Settings and Participants

We collected data from 150 PD patients (46 GBA-PD and 104 non-GBA-PD). All PD subjects met the criteria for PD diagnosis according to the Movement Disorders Society (MDS) clinical diagnostic criteria for PD and were treated by a movement disorder specialist. The majority of the patients (90 non-GBA and 17 GBA-PD) were enrolled in the Dutch Parkinson Cohort (DUPARC) study.²² Twenty-six GBA1 carriers who participated in a nationwide GBA1 screening were recruited to additionally participate in the DUPARC study protocol. Other patients with similar disease duration (14 non-GBA-PD and 3 GBA-PD) were recruited as part of another ongoing study in our laboratory, which follows the same clinical and imaging protocols.²³ All subjects underwent ¹⁸F-FEOBV PET imaging, a T1-weighted magnetic resonance imaging (MRI) scan, and a standardized clinical and neuropsychological examination. Exclusion criteria included the use of medication with anticholinergic effects (eg, clozapine),²⁴ inability to provide written informed consent, an estimated low premorbid intelligence level (estimated intelligence quotient [IQ] <70 on the Dutch Adult Reading test),²⁵ and MRI contra-indications (eg, ferrous objects in the body or claustrophobia). Subjects taking cholinesterase inhibitors (ChEI) were not excluded from this study, despite the potential influence on cognition. ChEIs are prescribed for patients with Parkinson's disease dementia (PDD), being part of our population. Given that the GBA1-related PD phenotype is associated with greater cognitive dysfunction, excluding participants taking ChEI would disproportionately impact the representation of GBA-PD cases. All subjects gave written informed consent, and the study was approved by the local ethics committee.

Genotyping

Saliva was obtained from PD patients using Oragene DNA OG-500 tubes (DNA Genotek). DNA isolation, next generation sequencing (NGS), and data analysis were performed by GenomeScan B.V., Leiden, the Netherlands. Primers were selected to unambiguously sequence the functional GBA1-gene and not the pseudogene, using long-range polymerase chain reaction (PCR). Fragments were amplified using PCR and sequenced using Illumina cBot and HiSeq 400 as described previously.²⁶ The allelic nomenclature of GBA1 variants is provided. Accordingly, GBA1 variants were classified as risk, mild, severe, or unknown following the classifications presented in the GBA1-PD browser (https://pdgenetics.shinyapps.io/GBA1Browser/).The classification of mild and severe variants is based on the type of GD that they cause in the homozygous state. Risk variants do not cause GD, but serve as risk factors of PD.^8-10

Clinical Assessment

Brain Image Acquisition and Pre-Processing of ¹⁸F-FEOBV PET Imaging

All participants underwent brain resonance imaging (MRI) and vesicular acetylcholine transporter (VAChT) PET imaging using ¹⁸F-FEOBV radiotracer while in the “on” medication state. Acquisition procedures are detailed in the supplementary materials S2. The ¹⁸F-FEOBV PET data was pre-processed using a previously published pipeline.²⁹ Statistical parametric mapping (SPM12; Wellcome Trust Centre for Neuroimaging, London, UK) software was used to realign the PET imaging frames within subjects to reduce the effect of subject motion during imaging sessions. PET-MRI registration was performed using T1-weighted MRI volumetric scans. Freesurfer software (Laboratory for Computational Neuroimaging, Athinoula A. Martinos Center for Biomedical Imaging, Boston, MA) was used to run the recon-all function (default setting) for all MRI images. Parametric images reflecting distribution volume ratios (DVR) were constructed of ¹⁸F-FEOBV in the brain using supratentorial white matter above the ventricles (morphologically eroded) as a reference region.²⁹ Müller-Gärtner partial-volume correction (PVC) approach was used on the parametric PET images. Following recently published guidelines, we conducted our analyses with and without PVC to verify any variation of the results.³⁰

Imaging Analyses

Statistical Analyses

We evaluated the differences in demographic, clinical variables using an independent sample t test or Mann–Whitney U test, for normally and non-normally distributed data, respectively. Differences between PD GBA1 variants were tested with an analysis of variance test for continuous variables and χ2 testing for dichotomous variables. Correlations between DVR VOI values with and without PVC were calculated using Spearman's rank correlation or Pearson correlation according to the distribution of the variables. The statistical threshold was set at p < 0.05. Statistical analyses were performed using SPSS (version 28.0) and RStudio (version 4.3.2).

