Brain microstructural alterations of depression in Parkinson's disease: A systematic review of diffusion tensor imaging studies

Abstract Depression, a leading cause of disability worldwide, is also the most prevalent psychiatric problem among Parkinson disease patients. Both depression and Parkinson disease are associated with microstructural anomalies in the brain. Diffusion tensor imaging techniques have been developed to characterize the abnormalities in cerebral tissue. We included 11 studies investigating brain microstructural abnormalities in depressed Parkinson's disease patients. The included studies found alterations to essential brain structural networks, including impaired network integrity for specific cortical regions, such as the temporal and frontal cortices. Additionally, findings indicate that microstructural changes in specific limbic structures, such as the prefronto‐temporal regions and connecting white matter pathways, are altered in depressed Parkinson's disease compared to non‐depressed Parkinson's disease and healthy controls. There remain inconsistencies between studies reporting DTI measures and depression severity in Parkinson disease participants. Additional research evaluating underlying neurobiological relationships between major depression, depressed Parkinson's disease, and non‐depressed Parkinson's disease is required to disentangle further mechanisms that underlie depression and related somatic symptoms, in Parkinson disease.

leading cause of disability worldwide, is also the most prevalent psychiatric problem among PD patients, substantially worsening their quality of life (Aarsland et al., 1999;Tandberg et al., 1996). Although depressive symptoms, such as sadness and anhedonia, occur in nearly half of the patients with PD (Gotham et al., 1986), the identification of these features is a clinically challenging task since psychomotor and somatic presentations of depression can highly resemble the typical hallmarks of PD, such as bradykinesia and hypomimia, as well as decreased concentration and appetite. Thus, depressed PD (DPD) patients are at risk of delayed diagnosis and under treatment (Schrag, 2006).
It is well documented that the development of mood disturbances can precede motor manifestations in PD cases, underscoring the fact that associated depression does not emerge only secondary to the disabilities caused by PD but possibly shares similar pathophysiology (Ishihara & Brayne, 2006). While the exact pathogenesis of PD remains unclear, the accumulation of Lewi bodies, mainly containing α-synuclein protein, as well as mitochondrial dysfunction and defective proteolysis, in regions such as the substantia nigra, are considered to be possible causes of neuronal cell death and the resultant reduction of striatal dopamine levels observed in PD brains (Calo et al., 2016;Lim & Zhang, 2013;Shulman et al., 2001;Wakabayashi et al., 2007). In this regard, neuroimaging techniques can provide invaluable insights into biochemical, structural, and functional alterations in the brains of DPD patients paving the way for the development of more precise and objective diagnostic and prognostic criteria in research and clinical settings (Won et al., 2019). Positron emission tomography (PET) and single-photon emission tomography (SPECT) studies demonstrate metabolic changes in the striatum, limbic thalamus, and frontal lobe, involving dopaminergic and serotonergic pathways (Chagas et al., 2013;Weintraub et al., 2005). Likewise, magnetic resonance imaging (MRI) studies of DPD demonstrate gray matter (GM) and white matter (WM) volume alterations (Scheuerecker et al., 2010;Wen et al., 2016). Research has also identified altered brain networks involving subcortical, frontolimbic, and corticocortical fibers of DPD patients compared to non-depressed PD (NDPD) patients (Ansari et al., 2019;van Mierlo et al., 2015).
