Longitudinal correlation of cerebrospinal fluid GFAP and the progression of cognition decline in different clinical subtypes of Parkinson's disease Authors: Yumei Liu, Jing Wang, Fangbo Ning, Guojun Wang, Anmu Xie Published: 2024 ABSTRACT AbstractGlial fibrillary acidic protein (GFAP) is an intermediate filament protein expressed mainly in astrocytes of the central nervous system (CNS), a potential biomarker of cognitive decline in Parkinson's disease (PD). The central motor subtypes of PD include tremor‐dominant (TD), postural instability and gait disorder (PIGD), and indeterminate subtypes, whose different course of disease requires the development of biomarkers that can predict progression based on motor subtypes. In this study, we aimed to assess the predictive value of cerebrospinal fluid (CSF) GFAP for PD motor subtypes in PD. Two hundred and sixteen PD patients were recruited in our study from the progression markers initiative. Patients were subgrouped into TD, PIGD, and indeterminate subtypes. Longitudinal relationships between baseline CSF GFAP and cognitive function and CSF biomarkers were assessed using linear mixed‐effects models. Cox regression was used to detect cognitive progression in TD patients. The baseline and longitudinal increases in CSF GFAP were associated with a greater decline in episodic memory, CSF α‐syn, and a greater increase of CSF NfL in TD and TD‐male subtypes. Cox regression showed that higher baseline CSF GFAP levels were corrected with a higher risk of developing mild cognitive impairment (MCI) over a 4‐year period in the PD with normal cognition (NC) group (adjusted HR = 1.607, 95% CI 1.907–2.354, p = 0.01). CSF GFAP might be a promising predictor of cognition decline in TD. Study Highlights WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC? Glial fibrillary acidic protein is an intermediate filament protein expressed mainly in astrocytes of the central nervous system, a potential biomarker of cognitive decline in Parkinson's disease. The relationships between parkinsonian phenotype, cognition function, and CSF GFAP are seldom researched. WHAT QUESTION DID THIS STUDY ADDRESS? (1) This study investigated the relationships between CSF GFAP levels and cognition change through longitudinal analyses in different motor subtypes for the first time. (2) This article analyzed the relationship between GFAP and cognition change in male and female PD patients separately. WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE? The baseline and longitudinal increases in CSF GFAP were associated with a greater decline in episodic memory, CSF α‐syn and a greater increase of CSF NfL in TD and TD‐male subtypes. Cox regression showed that higher baseline CSF GFAP levels were associated with a higher risk of developing mild cognitive impairment (MCI) over a 4‐year period in the PD with normal cognition (NC) group. HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE? Our results demonstrated that CSF GFAP levels are a promising predictive biomarker to monitor cognition progression in TD patients. In addition, only in TD‐male patients, CSF GFAP levels predict cognition decline. These findings helped us realize that the value of GFAP in predicting cognitive decline is different in different clinical subtypes of PD, providing new ideas for subsequent exploration of PD biomarkers. At the same time, we also found gender differences in the prediction of cognitive decline by GFAP, which may provide new ideas for exploring the pathogenic mechanisms of GFAP. Study Highlights WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC? WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC? WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC? Glial fibrillary acidic protein is an intermediate filament protein expressed mainly in astrocytes of the central nervous system, a potential biomarker of cognitive decline in Parkinson's disease. The relationships between parkinsonian phenotype, cognition function, and CSF GFAP are seldom researched. WHAT QUESTION DID THIS STUDY ADDRESS? WHAT QUESTION DID THIS STUDY ADDRESS? WHAT QUESTION DID THIS STUDY ADDRESS? (1) This study investigated the relationships between CSF GFAP levels and cognition change through longitudinal analyses in different motor subtypes for the first time. (2) This article analyzed the relationship