovery was correlated with the GMV changes in the superior and inferior cerebellar zones. The regaining of upper-limb motor functioning was correlated with the GMV changes of both superior and inferior cerebellum while that of lower-limb motor functioning with the GMV increase of the inferior cerebellum only.

Conclusions

The present findings support the hypothesis that macrostructural cerebellar neuroplasticity is correlated with individual differences in motor recovery after stroke.

Keywords

hemiparesis MRI gray matter volume rehabilitation voxel-based morphometry

Introduction

Post-stroke motor deficits deprive stroke survivors of an active life. Efficient strategies to enhance stroke recovery should be developed to improve the quality of life of stroke survivors. A premise of this important medical challenge includes a proper understanding of the neuroplastic mechanisms underlying stroke recovery.

Multiple factors underlie stroke recovery. The perilesional zones including the ischemic penumbra may gradually regain their functions following shrinkage of the lesions and edema.¹ More recently, we have gained knowledge on reorganization of the neural system (“neuroplasticity”) at the molecular, cellular, and network levels after neural injury.^2,3 At the cellular level, neuroplasticity may involve adult neurogenesis, synaptogenesis (involving pruning), angiogenesis, oligodendrogenesis, and astrogliosis modulation.⁴ In clinical settings, neuroplasticity can be probed with noninvasive neurotechnology such as transcranial magnetic stimulation,⁵ electroencephalography,⁶ and magnetic resonance imaging (MRI).

Neuroplasticity is induced by behavioral experiences in either an adaptive or maladaptive manner.⁷ The adaptive post-stroke neuroplasticity likely involves successful motor relearning enhanced by rehabilitation.^8,9 In the healthy brain, adaptive motor learning recruits the “motor control networks,” which include the primary motor and premotor cortices, supplementary motor areas (SMA), parietal cortex, basal ganglia, thalamus, and cerebellum.¹⁰ Changes in gray matter volumes (GMV) on MRI have been used as a macrostructural measure of learning-induced neuroplasticity in healthy people who underwent motor or cognitive training over a few months.^11-14 The macrostructural neuroplasticity associated with motor learning may in part reflect the formation of the “internal models”¹⁵ underpinning motor learning.

In stroke survivors, both functional^16,17 and structural^18-24 MRI studies have revealed neuroplastic changes in the motor control networks. For example, voxel-based morphometry (VBM) studies (n = 10-28) have reported both decreases and increases in GMV typically between acute (~a week after the onset) and subacute stages (several months) in patients with hemiparesis.^18-22 Decreased GMV was identified in the ipsilesional cortical areas,^18-22 corresponding to “post-stroke cortical atrophy,” and in the contralesional cerebellum,²² corresponding to “crossed cerebellar atrophy.”²⁵ By contrast, GMV increases, which may underpin the adaptive neuroplasticity leading to motor recovery, are only inconsistently described in the premotor cortex and SMA,^19,20 cognitive regions,^18,21,22 and subcortical nuclei,²⁰ probably because of limited sample sizes in those studies.

Here, to advance our understanding of post-stroke motor recovery, we analyzed structural MRI data obtained from subacute stroke patients. We did not rigidly specify the type of stroke for inclusion because of 2 reasons: to recruit a sufficiently large number of subacute stroke patients at a single hospital and to obtain a sufficiently large interindividual variability in motor deficits and recovery courses. We hypothesized that motor control networks including the cerebellum might show macrostructural neuroplasticity induced by motor re-learning after stroke. A cross-sectional data analysis revealed the ipsilesional cerebellum as the potential substrates supporting residual motor functioning after stroke. In a longitudinal data analysis, we hence paid particular attention to testing the macrostructural neuroplasticity in the ipsilesional cerebellum, given its role in motor learning¹⁵ among other motor and non-motor functions. We tested correlational relationship between the degree of motor recovery correlated with the level of motor recovery as indexed by the Fugl-Meyer assessment (FMA) and GMV changes in the brain regions (including the cerebellum) that showed macrostructural neuroplasticity after stroke.

Methods

Participants

Sixty-one patients with stroke at a subacute stage agreed to participate in this study upon admission to the Tokyo Metropolitan Rehabilitation Hospital. All participants provided written informed consent according to the study protocol, which was approved by the local ethics committee (Tokyo Metropolitan Rehabilitation Hospital). Inclusion criteria were: (1) clinically first-ever unilateral ischemic or hemorrhagic stroke (subarachnoid hemorrhage excluded), (2) transferred to the hospital for rehabilitation within 2 months after onset, and (3) no obvious cognitive impairment. Exclusion criteria were: (1) contraindication for physiotherapy, (2) contraindication for MRI, and (3) inability to provide a consent because of a cause other than aphasia. Research data were available from 54 post-stroke patients (mean age 62 ± 12 years, range 41-83 years) out of 61 patients (Table 1). The primary symptom at admission was right hemiparesis (n = 18), left hemiparesis (n = 28), right sensory disturbance (n = 1), left sensory disturbance (n = 1), right ataxia (n = 2), left ataxia (n = 1), dysphagia (n = 1), aphasia (n = 1), and left visual disturbance (n = 1).

Table 1.

Symptoms and the Sites of Lesions.

