Intravenous arachnoid granulation hypertrophy in patients with Parkinson disease

Abstract

Intravenous arachnoid granulations (AGs) are protrusions of the arachnoid membrane into the venous lumen and function as contributors to the cerebrospinal fluid (CSF) flow circuit. Patients with Parkinson disease (PD) often present with accumulation of alpha synuclein. Previous works have provided evidence for neurofluid circulation dysfunction in neurodegenerative diseases associated with changes in CSF egress, which may have implications regarding AG morphology. The present study aims to investigate group differences in AG volumetrics between healthy and PD participants, as well as relationships between AG characteristics and clinical assessments. Generalized linear models revealed significant increases in AG volumetrics and number in PD compared to healthy controls. Partial Spearman-rank correlation analyses demonstrated significant relationships between AG metrics and motor and cognitive assessments. Finally, AG volumetrics were positively correlated with objective actigraphy measures of sleep dysfunction, but not self-report sleep symptoms.

Introduction

Cerebrospinal fluid (CSF) plays an integral role in cerebral nutrient delivery¹ and waste clearance^2,3. Most CSF is produced in the atrium of the lateral ventricles by the choroid plexus complexes⁴ before flowing through the cerebral aqueduct and entering the subarachnoid space. Subsequent passage of CSF into the perivascular routes and interstitial spaces has been observed⁵, which supports interstitial fluid (ISF) flow as part of the hypothesized glymphatic system². Recent models show evidence of CSF egress from the central nervous system via different pathways⁶, including along the cranial nerves⁷ and the spinal nerves into lymphatics⁸, as well as along other peri-sinus structures to the dural and extracranial lymphatics⁹. Among these peri-sinus structures, arachnoid granulations (AGs) contribute fundamentally to CSF resorption into the blood circulation and CSF homeostasis¹⁰.

AGs present as a hypertrophy of normal arachnoid villi and are projections of the subarachnoid space into both the lacuna lateralis and the dural venous sinuses¹¹. The focus of this study is primarily concerned with intravenous AGs protruding into the superior sagittal sinus¹², also defined as type I AGs¹³. While AGs are not present at birth, arachnoid villi are found in all humans and even smaller mammalian brains before maturation as microscopic structures. However, current imaging techniques do not provide acceptable contrast and spatial resolution for in vivo visualization of arachnoid villi. By 18 months of age, AGs are visible on gross inspection in larger mammals and humans, soon forming identifiable nodules in the sinus lumen and continuing to grow in size and number¹⁴. In addition to being larger in size, intravenous AGs are also distinct from arachnoid villi as they demonstrate endothelial-lined tubules that extend from the subarachnoid space to the venous system¹⁵. A previous study indicated that intravenous AG diameter, stalk diameter, and number are all positively correlated with age, while the percentage of meningothelial covering of AG is inversely correlated with age, along with significant changes in the internal milieu of AGs over time¹⁶. Together, these findings underscore the complex changes in AG morphology across the lifespan.

Past research based on animal models has suggested that advancing age results in a decline in the efficient clearance of interstitial solutes from the brain parenchyma via CSF-ISF exchange along perivascular routes¹⁷. Animal models have also provided evidence that ISF circulation is increased during sleep¹⁸ and clearance efficiency is decreased when sleep is reduced or dysfunctional^19,20. Yet, the present literature lacks an understanding of the relationship between sleep and the CSF circuit in human models, including in Parkinson disease (PD), a neurodegenerative disorder with a high prevalence of sleep dysfunction²¹. PD is the second most common neurodegenerative disease of aging populations, presenting with cognitive, functional, and motor impairments²². Given that glymphatic clearance is likely upregulated during sleep²³ and sleep deprivation is known to exacerbate cognitive and behavioral symptoms²⁴, there is likely a bidirectional relationship between sleep impairments and symptomatology in PD. Further, previous work has uncovered significant relationships between sleep disorders and reduced CSF alpha-synuclein within patients with PD²⁵. As aging negatively impacts sleep patterns, the subsequent increase in sleep dysfunction, observed in PD, may contribute to dysregulation of CSF movement and decreased CSF clearance efficiency²⁶. Understanding the link between sleep dysfunction and impairments in CSF outflow, including via AGs, is therefore relevant to elucidating whether these changes may be a cause or a consequence of advancing age and disease progression.

