Identifying the white matter pathways involved in multiple sclerosis-related tremor using diffusion tensor imaging

Background Tremor affects up to 45% of patients with Multiple Sclerosis (PwMS). Current understanding is based on insights from other neurological disorders, thus, not fully addressing the distinctive aspects of MS pathology. Objective To characterize the brain white matter (WM) correlates of MS-related tremor using diffusion tensor imaging (DTI). Methods In a prospective case-control study, PwMS with tremor were assessed for tremor severity and underwent MRI scans including DTI. PwMS without tremor served as matched controls. After tract selection and segmentation, the resulting diffusivity measures were used to calculate group differences and correlations with tremor severity. Results This study included 72 PwMS. The tremor group (n = 36) exhibited significant changes in several pathways, notably in the right inferior longitudinal fasciculus (Cohen's d = 1.53, q < 0.001) and left corticospinal tract (d = 1.32, q < 0.001), compared to controls (n = 36). Furthermore, specific tracts showed a significant correlation with tremor severity, notably in the left medial lemniscus (Spearman's coefficient [rsp] = −0.56, p < 0.001), and forceps minor of corpus callosum (rsp = -0.45, p < 0.01). Conclusion MS-related tremor is associated with widespread diffusivity changes in WM pathways and its severity correlates with commissural and sensory projection pathways, which suggests a role for proprioception or involvement of the dentato-rubro-olivary circuit.


Introduction
Tremor has been historically recognized as one of the clinical manifestations of MS, 1 with an estimated prevalence of up to 45%, 2 though this may be an underestimation given the transient nature of the symptoms.Tremor is among the strongest predictors of impaired quality of life in patients with multiple sclerosis (PwMS), 3 and can impact activities of daily living (ADLs). 2 Current treatment modalities for MS-related tremor are lacking in their justification, measured efficacy, or outcomes. 4Deep brain stimulation is one of the few available options for severe medication-resistant tremor, though, it shows limited efficacy in treating MS-related tremor compared to essential tremor or Parkinson's disease-related tremor 5 while causing permanent adverse effects in up to 25% of the cases. 6One reason may be that MS differs fundamentally from other causes of tremor with variable topography of pathology in the brain and spinal cord.Many studies on MS-related tremor are conducted with prior assumptions of involved neurocircuitry drawn from other causes of tremor, often scrutinizing cerebello-thalamocortical networks. 4,7,8iffusion tensor imaging (DTI) is a reliable, validated technique that quantitatively assesses random water mobility, providing information about the tissue microstructure integrity measured in terms of diffusivity. 9In a medium without an oriented microstructure, diffusivity is similar in three axial directions.Mean diffusivity (MD) is the average diffusivity in the three directions and can be used to describe the structural integrity of the tissue.The directional diffusivities, radial diffusivity (RD), and axial diffusivity (AD) describe water movement in compact white matter tracts.It has been previously shown that increased RD is associated with myelin injury, 10 while AD correlates with axonal injury. 11Fractional anisotropy (FA) is a scalar value calculated from the directional diffusivities and reflects the directionality of diffusion.FA has been validated as a measure of myelination status. 12Damage to white matter (WM) causes an increase in mean diffusivity and a decrease in FA due to a loss of directional diffusion. 13 this study, we aimed to characterize the WM microstructural changes associated with MS-related tremor as well as describe how these changes correlate with tremor severity using diffusion tensor imaging (DTI), to offer a better understanding of MS-related tremor taking into consideration the distinct nature of the disease.

Subjects
In this prospective case-control study, PwMS diagnosed using 2017 McDonald's criteria, 14 having limb-predominant MS-related tremor not attributable to another etiology and British Medical Research Council Rating Scale motor strength in all extremities ≥ 4 were recruited, between 2019 and 2021, from the comprehensive MS care clinic at UT Physicians, McGovern Medical School, UTHealth.Subjects were sequentially enrolled, and subsequent adjustments to recruitment were made based on gender, age, and tremor severity to form the tremor group (MS-T).This method allowed us to minimize imbalance in our sample based on demographic and clinical characteristics.An additional group of PwMS matched by age and gender, without a clinically evident tremor, provided a comparison control group (MS-C) with a ratio between cases and controls of 1:1.The study was approved by the Institutional Review Board for the University of Texas Health Science Center at Houston -UTHealth (IRB HSC-MS-19-0305).All subjects participated in the study after giving written informed consent following approved procedures by the Committee for the Protection of Human Subjects of McGovern Medical School, UTHealth.