More recently, diffusion tensor imaging (DTI) techniques, or diffusion-weighted imaging (DWI), have been developed to characterize alterations in the microstructure and integrity of WM fiber tracts in psychiatric and neurological disorders, including PD (Basser et al., 1994;McKinstry et al., 1992). DTI measures the size and direction of water molecules' diffusion modeled using a voxel-based threedimensional Gaussian plot distribution (Sanjari Moghaddam et al., 2019;van Ewijk et al., 2012). In the absence of any hindrance, the diffusion coefficient will be equal (isotropic) in all directions, as in GM and pure water. However, the presence of myelin sheath and cellular membranes in the WM shall act as barriers; hence this coefficient differs in each direction (anisotropic), with maximum diffusion along the axon bundles and minimum diffusion perpendicular to the axon (Beaulieu, 2002). Two main DTI indexes indicating WM tracts' organization and direction and myelination are mean diffusivity (MD) and fractional anisotropy (FA). As a mean of the water molecules' diffusion coefficients in primary directions, MD is higher in free extracellular spaces reflecting the easier movement of water in these areas regardless of the direction and lower in an intact WM due to limited free diffusion (Tievsky et al., 1999).
On the other hand, the FA index represents the directionality of water molecules ranging from zero (completely isotropic diffusion) to one (completely anisotropic diffusion). Unlike MD, decreased FA generally indicates diminished WM integrity (Alexander et al., 2007). Two other DTI metrics, axial diffusivity (AD) and radial diffusivity (RD) quantify water diffusion in the direction of nerve fibers and perpendicular to them, respectively. Although AD is generally considered to be an indicator of axonal integrity, and RD an indicator of myelin sheath intactness, the interpretation of these measurements in neurological disorders should be made cautiously as other parameters, such as axonal diameter and density, can influence these indices (Sanjari Moghaddam et al., 2019;Wheeler-Kingshott & Cercignani, 2009).
Diffusion tensor imaging measures are robust biomarkers for progression in early-stage PD and potential biomarkers for other stages of this disease (Mitchell et al., 2021); however, inconsistent DTI findings have been reported in relation to symptom severity of DPD patients. Whereas several studies have reported decreased FA in the group of WM tracts located in the thalamic, limbic, frontal, and cerebellar areas (Ghazi Sherbaf et al., 2018;Matsui et al., 2007;Wu et al., 2018), other studies demonstrate no relationships between DTI indices and depression symptoms and an absence of specific changes in WM integrity of DPD patients compared with NDPD individuals (Lacey et al., 2019;Surdhar et al., 2012). Considering the potential clinical and research utility of DTI measures for the early assessment of patients with PD and depressive disorders, we conducted the current review. The intent was to systemically tabulate and assess the results of DTI studies that specifically compared white matter microstructure and white matter tracts between DPD patients, NDPD patients, and healthy controls (HC), and possible relationships to clinical measures of depression.

| Study selection and data extraction
Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) guidelines were followed to perform the current analysis (JPT et al., 2020); the study protocol was designed and registered at the International Prospective Register of Systematic Reviews (PROSPERO) website (Registration No. CRD42021258388). We searched PubMed and EMBASE databases using a combination of keywords ((("Diffusion Tensor Imaging" OR "diffusion weighted magnetic resonance imaging" OR "Diffusion Tractography" OR "diffusion weighted imaging" OR "diffusion weighted MRI" OR "DTI" OR "white matter" OR "grey matter" OR "gray matter")) AND (("Parkinson Disease" OR "Idiopathic Parkinson's Disease" OR "Lewy Body Parkinson's Disease" OR "Parkinson's Disease, Idiopathic" OR "Parkinson's Disease, Lewy Body" OR "Parkinson Disease, Idiopathic" OR "Parkinson's Disease" OR "Idiopathic Parkinson Disease" OR "Lewy Body Parkinson Disease" OR "Primary Parkinsonism" OR "Parkinsonism, Primary" OR "Paralysis Agitans"))) AND (("Depression" OR "Depressive Symptoms" OR "Depressive Symptom" OR "Symptom, Depressive" OR "Symptoms, Depressive" OR "Emotional Depression" OR "Depression, Emotional" OR "Depressions, Emotional" OR "Emotional Depressions")) to identify possibly relevant literature published up to June 2021. We included all case-control, cohort, and cross-sectional human studies examining DPD patients for WM microstructure changes in DTI measurements. Full-text manuscripts were available regardless of their publication date, geographic location, and participants' age group.