Here, an investigation of morphological changes of intravenous AGs was conducted using a combination of high-resolution 3D T₂-weighted MRI and deep-learning segmentation to test the hypothesis that AGs hypertrophy in the setting of PD. To address our first hypothesis of group differences in AG volumetrics, different measures of AG volume and count were compared between patients with PD and those of a similarly-aged healthy control cohort. Second, using correlation analyses, we tested hypotheses regarding AG hypertrophy and greater motor and cognitive symptom presentation in PD. Finally, as an exploratory analysis, correlation analyses were again conducted to investigate the hypothesis that AG volumes are increased in the setting of higher levels of sleep dysfunction in patients with PD.

Results

Demographics

In total, 74 participants completed the study. Thirty-two participants were diagnosed with PD (average years of symptoms = 6.2, average years since diagnosis = 4.4). Patients were on average 65.7 years old with a standard deviation of 8.5 years. Within the PD cohort, 12 participants were female and 20 were male. Moreover, 42 similarly-aged healthy controls were included in this study, with an average age of 67.4 years old and a standard deviation of 9.9 years. In the healthy control cohort, 22 participants were female and 20 were male. Details of the clinical and sleep assessments of the PD cohort can be found in Table 1. The PD cohort was composed of individuals with largely preserved cognitive function, with an average MoCA score of 24.4. We obtained baseline sleep data from 14 participants with PD who wore an actigraphy device ranging from 2 to 82 days (average = 23.5 days).

Table 1.

Description of the clinical presentation and sleep assessments conducted on the cohort of patients with Parkinson disease (PD)

	Mean [min–max]
Disease duration (n = 32)
Years since diagnosis	4.4 [0.1–18]
Years since onset	6.2 [0.4–19]
Clinical assessments (n = 32)
MoCA	24.4 [18–30]
SDMT	41.6 [24–57]
MiniBEST	21 [3–28]
UPDRS (Part III)	31.3 [8–61]
Self-reported sleep assessments (n = 32)
PROMIS SD	20.3 [9–31]
PROMIS SRI	15.8 [8–28]
Actigraphy (n = 14)
Sleep efficiency (%)	94.8 [91.6–97.1]
Wake after sleep onset (min/day)	26.1 [9.6–30.7]
Awakenings (number)	8.16 [5.0–12.8]

Clinical assessments were conducted off medication.

Group differences

We observed a significant increase of 73.3 mm³ in total AG volume (Fig. 1) in baseline PD scans compared to age and sex-matched healthy control scans (p_FDR = 0.004). In addition, data revealed significantly increased AG number (difference = 2.3, p_FDR = 0.019), maximum AG volume (difference = 24.4 mm³, p_FDR = 0.004), and mean AG volume (difference = 3.3 mm³, p_FDR = 0.004) in PD compared to age-matched controls (p_FDR = 0.004).

Fig. 1. Group differences in arachnoid granulation (AG) total volume (mm³).

A Significantly increased total AG volume in baseline Parkinson disease scans compared to healthy controls. B Significantly increased AG number in Parkinson disease compared to healthy controls. C Significantly increased maximum AG volume in Parkinson disease compared to healthy controls. D Significantly increased mean AG volume in Parkinson disease compared to healthy controls. Violin plots are shown with conventional boxplots and individual data points overlaid. Corrected p-values are shown.

Correlation with clinical assessment

Partial Spearman correlations adjusted for age (Table 2) revealed a significant positive relationship between total AG volume in patients with PD and MiniBEST (r = 0.46, p_FDR = 0.02), but no significant relationships with MoCA (r = 0.19, p_FDR = 0.57), SDMT (r = 0.35, p_FDR = 0.20) or UPDRS total score (r = −0.02, p_FDR = 0.98). We also found significant relationships between AG number in patients with PD with MiniBEST (r = 0.46, p_FDR < 0.05) and with SDMT (r = 0.50, p_raw = 0.01) before correction for false discovery rate. Data suggest inconclusive relationships between maximum AG volume in patients with PD and MoCA, SDMT, MiniBEST, and UPDRS-III total scores (Table 2).

Table 2.

