Diffusion along perivascular spaces reveals evidence supportive of glymphatic function impairment in Parkinson disease

Abstract

Background:

Reduced diffusion along perivascular spaces in adults with Alzheimer’s-disease-related-dementias has been reported and attributed to reduced glymphatic function.

Objectives:

To apply quantitative measures of diffusion along, and orthogonal to, perivascular spaces in a cohort of older adults with and without clinical symptoms of alpha-synuclein related neurodegeneration.

Methods:

181 adults with Parkinson disease (PD) or essential tremor (ET) additionally sub-classified by the presence of cognitive impairment underwent 3 Tesla MRI. Diffusion-tensor-imaging (spatial resolution=2×2×2 mm; b-value=1000 s/mm²; directions=33) measures of diffusion (mm²/s) parallel and orthogonal to perivascular spaces at the level of the medullary veins, and the ratio of these measures (ALPS-index), were calculated. Regions were identified by a board-certified neuroradiologist from T₁-weighted and T₂-weighted MRI. Evaluations of motor impairment and mild cognitive impairment (MCI) were interpreted by a board-certified neurologist and neuropsychologist, respectively. Multiple regression with false discovery rate correction was applied to understand how diffusion metrics related to (i) disease category (PD vs. ET), (ii) cognition (MCI status), and (iii) white matter disease severity from the Fazekas score.

Results:

The ALPS-index was reduced in PD compared to ET participants (p=0.037). No association between the ALPS-index and MCI status, but an inverse association between the ALPS-index and Fazekas score (p=0.002), was observed. The ALPS-index was inversely associated with age (p=0.007).

Conclusion:

Diffusion aberrations near perivascular spaces are evident in patients with alpha-synuclein related neurodegenerative disorders, and are related to age and white matter disease severity.

Introduction

Recent evidence, primarily from vertebrate animals, of fluid exchange between cerebrospinal fluid (CSF) and the interstitial space has led to speculation regarding the presence of a glymphatic clearance pathway. Animal models shown that this system is active during sleep and during administration of various types of anesthesia[1,2]. These studies suggest that poor fluid exchange between the CSF and interstitial space may be implicated in diverse types of central nervous system dysfunction.

Diffusion Tensor Imaging (DTI)-ALong the Perivascular Space (DTI-ALPS)[3] has been proposed as a metric with possible sensitivity to glymphatic function (Figure 1). This method relies on the fact that within cerebral white matter regions near the lateral ventricles, the perivascular space lies largely orthogonal to the projection and association white matter tracts. On susceptibility-weighted imaging, the medullary veins, which parallel perivascular fluid motion, can be seen coursing in the axial plane at the level of the superior aspect of the lateral ventricles. At moderate-to-high b-values typically used in DTI (i.e., b=1000 s/mm²), any contribution to the diffusion-weighted signal from flowing venous blood is suppressed. As such, excess diffusion measured in the direction of the medullary veins (transverse direction) and orthogonal to the primary fiber tract orientation has been hypothesized to contain contributions from perivascular, or glymphatic fluid transport. In the DTI-ALPS model, this is quantified as the amount of diffusion in the perivascular direction, in excess of the diffusion in the other nondominant tract. In the projection tract, this other nondominant direction is in the anterior-to-posterior direction and in the association tract, it is in the craniocaudal direction.

Figure 1.

Diffusion tensor imaging along perivascular spaces (DTI-ALPS). Figure 1A shows a minimum intensity projection of a susceptibility-weighted-image (SWI), demonstrating that the lateral projections of the medullary veins course in the right-left direction (x) at the level of the lateral ventricles. A map of fiber tracts at this location (Figure 1B) shows that where the medullary veins course in the right-left (x) direction, the principal fiber tract runs orthogonally, in the craniocaudal direction (z) in the projection region and in the anterior-posterior direction (y) in the association region. In Figure 1C-D subsequent magnifications of the projection and association regions along the medullary veins with the diffusion tensors represented as ellipsoids. Figure 1E shows a schematic of the principal fiber tract orientation in these two projection and association regions demonstrating that the perivascular space runs parallel to the medullary vein and therefore orthogonal to the principal fiber tract orientations. In Figure 1F, the orientations of the diffusion components that course orthogonal to the principal fiber tracts within these voxels can be represented as a diffusion component orthogonal to both the perivascular space and principal fiber tract (yellow) and parallel to the perivascular fluid motion (gray). The ALPS-index score is a measure of the ratio of the diffusion parallel to the perivascular space and the diffusion orthogonal to the perivascular space (yellow), without incorporating the diffusion information from the primary fiber tracts. The x-component of the diffusion direction (D_xx; arrow) in the projection region may have perivascular fluid motion relevance, given the expected direction of perivascular fluid motion is along medullary veins while the projection and association fibers run orthogonal to the medullary veins at this location.