Animal studies, conference abstracts, non-English articles, nonoriginal literature including book chapters, reviews, case studies, and letters were excluded.
Comprehensive review of the manuscripts entailed the extraction of the following data from the included studies: first author's name, publication year and location, number of DPD, NDPD, and HC subjects, demographic and medical data of participants, including age, sex, PD duration, depressive state, Parkinson medication, other medications, and common neurological findings, imaging parameters that included DTI field strength and DTI analysis method, and betweengroup DTI findings.

| Quality assessment
The quality of the included studies was evaluated by two reviewers (AJ and MAS) using the Newcastle-Ottawa scale (NOS) for nonrandomized studies, which assesses three methodological quality aspects of sample selection, case and control groups' comparability, and exposure determination method with a maximum score of four, two, and two for each domain, respectively (Stang, 2010 ; Table S1). Moreover, the risk of publication bias for each study was appraised through criteria developed by Viswanathan et al. in a design-specific manner (Viswanathan et al., 2008;Table S2). A third reviewer (SM) was consulted in case of any discrepancies in this process.

| Summary of reviewed studies
The search of databases yielded 547 articles. After automated and manual removal of duplicate records from search results (n = 100), the remaining articles underwent screening by two authors (AJ and MAS) based on their title or abstract, which resulted in the exclusion of an additional 419 records out of 447. Finally, 17 additional studies were excluded in the full-text screening based on: not being original research (n = 11); measured variables other than those of interest (n = 3); unavailability of full-text (n = 3). In the next phase, backward reference screening was also conducted on primarily included records to detect potentially missed eligible articles. Conflicts that emerged in this stage were resolved through discussion. Figure 1 illustrates the complete search and screening process resulting in the final selection of 11 studies for qualitative synthesis.
Eleven DTI studies were extracted in total. All studies were casecontrol studies except for the study by Won et al., which was a retrospective cohort (Won et al., 2019). Diagnosis of DPD and NDPD in studied patients was verified in all studies. The severity of depression in DPD was assessed using well-established mood rating scales, including the Hamilton Depression Rating Scale (HDRS; Hamilton, 1960), the Beck's Depression Inventory (BDI;Beck et al., 1960), and Geriatric Depression Scale (GDS; Meer & Baker, 1966). In addition to these mood rating scales, the studies of Li et al., Hu et al., and Huang et al. had an experienced psychiatrist diagnose depression using the Diagnostic and Statistical Manual of Mental Disorders (DSM) criteria Huang et al., 2014;Li et al., 2020). In the Huang et al. study, depression was diagnosed as post-morbid to PD (Wen et al., 2016). Studies extracted all confirmed DPD subjects and separated DPD subjects from NDPD subjects.
Study demographics and participants' clinical data are summarized in Table 1. Dissemination of key features, such as age and PD scale scores, can be seen in (Figure 2). Six studies were conducted in China from 2018 onwards (Gou et al., 2018;Hu et al., 2020;Huang et al., 2014;Li et al., 2010Li et al., , 2020Won et al., 2019). Subjects matched for age and sex were used in all included studies except for the study by Prange et al., which matched study subjects just for age, as well as the studies of Ansari et al. and Matsui et al. in which cases and controls were not matched (Ansari et al., 2019;Matsui et al., 2007;Prange et al., 2019). Disease duration was more than 5 years in all studies that reported this parameter. The mean age of participants in all studies was more than 50 years.
Most of the included studies did not report any information regarding the type or duration of PD or depression medication treatments (Ansari et al., 2019;Ghazi Sherbaf et al., 2018;Hu et al., 2020;Matsui et al., 2007;Prange et al., 2019;Won et al., 2019); however, three studies specified no anti-depressant treatment among their participants (Gou et al., 2018;Huang et al., 2014;Li et al., 2020). Regarding PD symptom scores, three studies did not provide Hoehn and Yahr (H&Y) scale scores (Huang et al., 2014;Li et al., 2010;Won et al., 2019). Nevertheless, Unified Parkinson's Disease Rating Scale (UPDRS-III) scores were reported in all studies except for the study by  Table 2).
