The effects of aquaporin-4 polymorphisms on glymphatic function and motor symptoms in Parkinson’s disease

Abstract

Growing evidence suggests that the glymphatic system, driven by aquaporin-4 (AQP4) water channels, plays a key role in Parkinson’s disease (PD). We examined the impact of specific AQP4 variants on glymphatic function using diffusion tensor imaging along the perivascular space (DTI-ALPS), and explored potential mechanisms underlying motor symptom severity and progression in PD. PD participants exhibited lower DTI-ALPS indices compared to controls. Reduced DTI-ALPS at baseline was linked to more severe motor symptoms and faster longitudinal motor decline, as assessed by the Movement Disorder Society-Unified Parkinson’s Disease Rating Scale, Part III (MDS-UPDRS III). One specific AQP4 variant was associated with decreased DTI-ALPS and increased MDS-UPDRS III scores, with DTI-ALPS mediating this relationship at baseline and follow-up. Some variants exhibited indirect protective effects on motor symptoms via enhanced glymphatic function. These findings indicate that AQP4 polymorphisms contribute to glymphatic dysfunction and motor progression in PD, supporting strategies for disease modification.

Introduction

Parkinson’s disease (PD) is a common neurodegenerative disorder that is clinically characterized by motor dysfunction. Pathophysiologically, PD is marked by the accumulation of protein aggregates, including Lewy bodies composed of misfolded α-synuclein (α-syn)^1–3. The glymphatic system—defined as a cerebrospinal fluid (CSF) —interstitial fluid (ISF) exchange pathway, has recently emerged as a critical mechanism for clearing metabolic waste and misfolded proteins from the central nervous system (CNS)^4,5. This system is primarily driven by aquaporin-4 (AQP4), a water channel protein encoded by the AQP4 gene and predominantly expressed at the perivascular end-feet of astrocytes^6,7.

Studies in transgenic mice have established that AQP4 is a major regulator of CNS water homeostasis, where it controls the exchange of CSF with ISF^8,9. Growing evidence suggests glymphatic dysfunction contributes to neurodegenerative processes, as impaired glymphatic clearance may compromise the removal of metabolic waste and misfolded proteins^10–16. Animal studies have demonstrated that AQP4-facilitated glymphatic impairment has been implicated in PD^16–18. Importantly, the regulation of AQP4 expression is highly complex and can be influenced by genetic variation¹⁹. Evidence from an in vitro luciferase reporter assay demonstrated that an AQP4 single-nucleotide polymorphism (SNP) could modulate the expression level of AQP4²⁰. Altered AQP4 expression has also been observed in aging brains²¹. In Alzheimer’s disease, specific AQP4 SNPs have been linked to increased Aβ deposition, disease stage progression, and cognitive decline, highlighting the potential clinical impact of AQP4 variation^22–24. Recent studies indicate that some AQP4 SNPs may be associated with cognitive performance and altered brain activity in PD^25,26. Together, these findings suggest that genetic variation in AQP4 could influence glymphatic function and thereby contribute to disease pathophysiology.

Diffusion tensor imaging analysis along the perivascular space (DTI-ALPS) has recently emerged as a promising non-invasive indicator of glymphatic system function in vivo in humans. While the DTI-ALPS index provides a useful proxy for interstitial fluid dynamics by quantifying the diffusivity of water molecules in the periventricular space²⁷, it remains an indirect measure. It cannot directly visualize CSF–ISF exchange and may be influenced by various microstructural factors, which should be considered when interpreting results. Accumulating evidence suggests that reduced DTI-ALPS index is associated with impaired glymphatic function. A previous study further confirmed the validity of this index by showing a significant correlation with traditional gadolinium-enhanced glymphatic MRI techniques²⁸. Moreover, the DTI-ALPS index has shown good reproducibility and consistency across studies^29,30, providing a reliable proxy for glymphatic system functioning. Decreased DTI-ALPS index values have been consistently reported in PD and other neurodegenerative conditions^27,31–33, highlighting its potential as a biomarker for glymphatic dysfunction in neurological disorders.

