White matter measures correlate with essential tremor severity—A pilot diffusion tensor imaging study

Abstract Background An evolving pathophysiological concept of essential tremor (ET) points to diffuse brain network involvement, which emphasizes the need to investigate white matter (WM) changes associated with motor symptoms of ET. Objectives To investigate ET‐related WM changes and WM correlates of tremor severity using tremor clinical rating scales and accelerometry. Methods Tract‐based spatial statistics (TBSS) approach was utilized to compare 3 Tesla diffusion tensor imaging (DTI) data from 12 ET patients and 10 age‐ and gender‐matched healthy individuals. Clinical scales, tremor frequency and amplitude as measured by accelerometry were correlated with DTI data. Results ET patients demonstrated mean (MD) and radial diffusivity (RD) abnormalities in tracts involved in primary and associative motor functions such as bilateral corticospinal tracts, the superior longitudinal fascicles, and the corpus callosum but also in nonmotor regions including the inferior fronto‐occipital and longitudinal fascicles, cingulum bundles, anterior thalamic radiations, and uncinate fascicles. A combined tremor frequency and amplitude score correlated with RD and MD in extensive WM areas, which partially overlapped the regions that were associated with tremor frequency. No significant relationship was found between DTI measures and clinical rating scales scores. Conclusions The results show that ET‐related diffusion WM changes and their correlates with tremor severity are preferentially located in the primary and associative motor areas. In contrast, a relationship between WM was not detected with clinical rating scales. Accelerometry parameters may, therefore, serve as a potentially useful clinical measures that relate to WM deficits in ET.


