GBA Variants in Parkinson's Disease: Clinical, Metabolomic, and Multimodal Neuroimaging Phenotypes

Alterations in the GBA gene (NM_000157.3) are the most important genetic risk factor for Parkinson's disease (PD). Biallelic GBA mutations cause the lysosomal storage disorder Gaucher's disease. The GBA variants p.E365K and p.T408M are associated with PD but not with Gaucher's disease. The pathophysiological role of these variants needs to be further explored.

Conclusions: This is the first study to comprehensively assess (neuro-)biological phenotypes of GBA variants in PD. Metabolomics and neuroimaging detected more significant group differences than clinical and behavioral evaluation. These alterations could be promising to monitor effects of disease-modifying treatments targeting glucocerebrosidase metabolism. Alterations in the GBA gene (NM_000157.3) represent the most common genetic risk factor for Parkinson's disease recognized to date. 1 GBA encodes glucocerebrosidase, a lysosomal enzyme involved in sphingolipid metabolism. 2 A number of known GBA mutations cause the autosomal-recessive lysosomal storage disorder Gaucher's disease in biallelic carriers. 3 It has been hypothesized that glucocerebrosidase plays a role in α-synuclein degradation and therefore aggregate formation may be facilitated when glucocerebrosidase function is impaired, 4,5 thereby increasing the risk to develop Parkinson's disease and dementia with Lewy bodies (DLB) in mono-and biallelic carriers. 4,6 Clinical presentation and long-term clinical course of patients with Parkinson's disease carrying GBA mutations is not overtly different from patients without known genetic risk factors; however, many studies have presented evidence for an earlier age at onset, 7,8 faster progression of motor symptoms and cognitive decline, 1,4,8-10 and more visual hallucinations or psychotic symptoms. 8,[11][12][13] Recently, more frequently occurring alterations in GBA that do not cause Gaucher's disease, termed GBA variants (rather than mutations), have also been recognized as genetic risk factors for Parkinson's disease: large multicenter studies and a meta-analysis have established significant associations between Parkinson's disease and the single nucleotide polymorphisms p. E365K and p.T408M (traditional nomenclature: p. E326K and p.T369M). 9,14,15 The 1000 Gen-omes>Project reports p.E365K in 1% and p.T408M in about 0.4% of the world population, whereas they were found in up to 5% and 3.9% of patients with Parkinson's disease, respectively. 9, 16 It has repeatedly been shown that carriers of p.E365K, similar to mutation carriers, suffer from a faster cognitive decline than noncarriers, which has not (yet) been demonstrated for p.T408M. 10,16,17 A meta-analysis of 13 Parkinson's disease cohorts suggested a faster disease progression in both variants. 16 Metabolic consequences of glucocerebrosidase dysfunction in affected patients with Parkinson's disease have rarely been explored in vivo. 18 Mass spectrometry-based analysis of dried blood spots showed reduced glucocerebrosidase enzymatic activity 19 as well as increased levels of the lysosphingolipid hexosylsphingosine not only in Gaucher's disease-related mutations but also in p.E365K and p.T408M. 18 Neuroimaging studies comparing patients with Parkinson's disease with and without GBA mutations reported reduced cerebral blood flow in the parieto-occipital cortex, which resembled the pattern typically seen in DLB. 8,20,21 To the best of our knowledge, no neuroimaging studies to date have focused on GBA variants.
In summary, current evidence suggests that the clinical and neurobiological phenotype of patients with Parkinson's disease carrying GBA alterations (mutations or variants) is not fundamentally different from GBA noncarrier patients, but may be more severe and/or progress slightly faster, especially with severe mutations, 8,9 whereas variants may have milder clinical effects. 9 The current study aimed to comprehensively assess the phenotypes of patients with Parkinson's disease carrying the GBA variants p.E365K or p.T408M compared with patients carrying wildtype GBA. Carriers and noncarriers were compared regarding clinical and family history, motor and nonmotor symptom severity, cognitive function, metabolomics, and multimodal neuroimaging using positron emission tomography (PET) and functional magnetic resonance imaging (fMRI).

