White matter alterations in Parkinson’s disease with normal cognition precede grey matter atrophy

Introduction While progressive MRI brain changes characterize advanced Parkinson’s disease (PD), little has been discovered about structural alterations in the earliest phase of the disease, i.e. in patients with motor symptoms and with normal cognition. Our study aimed to detect grey matter (GM) and white matter (WM) changes in PD patients without cognitive impairment. Methods Twenty PD patients and twenty-one healthy controls (HC) were tested for attention, executive function, working memory, and visuospatial and language domains. High-resolution T1-weighted and 60 directional diffusion-weighted 3T MRI images were acquired. The cortical, deep GM and WM volumes and density, as well as the diffusion properties of WM, were calculated. Analyses were repeated on data flipped to the side of the disease origin. Results PD patients did not show any significant differences from HC in cognitive functioning or in brain volumes. Decreased GM intensity was found in the left superior parietal lobe in the right (p<0.02) and left (p<0.01) flipped data. The analysis of original, un-flipped data demonstrated elevated axial diffusivity (p<0.01) in the superior and anterior corona radiata, internal capsule, and external capsule in the left hemisphere of PD relative to HC, while higher mean and radial diffusivity were discovered in the right (p<0.02 and p<0.03, respectively) and left (p<0.02 and p<0.02, respectively) in the fronto-temporal WM utilizing flipped data. Conclusions PD patients without cognitive impairment and GM atrophy demonstrated widespread alterations of WM microstructure. Thus, WM impairment in PD might be a sensitive sign preceding the neuronal loss in associated GM regions.


Introduction
While progressive MRI brain changes characterize advanced Parkinson's disease (PD), little has been discovered about structural alterations in the earliest phase of the disease, i.e. in patients with motor symptoms and with normal cognition. Our study aimed to detect grey matter (GM) and white matter (WM) changes in PD patients without cognitive impairment.

Methods
Twenty PD patients and twenty-one healthy controls (HC) were tested for attention, executive function, working memory, and visuospatial and language domains. High-resolution T1-weighted and 60 directional diffusion-weighted 3T MRI images were acquired. The cortical, deep GM and WM volumes and density, as well as the diffusion properties of WM, were calculated. Analyses were repeated on data flipped to the side of the disease origin.

Results
PD patients did not show any significant differences from HC in cognitive functioning or in brain volumes. Decreased GM intensity was found in the left superior parietal lobe in the right (p<0.02) and left (p<0.01) flipped data. The analysis of original, un-flipped data demonstrated elevated axial diffusivity (p<0.01) in the superior and anterior corona radiata, internal capsule, and external capsule in the left hemisphere of PD relative to HC, while higher mean and radial diffusivity were discovered in the right (p<0.02 and p<0.03, respectively) and left (p<0.02 and p<0.02, respectively) in the fronto-temporal WM utilizing flipped data. PLOS  Introduction changes; and axial diffusivity (AD) parallels axonal abnormalities. However, DTI metrics remain non-specific and the DTI outcomes should therefore be interpreted with caution [17]. Taken together, our multimodal study aimed to comprehensively test the following hypotheses: 1) cortical/subcortical GM alterations are not found or are only found in limited extents in PD-NCI; and 2) widespread WM alterations are found in PD-NCI.
If these hypotheses are correct, the WM alteration in PD would precede the major GM alterations.

Subjects
We recruited 25 PD patients from the Brno Movement Disorders Centre and 21 age-and gender-matched healthy controls (mean age: 57.9±7.24 years, age range: 41-71 years, 8 male). The patients that did not present any evident cognitive impairment at clinical examination were included to the study and were further examined utilizing detailed psychological tests. Five patients were excluded because of MRI technical problems (movement artefacts, restricted field of view, claustrophobia), drop-out rate was 20%. Twenty PD patients (mean age: 61.9 ±7.63; age range: 48-72 years, 11 male) that have included to the study were in the early stages of the disease with slight to moderate motor impairment (Hoehn-Yahr stage 1-1.5) and disease duration up to 5 years. In 10 patients, the right side was predominantly affected. Healthy individuals had no history of neurologic or psychiatric disease. All participants were righthanded ( Table 1).
The study was approved by ethics committee of St. Anne's University Hospital. Written informed consent was obtained from all participants.