GBA1 Genetic Screening

In the GBA-PD group of 46 subjects, 13 different GBA1 nonsynonymous variants were identified (S3). We found 26 subjects with a risk variant, 11 subjects with a mild variant, and 6 subjects with severe variants. Additionally, 3 subjects exhibited a variant that is classified as having an unknown severity.³⁴

Participant Demographics and Clinical Assessments

The 150 PD patients had a mean (standard deviation [SD]) age of 66.7 (8.6) years and a mean (SD) disease duration of 3.8 (1.8) years. Demographic and clinical data of the GBA-PD and non-GBA-PD groups are presented in Table 1. The GBA-PD group contained a significantly higher percentage of females compared to non-GBA-PD and exhibited a higher NMS-Q total score. GBA-PD had a significantly lower mean T-score on attention and executive functioning, reflecting worse performance.

Cholinergic Comparisons

GBA-PD Versus Non-GBA-PD

GBA-PD exhibited lower ¹⁸F-FEOBV binding than non-GBA-PD in various bilateral frontal regions, including inferior, medial, middle, and superior frontal gyri, as well as the precentral, paracentral, and postcentral gyri. Lower binding was also observed in the lingual gyri, precuneus, thalami—particularly the ventral lateral nucleus—and limbic structures, such as the insula, cingulate gyrus, and parahippocampal gyri. Additionally, the middle and inferior occipital gyrus, along with the angular gyrus and cuneus in the left hemisphere, showed lower VAChT binding in GBA1 carriers versus non-carriers (Fig 1). The data showed no significant voxels in opposite direction (¹⁸F-FEOBV binding GBA-PD > non-GBA-PD).

Partial volume correction with the Müller-Gärtner method did not change the voxel-wise results in any cortical region (S4). Additionally, bivariate correlation analyses of the DVRs of the subcortical volume-of-interest showed strong correlations between regions with and without partial volume correction, except for the pallidum (S5). Therefore, we decided to present all data without additional partial volume correction.

GBA-PD Versus Non-GBA-PD: Covariate Disease Duration and LEDD

Although not statistically significant, clinical characteristics comparing both PD groups demonstrated a slightly longer disease duration in GBA-PD (mean [SD] 4.2 [2.0] years) compared to non-GBA-PD (mean [SD] 3.6 [1.7] years). In addition, patients were tested in the “on” dopaminergic medication state. Bidirectional whole brain voxel-based analyses were therefore repeated, while controlling for both disease duration and LEDD. As for the 2 maps, GBA-PD exhibited lower ¹⁸F-FEOBV binding in comparable pattern compared to non-GBA-PD, involving the same brain regions (Fig 2). Reverse directions again showed no results.

Post Hoc: Stratification by Sex

The GBA-PD group had a significantly higher percentage of females compared to non-GBA-PD group. GBA-PD males presented with significantly lower performance on the cognitive domains attention, executive functions, and language compared to non-GBA-PD males. A significantly higher score on the MDS-UPDRS II, NMS-Q, and RBD questionnaire was found in GBA-PD females compared to non-GBA-PD females. See Supporting Information S1 for results of demographics and clinical characteristics when stratified by sex (S6 and S7).

Voxel-wise comparison did not reveal any significant lower VAChT binding in GBA-PD females versus non-GBA-PD females, using p < 0.05 FDR corrected at the voxel level. Therefore, we lowered the threshold to an uncorrected p < 0.005, resulting in a lower binding in the left > right inferior and middle occipital gyrus, left lingual gyrus, and bilateral precentral, and postcentral gyri, superior frontal gyri, and thalamus in female GBA-PD. We also found a lower VAChT binding in GBA-PD male versus non-GBA-PD male in the cingulate gyrus (anterior and posterior), both lingual gyri, parahippocampal gyri, superior and transverse temporal gyri, precentral and postcentral gyri, and medial and superior frontal gyri showed lower VAChT binding in GBA-PD. See Figure 3 for voxel-wise results stratified by sex. No significant results were observed in the reserve direction. The Dice coefficient showed an overlap of 0.2514, comparing the lower cholinergic binding patterns between female; (GBA-PD female vs non-GBA-PD female) versus male; (GBA-PD male vs non-GBA-PD male), indicating a very limited overlap in these distributions.

Correlations of Cognitive Performance with Voxel-Based Regional ¹⁸F-FEOBV Binding

Whole-brain voxel-based analyses were performed to explore the correlation between regional brain VAChT binding and T-scores on attention and executive functioning, which were statistically significantly different between GBA-PD and non-GBA-PD. Only positive correlations were observed, indicating that lower cholinergic binding was associated with worse cognitive performance. The t-values of the voxel-wise analyses represent the strength of these correlations in the respective brain regions.

Executive Functions

We found widespread regional cerebral cholinergic correlations in both GBA-PD and non-GBA-PD. Both correlation analyses showed the involvement of the bilateral frontal and temporal cortex, parahippocampal gyri, lingual gyri, anterior and posterior cingulate gyrus, insula, thalami, lentiform nucleus, caudate, and pons. See Figure 4.