between GFAP and cognition change in male and female PD patients separately. WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE? WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE? WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE? The baseline and longitudinal increases in CSF GFAP were associated with a greater decline in episodic memory, CSF α‐syn and a greater increase of CSF NfL in TD and TD‐male subtypes. Cox regression showed that higher baseline CSF GFAP levels were associated with a higher risk of developing mild cognitive impairment (MCI) over a 4‐year period in the PD with normal cognition (NC) group. HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE? HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE? HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE? Our results demonstrated that CSF GFAP levels are a promising predictive biomarker to monitor cognition progression in TD patients. In addition, only in TD‐male patients, CSF GFAP levels predict cognition decline. These findings helped us realize that the value of GFAP in predicting cognitive decline is different in different clinical subtypes of PD, providing new ideas for subsequent exploration of PD biomarkers. At the same time, we also found gender differences in the prediction of cognitive decline by GFAP, which may provide new ideas for exploring the pathogenic mechanisms of GFAP. INTRODUCTION Parkinson's disease (PD) is the second most prevalent neurodegenerative disorder, apart from Alzheimer's disease (AD), resulting from the progressive loss of dopamine neurons and α‐synuclein (α‐syn) aggregation. Cognitive dysfunction is one of the most common non‐motor symptoms of PD. It is reported that 30%–40% of patients with PD experience cognitive impairment and may develop dementia (PD–D). 1 Emerging evidence demonstrates that activation of neuroinflammation, such as reactive astrocyte proliferation, could contribute to development of PD. 2 , 3 Astrocytes could efficiently degrade fibrillar α‐syn by lysosomal degradation, which reduces the accumulation of α‐syn in neurons. 4 However, the α‐syn could lead to an imbalance or dysfunction of proteins within the astrocyte, ultimately destroying the function of the astrocyte. 5 Glial fibrillary acidic protein (GFAP) is an intermediate filament protein expressed mainly in astrocytes of the central nervous system (CNS). As a result of damage of the neuron, the cellular structure is destroyed, and GFAP is released in the cerebrospinal fluid (CSF) and then enters the bloodstream. 6 The expression of GFAP was significantly increased in neurodegenerative diseases, including AD, PD, and frontotemporal dementia (FTD). 7 , 8 , 9 , 10 , 11 In recent years, GFAP has received attention as a promising biomarker of cognition impairment. In AD patients, many studies showed that the levels of plasma GFAP were related to amyloid pathology and could predict conversion to AD dementia. 12 , 13 , 14 Bartl et al. found a possible association between baseline CSF GFAP level and Montreal Cognitive Assessment (MoCA) score in newly diagnosed untreated PD patients. 15 In PD patients, the level of serum GFAP is higher than in healthy control (HC) individuals, and plasma GFAP may have prognostic value in predicting the conversion of mild cognitive impairment (MCI) to dementia. 16 , 17 Tang and colleagues also demonstrated that there was a negative correlation between baseline plasma GFAP and Mini‐Mental State Examination scores and may serve as a predictor of cognitive impairment in PD patients. 18 Clinically, PD patients are primarily classified into tremor‐dominant (TD) type, indeterminate, and postural instability and gait disorder (PIGD) type based on clinical features of PD. Compared with patients with the TD subtype, the PIGD phenotype has also been related to worse cognitive performance in PD. 19 , 20 In the TD type, the neutrophil–lymphocyte ratio, a marker of inflammation, has positively correlated with motor severity. 