Clinical evaluation
Symptomatic medications to lessen tremor severity, if present, were held for at least 24 h prior to clinical assessments only if medically cleared by the treating physician and with additional informed consent from the patient.Currently, there is no consensus as to rating scales to evaluate MS-related tremor severity.We used The Essential Tremor Rating Assessment Scale (TETRAS), a previously validated measure used in essential tremor that correlates strongly with other measures previously usurped to assess MS-related tremor. 15,16TETRAS quantifies tremor severity and its impact on ADLs with a focus on upper extremity action tremor while also assessing head, face, voice, and lower limbs.Evidence indicates that TETRAS offers improved characterization of tremor, especially in instances of severe tremor. 16s an additional method of evaluating tremor, Archimedean spiral drawings were obtained and scored according to the Tremor Research Group instructions. 17Upper limb ataxia, which has a possible confounding effect due to overlap in etiology and neuroanatomical correlates with tremor, was scored using the Scale for Assessment and Rating of Ataxia (SARA). 18Lastly, limb spasticity and extremity motor strength were assessed using the Modified Ashworth Scale and the British Medical Research Council Rating Scale, respectively.Disability data as quantified by the expanded disability severity scale (EDSS) was assessed as previously described. 19All clinical assessments were performed by raters not involved in image analysis.
Image acquisition MRI was performed using a 3T Philips Ingenia research scanner (Philips Medical Systems, Best, Netherlands).High-resolution 3D T1-weighted, used for anatomical registration, were acquired at the start of each scanning session, as well as T2-weighted fluidattenuated inversion recovery (FLAIR) images.Diffusion-tensor images (DTI) were obtained using a single-shot spin-echo diffusion-sensitized echoplanar imaging sequence with a balanced Icosa21 tensor encoding scheme. 20DTI quality control throughout data acquisition was assured and postprocessed as detailed previously. 21Technical details on acquisition hardware and sequences can be found in the supplementary.

Image analyses
White matter lesions were segmented on the T2-FLAIR and T1-weighted images using BIANCA, 22 and validated by two trained raters.Voxel-wise lesion probability maps were created as previously described. 23A two-step approach was employed to identify and examine the potentially associated WM fiber pathways in an agnostic manner, summarized in Figure 1.The first step, tract selection, aimed to synthesize the tracts-of-interest (TOIs) using three independent techniques scoping different anatomical resolutions.The second step, tract segmentation, aimed to individually segment the TOIs for each subject.Further details on the image analyses experiments setup can be found in the supplementary.

Tract selection
• Atlas-based DTI segmentation: in this step, regional diffusivity values were obtained for various brain regions using an atlas-based method.By comparing the two groups, we identified WM regions potentially associated with MS-T.• Tract-Based Spatial Statistics (TBSS): in FMRIB Software Library, TBSS 24 was used to create a mean FA image and skeleton for each group by averaging their FA images in a standard space and voxelwise comparisons were made to identify any significant differences.Using atlasquery, the identified voxels were localized on the ICBM-DTI-81 white-matter labels atlas.• Correlational tractography: this technique uses connectometry, a more sensitive method compared to region-of-interest approaches, 25 to statistically infer the association with the study variable within WM pathways by analyzing the local voxel-wise connectomes.In MS-T, diffusion data were reconstructed in standard space.This analysis was conducted using quantitative anisotropy (QA) 26 and FA separately as a tracking index and TETRAS as the study variable while controlling for the effects of age and gender.
Tract segmentation.Tractography offers a method for WM segmentation that can effectively accommodate individual anatomical variations in tract dimensions and shapes.A recent study highlighted its feasibility in MS, as it does not introduce systematic differences when tracking in normal appearing WM, T2 lesions or T1 lesions. 27For each subject, 53 pathways were tracked using DSI Studio and a supervised automated method with a deterministic tractography algorithm 26 with augmented strategies to improve reproducibility. 28The diffusion data were reconstructed using generalized q-sampling imaging (GQI) 29 with a diffusion sampling length ratio of 2. GQI-based reconstruction has been validated on a similar acquisition protocol and has been shown to more consistently and accurately resolve crossing fibers compared to DTI-based reconstruction. 30The mean tract FA, MD, RD, and AD were obtained from the segmented pathways for MS-T and MS-C.

