Abstract

Objective

   To identify key cerebrospinal fluid (CSF) metabolomic biomarkers
   associated with Parkinson’s disease (PD) and prodromal PD, providing
   insights for intervention strategy development.

Methods

   Six hundred and thirty-nine participants from the Parkinson’s
   Progression Markers Initiative (PPMI) cohort were included: 300 PD
   patients, 112 healthy controls (HC), and 227 prodromal PD patients.
   Regression analysis and OPLS-DA identified metabolic biomarkers, while
   pathway analysis examined their relationship to clinical features. An
   XGBoost-based early prediction model was developed to assess the
   distinction between PD, prodromal PD, and HC. A two-sample
   bidirectional Mendelian randomization analysis was performed to examine
   the association between differential metabolites and Parkinson’s
   disease.

Results

   Sixty-four metabolites were significantly altered in PD patients
   compared to HC, with 58 elevated and 6 reduced. In prodromal PD, 19
   metabolites were increased, and 34 were decreased. Key metabolic
   pathways involved glutathione and amino acid metabolism. Dopamine
   3-O-sulfate correlated with PD progression, levodopa-equivalent dose,
   and non-motor symptoms. The XGBoost model demonstrated high specificity
   in predicting the onset of PD. The MR analysis results showed a
   positive correlation between higher genetic predictions of dopamine
   3-O-sulfate levels and the risk of Parkinson’s disease. In contrast,
   the reverse MR analysis found that Parkinson’s disease susceptibility
   significantly increased dopamine 3-O-sulfate levels.

Conclusion

   The differential expression of CSF metabolites reveals early cellular
   metabolic changes, providing insights for early diagnosis and
   monitoring PD progression. A bidirectional causal relationship exists
   between genetically determined PD susceptibility and metabolites.

   Keywords: Parkinson diseases, metabolomic biomarkers, early prediction
   model, bidirectional Mendelian randomization, PD susceptibility

Introduction

   Parkinson’s disease (PD) is a motor disorder characterized by the
   progressive loss of dopaminergic neurons in the substantia nigra and
   the abnormal aggregation of α-synuclein. The motor symptoms of PD are
   primarily tremor, bradykinesia, and rigidity. In addition to these
   motor deficits, PD also encompasses a range of non-motor and prodromal
   symptoms, such as sleep disturbances, olfactory dysfunction,
   psychiatric and mood changes, cognitive impairment, and autonomic
   dysfunction. These non-motor symptoms may appear either before or
   concurrently with the onset of motor symptoms ([31]Marinus et al.,
   2018).

   Although numerous hypotheses regarding the pathogenesis of PD have been
   proposed, there is currently no effective method to slow the
   progression of the disease. This complexity arises from the involvement
   of multiple brain regions and various neurotransmitter systems. These
   systems include the co-release of dopamine, a typical neurotransmitter,
   and other excitatory or inhibitory neurotransmitters, which contribute
   to clinical heterogeneity ([32]Barcomb and Ford, 2023). While the
   accuracy of clinical diagnosis has improved in the past decade,
   especially in the early stages, reaching up to 90.3% in some studies
   ([33]Virameteekul et al., 2023), predicting disease progression in the
   early stages remains challenging. This is primarily due to the overlap
   of early clinical features, the complexity of disease subtypes, and the
   limitations of diagnostic criteria ([34]Tolosa et al., 2021). The
   prodromal phase is considered a critical window for intervention,
   making early and accurate diagnosis essential. Predicting disease
   progression based on reliable and sensitive early biomarkers and
   quantifying different pathological states of PD remain key research
   focuses ([35]Theis et al., 2024).

   Early diagnosis of PD primarily involves clinical symptom assessment,
   biochemical testing, imaging techniques, and genetic analysis
   ([36]Parab et al., 2023; [37]Mitchell et al., 2021). Cerebrospinal
   fluid (CSF), which directly interacts with brain cells, offers an
   accurate reflection of the underlying molecular mechanisms of PD. While
   the α-synuclein seed amplification assay in CSF demonstrates high
   sensitivity and specificity, it reflects only part of the disease
   pathology, highlighting the need for additional biomarkers to fully
   characterize PD ([38]Postuma and Berg, 2016). CSF metabolomics, through
   the mapping and quantification of various small-molecule metabolites,
   provides a comprehensive insight into cellular metabolism and
   neurotransmitter alterations ([39]Stoessel et al., 2018). With recent
   advancements in liquid chromatography-mass spectrometry (LC-MS/MS), key
   biomarkers related to lipid metabolism, polyamines, amino acids, and
   purine metabolism have garnered increasing attention ([40]Trezzi et
   al., 2017; [41]Kremer et al., 2021).