Despite these advances, the precise mechanisms linking DTI-ALPS to the clinical manifestations of PD remain poorly understood. To address this gap, this study included newly diagnosed, drug-naïve patients with early-stage PD (Hoehn & Yahr stage I–II) to capture early alterations in glymphatic function before substantial neurodegeneration. Given the important role of AQP4 in perivascular CSF–ISF exchange and evidence that its polymorphisms can modulate expression²⁰, we hypothesize that AQP4 polymorphisms may affect glymphatic function and contribute to clinical symptoms in PD. Clarifying these associations may help explain the pathophysiological basis of DTI-ALPS alterations in PD and provide new insights into genotype-phenotype relationships.

Results

Demographic and clinical characteristics

Detailed demographic and clinical characteristics of the PD and Normal control participants are shown in Table 1. There were no significant group differences in age, sex distribution, years of education, body mass index, or Montreal Cognitive Assessment (MoCA) scores. PD patients had a median disease duration of 1.47 years (interquartile range [IQR]: 0.91–2.53), a median Movement Disorder Society-Unified Parkinson’s Disease Rating Scale Part III (MDS-UPDRS III) score of 20.85 (IQR: 14.0–26.0), and Hoehn and Yahr (HY) stage of 2.0 (IQR: 1.0–2.0). All PD patients were drug-naïve at baseline.

Table 1.

Demographic and clinical characteristics of the participants

	Parkinson’s disease (n = 128)	Normal control (n = 76)	p-value
Age	60.62 ± 9.54	59.41 ± 11.02	0.413
Sex (n, %)	Male: 81 (63.3%)	50 (65.8%)	0.678
	Female: 47 (36.7%)	26 (34.2%)
MoCA	28.0 (27.0–29.0)	28.0 (27.0–29.0)	0.723
body mass index (BMI)	25.95 (24.10–29.10)	25.71 (23.23–29.32)	0.458
years of education	16 (13.0–18.0)	16 (13.0–18.0)	0.479
Disease duration	1.47 (0.91–2.53)	—	—
MDS-UPDRS III	20.85 (14.0–26.0)	—	—
HY Stage	2 (1.0–2.0)	—	—
LEDD (mg/day)	0 (all drug-naïve at baseline)	—	—
DTI-ALPS
	Mean: 1.36 ± 0.26	1.47 ± 0.21	p < 0.001
	Left: 1.35 ± 0.25	1.47 ± 0.23	p < 0.001
	Right: 1.38 ± 0.28	1.48 ± 0.20	p < 0.01

Data were presented as mean ± standard deviation (SD), medians (interquartile range, 25–75^th percentile) and frequencies (percentages). Comparison between normal controls and Parkinson’s disease patients by Mann–Whitney U test, independent-samples t-tests, or chi-squared test.

MoCA Montreal Cognitive Assessment, BMI Body Mass Index, MDS-UPDRS III Movement Disorder Society-Unified Parkinson’s Disease Rating Scale, Part III, HY Stage Hoehn and Yahr Stage, LEDD The total daily levodopa equivalent dose, DTI-ALPS Diffusion Tensor Imaging Analysis Along the Perivascular Space.

Reduced DTI-ALPS in Parkinson’s disease and its association with motor symptoms

Compared to normal controls, patients with PD exhibited significantly lower DTI-ALPS values (Mean DTI-ALPS: p < 0.001; Left DTI-ALPS: p < 0.001; Right DTI-ALPS: p < 0.01) (Fig. 1a, Table 1), indicating impaired glymphatic function in PD.

Fig. 1. Reduced DTI-ALPS indices in Parkinson’s disease (PD) and its association with motor symptoms.

a Boxplots showing group comparisons of DTI-ALPS indices between normal controls (NC) (n = 76) and patients with PD (n = 128). PD patients exhibited significantly lower DTI-ALPS values than NC. b Partial correlation analyses between the DTI-ALPS indices and MDS-UPDRS III scores at baseline in patients with PD (n = 128). Significant negative correlations were observed for Mean DTI-ALPS, Left DTI-ALPS, and Right DTI-ALPS. PD Parkinson’s disease, NC Normal Controls, MDS-UPDRS III Movement Disorder Society-Unified Parkinson’s Disease Rating Scale, Part III, DTI-ALPS Diffusion Tensor Imaging Analysis Along the Perivascular Space.

