A total of 33 patients (23 males, 10 females) with idiopathic PD participated in this study. All subjects provided informed written consent according to the Declaration of Helsinki prior to undergoing the MRI scan. This study was approved by the Institutional Ethics Review Board.

All PD subjects were assessed in the medication ON state, using part III of the Movement Disorders Society Unified Parkinson’s Disease Rating Scale (MDS UPDRS-III) for symptom severity and the Hoehn and Yahr (H&Y) scale for disease staging. Motor assessments were administered by experienced research personnel who were certified by the International Parkinson and Movement Disorder Society. A summary of the clinical and demographical characteristics is available in Table 1. Clinical characteristics of male and female PD patients were compared by calculating axial, bradykinesia, rigidity, and tremor subscores of UPDRS-III, using the method described by Li et al. (2018).

The subjects were scanned with a 3.0 Tesla MR scanner (Philips Achieva 3.0 Tesla; Philips Medical Systems, Netherlands) with an eight-channel head coil. We acquired a full brain 3DT1-weighted scan for structural references with an inversion recovery MPRAGE sequence TI = 808 ms, TR = 1800 ms, and an isotropic voxel size of 1 mm³. T₂ relaxation data were collected using a modified 3D GRASE sequence with 32 echoes with 10 ms echo spacing, TR = 1,000 ms, and the scanning time of 10.5 min. Twenty slices were acquired at 5 mm slice thickness and reconstructed to 40 slices at 2.5 mm. The in-plane voxel size was 1x1 mm. All subjects were scanned on the same day as their UPDRS assessment. The multiecho GRASE sequence was analyzed using in-house written MatLab (MathWorks, Natick, MA) code that uses a non-negative least squares (NNLS) fitting method to approximate the multiexponential decay curve with a number of exponential basis functions on a voxel by voxel level. The algorithm includes correction for stimulated echoes as well as a regularizer to make the fit more robust against noise in the time domain (Prasloski et al., 2012), resulting in one whole cerebrum MWF map per subject. We used the John’s Hopkins University WM tract atlas provided in the FSL package (Jenkinson et al., 2012) and registered them to each subject’s first echo of the T2 relaxation data before calculating the average MWF per region of interest (ROI). A total of 20 WM ROIs were used to cover the majority of the WM and delineate major WM tracts to get the most reasonable coverage for biological interpretation (Figure 1 and Table 2). All registrations were performed with the advanced normalization toolbox (ANTs) (Avants et al., 2008).

The overall workflow of the analyses is shown in Figure 2. First, we calculated the average MWF of each white matter ROI (Figure 2A). Secondly, to assess sex differences in the myelin profiles, we performed logistical regression with sex as the dependent variable y and average MWF in each ROI as the independent variables X (Figure 2B). After assessing the coefficient vector β in the logistic regression model, we selected the subset of ROIs with significantly non-zero coefficients and then calculated the sum and laterality of MWF in this subset. A two-sample t-test was used to compare the differences between males and females in terms of these two MWF metrics.

To determine the parts of the UPDRS that corresponded to the sex differences in MWF, we performed multiple linear regression between UPDRS and MWF (Figure 2C). We first applied principal component analysis (PCA) on the different UPDRS motor subscores, with the optimal number of principal components determined by using cross-validation (Engemann and Gramfort, 2015) (which turned out to be 3). We then took the principal components of UDPRS motor scores as the independent variables A, and the linear combination of MWF values determined from the logistic regression as the dependent variable (i.e., Xβ from Figure 2B). After assessing the coefficient vector p in the multiple regression model, we examined whether or not there were principal components of UPDRS significantly associated with sex differences in MWF. Finally, we interrogated the weights (i.e., w) of individual UPDRS subscores (i.e., S) on the significant component (i.e., A(i), ith column of the matrix A) to evaluate the sex-related characteristics of clinical symptoms (Figure 2D). In addition, a two-sample t-test was used to examine the sex differences in clinical symptoms. Given the relatively small sample size, we further examined the robustness of our conclusions by using leave-one-out cross-validation. A false discovery rate (FDR) correction was applied for multiple comparisons. Statistical significance was set at p < 0.05.