Statistical analysis
Statistical analyses for group differences were conducted using jamovi (version 2.3), with a significance level set at p < 0.05.Demographic data's standard deviation was estimated using student's t-test, while the EDSS scores were compared using the Mann-Whitney U-test after testing for normality.Categorical variables were analyzed using contingency tables and chi-squared tests.
To compare the variance in mean diffusivity measures of the pathways between MS-T and MS-C, ANCOVA was employed while controlling for the effects of age and gender, following residual plotting and inspection.
Mean difference and effect size, quantified by Cohen's d (d), are reported.The p-values were adjusted for false discovery rate (q) using the Benjamini-Hochberg procedure.
In R (version 4.1.1),we studied the correlation between tract diffusivity measures and tremor severity in MS-T.Mean tract FA, MD, RD, and AD served as independent variables in the LASSO-partial ridge (LPR) regression. 31The tremor severity measures (TETRAS and its left, TETRAS-L, and right component, TETRAS-R, as well as left and right Archimedean spiral scores, ARCH-L and ARCH-R, respectively) were considered as dependent variables.LPR regression was performed using the HDCI package with 1000 paired bootstraps.The lambda value was selected using 10-fold cross-validation to achieve the most regularized model within one standard error of the minimum (lambda.1se),minimizing the mean squared error.This analysis also included age, gender, and ataxia severity (SARA) as non-penalized explanatory variables.The partial correlation coefficients (r s p) were calculated using Spearman's rank correlation for the selected pathways and their corresponding tremor measures while controlling for the effects of age, gender, and ataxia severity.
Raw data was generated at UTHealth.Anonymized derived data may be shared upon reasonable request to the corresponding or senior authors with researchers who provide a methodologically sound noncommercial proposal, subject to restrictions according to participant consent and data protection legislation.

Demographic and clinical data
A total of 72 PwMS were included in this study, Table 1 summarizes their demographic and clinical data.MS-T had higher total EDSS scores, particularly in ambulation and cerebellar functions compared to MS-C, and a higher lesion burden visually more prominent in the brain stem, the periventricular areas, splenium, and genu of corpus callosum.Lesion probability maps for both groups can be seen in Figure 2.
In MS-T, 89% of the subjects were right-handed (n = 32), 10 subjects were on tremor medications of which 9 held their medications 24 h prior to assessment, none of the control group subjects were on tremor medications.Data on clinical scores and tremor medications in MS-T are represented in the supplementary, Tables S1 and S2.

