Skip to main content
NCBI home page
Scientific Reports logoLink to Scientific Reports
. 2026 Mar 23;16:9588. doi: 10.1038/s41598-025-30521-4

Cerebrospinal fluid and plasma metabolites in Parkinson’s disease: a Mendelian randomization study

Jia-Li Wang 1,#, Qianhui Zhao 2,#, Ran Zheng 1, Yi Fang 1, Jin Cao 1, Bao-Rong Zhang 1,✉
PMCID: PMC13009158  PMID: 41872227

Abstract

Metabolites in cerebrospinal fluid (CSF) and plasma have been implicated in Parkinson’s disease (PD), but their causal roles remain unclear. This study aims to investigate the causal relationships between specific CSF and plasma metabolites and PD risk, using Mendelian Randomization (MR) analysis. We utilized MR with inverse-variance weighting (IVW) as the primary analytical method, supplemented by four additional MR models to validate robustness. Data included 1,400 plasma metabolites from the Canadian Longitudinal Study on Aging, and 338 CSF metabolites from the Wisconsin Alzheimer’s Disease Research Center and Wisconsin Registry for Alzheimer’s Prevention, USA. metabolites from the Canadian Longitudinal Study on Aging. PD outcome data were obtained from a GWAS meta-analysis by the International Parkinson’s Disease Genomics Consortium. MR analysis identified six CSF metabolites with suggestive causal relationships with PD. Dimethylglycine, gluconate, oxalate (ethanedioate), and an unknown metabolite (X-12015) were positively associated with PD risk, while (1-enyl-palmitoyl)-2-arachidonoyl-GPC (P-16:0/20:4) and an unknown compound (X-23587) were negatively associated. In plasma, 49 metabolites demonstrated suggestive causal relationships with PD risk. Key metabolites included hydroxy-3-carboxy-4-methyl-5-propyl-2-furanpropanoic acid (hydroxy-CMPF), carnitine C14, and 1-dihomo-linolenylglycerol (20:3) with positive associations, and tryptophan, succinate to acetoacetate ratio, and O-sulfo-L-tyrosine with negative associations. Notably, O-sulfo-L-tyrosine emerged as the most significant protective metabolite in plasma, showing robust associations across four MR models. This study highlights causal relationships between specific CSF and plasma metabolites and PD risk, underscoring O-sulfo-L-tyrosine as a potential biomarker and therapeutic target. These findings provide a foundation for further exploration of metabolic pathways in PD pathogenesis and intervention strategies.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-30521-4.

Keywords: Parkinson’s disease, Metabolomics, Mendelian randomization, Cerebrospinal fluid metabolites, Plasma metabolites

Subject terms: Parkinson's disease, Genetic databases

Introduction

Parkinson’s disease (PD) is a progressive neurodegenerative disorder primarily characterized by the loss of dopaminergic neurons in the substantia nigra, leading to symptoms such as tremors, bradykinesia, and rigidity1,2. Affecting millions globally, the prevalence of PD increases with age3. Despite extensive research, the exact etiology of PD remains unclear, involving a complex interplay of genetic, environmental, and lifestyle factors4,5.

Metabolomics, the comprehensive study of small molecule metabolites within biological systems, has emerged as a crucial tool in understanding PD6. This approach allows for the identification of metabolic changes that occur in response to disease processes, offering potential biomarkers for early diagnosis and targets for therapeutic intervention. Recent studies have underscored significant metabolic disruptions in both cerebrospinal fluid (CSF) and plasma of PD patients, pointing to systemic metabolic dysfunction associated with the disease7. Analyses of plasma metabolomics have revealed alterations in various metabolic pathways, including those related to amino acids, lipids, and energy metabolism8,9. Similarly, CSF metabolomics studies have identified changes in neurotransmitter metabolites, oxidative stress markers, and other pathways specific to the central nervous system10–12. Given that CSF directly reflects biochemical alterations within the brain, while plasma captures systemic metabolic changes, analyzing both biofluids together provides complementary insights into the central and peripheral metabolic disturbances associated with PD. These findings indicate that metabolic biomarkers in both CSF and plasma could serve as valuable indicators for PD13.

Mendelian Randomization (MR) is a robust analytical method that uses genetic variants as instrumental variables (IVs) to infer causal relationships between exposures (e.g., metabolite levels) and outcomes (e.g., PD). This method addresses confounding factors and reverse causation issues common in observational studies, providing more reliable evidence for causality. Applying MR to metabolomic data in PD can identify causal metabolic pathways and potential biomarkers, advancing our understanding of PD pathogenesis and aiding in the development of targeted therapeutic strategies and more effective clinical interventions14.

Patients and methods

Exposure data

In this study, we utilized plasma metabolite data from 8,299 participants aged 45–85 from the Canadian Longitudinal Study on Aging, which primarily includes individuals of European ancestry. The data included summary statistics for 1,091 blood metabolites and 309 metabolite ratios obtained via genome-wide association studies (GWAS)15. Additionally, the GWAS summary data for CSF metabolites were sourced from a study by Panyard et al., which included 338 metabolites in the CSF of 291 participants16. These data were derived from participants enrolled in the Wisconsin Alzheimer’s Disease Research Center, with a mean age of 64.7 years, and the Wisconsin Registry for Alzheimer’s Prevention, with a mean age of 62.0 years; all participants were of European ancestry.

