Comprehensive analysis of the cerebrospinal fluid and serum metabolome in neurological diseases

Abstract

Background Comprehensive characterization of the metabolome in cerebrospinal fluid (CSF) and serum by Nuclear Magnetic Resonance (NMR) spectroscopy may identify biomarkers and contribute to the understanding of the pathophysiology of neurological diseases. Methods Metabolites were determined by NMR spectroscopy in stored CSF/serum samples of 20 patients with Parkinson’s disease, 25 patients with other neuro-degenerative diseases, 22 patients with cerebral ischemia, 48 patients with multiple sclerosis, and 58 control patients with normal CSF findings. The data set was analysed using descriptive and multivariate statistics, as well as machine learning models. Results CSF glucose and lactic acid measured by NMR spectroscopy and routine clinical chemistry showed a strong correlation between both methods (glucose, R2 = 0.87, n = 173; lactic acid, R2 = 0.74, n = 173). NMR spectroscopy detected a total of 99 metabolites; 51 in both, CSF and serum, 16 in CSF only, and 32 in serum only. CSF concentrations of some metabolites increased with age and/or decreasing blood–brain-barrier function. Metabolite detection rates were overall similar among the different disease groups. However, in two-group comparisons, absolute metabolite levels in CSF and serum discriminated between multiple sclerosis and neurodegenerative diseases (area under the curve (AUC) = 0.96), multiple sclerosis and Parkinson’s disease (AUC = 0.89), and Parkinson’s disease and control patients (AUC = 0.91), as demonstrated by random forest statistical models. Orthogonal partial least square discriminant analysis using absolute metabolite levels in CSF and serum furthermore permitted separation of Parkinson’s disease and neurodegenerative diseases. CSF propionic acid levels were about fourfold lower in Parkinson’s disease as compared to neurodegenerative diseases. Conclusions These findings outline the landscape of the CSF and serum metabolome in different categories of neurological diseases and identify age and blood–brain-barrier function as relevant co-factors for CSF levels of certain metabolites. Metabolome profiles as determined by NMR spectroscopy may potentially aid in differentiating groups of patients with different neurological diseases, including clinically meaningful differentiations, such as Parkinson’s disease from other neurodegenerative diseases.