CSF Protein Level of Neurotransmitter Secretion, Synaptic Plasticity, and Autophagy in PD and DLB

Molecular pathways associated with α‐synuclein proteostasis have been detected in genetic studies and in cell models and include autophagy, ubiquitin‐proteasome system, mitochondrial homeostasis, and synaptic plasticity. However, we lack biomarkers that are representative for these pathways in human biofluids.

In the large majority of patients with α-synucleinopathies such as Parkinson's disease (PD) and dementia with Lewy bodies (DLB), the common pathologic feature is the aggregation of abnormal conformations of the protein α-synuclein. 1 Subsequently, several molecular pathways associated with α-synuclein proteostasis have been detected in genetic studies as well as in cell and animal models. These include impairment in autophagy with, for example, endolysosomal dysfunction, ubiquitin-proteasome system, mitochondrial homeostasis, and synaptic plasticity. 2 Despite this clear evidence for pathways involved, we lack validated biomarkers that are clearly disease-and/or pathway-specific in human biofluids. A recent study performed a targeted approach and explored cerebrospinal fluid (CSF) peptide/protein profiles associated with autophagic endolysosomal and ubiquitin-proteasome function using solid-phase extraction combined with parallel reaction monitoring mass spectrometry in 3 cohorts of PD patients, Alzheimer's disease (AD) patients, and healthy controls. Interestingly, CSF protein levels of lysosomalassociated membrane glycoprotein 1 and 2 (LAMP1 and LAMP2), cathepsin F (CTSF), ganglioside GM2 activator (GM2A), AP-2-complex subunit beta (AP2B1), and ubiquitin were lower in PD patients than in AD patients, prodromal AD patients, and healthy controls. 3 With advancing techniques, a recent unbiased quantitative proteomics approach using liquid chromatography-tandem mass spectrometry in data-independent acquisition mode reported decreased CSF levels of proteins associated with neurotransmitter secretion (granin family) as well as with adherens junctions in sporadic PD patients compared with healthy controls. 4 Specifically, CSF protein levels of neurosecretory protein VGF, secretogranin 2 (SCG2), secretogranin 3, and secretogranin 5, chromogranin-A (CHGA) and chromogranin-B, and cadherins were lower in PD. Following this line of evidence, we aimed to evaluate CSF protein profiles associated with pathways of α-synuclein proteostasis in 2 large monocentric cohorts with α-synucleinopathy including a total of 385 PD patients and 67 DLB patients compared with healthy controls. As impaired lysosomal chaperone-mediated autophagy represents a key mechanism in PD, we specifically included PD patients (80 of 385) and DLB patients (17 of 67) with variants in the glucocerebrosidase (GBA) gene. Given the accelerated α-synuclein pathology and pronounced clinical trajectories in DLB compared with PD and in patients with GBA variants compared with the wild-type condition, our primary hypothesis was that profiles in those patient groups would be more distinct.

Participants
Between 2005 and 2018, PD patients and DLB patients were recruited from the outpatient clinic and/or ward for Parkinson's disease at the University Hospital of Tuebingen for scientific lumbar puncture. In total, CSF samples from 305 sporadic PD patients without the GBA variant (PD GBA-wildtype ), 80 PD patients with the GBA variant (PD GBA ), 50 DLB patients without the GBA variant (DLB GBA-wildtype ), and 17 DLB patients with the GBA variant (DLB GBA ) were available. Clinical data and CSF levels of total α-synuclein of 210 PD patients with a median follow-up time of 4 years were available for longitudinal analysis. Spouses of patients and volunteers were clinically assessed to have no neurodegenerative disease and served as control participants (HCs, n = 31).

Genetic Analysis
Genetic screening of all exons of the GBA gene was performed by Sanger sequencing.
We performed GBA subgroup classification for variant severity according to established variant risks reported for PD. Of note, some variants that have been reported as nonrelevant for Gaucher disease have been proven to increase the risk for PD and therefore have been included in our analysis, for example, p.E326K and p.T369M. GBA variants were categorized as: GBA _low_risk , E326K, T297S, T369M; GBA _mild , N370S; GBA _severe , c1265_1319del55bp, D409H, Ex3 + 4 recombination, IVS2 + 1 g > a, L444P, P182L, R120Q, R359X, W184R.