To assess the direct association between glymphatic function and motor severity in PD, partial correlation analyses were conducted between the DTI-ALPS index and MDS-UPDRS III scores at baseline. Significant negative correlations were observed between DTI-ALPS and MDS-UPDRS III scores (Mean DTI-ALPS: r = −0.304, p < 0.001; Left DTI-ALPS: r = −0.302, p < 0.001; Right DTI-ALPS: r = −0.278, p < 0.01) (Fig. 1b). All correlations remained statistically significant after FDR correction (FDR q < 0.01).

Linear mixed-effects modeling revealed a significant interaction between baseline DTI-ALPS index and time (Time × DTI-ALPS: β = −0.72, 95% CI [−1.39, −0.06], p < 0.05), indicating that PD patients with lower DTI-ALPS experienced faster motor progression over time (Fig. 2). The main effect of DTI-ALPS was also significant (β = −10.83, 95% CI [−17.41, −4.25], p < 0.01), consistent with milder motor symptoms at baseline in patients with higher DTI-ALPS index. The main effect of time showed a non-significant trend toward increased MDS-UPDRS III scores over the follow-up period (β = 0.85, 95% CI [−0.08, 1.78], p = 0.072). In addition, older age was significantly associated with higher motor symptom severity (β = 0.25, 95% CI [0.03, 0.47], p < 0.05).

Fig. 2. Comparison of motor progression among Parkinson’s disease (PD) patients stratified by baseline DTI-ALPS index tertiles.

MDS-UPDRS III Movement Disorder Society-Unified Parkinson’s Disease Rating Scale, Part III, DTI-ALPS Diffusion Tensor Imaging Analysis Along the Perivascular Space.

Associations between AQP4 variants and motor symptoms

Among all tested SNPs, rs7240333 was significantly associated with higher MDS-UPDRS III scores (β = 7.28, p = 0.001, FDR q < 0.01) (Table 2). This suggests that carriers of the minor allele of rs7240333 may exhibit more severe motor symptoms. In addition, rs3763043 showed a nominal association with MDS-UPDRS III scores (β = 2.77, p = 0.089).

Table 2.

Associations between AQP4 SNPs and glymphatic function (DTI-ALPS Index) and motor symptoms (MDS-UPDRS III) in Parkinson’s disease

SNP	DTI-ALPS			MDS-UPDRS III
	β (95% CI)	Raw p-value	FDR p-value	β (95% CI)	Raw p-value	FDR p-value
rs1058427	0.20 (0.06, 0.35)	0.007	0.027	0.94 (−3.17, 5.05)	0.651	0.796
rs162009	0.08 (−0.04, 0.20)	0.194	0.427	1.17 (−2.09, 4.43)	0.479	0.753
rs335930	−0.01 (−0.14, 0.12)	0.870	0.909	1.45 (−2.05, 4.95)	0.413	0.753
rs3763040	−0.10 (−0.22, 0.02)	0.115	0.316	0.93 (−2.46, 4.32)	0.587	0.796
rs3763043	−0.06 (−0.18, 0.06)	0.324	0.510	2.77 (−0.43, 5.98)	0.089	0.492
rs3875089	0.28 (0.15, 0.41)	0.000	0.001	0.52 (−3.33, 4.36)	0.791	0.870
rs4800773	0.05 (−0.08, 0.17)	0.465	0.568	1.61 (−1.82, 5.03)	0.355	0.753
rs68006382	0.07 (−0.05, 0.20)	0.234	0.429	−1.37 (−4.69, 1.95)	0.415	0.753
rs71353405	0.08 (−0.13, 0.29)	0.448	0.568	−0.16 (−5.76, 5.44)	0.956	0.956
rs7240333	−0.28 (−0.43, −0.13)	0.000	0.002	7.28 (3.23, 11.33)	0.001	0.006
rs74163677	−0.01 (−0.22, 0.19)	0.909	0.909	−2.75 (−8.30, 2.81)	0.329	0.753

Bolded p-values indicate statistical significance after false discovery rate (FDR) correction.