Patients
From a larger, well-characterized cohort (KFO 219), DNA samples were available for 56 patients with Parkinson's disease. Inclusion criteria were Hoehn and Yahr stage 22 ≤ 3, age ≥ 40 years, and absence of dementia, 23 deep brain stimulation, or cerebral pathologies other than Parkinson's disease. All patients were recruited at the University Hospital of Cologne and diagnosed by a movement disorder specialist according to the UK Brain Bank criteria. 24 The study was approved by the local medical ethics committee (EK12-265) and registered with the German Clinical Trials Register (DRKS00005388); informed consent was obtained from each participant per the Declaration of Helsinki.
Patients in this cohort underwent an extensive study protocol with the assessment of motor, cognitive, neuropsychiatric, and other nonmotor symptoms; multimodal neuroimaging; and metabolomic and genetic analysis. Several of these procedures have been described in previous publications. [25][26][27] Clinical and Behavioral Data The motor part of the Unified Parkinson's Disease Rating Scale (UPDRS-III, including subscores for tremor and akinesia-rigidity to determine motor subtypes 28 ), collection of blood samples, and functional neuroimaging were performed after antiparkinsonian medication was discontinued for a minimum of 12 hours (levodopa) and up to 3 days (dopamine agonists). Clinical history, neuropsychological and other nonmotor data were obtained on dopaminergic medication. A cognitive test battery covered the domains executive function, memory, attention, language, and visual-spatial abilities, from which a global cognition z score was computed using age and education-adjusted standard norms. Self-rated scales were applied to measure apathy, depression, (hypo-)mania, impulsivity, and other nonmotor symptoms.
Groups were compared using IBM SPSS statistics version 25 (IBM Corp., Armonk, NY). Categorical data were analyzed with Fisher's exact test, and Mann-Whitney U test was used for ordinal variables. For continuous variables, normal distribution was tested by Shapiro-Wilk test; groups were compared by t test or U test as appropriate. Cognitive scores were compared by analysis of covariance to adjust for points on the Beck Depression Inventory, version II (BDI-II), as depression is known to heavily influence test performance. 29

Biospecimen Collection and Processing
Blood samples for metabolomic and genetic analyses were drawn after overnight fasting. A gene panel analysis was performed comprising 29 genes previously linked to parkinsonism or dystonia. Rare variants in Parkinson's disease genes (GBA, LRRK2, PARK7, PRKN, PINK1, SNCA, VPS35) were validated by Sanger sequencing, and confirmed alterations were registered.
Details of metabolomics processing and analysis in the KFO 219 cohort were previously published. 25 Plasma samples of 54 patients could be analyzed (Table 1). Gas chromatography coupled to mass spectrometry was applied to measure polar and nonpolar metabolite extracts. Metabolomic profiles were analyzed in an untargeted approach, comparing 71 metabolites between wildtype patients and GBA variant carriers using Welch's t test. In this exploratory analysis, suggestive group differences with uncorrected P < 0.05 are reported.

Neuroimaging Data Acquisition and Analysis
Image acquisition and preprocessing have previously been described in detail. 27 search volume to the bilateral striatum. An uncorrected threshold of P < 0.05 was defined as a suggestive group difference. [ 18 F]FDG PET scans were processed using the topographic rating algorithm implemented in ScAnVP (http://feinsteinneuroscience. org) to measure the expression of 2 distinct Parkinson's disease-related covariance patterns: the Parkinson's disease related pattern (PDRP), associated with disease progression and motor symptoms, 31,32 and the Parkinson's disease cognitive pattern (PDCP), associated with cognitive dysfunction. 33
In addition, 1 GBA wildtype patient was heterozygous for the likely benign variant c.1000C>T (p.R334C) in PRKN; 1 carrier of GBA:p.E365K was also heterozygous for the likely benign variant c.587C>T (p.P196L) in PINK1 and the PRKN mutation c.823C>T (p.R275W). Although the PRKN mutation p.R275W in biallelic carriers is associated with autosomal-recessive Parkinson's disease, 35 only a heterozygous carrier was detected here. Group comparisons were repeated without the PRKN mutation carrier, and because the results were largely unaltered, the subject was not excluded from final analyses.
Because carriers of the 2 variants p.E365K and p. 408M were similar concerning all measures of interest in this study (see Fig. 1, Tables 2, 3, S1), variant carriers were combined into 1 group for all comparisons with noncarriers. men. The 13 variant carriers were 66.7 AE 8.9 years old, and 11 (84.6%) were men. In each group, 3 patients had a first-degree relative with Parkinson's disease, corresponding to 7.1% of wildtype and 23.1% of variant carriers. These differences were not statistically significant; details are presented in Table 2.