Results of partial Spearman’s rank correlations, adjusted for age, between AG measures and clinical assessments (MoCA, MiniBEST, UPDRS) in participants with PD

	Total volume			Maximum volume			Mean volume			Number
	r	praw	pFDR	$\mathrm{r}$	praw	pFDR	r	praw	pFDR	r	praw	pFDR
Cognitive
MoCA	0.19	0.30	0.57	0.15	0.42	0.57	0.09	0.61	0.61	0.31	0.09	0.37
SDMT	0.35	0.10	0.20	0.17	0.44	0.51	0.14	0.51	0.51	0.50	0.01	0.06
Motor
UPDRS-III	−0.02	0.93	0.98	0.07	0.76	0.98	−0.01	0.98	0.98	−0.01	0.98	0.98
MiniBEST	0.46	0.02	<0.05	0.31	0.15	0.59	0.21	0.33	0.81	0.64	<0.01	<0.01

Significant relationships are underlined and bolded. Underlined refers to significant raw p-values that did not survive false discovery rate (FDR) correction.

MoCA Montreal Cognitive Assessment, SDMT Symbol Digit Modalities Test, UPDRS-III Universal Parkinson Disease Rating Scale part III, MiniBEST Mini Balance Evaluation Systems Test.

Correlation with brain and CSF volumetrics

Data indicate no significant relationships between AG metrics and total intracranial volume, gray matter volume, or white matter volume (Supplementary Table 1). Similarly, data suggest no significant relationships between total, maximum, and mean AG volume and AG number with CSF volumetrics, including total CSF volume, internal CSF volume, and external CSF volume (Supplementary Table 2).

Correlation with sleep assessments

Table 3 shows details of correlation analyses of sleep assessment data. Findings suggest non-significant relationships between AG measures and self-reported sleep assessments, with neither the SD or SRI sub-scores of the PROMIS measure being significantly related to AG measures, despite showing trends of increased sleep dysfunction (i.e., higher scores on SD and SRI tests) associated with larger AG volumetrics. Despite not reaching significance, these trends in subjective sleep reports were consistent with findings from actigraphy indicating that increases in AG volume and number in PD are related to sleep dysfunction.

Table 3.

Results of partial Spearman’s rank correlations, adjusted for age, between AG measures and sleep assessments (self-reported as well as actigraphy data), in participants with PD

	Total volume			Maximum volume			Mean volume			Number
	r	praw	pFDR	r	praw	pFDR	r	praw	pFDR	r	praw	pFDR
Self-reported assessments
SD	0.35	0.24	0.43	0.38	0.19	0.43	0.21	0.50	0.50	0.30	0.32	0.43
SRI	0.43	0.09	0.21	0.41	0.16	0.21	0.42	0.15	0.21	0.33	0.27	0.27
Actigraphy measures
SE	−0.88	<0.01	<0.01	−0.72	<0.01	0.01	−0.55	0.05	0.05	−0.69	<0.01	0.01
WASO	0.54	0.06	0.22	0.34	0.25	0.27	0.45	0.12	0.25	0.33	0.27	0.27
Awake	0.68	0.01	0.04	0.46	0.11	0.15	0.49	0.09	0.15	0.36	0.23	0.23

Significant relationships are underlined and bolded.

SD sleep disturbance, SRI sleep-related impairment, SE sleep efficiency, WASO wake after sleep onset, Awake and number of awakenings.

All individuals in the PD cohort were instructed to wear an actigraphy device for three months after MRI acquisition as part of a larger interventional study; however, incomplete compliance across participants required excluding a subset of participants from analysis. We selected a subset of the enrolled PD cohort (n = 14) who wore the actigraph for up to 82 days (range 2 to 82 days; Md = 28; M = 23.5, SD = 23.4) following MRI acquisition but prior to introduction of the intervention. This data was used as an estimate of baseline sleep parameters for exploratory analysis investigating the relationship between sleep behavior and AG volumetrics. We noted significant relationships between sleep efficiency and total AG volume (r = −0.88, p_FDR < 0.01), as well as with maximum AG volume (r = −0.72, p_FDR = 0.01) and AG number (r = −0.69, p_FDR = 0.01). We also found a significant relationship between number of awakenings and total AG volume (r = 0.68, p_FDR = 0.04). Further, total CSF volume was significantly related to SRI (r = −0.75, p_FDR < 0.01), sleep efficiency (r = 0.81, p_FDR < 0.01), wake after sleep onset (r = −0.84, p_FDR < 0.01), and number of awakenings (r = −0.82, p_FDR < 0.01) in this subset of participants. Internal and external CSF volume were not significantly related to any sleep assessments (Table 4).

Table 4.

