White Matter Tract Disruptions Predict Less Affected Hand Impairment Following Stroke: A Longitudinal Diffusion MRI Study

Firdaus Fabrice Hannanu University Grenoble Alpes: Universite Grenoble Alpes Bernadette Naegele Centre Hospitalier Universitaire Grenoble Alpes Marc Hommel University Grenoble Alpes: Universite Grenoble Alpes Alexandre Krainik Centre Hospitalier Universitaire Grenoble Alpes Olivier Detante Centre Hospitalier Universitaire Grenoble Alpes Assia Jaillard (  assia.jaillard@univ-grenoble-alpes.fr ) Centre Hospitalier Universitaire de Grenoble https://orcid.org/0000-0002-1045-671X


Introduction
Hand movements represent a speci c and essential function in humans required for everyday life activities. Following stroke, loss of hand functionality is one of the main factors affecting disability and remains a major target of rehabilitation interventions (Pennati et  Moreover, the impact of neuropsychological de cits such as apraxia and neglect has also been associated with LAH impairment (Chestnut & Haaland, 2008 Patients received standard medical care including thrombolysis and thrombectomy when indicated. Patients were randomized to receive autologous mesenchymal stem cells (cell-therapy group) or rehabilitation alone (control group) in the rst two weeks following stroke (Jaillard et al., 2020). Cell therapy was administered one month following stroke, after baseline clinical and MRI assessment. The ISIS-HERMES study was approved by the Institutional Review Board (CPP: 07-CHUG-25). Written informed consent was obtained from all patients before they participated in the study.
The main inclusion criteria were: age 18-70 years, rst-ever unilateral infarct in the internal carotid artery territory, moderate to severe neurological de cit de ned as NIHSS ≥7, and the ability to follow a rehabilitation program. Exclusion criteria were pre-stroke neurological or psychiatric disease, severe medical condition, apraxia, or neglect diagnosed with an extinction NIHSS subscore >1. Complete criteria are provided in Table S1. Clinical, behavioral, and MRI measures were acquired at three sessions: one-

Behavioral Measures
We explored LAH impairment using three behavioral tests. The Purdue Pegboard Test (PPT) (Lafayette Instrument Company, Indiana)(Rapin, Tourk, & Costa, 1966) was performed as described in http://www.equipement-ergotherapie.com/8-dexterité-manipulation.html, as a standardized quantitative test requiring motor (for grasping) and visuomotor (for reaching) components. The Hand dynamometer (Lafayette Instrument Company, Indiana; https://www.prohealthcareproducts.com/100-kg-220lb-handgrip-dynamometer-lafayette-instruments/) is a validated test to measure handgrip force (Grip). The Motor Screening Task (MST) measures movement time to assess sensorimotor de cits in CANTAB (https://www.cambridgecognition.com/cantab/cognitive-tests/attention/motor-screening-task-mot/). Raw scores were converted to percentiles to adjust for age and sex using published norms (Spreen & Strauss, 1998) and CANTAB norms. Scores below the 5 th percentile were considered as impaired. The frequency of LAH impairment was also assessed in patients without cognitive de cit, de ned as RBANS > 40.

MRI data acquisition
The MRI protocol included structural and diffusion sequences. Patients were scanned at the IRMaGe MRI facility (Grenoble) on a 3T Philips magnet (Achieva 3.0 TTX; Philips, the Netherlands) with a 32-channel head coil. High resolution (1 mm3) sagittal 3D-T1-weighted (TR 9.9 ms, TE 4.6 ms, ip angle 8°, TI 920 ms, inter shot time 1792 ms) and uid-attenuated inversion recovery (FLAIR) images (TR 8 s, TE 342 ms) were acquired. Diffusion-weighted images were acquired using single-shot echo-planar imaging (EPI) sequence (TR 11 milliseconds, TE 72 milliseconds, FOV 240 mm, slice thickness 2.0 mm, 70 axial slices, SENSE factor 2, fold-over direction anteroposterior, fat shift direction P, fat suppression, and voxel size 1.67*1.67*2 mm). We acquired 60 noncollinear directions with a b value of 1,000 s/mm 2 and 10 directions with a b value of 0 s/mm 2 that were averaged to give 1 average direction.