Clinical History and Symptoms
Clinical history, symptom severity, and motor subtypes were not significantly different between genotypes and are detailed in Table 2 36 was 455.8 AE 286.1 mg in noncarriers and 26% higher (577.1 AE 253.9 mg) in carriers; these group differences were not statistically significant. Likewise, no significant differences were detected with the nonmotor symptom scale.
Variant carriers scored lower in all 5 cognitive domains. The global cognition z score was significantly lower when BDI-II was included as a covariate (P = 0.039). Among the cognitive domains, executive and visual-spatial functions were the most affected, although group differences were narrowly not significant. BDI-II scores were significantly lower in carriers (5.8 AE 6.4) than in noncarriers (11.6 AE 7.8; P = 0.012), whereas groups were similar concerning apathy, (hypo-)mania, and impulsivity. Clinical and behavioral data are summarized in Table 2.

Metabolomics
Abundance of 1,5-anhydro-D-glucitol, asparagine, ornithine, glutamine, and glycine as well as the unknown metabolites with retention indices (RI) RI1169 and RI1568 was increased in GBA variant   carriers compared with noncarriers, whereas an unknown metabolite RI1120 showed decreased levels in carriers. All variant carriers were included in the metabolomics analysis. Results are presented in Figure 1 and Table S1.

Neuroimaging
Genotyping results in imaging subsamples can be found in Table 3. [ 18 F]FDopa uptake was reduced in the bilateral caudate nuclei, the antero-medial putamen ipsilateral, and nucleus accumbens contralateral to the more affected body side in variant carriers compared with noncarriers. PDRP expression was significantly higher in patients with GBA variants (3.07 AE 1.67) than in patients with wildtype GBA (0.99 AE 1.71, P = 0.0007), with similar scores in both variants (see Fig. 1C). PDCP expression was higher in carriers (0.64 AE 1.18) than in noncarriers (0.11 AE 1.41), but not significantly different between groups (P = 0.250). GBA variant carriers showed significantly reduced [ 18 F] FDG PET activity in the bilateral medial and lateral parietal lobe. FC was significantly reduced between the left and right caudate nuclei and the bilateral occipital cortex in carriers; the right nucleus accumbens showed reduced connectivity with the left superior parietal and right occipital fusiform cortex. More precisely, FC values were near zero in wildtype and negative in variant carriers, showing anticorrelations between activity fluctuations of the seed regions and the occipital/parietal cortex. FC of the left nucleus accumbens and left and right putamen was not different between groups. The results of voxel-wise group comparisons are detailed in Table 3 and depicted in Figure 2.

Further Analyses
Group characteristics in the subsamples for each analysis were similar to the whole sample (see Tables S2a-d). In patients with [ 18 F]FDG PET, the higher LEDD in carriers reached statistical significance (P = 0.017; Table S2b). To control for a potential influence of disease duration or antiparkinsonian medication, clinical, metabolomic, and imaging data were additionally compared with correction for disease duration and LEDD, which had only minimal effects on group differences; PDRP results remained significant with UPDRS-III as a covariate (P = 0.0015). The difference in BDI-II scores could not be explained by antidepressive medication or dopamine agonists, which were similar between groups (Tables 2, S2a-d). When Apathy Evaluation Scale scores were corrected for BDI-II, a trend for more apathy in variant carriers was observed (P = 0.083).