Symptoms	Participants T1/T2/T3	Lesions
Symptoms	Participants T1/T2/T3	Subcortical infarction	Cortical infarction	Brainstem infarction	Putaminal hemorrhage	Thalamic hemorrhage	Other hemorrhage
R paresis	18/15/8	11	1	1	4	1	0
L paresis	28/26/15	12	2	4	5	4	1
Others	8/6/0	0	1	6	0	0	1
Total	54/47/23	23	4	11	9	5	2

The participants T1/T2/T3 represent the number of participants available for the assessment at the at the admission (T1), 8 weeks later (T2), and before discharge (T3).

Study Design

We designed a 16-week longitudinal study, during which post-stroke patients underwent 3 assessments after admission to a single rehabilitation hospital. A rehabilitation program was administered to all patients. The content of the program was tailored to each patient’s condition upon the agreement of the in-charge rehabilitation team. Each patient underwent the tailor-made rehabilitation program for 2 or 3 hours a day. When possible, the assessments were performed at admission (T1), ~8 weeks later (T2), and before discharge (T3). Fifty-two participants completed both full motor assessment and MRI at least once, providing data for a cross-sectional VBM analysis. We also used a voxel-based lesion-symptom mapping (VLSM) analysis²⁶ to characterize the distribution of brain damage in the enrolled patients. The second MRI (~8 weeks later) was available from 46 participants while the third MRI was available from 25 participants before discharge (Figure 1). The cross-sectional VMB analysis identified the ipsilateral cerebellum, among the nodes of the motor control networks, as the correlates supporting residual motor functioning after stroke. We considered that the inclusion of data from patients with brainstem or cerebellar lesions would make the interpretation of the findings difficult in the longitudinal data analysis. We thus limited the longitudinal VBM analysis to patients with supratentorial regions (n = 43). Longitudinal assessment of both GMV and motor deficit was available from 41 patients.

Figure 1.

The flow of analyses. A total of 52 patients completed the assessment with both clinical batteries and MRI at least once. We limited the longitudinal VBM analysis to patients with supratentorial regions (n = 43). Longitudinal assessment of both GMV and motor deficit was available from 41 patients (62 data pairs).

Motor Assessment

As a measure of motor recovery, we used the FMA, which is a composite scale of stroke-related impairment with 5 domains: motor functioning (upper and lower limbs), sensory functioning, balance, range of joint motion, and pain. In this study, we mainly used the motor domain of the FMA (mFMA; maximum score = 100 points), which combined the score of the upper limb section (maximum score = 66 points) and lower limb section (maximum score = 34 points). We also analyzed the sensory FMA score (maximum score = 24 points) to supplement the findings. The FMA score was obtained by in-charge physicians who are specialized in rehabilitation medicine.

MRI Acquisition

MRI was conducted on a 3-Tesla scanner (Discovery 750 W, GE Healthcare, USA) equipped with a 24-channel head receiver coil. Three-dimensional T1-weighted (3D-T1W) structural MRI was obtained with a time-of-flight spoiled gradient echo sequence (repetition time = 7.66 ms, echo time = 3.10 ms, flip angle = 12°, and voxel size = 1.0 × 1.0 × 1.2 mm³).

MRI Preprocessing

SPM12 (http://www.fil.ion.ucl.ac.uk/spm) implemented on MATLAB (Mathworks, Inc., Natick, MA, USA) was used for MRI preprocessing. Because of marked differences in the locations and image properties of the lesions, the lesion masks were delineated manually by a neurologist with expertise in neuroimaging (TH), using MRIcron (https://www.nitrc.org/projects/mricron). The region masks were delineated slice-by-slice on 3D-T1W MRI and concatenated into a 3D volume. The lesion masks were used to create a lesion overlap map as a summary of the lesion location (Figure 2A). The lesion masks also enabled the use of cost-function masking in the spatial normalization process to the Montreal Neurological Institute space,²⁷ using the Clinical Toolbox (https://www.nitrc.org/projects/clinicaltbx/). The spatially normalized images were flipped along the anterior-posterior axis, so that the supratentorial lesion side was on the left side of the brain in all cases.

Figure 2.

Lesion overlap map and VLSM. (A) The colored region indicates the overlap of lesions across the patients. (B) VLSM. The colored areas indicate the lesions correlated with the level of motor impairment. The lesions are flipped to locate the supratentorial lesions to the left side.

Voxel-Based Lesion-Symptom Mapping

For VLSM, we included the datasets in which both structural MRI and mFMA were available at least once at the same timepoint (52 patients). The search volume was defined as areas with >5% overlap across the participants (see Figure 2A). In each voxel, VLSM tested the effects of having a lesion and those having no lesion on mFMA. Statistical inference was based on the Brunner–Munzel rank order with 1000 permutations (P < .05 with multiple comparisons within the search volume), using the non-parametric mapping toolbox in MRIcron.

Cross-Sectional VBM Analysis

For the cross-sectional VBM analysis, we examined a correlation between GMV and the level of motor impairment. To this end, we included the datasets for which both structural MRI and mFMA were available at least once. With this criterion, 118 data sets from the 52 patients were included. The analysis was performed using the computational anatomy toolbox (CAT12) and SPM12. The cost-function masking-normalized T1W MRI was segmented into GM, white matter (WM), cerebrospinal fluid, and non-brain voxels. We then used diffeomorphic anatomical registration using exponentiated Lie algebra (DARTEL) to create a study-specific GM template. The GM from each patient was further normalized to the study-specific template. The normalized GM image had a voxel size of 1.5 × 1.5 × 1.5 mm³. After spatial smoothing with a Gaussian kernel of 8-mm full width at half maximum, a generalized linear model (GLM) analysis was performed, with patients treated as a between-subject variable. In this cross-sectional analysis, mFMA was treated as a covariate of interest whereas age and total intracranial volume were considered covariates of no interest. Sphericity correction was performed to deal with dependency across the times. The correlation between GMV and mFMA was tested at P < .05 corrected for multiple comparisons.