Patients were at relatively mild stages of disease (H&Y stage 2) of moderate duration. Mean age, H&Y disease severity, and disease duration were not significantly different between male and female patients (Table 1). Male and female patients also did not significantly differ in total MDS UPDRS-III score, nor motor subscores of axial symptoms, bradykinesia, or rigidity. However, female patients had significantly higher tremor scores than male patients (p = 0.006).

When assessing sex differences in MWF using logistic regression, we found MWF of the following ROIs significant in predicting sex: thalamic radiation, cingulum, cingulum hippocampus, inferior fronto-occipital fasciculi (IFOF), inferior longitudinal fasciculi (ILF), and uncinate (p < 0.05, Figure 3A). The total MWF of the aforementioned significant ROIs was significantly higher in males compared to females (p = 0.02, Figure 3B). When comparing the MWF of each ROI individually, males appeared to have greater MWF than females across all ROIs, and this was significant (uncorrected) in the following ROIs (Figure 3C): R thalamic radiation (p = 0.03), L cingulum hippocampus (p = 0.02), R cingulum hippocampus (p = 0.02), splenium (p = 0.04), R IFOF (p = 0.02), and R uncinate (p = 0.04). Overall, the total MWF was larger in right-sided ROIs in both males (p = 0.006) and females (p = 0.07, not significant, but a trend). We further examined this laterality of MWF by taking the absolute difference between right and left-sided MWF among the significant ROIs (Figure 3A) and found that males had greater laterality of MWF (p = 0.03, Figure 3D).

When assessing the correspondence between UPDRS and the sex differences in MWF using multiple linear regression, the second principal component of the UPDRS score was significant (p = 0.02) in predicting the sex differences in MWF. The weights of this principal component are shown in Figure 4A. Note that the weights related to tremor and bradykinesia favored females, and weights related to rigidity and axial symptoms favored males. This was partially reflected in the raw data, where females had significantly greater total tremor score than males (p = 0.006, Table 1 and Figure 4B). However, there were no statistically significant differences in bradykinesia, rigidity, and axial symptom subscores between males and females. Weights related to motor performance (i.e., finger tapping, hand movements, pronation-supination hands, toe tapping and leg agility) tended to be different between left and right. Overall, the total motor performance score was larger (indicating greater symptom severity) on the left side in both males (p = 0.04) and females (p = 0.17, not significant, but a trend). We further examined this laterality of motor performance scores, by taking the absolute difference between all left and right-sided motor performance scores, and found that males had greater laterality of motor performance scores (p = 0.04, Figure 4C).

Using leave-one-out cross-validation, we found the MWF ROIs that were most frequently selected as significant in predicting sex were unchanged. While we found seven significant MWF ROIs using the whole dataset (Figure 3A), the top seven MWF ROIs in the leave-one-out cross-validation were the same except for the shift from right ILF (the eighth most significant ROI in the cross-validation analysis) to left ILF (Figure 5A). Sex differences in the individual/total MWF and the laterality of MWF remained significant in the cross-validation analysis (Figures 5B,C). Similar to the whole dataset analysis, we found the significant principal component of UPDRS scores (p = 0.03 ± 0.01) in predicting sex differences in MWF in the cross-validation analysis. The weight distribution of the significant principal component in the cross-validation analysis (Figure 6A) was very similar to that in the whole dataset analysis (Figure 4A). Sex differences in the total tremor score and the laterality of motor performance, remained significant in the cross-validation analysis (Figure 6B).

We found the principal component of UPDRS scores predicting sex differences in MWF had tremor and bradykinesia weights favoring females, and rigidity and axial symptom weights favoring males (Figure 4A). This demonstrates that the sex-specific differences in motor performance were associated with sex-specific differences in myelination. When looking at the raw clinical data, this observation was supported by noting female patients had significantly higher tremor scores compared to male patients (p = 0.006, Table 1).

The tremor-dominant phenotype of PD is classically regarded as a more benign subtype of PD, whereas rigidity and postural instability-dominant phenotypes have been regarded as having poorer prognosis. Clinical presentations are key features for different subtypes of diseases (Thenganatt and Jankovic, 2014; Wang et al., 2021). Female sex has been associated with the tremor-dominant phenotype and slower disease progression in prior studies (Haaxma et al., 2007; Solla et al., 2012). Female patients are also noted to have older age of disease onset, which has been positively associated with factors increasing cumulative estrogen exposure such as higher parity, later age of menopause, and longer fertile life span (Haaxma et al., 2007), suggesting a role for sex hormones in disease onset, clinical characteristics, and progression. While we did not observe a significant difference in the axial, rigidity, and bradykinesia subscores between male and female patients in the raw data, the weights related to rigidity and axial symptoms were biased toward males in the principal component that was significant in predicting sex differences in MWF.