Outcome data

The outcome data for PD were obtained from a recent GWAS meta-analysis conducted by Nalls et al.17. This meta-analysis, performed by the International Parkinson’s Disease Genomics Consortium in collaboration with 23andMe, included data from 16 cohorts with a total of 33,674 individuals: 15,056 PD patients and 18,618 control subjects. The GWAS primarily focused on identifying genetic variants associated with PD risk and adjusted for key covariates such as sex and age to ensure robust results. All individuals included in the meta-analysis were of European ancestry to minimize population stratification bias. It is important to note that the 23andMe data included in the meta-analysis was not publicly accessible, and therefore, this cohort was omitted from our analysis. This ensures that our study is based only on datasets that are publicly available and fully replicable.

Instrument selection

To ensure the validity of the genetic instrumental variables, we conducted a rigorous multi-step selection process for single-nucleotide polymorphisms (SNPs), as detailed in Fig. 1. First, considering that genome-wide significance thresholds may be too stringent for CSF and plasma metabolites, we adopted a relatively relaxed threshold (P < 5.00E-5) to identify instrumental variables (IVs) for plasma metabolites18,19. This threshold aligns with those used in similar metabolomics studies, aiming to enhance detection sensitivity while maintaining the number and representativeness of genetic instruments. Second, to ensure the independence of selected SNPs, we applied clumping based on the European samples from the 1000 Genomes Project linkage disequilibrium (LD) reference panel, excluding SNPs with R2 > 0.001 within a 10 kb window. This process reduces potential genetic correlations and ensures the random allocation of alleles. Specifically, clumping groups highly correlated SNPs within an LD region and retains only the most significant SNP, preserving the independence of the instrumental variables.

Fig. 1.

Fig. 1

Open in a new tab

Flowchart of Mendelian randomization study on metabolites and Parkinson’s disease. IVS: instrumental variables; SNPs: single nucleotide polymorphisms; MR: Mendelian randomization; IVW: inverse variance weighted.

Third, we harmonized the data to ensure consistent effect estimates, excluding SNPs with ambiguous alleles. For palindromic SNPs (e.g., A/T or G/C), we set the maximum minor allele frequency threshold at 0.3 to avoid orientation ambiguity. Fourth, we calculated F-statistics for each SNP and excluded those with F < 10 to avoid weak instrument bias. The calculation of F-statistics, as referenced in20, evaluates the strength of the association between the instrumental variable and the exposure, with F < 10 indicating weaker instruments that could compromise the reliability of MR analysis. Finally, MR analysis was conducted only on metabolites with more than two qualified SNPs to ensure robust statistical power. These steps collectively ensure the validity and reliability of the instrumental variables used in the MR analysis.

Mendelian randomization assumptions

MR was used to evaluate the causal relationship between exposure (plasma and CSF metabolites) and disease outcome (PD). MR analysis has three core assumptions: (1) SNPs are closely related to exposure, (2) SNPs are not associated with confounding factors, and (3) SNPs have no other (pleiotropic) way of affecting the results except through exposure21.

Statistical analysis

We conducted MR analyses using the TwoSampleMR package22 in R version 4.4.0. For the primary analysis, the inverse variance-weighted (IVW) method was employed to evaluate MR estimates, supplemented by four additional MR methods: MR-Egger, weighted median, simple mode, and weighted mode. The MR results were expressed as odds ratios (ORs) with corresponding 95% confidence intervals (CIs). The IVW method is considered robust for estimating causal effects, provided that genetic variations comply with the three instrumental variable assumptions and are not influenced by pleiotropy. To account for multiple testing, we applied Bonferroni and false discovery rate (FDR) corrections.

Sensitivity analysis

Several sensitivity analyses were conducted to validate the robustness of the MR estimates. These included the MR-Egger intercept22, Cochran’s Q test, leave-one-out test23, and funnel plots. Cochran’s Q test was utilized to detect potential heterogeneity, with a P < 0.05 indicating its presence. The MR-Egger intercept method assessed horizontal pleiotropy. Leave-one-out sampling identified influential IV outliers, and funnel plots visualized SNP heterogeneity. Forest and scatter plots were created to display the study results.

Reverse MR analysis

To ensure that reverse causality did not bias the findings, a reverse MR analysis was conducted for the metabolites that showed significant associations with PD risk in the primary (forward) MR analysis. In the reverse MR analysis, the IVW method was again utilized, maintaining consistency with the primary analysis. However, due to the availability of a larger number of instrumental variables for the reverse analysis, we applied a more stringent significance threshold (P < 5.00E-8) to select the genetic instruments. Apart from this adjustment, all other analytical methods and sensitivity tests were consistent with those employed in the forward MR analysis. By performing the reverse MR analysis, we aimed to confirm the directionality of the observed causal relationships and further ensure the robustness and reliability of our findings.

Metabolic pathway analysis

The metabolic pathway analysis was carried out using MetaboAnalyst 6.0, an online tool available at https://www.metaboanalyst.ca/. To maintain the accuracy and robustness of the results, only metabolites with a statistical significance of P < 0.05 were considered. This analysis leveraged two key databases: the Small Molecule Pathway Database (SMPDB) and the Kyoto Encyclopedia of Genes and Genomes (KEGG). A significance threshold of 0.05 was established for the pathway analysis. Functional enrichment and pathway analysis modules were applied to pinpoint relevant metabolite groups or pathways associated with the biological processes of interest.