Longitudinal VBM Analysis

To gain deeper insight into the findings from the cross-sectional analysis (see Results section), we studied the patients with supratentorial lesions who underwent MRI at least twice. From 43 patients who met these criteria, the data from T1 (n = 43), T2 (n = 41), and T3 (n = 24) were entered into a whole-brain generalized linear model analysis, treating participants as a between-subject variable and the times (T1, T2, and T3) as a within-subject variable. Age, total intracranial volume, and the stroke type (infarction or hemorrhage) were included in the design matrix as covariates. Sphericity correction was performed. Changes in GMV across the times were tested with F-statistics at P < .05 corrected for multiple comparisons.

Correlation Between Motor Recovery and Brain Structural Changes

To examine the temporal profile of GM changes overtime, GMV values were retrieved from each of the identified clusters in the longitudinal VBM analysis. A multiple regression model with the ordinary least square was then used to test whether the difference in GMV between the time points (ΔGMV) accounted for the corresponding difference in mFMA (ΔmFMA) as a dependent variable, using the python statsmodels. The explanatory variables included the age and the stroke type (infarction or hemorrhage) in addition to the ΔGMV. We also ran multiple regression analyses using the same set of explanatory variables and ΔmFMA of the upper limb and lower limb separately as a dependent variable. The correlation was also tested for the ΔFMA sensory scale. The variance inflation factor was between 1 and 2 in all the explanatory variables, excluding problematic collinearities. The model fit was assessed using the adjusted R², F-statistics, and Akaike and Bayesian Information Criteria (AIC and BIC, respectively). The AIC and BIC provide metrices to compare across different models to infer which model fits better (no thresholding value). The significance of the parameter estimates (β) was assessed using 1-sample t-tests for each explanatory variable. For the multiple regression analysis, we used 62 data pairs from 41 patients because ΔmFMA was not available from 2 patients. To visualize the association of ΔmFMA with ΔGMV in the areas where we found significant association in the first-round regression analyses, we computed the residuals after removing the effects of other regressors. This second-round multiple regression analyses yielded ΔGMV corrected for the effects of non-interest for each plot (corrected ΔGMV). The ΔmFMA (total, upper limb, or lower limb) was plotted against the corrected ΔGMV for each region.

Results

Motor Assessment

The mFMA was 62.3 ± 29.2 (mean ± standard deviation) at T1 (41 ± 15 days from stroke onset), 70.9 ± 27.0 at T2 (55 ± 6 days from T1), and 65.6 ± 27.7 at T3 (109 ± 22 days from T1). The upper-limb mFMA was 38.5 ± 23.9 at T1, 43.2 ± 23.0 at T2, and 38.9 ± 22.4 at T3 and lower-limb mFMA was 23.8 ± 7.8 at T1, 27.8 ± 6.0 at T2, and 26.7 ± 6.6 at T3.

Lesion Overlap Map and VLSM

The stroke lesions varied across the participants but were most consistently located in the basal ganglia, thalamus, and WM involving the internal capsule, corona radiata, and centrum semiovale (Figure 2A). In the VSLM analysis, mFMA was correlated with WM lesions involving the internal capsule and the corona radiata (Figure 2B).

Cross-Sectional VBM

In the cross-sectional VBM, mFMA was correlated with GMV in the ipsilesional cerebellum and the contralesional parietal cortex (Figure 3, Table 2A).

Figure 3.

Cross-sectional voxel-based morphometry (VBM) analysis. The warm colored areas indicate regions where gray matter volume (GMV) was correlated with residual motor ability (family-wise error corrected P < .05). The superior cerebellar region is on the ipsilesional side (IL), while the parietal region is on the contralesional side (CL). The color bar represents the t-value. Below, the GMV-derived parameters retrieved from the peak voxels of the IL cerebellum and CL parietal cortex are plotted against the mFMA. Gray dots represent the SPM predicted fitted responses plus errors and the black lines regression lines.

Table 2.

Voxel-Based Morphometry (VBM) Analysis.

A. Cross-sectional
Cluster	Side	Size (mm³)	Peak T value	P _FWE	MNI coordinate			Nomenclature
Cluster	Side	Size (mm³)	Peak T value	P _FWE	x	y	z	Nomenclature
Superior cerebellum	IL	5711	4.42	.00069	−18	−58	−16	Lobule VI
Parietal cortex	CL	6230	4.57	.00040	44	−33	42	SMG-SPL
B. Longitudinal VBM analysis.
Cluster	Side	Size (mm³)	Peak F-value	P _FWE	MNI coordinate			Nomenclature
Cluster	Side	Size (mm³)	Peak F-value	P _FWE	x	y	z	Nomenclature
Decreased
Thalamus	IL	2670	24.7	.0002	−15	−20	2	Thalamus
Thalamus	IL	2670	20.8	.0017	−16	−18	10	Thalamus
Increased
SMA	IL	176	20.2	.0023	−8	6	64	SMA
Superior cerebellum	IL	142	16.7	.017	−22	−63	−20	Lobule VI
		81	16.6	.018	−16	−45	−24	Lobule V
		27	15.6	.03	−20	−75	−32	Crus1
Inferior cerebellum	IL	125	17.7	.0095	−9	−68	−48	VIIIA
Inferior cerebellum	IL	7	15.2	.04	−30	−48	−48	VIIIA

Abbreviations: MNI, montreal neurological institute; IL, ipsilesional; CL, contralesional; SMG, supramarginal gyrus; SPL, superior parietal lobule; SMA, supplementary motor areas.