Attention

In the GBA-PD group, attentional performance showed widespread correlations with ¹⁸F-FEOBV binding in the frontotemporal regions (right precentral, inferior and middle frontal gyrus, cingulate gyrus, both parahippocampal gyri, temporal lobes, and uncus) and the posterior lobe of the cerebellum. A similar, yet more widespread, pattern was observed in non-GBA-PD, involving all cortical areas except the cerebellum. In the posterior cingulate, (pre)cuneus, and the thalami, correlations were more pronounced in non-GBA-PD versus GBA-PD. See Figure 5.

To assess the possible impact of ChEI, we calculated the overlap between the correlation maps of the complete GBA-PD group versus GBA-PD group excluding subjects taking ChEI, showing an overlap of, respectively, 0.8360 for executive function and 0.7545 for the attention domain. Additionally, we analyzed cognitive performance and regional ¹⁸F-FEOBV uptake between subjects taking ChEI to those without, regardless of their GBA1 status. The participants on ChEI (n = 8) did not differ from those not taking the medication (n = 142) in terms of age, disease duration, or educational level. However, subjects taking ChEI performed significantly worse across the attention, executive function, and memory domains (S8). Additionally, regional ¹⁸F-FEOBV uptake in ChEI users was lower in nearly all cortical areas, except for parts of the right > left occipital and temporal lobes. Lower VAChT binding was also observed in limbic regions, including the cingulate gyrus, insula, and parahippocampal gyrus, and the thalamus of ChEI users compared to those not taking the medication (S9).

Explorative Analysis: Genotype–Phenotype Correlations of Risk, Mild, and Severe GBA Variants

GBA-PD mild and severe variants exhibited a significant lower T-scores on the domain memory compared to GBA1 patients carrying a risk variant (S10). The mean ¹⁸F-FEOBV uptake values extracted from the mask derived from the significant results between GBA-PD and non-GBA-PD revealed a significant lower ¹⁸F-FEOBV binding when GBA1 risk variants (n = 26) and GBA1 mild and severe variants (n = 17) were compared to non-GBA-PD (n = 104). Differences between both groups of GBA1 variants were found (S11).

Clinical Phenotype

In this study, 56.5% of GBA-PD patients carried a risk variant and not a GD-related mutation, which is typically associated with a milder phenotype. These are the most common GBA1 variants in the Netherlands.²⁶ This might explain why we did not find a higher burden on most specific non-motor items. Additionally, previous data reported that GBA-PD and non-GBA-PD were clinically indistinguishable at the time of diagnosis.^{2, 36, 37} Further phenotypical distinction between GBA1 carriers and non-GBA carriers might become more apparent during later disease stages.^{36, 38}

Few studies have attempted to characterize the cognitive profile of GBA-PD using a complete cognitive test battery. One study found lower performance in working memory, executive function, and visuospatial domain in both GD associated GBA1 variants and E326K carriers compared to non-GBA-PD.¹⁷ Another study compared young-onset PD patients with L444P or N370S variants to patients without these variants and found significantly worse performance in the memory and visuospatial domains.¹⁵ The mean disease duration of patients in both studies was considerably longer, respectively, 15.4 and ~13 years compared to 3.8 years in our study. Previous longitudinal studies have reported diverging rates of cognitive deterioration in GBA-PD patients at ~3 years post-diagnosis.^{2, 36, 38} Our results suggest a profile of early cognitive decline in GBA-PD patients. The general progression of cognitive deficits in PD usually start with impairments in attention and executive function.³⁹

Cholinergic Denervation Pattern

A decrease in ¹⁸F-FEOBV binding was observed in GBA-PD versus non-GBA-PD, particularly in bilateral frontal regions, central areas, and limbic structures. Previous cross-sectional PET imaging data suggest a denervation gradient over time, progressing from posterior-to-anterior cortices, accompanied by cognitive worsening.^{40, 41} A recent PET study examining longitudinal changes in cortical cholinergic activity in PD found most severe denervation in posterior regions at baseline, with a posterior-to-anterior gradient and involvement of anteriorly located regions (in an anterior-to-posterior pattern) over time.⁴² The greatest regions of denervation between patients with de novo dementia with Lewy bodies (DLB) and PD patients 8 years post-diagnosis has been observed in more anterior regions of the brain, including the more anteriorly located regions, including the limbic network. The authors therefore proposed that the cholinergic changes in PD evolve over time and become similar to those in PDD and DLB.⁴³ Our current findings similarly demonstrate more severe cholinergic loss in anterior located and limbic regions in GBA-PD compared to non-GBA-PD, suggesting a more advanced pattern of cholinergic degeneration in GBA1 carriers, already present in early disease stages.