Many neurodegenerative conditions have been the targets of emerging studies of glymphatic dysfunction, due to (i) their largely unknown etiology and (ii) their association with proteinopathy and likely clearance dysfunction[4], as well as (iii) the finding that efflux of radioactive amyloid beta parallels fluid motion to and from the interstitial space to the CSF[2]. Alzheimer disease (AD) is a proteinopathy correlated with substantially increased cerebral β-amyloid plaque deposition. Decreased β-amyloid protein egress from the brain is closely tied to impaired glymphatic function[2]. The coupling of glymphatic function and clearance of the protein closely associated with AD has prompted speculation that the pathophysiology of other proteinopathies may relate to compromised glymphatic circulation. In support of this, the DTI-ALPS method has been shown to provide reduced diffusion measures in 31 adult participants with reduced cognitive performance as quantified from the established Mini-Mental State Exam (MMSE)[3].

Parkinson disease (PD) is another proteinopathy characterized by the abnormal accumulation of α-synuclein (Lewy Bodies). Alpha-synuclein pathology may begin peripherally, and travel centrally to the lower brainstem, midbrain and neocortex[5]. We assess whether perivascular fluid motion, as measured by the DTI-ALPS method, is different in a cohort of participants with PD compared to a population of essential tremor (ET) participants (motor disorder not associated with a proteinopathy). We applied the DTI-ALPS method in 181 adults with movement disorders in a similar anaesthetized state required per clinical indication for deep brain stimulation electrode placement. We evaluated diffusion along, and orthogonal to, perivascular spaces to test the hypothesis that diffusion measures of perivascular fluid motion were reduced in participants with (PD), relative to without (ET) proteinopathy.

Materials and Methods

Informed consent and participant recruitment

All participants provided informed, written consent in accordance with the local institutional review board (IRB) adhering to the ethical standards stipulated by the Declaration of Helsinki and its amendments. All participants presented to the neurology services at Vanderbilt University Medical Center between 2011 and 2017 with a clinical diagnosis of PD or ET confirmed by a neurologist according to established criteria[6,7]. All participants were sedated per clinical indication and received clinically-indicated scans for deep brain stimulation (DBS) bone marker placement. Participants underwent a motor disability assessment, neuropsychological cognitive battery, and neuroimaging using a non-contrast 3-Tesla head MRI.

Movement and assessment

Movement was assessed in an off-medication state. See supplemental description for details.

Cognitive assessment

Neuropsychological examination was performed by a licensed neuropsychologist to characterize cognitive functioning. See supplemental description for details.

Magnetic resonance imaging

See supplemental description for technical details regarding MRI sequence acquisition.

The DTI-ALPS approach was used for estimating the diffusion along perivascular spaces. Figure 1 outlines the contrast mechanism for this sequence. Diffusivity values in these regions can be used to calculate the ALPS-index index. Here, DTI was performed using a 2D spin echo single-shot echo-planar-imaging approach with 33 directions and TR/TE=10000/60 ms and isotropic spatial resolution=2.0 mm. The b-value used was 1000 s/mm², as was used in the initial DTI-ALPS study[3].

Image analysis

T₂-weighted images were reviewed by a board-certified radiologist with a certificate of added qualification in neuroradiology with 11 years of experience (C.M.) Grading of deep white matter (DWM) T₂–weighted hyperintensity was assessed on a 0–3 point scale according to Fazekas et al.[8]; DWM T₂-weighted hyperintensity was graded as 0 = absence, 1 = punctate foci, 2 = beginning confluence of foci, 3 = large confluent areas.

Diffusion tensors were reconstructed from diffusion-weighted images using the FMRIB Software Library (FSL) toolbox[9]. First, diffusion-weighted images were motion-corrected by linear registration to the middle volume using MCFLIRT and eddy current-corrected using the eddy function, and brain extracted with the BET utility with fractional intensity threshold set to 0.3. For all voxels within the brain, tensors, diffusion eigenvalues, and diffusion eigenvectors were calculated from the original diffusion images using the dtifit tool. Next, a board-certified neuroradiologist (C.M.) identified bilateral regions in which the lateral projections of the medullary veins traced orthogonal to the primary diffusion directions as described in the literature[3] and demonstrated in Figure 1. Separately within regions of interest (32 mm³) within the projection and association tracts, the separate diffusion parameters were calculated and the DTI-ALPS_MEAN score (ALPS-index) was calculated as the mean DTI-ALPS score across both hemispheres according to,