Longitudinal VBM

For the patients available for the longitudinal VBM analysis, the mean mFMA was 57.0 ± 27.9 at T1, 68.8 ± 26.2 at T2, and 66.3 ± 27.5 at T3, which were significantly different across the time points (F_[2,60] = 26.2, P < .001, partial η² = .47; Figure 4A). The mean mFMA 33.8 ± 23.2 at T1, 40.8.8 ± 22.7 at T2, and 39.3.3 ± 22.3 at T3 (F_[2,60] = 14.2, P < .001, partial η² = .32) for the upper limb section and 22.5 ± 7.4 at T1, 27.1 ± 5.9 at T2, and 26.7 ± 6.5 at T3 for the lower limb section (F_[2,60] = 28.9, P < .001, partial η² = .49).

Figure 4.

Longitudinal voxel-based morphometry (VBM) analysis. (A) Changes in the motor impairment (mFMA) for the total mFMA (UL&LL), mFMA of the upper limb (UL), and the lower limb (LL) across admission (T1, n = 43), second assessment (T2, n = 41), and third assessment (T3, n = 23). (B) F statistics identified 4 clusters with significant gray matter volume changes over time: the supplementary motor areas (SMA), thalamus, superior cerebellum (Sup cbll), and the inferior cerebellum (Inf cbll). For display purposes, the magenta areas show the clusters with peak an F value > 14.8 (P < .05 corrected for family wise error), while the cyan areas show the extent of the cluster with an F value > 7.7 (uncorrected P < .001). (C) Changes in GMV (arbitrary unit, a.u.) over time in stroke recovery. The error bars indicate the standard error of the mean. (D) The corrected ΔGMV (a.u.), which was ΔGMV after removing the effects of other covariates, is plotted against the ΔmFMA, ΔmFMA of the upper limb (ΔULmFMA) and the lower limb (ΔULmFMA) in the Sup cbll (black) and Inf cbll (gray).

In the longitudinal VBM analysis (Figure 4B, Table 2B for details), increased GMV was observed in the ipsilesional SMA (partial η² = .35 at the peak), and the ipsilesional superior (partial η² = .36) and inferior (partial η² = .39) cerebellum, while decreased GMV was observed in the ipsilesional thalamus (partial η² = .38).

Correlation Between ΔmFMA and ΔGMV

We tested whether a multiple regression model with age, stroke type (hemorrhage or infarction), and the GMV changes (ΔGMV) as explanatory variables accounted for the mFMA changes across the tine points (ΔmFMA). The multiple regression model revealed a significant fit with the data (adjusted R² = .23, partial η² = .30, P = .0025 by F_[6,55] = 3.92, AIC = 167.9, BIC = 182.8). The ΔmFMA showed correlation with the ΔGMV in the superior cerebellum (β = −.50, t = −3.46, p = .001) and the inferior cerebellum (β = .30, t = 2.03, P = .047), but not with the SMA ΔGMV (β = .23, t = 1.07, P = .095), the thalamus ΔGMV (β = −.23, t = −1.66, P = .102), age (β = −.08, t = −0.68, P = .50), or the stroke type (β = .13, t = 1.02, P = .30). The ΔmFMA showed a strong correlation with the ΔGMV after correcting for the effects other variables (corrected ΔGMV in Figure 3D, upper panel) in both the superior cerebellum (r = .93) and the inferior cerebellum (r = .89).

The multiple regression model fitted with the motor recovery scores when it was separately applied to the ΔmFMA of the upper limb (adjusted R² = .185, partial η² = .27, P = .007 by F_[6,55] = 3.31, AIC = 170.8, and BIC = 185.7),the lower limb (adjusted R² = .157, partial η² = .24, P = .016 by F_[6,55] = 2.89, AIC = 172.9, and BIC = 187.8). The upper-limb ΔmFMA was correlated with both the superior cerebellar ΔGMV (β = −.50, t = −3.47, P = .001) and the inferior cerebellar ΔGMV (β = .30, t = 2.03, P = .047). The lower-limb ΔmFMA was correlated only with the inferior cerebellar ΔGMV (β = .41, t = 2.71, P = .009). Consistently, the upper-limb ΔmFMA showed a strong correlation with the adjusted ΔGMV (Figure 3D, middle panel) in both the superior cerebellum (r = .95) and the inferior cerebellum (r = .89). The lower-limb ΔmFMA the was correlated with the adjusted ΔGMV (Figure 3D, lower panel) strongly in the inferior cerebellum (r = .89), but only weakly in the superior cerebellum (r = .31), in agreement with the multiple regression analysis.