| INTRODUC TI ON
Essential tremor (ET) is the most prevalent movement disorder affecting up to 5% of the general population above the age of 65 years (Louis & Ferreira, 2010), but its pathophysiology remains poorly understood. ET is characterized by a 4-12 Hz postural and kinetic tremor (Bareš, Husárová, & Lungu, 2012). For decades, studies have suggested various morphological changes underlying motor symptoms in ET (Gironell et al., 2012;Louis, Faust, & Vonsattel, 2012).
However, the relatively common occurrence of nonmotor symptoms such as cognitive, behavioral, and sleep disturbances in ET led to the replacement of the long-held "benign" mono-symptomatic model with a complex concept of ET (Bhalsing et al., 2014). The complexity of ET-related pathology was supported by findings of cerebellar abnormalities and widespread brain neuro-degeneration (Louis, 2014b(Louis, , 2014c. The cerebello-thalamo-cortical network (Sharifi, Nederveen, Booij, & van Rootselaar, 2014) (cerebello-thalamo-cortical [CTC], also known as "tremor network") has been proposed to play a substantial role in ET. In CTC network, the inferior olive nuclei project, via climbing fibers, to Purkinje cells in the contralateral cerebellar cortex, which send GABAergic inhibitory projections to the dentate and other deep cerebellar nuclei that in turn project to cortical brain areas via thalamic projections and to the red nucleus (Sharifi et al., 2014).
Diffusion tensor imaging (DTI) has been increasingly utilized as a noninvasive MRI technique, which allows quantifying microstructural substrate of disease-related white matter (WM) alterations based on water movement (Alexander, Lee, Lazar, & Field, 2007;Basser & Pierpaoli, 2011). The three-dimensional DTI tensor model provides fractional anisotropy (FA), mean (MD), radial (RD), and axial (AD) diffusivity values (Alexander et al., 2007;Pierpaoli & Basser, 1996). FA as the degree of anisotropy varies from 0 in nondirectional to 1 in highly oriented tissue (Alexander et al., 2007) and is influenced by the degree of myelination, axonal packing and size, coherence, and co-linearity of fiber organization (Alexander et al., 2007).
Changes in cellularity, edema, and/or necrosis affect MD, while RD and AD are sensitive to membrane coherence and axonal changes; respectively (Alexander et al., 2007).
Previous DTI studies in ET confirmed an involvement of the CTC (Jia, Jia-lin, Qin, Qing, & Yan, 2011;Klein et al., 2011;Nicoletti et al., 2010;Saini et al., 2012) but also of areas outside the "tremor network" (Klein et al., 2011;Saini et al., 2012) when a whole-brain DTI analysis using tract-based spatial statistics (TBSS) (Smith et al., 2006) was applied. The areas outside CTC were linked to nonmotor ET symptoms (Benito-León et al., 2017;Bhalsing et al., 2014Bhalsing et al., , 2015, which provides further evidence for the pathophysiological complexity of ET. The widespread and variable WM alterations in ET (Jia et al., 2011;Klein et al., 2011;Nicoletti et al., 2010;Saini et al., 2012) shown in prior studies emphasize the need to further investigate the correlation of WM microstructure changes with motor signs in ET and to identify potential morphological markers of tremor severity.
Identifying the relevant WM areas responsible for motor and nonmotor signs in ET would help explain the variable pharmacotherapy response in ET (Hedera, Cibulčík, & Davis, 2013) and potentially contribute to the development of reliable measures of therapeutic interventions (Ondo, 2016). Several DTI studies (Klein et al., 2011;Nicoletti et al., 2010;Pelzer et al., 2017;Saini et al., 2012) utilized clinical rating scales (e.g., Fahn-Tolosa-Marin [FTM] and Bain scales) as a measure of tremor severity; however, only two studies found a correlation between the FTM score and frontoparietal WM microstructure (Klein et al., 2011;Pelzer et al., 2017). The lower inter-rater reliability and the subjective nature of the clinical rating scales (Elble et al., 2013) can be mitigated by objective accelerometry measures (i.e., the tremor frequency and amplitude), which may provide a more accurate representation of the disease pathophysiological features. This is suggested by recent functional and morphometric MRI studies (Gallea et al., 2015;Popa et al., 2013). Using MRI morphometry, Gallea and colleagues showed that the cerebellar vermis volume was uniquely correlated to tremor frequency but not to FTM clinical scale (Gallea et al., 2015). Functional MRI data also demonstrated a relationship between connectivity within CTC and accelerometry tremor measures (Gallea et al., 2015;Popa et al., 2013). Although these recent MRI studies (Gallea et al., 2015;Popa et al., 2013) do imply a link between objective tremor measures and structural brain changes within CTC, there have been no studies that specifically frequency. No significant relationship was found between DTI measures and clinical rating scales scores.

Conclusions:
The results show that ET-related diffusion WM changes and their correlates with tremor severity are preferentially located in the primary and associative motor areas. In contrast, a relationship between WM was not detected with clinical rating scales. Accelerometry parameters may, therefore, serve as a potentially useful clinical measures that relate to WM deficits in ET.