MDS-UPDRS III Movement Disorder Society-Unified Parkinson’s Disease Rating Scale, Part III, DTI-ALPS Diffusion Tensor Imaging Analysis Along the Perivascular Space.

Associations between AQP4 variants and DTI-ALPS

Linear models adjusted for covariates revealed that rs7240333 was significantly associated with reduced DTI-ALPS index (β = −0.28, p < 0.001, FDR q < 0.01). Additionally, both rs3875089 (β = 0.28, p < 0.001, FDR q < 0.01) and rs1058427 (β = 0.20, p < 0.01, FDR q < 0.05) showed a significant positive association with the DTI-ALPS index (Table 2).

DTI-ALPS mediates the association between AQP4 variants and motor symptoms

To examine whether glymphatic dysfunction mediates the association between AQP4 genetic variants and motor symptom severity or longitudinal progression in PD, mediation analyses were performed (Fig. 3).

Fig. 3. Mediation model of AQP4 SNPs, DTI-ALPS index, and MDS-UPDRS-III motor symptoms.

Path (a) represents the effect of the SNP on the DTI-ALPS index; Path (b) captures the effect of the DTI-ALPS index on MDS-UPDRS III scores; Path (c′) indicates the direct effect of the SNP on MDS-UPDRS III scores. MDS-UPDRS III Movement Disorder Society-Unified Parkinson’s Disease Rating Scale, Part III, DTI-ALPS Diffusion Tensor Imaging Analysis Along the Perivascular Space.

For rs7240333, significant indirect effects via DTI-ALPS was observed both at baseline (β = 1.59, 95% CI [0.30, 3.50], p = 0.006) and during follow-up (β = 2.58, 95% CI [1.65, 3.69], p < 0.001), indicating a glymphatic-mediated pathway linking this SNP to motor symptoms severity and progression. The direct effect was significant at baseline (β = 4.05, 95% CI [0.04, 8.10], p = 0.046), but not at follow-up (β = 3.89, 95% CI [1.73, 9.50], p = 0.175), suggesting partial mediation at baseline. The proportion of the total effect mediated by DTI-ALPS at baseline was 27% (95% CI [0.05, 0.97], p = 0.008).

For rs1058427, significant indirect effects were observed (baseline: β = −1.75, 95% CI [−3.60, −0.29], p = 0.012; follow-up: β = −3.46, 95% CI [−5.00, −2.13], p < 0.001), while the direct effects were non-significant (baseline: p = 0.210; follow-up p = 0.130), indicating full mediation. A similar pattern was observed for rs3875089, where significant mediation effects were found at both baseline and follow-up (baseline: β = −2.68, 95% CI [−4.99, −0.98], p = 0.004) and follow-up (β = −4.60, 95% CI [−6.38, −3.03], p < 0.001).

AQP4 variants moderate the association between DTI-ALPS and motor symptoms

Moderation analyses were conducted only for SNPs without significant direct effects on motor outcomes, to assess whether AQP4 variation modulates the association between DTI-ALPS and motor symptoms in PD patients. Based on a significant interaction term between rs1058427 and DTI-ALPS (β = −11.63, SE = 4.73, p = 0.014), rs1058427 was identified as a moderator of the relationship between DTI-ALPS and MDS-UPDRS III scores. Specifically, the negative association between DTI-ALPS and MDS-UPDRS III scores was more pronounced among minor allele carriers. Similarly, a significant moderation effect was observed for rs3875089 (β = −10.65, SE = 4.66, p = 0.022), highlighting the potential role of AQP4 genetic variation in modifying the association between glymphatic function and motor severity. A visual representation of the moderation effect is provided in Fig. 4, illustrating the different slopes of the DTI-ALPS-MDS-UPDRS III relationship by SNP carrier status.