To interrogate if these longitudinal brain structural changes were also associated with sensory recovery, we tested the multiple regression model using the ΔFMA sensory scale (19.3 ± 4.5 at T1, 21.1 ± 3.8 at T2, and 21.4 ± 4.4 at T3), which differed across the time points (F_[2,60] = 10.6, P < .001, partial η² = .26). The multiple regression model however failed to explain the ΔFMA sensory scale with the ΔGMV (adjusted R² = .15, partial η² = .15, P = .146 by F_[6,55] = 1.67, AIC = 179.6, and BIC = 194.5).

Discussion

We analyzed structural MRIs and FMA in subacute poststroke patients from multiple perspectives. The lesion overlap map and the VLSM analysis indicated that the WM lesions were primarily responsible for the variance of motor deficits in our patients. The cross-sectional VBM analysis revealed correlation of their residual motor ability with GMV in the ipsilesional cerebellum and the contralesional parietal cortex. The longitudinal VBM analysis revealed decreases in GMV in the ipsilesional thalamus and increases in GMV in the “motor control networks” including SMA and the ipsilesional superior and inferior cerebellum. Moreover, we found that the GMV changes of the superior and inferior cerebellum were correlated with motor recovery consistent with the “double motor representations” of the cerebellum. The upper-limb ΔmFMA was correlated with both superior and inferior cerebellar ΔGMV while the lower-limb ΔmFMA was correlated only with the inferior cerebellar ΔGMV. The novel evidence of the macrostructural cerebellar neuroplasticity underlying stroke recovery supports our a priori hypothesis about the role of the motor control network, especially the cerebellum, in motor relearning underlying stroke recovery.

Post-Stroke Atrophy

The reduction of GMV from several months to a year after the stroke onset was previously reported in the thalamus^22,28 and the basal ganglia.^18,19 In the longitudinal VBM analysis, we also found significant decreases in GMV in the thalamus. These findings fit with the concept of accelerated brain atrophy in the first 3 months after stroke.²⁸ Conversely, Abela et al²⁰ found increased, rather than decreased, GMV in the ipsilesional basal ganglia and thalamus between 3 and 9 months after the stroke onset. Note, however, that the study by Abela et al²⁰ examined stroke involving the motor cortex, which is likely associated with different recovery mechanisms than that for the subcortical damage mainly observed in the present cohort.

Macrostructural Cerebellar Neuroplasticity Correlated With Motor Recovery

The cross-sectional and longitudinal VBM analyses consistently revealed that macrostructural neuroplasticity of the ipsilesional cerebellum might underlie motor recovery after stroke. This finding deemed particularly important because of the correlation between the ΔmFMA and ΔGMV in the ipsilesional cerebellum. Previous studies have reported inconsistent findings on post-stroke macrostructural neuroplasticity in the cerebellum. Decreased GMV in the contralesional cerebellum has been widely reported after stroke.^18-22 Notably, Yu et al²² found that the atrophy progression of the contralesional cerebellum was correlated with poor motor recovery. This phenomenon is termed the cross cerebellar atrophy coupled with ipsilesional supratentorial atrophy.²⁵ A similar cerebellar GMV reduction was reported in patients with pontine infarction, which likely disrupted the mossy fibers to the cerebellum.²⁹ Nevertheless, Fan et al¹⁸ reported a moderate increase in GMV in the ipsilesional cerebellum (lobule VI) after stroke, although this cerebellar GMV change was not correlated with motor recovery.

The role of the cerebellum in motor control and learning is most well established among various functions of the cerebellum.³⁰ Anatomically, the cerebellar hemisphere is subdivided into the anterior lobe (lobules I-VI) and posterior lobe (lobules VII-X), which are separated by the primary fissure. The cerebellum has 2 separate whole-body motor representations (double motor representations): one in the anterior lobe (lobules IV and V) and the other in the posterior lobule (including lobule VIII).^31,32 In our cross-sectional analysis, mFMA was correlated with GMV in the anterior cerebellar lobule V and VI. This zone coincided with the superior cerebellar zone (Supplemental Material Information), which showed macrostructural neuroplasticity in the longitudinal VBM analysis. Additionally, we found macrostructural neuroplasticity in the inferior cerebellar zone, corresponding to the lobule VIIIA in the posterior lobe. The present superior and inferior cerebellar zones agree with double motor representations in the anterior and posterior–inferior lobes of the cerebellum.^31,32 The correlation of upper limb ΔmFMA with both superior and inferior cerebellar ΔGMV is consistent with the wide upper-limb representations in the cerebellum. Notably, we found association between ΔmFMA of lower limb and ΔGMV only of the inferior cerebellum, which mainly represents lower limb movement.³¹

How can the neuroplastic changes of the motor cerebellum support stroke recovery? Only a few among the previous longitudinal VBM studies^18-24 explicitly reported rehabilitation during the follow-up period. The participants in the present study underwent intensive in-hospital rehabilitation programs in the subacute stage of stroke. Thus, the neuroplastic changes would include the effects of both spontaneous recovery and rehabilitation. Consistent with this concept, the superior cerebellar zone corresponds to the cerebellar region, where previous task functional MRI (fMRI) studies found increased motor task-related activity after rehabilitation training.^16,17 Luft et al¹⁷ also reported increased brain activity in the ipsilesional cerebellum and contralesional frontoparietal cortices after 6-week training with rhythmic cueing. Furthermore, Johansen-Berg et al¹⁶ found increased activity in the bilateral cerebellum and the contralesional frontoparietal cortices after a 2-week constrained-induced therapy. These studies of longitudinal task fMRI combined with rehabilitative interventions support a role for the ipsilesional cerebellum and contralesional frontoparietal cortex in motor retraining. Note, however, that effective rehabilitation training should influence motor performance, which in turn modulates brain activity during fMRI. Thus, it is difficult for these task fMRI studies alone to segregate between neuroplastic changes and the resulting improvement in motor performance. Given that structural measurement is not influenced by task performance at measurement, our findings complement these fMRI studies^16,17 to support a role for cerebellar neuroplasticity in rehabilitation-mediated motor recovery.