$$DTI-ALP{S}_{MEAN}=\frac{D_{\left|\right|}}{D_{\perp }}=\frac{mean\left(D_x^{proj},D_x^{assoc}\right)}{mean\left(D_y^{proj},D_z^{assoc}\right)}$$ [1]

where D_|| denotes the diffusion parallel to the perivascular space of the medullary veins through both projection and association regions (mm²/s), and D_⊥ denotes diffusion orthogonal to the perivascular space and primary fiber tract duration in the projection and association regions (mm²/s) (Figure 1). The primary metric from the DTI analysis that was preserved for hypothesis testing was the mean DTI-ALPS score (unitless) or ALPS-index as defined in Eq. 1.

Statistical analysis and hypothesis testing

All statistical analyses were performed using R version 3.5.3 (R Foundation for Statistical Computing, Vienna). To account for multiple comparisons, the results from all regression analyses were considered significant at the level of false discovery rate (FDR) of 0.05. The FDR corrections were applied separately to the results from each hypothesis.

First, group differences in demographic and clinical parameters present in both ET and PD participants were evaluated for continuous parameters using a Wilcoxon rank-sum test or for categorical parameters a Fisher exact test with criteria for significance being two-sided p<0.05. We first evaluated mean differences in the ALPS-index between males and females using a Wilcoxon rank-sum test and the dependence of the ALPS-index with age using a Spearman rank analysis.

The primary hypothesis of this study was that the ALPS-index, which has been proposed as a measure of glymphatic function, is lower in a cohort of PD participants compared to a similar population of participants without a proteinopathy-related neurodegenerative process (ET). To ensure that differences in ALPS-index scores between PD and ET were not a result of potential confounding factors, we first matched the two cohorts by age, sex, and DWM Fazekas score. The matching was performed using the nearest neighbor method based on the propensity score of age, sex and DWM Fazekas score, where 72 PD patients were selected to match the 36 ET patients in the cohort at a 2:1 ratio. We over-matched ET patients to add power to our study by including more subjects, and the ratio of 2:1 was chosen since there is little power gain for ratios beyond this[10]. Subsequently, we applied a general linear model (GLM) regression analysis using the mean ALPS-index score as the dependent variable and the diagnosis status (PD or ET) as the independent variable.

Supplementary hypotheses were that (i) a lower ALPS-index is associated with the presence of cognitive impairment, and (ii) the ALPS-index is reduced in all participants with elevated white matter pathology, as indicated by higher DWM Fazekas score. To evaluate the supplementary hypothesis that a lower ALPS-index score is associated with the presence of cognitive impairment, the above GLM analysis was repeated using the ALPS-index score as the dependent variable, the MCI status (binary classification of MCI or non-MCI) as the independent variable, and age, sex, diagnosis status (PD or ET) and DWM Fazekas score as covariates. To test the hypothesis that the ALPS-index is reduced in participants with more white matter pathology, we performed a GLM analysis using the ALPS-index score as the dependent variable, DWM Fazekas score as the independent variable, and age, sex and diagnosis status (PD or ET) as covariates.

Data Availability

The datasets generated during and or analyzed during the current study are available from the corresponding author on reasonable request.

Results

Demographics

A total of 181 participants (144 PD: male=97, female=47; 37 ET: male=21, female=16) completed the motor, cognitive, and neuroimaging assessments (Table 1). Based on the neuropsychological criteria for cognitive impairment, 66 participants met MCI criteria (58 PD (40%) and 8 ET (21%)). No participants met clinical criteria of dementia, as this was exclusionary for participation in the MRI assessment. Age at imaging (PD: 62.2 ± 9.0 years; ET: 67.8 ± 6.2 years) and disease duration (PD: 9.2 ± 4.5 years; ET: 15.3 ± 12.0 years) were significantly different between groups, as expected, as these conditions affect individuals at different ages. Given these differences, we applied propensity score matching based on age, sex and DWM Fazekas score and we used the matched pairs of PD and ET for comparison between groups.

Table 1.

Demographics and clinical data.

