Parkinson's disease (PD) is an idiopathic disease of the central nervous system characterized by both motor and non-motor symptoms. It is the second most common neurodegenerative disease worldwide with rising incidence and prevalence, partly due to changing population demographics (1), with world-wide PD cases expected to increase from an estimated 4.1 million in 2005 to ~8.7 million by 2030 (2). Biomarkers play a critical role in the management of patients with PD, including for diagnosis and monitoring disease progression. Neuroimaging techniques may provide such biomarkers, but these must be investigated in large-scale, multicenter, and longitudinal samples. For example, the Parkinson's Progression Markers Initiative (PPMI) study (3) (https://www.ppmi-info.org/) is a comprehensive observational, international, multicenter study designed to identify PD progression biomarkers, both to improve understanding of disease etiology and course and to provide crucial tools to evaluate the efficacy of PD-modifying therapeutics.

In recent years, PD has been studied with a range of MRI techniques, such as structural imaging, functional MRI (fMRI), and diffusion weighted MRI (dMRI). These have been shown to be useful not only in PD but also in other forms of parkinsonism (4). For instance, dMRI has been used extensively to study structural changes in brain white matter (WM), such as axonal caliber, density, myelination, and orientation, along with WM-dependent structural connectivity in early PD (5). One of the most used dMRI analysis methods that can help to understand the pathophysiology and evolution of PD is diffusion tensor imaging (DTI) (6–9). A recent meta-analysis of DTI findings showed consistent differences indicative of neurodegeneration between PD and healthy controls (HC) in subcortical regions (including the substantia nigra), cingulate and temporal cortex, and the corpus callosum (10). Paradoxical changes in the corticospinal tract were also observed and could be indicative of either compensatory changes or selective neurodegeneration of proximate WM fibers. The latter highlights a known limitation related to standard DTI, notably that DTI is not able to resolve crossing fibers within brain voxels (11). To overcome this limitation, more advanced dMRI methods must be used (12). For example, the distribution of fiber orientations (i.e., fiber orientation distribution, FOD) within a voxel can be directly estimated from dMRI data without prior assumptions regarding the number of fiber populations present (13).

Combining dMRI and tractography enables the assessment of brain structural connectivity, which represents the WM tracts that physically interconnect brain regions in vivo (14). Additionally, tractography, in combination with graph theory, can been used to reconstruct and analyze the structural whole-brain connectivity in PD (10, 15), as well as other neurodegenerative disorders (16, 17). These studies in PD have revealed the existence of novel PD-specific anatomic networks, characterized by alterations in cortical–subcortical loops, reduced modular organization, and compensatory changes (15, 18).

In the current study, structural connectivity and WM networks were analyzed using the FOD from dMRI and graph theory in a cohort of patients with PD and a cohort of HCs recruited to the PPMI study (3). To examine whether structural connectivity and WM network analysis could yield information about PD and its progression, we investigated the differences in structural connectivity and WM networks between these groups in a cross-sectional analysis. We also investigated longitudinal changes in the PD cohort over 36 months.

In the current study, we analyzed structural connectivity and WM networks in participants with PD using both cross-sectional comparison with HCs and longitudinal evaluation across 36 months. We found significant differences in these metrics between PD and HCs, suggesting these metrics may reflect the impact of PD pathology on the brain's structure and networks. We also found longitudinal changes over a 36-month period, suggesting that structural and network metrics may be sensitive to disease progression. We discuss these results in more detail below.

Compared with the HC group, participants with PD showed lower structural connectivity in several brain areas, which were confirmed by a large effect-size. The PD group showed lower connectivity in several WM areas, such as, the corpus callosum (mainly in the genu), fornix, sagittal stratum, external capsule, cingulum, and uncinate fasciculus. Decreased structural connectivity in the corpus callosum might indicate degeneration of interhemispheric axonal connections between frontal areas. The corpus callosum is a critical structure for interhemispheric information transfer and thereby plays an important role in cognitive function (35). Indeed, other dMRI studies have shown that deterioration of the genu of the corpus callosum is linked to PD dementia (36), as well as executive and attention dysfunctions (37). Additionally, comparison of DTI metrics [such as FA and mean diffusivity (MD)] in the corpus callosum in PD has been shown to differentiate PD with and without cognitive impairment (38). In addition, structural alteration of the corpus callosum may be involved in predominant gait disorders (39) and impulse control disorders (40).

As already mentioned, other WM areas where connectivity was lower in PD participants were the fornix, cingulum, external capsule, and uncinate fasciculus. Altered DTI-related metrics have been found in these WM areas in PD using tract-based spatial statistics (TBSS) (41–45). Others have suggested that cognitive symptoms, such as episodic memory and visuospatial processing, as well as axial motor control, in individuals with PD may involve the cingulum (46). These findings might suggest that the lower connectivity inside the cingulum reported here might be responsible for the presence of such symptoms in this PD cohort (47). On the other hand, the fornix has important axonal connections with the subiculum of the hippocampus (48) and altered connectivity in this area is known to be associated with hippocampal damage (49, 50) as well as memory impairment in patients with PD (37). Together, these findings indicate that further analyses of structural connectivity metrics and their relationships with different PD symptoms are warranted.