Discussion
This is the first study to date that provides detailed phenotypical data about the 2 main GBA variants that do not cause Gaucher's disease but are associated with Parkinson's disease, p.E365K and p.T408M. The results point to similarities with Gaucher's diseaserelated GBA mutations and are suggestive of a more severe course of the disease.
The present cohort had a relatively high proportion of GBA variant carriers (23.2%) compared with less than 10% reported in previous studies, whereas the rate of GBA mutations (1/56) was comparably low. 9,16 The total frequency of GBA alterations was not significantly higher than in the Dutch PROPARK cohort 9 and considering the sample size is probably coincidental. Mutations are much rarer than variants, most of them occurring far less frequently than p.N409S, 9 so they may be missed in a small cohort. The high proportion of variant carriers makes it extremely unlikely that exclusion criteria (eg, dementia) were biased against them. Carriers had a trend for more first-degree relatives with Parkinson's disease than noncarriers (23.1% vs 7.1%); similar rates of a positive family history are regularly reported in GBA mutations. 1,7 A metaanalysis has not found higher UPDRS-III scores in variant carriers, 16 whereas LEDD may be increased 11 and should be investigated longitudinally in larger cohorts. Surprisingly, GBA carriers showed less symptoms of depression than wildtype patients. More depression in mutation carriers has been reported, 37 but often no effect was found. 11,12,38 In non-Gaucher's-related variants, no change in depression has been described. 11,16,39 When BDI-II was taken into account, significantly reduced global cognition was detected. However, cognition z scores in both groups indicated function within the normal range (−0.13 in wildtypes vs −0.34 in carriers). An increased risk for cognitive deficits has previously only been demonstrated for p. E365K. In this study, both variants scored similarly in global cognition and BDI-II, thus contributing equally to the group difference.
Although carriers and noncarriers presented only minor clinical differences, metabolomics and neuroimaging highlighted interesting effects of GBA variants. Levels of 8 metabolites were suggestively altered in GBA variant carriers compared with noncarriers, that is, 1,5-anhydro-D-glucitol, asparagine, ornithine, glutamine, glycine, and 2 unidentified metabolites (RI1169 and RI1568) being increased, and the unknown metabolite RI1120 being decreased. Further elucidation is required for the identification and validation of the unknown metabolites. These preliminary results provide new hints to which pathways may be altered in GBA-related Parkinson's disease. Increases of anhydroglucitol and other polyol pathway metabolites have been shown in the brain tissue of a DJ-1 knockout Parkinson's disease mouse model, together with decreased levels of glycolysis intermediates. 40 Therefore, the polyol pathway might be dysregulated in GBA variant carriers. Ornithine has been reported to be higher in the cerebrospinal fluid of patients with Parkinson's disease than in healthy controls, 41 whereas blood plasma levels were increased in a rotenone-lesioned rat model. 42 The authors suspected a link to impairments in mitochondrial transport and the urea cycle. Asparagine, glutamine, and glycine have all been found to be significantly increased in blood plasma of patients with Parkinson's disease compared with healthy controls. 43 Higher levels in GBA variant carriers may indicate that changes observed in Parkinson's disease are more severe in patients with genetically determined glucocerebrosidase dysfunction. The results of this untargeted, hypothesis-generating approach are preliminary but can help focus future investigations on certain pathways or metabolites.
Neuroimaging studies of GBA-related Parkinson's disease so far have almost exclusively examined Gaucher's disease-causing mutations, and no publications could be found that applied imaging techniques specifically in the p.E365K and p.T408M variants. Recently, metabolic networks have been investigated in patients with Parkinson's disease carrying Gaucher'srelated GBA mutations, showing similar results of significantly increased PDRP scores and slightly, not significantly increased PDCP expression values. 44 Increased PDRP scores are in line with a more severe course of the disease and suggest that subclinical neurobiological changes are present in GBA variant carriers even when clinical differences are not (yet) significant. The PDRP is already expressed in very early stages of the disease, even before the onset of motor symptoms in hemiparkinsonism 32 and prodromal Parkinson's disease. 45 Conversely, increased PDCP expression values may only reach significance with more advanced impairment: they steadily increase with disease progression and cognitive decline, but a significant group difference was described only between patients with multidomain cognitive impairment and healthy controls. 34 Integrity of the dopaminergic system has not previously been examined in carriers of GBA variants. In mutation carriers, results were variable: [ 18 F]FDopa uptake was reduced in patients with Gaucher's disease with and without Parkinson's disease compared with healthy controls 20 ; compared with noncarriers, [ 123 I] FP-CIT binding was lower in patients with more advanced stage Parkinson's disease carrying severe but not mild mutations, 8 whereas a recent study reported higher dopamine transporter density in carriers' more affected striatum at an early stage. 