Results

We conducted an MR study on CSF and plasma metabolites in a European population (Fig. 2). From a total of 338 CSF metabolites and 1400 plasma metabolites, we selected between 3 and 404 independent SNPs as IVs. Each of these IVs had an F value greater than 10, ensuring no weak IV bias. Detailed summary information for these SNPs is provided in Tables S1 and S2.

Fig. 2.

Fig. 2

Open in a new tab

Circular heatmap of combined cerebrospinal fluid and plasma metabolites associated with Parkinson’s disease in Mendelian randomization analysis. The circular heatmap displays the odds ratios and p-values of cerebrospinal fluid and plasma metabolites associated with Parkinson’s Disease. Red indicates risk factors and blue indicates protective factors. OR: Odds Ratio; IVW: Inverse Variance Weighted; WM: Weighted Median.

In CSF metabolites, a total of six metabolites were identified to have suggestive causal relationships with the PD risk, including two unknown metabolites (Fig. 3 and Table S8). Among them, levels of dimethylglycine (OR [95% CI] = 1.06 [1.01–1.12], P = 0.023), gluconate (OR [95% CI] = 1.12 [1.01–1.25], P = 0.034), oxalate (ethanedioate) (OR [95% CI] = 1.07 [1.00-1.13], P = 0.043), and the unknown metabolite X-12,015 (OR [95% CI] = 1.06 [1.00-1.11], P = 0.044) were positively associated with the PD risk. Conversely, levels of (1-enyl-palmitoyl)-2-arachidonoyl-gpc (p-16:0/20:4) (OR [95% CI] = 0.93 [0.88–0.99], P = 0.019) and the unknown metabolite X-23,587 (OR [95% CI] = 0.93 [0.88–0.98], P = 0.005) were negatively associated with the PD risk (Fig. 3).

Fig. 3.

Fig. 3

Open in a new tab

Mendelian randomization results of cerebrospinal fluid metabolites and Parkinson’s disease. This figure summarizes the causal relationships between cerebrospinal fluid metabolites and Parkinson’s Disease, determined by Mendelian randomization analysis. The size of each point represents the odds ratio, and horizontal lines indicate 95% confidence intervals. OR: Odds Ratio; CI: Confidence Interval; IVW: Inverse Variance Weighted.

In the MR analysis of plasma metabolites, a total of 49 metabolites were identified to have suggestive causal relationships with the PD risk (Fig. 4 and Table S9), including five unknown metabolites. Among them, 21 metabolites were positively associated with the PD risk. The top five metabolites with the smallest p-values are hydroxy-3-carboxy-4-methyl-5-propyl-2-furanpropanoic acid (hydroxy-CMPF) (OR [95% CI] = 1.36 [1.15–1.60], P = 0.0003), carnitine C14 (OR [95% CI] = 1.38 [1.15–1.66], P = 0.0005), 1-linoleoyl-GPG (18:2) (OR [95% CI] = 1.26 [1.09–1.47], P = 0.002), glucose to maltose ratio (OR [95% CI] = 1.19 [1.04–1.36], P = 0.012), and cis-3,4-methyleneheptanoate (OR [95% CI] = 1.21 [1.04–1.42], P = 0.015). Conversely, 23 metabolites were negatively associated with the PD risk. The top five protective metabolites with the smallest p-values are tryptophan (OR [95% CI] = 0.79 [0.69–0.90], P = 0.0006), succinate to acetoacetate ratio (OR [95% CI] = 0.79 [0.66–0.93], P = 0.0055), N, N,N-trimethyl-alanylproline betaine (TMAP) (OR [95% CI] = 0.82 [0.71–0.95], P = 0.0067), glucuronide of piperine metabolite C17H21NO3 (3) (OR [95% CI] = 0.80 [0.68–0.94], P = 0.0073), and linoleoylcholine (OR [95% CI] = 0.73 [0.58–0.92], P = 0.0078).

Fig. 4.

Fig. 4

Open in a new tab

Mendelian randomization results of plasma metabolites and Parkinson’s disease. This figure summarizes the causal relationships between plasma metabolites and Parkinson’s Disease, determined by Mendelian randomization analysis using the inverse variance weighted method. The size of each point represents the odds ratio, and horizontal lines indicate 95% confidence intervals. OR: Odds Ratio; CI: Confidence Interval.

Additionally, several metabolites demonstrated significant results across multiple MR methods, highlighting their robustness. 1-dihomo-linolenylglycerol (20:3) levels showed a significant positive association with PD risk in the IVW method (OR [95% CI] = 1.15 [1.01–1.32], P = 0.038). Other methods also produced significant results, including MR-Egger (P = 0.027, OR = 1.56) and the weighted median (P = 0.007, OR = 1.29). Carnitine C14 levels, one of the top metabolites associated with increased PD risk, remained significant in the IVW method (OR [95% CI] = 1.38 [1.15–1.66], P = 0.0005), MR-Egger (P = 0.034, OR = 1.82), and the weighted median (P = 0.008, OR = 1.42). These consistent findings across methods support the potential causal role of these metabolites in PD pathogenesis. Notably, O-sulfo-l-tyrosine levels exhibited a strong negative association with PD risk across all four MR methods. In the IVW method, it was significantly associated with reduced PD risk (OR [95% CI] = 0.87 [0.77–0.99], P = 0.032). Other methods also showed significant results, including MR-Egger (P = 0.011, OR = 0.71), the weighted median (P = 0.031, OR = 0.82), and the weighted mode (P = 0.013, OR = 0.69). This consistent pattern strongly indicates its protective role against PD and highlights its potential as a key metabolic factor warranting further investigation.