Fig. 4. Moderating effects of AQP4 SNPs on the association between DTI-ALPS and motor symptoms severity.

Regression lines depict the interaction effect of SNP carrier status on the ALPS-MDS-UPDRS III relationship. MDS-UPDRS III Movement Disorder Society-Unified Parkinson’s Disease Rating Scale, Part III, DTI-ALPS Diffusion Tensor Imaging Analysis Along the Perivascular Space.

Discussion

To our knowledge, this is the first study to explore the role of AQP4 gene polymorphisms in glymphatic function and motor symptoms in drug-naïve patients with sporadic PD. Our findings provide novel in vivo human evidence linking AQP4 genetic variation, glymphatic function (as measured by the DTI-ALPS index), and motor symptom outcomes in PD.

Among the tested AQP4 polymorphisms, rs7240333 was significantly associated with both reduced DTI-ALPS index and increased MDS-UPDRS III scores. Mediation analysis revealed that the DTI-ALPS index significantly mediated the effect of rs7240333 on both baseline motor severity and longitudinal MDS-UPDRS III progression. Our findings suggest that AQP4 genetic variants may contribute to motor symptom severity in PD through their influence on perivascular clearance efficiency, as reflected by alterations in the DTI-ALPS index. This underscores the plausibility that glymphatic function may serve as a potential mechanistic link between AQP4 genetic variation and clinical motor phenotype. The divergent effects of different SNPs are likely driven by their distinct functional consequences. The interpretation of the identified SNP is further supported by its genomic locations within the AQP4 gene³⁴. Specifically, rs7240333 is located in the 3′ untranslated region (3′ UTR), a region commonly involved in post-transcriptional regulation, including mRNA stability and microRNA binding^35,36. Variants in this region may alter mRNA turnover or translation efficiency of AQP4, thereby affecting perivascular water transport. Notably, previous studies have also reported associations of rs7240333 with greater sleep disturbances and shorter sleep duration, suggesting that it may influence glymphatic clearance indirectly through its effects on sleep, a critical period for glymphatic system activity³⁷.

Both rs1058427 and rs3875089 demonstrated indirect protective effects through enhanced glymphatic function (higher DTI-ALPS index), despite lacking direct associations with motor symptoms. In addition, moderation analyses showed that minor allele carriers of these two SNPs demonstrated greater sensitivity of motor symptom severity to changes in DTI-ALPS, implying that AQP4 variants may shape individual responsiveness to glymphatic dysfunction. Together, these findings indicate that rs1058427 and rs3875089 may contribute to improved glymphatic function and heightened responsiveness to glymphatic alterations. While rs1058427 is situated in the 3′ UTR involved in post-transcriptional regulation, rs3875089 is located upstream in the 5′ UTR/promoter region, where it may regulate AQP4 gene expression by altering transcription factor binding or promoter activity. Consistent with these functional implications, previous studies have also reported protective effects of rs3875089 across disease contexts, linking its minor allele to slower cognitive decline and dementia progression in Alzheimer’s disease²⁴, more favorable outcomes following traumatic brain injury^38,39, better maintenance of cognitive performance, and resistance to language deficits under conditions of increased sleep disturbances³⁷. Collectively, these findings suggest that rs3875089 may modulate AQP4 expression and perivascular clearance efficiency, thereby promoting resilience against pathological insults. Although the precise mechanisms remain to be clarified, these observations highlight the potential role of AQP4 as a cross-disease modulator involved in diverse pathological processes.