Clinical Investigations
All participants were examined by a neurologist who specialized in movement disorders. Diagnosis of PD was defined according to the UK Brain Bank Society criteria. 5 Diagnosis of DLB was made according to the revised consensus criteria for DLB developed by the third report of the DLB consortium. 6 PD and DLB patients were assessed in the dopaminergic ON state, and levodopa-equivalent daily dose (LEDD) was calculated. We assessed severity of motor symptoms using part III of the Unified Parkinson's disease Rating Scale (UPDRS-III), from 2006 to 2008 the old version and from 2009 on the MDS-UPDRS. 7 Disease stage was categorized by the modified Hoehn and Yahr Scale (H&Y). 8 Cognitive function was tested using the Montreal Cognitive Assessment (MoCA) 9 and/or the Mini-Mental Status Examination (MMSE). 10 Because the MoCA was available only from 2009 on, all previously obtained MMSE scores were converted to MoCAequivalent scores. 11 Mood disturbances were assessed using Beck's Depression Inventory II (BDI-II). 12

Collection of CSF
Spinal tap was performed between 9:00 AM and 1:00 PM. Samples were directly taken from the bedside and centrifuged within 60 minutes and frozen at À80 C within 90 minutes. Samples with abnormal routine CSF diagnostics (white blood cell count > 6 cells/μL, immunoglobulin subtype G index > 0.7, erythrocytes > 1/μL) were excluded.

CSF Measurements of Proteins Related to
α-Synuclein Proteostasis

Sample Preparation
Measurement of protein concentrations was performed as previously described. 3,13 Briefly, 50 μL of CSF was mixed with 50 μL of an internal standard mixture containing stable isotope-labeled peptides (JPT Peptide Technologies GmbH, Germany; Thermo Fisher Scientific Inc., Waltham, MA), 13C-labeled ubiquitin (Silantes, GmbH, Germany), and bovine serum albumin (Sigma-Aldrich Co., St. Louis, MO) diluted in 50 mM NH 4 HCO 3 . Reduction and alkylation were performed by the addition of 50 μL of 15 mM 1,4-dithiothreitol in 50 mM NH 4 HCO 3 , shaking for 30 minutes at 60 C, cooling down at room temperature for 30 minutes, and the addition of 25 μL of 70 mM iodoacetamide in 50 mM NH 4 HCO 3 , followed by shaking at room temperature in the dark for 30 minutes. Samples were digested by the addition of 25 μL of 0.08 μg/μL sequencing-grade modified trypsin (Promega Co., Madison, WI) diluted in 50 mM NH 4 HCO 3 and incubated at 37 C shaking for 18 hours. Digestion was ended by the addition of 25 μL of 10% trifluoroacetic acid. Solid-phase extraction was performed using Oasis HLB 96-well μElution Plates (2 mg sorbent and 30-μm particle size; Waters Co., Milford, MA) by conditioning (2 Â 300 μL methanol), equilibration (2 Â 300 μL H 2 O), loading of samples, washing (2 Â 300 μL H 2 O), and elution (2 Â 100 μL methanol). Samples were dried by vacuum centrifugation and stored at À80 C.

Parallel Reaction Monitoring Mass Spectrometry
Samples were dissolved by the addition of 50 μL of 50 mM NH 4 HCO 3 and shaken at room temperature for 1 hour. A total of 40 μL of sample was injected and separated using a Dionex UltiMate 3000 standard-LC system (Thermo Fisher Scientific Inc., Waltham, MA) and a Kinetex EVO C18 column (length, 150 mm; inner diameter, 2.1 mm; particle size, 1.7 μm; Phenomenex Inc., Torrance, CA) with a SecurityGuard ULTRA cartridge prefilter (Phenomenex Inc. using a flow rate of 300 μL/min, the gradient went from 3% to 5% B over 1 minute, followed by 5% to 26% B over 48 minutes. The column temperature was set to 50 C. Separation by high-performance liquid chromatography was performed in online mode coupled to a Q Exactive Hybrid Quadrupole-Orbitrap mass spectrometer (Thermo Fisher Scientific Inc.). Using a HESI-II ionization probe (Thermo Fisher Scientific Inc.), electrospray ionization was performed in positive ion mode with the following settings: spray voltage, +4.1 kV; heater temperature, 400 C; capillary transfer tube temperature, 380 C; sheath gas flow rate, 25; auxiliary gas flow rate, 10; and S-Lens RF level, 60. Acquisition of data was performed using single microscans in parallel reaction monitoring (PRM) mode with an isolation window of m/z 2 centerd on the second isotope of the precursor ion. The resolution setting was 70 k, with an AGC target of 1 Â 106 and a 256-millisecond injection time. Fragmentation was performed using higher-energy collision-induced dissociation.

Data Extraction
Skyline v.19.1 was used to calculate and export fragment ion peak areas and to monitor fragment ion traces and ratios. The ratio between tryptic peptide and isotope-labeled peptide peak area was used for quantification.