21 However, the relationships between parkinsonian phenotype, cognition function, and CSF GFAP are seldom researched. The heterogeneous clinical phenotypes and varying disease course seen in PD patients highlight the need for biomarkers. The Parkinson's Progression Markers Initiative (PPMI) repository is a collaborative effort involving multiple centers and contains comprehensive longitudinal clinical, imaging, and biological data from patients with PD. The aim of our study was to describe the association between CSF GFAP and cognitive decline across various domains during the progression of PD in each subtype using PPMI database, which could further elucidate the potential pathogenesis underlying cognitive decline in PD via CSF GFAP. MATERIALS AND METHODS Participants The analyzed data used in this study were collected from the PPMI. 22 PD participants enrolled must be ≥30 years old at baseline, be untreated for PD, have no dementia (cognitive assessment (MoCA) total score ≥ 22)) at baseline, diagnosed with PD within 2 years, exhibit dopamine transporter (DAT) deficits on baseline I‐123 FP‐CIT SPECT images, Hoehn and Yahr (H&Y) stage I or II at the beginning of the study. Patients were classified into TD (ratio ≧ 1.15), indeterminate (ratios >0.90 and <1.15), and PIGD subtypes (ratio ≦ 0.90) based on the ratio of the mean tremor scores to the mean PIGD scores from the baseline MDS‐UPDRS scores. 23 In the longitudinal study, all enrolled participants had at least one additional CSF GFAP during the 5‐year follow‐up. Eleven TD, four PIGD, and two indeterminate patients were excluded due to no follow‐up GFAP data. Measurements of CSF biomarkers Roche NTK was used to quantify CSF GFAP, neurofilament light (NfL), and α‐syn on a cobas e411 analyzer at Covance Greenfield laboratories. 15 , 24 Automated cobas e601 analyzer was used to measure the levels of CSF amyloid‐β42 (Aβ42), total tau (t‐tau), and phosphorylated tau (p‐tau) by electrochemiluminescence (ECL) method. 25 Cognitive assessments Cognitive assessment included global cognitive condition (MoCA total score), processing speed/attention (Symbol Digit Modality Test, SDMT), visuospatial abilities (Benton Judgment of Line Orientation, JoLO), executive function/working memory (Letter Number Sequencing, LNS), language (semantic fluency test), and episodic memory (Hopkins Verbal Learning Test, HVLT) with total recall, delayed recall, retention, and recognition discrimination. Cognitive status was defined based on the criteria used in previous studies: normal cognition (NC, MoCA > 26), MCI (22 ≤ MoCA ≤ 26) and dementia (MoCA < 22). 26 Statistical analyses CSF GFAP levels were not normally distributed in the Kolmogorov–Smirnov test, they were log10‐transformed to obtain a normal distribution. Outliers above or below three standard deviations were excluded. The three motor subtypes (TD vs. PIGD vs. Indeterminate) were compared for similar parameters by analysis of variance (ANOVA) and Kruskal–Wallis test for continuous and categorical variables, respectively. The prediction of cognitive changes using the baseline or longitudinal change in CSF GFAP was measured by the linear mixed‐effects model. The models included random intercepts and slopes for time and an unstructured covariance matrix for random effects, which was regarded as a predictor of the interactions between time and dependent variables. Kaplan–Meier survival curves were used to compare the cumulative incidence of cognitive progression among different groups during the follow‐up. The association between baseline CSF GFAP levels and the occurrence of cognitive progression during follow‐up was analyzed using Cox regression model. All statistical analyses were performed using the software R 4.1.2. All regression analyses were corrected for age, gender, educational level, age‐onset, race, and APOE_e4 carrier status. ApoE genotyping was determined with the use of allele specific oligonucleotide probes labeled with fluorogenic reporter dyes (TaqMan method). The flowchart of the main analysis is shown in Figure 1. Flowchart of data analysis. CSF, cerebrospinal fluid; GFAP, glial fibrillary acidic protein; PIGD, Postural Instability Gait Disorder; PPMI, Parkinson's Progression Markers Initiative; TD, tremor‐dominant. RESULTS Study participant characteristics The baseline characteristics of the participants are shown in Table 1. Of the 216 PD patients, 152 patients were subgrouped to the TD subtype (70%), 40 patients to the PIGD subtype (19%), and 24 to the indeterminate subtype (11%). There was no significant difference in age, sex, or educational years among the three groups. TABLE 1Characteristics of participants in PPMI.VariablesTD