Numerous human neuroimaging studies have reported changes in cerebellar activity and connectivity in association with motor or procedural learning, even in older adults.^15,33,34 Training-induced increases in cerebellar GMV have also been reported in patients with cerebellar degeneration (2-week postural training)³⁵ and in healthy people (2-month video-game training).³⁶ These studies provide empirical evidence supporting the concept that cerebellar GMV changes reflect the formation of new cerebellar networks, corresponding to the “internal models” or feedforward controllers of newly learned movement or procedure.^15,37 Hence, a possible interpretation of the cerebellar macrostructural neuroplasticity is the formation of internal models, corresponding to the local neural circuits encoding relearned motor patterns and procedures.

Ipsilesional SMA and Contralesional Parietal Cortex

It is rather unlikely that the cerebellum alone is responsible for motor recovery. The SMA, which plays an important role in motor learning, showed neuroplastic changes in the present study. This is reasonable because the SMA was preserved in our patient group (Figure 2A). Additionally, the contralesional parietal cortex correlated with residual motor ability. Thus, the fronto-parietal cortex is potentially involved in stroke recovery. The cerebellum is thought to be involved in error-based learning by receiving inputs from the inferior olivary nucleus (climbing fibers) and the cerebral cortex via the pontine nucleus (mossy fibers). After processing in the cerebellar local circuits, the cerebellar output projects to the cerebral cortex via the thalamus. Hence, cerebellar neuroplasticity should accompany neuroplasticity in the cortico-cerebellar circuits. Indeed, the integrity of cortico-cerebellar connectivity was shown to reflect residual motor performance in post-stroke patients.²⁴

Limitations

Longitudinal data were not available from all patients. Furthermore, mFMA at T3 tended to be lower than that at T2, indicating that more severely impaired patients required longer in-hospital rehabilitation. The heterogeneity of our study cohort was also a limitation, although this may have introduced adequate data variations, thus helping to detect findings that were probably underestimated in previous studies. In relation to the heterogeneity issue, we needed to flip the lesion side to obtain clear results; thus, we cannot discuss differences in the effects of the lesion side on motor recovery. We also failed to find the GMV changes correlated with sensory recovery probably because we mainly recruited patients with motor impairment, resulting in small variations in the measurements correlated with the sensory score. A larger and more comprehensive dataset is required to fully explore all the aspects of, and potential contributing factors to, the mechanisms underlying recovery after stroke. An important step toward this is to form a multi-site consortium such as the ENIGMA Stroke Recovery Working Group.³⁸

The neurobiological mechanisms underlying macrostructural neuroplasticity remain unclear.³⁹ At the cellular level, the macrostructural neuroplasticity is supposed to involve axonal sprouting, synaptogenesis (also involving pruning), angiogenesis, oligodendrogenesis, and modulation of astroglia. There is still a gap between these cellular level mechanisms and GMV increases or decreases; researchers are now trying to fill the gap. In an animal model study, rehabilitative training induced cerebellar GM changes involving synaptogenesis and increased glial cell size.⁴⁰ We have also established an animal model in which operant conditioning induces macrostructural neuroplasticity in the cerebellum.⁴¹ Using this model, we are currently examining the neurobiological mechanisms associated with cerebellar GMV changes during motor training.

Conclusions

The present study supported that the involvement of the macrostructural neuroplasticity of the motor control network, the cerebellum in particular, underlies motor recovery after stroke. The macrostructural neuroplasticity likely corresponds to the formation of the internal models, thereby generally supporting the statement that “Thus rehabilitation needs to emphasize techniques that promote formation of appropriate internal models and not just repetition of movement.⁸” Theoretically, this process is achieved through motor learning relying on both motor commands and sensory information about the motor outcomes. Hence, rehabilitation techniques should emphasize both the motor efferents and sensory afferents.

Supplemental Material

sj-docx-1-nnr-10.1177_15459683231207356 – Supplemental material for Macrostructural Cerebellar Neuroplasticity Correlates With Motor Recovery After Stroke

Supplemental material, sj-docx-1-nnr-10.1177_15459683231207356 for Macrostructural Cerebellar Neuroplasticity Correlates With Motor Recovery After Stroke by Takashi Hanakawa, Fujiko Hotta, Tatsuhiro Nakamura, Keiichiro Shindo, Naoko Ushiba, Masaki Hirosawa, PT, MS, Yutaka Yamazaki, OT, BS, Yoshinao Moriyama, PT, PhD, Syota Takagi, PT, BS, Katsuhiro Mizuno and Meigen Liu in Neurorehabilitation and Neural Repair

Footnotes

Acknowledgements

We thank Edanz () for editing a draft of this manuscript.