Together, these results demonstrate that glymphatic efficiency, influenced by AQP4 polymorphisms, shapes motor symptom severity and progression in PD. Importantly, the DTI-ALPS index was adjusted for FA and MD, thereby minimizing potential confounding from white matter microstructural changes. This suggests that observed DTI-ALPS index reductions are more likely reflective of altered perivenous water diffusivity, rather than generalized white matter degeneration, aligning with previous methodological recommendations⁴⁰. Collectively, these findings extend previous animal research^16,17 showing AQP4’s involvement in waste clearance pathways and reinforce the hypothesis that impaired AQP4-regulated glymphatic flow plays a critical role in PD pathophysiology by modulating disease progression. We observed a reduced DTI-ALPS index in PD patients compared to healthy controls, which was significantly associated with more severe motor symptoms at baseline, consistent with previous findings^31,33,41. Although baseline motor symptoms were relatively mild given the early-stage nature of the cohort, only 12.1% of participants had baseline MDS-UPDRS III scores <10, suggesting that floor effects are unlikely to have substantially biased the observed associations. Moreover, the availability of longitudinal data further reduces the potential impact of baseline floor effects on the overall findings. Individuals with impaired glymphatic function exhibited a faster progression of motor symptoms over time, in agreement with recent studies^42,43.

Despite these advances, some limitations need to be acknowledged. First, although DTI-ALPS offers a promising non-invasive estimate of glymphatic activity and was derived using a validated and reproducible pipeline, it remains an indirect proxy. Second, although our mediation model incorporated longitudinal clinical trajectories, causal inferences should still be interpreted with caution. Future studies in larger cohorts using advanced imaging approaches to assess glymphatic function are needed to validate these findings and clarify the role of AQP4-related glymphatic dysfunction in PD.

This study establishes AQP4 genetic variation as a key modulator of glymphatic function and motor progression in PD. rs7240333, through its putative effects on AQP4 expression, exacerbates glymphatic impairment and accelerates symptom severity, whereas rs1058427 and rs3875089 showed protective effects via enhanced glymphatic function. These findings support the potential of AQP4 genotype-based stratification to guide glymphatic-targeted interventions and highlight glymphatic enhancement as a promising strategy to mitigate disease progression.

Methods

Participants

All data used in this study were obtained from the Parkinson’s Progression Markers Initiative (PPMI) database (http://www.ppmi-info.org/). Clinical assessments included the Movement Disorder Society-Unified Parkinson’s Disease Rating Scale (MDS-UPDRS) and Hoehn and Yahr (H&Y) staging, with Part III of the MDS-UPDRS (during off-state) specifically used to evaluate motor symptom severity. Cognitive status was assessed using the Montreal Cognitive Assessment (MoCA). All PD participants were drug-naïve at baseline. The total daily levodopa equivalent dose (LEDD) was recorded throughout the follow-up visits. The major inclusion criteria for PD patients in the PPMI were as follows: (1) age ≥30 years; (2) drug-naïve; (3) Hoehn and Yahr (H&Y) stage I or II; (4) intact cognition at baseline (MoCA score ≥26). Only individuals with available whole-genome sequencing (WGS) and diffusion tensor imaging (DTI) data were included for analysis. Patients carrying pathogenic mutations in GBA or LRRK2 were excluded to minimize potential genetic confounding. A total of 128 newly diagnosed, drug-naïve sporadic PD patients and 76 normal controls were included in this study. The PPMI study conducted standardized longitudinal follow-up visits at predefined time points. Clinical and imaging assessments were scheduled at baseline (BL) and at 6 months (V02), 12 months (V04), 18 months (V05), 24 months (V06), 30 months (V07), 36 months (V08), 42 months (V09), 48 months (V10), and 60 months (V12) after baseline during the first 5 years. From year 6 onward, annual follow-up visits were conducted at 72 months (V13), 84 months (V14), 96 months (V15), 108 months (V16), 120 months (V17), 132 months (V18), and 144 months (V19) after baseline. For the current analysis, we used the annual follow-up data from BL, V04, V06, V10, V12, V13, and subsequent visits when available. Among the included PD patients, 104 PD patients had ≥3 visits (baseline plus ≥2 follow-ups) available for longitudinal analysis. The median follow-up duration was 8.3 years (IQR: 5.1–10.2 years), with a maximum follow-up of 12.0 years. All patients signed an informed consent form before they participated in the PPMI study.