Tract selection
The atlas-based DTI segmentation highlighted 57 WM regions exhibiting significant (q < 0.05) group differences in mean diffusivity of which 18 regions correlated with tremor severity, most prominently in the genu of the corpus callosum (left: r s p = 0.53 and right: r s p = 0.54), as well as the anterior corona radiata (left: r s p = 0.53 and right: r s p = 0.54).TBSS revealed significant widespread group differences in FA, while correlational tractography highlighted specific pathways that correlated with tremor severity, notably the CC-FMN (t = 2.10, q < 0.001), TR-A-L (t = 2.04, q < 0.001), CC-FMJ (t = 1.94, q < 0.001) and the ML-L (t = 1.85, q < 0.001).The results are summarized in Table 2.The TOIs were constructed by evaluating the significant areas from the different techniques used, which concluded in 53 pathways (Supplementary, Fig S1).
Tract segmentation Group comparisons.A comparison was conducted between the two groups, for the segmented tracts using the mean pathway values of FA, MD, RD, and AD.Group differences in FA are depicted in Figure 3.The greatest reduction in FA in MS-T compared to MS-C was found in the right inferior longitudinal fasciculus (ILF-R) (d = 1.53, 95% CI: 0.99-2.07,q < 0.001) followed by the left corticospinal tract (CST-L) (d = 1.32, 95% CI: 0.80-1.84,q < 0.001) and the body of the corpus callosum (CC-B) (d = 1.31, 95% CI: 0.79-1.84,q < 0.001).Other pathways with significant decreases were the left and right   MS-C > MS-T at 0.3 FA threshold, p-FWE < 0.05 In terms of MD, the most prominent increase in MS-T compared to MS-C was again seen in the ILF-R (d = -1.6,95% CI: −2.15 -−1.06, q < 0.001), followed by the CST-L (d = -1.52,95% CI: −2.06 -−0.98, q < 0.001) and the IFOF-R (d = -1.49,95% CI: −2.03 -−0.96, q < 0.001).Other pathways with significant increases included the right anterior thalamic radiation (TR-A-R), followed by the parietal section of the left superior longitudinal fasciculus, and the IFOF-L as depicted in Figure 4. Group differences in FA and MD for all 53 pathways are listed in the supplementary, Table S3.
Correlations with tremor severity.All 53 segmented tracts were evaluated.The correlations between tremor severity measurements and the diffusivity measures of the pathways were calculated while controlling for the effect of ataxia, age, and gender.A significant negative correlation between FA and TETRAS was noted in the left medial lemniscus (ML-L) (r s p = -0.56,95% CI: −0.75, −0.22, p < 0.001), and the forceps minor of corpus callosum (CC-FMN) (r s p = -0.45,95% CI: −0.72, −0.04, p < 0.01), followed by the forceps major of corpus callosum (CC-FMJ) (r s p = -0.35,95% CI: −0.64, −0.01, p = 0.05) as shown in Figure 5.To explore the influence of laterality and handedness we interchanged the left and right medial lemniscus FA measures for left-handed participants (n = 4).Interestingly, this modification revealed a slightly stronger correlation with TETRAS (r s p = -0.59,p < 0.001).

Discussion
Our results showed significant differences in diffusion metrics within several pathways in MS-T compared to MS-C.The greatest reduction in FA was seen in the right inferior longitudinal fasciculus, followed by the left corticospinal tract, and body of corpus callosum.In terms of MD and RD, the greatest increase was again seen in the right inferior longitudinal fasciculus, followed by the left corticospinal tract and right inferior fronto-occipital fasciculus, suggesting that demyelination largely drives dysfunction within several brain pathways.Interestingly, correlations between tremor severity and diffusivity measures for FA, MD, RD, and AD were seen in the left medial lemniscus, the forceps major of corpus callosum, and the forceps minor of corpus callosum, which all showed significant group differences, as well.Indeed, correlations between TETRAS and FA were stronger for the medial lemniscus when transposing the right-left brain orientation of left-handed participants.We hypothesize, in light of our findings, that MS-related tremor associates with widespread structural disconnectivity secondary to lesion-driven The medial lemniscus is part of the dorsal columnmedial lemniscus system, which is responsible for relaying the peripheral mechanoreceptive sensations. 33Heenan et al. 34 suggested that, mechanistically, MS-related tremor is secondary to sensory feedback maladaptation and defects in internal estimates of limb dynamics.Indeed, peripheral limb cooling was shown in one study to reduce tremor in PwMS, though the mechanisms remain unexplored. 35Feys et al. 36 reported correlations between tremor amplitude and lesion load in the contralateral pons, suggesting the possible involvement of the descending rubro-olivary pathway.Lesions in the midbrain can cause dystonia frequently associated with tremor, while thalamic or striatopallidal lesions can cause pure focal dystonia.A study showed that the resting and postural components of dystonia-related tremor occurred exclusively with lesions disrupting the dentato-rubro-olivary loop as well as the medial lemniscus. 37Therefore, two potential explanations exist for the strong correlation between tremor severity and the medial lemniscus diffusivity measures in our sample.One explanation could be the involvement of the dentato-rubro-olivary circuit 38 in MS-related tremor due to its intimate anatomical proximity. 33The second explanation could be a reduction in sensory afferents modulating movement, 39,40 as the medial lemniscus is the main pathway for proprioception and conveys sensory information to various circuits to regulate smooth motor action by providing positional and gravitational information.

