Neurocognitive and psychiatric disorders‐related axonal degeneration in Parkinson's disease

Abstract Neurocognitive and psychiatric disorders have significant consequences for quality of life in patients with Parkinson's disease (PD). In the current study, we evaluated microstructural white matter (WM) alterations associated with neurocognitive and psychiatric disorders in PD using neurite orientation dispersion and density imaging (NODDI) and linked independent component analysis (LICA). The indices of NODDI were compared between 20 and 19 patients with PD with and without neurocognitive and psychiatric disorders, respectively, and 25 healthy controls using tract‐based spatial statistics and tract‐of‐interest analyses. LICA was applied to model inter‐subject variability across measures. A widespread reduction in axonal density (indexed by intracellular volume fraction [ICVF]) was demonstrated in PD patients with and without neurocognitive and psychiatric disorders, as compared with healthy controls. Compared with patients without neurocognitive and psychiatric disorders, patients with neurocognitive and psychiatric disorders exhibited more extensive (posterior predominant) decreases in axonal density. Using LICA, ICVF demonstrated the highest contribution (59% weight) to the main effects of diagnosis that reflected widespread decreases in axonal density. These findings suggest that axonal loss is a major factor underlying WM pathology related to neurocognitive and psychiatric disorders in PD, whereas patients with neurocognitive and psychiatric disorders had broader axonal pathology, as compared with those without. LICA suggested that the ICVF can be used as a useful biomarker of microstructural changes in the WM related to neurocognitive and psychiatric disorders in PD.


| INTRODUC TI ON
Parkinson's disease (PD) is primarily considered to be a neurodegenerative movement disorder that is characterized by bradykinesia, rigidity, resting tremor, and postural abnormalities, with the pathological hallmarks of the loss of dopaminergic neurons in the substantia nigra (Fearnley & Lees, 1991) and the widespread aggregation of α-synuclein in the form of Lewy pathology (Braak et al., 2003).
In addition to dopamine-related impairments, it has become evident that PD involves multiple neurotransmitter pathways, including the noradrenergic, serotonergic, and cholinergic systems, within the brain that are associated with a wide variety of motor and non-motor symptoms (Schapira, Chaudhuri, & Jenner, 2017). Among the nonmotor symptoms, neurocognitive and psychiatric symptoms (NCPs) have received attention because these signs are highly prevalent, contribute to severe disability (Weintraub, Moberg, Duda, Katz, & Stern, 2004), impair quality of life (Barone et al., 2009), and shorten life expectancy (Chaudhuri, Healy, Schapira, & National Institute for Clinical Excellence, 2006).
Defects to the structure of the white matter (WM) are associated with many disorders affecting cognition and physiological states (Fields, 2008). Diffusion tensor imaging (DTI) has been widely used to investigate pathological changes in the WM of PD patients by probing the diffusivity of water molecules within the WM tracts (Cochrane & Ebmeier, 2013). Previous studies that utilized DTI to evaluate the WM of PD patients with NCPs reported conflicting observations, such as lower fractional anisotropy (FA) and higher mean diffusivity (MD) (Agosta et al., 2014;Hattori et al., 2012;Huang et al., 2014;Yao et al., 2016;Zhong et al., 2013), as well as unaltered FA and MD (Melzer et al., 2013;Surdhar et al., 2012). Sample size, demographic, and clinical data variances, however, are likely to have contributed to these discrepant findings but may also be due to the limitations of DTI. First, DTI assumes that water diffusion has a Gaussian probability distribution (Basser & Jones, 2002).
However, water molecules in biological structures are hindered by barriers, such as the cell membrane and the internal membranes that encase the organelles; thus, DTI is not appropriate for modeling of non-Gaussian diffusion in living tissues (Assaf & Pasternak, 2008).
Second, DTI-derived indices have been shown to be sensitive, but not specific, to microstructural changes. A decrease in FA accompanied by an increase in MD may be attributed to reduced axon density, increased axonal dispersion, and demyelination (Assaf & Pasternak, 2008).
On the contrary, neurite orientation dispersion and density imaging (NODDI) can be used to address the limitations of DTI. NODDI was developed on the basis of a multicompartment (intracellular, extracellular, and cerebrospinal fluid) diffusion model with the particular advantage of providing useful microstructural indices from multishell diffusion magnetic resonance imaging (MRI) data that can be acquired in a clinically feasible amount of time (Zhang, Schneider, Wheeler-Kingshott, & Alexander, 2012). The intracellular volume fraction (ICVF) and the orientation dispersion index (ODI) are indices of NODDI that are assumed to reflect the packing density of axons in WM and the spatial organization of the axons, respectively, which are two disentangled facets of FA (Zhang et al., 2012). Previous studies have demonstrated the superiority of NODDI over DTI for the detection of PD pathology in the substantia nigra (Kamagata et al., 2016), nigrostriatal pathway (Andica et al., 2018), and gray matter (GM) (Kamagata et al., 2017). A simplified model of NODDI, neurite density imaging, has also been used to demonstrate reduction in the density of some WM areas of patients with PD (Surova et al., 2016).
In this study, we used NODDI to characterize in vivo WM pathology associated with the NCPs of PD patients. Using tract-based spatial statistics (TBSS) and tract-of-interest (TOI) analyses we evaluated differences in NODDI metrics between healthy controls and PD patients with and without NCPs in addition to DTI metrics. Also, recent studies of PD patients (Aquino et al., 2014;Kamagata et al., 2017;Melzer et al., 2015) have reported multimodal MRI data in terms of a robust biomarker to further elucidate the pathology of the disease. Those data, however, were analyzed separately and only detected single-dimensional information by each measure; thus, as a consequence, failed to model common variance across features.
In an attempt to overcome these limitations, linked independent component analysis (LICA) was applied in the present study (Groves, Beckmann, Smith, & Woolrich, 2011;Groves et al., 2012) as an integrated approach by fusing different modalities to model inter-subject variabilities across the measured indices.