The MWF of the thalamic radiation, cingulum, cingulum hippocampus, IFOF, ILF, and uncinate were significant in predicting sex (Figure 3A). Males had higher overall MWF values (Figure 3B), and specifically higher MWF in the R thalamic radiation (p = 0.03), L cingulum hippocampus (p = 0.02), R cingulum hippocampus (p = 0.02), splenium (p = 0.04), R IFOF (p = 0.02), and R uncinate (p = 0.04), compared to females. Sex differences in myelin in the PD population have yet to be studied. However, there is emerging evidence that sex hormones play a significant role in the sexual differentiation of WM microstructure, which may be driven by differences in myelin (van Hemmen et al., 2016). van Hemmen et al. compared WM microstructure in karyotype XY men, karyotype XX women, and karyotype XY women with complete androgen insensitivity syndrome (CAIS), and found that karyotype XY women with CAIS showed female-typical FA throughout many WM ROIs, which was predominantly due to female-typical RD. This suggests a significant effect of sex hormones on human neural sexual differentiation, and specifically on WM and myelin microstructure. Male participants of this study had lower RD (suggestive of greater myelination) in the left internal and external capsule, corona radiata, cerebral peduncle, uncinate, superior fronto-occipital fasciculus, pallidum, and amygdala, compared to female participants. Other studies have also found diffusion changes suggestive of sex differences in myelin. Similar to the results of our study, Menzler et al. found higher values of FA and lower RD in the thalamus, corpus callosum, and cingulum of male subjects compared to females. Jung et al. found that males had lower RD in the left anterior thalamic radiations and left uncinate fasciculus when compared to female subjects. However, most existing studies of sex differences in WM, including the above mentioned, are in the young, healthy adult population. In order to determine sex-specific myelin changes in PD, more studies of sex differences in myelin in the healthy elderly population are needed. Inano et al. examined the effects of age and sex on FA, axial diffusivity (AD), and RD in WM, in a large study of 857 healthy subjects with a wide age range (25–85, mean age 56, 32). Global FA was negatively correlated with age, whereas global RD was strongly positively correlated with age, suggesting that myelin degeneration was implicated in age-related changes in WM integrity. Interestingly, this study did not find a significant interaction between age and sex, suggesting that there are no sex differences in the aging process of WM. The WM ROIs identified in our study have some overlap with previously noted sex differences in myelin among the healthy young adult population. This suggests some of the observed differences may be in part due to normal sexual dimorphism in WM, and some due to PD-specific differences in myelin. Unlike the healthy elderly participants studied by Inano et al. (2011), we found sex differences in myelin profiles among our study population with PD, which may reflect PD-specific sex differences. However, due to the overall paucity of studies investigating sex differences in myelin, it is difficult to differentiate which of the observed ROIs are representative of normal sexual dimorphism of myelin, versus PD-specific sexual dimorphism of myelin. Given that sex differences in MWF observed in our study were associated with previously established sex differences in clinical symptom profile (tremor-dominant phenotype in females, rigidity and postural instability phenotype in males), we suggest that there is a role of myelin differences in clinically-observed sex differences in PD.

Clinically, PD is well-recognized as an asymmetric disease, with a characteristic unilateral onset of motor symptoms, associated with more severe nigrostriatal degeneration on the contralateral side. Claassen et al. described a “left hemisphere susceptibility” in PD, which persists after accounting for side of disease onset as well as handedness (Claassen et al., 2016). The results of the present study may add to this previous notion of left-sided susceptibility in PD. We found that MWF was larger in right-sided ROIs (indicating more myelination on the right) in both males (p = 0.006) and females (not significant, but a trend). This laterality of myelination was greater in males when taking the absolute difference between right and left-sided ROIs (p = 0.03, Figure 3D). Motor performance scores were also larger (indicating greater symptom severity) on the left side in both males (p = 0.04) and females (not significant, but a trend), and again, males had greater left-sided laterality of motor performance scores (p = 0.04, Figure 4C).