subtypePIGD subtypeIndeterminate subtype p‐Value n = 152 n = 40 n = 24 Demographic characteristics Sex (n)Male10123190.207Female51175Age61.59 ± 9.5260.34 ± 9.7558.21 ± 11.020.41Educate years16.28 ± 2.5615.95 ± 2.9115.04 ± 2.240.091MoCA26.96 ± 2.3826.95 ± 1.7827.46 ± 1.960.471JoLO13.07 ± 1.8913.1 ± 1.8213.83 ± 1.040.172HVLT immediate recall24.58 ± 4.9125.1 ± 4.2425.88 ± 4.310.354HVLT delayed recall8.48 ± 2.58.7 ± 2.678.79 ± 2.260.845HVLT retention0.86 ± 0.20.88 ± 0.190.84 ± 0.180.742HVLT discrimination9.33 ± 3.2310.45 ± 1.3810.13 ± 2.150.27HVLT delayed recognition11.19 ± 1.1411.35 ± 0.9211.42 ± 0.650.775LNS10.59 ± 2.6810.68 ± 2.4411.13 ± 2.030.44Semantic fluency total score49.3 ± 12.1753.63 ± 12.0152.17 ± 9.870.033SDMT41.48 ± 9.2243.38 ± 8.9744.29 ± 7.310.246CSF biomarkersAβ1‐42, pg/mL908.14 ± 422.91823.03 ± 319.1901.08 ± 362.850.628a‐synuclein, pg/mL1551.27 ± 749.471415.49 ± 652.031337.89 ± 622.70.255tau, pg/mL170.07 ± 58.09160.36 ± 51.99156.83 ± 42.790.516p‐tau, pg/mL15.19 ± 5.314.12 ± 5.1213.1 ± 3.930.169Abbreviations: Aβ1‐42, amyloid β1‐42; HVLT, Hopkins Verbal Learning Test; JoLO, Benton Judgment of Line Orientation Score; LNS, Letter Number Sequencing; MCA, Montreal cognitive assessment; PIGD, Postural Instability Gait Disorder; PPMI, Parkinson's Progression Markers Initiative; p‐tau, phosphorylated tau; SDMT, Symbol Digit Modality Test; TD, tremor‐dominant. Characteristics of participants in PPMI. Abbreviations: Aβ1‐42, amyloid β1‐42; HVLT, Hopkins Verbal Learning Test; JoLO, Benton Judgment of Line Orientation Score; LNS, Letter Number Sequencing; MCA, Montreal cognitive assessment; PIGD, Postural Instability Gait Disorder; PPMI, Parkinson's Progression Markers Initiative; p‐tau, phosphorylated tau; SDMT, Symbol Digit Modality Test; TD, tremor‐dominant. Relationships of baseline CSF GFAP levels and cognitive change The relationships between baseline CSF GFAP levels and cognitive change are shown in Figure 2 and Table S1. In TD patients, higher baseline CSF GFAP levels were related to a faster decline in episodic memory (HVLT Delayed Recall, β = −0.13, p = 0.008; HVLT Retention, β = −0.158, p = 0.009; HVLT discrimination, β = −0.139, p = 0.033) and α‐syn (β = −0.099, p = 0.043). In addition, the results for TD‐male patients with higher baseline CSF GFAP levels were also related to a faster decline in episodic memory (HVLT Delayed Recall, β = −0.137, p = 0.028; HVLT Retention, β = −0.164, p = 0.025; HVLT discrimination, β = −0.19, p = 0.024) (Table S2). Associations of baseline GFAP levels on cognition measurements during follow‐up in different motor subtypes of PD. Aβ1‐42, amyloid β1‐42; GFAP, glial fibrillary acidic protein; HVLT, Hopkins Verbal Learning Test; JoLO, Benton Judgment of Line Orientation Score; LNS, Letter Number Sequencing; MoCA, Montreal cognitive assessment; NfL, neurofilament light; PIGD, Postural Instability Gait Disorder; p‐tau, phosphorylated tau; SDMT, Symbol Digit Modality Test; TD, tremor‐dominant. TD‐NC patients with higher CSF GFAP levels were related to a faster decrease in α‐syn (β = −0.173, p = 0.017). We also found that TD‐MCI patients with higher baseline CSF GFAP levels were also associated with faster decline in episodic memory (HVLT Delayed Recall, β = −0.253, p = 0.012; HVLT Retention, β = −0.354, p = 0.001) and total scores of Semantic fluency (β = −0.216, p = 0.008) (Table S3). The results of a Kaplan–Meier analysis and the log‐rank test indicated that the higher baseline CSF GFAP levels had a higher risk of cognitive transition to MCI in the PD–NC group (p = 0.01). No significant predictive effect of CSF GFAP on the risk of dementia was found (Figure 3). Effects of longitudinal changes in CSF GFAP levels on cognition measurements during follow‐up in different motor subtypes of PD. Aβ1‐42, amyloid β1‐42; GFAP, glial fibrillary acidic protein; HVLT, Hopkins Verbal Learning Test; JoLO, Benton Judgment of Line Orientation Score; LNS, Letter Number Sequencing; MoCA, Montreal cognitive assessment; NfL, neurofilament light; PIGD, Postural Instability Gait Disorder; p‐tau, phosphorylated tau; SDMT, Symbol Digit Modality Test; TD, tremor‐dominant. Associations between longitudinal CSF GFAP level and longitudinal changes in CSF biomarkers and cognitive function The relationships between longitudinal changes in CSF GFAP levels and cognition changes