To explore the possibility of reverse causality, we performed a reverse MR analysis to determine whether an increased PD risk could influence the levels of certain metabolites (Fig. 5). In the CSF metabolites, PD risk was associated with a decrease in gluconate levels (OR [95% CI] = 0.95 [0.91–0.99], p = 0.0276). In the plasma metabolites, PD risk was associated with decreases in 1-stearoyl-2-oleoyl-GPI (18:0/18:1) levels (OR [95% CI] = 0.94 [0.89–0.99], p = 0.0123), argininate levels (OR [95% CI] = 0.94 [0.89–0.99], p = 0.0153), and 2,3-dihydroxy-2-methylbutyrate levels (OR [95% CI] = 0.95 [0.90-1.00], p = 0.0331). These findings suggest that an increased PD risk may lead to alterations in specific CSF and plasma metabolites, underscoring the impact of reverse causality on our study.

Fig. 5.

Fig. 5

Open in a new tab

Reverse Mendelian randomization results of metabolites and Parkinson’s disease. This figure shows the reverse Mendelian randomization results of metabolites and Parkinson’s Disease, determined using the inverse variance weighted method based on metabolites that showed significant associations (P < 0.05) in the primary analysis. “Forward” indicates the causal effect of metabolites on Parkinson’s Disease, and “reverse” indicates the causal effect of Parkinson’s Disease on metabolite levels. OR: Odds Ratio; CSF: Cerebrospinal Fluid; CI: Confidence Interval.

The sensitivity analyses confirmed the robustness of the MR estimates. Despite the IVW methods being susceptible to weak IV bias, the MR-Egger, weighted median, simple mode, and weighted mode methods showed directions consistent with the MR estimates. Heterogeneity was evaluated using Cochran’s Q statistic in MR-Egger regression, and results with notable heterogeneity were excluded (Tables S3 and S4). The MR-Egger intercept was used to detect pleiotropy, and any results indicating pleiotropy were excluded (Tables S5 and S6). The leave-one-out sensitivity analysis (Figures S1-S2) and forest plots (Figures S3-S4) did not reveal any anomalous SNPs, confirming the stability of the MR estimates. Scatter plots (Figures S5-S6) and funnel plots (Figures S7-S8) further supported these findings, showing directions consistent with the IVW approach and symmetrical distribution of IVs.

Metabolic pathway analysis

In this study, we utilized MetaboAnalyst 6.0 to conduct a comprehensive metabolic pathway analysis of plasma metabolites in PD patients (Table S7 and Figures S9). Due to the limited number of annotated metabolites available in CSF, pathway analysis was not performed for CSF samples. Our results identified 15 significant metabolic pathways, of which only the urea cycle (4/23, P = 0.0006; D-alanine, pyruvic acid, urea, adenosine 5’-diphosphate (ADP); FDR = 0.0439) and ammonia recycling (4/25, P = 0.0009; pyruvic acid, L-asparagine, L-histidine, ADP; FDR = 0.0439) remained statistically significant after FDR correction. These pathways demonstrate critical metabolic disruptions in PD, highlighting their potential involvement in disease pathogenesis. The other metabolic pathways identified included alanine metabolism, glucose-alanine cycle, glycine and serine metabolism, trehalose degradation, bile acid biosynthesis, fatty acid biosynthesis, phosphatidylethanolamine biosynthesis, riboflavin metabolism, transfer of acetyl groups into mitochondria, betaine metabolism, phosphatidylcholine biosynthesis, glutathione metabolism, and glutamate metabolism.

Discussion

Key findings from MR analysis

The findings from our MR study provide robust evidence for the metabolic underpinnings of PD. By leveraging genetic variants as instrumental variables, we inferred causal relationships between specific CSF and plasma metabolites and PD risk. Our MR analysis identified 49 metabolites with suggestive causal relationships with PD risk: 21 positively associated, 23 negatively associated, and 5 unknown compounds. In the context of CSF metabolites, we identified six with suggestive causal relationships to PD: dimethylglycine, gluconate, oxalate (ethanedioate), and the unknown metabolite X-12,015 were positively associated with PD risk, whereas (1-enyl-palmitoyl)-2-arachidonoyl-GPC (P-16:0/20:4) and the unknown metabolite X-23,587 were negatively associated. Notably, in the plasma metabolites, we identified hydroxy-3-carboxy-4-methyl-5-propyl-2-furanpropanoic acid (hydroxy-CMPF), carnitine C14, and 1-dihomo-linolenylglycerol (20:3) as having the positive associations with PD risk, while metabolites such as tryptophan, succinate to acetoacetate ratio, and O-sulfo-L-tyrosine were found to be negatively associated with PD risk. Among these, O-sulfo-L-tyrosine stood out as the most prominent protective metabolite in plasma, demonstrating consistent and significant results across four different MR models, underscoring its potential role as a key metabolic marker in PD pathophysiology.

