Integrated proteomics analysis in cerebrospinal fluid and saliva reveals the changes of endopeptidase activity in Parkinson`s disease

Abstract Background The prodromal phase of Parkinson's disease can reach 10–20 years, and clinically meaningful biomarkers associated with Parkinson's disease (PD) have not been developed based on biofluid samples. Here, we aimed to identify novel biofulid candidate biomarkers by integrating CSF and saliva proteomes in PD. Methods We used a high-throughput tandem mass spectrometry to analyze 120 CSF samples and 203 saliva samples. Weighted gene co-expression network analysis (WGCNA) were performed to determine the protein features that are significantly correlated with the clinical parameters. Additionally, We used machine learning techniques to identify candidate biomarkers for PD diagnose. Results In total, we identified 2,585 and 4,301 proteins in CSF and saliva, respectively. Among these proteins, 10 differentially expressed proteins (DEPs) were in common between CSF and saliva proteome, mainly involved in the negative regulation of endopeptidase activity as well as hyaluronan metabolic. Interestingly, persistent activation of the negative regulation of endopeptidase activity during the progression of PD. WGCNA analysis revealed a significant negative correlation between AGT protein and UPDRS score in both CSF and saliva. In addition, machine learning identified a combination of 5 protein (GAPDH, GNS, ITIH2, CTSL, and GPX3) as biomarkers for PD, with an area-under-the-curve (AUC) of 0.877–0.958. Conclusion In summary, we integrated and analyzed the proteomes of CSF and saliva proteomes, confirming that CSF and saliva proteome could both reflect the occurrence and development of PD. Furthermore, we found that endopeptidase activity might be a potential pathogenesis of PD, especially during the progression of disease. In addition, we also discovered valuable candidate biomarkers for PD diagnosis.