Our data on de novo PD demonstrated predominantly posterior cholinergic loss compared to controls in both GBA-PD and non-GBA-PD, consistent with previously reported findings.^{21, 40, 44-47} In addition, prior research has shown that variable posterior cholinergic denervation is also present in non-demented PD.⁴⁸ This indicates that the clinical onset of cognitive impairment because of cholinergic changes in PD may depend more on a widespread network dysfunction rather than localized cortical changes.^{46, 49} It has even been proposed that cholinergic innervation of posterior subregions may be affected already before PD diagnosis.⁴⁷ Therefore, the absence of significant differences in posterior cholinergic binding in the comparison between GBA-PD and non-GBA-PD may reflect a similar degree of cholinergic loss in these regions in both groups, resulting in no detectable differences.

GBA-PD performed significantly worse in the cognitive domains of executive function and attention. To gain a clearer understanding of the cholinergic system's role in the observed clinical differences between GBA-PD and non-GBA-PD, we conducted voxel-based correlation analyses. These correlation maps visualized cholinergic brain regions associated with cognitive domain performance in GBA-PD and non-GBA-PD separately. For executive function, ¹⁸F-FEOBV binding correlation maps revealed that the same brain regions were involved in both GBA-PD and non-GBA-PD patients, indicating that executive impairments may arise from similar mechanisms in PD, regardless of the presence of the GBA1 mutation. Notably, these regions were more severely affected in our GBA-PD cohort, which also exhibited greater executive deficits. The same holds true for attention. Widespread cerebral cholinergic denervation areas were found in both GBA-PD and non-GBA-PD. The extensive topography in both groups confirms widely distributed cholinergic involvement in attentional function.⁵⁰ However, in non-GBA-PD, we found an even more pronounced regional cholinergic topography correlated with attention, possibly because of the larger sample size of non-GBA-PD patients, allowing more statistical power.

Evidence on sex distribution in GBA-associated PD is inconsistent across studies. However, a meta-analysis suggested a higher prevalence of females with GBA-PD in Europe and Northern America in GBA-PD.⁵¹ In line with this, we found a higher percentage of females in GBA-PD compared to non-GBA-PD. To investigate potential differences in clinical presentation and underlying cholinergic pathologies, we repeated the analysis stratified by sex. Both GBA-PD females and males exhibited lower cholinergic VAChT binding compared to their non-GBA-PD counterparts. Interestingly, we found notable differences in the overlap between both maps. This indicates a distinct cholinergic profile in GBA1-male and GBA1-female compared to sex-stratified non-GBA-PD. We propose that future research investigating the clinical features of GBA-PD and their underlying processes should therefore also concentrate on sex-specific effects.

Genotype–phenotype studies have shown that PD patients with severe or mild GBA1 variants present with more frequent motor and non-motor symptoms.^{11, 12} However, our exploratory analyses revealed no significant differences in ¹⁸F-FEOBV uptake between mild, severe, and risk variants of GBA-PD. Given the small sample sizes in this subgroup analysis, potential differences may have been missed, emphasizing the need for further investigation with well-powered studies.

Finally, future studies should also investigate GBA1 variants in non-PD individuals, as this population is essential for properly analyzing potential prodromal PD subjects.

Strengths and Limitations

Strengths of this study include the large sample size of PD patients with and without GBA1 mutations and an extensive and detailed subject assessment. The assessment comprised a clinical evaluation, the use of ¹⁸F-FEOBV PET, and a comprehensive neuropsychological test battery, using T-scores for the evaluation of cognitive functioning. The T-scores were already adjusted for age, gender, and educational level, eliminating these variables as possible by recovering components. Another strength is the use of full-gene GBA1 sequencing. A limitation of the study is that motor assessments were performed in the dopaminergic “on” state. Another limitation is the relatively small sample size of the subgroups of GBA1 variants. We did not exclude subjects taking ChEIs despite the possible positive impact on cognition, which might be considered as a limitation. However, excluding participants taking ChEI would have disproportionately impact the representation of GBA-PD cases, introducing selection bias. Notably, patients on ChEIs still performed statistical significantly worse in the executive function, attention, and memory domain, along with lower regional ¹⁸F-FEOBV uptake compared to the patients not taking ChEIs. Furthermore, the correlation maps between ¹⁸F-FEOBV uptake and cognition did not differ between the overall GBA-PD group versus the GBA-PD group without the subjects taking ChEIs. Last, ¹⁸F-FEOBV is a selective presynaptic tracer, and there is no proven effect of rivastigmine on the tracer signal.⁴³