Only one study reported morphometric data in addition to DTI data . All included studies described abnormalities observed in the brain microstructure or network properties of the brain. Tractography, a 3D reconstruction method to evaluate neural tracts using DTI data (Berman et al., 2008), was the most used analysis method. Other applied methods included region of interest (ROI), tract-based spatial statistics (TBSS), a voxel-based hypothesis-free technique that utilizes FA but can only be applied to WM, and voxelbased analysis (VBA) that contrasts the DTI data in each voxel to a template and, unlike TBSS, is applicable to GM and WM. The DTI measures assessed in the included studies are detailed in (Figure 2).
Most all studies analyzed the whole brain or multiple major WM regions or tracts, whereas Li et al. analyzed only the thalamus region (Li et al., 2010) (Table 3).
Besides comparisons made between DPD and NDPD patients, all included studies except Prange et al. (2019), also made comparisons with other diagnostic groups, including three studies that compared DPD patients to HC (Ghazi Sherbaf et al., 2018;Hu et al., 2020;Li et al., 2020), one study that compared apathetic PD versus non-apathetic PD cases and apathetic PD versus HC cases (Prange et al., 2019), and one study that compared DPD with rapid eye movement sleep behavior disorder(RBD) versus DPD without RBD (Ghazi Sherbaf et al., 2018; Table 1). Moreover, four studies (Gou et al., 2018;Huang et al., 2014;Li et al., 2010;Won et al., 2019) specifically assessed correlations between diffusivity or network measures with H&Y, UPDRS, and GDS measures of DPD (Table 2).

| Overview
As illustrated in Table 2, studies using ROI analysis showed no significant difference in FA measures between groups (Lacey et al., 2019;Li et al., 2020); however, the study of Matsui et al. revealed a significant FA decrease in the frontal ROIs (Matsui et al., 2007). Similarly, both increased MD and AD values   7.5 ± 7.5 6.2 ± 6.7 6.3 ± 6.8 8.5 ± 7.5 - Additionally, two studies reported no differences in RD between DPD and NDPD groups (Lacey et al., 2019;Li et al., 2020). TBSS analysis revealed no AD and RD groups differences ( Two studies conducting deterministic tractography analyses demonstrated alterations in network integration variables, mainly a significant increase in clustering coefficient (γ) and shortest path length (λ) and a significant decrease in global efficiency in DPD compared to NDPD and HC (Gou et al., 2018;Hu et al., 2020). Those findings suggest a higher capability of a local combination of data and slower interactional speeds between brain regions with a reduced capacity to process information. ( Figure 3).
As noted above, Hu et al. reported an increase in the clustering coefficient (γ) and local efficiency in DPD compared to NDPD patients . However, NDPD patients showed a decrease in both parameters compared to HC. Although small worldness (σ:γ/λ) was decreased in all PD groups compared to NDPD and HC in the study by Hu et al. , Gou et al. found this to be increased in NDPD compared to HC subjects (Gou et al., 2018). Betweenness centrality was not reported by any of the network studies.

| Between-group contrasts: brain lobes
Alterations in diffusivity in the frontal, temporal and parietal regions were demonstrated in four studies comparing DPD and NDPD subjects, including two network studies. Probabilistic tractography between the postcentral gyrus, the right hippocampus, and the anterior end of the temporal lobe, the temporal pole, revealed a significant increase in degree centrality of structural connectivity in DPD patients compared to NDPD subjects (Won et al., 2019). ROI analysis in the study of Li et al. revealed a significant increase in the index of axonal integrity in the right cingulate gyrus but no significant differences in AD and RD for both the right cingulate gyrus and the left hippocampus between diagnostic groups . This study also showed that microstructural impairments could be seen in the frontolimbic and hippocampus regions which are associated with mood reg- No differences in white matter integrity in specific ROIs or in the whole brain level.