46 Discrepancies between these and our results may be attributed to different methodologies (ie, higher statistical threshold, region-of-interest based approach, radiotracer and target), and group differences may evolve with progression. Here, reduced uptake was detected at a low threshold, possibly because of the small number of variant carriers with [ 18 F]FDopa PET (n = 7). Striatal dopaminergic loss progresses from posterior to anterior in Parkinson's disease, 47 and the caudate nucleus is involved in cognition, particularly executive functions. 48 Interestingly, variant carriers displayed FC anticorrelations between the caudate nuclei and the occipital cortex, whereas noncarriers' FC values were near zero. An almost identical pattern of caudate-occipital anticorrelations was previously reported in Parkinson's disease dementia (PDD) compared with positive correlations in healthy controls. 49 The striking similarity between these findings strongly suggests a link to more cognitive deficits in patients with Parkinson's disease with GBA alterations. In addition, altered connectivity patterns detected here could relate to GBA mutation carrier's susceptibility to psychotic symptoms: in Parkinson's disease with hallucinations, occipital and striatal regions show reduced connectivity, 50 and visual areas are deactivated during the hallucinatory experience. 51 In schizophrenia with visual hallucinations, abnormal FC was described between the nucleus accumbens and higher visual areas. 52 Functional MRI has, to our knowledge, never before been applied in GBA-related Parkinson's disease.
Parieto-occipital cortex hypoactivity in our cohort was very similar to previous findings in Parkinson's disease with GBA mutations 20 and in DLB 8 as well as in Parkinson's disease with visual hallucinations. 53 In line with this, cortical Lewy body load has been shown to be higher in patients with Parkinson's disease carrying GBA mutations. 54 In our cross-sectional study of midstage and early-stage patients, hallucinations or psychotic symptoms were not observed, but imaging findings indicate an increased susceptibility to these symptoms in variant carriers, which has repeatedly been described in GBA mutation carriers. 1 It has been shown that hallucinations and temporo-parietooccipital hypometabolism precede PDD, 53 but to date no imaging predictors of hallucinations in Parkinson's disease have been described.
Despite a high degree of similarity between the 2 variants investigated here (Tables 1-3, S1-S2, Fig. 1), future studies with more participants should address potential differences between them. To date, most neuroimaging studies are performed without genetic testing. The unexpectedly high number of carriers found here is a reminder that, especially in smaller cohorts, the rate at which certain genetic risk variants occur is somewhat random. Depending on the methods used, results could potentially be confounded by the genetic risk profile of included patients. The results of previous studies in the same cohort investigated here are unrelated to the differences detected between carriers and noncarriers of GBA variants. [25][26][27] One reason why GBA has lately been drawing attention among researchers and clinicians is the hope for an-at least partially-causative therapy targeting glucocerebrosidase metabolism. Ambroxol, for example, enhances glucocerebrosidase activity in the cerebrospinal fluid of patients with Parkinson's disease, 55 and a randomized controlled trial in patients with PDD with and without GBA mutations or variants is currently recruiting. 56 In conclusion, metabolomic and neuroimaging findings are suggestive of a more severe Parkinson's disease pathology and demonstrate similarities with PDD and DLB even in carriers with only minimal cognitive decline. Group differences were apparent at the (neuro-)biological level, but not significant at the clinical level. The lack of clinical differences is congruent with the hypothesis that more severe mutations have a more profound effect on the clinical course, 8 and evidence suggesting that patients with non-Gaucher's-related variants fall in between carriers of mild mutations and noncarriers. 9,16 Similarities with DLB are thought to be most apparent with severe mutations, 1 but here were also seen in patterns of cortical hypoactivity. We demonstrate for the first time that even in the absence of a significantly more severe clinical syndrome, subclinical findings are present in "mild" GBA variants. To the best of our knowledge, this is the most in-depth description to date of clinical and biological phenotypes of patients with Parkinson's disease carrying the GBA variants p.E365K and p.T408M. Metabolomic changes have to be validated, and longitudinal studies are needed to investigate whether the observed neuroimaging changes progress or if they are followed by significantly more severe motor symptoms, clinical dementia, and psychotic symptoms. Similar approaches could be applied to other genetic variants associated with Parkinson's disease, and the observed alterations could be promising to monitor effects of targeted diseasemodifying treatments.