Results of partial Spearman’s rank correlations, adjusted for age, between CSF volumetrics and sleep assessments (self-reported as well as actigraphy data), in participants with PD

	CSF volume			Internal CSF volume			External CSF volume
	r	praw	pFDR	r	praw	pFDR	r	praw	pFDR
Self-reported assessments
SD	−0.17	0.57	0.57	0.33	0.28	0.44	−0.31	0.30	0.44
SRI	−0.75	<0.01	<0.01	−0.42	0.15	0.17	−0.41	0.17	0.17
Actigraphy measures
SE	0.81	<0.01	<0.01	0.35	0.23	0.23	0.46	0.11	0.17
WASO	−0.84	<0.01	<0.01	−0.38	0.20	0.20	−0.50	0.08	0.13
Awake	−0.82	<0.01	<0.01	−0.39	0.19	0.19	−0.50	0.08	0.12

Significant relationships are underlined and bolded.

SD sleep disturbance, SRI sleep-related impairment, SE sleep efficiency, WASO wake after sleep onset, Awake and number of awakenings.

Discussion

In the present study, we investigated arachnoid granulation hypertrophy in patients with PD compared to healthy controls using a high-resolution 3D T₂-weighted MRI combined with a deep-learning segmentation model. We conducted correlation analyses to test hypotheses pertaining to the relationship between AG enlargement in PD with cognitive and motor impairments, and as an exploratory analysis, we investigated the relationship between AG enlargement and multiple measures of sleep disruption. The observed data provide evidence of enlarged AGs in the PD population, which correlate with multiple clinical metrics and sleep measures. These findings provide potential preliminary evidence that AG morphology relates to sleep dysfunctions in PD pathophysiology.

Intravenous arachnoid granulations have long been observed in the superior sagittal sinus¹². Various studies, including both post-mortem¹⁰ and in vivo¹³, have evaluated AGs within healthy adults and have demonstrated that AGs increase with age in healthy individuals²⁷. A previous study has hypothesized that AG hypertrophy results from increased CSF pressure in the subarachnoid space¹⁴ and/or decreased CSF outflow via the meningeal and cranial nerves¹⁷. This consequently results in increased CSF pressure and a need for another pathway to rid the brain of metabolic waste, causing AGs to hypertrophy to compensate for decreased egress efficiency and aid in further CSF bulk outflow. Other etiologies, such as incidental findings of giant AG (i.e., large enough to fill the dural sinus lumen and cause local dilation), have been reported in patients presenting with venous hypertension and headaches, potentially due to the abnormal growth of AG filling dural sinuses, and consequently altering venous outflow dynamics²⁸. In the present study, we observed significantly increased AG volume in patients with PD compared to healthy controls. This increase may serve as a compensatory mechanism for impaired CSF clearance via the recently proposed glymphatic pathway; however, independent measures of perivascular and interstitial flow are required to rigorously evaluate this possibility.

AGs were originally described as a main site for CSF egress²⁹. However, studies using intrathecal injection of gadolinium-based contrast agents challenged this idea^30,31, as the contrast agent reached maximum concentration in the plasma before peak concentration was detected in peri-sinus structures. This implicated other efflux routes as primary sites for CSF-mediated metabolic waste clearance. Although the results of these studies improve the present understanding of CSF egress, there are some limitations that must be addressed. First, gadolinium-based contrast is known to cross the blood-CSF barrier and subsequently enter the brain parenchyma³². In addition, it may not be valid to equate exogenous contrast movement with fluid and waste movement, as the molecular properties of gadolinium-based contrast agents may influence egress patterns along such small conduits.

Total, maximum, and mean AG volume, as well as AG number, were not significantly correlated with brain or CSF measures, indicating that increases in AG volumetrics and count in patients with PD likely occur independently of global brain atrophy. Past literature has described overall loss of gray matter³³ and ventricular expansion³⁴ in PD compared to controls, but few studies have evaluated global CSF volumetric changes as they relate to white and gray matter changes in the PD population. Across the healthy lifespan, however, intracranial CSF volume has been shown to increase linearly due to brain volume reduction³⁵, but whether these changes are replicated in PD remains unknown. Further, the present literature lacks an understanding of how AG volume relates to CSF volume and brain matter volume. Histologic evidence of the composition of AG reports some soft tissue elements with CSF flow turbulence³⁶, as well as CSF-incongruent fluid³⁷. As the morphology of AG is not completely understood, these reports reveal a gap in the interpretation of AG structure and function in relation to CSF and brain volumes. Thus, AG hypertrophy in PD compared to healthy control participants may reflect a unique difference between these cohorts not otherwise explained by potential changes in CSF or brain volume.