Insights into tryptophan and mitochondrial metabolism

These findings highlight the multifaceted metabolic alterations in PD. Our study found that tryptophan is negatively associated with PD risk. This amino acid is not only crucial for protein synthesis but also serves as a precursor to serotonin, a neurotransmitter that regulates numerous functions in the central nervous system. In PD patients, serotonin system dysfunction is associated with non-motor symptoms such as depression and sleep disturbances24. Additionally, tryptophan metabolites possess antioxidative properties, reducing oxidative stress by decreasing the production of free radicals and enhancing the activity of antioxidant enzymes, thereby protecting neurons25. The kynurenine pathway of tryptophan metabolism also generates anti-inflammatory metabolites, which is important given the role of inflammation in PD pathology26.

Our study found that a lower succinate to acetoacetate ratio is associated with reduced PD risk. Previous research has shown that this ratio reflects the metabolic state of cells, particularly mitochondrial function, which is often compromised in PD. Mitochondrial dysfunction leads to disrupted energy metabolism, a hallmark of PD27. An increase in succinate may indicate a block in the tricarboxylic acid cycle, contributing to oxidative stress by generating reactive oxygen species. Conversely, an increase in acetoacetate suggests enhanced ketone body metabolism, which can serve as an alternative energy source when mitochondrial function is impaired, thereby reducing oxidative stress by decreasing reactive oxygen species production28.

The protective role of O-sulfo-L-tyrosine

Among the identified metabolites, O-sulfo-L-tyrosine emerged as the most significant protective factor in plasma, showing consistent and robust associations across four MR models (IVW, MR-Egger, weighted median, and weighted mode). While this relationship was not observed in our CSF data, its protective role is strongly supported by findings in other studies using larger integrated CSF datasets29. These results underscore O-sulfo-L-tyrosine’s potential as a metabolic biomarker and provide new insights into its role in PD pathophysiology. Mechanistically, O-sulfo-L-tyrosine is generated from tyrosine through sulfation catalyzed by tyrosylprotein sulfotransferases, and its circulating concentration is strongly influenced by the availability of tyrosine as a substrate30. Therefore, lower O-sulfo-L-tyrosine levels in PD may primarily reflect reduced tyrosine pools, rather than changes in sulfation activity. This interpretation is consistent with previous reports of reduced tyrosine and phenylalanine levels and a decreased tyrosine/phenylalanine ratio in PD patients31, as well as emerging evidence that phenylalanine may exert protective effects32. Given that tyrosine is the direct biochemical precursor of L-DOPA and dopamine, diminished tyrosine availability could compromise dopaminergic neurotransmission, with O-sulfo-L-tyrosine serving as a proxy marker of aromatic amino acid sufficiency. Thus, our MR findings support a protective association of O-sulfo-L-tyrosine with PD.

Lipid metabolism-related plasma markers

Notably, among the plasma metabolites, carnitine C14, 1-dihomo-linolenylglycerol (20:3), and 1-linoleoyl-GPG (18:2) demonstrated positive associations with PD risk. Carnitine C14 is a long-chain acylcarnitine and a marker of impaired mitochondrial fatty acid β-oxidation. Its accumulation can disrupt mitochondrial energy metabolism and promote inflammation via activation of NF-ĸB and MAPK pathways, contributing to neurodegenerative processes33. Likewise, 1-dihomo-linolenylglycerol (20:3) and 1-linoleoyl-GPG (18:2) are glycerolipid derivatives that may alter phospholipid composition and membrane dynamics. Such alterations can disrupt synaptic vesicle trafficking and facilitate α-syn aggregation, a hallmark of PD pathology34. Supporting this, plasma lipidomics studies in PD patients have consistently reported dysregulation of glycerolipids, including diacylglycerols and triacylglycerols, which may contribute to neuroinflammation and mitochondrial stress35,36. These mechanistic insights reinforce the plausibility of a causal role for these lipid-related metabolites in PD pathogenesis.

Metabolic pathways associated with PD

Due to the limited number of metabolites detected in CSF, pathway analysis was performed only for plasma metabolites. Our metabolic pathway analysis identified 15 enriched pathways, with the urea cycle and ammonia recycling remaining statistically significant after FDR correction. These pathways emphasize critical disruptions in nitrogen metabolism, which may contribute to neurotoxicity and disease progression in PD. Other pathways, such as alanine metabolism, glutathione metabolism, and phosphatidylcholine biosynthesis, also revealed potential metabolic impairments linked to oxidative stress and mitochondrial dysfunction. Although these pathways did not pass FDR correction, they provide valuable insights into broader systemic alterations associated with PD, warranting further investigation.