Neuropsychological scores
In addition to a neurological examination, comprehensive neuropsychological testing (Mattis Dementia Rating Scale (MDRS), Tower of London, Stroop Test, Rey-Osterrieth Complex Figure Test, Wechsler Memory Scale III, Wechsler Adult Intelligence Scale-, Third Edition (WAIS III) and Verbal Fluency Test) was performed by an experienced psychologist (I.Z.) on the HC group and the PD patient group after they had taken their regular medication.
For further comparisons, individual Z-scores were calculated for the separate subscales of the different tests by subtracting the predefined means from the individual raw scores and then dividing by the predefined standard deviations (Z-score = (raw score-mean STD ) / SD STD ). The obtained Z-scores of the selected subscales were then averaged to produce cognitive domain specific (visuospatial, memory, attention, language, and executive) Z-scores. These cognitive domain scores were used for interaction analyses to evaluate group differences between the slopes of fitted correlation lines (based on: http://core.ecu.edu/psyc/wuenschk/ docs30/CompareCorrCoeff.pdf).
Since the motor symptoms of PD usually start on one side, lateralised volumetric differences were expected. The literature of lateralised neurological diseases demonstrated importance of the flipping of the images along the x-axis for pooling the symptoms virtually to one side [18]. So, as first step flipped analyses were performed to examine connections between GM/WM alterations and the side of the symptom onset. Three different versions of the same analysis were carried out: 1) images of patients with symptoms that started on the left side were mirrored about the x-axis (as if all the patients had symptoms that started on the right side-RightMirr); 2) images of patients with symptoms that started on the right side were mirrored about the x-axis (as if all the patients had symptoms that started on the left side-Left-Mirr); and 3) non-mirrored images were used (symptoms started on the left and right sided mixed-OrigSided).
While the flipping method can only be used if there is no laterality in the investigated parameters, previous studies showed subtle differences between the left and the right side of the brain even in healthy subjects [19,20]. Such asymmetry can profoundly affect the results of flipping analyses and thus challenges data interpretation. For this reason, a second method (tbss_sym) was also used to investigate the normal and altered asymmetry of the white matter diffusion parameters (see details below).
Tract-based spatial statistics. The diffusion data were corrected for eddy currents and movement distortions using eddy_correct tool in FSL [21]. Non-brain parts were removed from all images using the Brain Extraction Tool (BET, [22]). Diffusion tensors at each voxel were fitted by the algorithm included in the Diffusion Toolbox (FDT) of the FMRIB Software Library (FSL v. 4.0, www.fmrib.ox.ac.uk/fsl; [23]). Fractional anisotropy (FA), mean diffusivity (MD), and diffusivity parallel (AD, axial, λ1) and perpendicular (RD, radial, (λ2+λ3)/2) to the principal diffusion direction were computed for the whole brain.
The Tract-Based Spatial Statistics (TBSS [24]) tool was used to perform a voxel-wise statistical analysis of the diffusion tensor maps ("OrigSided"). All of the subjects' FA maps were aligned into a standard space by co-registration to the most "typical" subject, which was the  best target from all FA images based on the least number of transformations to the FA of all other subjects with non-linear registration. This option was chosen in order to achieve a better alignment of WM tracts in our population. A mean FA skeleton was created, representing the cores of all tracts common to the group. Each individual's FA data were then projected onto this skeleton and thresholded at 0.2 FA. Voxel-wise cross-subject statistics analysis was performed using 5000 permutations, testing with the Threshold-Free Cluster Enhancement (TFCE) approach [25] adjusted for age and gender. Analyses were repeated on data flipped to the side where the symptoms were observed at first [18] ("RightMirr" patients, "LeftMirr" patients).
Symmetry in diffusion characteristics was analysed using tbss_sym util with age and gender as covariates on original data without flipping. Three analyses were performed: (A.) symmetry analysis was conducted in healthy controls to depict the normal asymmetry; (B.) symmetry analysis in PD groups: all PD patients (PD_all), patients with right side (PD_right) and patients with left side (PD_left) symptom onset. (C.) 3-group comparison: healthy controls, PD_left and PD_all. Obviously, this analysis is restricted to only those parts of the white matter tracts that are already sufficiently close to being symmetric-i.e. where there is reasonable correspondence in general tract structure between left and right in the brain.
The Johns Hopkins University white-matter atlas was used to identify the anatomical locations of altered regions.
Brain volumetry. Total brain volume was estimated with SIENAX [26], part of FSL. SIE-NAX starts by extracting brain and skull images from the single whole-head input data. Next, tissue-type segmentation with partial volume estimation is carried out [27] in order to calculate the total volume of brain tissue. The volumetric comparison was performed using the Statistical Package for Social Sciences (SPSS 22.0 for OS X, SPSS Inc., http://www.spss.com) with age and gender as covariates, corrected for multiple comparisons.
Voxel-based morphometry. Local GM changes were detected by an "optimised" voxelbased-style protocol (VBM) [28] using FSL [22]. After brain extraction, tissue-type segmentation was carried out by the FAST algorithm [27]. The resulting GM partial volume images were registered to a standard space (MNI152) using linear transformation [29] followed by a non-linear registration [30]. The resulting images were averaged to create a study-specific template, to which the native GM images were then non-linearly re-registered. The registered partial volume images were then corrected for local expansion or contraction by dividing by the Jacobian of the warp field. The modulated segmented images were then smoothed with an isotropic Gaussian kernel with a sigma of 3 mm. Final statistical steps with TFCE were the same as described with TBSS.
Volumetry of deep brain grey matter structures. We used FIRST, a model-based segmentation/registration tool for volume comparison of the subcortical structures of the two groups [31]. This approach uses deformable surface meshes specific to subcortical structures, namely the amygdala, caudate nucleus, hippocampus, pallidum, putamen, and thalamus. Given the observed intensities in a T1-weighted image, FIRST searches through linear combinations of shape modes of variation for the most probable shape instance based on learned models. We decided not to include the nucleus accumbens due to inappropriate segmentation. Table 1 summarizes the basic demographic data and clinical and neuropsychological scores. Patients and controls were matched for age, gender, and years of education. No statistically significant group difference was found in the neuropsychological tests. Hoehn-Yahr stage, age and gender were compared for subgroups of patients (PD_left, PD_right) and no differences were found (p>0.95).