K E Y W O R D S
diffusion tensor imaging, essential tremor, Guillain-Mollaret triangle, radial diffusivity, tremor network, TremScore investigated the correlation between the WM changes in ET and tremor amplitude and frequency.
In this study, we utilized TBSS (Smith et al., 2006), as a method that allows exploring whole-brain DTI metrics, to examine the relationship between WM changes and both clinical rating scales (i.e., FTM and Bain scale) and accelerometry tremor measures (i.e., amplitude, frequency and their combined score).
Participants had no history of other neurologic or psychiatric disorder, cognitive deficit (Mini Mental State Examination Score >25), traumatic brain injury, or seizure. All patients were off their tremor medication for 10 days prior to the study. All participants signed an Physiological tremor measures were recorded in the vertical axis using a tri-axial accelerometer, that is, Tremorometer (160-Hz sampling rate, 2 mg resolution; FlexAble-Systems, Fountain Hills, AZ) (Caligiuri & Tripp, 2004). A large study conducted on 242 patients with various tremor types (i.e., Parkinson's disease, ET, neuroleptic-and lithium-induced tremor) confirmed ability of the Tremorometer to reliably and reproducibly generate all components of the TremScore such as amplitude, frequency, and percentage of the tremor presence. Study also confirmed clinical utility of TremScore components to differentiate tremors of distinct etiologies (Caligiuri & Tripp, 2004). In addition, Tremorometer was utilized along with MEG to assess motor cortex during voluntary and involuntary movements (Bowyer et al., 2007). In our study, the amplitude and frequency of tremor for both upper limbs were calculated during postural task while the subject was sitting with both arms extended and hands in a pronated position, parallel to the ground. The posture was maintained for 21 s, which was duration of one trial. The amplitude and the peak frequency were calculated using spectral analysis (Fast Fourier transformation) (Caligiuri & Tripp, 2004). The TremScore was automatically generated as following: weighted tremor frequency mean (WFM,  Notes. Variables are reported as minimum (min), maximum (max), mean, standard deviation (SD), and standard error of the mean (SEM). FTM: modified Fahn-Tolosa-Marin scale. * Resting tremor present in one subject with score 1 (range 0-4).
TA B L E 1 Clinical characteristics and accelerometry outcomes in essential tremor group (Caligiuri & Tripp, 2004). Amplitude was internally converted from mm/s 2 to the mg equivalent for calculating the TremScore.
The tremor is considered more severe if the frequency is higher (WFM), its amplitude is higher (TA) or percent of time present is higher (SPREAD). The value is divided by 10 to keep the numbers in a reasonable range.
Three trials were recorded for each hand and averaged.
All scores were then averaged between two hands and were examined for outliers that were excluded from analysis (Table 1).  ox.ac.uk/fsl). Fieldmaps were used to diminish EPI distortion caused by inhomogeneities in applied magnetic field. Diffusion weighted data were corrected for movement and eddy current (ECC) distortions (Smith, 2002). Average motion parameters in the x-, y-, z-axis were calculated for each subject.

| DTI data analysis
Preprocessed images were fitted to the tensor model at each voxel to generate FA, MD, AD, and RD images using DTIFIT in FSL (Jenkinson et al., 2012). TBSS (Smith et al., 2006) was chosen as an optimal full-automated analysis tool that allows investigating the whole-brain without selection of specific WM region beforehand.
All FA images were registered into a common space using nonlinear registration with FMRIB58_FA as a registration target image (Andersson, Jenkinson, & Smith, 2007a, 2007b  The nonparametric voxel-wise statistics in "randomize" program in FSL was used to calculate differences between patients and controls adjusted for age using threshold-free cluster enhancement (Smith & Nichols, 2009 (Table 3). Since age has a critical confounding effect on both DTI (Salat et al., 2009) and clinical tremor measures (Elble, Higgins, Leffler, & Hughes, 1994), all regression tests were adjusted for age.

| Cross-sectional DTI differences
Parameters characterizing movement during the scanning did not

| Association between DTI and tremor severity parameters
Voxel-by-voxel TBSS analysis adjusted for age demonstrated a sig-

