Changes in the cerebrospinal fluid proteome precede and stratify the course of Parkinson's Disease

The proteome encodes for various information. By quantifying 4071 human proteins in the cerebrospinal fluid using high-throughput affinity proteomics, this study aimed to extract proteomic signatures associated with Parkinsons disease using unbiased machine-learning and to examine their impact on Parkinsons disease course. From the Parkinsons Progression Markers Initiative, we first included 312 drug-naive patients with Parkinsons disease without GBA1, LRRK2, and SNCA mutations (non-genetic Parkinsons disease) and 161 healthy controls. Differentially expressed protein analysis identified 149 proteins that were likely to be differentially expressed with a false discovery rate of [≤] 0.2. Subsequently, a logistic regression analysis with the least absolute shrinkage and selection operator created a 55-protein-based model that quantified the degree to which non-genetic Parkinsons disease-associated cerebrospinal fluid proteomic signatures were present in each participant in the form of a score named a "non-genetic Parkinsons disease-associated proteomic score." This score accurately distinguished non-genetic Parkinsons disease from healthy controls in both a derivation cohort (area under the curve = 0.91 [95% confidence interval, 0.89-0.94]) and an independent validation cohort comprising 38 drug-naive patients with non-genetic Parkinsons disease and 15 healthy controls (area under the curve = 0.87 [95% confidence interval, 0.76-0.99]). Notably, this score was also significantly increased in patients with Parkinsons disease harboring GBA1, LRRK2, or SNCA mutations (n = 258), albeit to varying degrees depending on the type of mutation. Furthermore, the score of genetic prodromal individuals (n = 365) was intermediate between that of healthy controls and patients with Parkinsons disease. Next, cross-sectional correlation analyses revealed that regardless of the presence or absence of genetic mutations, this score was significantly correlated with several baseline clinical parameters and biomarkers. Furthermore, longitudinal survival analyses revealed that even after adjustment of baseline characteristics, this score significantly predicted progression to important clinical milestones including mild cognitive impairment, dementia, and Hoehn and Yahr stage IV. Finally, longitudinal analyses using linear-mixed effects models confirmed that even after adjustment of baseline characteristics, the score was significantly associated with subsequent motor and cognitive decline. Collectively, our study demonstrated that cerebrospinal fluid proteomic signatures associated with non-genetic Parkinsons disease could be quantified using a novel non-genetic Parkinsons disease-associated proteomic score. Furthermore, we identified the presence of these signatures in genetic Parkinsons disease to varying degrees depending on the type of mutation. Moreover, these signatures appeared from the prodromal stage and were robustly linked to both subsequent motor and cognitive decline in early Parkinsons disease, indicating that the cerebrospinal fluid proteome encodes important information for both the onset and progression of Parkinsons disease.