Total AG volume and AG number were significantly correlated with MiniBEST total scores in patients with PD, despite non-significant relationships with UPDRS-III scores. This could in part reflect the fact that the UPDRS-III reflects global motor dysfunction, with less sensitivity for motor symptoms in early PD³⁸ (mean score for our sample was consistent with Hoen and Yahr³⁹ stage 2.5, i.e., “mild bilateral involvement with recovery on retropulsion test”), whereas the MiniBEST is considered the strongest individual predictor of falls in patients with PD⁴⁰. Therefore, it is possible that the MiniBEST provides a more sensitive measure of gait and dynamic instability than the UPDRS-III, as it addresses specific impairments, suggesting that changes in AG volumetrics are correlated with the severity of specific balance and postural deficits or early motor symptomatology in PD, instead of global disease severity. Similarly, total AG volume (before correction for false discovery rate) and AG number were significantly related to scores on the SDMT in patients with PD, but not MoCA, a screening assessment of global cognitive status. Since global cognitive status was largely preserved in our sample, it is likely that low variability in the MoCA and higher sensitivity of the SDMT to frontal-subcortical network dysfunction associated with PD pathology accounted for these findings.

It should first be noted that this study did not measure perivascular or interstitial flow, which are the fundamental components of the proposed glymphatic system. However, results of the experiments conducted in this study may be considered alongside the growing literature investigating changes of glymphatic-related markers in PD. Previous studies have implicated the role of the CSF circuits in brain clearance⁴¹ and have suggested glymphatic system dysfunction in patients with PD^42–44. By utilizing diffusion tensor imaging (DTI-ALPS)^45–47, a metric of water diffusion at the level of the medullary veins and orthogonal to primary fiber tracks, past research has reported a decrease in DTI-ALPS in patients with PD compared to healthy controls⁴⁸. In addition, recent work suggested that fluid movement within the posterior aspects of the suprasellar cistern is reduced in 32 individuals with PD relative to 27 healthy controls⁴⁹. Further investigations revealed that a lower DTI-ALPS score was positively correlated with disease duration in PD⁵⁰, which may reflect downstream dysfunction. In this study, the finding of increased AG volumetrics in PD compared to controls may highlight the role of AGs as a compensatory response to decreased perivascular fluid movement in patients with PD.

Past studies have suggested water diffusion parallel to the medullary veins to be significantly lower in individuals with sleep disruption compared to healthy controls⁵¹. In those experiencing chronic sleep dysfunction, recent work has also shown enlargement of the perivascular spaces, which are critical to glymphatic function⁵². The increased volume of perivascular spaces and peri-sinus structures, such as AG, may not only be a marker of aging and neurodegenerative disease, but also of chronic sleep dysfunction. Taken together, these studies lend further support for reduced CSF clearance in both neurodegenerative states and in sleep dysfunctions, which could contribute to PD pathology progression. As reported in our study, total AG volume significantly relates to actigraphy-based sleep efficiency, an objective measure of sleep disturbance. In addition, mean and maximum AG volume, as well as AG number, significantly relate to actigraphy-based sleep efficiency before correction for false discovery rate. However, this finding was shown in a subsample of 14 participants, and subjective sleep reports (i.e., PROMIS SD and SRI sub-scores) were not significantly correlated with any changes in AG metrics, despite replicating the trends shown with actigraphy. Thus, our findings are suggestive of a relationship, but confirmation and clarification of causal relationship requires further empirical investigation.

Overall, the data indicate significant increases in total, average, and maximum AG volume, as well as AG number, in a cohort of patients with PD who present with mildly impaired-to-normal range cognition. This suggests that pathological changes in AG may develop in PD even in the absence of substantial neurocognitive decline. However, our findings do support relationships between AG volumetry and various assessment measures more sensitive to neurobehavioral dysfunction in PD (i.e., MiniBEST, SDMT, actigraphy-based sleep quality), suggesting pathological AG changes may be associated with prodromal or mild clinical decline in PD.

While results indicate significant relationships between AG hypertrophy and sleep difficulties, future studies are needed to determine the causal mechanism of increased AG volume in PD and its impact on motor and cognitive impairment and sleep. Potential explanations include AG hypertrophy due to impaired downstream clearance of CSF and/or AG enlargement as a compensatory response to handle increased CSF-mediated waste clearance of protein aggregation in PD pathology. In addition, as glymphatic clearance may occur at higher rates during sleep⁵³, chronic sleep dysfunction may contribute to less efficient CSF clearance.