| D ISCUSS I ON
Our data show significant correlations between objective tremor measures (combined TremScore, tremor frequency) and diffusivity metrics primarily in the WM regions that are considered parts of the CTC network but also in areas beyond the tremor network. Higher MD and RD diffusivities in ET patients compared to healthy controls were found in extensive WM areas. In addition, subtle AD changes were detected in the forceps minor.
The extent and right-sided preference of ET-related changes are in agreement with a prior TBSS-based DTI study . Given the right-handedness of most of the participants (82%), the dominance of contralateral (left) hemisphere is expected (Gut et al., 2007). Therefore, the WM in the left hemi- to RD abnormalities implies an abnormal myelination (Song et al., 2002) in the ET pathology. Similar to prior studies, we did not detect significant changes in FA, while demonstrating differences in diffusivity measures . Although FA represents a very robust WM measurement (Basser & Pierpaoli, 2011), it specifically shows a relative variation between the levels of diffusion measured in different direction (Alexander et al., 2007;Basser & Pierpaoli, 2011). FA may remain relatively stable even if individual diffusivity values undergo relatively large changes.
White matter areas related to objective tremor measures as well as changes detected in patients compared to controls were anatomically located within the same tracts associated preferentially with motor functions. ET-related deficits were demonstrated in the CST, the primary motor projection of the cortical motor areas to the spinal cord (Martin, 2005). WM alteration within CST in ET has been demonstrated previously (Gallea et al., 2015;Saini et al., 2012) and corroborate with functional MRI studies showing altered brain activity in motor cortical areas at rest (Fang et al., 2016), during the grip-force (Neely et al., 2015) and finger-tapping task in ET patients . Similar findings were also shown with complex motor tasks such as extension of the right hand and arm to induce tremor (Buijink, (Louis, 2014b). Also, ET in younger versus elder patients can have a different disease nature supporting the hypothesis of ET being not a single disease but rather a family of diseases with a large heterogeneity (Louis, 2014a). The lower RD values, are often associated with higher myelination (Song et al., 2002), which might lead to increased transmission velocity between brain regions but does not necessarily indicate an increased communication efficiency (Laughlin & Sejnowski, 2003) and may, in fact, represent a compensation for aberrant function elsewhere in the brain (Mandl et al., 2010). While patients demonstrated changes suggestive of lower myelination levels compared to controls, the opposite correlation between tremor score in patients group suggests a compensatory increase in structural connectivity to balance clinical deficits. Importantly, areas that correlate with objective tremor measures extend to the midbrain, specifically areas around the cerebral peduncles and red nucleus thus including the WM within dentate-rubro-olivary pathway (Guillain-Mollaret triangle), which is considered a tremorgenic loop (Murdoch, Shah, & Jampana, 2016). These findings are consistent with previous studies showing abnormalities in DTI parameters directly within red nucleus (Jia et al., 2011), in retrorubral WM (Shin, Han, Kim, & Lee, 2008) and the Guillain-Mollaret triangle (Nicoletti et al., 2010) in ET patients. However, no correlation with tremor clinical parameters was reported in these studies (Jia et al., 2011;Nicoletti et al., 2010;Shin et al., 2008). Our data show a unique correlation of CST changes exclusively with combined TremScore but not with frequency or FTM scale. Similarly, forceps minor located in the ventral part of the corpus callosum showed a significant correlation with both objective combined score and frequency but not the clinical rating scales. While the absence of correlation between diffusion measurements and clinical ET rating scales in our study corroborates two previous DTI studies conducted on the larger sample of 22 and 25 ET patients (Nicoletti et al., 2010;Saini et al., 2012), others showed significant association between frontoparietal WM and Fahn score in smaller samples of 14 and 19 ET patients, respectively (Klein et al., 2011;Pelzer et al., 2017). These discrepancies may be explained by lower inter-rater reliability and the subjective nature of the clinical rating scales (Elble et al., 2013). We cannot exclude that the correlation with clinical rating scales could be detected on larger ET populations, although our sample shows significant relationships between WM and objective tremor measures but no association with clinical scales after exclusion of outliers and thus suggests that objective tremor measures reflect WM deficits.