   This study aims to explore the differences in various metabolites,
   particularly lipid metabolites, at different clinical stages of PD
   (including healthy controls, prodromal PD, and clinically diagnosed PD
   patients) using CSF metabolomics as a data-driven source. The study
   further aims to predict the risk of PD progression. The objectives of
   this study are as follows: (1) to identify cerebrospinal metabolic
   biomarkers at different stages of PD progression; (2) to assess the
   reliability of predictive models by developing a clinical risk model
   for PD; (3) to link lipid metabolism biomarkers with clinical
   manifestations to provide clinical utility; (4) to uncover potential
   mechanisms underlying PD progression through metabolic biomarkers and
   associated molecular pathways; (5) MR analysis was conducted using
   publicly available genome-wide association data to evaluate the causal
   relationship between differential metabolites and Parkinson’s disease.

Materials and methods

Study participants

   This study utilized data from the Parkinson’s Progression Markers
   Initiative (PPMI) database, a large-scale clinical observational study
   aimed at identifying biomarkers of PD progression from the prodromal
   phase through to disease onset. A total of 639 participants were
   included in the analysis, with data collection completed by January
   2020. The sample size was determined based on previous studies
   ([42]Huntwork-Rodriguez et al., 2023). Participants were classified
   into three groups based on predefined inclusion criteria: (1) PD
   patients: individuals diagnosed with PD, who were undergoing levodopa
   treatment. (2) Healthy controls: individuals with no history of
   neurological disorders, no first-degree family history of PD, and
   normal dopamine transporter (DAT) single-photon emission computed
   tomography (SPECT) imaging. (3) Prodromal participants: individuals who
   had not been clinically diagnosed with PD but exhibited one or more of
   the following risk factors: rapid eye movement sleep behavior disorder
   (RBD), olfactory dysfunction, dopamine transporter (DAT) deficiency, or
   genetic variants associated with an increased risk of PD. The prodromal
   cohort has as inclusion criteria age >60 years (with the exception of
   SCNA and other rare mutations).

   Baseline demographic information, motor and non-motor assessments, and
   biochemical test results were collected for all participants. All
   participants underwent lumbar puncture for CSF collection, followed by
   metabolite and lipid analysis. Data for this study were accessed via
   the PPMI online database.[43]^1 The PPMI study received ethical
   approval from the institutional review boards of over 50 research
   centers globally. Detailed information regarding the ethics committees
   of the clinical centers can be found in [44]Supplementary Table 1. All
   participants provided written informed consent before inclusion in the
   study, in accordance with ethical guidelines ([45]Brumm et al., 2023).
   The methodology of this study complies with the relevant guidelines of
   the PPMI Data and Publications Committee (DPC), and the manuscript was
   submitted to the DPC for review. Genetic association summary data were
   obtained from GWAS, with dopamine 3-O-sulfate GWAS data sourced from
   the Wisconsin Alzheimer’s Disease Research Center (WADRC) and the
   Wisconsin Registry for Alzheimer’s Prevention (WRAP) cohorts, two
   European populations ([46]Panyard et al., 2021). The data included 412
   cerebrospinal fluid metabolites from 291 samples. Parkinson’s disease
   GWAS data were sourced from the International Parkinson’s Disease
   Genomics Consortium ([47]Nalls et al., 2019), comprising 33,674 PD
   cases and 449,056 control samples.

Cerebrospinal fluid metabolite and lipid analysis

   CSF samples collected from participants were analyzed using liquid
   chromatography–tandem mass spectrometry (LC-MS/MS), employing both
   targeted metabolomics/lipidomics and untargeted metabolomics
   approaches. A total of 348 compounds were identified, including
   sphingolipids, polyamines, cholesterol, gangliosides, ceramides, amino
   acids, caffeine, and purine metabolites. To minimize batch effects,
   internal references were established separately for PD patients and