A metabolomics study in aqueous humor discloses altered arginine metabolism in Parkinson's disease

Abstract

Background: The lack of accessible and informative biomarkers results in a delayed diagnosis of Parkinson's disease (PD), whose symptoms appear when a significant number of dopaminergic neurons have already disappeared. The retina, a historically overlooked part of the central nervous system (CNS), has gained recent attention. It has been discovered that the composition of cerebrospinal fluid influences the aqueous humor composition through microfluidic circulation. In addition, alterations found in the brain of patients with PD have a correlate in the retina. This new paradigm highlights the potential of the aqueous humor as a sample for identifying differentially concentrated metabolites that could, eventually, become biomarkers if also found altered in blood or CSF of patients. In this research we aim at analyzing the composition of the aqueous humor from healthy controls and PD patients.

Methods: A targeted metabolomics approach with concentration determination by mass spectrometry was used. Statistical methods including principal component analysis and linear discriminants were used to select differentially concentrated metabolites that allow distinguishing patients from controls.

Results: In this first metabolomics study in the aqueous humor of PD patients, elevated levels of 16 compounds were found; molecules differentially concentrated grouped into biogenic amines, amino acids, and acylcarnitines. A biogenic amine, putrescine, alone could be a metabolite capable of differentiating between PD and control samples. The altered levels of the metabolites were correlated, suggesting that the elevations stem from a common mechanism involving arginine metabolism.

Conclusions: A combination of three metabolites, putrescine, tyrosine, and carnitine was able to correctly classify healthy participants from PD patients. Altered metabolite levels suggest altered arginine metabolism. The pattern of metabolomic disturbances was not due to the levodopa-based dopamine replacement medication because one of the patients was not yet taking levodopa but a dopamine receptor agonist.

Keywords: Biogenic amines; Carnitine; Eye; Levodopa; Linear discrimination; Mass spectrometry; Putrescine; Sensitivity; Spermidine.

Fig. 1

Univariate analysis of metabolites in the AH of the two groups. Median in box-and-whisker plots (whiskers indicate the furthest data points in lower and upper ranges determined by 1.5 times the interquartile range in the two groups, control and parkinsonian). (A) Biogenic amines. (B) Amino acids. (C) Acylcarnitines

Fig. 2

Forest plot showing confidence intervals at the 95% level of the AUC of the differentially concentrated metabolites. The minimal and maximal interval values are in the last two columns of Table 1

Fig. 3

Multivariant metabolite analyses. (A) Principal component analysis diagram. The space and data for controls are in yellow and the space and data for PD individuals are in blue. Principal component (PC) 1, PC2, and PC3 explain, respectively, the 63.8%, the 12.0%, and the 9.0% of the between-group variance. (B) Dendrogram and heatmap. It shows the similarity index between the differentially concentrated metabolites in PD patients. The similarity index was obtained using Pearson’s correlation coefficient (the darker the blue, the higher the similarity index). Compounds displaying a similarity index greater than 0.5 are in blue the darker the red, the lower the similarity index). (C) Linear discrimination analysis. Probability of being classified as a control using the LDF1 function with cross-validation. The colors of the bars indicate the classification result (gold: control (C); blue: PD patient (P)). The overall performance of the cross-validated function is 93.75%

Fig. 4

Metabolic pathways in which differentially concentrated compounds are involved. The diameter of every point red indicates the enrichment ratio (value in PD versus value in control). The greater the reddish color, the lower the p-value. The enrichment ratio is calculated as the ratio between the number of metabolites in the dataset that are involved in a specific metabolic pathway and the total number of metabolites associated with that pathway; the p-value is obtained using the hypergeometric test

Fig. 5

Metabolic pathway map summarizing the results. The colored boxes indicate the different metabolic processes included in the scheme. Metabolites differentially concentrated in PD samples are in red