Finally, while our work supports significant differences in AG volumetry between PD and healthy controls, future work could expand upon the scope of our findings to compare AGs in PD to other disease populations. This would allow for exploration of AG morphostructural and functional differences between various neurodegenerative disorders to better understand how they contribute to brain waste clearance and clinical impairments.

The deep-learning algorithm provided a method to automatically segment AG structures in vivo using high-resolution 3D T₂-weighted MRI. However, manual correction was necessary to ensure a high level of accuracy and precision in delineating AGs. Moreover, despite the use of high-resolution 3D T₂-weighted MRI, the current MRI resolution could only provide visualization of AG volumes above 2 mm³. Therefore, smaller AGs could not be assessed, resulting in a limitation of accurately counting the number of AG protruding the lumen of the superior sinus. To address this limitation, we investigated additional metrics, including total, mean, and maximum volume (i.e., larger AG size within an individual). Nevertheless, there are other anatomical variables, such as pedicle diameter, AG sphericity, and percent meningothelial coverage, amongst others, that we did not evaluate with the current method and may contribute to differences in AG morphology.

In addition, in this study, we conducted volumetrics analysis of intravenous AG structures. Although volume provides a comprehensive measure of structural changes, it does not allow us to assess functional changes. Findings presented in this study motivate future structural imaging analysis of other anatomical variables and functional imaging analysis of AGs to understand the impact of increased AG volume in patients with neurodegenerative proteinopathy. Finally, in our cross-sectional study design, while we are able to evaluate correlation between AG hypertrophy and clinical presentation, we are unable to infer causality. Future work should aim to investigate AG morphological changes longitudinally to uncover causal relationships between AG enlargement, PD pathology, and sleep dysfunction.

Furthermore, total AG volume was only significantly correlated with objective measures of sleep from actigraphy but not subjective measures of sleep using self-report PROMIS questionnaires. Additional investigation of whether AG volumetrics may relate to various sleep measures in a larger sample and in a controlled setting is necessary to validate the relationships detected in the current study. The present observational study using actigraphy was conducted in a small sample and extracted from a group of individuals who happened to wear an actigraph pre-intervention; as such, the study our data was taken from had not been designed for the purpose of investigating sleep quality in individuals with PD at baseline. While both subjective and objective measures of sleep quality can inform a multimodal assessment of sleep in our population, other, more robust measures, such as polysomnography, could be employed in future studies to better measure sleep disruption and build upon our results. Our finding of sleep dysfunction in PD relating to AG hypertrophy motivates further study in a research setting designed and aimed to specifically investigate the relationship between the CSF flow circuit and sleep dysfunction in humans, in both healthy and disease states.

We observed increased volume of intravenous arachnoid granulations in patients with Parkinson disease. Arachnoid granulation hypertrophy was also significantly related to motor impairment (on MiniBEST) and poorer cognitive functioning (on SDMT). In our sample, data revealed a significant relationship between AG volume and actigraphy-based sleep dysfunction, potentially suggesting a critical role of sleep in neurofluid regulation. Follow-up studies should investigate the causal role of arachnoid granulations in pathology progression, as well as how impaired sleep may influence this relationship.

Methods

Demographics

All participants provided informed, written consent in accordance with the Vanderbilt University Medical Center Institutional Review Board (IRB) and consistent with the Declaration of Helsinki and its amendments (IRB Study #191206). PD and healthy control participants completed a 3-Tesla MRI between January 2020 and September 2023 at the Vanderbilt University Medical Center (Table 5).

Table 5.

Description of the demographics and brain (total intracranial volume, gray matter volume, and white matter volume) and cerebrospinal fluid (CSF) volumetrics (total, internal, and external) in patients with Parkinson disease (PD) and healthy controls

	Parkinson disease	Healthy	p
Demographics
Number of participants	32	42	–
Age (year) [min–max]	67.4 [56–79]	64.4 [50–86]	0.09
Sex (Female/Male)	12/20	22/20	0.17
Brain volume
Intracranial (cm3)	1434.1 [1167.3–1750.0]	1389.7 [1060.6–1639.6]	0.28
Gray matter (cm3)	720.0 [582.4–895.0]	706.5 [532.8–845.9]	0.61
White matter (cm3)	430.9 [323.0–588.8]	422.7 [286.4–561.5]	0.55
CSF volume
Total (cm3)	264.2 [177.3–403.4]	242.5 [144.8–389.5]	0.05
Internala (cm3)	42.9 [15.8–95.6]	37.8 [16.6–72.8]	0.70
Externalb (cm3)	219.6 [160.7–331.3]	203.4 [123.3–350.2]	0.07

^aInternal CSF volume comprises lateral, third, and fourth ventricles.