Author Contributions

T. Hanakawa, F. Hotta, K. Mizuno, and M. Liu designed the study; F. Hotta, K. Shindo, N. Ushiba, M. Hirosawa, Y. Yamazaki, Y. Sato, and S. Takagi collected the data; T. Hanakawa and T. Nakamura analyzed the data; T. Hanakawa drafted the paper. All authors have been involved in revising this manuscript for intellectual content and have approved the final version of the manuscript.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported in part by the Japan Agency for Medical Research and Development (17dm0107033h0005, 18 dm0207070, 18 dm0307003) and KAKENHI (19H03536 and 19H05726) from the Japan Society for the Promotion of Science and Japan Science and Technology Agency, CREST, grant number JPMJFR206G to TH.

ORCID iD

Takashi Hanakawa

Supplementary material for this article is available on the Neurorehabilitation & Neural Repair website along with the online version of this article.

References

Fisher

Bastan

. Identifying and utilizing the ischemic penumbra. Neurology. 2012;79(13 suppl. 1):79-86. doi:10.1212/WNL.0b013e3182695814

Murphy

Corbett

. Plasticity during stroke recovery: from synapse to behaviour. Nat Rev Neurosci. 2009;10(12):861-872. doi:10.1038/nrn2735

Joy

Carmichael

. Encouraging an excitable brain state: mechanisms of brain repair in stroke. Nat Rev Neurosci. 2021;22(1):38-53. doi:10.1038/s41583-020-00396-7

Otero-Ortega

Gutiérrez-Fernández

Díez-Tejedor

. Recovery after stroke: new insight to promote brain plasticity. Front Neurol. 2021;12:768958. doi:10.3389/fneur.2021.768958

Fridman

Hanakawa

Chung

Hummel

Leiguarda

Cohen

. Reorganization of the human ipsilesional premotor cortex after stroke. Brain. 2004;127(4):747-758. doi:10.1093/brain/awh082

Kawano

Hattori

Uno

, et al. Electroencephalographic phase synchrony index as a biomarker of poststroke motor impairment and recovery. Neurorehabil Neural Repair. 2020;34(8):711-722. doi:10.1177/1545968320935820

Nudo

. Recovery after brain injury: mechanisms and principles. Front Hum Neurosci. 2013;7:887. doi:10.3389/fnhum.2013.00887

Krakauer

. Motor learning: its relevance to stroke recovery and neurorehabilitation. Curr Opin Neurol. 2006;18(1):84-90.

Maier

Ballester

Verschure

PFMJ

. Principles of neurorehabilitation after stroke based on motor learning and brain plasticity mechanisms. Front Syst Neurosci. 2019;13:74. doi:10.3389/fnsys.2019.00074

10.

Lohse

Wadden

Boyd

Hodges

. Motor skill acquisition across short and long time scales: a meta-analysis of neuroimaging data. Neuropsychologia. 2014;59(1):130-141. doi:10.1016/j.neuropsychologia.2014.05.001

11.

Draganski

Gaser

Kempermann

, et al. Temporal and spatial dynamics of brain structure changes during extensive learning. J Neurosci. 2006;26(23):6314-6317. doi:10.1523/JNEUROSCI.4628-05.2006

12.

Draganski

Gaser

Busch

Schuierer

Bogdahn

May

. Changes in grey matter induced by training Newly honed juggling skills show up as a transient feature on a brain-imaging scan . Nature. 2004;427:311-312. doi:10.1038/427311a

13.

Hosoda

Tanaka

Nariai

Honda

Hanakawa

. Dynamic neural network reorganization associated with second language vocabulary acquisition: a multimodal imaging study. J Neurosci. 2013;33(34):13663-13672. doi:10.1523/JNEUROSCI.0410-13.2013

14.

Hosoda

Tsujimoto

Tatekawa

Honda

Osu

Hanakawa

. Plastic frontal pole cortex structure related to individual persistence for goal achievement. Commun Biol. 2020;3(1):194. doi:10.1038/s42003-020-0930-4

15.

Kim

Ogawa

Schweighofer

Imamizu

. Neural substrates related to motor memory with multiple timescales in sensorimotor adaptation. PLoS Biol. 2015;13(12):e1002312. doi:10.1371/journal.pbio.1002312

16.

Johansen-Berg

Dawes

Guy

Smith

Wade

Matthews

. Correlation between motor improvements and altered fMRI activity after rehabilitative therapy. Brain. 2002;125(12):2731-2742. doi:10.1093/brain/awf282

17.

Luft

McCombe-Waller

Whitall

, et al. Repetitive bilateral arm training and motor cortex activation in chronic stroke: a randomized controlled trial. JAMA. 2004;292(15):1853-1861. doi:10.1001/jama.292.15.1853

18.

Fan

Zhu

Chen

, et al. Dynamic brain structural changes after left hemisphere subcortical stroke. Hum Brain Mapp. 2013;34(8):1872-1881. doi:10.1002/hbm.22034

19.

Dang

Liu

Xing

, et al. Longitudinal cortical volume changes correlate with motor recovery in patients after acute local subcortical infarction. Stroke. 2013;44(10):2795-2801. doi:10.1161/STROKEAHA.113.000971

20.

Abela

Seiler

Missimer

, et al. Grey matter volumetric changes related to recovery from hand paresis after cortical sensorimotor stroke. Brain Struct Funct. 2015;220(5):2533-2550. doi:10.1007/s00429-014-0804-y

21.