The mechanisms of sex differences in PD have not yet been fully elucidated, but a key role of sex hormones is likely. Estrogen has been associated with neuroprotective effects on the nigrostriatal dopaminergic system, with various proposed mechanisms, including inhibition of reactive glia, inhibition of the inflammatory cytokine cascade, antioxidative effects, and anti-apoptotic effects. Early menopause, whether natural or surgically induced, is associated with increased risk of parkinsonism, and the risk is increased with younger age at oophorectomy (Rocca et al., 2008). Neuroprotective effects of estrogen have also been observed in mouse models with dopaminergic striatal neurotoxicity induced by 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) (D’Astous et al., 2004). However, there are also studies that have found the opposite association between estrogen and PD, or even a lack of association (Smith and Dahodwala, 2014). Given there are complex interplays in biological systems, as well as differential effects of hormones depending on how (ex. cyclical vs. continuous), when (ex. pre vs. post-menopausal), and in what form they are administered (stereoisomers of estradiol, estriol), definitive conclusions regarding estrogen and clinically observed sex differences PD are difficult to establish.

Sex-specific differences in clinical characteristics, with a predilection toward a more benign clinical phenotype in female PD patients, may support the neuroprotective role of estrogen. The effects of estrogen on myelination and cellular pathology in PD is unclear. However, given robust clinical differences in PD between male and female patients, these differences may in part be explained by the effects of sex hormones. In our study, though females had the more benign tremor-dominant phenotype, male PD patients had overall higher MWF values, and specifically in the R thalamic radiation, bilateral cingulum hippocampus, splenium, R IFOF, and R uncinate, indicating more myelination in these WM ROIs.

In this study, we used MWF instead of DTI metrics to assess myelin integrity in PD. We have previously shown that using DTI measures to define ROIs, but then interrogating the MWF values within these ROIs is more sensitive and specific than DTI for predicting biological sex in healthy controls (Baumeister et al., 2020). Finally, we have previously shown that white matter profiles assessed via MWF were able to predict clinical subtypes of PD, but FA was not (Baumeister et al., 2019). Thus, based on our prior work demonstrating superiority of MWF in predicting sex and clinical aspects of PD, we felt justified in using MWF for looking at the relationships between sex differences and myelin and PD.

There are a number of limitations to our study. Our sample size is relatively small. Myelin water imaging is a relatively new sequence that does not have the same availability as other more popular, but more qualitative, sequences such as FLAIR. However, even with these relatively small numbers we observed statistically significant results. The number of subjects was not balanced between the sexes, but we note that our sample has roughly the same 2: 1 male to female ratio as the incidence of the disease (Miller and Cronin-Golomb, 2010). While we found relatively broad overall changes in myelin between the sexes; a larger sample size may determine if changes in particular ROIs are associated with specific clinical features.

We have implicitly implied a causal direction between changes in myelin and clinical presentation. While this seems likely, within the limitations of cross-sectional data it is difficult to establish that the observed sex differences in MWF of WM ROIs caused differences in clinical characteristics of male and female PD patients. We cannot discount that other factors might jointly affect myelin and clinical features. Moreover, we cannot discount that physical activity induces plasticity in myelin (Fields, 2015), and thus causal directions between clinical features and myelin may become less clear.

The cerebellum and cerebello-thalamo-cortical pathway are frequently postulated as contributors to tremor genesis in PD and essential tremor patients, but the cerebellum was excluded in the WM ROI atlas. While we took a whole-brain network-perspective and identified widespread changes in WM microstructure, we cannot exclude specific changes in the cerebellum because the myelin water imaging sequence employed does not include the cerebellum. This might be corrected as the MWF technique matures and becomes more widespread.

Sex differences are apparent in the clinical characteristics of PD. Recent evidence suggests a potential role for myelin changes in the cellular pathology of PD, as well as in driving sex differences in WM microstructure in healthy adults. This necessitates sex-specific investigation of microstructural and cellular pathology in PD. In this study, we found sex differences in myelin associated with sex differences in UPDRS motor symptom profile. While preliminary, our results suggest that further investigation of the role of biological sex in myelin pathology and clinical presentation in PD is warranted.