^bExternal CSF volume comprises all subarachnoid space and other CSF compartment surrounding brain space.

Participants with PD were recruited as volunteers from the community seen in neurology clinics and were included if they were aged 55–80 and had PD as defined by the UK Brain Bank Criteria², which requires a diagnosis of Parkinsonian Syndrome (bradykinesia and muscular rigidity, 4–6 Hz rest tremor, and/or postural instability not caused by primary visual, vestibular, cerebellar or proprioceptive dysfunction) and at least three supportive prospective criteria for PD. Participants with any contraindication or inability to tolerate brain MRI, or with a history or signs of cerebrovascular disease, clinically significant neurological disorder (e.g. motor neuron disease, normal pressure hydrocephalus, brain tumor, etc.), severe or repeated head injury, encephalitis, untreated obstructive sleep apnea, or a Clinical Dementia Rating scale score ≥1 were excluded from the current study. Healthy controls were recruited from the Vanderbilt Glymphatic Imaging Project dataset (VGIP), in which all participants (age = 18–83 years) were scanned between January 2020 and September 2021 at Vanderbilt University Medical Center. Inclusion criteria involved compatibility with 3 Tesla MRI. Healthy participants were excluded if they presented with a history of cerebrovascular disease, anemia, psychiatric, or neurological disorder including but not limited to prior overt stroke, sickle cell anemia, schizophrenia, bipolar disorder, Alzheimer’s disease, PD, or multiple sclerosis. The presence of non-specific white matter lesions was not an exclusion criterion, as these lesions are prevalent with normal aging, and we sought our cohort to be representative. Clinical history was reviewed by a board-certified Neurologist (DOC; experience = 15 years) and anatomical imaging and angiography by a board-certified neuroradiologist (CDM; experience = 13 years).

Clinical assessments

Cognitive and motor impairments were assessed in patients with PD (Table 1) using the Montreal Cognitive Assessment (MoCA), Symbol Digit Modality Test (SDMT)⁵⁴, Mini Balance Evaluation Systems Test (MiniBEST)⁵⁵, and Universal Parkinson Disease Rating Scale – part III (UPDRS)⁵⁶. The MoCA is a screening assessment of cognitive function. The SDMT is a cognitive test that assesses processing speed and graphomotor speed. The UPDRS – part III is the motor examination portion used to measure disease severity in PD. Finally, the MiniBEST is a shortened version of the BEST assessment that measures dynamic balance, gait, postural reactivity, and mobility. All clinical assessments were performed on the day of MRI with the patient off medication for a period of 16 h prior to the examination.

Sleep assessments

Sleep assessment was accomplished through multiple methods to more comprehensively characterize the clinical domain. Clinical actigraphy units (ActiGraph wGT3X-BT)⁵⁷ were worn by patients following neuroimaging and clinical assessment visit, to characterize sleep-wake behavior (see Correlation with sleep assessments section for details on timing and length of assessment window). Resultant variables included in this study were sleep efficiency (SE; total estimated sleep time divided by total estimated sleep opportunity period), wake after sleep onset (WASO; total minutes awake during estimated sleep opportunity period), and number of awakenings (the number of times individuals would wake up throughout the night after falling asleep), all pre-processed and scored using ActiLife software (version 6.13.15). Assessment of sleep-related symptom burden was accomplished using the Patient-Reported Outcomes Measurement Information System (PROMIS)⁵⁸. Sleep-wake-related PROMIS short forms for two constructs were employed: sleep disturbance (SD) and sleep-related impairment (SRI). SD assesses perception of sleep quality, sleep depth, and restoration associated with sleep, while SRI focuses on perceptions of alertness, sleepiness, and tiredness during usual waking hours. PROMIS sleep questionnaires were administered on the day of MRI scan, following standard instructions requesting the participant to reflect on the last week of symptoms. In this study, raw total scores were used for analyses.