Cai

Xin

, et al. Contralesional cortical structural reorganization contributes to motor recovery after sub-cortical stroke: a longitudinal voxel-based morphometry study. Front Hum Neurosci. 2016;10:8. doi:10.3389/fnhum.2016.00393

22.

Yang

Song

, et al. Changes in structure and perfusion of grey matter tissues during recovery from Ischaemic subcortical stroke: a longitudinal MRI study. Eur J Neurosci. 2017;46(7):2308-2314. doi:10.1111/ejn.13669

23.

Takenobu

Hayashi

Moriwaki

Nagatsuka

Naritomi

Fukuyama

. Motor recovery and microstructural change in rubro-spinal tract in subcortical stroke. NeuroImage Clin. 2014;4:201-208. doi:10.1016/j.nicl.2013.12.003

24.

Schulz

Frey

Koch

, et al. Cortico-cerebellar structural connectivity is related to residual motor output in chronic stroke. Cereb Cortex. 2017;27(1):635-645. doi:10.1093/cercor/bhv251

25.

Tien

Ashdown

. Crossed cerebellar diaschisis and crossed cerebellar atrophy: correlation of MR findings, clinical symptoms, and supratentorial diseases in 26 patients. Am J Roentgenol. 1992;158(5):1155-1159. doi:10.2214/ajr.158.5.1566683

26.

Gitelman

Levy

Hulvershorn

Parrish

. Identification of critical areas for motor function recovery in chronic stroke subjects using voxel-based lesion symptom mapping. Neuroimage. 2010;49(1):9-18. doi:10.1016/j.neuroimage.2009.08.044

27.

Brett

Leff

Rorden

Ashburner

. Spatial normalization of brain images with focal lesions using cost function masking. Neuroimage. 2001;14(2):486-500. doi:10.1006/nimg.2001.0845

28.

Brodtmann

Khlif

Egorova

Veldsman

Bird

Werden

. Dynamic regional brain atrophy rates in the first year after ischemic stroke. Stroke. 2020;51:e183-e192. doi:10.1161/STROKEAHA.120.030256

29.

Wei

Wang

Liu

, et al. Progressive gray matter atrophy and abnormal structural covariance network in ischemic pontine stroke. Neuroscience. 2020;448:255-265. doi:10.1016/j.neuroscience.2020.08.033

30.

Stoodley

Schmahmann

. Functional topography in the human cerebellum: a meta-analysis of neuroimaging studies. Neuroimage. 2009;44(2):489-501. doi:10.1016/j.neuroimage.2008.08.039

31.

Mottolese

Richard

Harquel

Szathmari

Sirigu

Desmurget

. Mapping motor representations in the human cerebellum. Brain. 2013;136(1):330-342. doi:10.1093/brain/aws186

32.

Boillat

Bazin

van der Zwaag

. Whole-body somatotopic maps in the cerebellum revealed with 7T fMRI. Neuroimage. 2020;211:116624. doi:10.1016/j.neuroimage.2020.116624

33.

Jackson

Maldonado

Eakin

Orr

Bernard

. Cerebellar and prefrontal-cortical engagement during higher-order rule learning in older adulthood. Neuropsychologia. 2020;148:107620. doi:10.1016/j.neuropsychologia.2020.107620

34.

Mawase

Bar-Haim

Shmuelof

. Formation of long-term locomotor memories is associated with functional connectivity changes in the cerebellar-thalamic-cortical network. J Neurosci. 2017;37(2):349-361. doi:10.1523/JNEUROSCI.2733-16.2016

35.

Burciu

Fritsche

Granert

, et al. Brain changes associated with postural training in patients with cerebellar degeneration: a voxel-based morphometry study. J Neurosci. 2013;33(10):4594-4604. doi:10.1523/JNEUROSCI.3381-12.2013

36.

Kühn

Gleich

Lorenz

Lindenberger

Gallinat

. Playing super mario induces structural brain plasticity: gray matter changes resulting from training with a commercial video game. Mol Psychiatry. 2014;19(2):265-271. doi:10.1038/mp.2013.120

37.

Wolpert

Miall

Kawato

. Internal models in the cerebellum. Trends Cogn Sci. 1998;2(9):338-347. doi:10.1016/S1364-6613(98)01221-2

38.

Liew

Zavaliangos-Petropulu

Jahanshad

, et al. The ENIGMA Stroke Recovery Working Group: big data neuroimaging to study brain–behavior relationships after stroke. Hum Brain Mapp. 2022;43:129-148. doi:10.1002/hbm.25015

39.

Zatorre

Fields

Johansen-Berg

. Plasticity in gray and white: neuroimaging changes in brain structure during learning. Nat Neurosci. 2012;15(4):528-536. doi:10.1038/nn.3045

40.

Klintsova

Scamra

Hoffman

Napper

RMA

Goodlett

Greenough

. Therapeutic effects of complex motor training on motor performance deficits induced by neonatal binge-like alcohol exposure in rats: II. A quantitative stereological study of synaptic plasticity in female rat cerebellum. Brain Res. 2002;937(1-2):83-93. doi:10.1016/S0006-8993(02)02492-7

41.

Hori

Ihara

Sugai

, et al. Ventral striatum links motivational and motor networks during operant-conditioned movement in rats. Neuroimage. 2019;184:943-953. doi:10.1016/j.neuroimage.2018.10.018

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.15 MB