Identification of lipid biomarker from serum in patients with chronic obstructive pulmonary disease

Abstract

Background: Chronic obstructive pulmonary disease (COPD) is the third leading cause of death in the United States with no effective treatment. The current diagnostic method, spirometry, does not accurately reflect the severity of COPD disease status. Therefore, there is a pressing unmet medical need to develop noninvasive methods and reliable biomarkers to detect early stages of COPD. Lipids are the fundamental components of cell membranes, and dysregulation of lipids was proven to be associated with COPD. Lipidomics is a comprehensive approach to all the pathways and networks of cellular lipids in biological systems. It is widely used for disease diagnosis, biomarker identification, and pathology disorders detection relating to lipid metabolism.

Methods: In the current study, a total of 25 serum samples were collected from 5 normal control subjects and 20 patients with different stages of COPD according to the global initiative for chronic obstructive lung disease (GOLD) (GOLD stages I ~ IV, 5 patients per group). After metabolite extraction, lipidomic analysis was performed using electrospray ionization mass spectrometry (ESI-MS) to detect the serum lipid species. Later, the comparisons of individual lipids were performed between controls and patients with COPD. Orthogonal projections to latent structures discriminant analysis (OPLS-DA) and receiver operating characteristic (ROC) analysis were utilized to test the potential biomarkers. Finally, correlations between the validated lipidomic biomarkers and disease stages, age, FEV1% pack years and BMI were evaluated.

Results: Our results indicate that a panel of 50 lipid metabolites including phospholipids, sphingolipids, glycerolipids, and cholesterol esters can be used to differentiate the presence of COPD. Among them, 10 individual lipid species showed significance (p < 0.05) with a two-fold change. In addition, lipid ratios between every two lipid species were also evaluated as potential biomarkers. Further multivariate data analysis and receiver operating characteristic (ROC: 0.83 ~ 0.99) analysis suggest that four lipid species (AUC:0.86 ~ 0.95) and ten lipid ratios could be potential biomarkers for COPD (AUC:0.94 ~ 1) with higher sensitivity and specificity. Further correlation analyses indicate these potential biomarkers were not affected age, BMI, stages and FEV1%, but were associated with smoking pack years.

Conclusion: Using lipidomics and statistical methods, we identified unique lipid signatures as potential biomarkers for diagnosis of COPD. Further validation studies of these potential biomarkers with large population may elucidate their roles in the development of COPD.

Keywords: Biomarkers; Chronic obstructive pulmonary disease (COPD); Lipidomics; OPLS-DA; Receiver operating characteristic.

Fig. 1

Lipidomic analysis of serum samples from patients with different severity of COPD. a. Relative lipid species composition in serum samples by electrospray ionization mass spectrometry (ESI-MS); b. Venn diagram of 50 lipid species with either FDR-adjusted significance (p < 0.05, blue with 17 lipid species) or two-times fold change (green with 43 lipids species); c. Detailed fold change of 50 lipid species: 10 lipid species with both two-times fold change and p < 0.05 (orange bar), 32 lipid species with two-times fold change only (green bar), 7 lipid species with significant change (blue bar) only

Fig. 2

OPLS-DA evaluation of potential biomarker candidates. OPLS-DA analysis of1 0 lipid ratios candidates: R2 = 0.619 and Q2 = 0.735. Ellipses display 95% confidence regions. Serum samples from control are in red and samples from patients with COPD are in green

Fig. 3

The ROC analysis from representative lipid species and lipid ratio biomarker candidates. The performance of each biomarker model was evaluated by the area under the ROC curve (AUC) and the determination of specificity (X-axis) and sensitivity (Y-axis) at the optimal cut-off point defined by the minimum distance to the top-left corner. a. Representative lipid species with AUC value of PC (32:1), C 16:1 CE, PI (36:6) were 0.81, 0.885, 0.95 (Top); b. Representative lipid ratio with AUC value of PI (36:2)/C 16:1 CE, PC (32:0)/C 16:1 CE, PC (40:4)/ePC (38:2) were 0.99, 0.95, 0.95 (Bottom)

Fig. 4

Representative lipid ratio PI (38:4)/C16:1 CE with FEV1%, Pack-Years, and Age. a. The lipid ratio could distinguish COPD patients and healthy people; The lipid ratio of PI (38:4)/C16:1 CE was not related to either smoking status (b) or the age of the patients (c)

Fig. 5

Potential mechanistic scheme of PC/CE ratio identified in sera of patients with COPD