In this study, we identified several causal metabolites associated with PD risk, including urea, the alanine to asparagine ratio, the ADP to choline ratio, and the histidine to pyruvate ratio. Notably, urea, the final product of the urea cycle, reflects ammonia detoxification capacity, and its elevated levels have been reported in multiple brain regions of PD patients, potentially contributing to neurotoxicity, cognitive impairment, and motor dysfunction37,38. Alterations in amino acid-related ratios, such as the alanine to asparagine and histidine to pyruvate ratios, suggest disturbances in nitrogen metabolism and amino acid catabolism, which may exacerbate ammonia accumulation. Although the ADP to choline ratio is not directly involved in the urea cycle, ADP is crucial for cellular energy metabolism, and the urea cycle itself is an adenosine 5’-triphosphate (ATP)-dependent process. Thus, impaired energy metabolism may indirectly compromise urea synthesis and ammonia clearance. Collectively, these findings support the involvement of urea cycle dysregulation and impaired ammonia recycling in PD pathogenesis, consistent with clinical evidence linking elevated blood urea nitrogen, proteinuria, and renal dysfunction to greater PD severity and faster disease progression39.

Our findings also align with and expand upon previous MR studies on CSF and plasma metabolites in PD. Notably, Liao et al.40 identified oxalate in CSF as positively associated with PD risk, consistent with our results and reinforcing the role of metabolic toxicity. In contrast, other metabolites such as ribitol and lysine were not replicated, likely due to differences in outcome datasets or instrument selection. In plasma, we confirmed several associations reported by Zhang et al., particularly lipid-related metabolites such as carnitine derivatives and amino acid-related compounds including tryptophan41. Additionally, both studies pointed to mitochondrial and nitrogen metabolic dysfunction as key features of PD. Our pathway analysis uniquely identified the urea cycle and ammonia recycling as significantly enriched pathways, which were not reported in earlier plasma-based MR studies. By integrating CSF and plasma data, our study provides a more comprehensive view of metabolic disruptions in PD and highlights distinct metabolic signals captured through CSF profiling. Our study extends prior work with broader metabolite coverage and more robust MR analyses.

Clinical implications and future directions

This study represents the first application of MR to investigate CSF and plasma metabolites in PD. By addressing confounding factors and reverse causation, this approach provides valuable insights into the metabolic dysregulation associated with PD, though it does not fully eliminate potential biases inherent to MR methods.

Certain limitations should be acknowledged. Firstly, our study population consisted primarily of individuals of European descent, which may restrict the generalizability of our findings to other ethnic groups. Future research should include more diverse populations to validate these results across different genetic backgrounds. Secondly, our analysis was limited to metabolites with available genetic data, potentially overlooking other relevant metabolites and pathways involved in PD. Expanding metabolomic profiling to include a broader spectrum of metabolites could yield a more comprehensive understanding of the disease. Thirdly, although this study utilized CSF metabolite-QTL data from Panyard et al. and PD GWAS data from Nalls et al. (2019), more recent and larger-scale studies have since become available. Notably, a CSF metabolomics GWAS based on approximately 2,000 individuals and an updated PD GWAS identifying 59 additional risk loci may offer more comprehensive resources for future replication and refinement of our findings.

Additionally, we employed a relatively relaxed significance threshold (P < 5.00E-5) due to the limited number of SNPs, consistent with the methodological practices reported in similar metabolite-based MR studies19. For comparison, results using a stricter threshold (P < 5.00E-8) for plasma are reported in Supplementary Table S10. While this approach helps ensure sufficient SNP selection for analysis, it may increase the likelihood of false positives. Furthermore, we performed multiple testing corrections using both FDR and Bonferroni methods (Tables S8 and S9); however, as in other similar studies18, none of the associations in our study remained significant after correction, likely due to the large number of metabolites tested. To address these limitations, future studies should consider adopting stricter statistical thresholds and implementing multiple testing corrections to minimize the likelihood of false discoveries. Expanding the dataset, either by increasing sample sizes or integrating additional cohorts, would also enhance the statistical power and reliability of the results. These measures will not only improve the robustness of metabolomic MR studies but also ensure comparability with existing research methodologies.

In conclusion, our MR study identified 6 CSF metabolites and 49 plasma metabolites with suggestive causal relationships with PD. Of these, 23 plasma metabolites and 1 CSF metabolite were protective, while 21 plasma metabolites and 3 CSF metabolites were risk factors. Notably, O-sulfo-L-tyrosine emerged as the most significant protective metabolite in plasma, highlighting its potential as a diagnostic biomarker and therapeutic target. Our metabolic pathway analysis further revealed critical disruptions in nitrogen metabolism, particularly in the urea cycle and ammonia recycling pathways, which may contribute to PD pathogenesis. These findings lay a foundation for future research to validate and translate these metabolic insights into clinical applications, ultimately enhancing our understanding of PD and improving patient outcomes.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary Material 1 (3.4MB, zip)

Acknowledgements

We express our gratitude to the participants of the UK Biobank for their contribution to this research, which utilized the UK Biobank Resource. We also extend our sincere gratitude to Blauwendraat and colleagues for providing the essential data that greatly contributed to our research.

Author contributions

B.Z. contributed to the conception and design of the study. J.W. and Q.Z. contributed equally to this work. J.W., R.Z., Y.F., and J.C. contributed to the acquisition and analysis of data. J.W. drafted the initial text and prepared the figures. Q.Z. performed additional statistical analyses (including multiple-testing corrections and sensitivity analyses), assisted in preparing supplementary tables and figure revisions, and expanded the Discussion with updated mechanistic insights on O-sulfo-L-tyrosine, tyrosine metabolism, and nitrogen pathways. All authors read and approved the final manuscript.

Data availability

All data generated or analyzed during this study are included in this published article and its Supplementary Materials.