Genetic analysis

Whole-genome sequencing data were processed using PLINK⁴⁴ to perform SNP pruning and quality control for AQP4 genetic variants. The AQP4 gene variants analyzed in this study are referenced by their dbSNP reference SNP IDs (rsIDs; e.g., rs3875089, rs4800773, rs3763040; https://www.ncbi.nlm.nih.gov/snp/), which are unique identifiers used in public SNP databases. SNPs that deviated from Hardy-Weinberg equilibrium (p < 0.05) or with a minor allele frequency of <5% were excluded. Linkage disequilibrium-based pruning was performed to reduce statistical redundancy while maintaining coverage of the AQP4 gene. SNP pruning parameters were set as: window size 10, increment 5, and variance inflation factor 2. Two SNP pairs exhibited high linkage disequilibrium (rs1058427 and rs12968026: r2 = 0.96; rs335930 and rs455671: r2 = 0.86), and in each pair, one SNP was retained as a representative (rs12968026 and rs455671 were removed). Finally, 11 SNPs were used for further analysis (Fig. 5). A dominant model was applied for all AQP4 SNPs. We assigned a “carrier” status to participants who are either homozygous in the minor allele or heterozygous, and a “noncarrier” status to participants who are homozygous in the common allele.

Fig. 5.

Illustration of the position of the SNPs used for the analysis within the AQP4 gene.

MRI data acquisition

Diffusion tensor imaging (DTI) data were acquired on a 3T MRI scanner using an echo planar imaging (EPI) sequence (TR = 900 ms, TE = 88 ms, flip angle = 90°, b = 1000 s/mm², 72 slices, slice thickness = 2 mm, voxel size = 2 × 2 × 2 mm³). Additional details regarding DTI acquisition protocols are available on the PPMI website (http://www.ppmi-info.org/study-design/research-documents-and-sops/).

DTI-ALPS processing

DTI data were preprocessed using MRtrix3 (https://www.mrtrix.org/) and FMRIB software library version 6.0.1 (FSL, http://www.fmrib.ox.ac.uk/fsl/)^45,46. The preprocessing pipeline included denoising, Gibbs ringing artifact removal, eddy current and motion correction, and diffusion tensor fitting. The DTI analysis along the perivascular space (DTI-ALPS) index was computed to evaluate glymphatic function by measuring the diffusivity of water molecules in the direction of the perivascular space (PVS)²⁷. In the DTI-ALPS method, the diffusivity of water molecules is measured at the level of the lateral ventricular body to noninvasively assess glymphatic function. At this level, the PVS is assumed to align with the orientation of medullary veins, running in the right-left direction, defined as the x-axis, and perpendicular to the ventricular wall. In the same region, projection fibers run in the head-foot direction (y-axis), while association fibers run in the anterior-posterior direction (z-axis), both orthogonal to the PVS. In this study, the DTI-ALPS index was calculated based on a previously validated and reproducible analysis pipeline, which demonstrated high inter-vendor, inter-rater, and test-retest reliability²⁹. ROI analysis was done using the provided ROIs drawn on FSL’s JHU-ICBM-FA-1mm. After registering individual diffusion-weighted images (DWI) to a standard fractional anisotropy (FA) template, ROIs were placed in a standardized manner in MNI space within projection and association fiber regions. The average value across the left and right hemispheres was used, which helps mitigate the effects of head tilt and other factors. The mean DTI-ALPS index was used in subsequent statistical analyses as the primary measure of glymphatic function. A higher DTI-ALPS index represents greater diffusivity along the perivascular space. The DTI-ALPS index was calculated as follows: $\mathrm{DTI}-\mathrm{ALPS\; index}=\frac{\mathrm{mean}({\mathrm{D}}_{\mathrm{xproj}},{\mathrm{D}}_{\mathrm{xassoc}})}{\mathrm{mean}({\mathrm{D}}_{\mathrm{yproj}},{\mathrm{D}}_{\mathrm{zassoc}})}$, where Dxproj and Dxassoc represent x-axis diffusivity in the projection and association fiber areas, respectively, and Dyproj and Dzassoc represent y-axis and z-axis diffusivity in the same areas.