| Study participants
The cohort of this retrospective case-control study included 39 PD patients who remained free from atypical parkinsonism and exhibited a good response to anti-parkinsonian therapy for 18 months or more after diagnosis. PD was diagnosed by three movement disorders specialists (TH, GO, and YS) based on the clinical diagnostic criteria for PD of the Movement Disorder Society (MDS; Postuma et al., 2015). At the time of MRI and clinical examination, all PD patients were taking levodopa in combination with a dopamine decarboxylase inhibitor (benserazide or carbidopa). Disease severity was then assessed on the basis of the non-motor and motor scores of the MDS-Unified Idiopathic PD Rating Scale (MDS-UPDRS) parts I and III, respectively (Goetz et al., 2008).
The NCPs of PD patients might be caused by overlapping pathophysiology, whereas some symptoms might be similar (Fields, 2017). Thus, PD patients were categorized into groups with or without NCPs (PD-wNCPs and PD-woNCPs, respectively), where the total MDS-UPDRS score of items I.1-I.6 (I.

| Diffusion MRI pre-processing
The diffusion-weighted data were corrected for susceptibilityinduced geometric distortions, eddy current distortions, and inter-volume subject motion using the EDDY and TOPUP toolboxes (Andersson & Sotiropoulos, 2016). We then visually assessed all DWI datasets in the axial, sagittal, and coronal views. All datasets were free from severe artifacts, such as gross geometric distortion, signal dropout, and bulk motion. The resulting images were fitted to the NODDI model (Zhang et al., 2012) using the NODDI MATLAB Toolbox 5 (http://www. nitrc.org/proje cts/noddi_toolbox). Maps of ICVF, ODI, and isotropic volume fraction (ISO, Figure S1) were generated using Accelerated Microstructure Imaging via Convex Optimization (Daducci et al., 2015). The diffusion tensor was estimated using ordinary least squares applied to the diffusion-weighted images with b = 0 and 1,000 s/mm 2 . FA, MD, axial diffusivity (AD), and radial diffusivity (RD) maps ( Figure S1) were then generated for all subjects using the DTIFIT tool of the FMRIB Software Library 5.0.9 (FSL, Oxford

TA B L E 1 Demographic characteristics of the participants
Center for Functional MRI of the Brain, UK; www.fmrib.ox.ac.uk/ fsl) to fit the tensor model to each voxel of the DWI data (Basser, Mattiello, & LeBihan, 1994).

| TBSS analysis
Voxel-wise statistical analysis was carried out using TBSS (Smith et al., 2006)  First, we performed nonlinear registration of FA images of all subjects into 1 × 1 × 1 × mm 3 Montreal Neurological Institute (MNI 152) common space (a normalized/averaged brain) using the FMRIB's nonlinear registration tool (Jenkinson et al., 2012). Second, the transformed FA images were average to create a mean FA image.
Third, the mean FA was thinned to create a mean FA skeleton, which represented the centers of all tracts common to the groups. The threshold of the mean FA skeleton was set to >0.20 to include the major WM pathways and exclude the peripheral tracts and GM. The aligned FA map of each subject was then projected onto the skeleton. Finally, the NODDI (ICVF and OD) and DTI (MD, AD, and RD) maps were projected onto the mean FA skeleton after applying the warping registration field of each subject to the standard space. and SLF temporal part) (Hua et al., 2008;Wakana et al., 2007). The average diffusion metric was averaged over all WM skeleton voxels comprising a given region, as delineated by the atlas, for all subjects.