Code availability

The underlying code for this study is available at https://github.com/598047655/MR_metabolomics.

Declarations

Standard protocol approvals, registrations, and patient consents

Ethical review and approval were waived for this study because no individual patients were directly involved in the overall process of our study. Our study was based only on publicly available GWAS data.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Jia-Li Wang and Qianhui Zhao contributed equally to this work.

References

  • 1.Pasciuto, E. et al. Microglia require CD4 T cells to complete the fetal-to-adult transition. Cell182, 625–640e24. 10.1016/j.cell.2020.06.026 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Brochard, V. et al. Infiltration of CD4 + lymphocytes into the brain contributes to neurodegeneration in a mouse model of Parkinson disease. J. Clin. Invest. JCI36470. 10.1172/JCI36470 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Dorsey, E. R., Sherer, T., Okun, M. S. & Bloem, B. R. The emerging evidence of the Parkinson pandemic. JPD8, S3–S8. 10.3233/JPD-181474 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Yahalom, G. et al. Age at onset of Parkinson’s disease among Ashkenazi Jewish patients: contribution of environmental factors, LRRK2 p.G2019S and GBA p.N370S mutations. JPD10, 1123–1132. 10.3233/JPD-191829 (2020). [DOI] [PubMed] [Google Scholar]
  • 5.Chairta, P. P. et al. Prediction of Parkinson’s disease risk based on genetic profile and established risk factors. Genes12, 1278. 10.3390/genes12081278 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Li, X., Fan, X., Yang, H. & Liu, Y. Review of metabolomics-based biomarker research for Parkinson’s disease. Mol. Neurobiol.59, 1041–1057. 10.1007/s12035-021-02657-7 (2022). [DOI] [PubMed] [Google Scholar]
  • 7.Trupp, M. et al. Metabolite and peptide levels in plasma and CSF differentiating healthy controls from patients with newly diagnosed Parkinson’s disease. J. Parkinson’s Disease. 4, 549–560. 10.3233/JPD-140389 (2014). [DOI] [PubMed] [Google Scholar]
  • 8.Shao, Y. et al. Comprehensive metabolic profiling of Parkinson’s disease by liquid chromatography-mass spectrometry. Mol. Neurodegeneration. 16, 4. 10.1186/s13024-021-00425-8 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Zhao, H. et al. Potential biomarkers of Parkinson’s disease revealed by plasma metabolic profiling. J. Chromatogr. B. 1081–1082, 101–108. 10.1016/j.jchromb.2018.01.025 (2018). [DOI] [PubMed] [Google Scholar]
  • 10.Kremer, T. et al. Longitudinal analysis of multiple neurotransmitter metabolites in cerebrospinal fluid in early Parkinson’s disease. Mov. Disord.36, 1972–1978. 10.1002/mds.28608 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Willkommen, D. et al. Metabolomic investigations in cerebrospinal fluid of Parkinson’s disease. PLoS ONE. 13, e0208752. 10.1371/journal.pone.0208752 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.LeWitt, P. A. et al. 3-hydroxykynurenine and other Parkinson’s disease biomarkers discovered by metabolomic analysis. Mov. Disord.28, 1653–1660. 10.1002/mds.25555 (2013). [DOI] [PubMed] [Google Scholar]
  • 13.Stoessel, D. et al. Promising metabolite profiles in the plasma and CSF of early clinical Parkinson’s disease. Front. Aging Neurosci.10, 51. 10.3389/fnagi.2018.00051 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Ibanez, L. et al. A comprehensive analysis of dementia cerebrospinal fluid biomarkers using GWAs, polygenic risk scores and Mendelian randomization in Parkinson’s disease: Genetics/atypical and other dementias. Alzheimer’s Dement.16, e043233. 10.1002/alz.043233 (2020). [Google Scholar]
  • 15.Chen, Y. et al. Genomic atlas of the plasma metabolome prioritizes metabolites implicated in human diseases. Nat. Genet.55, 44–53. 10.1038/s41588-022-01270-1 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Panyard, D. J. et al. Cerebrospinal fluid metabolomics identifies 19 brain-related phenotype associations. Commun. Biol.4, 63. 10.1038/s42003-020-01583-z (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Nalls, M. A. et al. Identification of novel risk loci, causal insights, and heritable risk for Parkinson’s disease: a meta-analysis of genome-wide association studies. Lancet Neurol.18, 1091–1102. 10.1016/S1474-4422(19)30320-5 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Duan, A. et al. Metabolome-wide Mendelian randomization assessing the causal role of serum and cerebrospinal metabolites in traumatic brain injury. Biomedicines12, 1178. 10.3390/biomedicines12061178 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Zhou, Z. & Leng, H. Deciphering the causal relationship between plasma and cerebrospinal fluid metabolites and glioblastoma multiforme: a Mendelian randomization study. Aging10.18632/aging.205818 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Botelho, J., Machado, V., Mendes, J. J. & Mascarenhas, P. Causal association between periodontitis and Parkinson’s disease: a bidirectional Mendelian randomization study. 8 (2021). [DOI] [PMC free article] [PubMed]
  • 21.Lawlor, D. A. et al. Mendelian randomization: using genes as instruments for making causal inferences in epidemiology. Statist Med.27, 1133–1163. 10.1002/sim.3034 (2008). [DOI] [PubMed] [Google Scholar]
