Proteome‐wide effects and causality of familial Parkinson’s disease

Abstract

AbstractBackgroundThe causal mediators of phenotype‐associated genetic variation in ∼75 genetic loci associated with Parkinson’s disease (PD) risk, including LRRK2 and GBA1 key loci, remain unresolved. We integrated the cerebrospinal fluid (CSF) proteome and genomic data to uncover causal associations in genetically‐defined PD.MethodsCSF proteome from the Parkinson’s Progression Markers Initiative (PPMI) cohort (N = 1158) was quantified using the 5K SOMAscan assay. We integrated whole genome sequencing data of 683 samples to identify genetic variants associated with protein levels (protein quantitative trait loci; pQTLs). pQTLs within a region of ± 1Mb from the sentinel SNP as cis‐pQTLs, otherwise, as trans‐pQTLs. We sought pleiotropic locus associated with multiple proteins. We used the latest meta‐analysis from Nalls et al.[1] as an outcome to implement two‐sample Mendelian randomization using TwoSampleMR[2]. We identified colocalized proteins for PD using coloc[3,4,5]. We identified associated proteins with sporadic PD (sPD), and PD with mutations in LRRK2, and GBA1. The pathway analysis was performed using pathfindR[6]. We built elastic‐net logistic regression models to select protein combinations for sPD, LRRK2+ PD, and GBA1+ PD prediction using glmnet[7]. Plasma proteome data from the Penn Udall and Multi‐site PDBP were used as replication cohorts[8].ResultsWe found 701 cis‐ and 420 trans‐pQTLs, and 23 causal proteins; eight of which colocalized with PD risk loci. We found 92 pleiotropic loci, including the APOE and LRRK2 loci. We found 59, 271, and 127 proteins associated with sPD, LRRK2+ PD, and GBA1+ PD, respectively. The predictive models with 33, 154, and 66 proteins for precisely classifying sPD vs. controls (AUC = 0.835; PR‐AUC = 0.898; F1‐score = 0.852), LRRK2+ PD vs. sPD (AUC = 0.95; PR‐AUC = 0.918; F1‐score = 0.828), and GBA1+ PD vs. sPD (AUC = 0.877; PR‐AUC = 0.776; F1‐score = 0.688), respectively. LRRK2+ and GBA1+ PD associated proteins were enriched for dopaminergic synapse, proteasome, and lysosome pathways, and sPD‐associated proteins were enriched for neurotrophin, endocytosis, and chemokine signaling. We found 25 PD‐associated proteins in the Upenn Udall Discovery cohort and replicated 11 proteins in the PDBP replication cohort.ConclusionsOur systems biology approach discovered causal PD proteins which could be used as drug targets and offers a novel methodology to improve our understanding of PD pathogenesis and prediction.