Limitations
Our study has some limitations, most notably the absence of a validated MS-related tremor scale to precisely characterize and measure tremor severity.TETRAS offers improved characterization of severe tremor and correlates well with other scoring systems previously usurped to quantify MS-related tremor, such as the Fahn-Tolosa-Marin scale. 15,16DSS differences between groups were primarily driven by higher cerebellar and ambulation subscores, due to the need for assistance devices secondary to ataxia, though all correlations were adjusted for SARA.Diffusion MRI fiber tracking is a user-defined process, which might affect reproducibility of results.
We mitigated this issue methodologically by using a semi-automated approach to standardize the procedure. 26,28Another limitation is that our DTI sequence is not optimized fully for GQI-based reconstruction, although it has been validated on a similar acquisition protocol. 30Additionally, in this initial work, we did not explore in-depth the impact of MS-related lesions, either in the brain or spinal cord, on the described tracts.Lastly, inferences from our findings are restricted by the limited sample size and the exploratory nature of the study.

Conclusions
The results showed significant widespread changes in diffusivity measures along several association, projection, and commissural pathways associated with MS-related tremor.The greatest dysfunction between groups was seen in the inferior longitudinal fasciculus followed by the corticospinal tract.In the tremor group, we found a correlation between tremor severity and diffusivity measures of certain projection and commissural pathways.The correlations between tremor severity and diffusivity measures in the medial lemniscus might suggest a potential involvement of the dentato-rubro-olivary circuit in MS-related tremor or a reduction in sensory afferents affecting movement due to the medial lemniscus being the main pathway for proprioception.Further confirmatory studies with larger sample sizes are needed to validate the findings, studying both causality and temporality between tract-specific dysfunction and tremor incidence as well as severity.

Figure 2 .
Figure 2. Lesion probability maps.Depicting the average lesion burden distribution in MS-C and MS-T, t-statistic at p < 0.05 calculated from permutation tests to adjust for multiple comparisons.

Figure 3 .
Figure 3. Group differences in mean tract FA.Depicting axial slices in tractogram.The pathways are color-coded using their corresponding effect sizes (Cohen's d).Effect sizes were calculated via post-hoc comparisons from ANCOVA by comparing tract mean fractional anisotropy between MS-T and MS-C.

Figure 4 .
Figure 4. Group differences in mean tract MD.Depicting axial slices in tractogram.The pathways are color-coded using their corresponding effect sizes (Cohen's d).Effect sizes were calculated via post-hoc comparisons from ANCOVA by comparing tract mean diffusivity between MS-T and MS-C.

Figure 5 .
Figure 5. Correlations between mean tract FA and tremor severity.Raincloud plots represent the distribution of measures for both groups (MS-C in green and MS-T in orange).Effect size (d) calculated via post-hoc comparisons from ANCOVA comparing mean tract fractional anisotropy.Scatter plots showing the relationship between tremor severity and mean tract FA.Blue dashed lines represent 95% confidence interval.Partial correlation coefficients (r s p) calculated using Spearman's rank correlation.Tremor severity measured by TETRAS.Panel A: left medial lemniscus (ML-L).Panel B: forceps minor of corpus callosum (CC-FMN).Panel C: forceps major of corpus callosum (CC-FMJ).

Table 1 .
Demographic and clinical measures.
a b TBSS groupwise comparisons results localized using the ICBM-DTI-81 atlas.c Correlational tractography results correlating with TETRAS controlling for age and gender, t-statistic represents strength of correlation.