TBSS
The OrigSided AD in the PD group was significantly higher in the superior and anterior corona radiata, internal capsule, and external capsule on the left side than in the HC group (p<0.009, corrected). No other DTI parameter showed significant changes, although a slight trend was detected for MD and RD. Increased MD (p<0.059, corrected, in the left sided: superior corona radiata, superior longitudinal fascicle, external capsule, internal capsule, temporal WM, and prefrontal WM) and increased RD (p<0.090, corrected, in the left sided: cortico-spinal tract, superior corona radiata, body of corpus callosum, and prefrontal WM) was found in PD compared to HC (Fig 1/A). These results became significant in the LeftMirr (Fig 1/B) (MD: Hemispheric comparison in healthy individuals showed a rightward shift of AD, MD and RD dominance (p<0.05, corrected) in the frontal, parietal and temporal white matter, and a rightward shift of FA dominance along the superior longitudinal fasciculus. (The leftward shift in dominance of these parameters is the reverse of the above findings). Hemispheric comparison in PD patients revealed a slight anterior-ward shift of the above parameters, mainly affecting the secondary and primary somatosensory and primary somato-motor areas (S2 Fig). In case of the 3-group comparison (S3 Fig), in PD_right, the asymmetry of in RD values was detected in the left hemisphere (RD was higher in the temporal and parietal lobe), and FA was lower in the left temporal lobe compared to the controls (p<0.05, corrected). In PD_left, the analysis revealed a trend for the same results, but only in the temporal lobe (p<0.07, corrected). After plotting the subject-wise diffusion data under the ROIs of the 3-group comparison result (S4 Fig), the disappearing of the normal asymmetry of the RD and FA parameters could be seen in PD patients.
A tendency for decreased focal GM intensity was found with VBM in the left superior parietal lobe in OrigSided PD (p<0.059, corrected). It became significant in the RightMirr (Fig 2  (p<0.02, corrected) and (p<0.014, corrected) in the LeftMirr comparison.
No significant differences were found with FIRST in the volumes of subcortical GM structures between PD and HC (S5 and S6 Figs).