| Voxel-based morphometry (VBM)
VBM was used to obtain WM volumetry. com/produ cts/matlab.html; Ashburner & Friston, 2000). We then spatially normalized the segmented WM and GM images to the customized template in the standardized anatomic space with the use of the "Diffeomorphic Anatomical Registration Using Exponentiated Lie Algebra" (DARTEL) algorithm (Ashburner, 2007). To preserve the WM and GM volumes within each voxel, the Jacobian determinants derived from the spatial normalization, acquired with the DARTEL algorithm, and an 8 mm FWHM Gaussian kernel were used to modulate and smooth the images, respectively.   (Cohen, 1992).

| Study participants
The demographic and clinical details of all groups are summarized in Table 1. There was no significant difference in age and sex among the three groups (healthy controls, PD-woNCPs, and PD-wNCPs) or with regard to disease duration, Hoehn and Yahr stage, MDS-UPDRS part III score, and levodopa equivalent daily dosage between the PD-woNCPs and PD-wNCPs groups.  Table 2. There were no significant differences in FA values between the PD-woNCPs and PD-wNCPs groups, in AD values between the PD-woNCPs group and healthy controls, and in ODI values between the PD-woNCPs group and healthy controls and between PD-woNCPs and PD-wNCPs groups.

| Voxel-based morphometry
There was no significant difference in total and focal WM volumes among the three groups.

| Correlation analysis
There were no significant correlations between all metrics in all analyses and disease duration or MDS-UPDRS part III in the whole PD group and MDS-UPDRS part I.1-I.6 in the PD-wNCPs group.

| D ISCUSS I ON
In the present study, NODDI, a novel technique for analyzing multishell DWI data, was applied to investigate microstructural changes in the WM related to the NCPs of PD patients. The major findings of this study were that (a) compared to DTI, NODDI gave a pattern of results that was more regionally specific, (b) both TBSS and TOI anal-

F I G U R E 1
Comparison of DTI (FA, MD, AD, and RD) and NODDI (ICVF) measures among the healthy control, PD-woNCPs and PD-wNCPs groups. Using TBSS analysis, significantly lower FA and ICVF (blue-light blue) and significantly higher MD and RD (red-yellow) values were observed in the PD-woNCPs group compared with the healthy controls. Compared with the healthy controls, the PD-wNCPs group had significantly lower FA and ICVF and higher MD, RD, and AD values. Compared with the PD-woNCPs group, the PD-wNCPs group had significantly lower ICVF and higher MD, RD, and AD values. There were no significant differences in ODI values between the groups. The FA skeleton with FA > 0.2 is shown in green. To aid visualization, results are thickened using the fill script implemented in FMRIB Software Library. Abbreviations: DTI, diffusion tensor imaging; FA, fractional anisotropy; HC, healthy control; ICVF, intracellular volume fraction; MD, mean diffusivity; NODDI, neurite orientation dispersion and imaging; ODI, orientation dispersion index; PD-woNCPs, Parkinson's disease without neurocognitive-psychiatric symptoms; PD-wNCPs, Parkinson's disease with neurocognitive and psychiatric symptoms; RD, radial diffusivity [Color figure can be viewed at wileyonlinelibrary.com] The anterior brain has been implicated as a region that is more prone to Lewy pathology than the posterior brain (Cochrane & Ebmeier, 2013;Luk & Lee, 2014). Similarly, visualization of the TBSS results showed that axonal loss occurred predominantly in the anterior aspect of the brain of the PD-woNCPs group, as compared with that of the healthy controls. This finding is consistent with that of a previous meta-analysis showing that the changes in diffusion measures occur consistently in the frontal lobe of PD patients (Cochrane & Ebmeier, 2013).