are shown in Figure 4 and Table S4. In the longitudinal analysis, TD patients with a higher increasing rate of CSF GFAP were associated with the faster worsening in episodic memory (HVLT immediate recall, β = −0.096, p = 0.04; HVLT Delayed Recall, β = −0.136, p = 0.001; HVLT retention, β = −0.14, p = 0.011; HVLT discrimination, β = −0.156, p = 0.007; HVLT Delayed Recognition, β = −0.109, p = 0.049) and processing speed and attention (SDMT, β = −0.107, p = 0.013). In addition, we found that a greater increasing rate of GFAP was related to a greater decrease in α‐syn (β = −0.158, p < 0.001) and a greater increase in NfL (β = 0.124, p < 0.001). However, we only found PIGD patients with a greater increasing rate of GFAP were associated with a greater decline in episodic memory (HVLT immediate recall, β = −0.189, p = 0.046; HVLT Delayed Recall, β = −0.343, p = 0.002; HVLT retention, β = −0.338, p = 0.015). Similar to TD, we also found a greater increasing rate of GFAP was associated with faster worsening in episodic memory, processing speed and attention, a faster decline in α‐syn, and a faster increase of NfL in TD‐male patients (Table S5). Kaplan–Meier curve for cognitive phase conversion in the follow‐up among PD participants. (a) PD–NC converted to PD–MCI, (b) PD–NC converted to PD–D, (c) PD–MCI converted to PD–D. MCI, mild cognitive impairment; NC, normal cognition; PD, Parkinson's disease; PD–D, Parkinson's disease with dementia. The baseline NfL concentrations were divided three tertiles: Tertile 1, 1 standard deviation below mean; tertile 2, mean; and tertile 3, 1 standard deviation above mean. TD‐NC patients with a greater increasing rate of GFAP were also related to the faster decline of α‐syn (β = −0.25, p < 0.001). Furthermore, TD‐MCI patients with a greater increasing rate of GFAP were associated with a faster decline in episodic memory (HVLT Delayed Recall, β = −0.298, p < 0.001; HVLT retention, β = −0.35, p < 0.001; HVLT discrimination, β = −0.211, p = 0.048), executive function/working memory (LNS, β = −0.186, p = 0.029), language (Semantic fluency total score, β = −0.23, p < 0.001), and processing speed/attention (SDMT, β = −0.201, p = 0.019). TD‐MCI patients with a greater increasing rate of GFAP were also related to a greater increase of tau (β = 0.021, p = 0.822), p‐tau (β = 0.237, p < 0.001) (Table S6). DISCUSSION In this study, we comprehensively analyzed the longitudinal associations of CSF GFAP with cognitive decline and changes in CSF biomarkers in different motor subtypes. In the longitudinal analysis, higher baseline CSF GFAP and a greater increasing rate of CSF GFAP were associated with faster worsening in episodic memory and processing speed/attention, greater decline of α‐syn, and greater increase of NfL in TD patients and TD males. GFAP is the cytoskeleton of astrocytes, indispensable for brain development and vital in regulating neuronal signaling. 27 , 28 In addition, GFAP might play an essential role in the motility and migration of astrocytes, as demonstrated in glioma cell lines. 29 However, the role of GFAP on astrocyte motility and migration under physiological conditions is unknown. 29 A study showed that GFAP expression plays a vital role in maintaining astrocytic foot processes to induce blood–brain barrier properties in endothelial cells and maintain normal CNS myelination. 30 Furthermore, emerging evidence demonstrated that astrocytes could efficiently take up and degrade α‐syn. 31 , 32 Our study found that the longitudinal changes of CSF GFAP were associated with a faster worsening of episodic memory and processing speed/attention. It is reported that GFAP adenovirus‐induced reactive astrocyte proliferation could exacerbate PD pathology mediated by A53T‐α‐syn. 33 Furthermore, Herrera et al. indicated that astrocytes might adversely affect hippocampal function and cognitive behavior. 34 We speculate that the abnormal aggregation of α‐syn is increasing with the development of PD and could activate inflammatory responses in glial cells. 35 Besides, when astrocytes take up a large amount of α‐syn, the structure of astrocytes gets damaged, and astrocytes rapidly release GFAP into CSF and then spread to peripheral blood. 