Discussion
In this study, we investigated cortical/subcortical GM and WM changes in PD patients with normal cognition as compared to healthy individuals. The outcomes revealed limited Our results revealed early susceptibility of both GM and WM of left hemisphere in early stages PD-NCI and align with GM outcomes of Claassen and co-workers [32]. They found no influence of handedness and motor symptom asymmetry on results [32], and the current literature does not contain sufficient explanation of this phenomenon. On one hand, the analyses that included the symptom asymmetry strengthen our conclusions of critical vulnerability of the left hemisphere in early PD, since the results were more pronounced and reached higher level of significance. On the other hand, the normally existing brain asymmetry might disappear in PD due to on-going pathology in the brain, suggesting that PD might be a disconnection syndrome as Caminiti and co-workers [33] claimed in recent study. They revealed loss off functional and structural connection in PD patients' central nerve system caused by axonal loss, which hypothesis could also support our findings.
Our findings of left-sided superior parietal atrophy are in agreement with previous studies with mild posterior atrophy already present in early PD [9,34]. The superior parietal cortex is critically involved in attention and visual processing in healthy subjects We detected no global cortical GM and focal subcortical volume change using different MRI analysis methods. Our results support the conclusions of several studies that did not detect subcortical volume changes [44][45][46][47][48], suggesting that volumetric alteration of deep brain GM structures may occur later in the disease course. Indeed, multiple studies described volumetric abnormalities of subcortical GM structures (e.g. striatum [49,50], caudate nucleus [51], and thalamus [49]) in patients with advanced PD, indicating that subcortical atrophy occurs later during the disease course.
In contrast to the subtle GM atrophy, we observed a robust impairment of subcortical WM [36, 52,53]. WM damage in PD, mostly related to cognitive decline, was reported in several studies [7,[54][55][56]; normal or minor changes in non-demented PD patients were also reported [5]. Limited GM alteration accompanied by widespread WM changes were reported in a few studies [9,57]. The critical conflicts between studies might stem from differences in imaging and analysis protocol or sample size and characteristics [54].
Similarly to some of previous PD studies [54,58,59], but contradictory to the others [60,61], we detected an increase in AD in PD-NCI compared to HC. AD is considered a measure of axonal consistency with higher values potentially indicating an increase of axonal co-linearity, however axonal loss in co-occurrence of inflammation [62] may provide additional explanation to elevated AD in the PD population. The activation of inflammatory processes related to activation of microglia caused by alpha-synuclein aggregation may be associated with WM abnormalities in PD [63]. Studies demonstrated decrease of AD in early stages of axonal damage [16] that is followed by pseudo-normalization of AD due to clearance of axonal and myelin debris in the WM [64] and pointed to evolving AD abnormalities during the disease on course.
While RD quantifies diffusion perpendicular to the axonal axis, with an increase indicating an abnormally low myelin content [65], the diffusion parameters indirectly reflect changes in the microstructural tissue. The elevation of RD, which quantifies diffusion perpendicular to the axonal axis and indicates an abnormally low myelin content [17], was congruently detected in various PD stages [7,58,59]. Theilmann et al [55] detected similar co-occurrence of elevated RD with higher AD and suggested that these changes may reflect early cell loss and gliosis in PD. It has been reported that PD is primarily caused by a synaptic dysfunction with axonal transport problems leading to subsequent cell death [66]. Also, AD and RD alterations may together contribute to increased MD that could refer to loss of tissue density [67] and it is in agreement with previously published studies [36,52]. The pathology underlying the loss of microstructural WM integrity could be explained with immunocytochemical evidence for the presence of ubiquitin and alpha-synuclein inclusions in the axons of Lewy body disease cases, which may impair axonal transport before cell body damage [68]. The impairment of WM might underlie the disturbance of large-scale network connectivity in PD-NCI [69]. While DTI studies congruently pointed to complex PD-related WM alteration, the interpretation of actual tissue abnormalities remains challenging due to non-specificity of DTI parameters, which indirectly refer to microstructural tissue change by calculating the changes in water diffusion per se [65].
WM lesions (WML) observed in structural MRI studies have been associated with cognitive impairment in PD [70] (for review see: [71]); however, our participants were relatively young and exhibited only minor WML on anatomical images. WM changes assessed by TBSS were found in PD but not in HC, with a comparable burden of ischemic WML, indicating that the DTI-assessed WM changes reflect the initial stages of neurodegeneration [72,73]. A recent follow-up DTI study confirmed more pronounced changes in diffusivity parameters during aging in PD than in HC [74].
Our findings in the PD-NCI cohort expand on the published works of other groups studying WM and GM integrity [40,57,75] by adding an analysis of subcortical GM, including the basal ganglia. Our results of nearly normal cortical GM and normal subcortical GM volume do not support the hypothesis that WM alterations reflect Wallerian degeneration and are secondary to early cortical/subcortical GM atrophy [76].