	Total (n=181)	PD (n=144)	ET (n=37)	p-value
Age at imaging (years)	63.4 (8.8)	62.2 (9.0)	67.8 (6.2)	<0.001
Age at diagnosis (years)	52.9 (10.5)	53.0 (9.8)	52.5 (13.0)	0.645
Disease duration (years)	10.4 (7.2)	9.2 (4.5)	15.3 (12.0)	0.020
Gender (M/F)	118/63	97/47	21/16	0.250
MCI / Non-MCI	66/115	58/86	8/29	0.037

Data are shown as mean (standard deviation). PD=Parkinson disease. ET=essential tremor. MCI=mild cognitive impairment.

Group statistics

DWM Fazekas score was significantly different between PD (0% score=0, 91% score=1, 6% score=2, and 3% score=3) and ET participants (3% score=0, 70% score=1, 22% score=2, and 5% score=3) (p=0.003, Fisher exact test). The mean ALPS-index score was 1.452 ± 0.158 in PD participants and 1.483 ± 0.161 in ET participants (p=0.349, Wilcoxon rank sum test). For PD participants, there was no difference in the mean ALPS-index score in males (1.441 ± 0.145) versus females (1.477 ± 0.181) (p=0.134, Wilcoxon rank sum test). In ET participants, there was also no difference in mean score in males (1.446 ± 0.156) versus females (1.532 ± 0.159) (p=0.101, Wilcoxon rank sum test). We observed a significant decline in the ALPS-index score with age (Spearman’s rho=−0.203; p=0.007) (Figure 2). When the diffusion along perivascular spaces (D_||) compared to orthogonal to perivascular spaces and primary fiber tracts (D_⊥) were considered separately, the relationship between D_|| was not significant with age (rho=0.036; p=0.636) but the relationship between D_⊥ and age was significant (rho=0.192; p=0.009).

Figure 2.

Relationships between ALPS-index and age. Trend lines depict linear regression with 95% confidence intervals (shaded).

ALPS-index and disease status, cognition, and white matter disease

A significant difference was observed in the ALPS-index between the matched PD and ET participants (p=0.037) (Figure 3). To understand whether this relationship was driven by the D_|| or D_⊥, the regression was repeated using these separate components of the ALPS-index score as the dependent variable. Here, D_|| and D_⊥were not significantly different between groups (p=0.926 and 0.134, respectively).

Figure 3.

Boxplots showing the distribution of the ALPS-index values for the matched participants with essential tremor (ET) and Parkinson disease (PD). The p-value was obtained from the general linear model (GLM) analysis.

We did not observe a significant association between the ALPS-index score (p=0.131), D_|| (p=0.196), or D_⊥ (p=1) and MCI status after controlling for age, sex, diagnosis status (PD or ET) and DWM Fazekas.

When examining the relationship between the ALPS-index and white matter pathology, we observed a significant inverse association between the ALPS-index and the DWM Fazekas score (p=0.002) while controlling for age, sex and diagnosis status (PD or ET). Both D_|| and D_⊥were significantly positively correlated with the DWM Fazekas score (p<0.001), but D_⊥ had a stronger association than D_||, resulting in an overall inverse association between the mean ALPS-index score and DWM Fazekas score.

Discussion

We investigated how a recently-proposed noninvasive imaging measure of diffusion along perivascular spaces, previously hypothesized to be attributable to glymphatic function, differed between adults with movement disorder presentations distinguished by the presence of (PD), or absence of (ET), proteinopathy. Investigation of this cohort was motivated by the hypothesis that cerebral protein (α-synuclein) accumulation may be partly attributable to altered perivascular fluid transport, which has been proposed to be related to glymphatic clearance dysfunction. Regression analyses were applied to control for variations in age, sex, and white matter disease. Sedation is a known modifier of glymphatic function. As glymphatic function is widely recognized to be inactive during the awake state, imaging during sedation may offer a unique window to detect imaging measures of glymphatic function. Sedating healthy control patients for comparison purposes is not feasible. Therefore, the comparison ET cohort was selected as a) the sedating procedure in this group was identical to that of the PD cohort and b) ET is distinguished from PD as it is not recognized as associated with proteinopathy. These findings demonstrate that the ALPS-index is reduced in individuals with PD relative to ET, consistent with the primary hypothesis. Decreased ALPS-index measures are reflective of less robust fluid transport in the direction of perivascular fluid motion. These data support the conclusion that altered perivascular fluid transport in PD may be a mechanism of underlying (α-synuclein) accumulation.