Statistics
Statistical analysis was performed using IBM SPSS 26.0 software. Dichotomous data were analyzed using the Fisher's exact test or the chi-square test. Homogeneity of variance of the variables was evaluated using Levene's test, which showed P > 0.05. Intergroup comparisons of demographics, clinical markers, and CSF protein levels were calculated using analysis of variance/analysis of covariance (ANOVA/ANCOVA) including sex, age, and disease duration as covariates as appropriate. Receiver operating characteristic (ROC) curves and area under the curve (AUC) were calculated to evaluate discriminant capabilities of CSF proteins between patient groups and controls. Pearson's correlation was used to evaluate associations between CSF protein levels.
In PD patients, for each CSF protein associated with pathways related to α-synuclein proteostasis, we assessed the optimal cutoff for pathological levels by calculating Youden's index (Table S3). Values greater than the optimal diagnostic cutoff were considered normal (PD normal_proteostasis_profile ), whereas those below indicated pathologically reduced protein (PD pathological_proteostasis_profile ).
Longitudinal analysis of CSF levels of total α-synuclein and clinical trajectories between PD pathological_proteostasis_profile versus PD normal_proteostasis_profile was done using a linear mixed model with the fixed factors group (PD pathological_proteostasis_profile , PD normal_proteostasis_profile ) and time (time of follow-up in years), their interaction, and the random variable subject, modeled by random intercepts. The addition of random slopes did not improve model fit. CSF levels of total α-synuclein, MoCA, UPDRS-III, H&Y stage, BDI-II were defined as dependent variables. We analyzed the fixed effect of group, time, and the interaction of group and time on the dependent variable.

Standard Protocol Approval, Registrations, and Patient Consents
The study was approved by the Ethics Committee of the University of Tuebingen (26/2007BO1, 404/ 2010BO1, 199/2011BO1, 702/2013BO1). All participants gave written informed consent.

Demographics and Clinical Characteristics
PD GBA was younger in mean age (64 vs 66 years; P = 0.034) and in mean age at onset (56 vs 60 years; P = 0.002) and had a longer mean disease duration (8 vs 7 years; P = 0.018) compared with PD GBA_wildtype . UPDRS-III, H&Y staging, MoCA, BDI-II score, and LEDD were not significantly different between PD GBA and PD GBA_wildtype . Demographics and clinical variables did not differ significantly between DLB GBA_wildtype and DLB GBA . All PD and DLB groups were older than healthy control participants (P ≤ 0.001; Table 1).
Within the group of PD patients, PD GBA_severe patients had lower CSF levels of SCG2 compared with PD GBA_wildtype patinets (0.80 vs 1.02) and lower CSF levels of VGF compared with PD GBA_low_risk and PD GBA_wildtype patients (0.70 vs 1.05 vs 1.04, respectively; P < 0.05; Table S1).

Synaptic Plasticity
CSF levels of NPTX1 showed a trend toward being lower in PD GBA (P = 0.087) and were significantly lower in DLB GBA_wildtype and DLB GBA compared with healthy controls (P ≤ 0.001; Table 1).
Stratification according to GBA variant severity revealed no significant differences between the PD and DLB groups (Tables S1 and S2).
Stratification according to GBA variant severity revealed no significant differences between the PD groups (Table S1).
In addition to the abovementioned proteins, the lysosomal protein ubiquitin discriminated between PD GBA patients and healthy controls (AUC, 0.657), whereas the lysosomal protein LAMP2 discriminated between DLB GBA patients and healthy controls (AUC, 0.735). LAMP2 also discriminated between DLB GBA and DLB GBA_wildtype (AUC, 0.695); see Figure 1.
Adding α-synuclein as a covariable to the model did not improve the AUC of the proteins in any of the models.

Associations of CSF Proteostasis Protein Levels with Demographics and Clinical Scales
Male sex and higher age were weakly associated with lower CSF levels of AP2B1, CHGA, SCG2, VGF, and NPTX1 (P < 0.05). Higher H&Y staging and lower MoCA were associated very weakly with lower CSF levels of SCG2, VGF, and NPTX1 (P < 0.05). LEDD was not associated with any of the proteins.

