Proteogenomic network analysis reveals dysregulated mechanisms and potential mediators in Parkinson's disease

Abstract

Parkinson's disease is highly heterogeneous across disease symptoms, clinical manifestations and progression trajectories, hampering the identification of therapeutic targets. Despite knowledge gleaned from genetics analysis, dysregulated proteome mechanisms stemming from genetic aberrations remain underexplored. In this study, we develop a three-phase system-level proteogenomic analytical framework to characterize disease-associated proteins and dysregulated mechanisms. Proteogenomic analysis identified 577 proteins that enrich for Parkinson's disease-related pathways, such as cytokine receptor interactions and lysosomal function. Converging lines of evidence identified nine proteins, including LGALS3, CSNK2A1, SMPD3, STX4, APOA2, PAFAH1B3, LDLR, HSPB1, BRK1, with potential roles in disease pathogenesis. This study leverages the largest population-scale proteomics dataset, the UK Biobank Pharma Proteomics Project, to characterize genetically-driven protein disturbances associated with Parkinson's disease. Taken together, our work contributes to better understanding of genome-proteome dynamics in Parkinson's disease and sets a paradigm to identify potential indirect mediators connected to GWAS signals for complex neurodegenerative disorders.

Fig. 1. An overview of the study design.

The study consists of three stages, including proteogenomic analysis, network analysis, and epigenomic data analysis.

Fig. 2. Manhattan plots for PD case-control GWAS using the EHR-based clinically defined cohorts from the UKB and the UKB-FinnGen meta-analysis in European samples.

(a) UKB case-control GWAS, (b) UKB-FinnGen GWAS meta-analysis. A solid blue line indicates the significance threshold P < 5×10^-8. The dashed line shows P = 10^-6.

Fig. 3. Chromosomal positions of the identified pQTLs relative to their associated proteins.

a Position of pQTLs and their associated proteins; b PD-associated SNPs and their associated proteins.

Fig. 4. Overview of blood plasma pQTL architecture in PD.

a histogram of the number of PD-associated pQTLs per protein; b histogram of the number of associated proteins per pQTL; c–d number of associations in panels a, b separated by the association type.

Fig. 5. Network analysis results.

a hypothesis-free analysis: aggregated view of the top 1% of the identified PD-associated protein modules. Red nodes represent GWAS signals, green nodes represent pQTL-associated proteins in PD, dark blue nodes are pQTL-associated proteins which are PD-associated GWAS loci, and the size of the nodes is proportional to the gene-based scores generated by MAGMA; b hypothesis-driven analysis: aggregated view of the identified loci from UKB-FinnGen meta-analysis GWAS (shown in red) and their first-degree neighbors in the PPI network. Green nodes represent pQTL-associated proteins in PD.

Fig. 6. Expression profiles of PD-associated genes among PD, LBD, and healthy individuals.

a Expression of PD-associated genes identified through UKB GWAS and UKB-FinnGen meta-analysis GWAS; b–d Expression of the gene encoding the identified proteins from the network analyzes in dopamine (DA) neurons, microglia, and astrocytes, respectively. Source data are provided as a Source Data file.

Fig. 7. Genotype-specific abundance of LGALS3 among carriers of the protective SNP rs11158026 among.

a incident+prevalent PD cases; b prevalent PD cases only; c incident PD cases only; d cases and controls. NPX: Normalized Protein Expression. In (d) the p-values indicated at the bottom of the figure represent the differences between the NPX values once combining incident, prevalent and controls in each genotype. All data panels are represented as median values plus the first and third quartiles. Two-sided ANCOVA test was performed to evaluate the genotype-specific abundance of LGALS3 adjusted for age and sex. Source data are provided as a Source Data file.