MRI acquisition

All participants underwent the same acquisition protocol. Participants were scanned using 3-Tesla MRI (Philips Medical Systems, Best, The Netherlands) with body coil radiofrequency transmission and phased array 32-channel reception. The scan protocol included standard non-contrasted anatomical imaging to ensure inclusion criteria were met along with a 3D T₁-weighted magnetization-prepared-rapid-gradient-echo (MPRAGE) scan with field-of-view = 150 × 256 × 180 mm, repetition time = 3.7 ms, echo time = 8.1 ms, flip angle = 8°, spatial resolution = 1 × 1 × 1 mm and a 3D T₂-weighted (sagittal acquisition) volumetric isotropic turbo-spin-echo acquisition (VISTA) sequence with field-of-view = 250 × 250 × 188.8 mm, repetition time = 2500 ms, echo time = 331 ms, flip angle = 90°, spatial resolution = 0.78 × 0.78 × 0.78 mm (see Fig. 2).

Fig. 2. Case examples of AG hypertrophy.

Panels A and B represent the maximum intensity projection (top) and coronal slices (bottom) of T₂-weighted scans, highlighting regions with high concentration of fluids, such as lateral ventricles and subarachnoid space, overlaid with 3D rending of the superior sinus (red), and intravenous arachnoid granulations (blue) from two participants in the axial, sagittal, and coronal planes. Panel A illustrates a participant with increased average AG volume (1, 2, and 3). Panel B illustrates a participant with normal average AG volume (1) and (2), but an increase of maximum AG volume (3). Both participants (A, B) had the same number of detected AGs.

Image processing

Quantification of total intracranial volume (TICV), brain tissue volumes (total volume, gray matter, and white matter) as well as CSF volumes, including internal CSF (i.e., lateral, third, and fourth ventricle), external CSF (i.e., subarachnoid space and basal cisterns⁵⁹), and total CSF, was obtained by feeding the T₁-weighted MRI to AssemblyNet, an ensemblist deep-learning algorithm estimating brain structure volumes from structural MRI⁶⁰. AGs were measured using the following metrics: total volume (sum of AG individual volumes), number of AG detected, mean volume (total AG volume divided by number of AG), and maximum volume (represents the volume of the largest AG detected in each participant).

Arachnoid granulations were measured using a novel non-invasive method combining high-resolution T₂-weighted MRIs and deep-learning⁶¹. T₂-weighted MRIs were corrected for field inhomogeneity and aligned to the MNI ICBM 152 symmetric template using affine registration⁶². Following pre-processing, a deep-learning model was trained and validated using 82 manually delineated scans from healthy controls and participants with various neurological conditions (i.e., Alzheimer disease, Parkinson disease, and Huntington disease) under the supervision of a board certified neuroradiologist. This software is freely available for research purposes (https://github.com/Center-of-Imaging-Biomarker-Development/spesis). Quality assessment of automatic segmentation was performed, and manual correction of segmentation was applied blinded from the neurological conditions of the patients.

Distribution of age and sex ratio differences between healthy controls and patients with PD were assessed using Kruskal–Wallis test and Chi-squared test, respectively. A group-wise analysis was conducted to investigate the primary hypothesis of differences in AG metrics between the two investigated cohorts and was performed using a generalized linear model. AG metrics (i.e., number, total, mean, and maximum AG volume) were used as dependent variables; pathological group (i.e., healthy control and PD) was defined as the independent variable; and age, sex, and intracranial volume (ICV) were used as covariates. As a secondary analysis, the relationship between AG features and clinical (i.e., cognitive and motor) assessments was determined using partial Spearman-rank correlation⁶³, adjusting relationship for age. The exploratory analysis of the relationship between AG features and sleep dysfunction, assessed using self-reported and actigraphy measures, was also conducted using the partial Spearman-rank correlation method, adjusted for age. All p-values are reported as uncorrected (noted p_raw) and corrected for false discovery rate (noted p_FDR) using the Benjamini–Hochberg procedure⁴⁸.

Author contributions

M.L., C.D.M., D.O.C., M.J.D., and K.H. contributed to study design and interpretation of data. J.J.E., M.J.D., A.K.S., J.E., and T.P. contributed to data acquisitions. C.D.M. and P.T. contributed to the review of the MRI data. M.L., C.D.M., M.J.D., and K.H. contributed to analysis and development of new methods. All authors contributed to writing the manuscript. All authors read and approved the final manuscript.

Data availability

The datasets used and analyzed during the current study are available from the corresponding author upon reasonable request.

Code availability

The underlying code for this study is available in GitHub and can be accessed via this link: https://github.com/Center-of-Imaging-Biomarker-Development/spesis.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41531-024-00796-x.