First proposed in 2012 by Iliff et al, the glymphatic system was hypothesized to reflect a previously unrecognized exchange of fluid between the subarachnoid CSF and the interstitial space of the brain parenchyma via the perivascular space and intervening aquaporin 4 channels[1]. This exchange of fluid has since been recognized as a likely important mechanism of waste clearance from the brain parenchyma[1,2]. Subsequent murine[11], non-human primate[12], and human studies[13–15] have further supported the existence of the glymphatic system and provided a new construct through which the potential pathophysiology of neurologic disease may be understood. Amyloid beta (Aβ) plaque accumulation in particular appears to be closely associated with glymphatic dysfunction[1,2], with one study reporting measurable increases in Aβ in the thalamus and hippocampus following just one night of sleep deprivation[16]. As the glymphatic system appears to function as a central waste removal system, impaired clearance of various proteins such as Aβ, α-synuclein, and tau, may result in increasing concentrations of protein aggregates in the interstitial space and neurons. Some degenerative diseases, including α-synucleinopathies, have been speculated to have an associated “interstitial fluidopathy”, referring to underlying pathology related to abnormal interstitial fluid dynamics[17].

PD pathophysiology is closely tied to abnormal accumulation of α-synuclein whereas ET pathophysiology is not characterized by abnormal accumulation of protein. A decrease in the ALPS-index in participants with PD relative to ET could parallel earlier ALPS-index results which demonstrated that lower ALPS-index scores were observed in AD and MCI patients with worse mini-mental-state-exam (MMSE) scores[3]. Such a result would support a broader argument that protein aggregation is associated with impaired perivascular fluid motion, and possibly glymphatic function in humans. Furthermore, this would provide additional support to the notion that the DTI-ALPS method is sensitive to glymphatic function.

Our results also point toward a statistically significant inverse relationship between white matter disease severity, as measured by Fazekas white matter scoring, and the ALPS-index score. White matter disease severity, commonly characterized as chronic microvascular ischemia, is therefore closely associated with impaired measures of perivascular motion, as measured by the DTI-ALPS method. We also observed a significant decline in the ALPS-index score with increased age. The ALPS-index score was not related to cognitive impairment assessment after controlling for white matter disease burden, age and sex.

Taken together these findings support a more sophisticated understanding of how the DTI-ALPS method may offer insight into the fluid dynamics in the brain. The presence of proteinopathy is demonstrated to be independently associated with altered measures of fluid motion in the brain parenchyma.

These findings should also be considered in light of several strengths relative to prior work. Our study benefits from a much larger patient number (n=181) as compared to the initial DTI-ALPS investigation (n=31)[3]. Furthermore, we build upon the prior study by employing robust regression analysis to further define how white matter burden, sex, age, and neuropsychological status affect DTI-ALPS measures. Another distinguishing feature of this study is that the patients uniformly underwent MR imaging under general anesthesia, eliminating motion artifacts and potential variables such as sleep or awake state which could confound results.

Several limitations of this study should be considered. Our study is limited in that the DTI-ALPS method offers an indirect proposed mechanism of assessing glymphatic function at one anatomic level of the brain. As there are limited, if any, methods for measuring glymphatic function in humans without intrathecal contrast injections, we investigated the relevance of this method in a larger cohort of participants to better understand the contrast mechanism. It must be noted that the contrast mechanism, as we demonstrate here, is also related to age and white matter status and care should be taken when assigning DTI-ALPS measures to glymphatic function. Second, this study could have regional bias as it only focused on perivascular fluid motion around the lateral ventricles due to the orthogonal direction of vessels, neurons, and the ventricles and not in other areas of the brain. It is not known how diffusion orthogonal and parallel to perivascular spaces may differ in other brain regions. Regional α-synuclein levels were not cross-correlated with DTI-ALPS values as α-synuclein values were not available in this cohort. Finally, variation within each group is substantial and the difference in mean values is modest. Therefore, despite significant difference among cohorts, this method cannot be used on an individual level. Further work is needed to develop methods capable of differentiating patients on an individual level.

In conclusion, we investigated how a newly-proposed noninvasive imaging measure of potential glymphatic function varied between individuals with PD, an α-synuclein proteinopathy, and individuals with ET which is not associated with abnormal protein accumulation. We observed decreased ALPS-index scores in PD compared to ET patients. We also observed an inverse relationship between white matter disease severity and age with ALPS-index scores. These results support the conclusion that altered perivascular fluid transport in PD may be a mechanism of underlying (α-synuclein) accumulation.

Highlights:

• The DTI-ALPS method is purported to contain information related to glymphatic function

• ALPS-index scores are reduced in Parkinson patients compared to essential tremor patients

• Findings parallel prior results demonstrating reduced ALPS-index scores in Alzheimer’s disease

• Altered perivascular fluid transport in PD may be a mechanism of underlying (α-synuclein) accumulation