The different association patterns with WM changes of tremor frequency and amplitude suggests that tremor frequency and amplitude are generated by different mechanisms involving different brain structures which may explain our finding that combined score was more extensively reflected in WM than frequency per se, while amplitude alone failed to show any significant correlations.
White matter alteration in ET compared to control group, as well as association with objective measures, were also reported in the SLF that is partially involved in motor functions (Makris et al., 2005). The SLF represents a complex fiber bundle that facilitates executive functioning but also encodes the position of the body in the surrounding space and monitors complex motor activities (Makris et al., 2005). Thus, our data provide compelling evidence of an involvement of primary and associative motor tracts in ET.
In addition, our data also shown WM changes in ET patients within fiber bundles such as ATR, IFOF, ILF, and UF as well as CB.
WM microstructure in abovementioned bundles was related to various neuropsychological outcomes in ET (Bhalsing et al., 2015), while our study linked WM blobs within these tracts to objective tremor measurements reflecting abnormal motor functions.
These relationships were in partial agreement with an association between FTM scale and fronto-parietal WM deficits demonstrated previously (Klein et al., 2011). In ATR (Mamah et al., 2010), which spreads from thalamus to frontal regions, were the tremorrelated areas located directly in the thalamus. Indeed, previous F I G U R E 2 White matter differences between essential tremor patients and healthy controls. Mean (MD) and radial diffusivity (RD) differences between groups based on a voxel-vise comparison of skeleton voxels using tract-based spatial statistics (TBSS) within FSL. Clusters with significantly higher MD and RD in patients with essential tremor compared to healthy controls are in red-yellow (p FWEcorr < 0.05), voxels that belong to TBSS-skeleton in green. x, y, z values are showing MNI coordinates of selected slice. Significant voxels were spatially smoothed using "fill" tool in TBSS to enhance visualization of the results F I G U R E 3 Associations between clinical characteristics and diffusion metrics. Clusters with significant associations with tremor severity in patients with essential tremor are shown in red-yellow (p FWEcorr < 0.05), voxels that belong to tract-based spatial statistics (TBSS)-skeleton in green. Significant voxels were spatially smoothed using "fill" tool in TBSS to enhance visualization of the results. (a) Mean diffusivity areas related to the TremScore. (b) Radial diffusivity values associated with TremScore. (c) Relationship between frequency and radial diffusivity PET studies showed increased thalamic blood flow during rest (Wills, Jenkins, Thompson, Findley, & Brooks, 1994) and passive movement performance (Jenkins et al., 1993;Wills, Jenkins, Thompson, Findley, & Brooks, 1995)  Contrary to one whole-brain DTI study  but in agreement with another (Klein et al., 2011), we did not find any significant changes in the cerebellum or cerebellar peduncles using TBSS method (Smith et al., 2006). Klein et al. (2011) suggested that differences between region-of-interest (ROI) and whole-brain methods are explicitly in the cerebellar peduncles caused by an inability of the FAdriven registration techniques to properly align these structures, particularly to distinguish inferior and medial cerebellar peduncles. While TBSS is more suitable for exploratory studies as it does not exclude any region beforehand, the stringent multiple comparisons used for our statistical analysis (Smith & Nichols, 2009) might eliminate results potentially revealed by ROI analyses. Indeed, Klein et al. (2011) did not report cerebellar changes using whole-brain analysis but confirmed an involvement of cerebellar peduncles using ROI analysis method.

| LI M ITATI O N S
The small number of subjects is the main limitation of this study.
Further studies are needed to confirm our findings. Small sample also did not allow us to separate subjects based on a family history of ET, response to alcohol, head or resting tremor presence. Since our study did not provide a comprehensive neuropsychological assessment, we cannot evaluate the relationship between cognition and WM changes and completely exclude the possibility that WM alteration is to certain level caused by cognitive deficits.

| CON CLUS IONS
Our results showed widespread WM alterations in ET patients, which further supports the neurodegenerative nature of the disease.
Accelerometry tremor parameters showed robust correlation with changes in WM of the primary and associative motor areas, which were not detectable with clinical rating scales. Objective tremor metrics may therefore serve as important clinical measurements that correlate with WM alterations in ET.

ACK N OWLED G M ENTS
We thank Diane Hutter, RN for her help with patient's recruitment.
We are grateful to the subjects who participated in this study as

CO N FLI C T O F I NTE R E S T S
The authors report no conflict of interests.