Limitations
The possible limitations of our study should be noted. The small sample size limits power of the statistical analyses and may cause that we failed to reach statistical significance in certain measures. The diagnosis of PD lacks histopathological verification. The subjects were assessed while taking their medication, which could influence cognitive outcomes. The distinct analysis methods used in our study might have different levels of sensitivity. We cannot exclude the possibility that there were subtle changes in the GM that were not revealed by the methods we used. On the other hand, the WM impairment was robust, as in the report by Rae et al. [54]. The interpolation of DTI data makes it possible to see subtle anatomical details, although it might have potential disadvantages. While high-order interpolations may cause ringing artefacts, especially in border areas between compartments with different T2 relaxation times, no significant interpolation effects were detected on DTI indices [77]. While TBSS requires at least one diffusion-unweighted scan, averaging of multiple b0 images improves SNR and thus estimation of diffusion tensor [78]. Lower SNR of b0 scans concerns the Rician noise distribution and the presence of the rectified noise floor that underestimates mean diffusivity particularly in regions with high anisotropy [79]. However, the potential bias introduced by a limited number of b0 is common for all participants in the study, thus unlikely caused differences between patients and healthy individuals. Our findings are preliminary; future prospective longitudinal studies are needed to assess the long-term progression of WM and cortical and subcortical atrophy.

Conclusions
Our findings confirmed widespread WM pathology in PD. Given the small extent of GM atrophy, the critical WM deficit in PD-NCI challenges the traditional view that WM degeneration, including loss of axons and myelin, occurs secondary to cortical pathology. And this WM alteration might lead to structural disconnection with the loss off normal existing brain asymmetry resulting dysfunctional brain networks. Hence, WM alterations in PD might be a sensitive preceding sign of the neuronal loss in associated GM regions [4].

Acknowledgments
The authors wish to thank Jana Víšková and Petr Mikulenka who assisted with the data acquisition. Peter Mikulenka participated in the first stages of data processing; he cannot be listed among the authors as he left the department and could not be contacted to approve the manuscript. Support was provided by the project CEITEC 2020 (LQ1601) from the Ministry of Education, Youth and Sports of the Czech Republic and from the National Center for Advancing Translational Sciences of the National Institutes of Health, Award Number UL1TR000114 (to AS) and European Union's Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No 691110 (MICROBRADAM). We acknowledge the core facility MAFIL of CEITEC supported by the MEYS CR (LM2015062 Czech-BioImaging). On behalf of all authors, the corresponding author states that there is no conflict of interest.