No differences in white matter integrity in specific ROIs or in the whole brain level.
No differences in white matter integrity in specific ROIs or in the whole brain level.  (Gou et al., 2018;Lacey et al., 2019;Li et al., 2020) or for whole-brain (Gou et al., 2018;Lacey et al., 2019). Assessing the relative mobility of water molecules, DPD patients exhibited increased water mobility compared to NDPD subjects across different brain regions, such as the bilateral fornix (cres)/bilateral inferior front occipital fasciculus, stria terminalis, the body of corpus callosum, bilateral corticospinal tract, and arcuate fibers in the left amygdala, when compared to HC in two studies (Gou et al., 2018;Li et al., 2020) but no evidence of differences compared to NDPD subjects in the uncinate fasciculus, longitudinal fasciculus, forceps minor (Lacey et al., 2019) or for whole-brain (Gou et al., 2018;Huang et al., 2014;Lacey et al., 2019).

Conversely
The axonal and myelin integrity indices did not differ between DPD and HC subjects  or between DPD and NDPD subjects (Lacey et al., 2019), although the study by Li et al. showed higher AD values between DPD and HC subjects .
Decreased connectivity for DPD participants in the inferior longitudinal fasciculus and the corpus callosum was reported in two studies (Ansari et al., 2019;Ghazi Sherbaf et al., 2018; Table 3).

| Between-group contrasts: subcortex
Several studies have reported damage to subcortical structures of the brain, including the thalamus that is considered to be a major neural no evidence was found for MD differences comparing DPD and NDPD subjects (Huang et al., 2014).
PD disability assessments, including the H&Y and UPDRS, were conducted on most study subjects. However, PD disability scores were not reported to have any association with diffusivity/network metrics. GDS is a clinical score representing the degree of depression.
Patients with a GDS score of 6 or more are regarded as depressive subjects (Yesavage et al., 1982). Won et al. showed a strong correlation between the GDS score and selected imaging features (Won et al., 2019). One of the network studies, Gou et al. (2018) (Gou et al., 2018).
HDRS, also referred to as the HAMD, is a questionnaire that is used to assess depression severity, tapping both mood and somatic symptoms, and also is commonly used as a measure of recovery in treatment trials or naturalistically over time (Biegler, 2018).  Table 2).
One study did find that increasing depression symptom severity was positively correlated with focal FA reduction, affecting mainly the right substantia nigra and posterior putamen (Prange et al., 2019) (2018) also found similar overlapping regional findings as Huang et al. (2014) in the uncinate and longitudinal fasciculus, suggesting damage to the limbic system and its functions (Gou et al., 2018;Huang et al., 2014). The network studies demonstrated an increase in characteristic path length and reduction in global efficiency, which can be conceptualized as a breakdown of brain network unity (Gou et al., 2018;Hu et al., 2020). Consistent with this, an fMRI study by Quin et al. also showed an increase in the characteristic path length of the functional network in DPD subjects compared to HC (Qian et al., 2017). These alterations suggest lower efficiency and additional costs of data transfer between different functional areas in DPD patients. Impaired brain network data transmission could result in, or amplify, motor retardation, a prominent somatic symptom both of depression and PD.
The study by Lacey et al. (2019) revealed no significant WM microstructural differences in the thalamic region among DPD patients. In contrast, two other studies did find alterations in the thalamus and its radiations in DPD patients (Li et al., 2010;Prange et al., 2019) that is supported by studies using other neuroimaging modalities, such as fMRI, that indicate decreased connectivity in the left mediodorsal thalamus (Cardoso et al., 2009). As the thalamus processes and relays sensory information to distinct cortical regions, connections with the medial temporal lobes contribute to processing sensory input associated with learning and memory within the medial limbic circuit. This suggests an additional focus on thalamic microstructural and related network alterations in DPD patients, which may contribute to associated cognitive impairment, particularly affecting memory and attention domains.