Statistical analysis

Statistical analyses were performed using Python 3.11, with libraries including statsmodels, scipy, pandas, and semopy. Data visualization was conducted using the Seaborn package. The one-sample Kolmogorov–Smirnov test was used to assess the normality of continuous variables. All statistical tests were two-tailed, and p-values less than 0.05 were considered statistically significant. To control for multiple comparisons, p-values were adjusted using the Benjamini-Hochberg false discovery rate (FDR) procedure, and the resulting values are reported as FDR q-values.

First, group differences in the DTI-ALPS index between PD patients and normal controls were assessed using multivariable linear regression models, adjusting for age, sex, body mass index (BMI), years of education, and MoCA. To exploratorily investigate the direct association between glymphatic function and motor symptoms, partial correlation analyses were performed between the DTI-ALPS index and MDS-UPDRS III scores. Covariate effects were regressed out before correlation, including age, sex, BMI, years of education, race, disease duration, and white matter integrity measures (fractional anisotropy (FA) and mean diffusivity (MD)). FA and MD were extracted from the same ROIs used for ALPS calculation, and their averages were specifically included to control for potential confounding effects of white matter microstructural integrity on the DTI-ALPS index. To further examine the longitudinal relationship between the DTI-ALPS index and motor progression, a linear mixed-effects model was applied with MDS-UPDRS III score as the dependent variable. Time (follow-up year), baseline DTI-ALPS, and their interaction (Time × DTI-ALPS) were modeled as fixed effects, along with covariates. A random intercept was included for each subject to account for within-subject correlations over repeated follow-up assessments.

Second, linear regression models were used to test the association between AQP4 genetic variants and both glymphatic function (measured by the DTI-ALPS index) and motor symptom severity (measured by MDS-UPDRS III scores) at baseline. All models were adjusted for age, sex, BMI, years of education, race, and disease duration. Models for the DTI-ALPS index additionally included FA and MD as covariates.

Third, mediation analysis was performed to test whether DTI-ALPS mediated the relationship between AQP4 SNPs and motor symptoms (MDS-UPDRS-III scores), as well as motor progression over time. For both cross-sectional and longitudinal models, the baseline DTI-ALPS index was used as the mediator, with SNP carrier status as the independent variable. In the longitudinal models, MDS-UPDRS-III progression was modeled as the dependent variable, estimated through its interaction with follow-up time using a linear mixed-effects model. Models were adjusted for age, sex, BMI, years of education, race, FA, and MD. Disease duration was included only in the cross-sectional models, while LEDD and baseline MDS-UPDRS-III scores were additionally controlled for in the longitudinal models. 95% confidence intervals (estimated with 1000 bootstrap resamples) and p-values were calculated to assess statistical significance.

Fourth, structural equation modeling was used to assess whether AQP4 SNPs moderated the relationship between DTI-ALPS and motor symptoms. Interaction terms between the DTI-ALPS index and SNP carrier status were included as predictors in models with MDS-UPDRS III scores as the dependent variable. The same covariates as above were included in the moderation models.

Author contributions

Jianmei Qin: Conceptualization; Investigation; Data curation, Formal analysis; Methodology; Visualization; Writing - original draft; Writing - review & editing; Yi Fang: Conceptualization; Investigation; Data curation, Methodology; Writing - original draft; Writing – review & editing; Xiaojie Duanmu, Jiaqi Wen, Haoting Wu and Chenqing Wu: Conceptualization; Investigation; Methodology; Writing - original draft; Writing – review & editing; Qianshi Zheng, Weijin Yuan, Zihao Zhu, Jingwen Chen, Jingjing Wu, Mengze Jiang: Conceptualization; Investigation; Methodology; Writing – review & editing; Tao Guo, Cheng Zhou, Baorong Zhang, Xiaojun Xu, Xiaojun Guan, and Minming Zhang: Funding acquisition; Investigation; Methodology; Project administration; Writing - review & editing.

Data availability

The data supporting the findings of this study are publicly available from the Parkinson’s Progression Markers Initiative (PPMI) database at https://www.ppmi-info.org.

Competing interests

The authors declare no competing interests.