Metabolomic profiling of saliva from cystic fibrosis patients

Abstract

The development of targeted therapies that correct the effect of mutations in patients with cystic fibrosis (CF) and the relevant heterogeneity of the clinical expression of the disease require biomarkers correlated to the severity of the disease useful for monitoring the therapeutic effects. We applied a targeted metabolomic approach by LC-MS/MS on saliva samples from 70 adult CF patients and 63 age/sex-matched controls to investigate alterations in metabolic pathways related to pancreatic insufficiency (PI), Pseudomonas aeruginosa (PA) colonization, CF liver disease (CFLD), and CF related diabetes (CFRD). Sixty salivary metabolites were differentially expressed, with 11 being less abundant and 49 more abundant in CF patients. Among these, the most relevant alterations involved salivary ADMA, N-acetylornithine, methionine and methionine sulfoxide levels. Furthermore, methionine was significantly lower in CF patients with PI and salivary histamine levels were significantly lower in patients colonized by PA. Moreover, ADMA as well as N-acetylornithine and methionine were significantly lower in CF patients with CFRD than in patients without CFRD. Finally, the levels of DOPA resulted significantly lower in saliva from patients with liver disease. Our study revealed an imbalance in arginine methylation and tryptophan pathway related to CFRD and PI as well as alterations in dopaminergic pathway and Krebs cycle related to CFLD. This study also highlights different salivary metabolites as new potential biomarkers in a non-invasive sample that could represent a useful tool for the stratification and management of CF patients.

Keywords: Pseudomonas aeruginosa colonization; CF liver disease; CF related diabetes; Cystic fibrosis; Metabolomics; Pancreatic insufficiency.

Fig. 1

Comparative metabolome analysis of patients with CF (n = 70) and healthy controls (CTRL, n = 63). (A) The supervised partial least squares-discriminant analysis (PLS-DA) shows the segregation of the CF and CTRL groups. (B) The heatmap reports metabolites concentrations in each group. The intensity of the colored boxes represents the relative abundance of each molecule, after log10-transformation, and Pareto-scaling. (C) The discriminant features in CF were identified according to the Variable Importance in Projection (VIP) score. The 30 most important molecules were selected in the plot. The intensity of the colored boxes represents the relative abundance in each group after log10-transformation, and Pareto-scaling. (D) Volcano plot analysis of the significantly changed metabolites in the comparison CF versus CTRL. The purple and orange dots represent the significantly increased and decreased metabolites, respectively. Non-colored dots refer to all the other molecules identified in the dataset with not significant difference between the groups. (E) Pathway analysis plots with details of the most significant up-regulated pathways (left plot) and down-regulated pathways (right plot) in CF patients.

Fig. 2

Metabolomic signatures of CF samples according to their clinical phenotype (I). Heatmaps show the average metabolite abundance and compare CTRL and each of the two CF groups. Patients were subdivided as affected and non-affected by a specific clinical feature, as follows: pancreatic insufficiency (PI) and sufficiency (PS); with and without colonization by P. aeruginosa (PA and without_PA); with and without CF liver disease (LD and without_LD); with and without CF related diabetes (RD and without_RD).

Fig. 3

Metabolomic signatures of CF samples according to their clinical phenotype (II). PLS-DA shows group segregation in the two CF groups. Patients were subdivided as affected and non-affected by a specific clinical feature, as follows: pancreatic insufficiency (PI) and sufficiency (PS); with and without colonization by P. aeruginosa (PA and without_PA); with and without CF liver disease (LD and without_LD); with and without CF related diabetes (RD and without_RD).

Fig. 4

Metabolomic signatures of CF samples according to their clinical phenotype (III). VIP molecules were identified to find metabolites discriminating the specific CF phenotype. Patients were subdivided as affected and non-affected by a specific clinical feature, as follows: pancreatic insufficiency (PI) and sufficiency (PS); with and without colonization by P. aeruginosa (PA and without_PA); with and without CF liver disease (LD and without_LD); with and without CF related diabetes (RD and without_RD).

Fig. 5

Metabolomic signatures of CF samples according to their clinical phenotype (IV). Abundance violin plots of the most significant VIP metabolites (VIP > 2) per CF phenotype. Patients were subdivided as affected and non-affected by a specific clinical feature, as follows: pancreatic insufficiency (PI) and sufficiency (PS); with and without colonization by P. aeruginosa (PA and without_PA); with and without CF liver disease (LD and without_LD); with and without CF related diabetes (RD and without_RD).

Fig. 6

Correlation analysis of the saliva metabolome with the FEV1(%) parameter of CF patients. (A) Correlation matrix shows the possible association of each molecule (divided by chemical class) with the FEV1(%) parameter. (B) Significant hits obtained by simple linear regression analysis were plotted showing positive correlation with the FEV1(%) parameter.