  • 22.Hemani, G. et al. The MR-Base platform supports systematic causal inference across the human phenome. eLife7, e34408. 10.7554/eLife.34408 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Estiar, M. A. et al. Lack of causal effects or genetic correlation between restless legs syndrome and Parkinson’s disease. Mov. Disord. 36, 1967–1972. 10.1002/mds.28640 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Coles, L. D. et al. Repeated-dose oral N‐acetylcysteine in Parkinson’s disease: pharmacokinetics and effect on brain glutathione and oxidative stress. J. Clin. Pharma. 58, 158–167. 10.1002/jcph.1008 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Oxenkrug, G. Peripheral tryptophan-kynurenine metabolism associated with metabolic syndrome is different in Parkinson’s and Alzheimer’s diseases (2017). [PMC free article] [PubMed]
  • 26.on behalf of AddNeuroMed consortium et al. Metabolic phenotyping reveals a reduction in the bioavailability of serotonin and kynurenine pathway metabolites in both the urine and serum of individuals living with Alzheimer’s disease. Alz Res. Therapy. 13, 20. 10.1186/s13195-020-00741-z (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Toomey, C. E. et al. Mitochondrial dysfunction is a key pathological driver of early stage Parkinson’s. Acta neuropathol. commun.10, 134. 10.1186/s40478-022-01424-6 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Pérez-Cerdá, C. et al. A new case of succinyl‐CoA: acetoacetate transferase deficiency. J. Inher Metab. Disea. 15, 371–373. 10.1007/BF02435979 (1992). [DOI] [PubMed] [Google Scholar]
  • 29.Wang, C. et al. Genetic architecture of cerebrospinal fluid and brain metabolite levels and the genetic colocalization of metabolites with human traits. Nat. Genet.10.1038/s41588-024-01973-7 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Sakakibara, Y., Suiko, M., Nakajima, H. & Liu, M. C. Sulphation of l-tyrosine in mammalian cells: a comparative study. Biochem. J.305, 993–998. 10.1042/bj3050993 (1995). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Hirayama, M. et al. Serum tyrosine-to-phenylalanine ratio is low in Parkinson’s disease. J. Parkinson’s Disease. 6, 423–431. 10.3233/JPD-150736 (2016). [DOI] [PubMed] [Google Scholar]
  • 32.Cheng, L. et al. Exploring the causal link between serum amino acids and Parkinson’s disease: a Mendelian randomization approach. Sci. Rep.14, 30271. 10.1038/s41598-024-81787-z (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Rutkowsky, J. M. et al. Acylcarnitines activate proinflammatory signaling pathways. Am. J. Physiology-Endocrinology Metabolism. 306, E1378–E1387. 10.1152/ajpendo.00656.2013 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Alza, N. P., Conde, M. A., Scodelaro-Bilbao, P. G. & Salvador, G. A. Neutral lipids as early biomarkers of cellular fate: the case of α-synuclein overexpression. Cell. Death Dis.1210.1038/s41419-020-03254-7 (2021). [DOI] [PMC free article] [PubMed]
  • 35.Chan, R. B. et al. Elevated GM3 plasma concentration in idiopathic Parkinson’s disease: A lipidomic analysis. PLoS ONE. 12, e0172348. 10.1371/journal.pone.0172348 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Hu, L. et al. Integrated metabolomics and proteomics analysis reveals plasma lipid metabolic disturbance in patients with Parkinson’s disease. Front. Mol. Neurosci.1310.3389/fnmol.2020.00080 (2020). [DOI] [PMC free article] [PubMed]
  • 37.Scholefield, M. et al. Severe and regionally widespread increases in tissue urea in the human brain represent a novel finding of pathogenic potential in Parkinson’s disease dementia. Front. Mol. Neurosci.14, 711396. 10.3389/fnmol.2021.711396 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Yasunishi, M. et al. A case of hyperammonemia resulting from urinary tract infection caused by urease-producing bacteria in a Parkinson’s disease patient with drug-induced urinary retention. Jpn j. Geriat. 54, 560–566. 10.3143/geriatrics.54.560 (2017). [DOI] [PubMed] [Google Scholar]
  • 39.Yang, G. et al. Study on the correlation between blood urea nitrogen, creatinine level, proteinuria and Parkinson’s disease. Neurol. India. 71, 1217–1221. 10.4103/0028-3886.391388 (2023). [DOI] [PubMed] [Google Scholar]
  • 40.Liao, J. et al. Causal association between cerebrospinal fluid metabolites and Parkinson’s disease: A two-sample bidirectional Mendelian randomization study. Behav. Brain. Res.482, 115426. 10.1016/j.bbr.2025.115426 (2025). [DOI] [PubMed] [Google Scholar]
  • 41.Zhang, A. et al. Causal association between plasma metabolites and neurodegenerative diseases. Prog. Neuropsychopharmacol. Biol. Psychiatry. 134, 111067. 10.1016/j.pnpbp.2024.111067 (2024). [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Material 1 (3.4MB, zip)

Data Availability Statement

All data generated or analyzed during this study are included in this published article and its Supplementary Materials.

The underlying code for this study is available at https://github.com/598047655/MR_metabolomics.

Articles from Scientific Reports are provided here courtesy of Nature Publishing Group

RESOURCES