MRI data analysis
Structural images were used to manually delineate lesion masks and compute lesion volumes using MRIcron (https://www.nitrc.org/projects/mricron). Diffusion-weighted images were processed with the Diffusionist toolkit derived from FSL software, as previously described (Soulard et al., 2020). Each DWI image was visually checked and removed if corrupted. Then, after correction of eddy-current distortions, the diffusion tensor was estimated.
We used FA to assess white matter disruptions. Voxel-wise FA images were constructed from the resulting tensors. Linear and nonlinear registration transformations were applied to the FSL FA template in the MNI-152 space by incorporating the knowledge of each brain lesion using manually delineated lesion masks (Renard, Urvoy, & Jaillard, 2015). FA was estimated only in the template's skeleton and outside the lesion mask. We estimated FA with atlas-based regions of interest (ROI) approach using the human brain white matter JHU atlas (Oishi et al., 2008) to analyse the descending motor tracts at different levels.
Average FA values were estimated in 20 ROIs listed in Table 2 and represented in Fig. S2. As we did not acquire sequences for susceptibility-induced distortions correction that are needed to obtain reliable diffusion values in the pons region, we excluded the pons ROI from our analysis. Diffusionist toolkit and related documentation can be found at http://mri-diffusionist.com/.

Statistical analysis.
LAH impairment was explored using descriptive statistics. To explore the mechanisms of LAH impairment, we rst assessed the relationship between PPT, handgrip, and MST percentiles and clinical scores using Spearman correlations. Linear associations between ROI-derived FA and LAH raw scores were assessed using partial correlations controlling for age and sex, with bootstrap based on 1000 replications.
We used linear mixed models (LMMs) to determine which tracts (ROIs) in uence LAH performances over time. The effects of ROIs, ROI by session interaction, lesion side, volume, handedness, education BMI, height, and weight were tested and included in the model only if signi cant. As cell therapy could have in uenced outcomes, all LMMs were adjusted for cell therapy. Bonferroni correction was applied for post hoc multiple comparisons. Statistical signi cance was determined with the F-test (p<0.05) and model t was estimated with the Akaike Information Criterion (AIC) and by examining the distribution of residuals (Steyerberg et al., 2001). Model accuracy was assessed using R 2 . Statistical data analyses were performed using SPSS 23.0.

Factors associated with LAH impairment
At M0, PPT, Grip and MST signi cantly correlated with lesion volume and most clinical and cognitive scores (Table S2). LAH impairment was strongly correlated with mRS at M6 and M24 (Table S3). There was no signi cant effect of lesion side on LAH performances.
LAH and FA correlations are presented in Table 2. PPT correlated with all ipsilesional (i-) CST ROIs, contralesional (c-) CP, and bilateral SLF, corpus callosum, ALIC, and ACR. Grip correlated with the same ROIs except for i-SLF and i-ALIC. MST correlated with i-PLIC, i-CP, c-SLF, genu-CC, bilateral ACR, and c-ALIC. There were no signi cant correlations with the cerebellar peduncles.