Our study suggests that FA in WM clusters may be decreased in DPD, which indicates WM damage. However, DTI findings from several previous studies (Sanjari Moghaddam et al., 2020) do not support this relationship. In this context, we did not find that axonal and myelin integrity indices were different for DPD subjects, except for one study (Li et al., 2010). It is important to note that DTI determines microstructural WM cluster anomalies without correlation to WM volume changes, which overall do not appear to differ in PD patients compared to HC (Rektor et al., 2018). As a result, the specific relationship of myelin and axonal structure alterations related to depression in PD remains indefinite and requires further investigation.
The cerebellum is now regarded as a critical structure for cognitive and executive functions, in addition to motor coordination (Stoodley & Schmahmann, 2010). A recent fMRI resting-state connectivity study has implicated the cerebellum as having a role in the underlying pathophysiology of depression in PD (Wang et al., 2018).
As DTI studies included in this analysis mostly did not report cerebellar relationships, future DTI studies seeking to better understand microstructural abnormalities in DPD patients need to assess cerebellar and cerebro-cerebellar networks systematically.
The studies of Li et al. and Hu et al. included cognitive indices but enrolled only subjects having mean MMSE scores higher than 24, suggesting normal cognitive functioning in the DPD patients that substantially limits conclusions regarding any interactions between regional microstructural alterations, depression, and cognition. The overall lack of cognitive impairment observed in subjects included in this analysis also makes it difficult to comment upon the interaction of PD and depression on regional microstructure alterations Li et al., 2020). Thus, future studies will need to extend investigations of DPD patients to include more clinically distinct populations not selected based on an intact cognitive profile to understand better the inter-relationships between depression and cognitive impairment from microstructural alterations. Other potentially confounding factors, such as possible structural brain differences, disease stage, age/sex differences, and concurrent medication effects, will need to be more fully addressed. Overall, 229 DPD subjects were included in this review. Thus, available studies that specifically assessed depression in PD are mostly underpowered, and the totality of DPD subjects studied is very limited, necessitating larger samples to establish more clear relationships.
As mood disturbances in PD can be difficult to disentangle from PD symptom expression, it would be necessary for future investigations to systematically evaluate whether depression associated with PD is different from the neurobiological underpinnings of major depression not associated with PD. Critical assessment of imaging findings, including DTI, evaluating differences among DPD, NDPD, individuals with major depression without PD, and HC without depression, will help better understand the neurobiological substrates of depression and PD.
The current study systematically reviewed brain network and microstructure relationships between associated depression and PD. However, this analysis has several limitations. First, DTI findings are affected by intrinsic technical factors like low anatomical resolution, low SNR and distortions (e.g., vibrational artifacts caused by mechanical deficiencies of the scanner or subject-initiated motion during scanning), and variable capability to resolve crossing fiber tracts that restrict its diagnostic power (Solders et al., 2017). DTI findings are also subject to sequence acquisition parameters, scanner field strength, and maximum b values (Zhang & Burock, 2020

| CONCLUSION
DTI findings to date suggest altered WM microstructure and network disruptions of the cerebrum and cerebellum in PD patients with depression. Studies using various diffusion and network analytic approaches overall found alterations to essential brain structural networks, including impaired network integrity for specific cortical regions, such as the temporal and frontal cortices. Additionally, findings indicate that microstructural changes in specific limbic structures, such as the prefronto-temporal regions and connecting WM pathways, are altered in DPD compared to NDPD. There remain inconsistencies between studies reporting DTI measures, such as FA and MD, and depression severity in PD participants. Additional research evaluating underlying neurobiological relationships between major depression, DPD, and NDPD is required to disentangle further mechanisms that underlie depression, and related somatic symptoms, in PD.