36 Increased GFAP levels could in turn activate other inflammatory factors, such as TNF‐α and promote the progression of PD. In addition, our study indicated that longitudinal increases in CSF GFAP were associated with a greater decrease of CSF α‐syn and a greater increase of CSF NfL in PD‐TD patients for the first time. The exact mechanism remains unclear, and further studies are required. Some studies demonstrated that CSF α‐syn increases as disease progression, and the increases in CSF α‐syn were associated with motor and cognitive deterioration, 37 , 38 while others reported a decrease in CSF α‐syn. 39 CSF α‐syn is decreased because of intracellular aggregation of α‐syn. Furthermore, some studies have indicated that CSF NfL levels are significantly increased in PD–MCI and PD–D patients compared with HCs. 40 Emerging studies provided the possibility of CSF NfL dictating predictive values for general early cognitive decline in PD. 41 , 42 A previous study has found that higher baseline CSF GFAP predicted a greater decline of global cognition over time in early PD patients. 43 However, the previous study did not investigate CSF GFAP and PD motor subtypes. In our study, CSF GFAP could serve as a crucial biomarker of cognition decline in TD. What's more, the increases in CSF GFAP were related to the greater decline in α‐syn and the greater increase in NfL. In addition, this study is the first to find gender differences in the relationship between GFAP and cognition change in TD patients. In the longitudinal analysis, the greater increasing rate of CSF GFAP was related to faster worsening in episodic memory in TD‐male patients, but no such relationship was found in female patients. This may be due to the protective effect of estrogen. Compared with male individuals, female individuals have a lower incidence. 44 , 45 Emerging studies showed that estrogen has anti‐neurotoxic, anti‐inflammatory, anti‐apoptotic, and antioxidant effects. 46 Furthermore, estrogen has been found to regulate nigrostriatal dopaminergic activity. 47 Previous studies have reported that decreased estrogen levels induced by menopause and hysterectomy could promote the progression of PD. 48 However, the pathological mechanism of estrogen is unclear. The reasons are speculated as follows: (1) Estrogens have an antioxidant effect due to the hydroxyl group at the C3 position on the A ring of the steroid structure. 49 (2) Estrogens could interact with BDNF to exert a protective effect. 50 (3) Estrogen could protect dopaminergic neurons by regulating the neuro‐inflammatory reaction of glia. 51 The advantages of our study were as follows: (1) This study investigated the relationships between CSF GFAP levels and cognition change through longitudinal analyses in different motor subtypes for the first time. (2) This article analyzed the relationship between GFAP and cognition change in male and female PD patients separately. However, this study has some limitations. Firstly, the data for this study are required from the PPMI, which is comprised of Caucasians, which may reduce the generalizability. Secondly, there are some missing values of CSF GFAP during follow‐up, which affect the results of the longitudinal analyses. Thirdly, due to the small number of samples of TD, PIGD and intermediate types, especially PIGD and intermediate types, the results analyzed in this paper may have certain biases, and a large number of studies are needed to confirm their accuracy in the future. CONCLUSION In conclusion, our study found that CSF GFAP levels are a promising predictive biomarker to monitor cognition progression in TD patients. However, larger samples are needed to research the predictive value of CSF GFAP levels. AUTHOR CONTRIBUTIONS Y.L. wrote the manuscript. Y.L. and J.W. designed the research. Y.L., J.W., F.N., and A.X. performed the research. Y.L., G.W., and A.X. analyzed the data. FUNDING INFORMATION This research was supported by the Natural Science Foundation of Shandong Province (ZR2020QH123) and the National Natural Science Foundation of China (81971192). CONFLICT OF INTEREST STATEMENT The authors declared no competing interests for this work. ETHICS STATEMENT Ethics approval was obtained from the individual PPMI study sites. CONSENT TO PARTICIPANTS Informed consent was obtained was obtained from all participants prior to inclusion. CONSENT FOR PUBLICATION Not applicable. Supporting information Table S1 Table S2 Table S3 Table S4 Table S5 Table S6