An Innovative Lipidomic Workflow to Investigate the Lipid Profile in a Cystic Fibrosis Cell Line

Abstract

Altered lipid metabolism has been associated to cystic fibrosis disease, which is characterized by chronic lung inflammation and various organs dysfunction. Here, we present the validation of an untargeted lipidomics approach based on high-resolution mass spectrometry aimed at identifying those lipid species that unequivocally sign CF pathophysiology. Of n.13375 mass spectra recorded on cystic fibrosis bronchial epithelial airways epithelial cells IB3, n.7787 presented the MS/MS data, and, after software and manual validation, the final number of annotated lipids was restricted to n.1159. On these lipids, univariate and multivariate statistical approaches were employed in order to select relevant lipids for cellular phenotype discrimination between cystic fibrosis and HBE healthy cells. In cystic fibrosis IB3 cells, a pervasive alteration in the lipid metabolism revealed changes in the classes of ether-linked phospholipids, cholesterol esters, and glycosylated sphingolipids. Through functions association, it was evidenced that lipids variation involves the moiety implicated in membrane composition, endoplasmic reticulum, mitochondria compartments, and chemical and biophysical lipids properties. This study provides a new perspective in understanding the pathogenesis of cystic fibrosis and strengthens the need to use a validated mass spectrometry-based lipidomics approach for the discovery of potential biomarkers and perturbed metabolism.

Keywords: OMICS; biomarker; cell structure; cystic fibrosis; lipidomics; membrane composition; sphingolipid; untargeted analysis.

Figure 1

Cumulative frequency distribution of the coefficient of variations (CVs) (%) in (A) pool samples, (B) healthy HBE, and (C) cystic fibrosis (CF) cell extracts obtained for the precision evaluation of different normalization protocols comprehensive of results from both polarities. The dotted line indicates the separation between features within 30% of the CV, which is intended as the maximum permitted for the validation. The graphs showed the better performance of Lowess coupled to µg proteins as the normalization technique, reaching (A) 89%, (B) 64%, and (C) 68% in acceptable features (with a CV% inferior to 30%). (D) Graphs show the mean ± SD of the percentage of acceptable features (with a CV% inferior to 30%) between the different normalization techniques.

Figure 2

Lipid content comparison between healthy epithelial (H, n = 3 independent biological replicates) vs. cystic fibrosis (CF, n = 3 independent biological replicates) cells. Graphs represent the lipid amount (Amount, mean ± SD), which indicates the sum of the metabolites intensities within a class after normalization (see Equation (1)). Two-tailed unpaired t-tests were performed in each lipid class to establish a statistical difference (* p ≤ 0.05; ** p ≤ 0.01; *** p ≤ 0.001; **** p ≤ 0.0001).

Figure 3

The volcano plot is a combination of fold change and t-tests: X-axis is log₂(fold change, FC), and Y-axis is −log₁₀ (adjusted for false discovery rate). Dots indicate features that presented both a FC >2 and p-value < 0.05. Lipids in pink and green are reduced (n.118) and augmented (n.514) in CF vs. H, respectively.

Figure 4

(A) Partial least squares discriminant analysis (PLS-DA) chemometric analysis. (B) Box and whiskers plots (line at median, and box stretched from the 25–75th percentiles; whiskers indicated the 10–90th, whereas outliers were plotted as single points) of the discriminant lipids (n = 624) subdivided for lipid classes and evaluated by their IF scores (see Equation (3)). Grey boxes designated lipid classes that displayed an IF< cut-off (visualized as a dotted line and calculated as the lower confidence limit of the median of the features considered).

Figure 5

(A) Enrichment analysis (top 40) of CF vs. H phenotypes. The dotted line indicates the cut-off value of significant enrichments (q < 0.05). Bar length is related with the enrichment (−log q-values corrected for false discovery rate, FDR), whereas colors are dependent to the type of the enrichment: lipid function, cellular component, and physical-chemical properties. (B) Distribution of the acyl chain unsaturation from all lipid fraction (%) in CF vs. H discriminant lipid group (top 100). (C) Distribution of the ester and ether linkages in phospholipids (%) in CF vs. H in the discriminant lipid group (top 100). In (B) the chi-square test and in (C) binomial test, revealed a p-value < 0.05.

Figure 6

Overview of the lipid biosynthetic and metabolic pathways. Colored dots represented the lipid changes in CF bronchial epithelial cells.