In other WM areas, such as the right posterior limb of internal capsule and the left anterior corona radiata, the PD group exhibited higher connectivity than HC. Indeed, compared with the HCs, PD patients showed higher structural connectivity in only one path: between the left caudate and left medial frontal gyrus (orbital part). These findings are in line with the results reported by Mishra et al., which identified the presence of a distinctive PD-specific structural connectome, along with the unanticipated novel finding of increased structural connectivity between known PD-relevant brain regions including the caudate (15). These results support the hypothesis that structural network changes may underlie altered functional network capacity observed in PD. Additionally, the predominant structural connectivity differences occurred in the left hemisphere, suggesting lateral asymmetry. Several other studies have similarly shown that cortical atrophy (51) and nigral microstructural changes (52) are left lateralized in PD cohorts, with relative preservation of the right hemisphere until later in disease.

Altered connectivity between baseline and after 36 months was found in different paths at medium effect-size. It should be noted that these changes likely reflect a combined effect of both PD progression and aging, as changes in structural connectivity are known to occur with aging and may be attributed to brain reorganization that prioritizes selective connectivity to maintain function (53, 54). Interestingly, compared with PD at baseline, higher structural connectivity was found in PD subjects after 36 months inside 8 different paths (see Table 3 and Figure 4). WM areas with improved connectivity include the corpus callosum, fornix, and uncinate fasciculus. One explanation of these findings might be that increasing connectivity is indicative of compensatory changes, which have already been reported as compensation in later stages of the disease and the whole-brain/network scale that are probably due to the heterogeneous nature of PD (55, 56). However, on the other hand, we also found other 5 paths where PD patients at 36 months of follow-up had lower connectivity than at baseline.

Studies that have examined altered structural connectivity patterns in PD via graph theory analyses (57–59) have reported conflicting findings. This may be related to the differences in preprocessing and application of tractography methods (e.g., choice of tracking algorithm (probabilistic, deterministic, or other) and/or by the choice of edge weights) or could also be attributed to heterogeneity in PD populations. Here, we report modest between-group and within-group (PD) differences in network density (p < 0.05; uncorrected). The network density metric is equal to the numeric ratio of the actual connections to the possible connections. It indicates the degree of closeness of the relationships between the nodes in the network. In this study, the mean network density in the PD group was found to be lower than for controls. A similar trend was found for the longitudinal changes seen in PD.

Significant between-group and within-group differences were observed in betweenness centrality, nodal degree, and the strength in several nodes. Differences in the betweenness centrality metric may be associated with altered functional connectivity in PD (60) and were found in nodes mainly located in the frontal and temporal lobes. Differences in nodal degree were found mainly in caudate, frontal, occipital, and temporal lobes for HC vs. PD, as well as in the insula and putamen in the longitudinal comparison. Differences in strength were found mainly in nodes located in the caudate (baseline <Month 36), putamen, and amygdala, which are regions where similar changes have previously been observed in PD patients (61). Taken together, our results demonstrate that alterations in brain networks may be associated with PD and its progression.

Lastly, we also found significant correlations (FWE <0.05) at baseline between structural connectivity and MoCA score. These corresponded to two long-range connections, with an interhemispheric connection positively correlated with MoCA score and an intra-hemispheric connection negatively correlated with the MoCA score.

There are several strengths and limitations to this study. A major strength here is the use of a robust pre- and post-processing pipeline of the imaging data. One limitation is related to the single-shell data acquisition for dMRI, which has several drawbacks, including known inaccuracies related to partial volume effects (62) and difficulties to estimate the correct fiber orientation in voxels with complex fiber structure (e.g., crossing fibers) (63). In this study, we overcome these limitations by estimating the response function(s) for spherical deconvolution and the FOD (64) in order to more accurately estimate tractography. Additionally, the tcksift2 (30) algorithm was applied to our tractographies to improve the accuracy of these reconstructions further. A final limitation is the lack of data available for longitudinal analysis of a HC group (not available in PPMI), and thus changes observed in the PD group cannot be attributed to the effects of PD pathology alone but are rather the combined result of aging and PD pathology. Using the data available, both group differences at baseline and longitudinal changes in the PD cohort were considered using advanced dMRI metrics, which is a strength of this study.

In conclusion, we report altered structural and network metrics in several brain regions, such as corpus callosum, fornix, and cingulum, which showed lower connectivity in PD. We also report higher connectivity in the PD group, after 36 months, potentially due to compensatory processes as a result of the combined effects of PD pathology and aging. This data indicate that structural and network metrics could yield new diagnostic and progression markers for PD.

The data analyzed in this study was obtained from the Parkinson's Progression Markers Initiative (PPMI) database, the following licenses/restrictions apply: Investigators seeking access to PPMI data must sign the Data Use Agreement, submit an Online Application and comply with the study Publications Policy. Requests to access these datasets should be directed to PPMI, https://ida.loni.usc.edu/collaboration/access/appLicense.jsp.

Ethical review and approval was not required for the study on human participants in accordance with the local legislation and institutional requirements. The patients/participants provided their written informed consent to participate in this study.

Conception, design, and original draft preparation: MB and AMS. Downloading data from PPMI database: MB. Data analysis and interpretation: MB, EK, and AMS. Manuscript writing, review, and editing: MB, EK, NJR, AM, MS, and AMS. All authors approved the final manuscript.