PD Patients with a Pathological CSF Protein Profile Related to α-Synuclein Proteostasis Versus Those with a Normal Profile
To discriminate between PD patients with a pathologically low CSF proteostasis profile (PD pathological_proteostasis_profile ) from those patients with a profile similar to healthy controls (PD normal_proteostasis_profile ), Youden's Index was calculated for each CSF protein related to α-synuclein proteostasis (Table S3). We detected 101 PD patients with a pathological CSF protein level below the cutoff in all 6 proteins, whereas 284 showed a normal profile. PD pathological_proteostasis_profile patients were younger in mean age (62 vs 67 years; P ≤ 0.001) and in mean age at onset (56 vs 60 years; P ≤ 0.001) compared with PD normal_proteostasis_profile patients. Also, they were more often male (78% vs 62%; P = 0.002), had higher mean H&Y scores (2.2 vs. 2.1; P = 0.012), and lower mean CSF level of total α-synuclein (380 vs. 690 pg/ mL; P ≤ 0.001). Mean UPDRS-III, MoCA, and BDI-II scores were not significantly different between PD pathological_proteostasis_profile and PD normal_proteostasis_profile ( Table 2).

Longitudinal Analysis in PD Patients with a Pathological CSF Protein Profile Related to α-Synuclein Proteostasis Versus Those with a Normal Profile
Linear mixed-model analysis revealed that CSF levels of total α-synuclein were influenced by the fixed factor group (PD pathological_proteostasis_profile vs PD normal_proteostasis_profile ) but not by the factor time and the interaction group/ time. The predicted mean was 397 pg/mL in PD pathological_proteostasis_profile vs. 662 pg/mL in PD normal_proteostasis_profile ; P ≤ 0.001.
H&Y stage and MoCA score were influenced by the factor time. For both groups the yearly increase in H&Y was 0.05 points and the yearly decrease in MoCA score was 0.2 points. No effects of group, time, or group/time interaction were seen for UPDRS-III and BDI-II.