Mechanisms of LAH impairment
Our ndings showed that several tracts including the i-CST, c-ALIC, CC, and i-SLF, depending on the task, were associated with LAH impairment.
We found moderate to strong correlations between the three LAH scores and FA in the ipsilesional CST, while no correlation was observed with contralesional CST-CSR and CST-PLIC, suggesting that LAH impairment is driven by the ipsilesional CST. Furthermore, i-CST was a strong predictor of PPT and grip impairment. These ndings suggest that LAH impairment requiring motor processing such as PPT and grip relates to the damaged uncrossed bers of the ipsilesional CST ( Fig. 1 & S4A). These bers terminate in the ventromedial intermediate zone to propriospinal neurons connected to distal motoneurons through intersegment spinal interneurons and may be involved in the motor control of dexterous hand movements (Tohyama et al., 2017). Taken together, our results support the hypothesis that damaged uncrossed ipsilateral CST bers contribute to the motor components of LAH impairment.
LAH scores were also correlated with FA in bilateral ALIC and ACR, and CC. LMMs showed that PPT and handgrip impairment was associated with c-ALIC. Although ACR has been linked to cognition and particularly to attention in adults with brain injury (Niogi et  Therefore, motor control components of LAH impairment may also implicate the contralesional CST through the transcallosal bers and c-ALIC bers from premotor and/or prefrontal areas ( Fig. 1 and S4.B).
We found that LAH scores correlated with CC FA, which predicted MST. A role of the CC is motor coordination of bimanual (Andres et al., 1999) and unilateral hand motor movements through the balance of excitatory and inhibitory interhemispheric interactions (Chettouf et al., 2020). The ipsi-and contralesional motor areas exert a reciprocal in uence through transcallosal bers (Leichnetz, 1986), as evidenced in tracer studies showing reciprocal transcallosal connections for both MI and SMA (Gould, Cusick, Pons, & Kaas, 1986). In nonhuman primates, SMA lesions impaired the LAH motor program through transcallosal connections to contralesional SMA (Brinkman, 1984). Of note, c-ALIC and CC predicted worse LAH performances when accounting for the effects of i-CST and i-SLF, suggesting that LAH impairment may be compounded by the imbalance between ipsilesional and contralesional motor regions.
Our results showed that LAH impairment correlated with decreased FA in bilateral SLF and that i-SLF predicted PPT and MST. These ndings support the theory that LAH impairment relates to unilateral movement bilateral hemispheric control (Chettouf et  We found consistent correlations between LAH impairment and spatial and attentional cognitive de cits that may result from SLF, ACR or ALIC disruptions, depending on the cognitive domain. Interestingly, excluding patients with cognitive de cit did not signi cantly improve LAH performances, suggesting that cognitive impairment was not a major cause of LAH impairment in this study. Nevertheless, as patients with severe apraxia or neglect were excluded from our study, we may have underestimated the in uence of cognitive impairment on LAH impairment related to apraxia and neglect (Sunderland et al., 1999).

The role of the lesion volume
Lesion volume correlated with LAH impairment and predicted PPT and MST, consistently with nonhuman macaque experiments (Darling et al., 2011). Surprisingly, few studies, if any, have explored the relationships between lesion volume and LAH impairment in humans. Our results showing that LAH tasks with visuomotor processing were compounded by lesion volume provide empirical support for the notion of mass action through redundancy in the visuomotor system (Glassman, 1987).

Limitations
The small sample size is the main limitation of this study. However, this is the rst study exploring the microstructural white matter disruptions to understand the underlying mechanisms of LAH impairment following stroke. Also, the longitudinal design with repeated measures for both behavioral and FA measures, the homogeneity of our population in terms of age, absence of leukoaraïosis, and stroke severity and territory, and the high model accuracy suggest that the sample size was adequate for this study. Nevertheless, the small sample may explain why we did not observe any effect of the lesion side, in contrast with others (Varghese & Winstein, 2019). Moreover, this study was part of a randomized clinical trial assessing cell therapy, which might have in uenced outcomes. To account for this limitation, all LMMs were adjusted for cell therapy.

Conclusion
This study showed that motor-related tract disruptions predict LAH impairment, with a pattern that varies according to the motor and visuomotor processing: tasks with motor processing were associated with the ipsilateral CST implicating the involvement of uncrossed CST bers, while tasks with visuomotor processing were related to the SLF supporting hand motor control and to lesion volume. In addition, the contralesional hemisphere may play a role in LAH impairment by in uencing the planning and execution of hand movements through prefrontal/premotor areas and transcallosal interactions. Taken together, our ndings suggest that LAH impairment requires the summation of tract disruptions, supporting the concept of cerebral redundancy for the motor system. Our results provide a theoretical basis for integrating LAH impairment in rehabilitation programs to improve functional recovery and for research interventions, such as neuromodulation.  peduncles, superior longitudinal fasciculus (SLF), genu and body of the corpus callosum (CC), anterior corona radiata (ACR), anterior limb of the internal capsular (ALIC),  Figure 1 Schematic representation of crossed and uncrossed bers of the corticospinal tract (CST). ACR and ALIC bers emerging from the SMA and PMC and SCR-PLIC bers emerging from PMC and MI merge in the CP to form the CST, which continues in the pons and medulla. Then the CST is divided into 2 parts. 1) Crossed CST (solid red and blue lines): most CST bers decussate in the medullar pyramids to descend in the contralateral spinal cord and terminate in the contralateral anterior spinal horn to distal extremity