TA B L E 2 Tract-based spatial statistics analysis of DTI and NODDI indices in patients with Parkinson's disease and healthy controls
Notably, axonal pathology occurred throughout the WM and predominantly in the posterior aspect of the brain (such as forceps  Aarabi, 2018). Furthermore, in TOI analysis, WM microstructural changes (i.e., lower ICVF and higher MD and RD) in PD-wNCPs, as compared with those in PD-woNCPs, were consistently found in the SLF by TOI analysis. This finding suggests that NCPs in PD is related to impaired long WM nerve fibers, especially the SLF. SLF is an association fiber bundle connecting the frontal, occipital, parietal, and temporal lobes and is closely related to the functions of the frontal lobe (Jiang, Shi, Niu, Xie, & Yu, 2015). Indeed, the NCPs of PD patients have been linked to the impairment of multiple neurotransmitter (dopaminergic Biundo, Weis, & Antonini, 2016;Schapira et al., 2017), noradrenergic (Delaville, Deurwaerdere, & Benazzouz, 2011;Schapira et al., 2017), and cholinergic (Nagasaka, Watanabe, & Takashima, 2017;Perez-Lloret & Barrantes, 2016) pathways that are projected to the frontal lobe. Furthermore, previous studies have consistently showed alterations in the diffusion metrics in the SLF of PD patients with cognitive impairment Kamagata et al., 2012Kamagata et al., , 2013, apathy (Lucas-Jimenez et al., 2018), and depression (Huang et al., 2014).
Here, comprehensive analysis was also performed by fusing data of the measured indices (DTI [FA, MD, AD, and RD], NODDI [ICVF and ODI], and VBM [WM volume]) using LICA (Groves et al., 2011(Groves et al., , 2012. LICA is a probabilistic technique based on a Bayesian framework that provides an effective way for simultaneously modeling covariances across modalities (Groves et al., 2011(Groves et al., , 2012 Increased MD, an index of increased accumulation of water content and spacing between membrane layers due to neuronal loss (Zhang et al., 2012), in WM tracts had overlapped, at least to some extent, with those displaying lower ICVF. MD was shown to be more accurate than FA (Wiltshire et al., 2010), and a recent investigation in patients with premanifest Huntington's disease using NODDI also showed widespread reduction in ICVF F I G U R E 2 Significant tracts from tract-of-interest analysis comparing diagnostic groups. (a) Mean of each measure in the PD-woNCPs and PD-wNCPs groups (represented as the percentage difference from the healthy controls). Significant tracts (*p < .05, **p < .01, ***p < .001) are displayed in color, whereas non-significant tracts are shown in gray. ★ Tracts with significant differences between the PD-woNCPs and PD-wNCPs groups. that overlapped with increased MD . A strong negative correlation has been demonstrated between the ICVF and MD maps in the human brain (Fukutomi et al., 2018). On the contrary, other DTI indices have been shown to be inconsistent.
For example, in TBSS and TOI analyses, the PD-woNCPs and PD-wNCPs groups had lower FA and higher RD values in broad areas of the WM, as compared with the healthy controls. FA and AD failed to demonstrate differences between the PD-woNCPs and PD-wNCPs groups and between the healthy control and PD-woNCPs groups, respectively. Likewise, the studies by Kamagata et al. (2012) and Melzer et al. (2013) also failed to show the differences between PD with versus. without dementia or between PD with mild cognitive impairment versus. PD with dementia, respectively. DTI indices are reportedly sensitive, but not specific, to microstructural changes (Andersson & Sotiropoulos, 2016).
Decreases in FA may arise not only from the axonal loss but also from demyelination and changes in the size of axons   (Huang et al., 2014).
Furthermore, although we only included PD patients with no significant difference in MDS-UPDRS part III and Hoehn and Yahr score, the presence of motor symptoms may have interfered with the ability to evaluate non-motor abnormalities (Schapira et al., 2017).
Finally, in this study, we only obtained one b = 0 image; thus, it is not possible to correct the signal drift, which causes a global signal decrease with subsequently acquired images. This might have affected the estimation of the diffusion parameters (Vos et al., 2017).

| CON CLUS IONS
The results of this study support the view that axonal loss is a major factor underlying NCPSs-related WM microstructural changes in PD.
Decreased axonal density, as a result of α-synuclein deposition and multiple neurotransmitter deficiencies, was found to be broader (posterior predominant) in the PD-wNCPs group than in the PD-woNCPs group. Furthermore, the LICA results showed that ICVF can be used as a useful measure for the evaluation of microstructural changes in the WM related to the NCPs of patients with PD. Taken together, these results suggest the potential of NODDI as an imaging biomarker for NCPs-related microstructural changes in the WM of PD patients.

D ECL A R ATI O N O F TR A N S PA R EN C Y
The authors, reviewers, and editors affirm that in accordance with the policies set by the Journal of Neuroscience Research this manuscript presents an accurate and transparent account of the study being reported and that all critical details describing the methods and results are present.

ACK N OWLED G M ENTS
We thank Mana Kuramochi for her research assistance. We also thank all the patients and healthy controls who participated in this study.

S U PP O RTI N G I N FO R M ATI O N
Additional supporting information may be found online in the Supporting Information section.