Discussion
By assessing CSF protein levels representative for major pathways associated with α-synuclein proteostasis, we show: (1) Proteins associated with neurotransmitter secretion (CHGA, SCG2, VGF), synaptic plasticity (NPTX1), and endolysosomal autophagy (AP2B1, GM2A, LAMP2, ubiquitin) were lower in PD patients and/or DLB patients compared with healthy controls. (2) These patterns were more pronounced in DLB than in PD and were accentuated by GBA variant status in both entities, PD and DLB.
(3) In all analyzed cohorts, CSF levels of these proteostasis-related proteins were positively associated with CSF levels of total α-synuclein, meaning that lower levels of these proteins were associated with lower levels of total α-synuclein. (4) These findings could be confirmed longitudinally. PD patients with low CSF profiles of proteins related to proteostasis showed lower CSF levels of α-synuclein longitudinally compared with PD patients with a normal proteostasis profile. (5) Lower CSF levels of these proteostasis-related proteins were associated with lower MoCA scores and with higher H&Y staging linking biochemical profiles with clinical outcome parameters.
Decreased CSF levels of the same proteins associated with neurotransmitter secretion (CHGA, SCG2, VGF) as found in our study have recently been identified in an unbiased quantitative proteomics approach in PD patients compared with controls. 4 Results were consistent across 3 US cohorts, one of them longitudinally. These proteins play an essential role in the regulated secretory pathway, which controls the delivery of neurotransmitters and biologically active peptides. They are co-stored in dense-core vesicles and regulate levels of catecholamines such as dopamine. Secretoneurin, a SCG2-derived peptide stimulates the release of dopamine from nigrostriatal neurons. 14,15 A small study reported CSF levels of CHGA to be reduced in drug-naive PD patients compared with healthy controls. As we did not find an association between LEDD and CSF level of CHGA, these findings suggest that this pattern is already present at diagnosis and not driven by dopaminergic treatment. 16 Interestingly, postmortem findings show that CHGA accumulates in cortical Lewy bodies in a developmental stage-related manner. 17,18 Using an unbiased proteomics approach in 2 cohorts of patients with DLB, it was recently shown that CSF levels of SCG2 and VGF are also decreased in DLB compared with age-matched control participants. 19 We can confirm this finding. As VGF is associated with learning and memory, the same group investigated CSF levels of VGF not only in DLB but also in AD. They found lower CSF levels of VGF in patients with DLB Although CSF levels of VGF were positively associated with CSF levels of α-synuclein, no association was found with CSF levels of Abeta 1-42 . In DLB patients, lower levels of VGF were related to more advanced cognitive impairment. 20 VGF is a secreted neuronal precursor protein, and its expression is induced by brainderived neurotrophic factor, a protein that has been repeatedly shown to be decreased in α-synucleinopathies and has been linked to dopaminergic neuronal survival. 21 VGF is processed into several bioactive peptides that are implicated in energy homeostasis, synaptic biogenesis and synaptic plasticity. 14 Impaired synapse plasticity is emerging as an important early event in the development of α-synucleinopathies and found already in the period of brain development, as shown in a LRRK2 p.G2019S mouse model. 22 Neuronal pentraxin-1 (NPTX1) belongs to a family of proteins that are involved in synaptic plasticity. Using wholegenome expression profiling, it was shown that NPTX2 is colocalized with α-synuclein aggregates in the substantia nigra and cerebral cortex in PD. 23 Decreased CSF levels of NPTX1 and NPTX2 have also been shown in patients with frontotemporal dementia and even in presymptomatic mutation carriers (GRN, C9orf72, or MAPT). 24 With the identification of variants in the GBA gene as major genetic risk factor for PD and DLB, it became clear that impairment in lysosomal-mediated autophagy represents a key event in the pathogenesis of α-synucleinopathies. 25,26 We think that the ROC-based differentiation between DLB GBA and DLB wildtype by CSF LAMP2 protein profiles further supports the importance of lysosomal dysfunction based on GBA mutations. We now seem to make progress in the field of pathway-specific biofluid markers in patient cohorts. This was further underpinned by the finding that genetic variants in other genes associated with lysosomal storage disorders also increase the risk of developing PD. 27 A recent study also found AP2B1, GM2A, and LAMP1 and LAMP2 to be decreased in PD compared with prodromal AD, AD, and healthy controls. 3 Importantly, lower CSF levels of these proteostasisrelated proteins (neurotransmitter secretion, synaptic plasticity, and endolysosomal autophagy) were associated with lower CSF levels of total α-synuclein crosssectionally and longitudinally. Previous analyses in PD and DLB demonstrated CSF levels of total α-synuclein to be decreased in both diseases compared with healthy controls with even more pronounced decrease in patients with GBA variants. [28][29][30][31] This mechanistic link between proteostasis-associated proteins and α-synuclein as pathologic hallmark was previously shown in postmortem studies and cell models 32-37 but seems now also detectable in vivo in patient-derived CSF. It is of importance that results and hypothesis coming from cell and animal models can be translated into patient-derived biofluid-marker studies. This field is just beginning to emerge, and we often failed to translate findings/hypothesis into patient cohort. One might imagine that composite profiles of several biofluid markers that are representative of different pathways upstream or downstream of α-synuclein might help to stratify patients for future pathway-specific trials. Of note, the pattern of reduced CSF protein profiles was more pronounced in DLB than in PD and accentuated by GBA variant status in both entities. This highlights the currently widely discussed pathomechanistic continuum between PD and DLB 38 with additional acceleration because of GBA variants. It will be interesting to evaluate these protein profiles in patients with biallelic variants in recessively inherited PD genes such as parkin and PINK1, as these patients often show nigral degeneration without the typical α-synuclein aggregation.
The association of these proteostasis-related proteins with main clinical outcomes such as disease staging by H&Y and cognitive function measured by MoCA warrants further investigation. It will be valuable to assess these proteins in prospective longitudinal cohorts in which CSF along with more complex clinical data and longitudinal clinical trajectories is available from de novo PD patients with several years of follow-up. Such analysis will help to evaluate whether these proteins might serve as biological progression markers that relate to clinical outcomes.
Strengths of our study are: (1) the large monocentric standardized collection of high-quality CSF samples minimizing variance in sample collection and processing; and (2) the validation of findings in longitudinal PD samples and in another α-synucleinopathy, namely, DLB, which represents a histopathological continuum with even more pronounced profiles.
The study had these limitations. (1) As our cohort comprised early-to mid-stage patients, we cannot conclude whether findings are causation or consequence. Therefore, we suggest to explore these findings in de novo and, ideally, prodromal cohorts.
(2) The investigated proteins may not be fully representative of the involved pathways that have been discussed. (3) The number of healthy controls is smaller than the number of PD and DLB patients. (4) We lack a direct validation in brain samples from the same patients from whom these CSF profiles were assessed. (5) It is technically possible to measure at least some of the pathway-specific proteins in blood samples, which would be less invasive. However, we do not know whether blood levels of these proteins reflect the CNS milieu. Therefore, it would be important to analyze CSF/blood pairs in future studies to answer this question.
Taken together, our findings suggest that proteins associated with neurotransmitter secretion, synaptic plasticity, and endolysosomal autophagy might serve as biomarkers related to α-synuclein proteostasis in PD and DLB.

Data Availability Statement
Anonymized data are available upon request (kathrin.brockmann@uni-tuebingen.de)