Metabolic profiling of idiopathic pulmonary fibrosis in a mouse model: implications for pathogenesis and biomarker discovery

Abstract

Background: Alterations in metabolites and metabolic pathways are thought to be important triggers of idiopathic pulmonary fibrosis (IPF), but our lack of a comprehensive understanding of this process has hampered the development of IPF-targeted drugs.

Methods: To fully understand the metabolic profile of IPF, C57BL/6 J male mice were injected intratracheally with bleomycin so that it could be used to construct a mouse model of IPF, and lung tissues from 28-day and control IPF mice were analyzed by pathology and immunohistochemistry. In addition, serum metabolites from IPF mice were examined using LC-ESI-MS/MS, and the differential metabolites were analyzed for KEGG metabolic pathways and screened for biomarkers using machine learning algorithms.

Results: In total, the levels of 1465 metabolites were detected, of which 104 metabolites were significantly altered after IPF formation. In IPF mouse serum, 52% of metabolite expression was downregulated, with lipids (e.g., GP, FA) and organic acids and their derivatives together accounting for more than 70% of the downregulated differentially expressed metabolites. In contrast, FA and oxidised lipids together accounted for 60% of the up-regulated differentially expressed metabolites. KEGG pathway enrichment analyses of differential metabolites were mainly enriched in the biosynthesis of unsaturated fatty acids, pentose phosphate pathway, and alanine, aspartate, and glutamate metabolism. Seven metabolites were screened by machine learning LASSO models and evaluated as ideal diagnostic tools by receiver operating characteristic curves (ROCs).

Discussion: In conclusion, the serum metabolic disorders found to be associated with pulmonary fibrosis formation will help to deepen our understanding of the pathogenesis of pulmonary fibrosis.

Keywords: biomarkers; machine learning; metabolites; mice; pulmonary fibrosis.

Figure 1

Bleomycin mouse models. (A) Study design, intratracheal instillation of bleomycin or saline (NaCl) was administered to anesthetized mice at day 0. Lung tissues and serum were collected from mice at 28 days of bleomycin induction and from control mice for histopathological and metabolomic assays; (B) Lung histopathology after a single dose of bleomycin treatment (Masson, TGF-β1, and ɑ-SMA staining). Representative Masson’s trichrome-stained (a and b), TGF-β1-stained (c and d), and ɑ-SMA-stained (e and f) lung tissue sections are shown. The mice were sacrificed at 28 days (b, d, and f) after bleomycin instillation. Except for the PBS control (a, c, and e), all panels showed lung sections from bleomycin-treated mice. The lungs were inflated with 10% buffered formaldehyde. Original magnification, 20X; (C) The results of Masson, TGF-β1 and ɑ-SMA immunohistochemistry. Positive area ratio (%), reflecting the proportion of the positive area. Statistical significance by two-tailed Student’s t-tests: ^∗p < 0.05; ^∗∗p < 0.01.

Figure 2

Metabolome profiling of serum. (A) The clustering analyses of orthogonal partial least-squares discriminant analysis (OPLS-DA) of Control and Model group (R2X = 0.349, R2Y = 0.990, Q2 = 0.527); (B) Volcano plot of metabolites, with purple representing downward trend, gray representing insignificance, and red representing upward trend; (C) Heatmap of the relative abundance of representative differential serum metabolites among Control and Model groups. GP, glycerophospholipids; FA, fatty acyls.

Figure 3

Pathway enrichment analysis of differential metabolites in the serum of mice with pulmonary fibrosis Results of differential metabolite enrichment analysis via the MBRole 2.0 website. The color of the dot represents the p-value, and the redder the bar, the more significant the pathway.

Figure 4

LASSO regression algorithm for screening biomarkers and evaluating their performance. (A) The optimal set of retinal parameters was selected by tenfold cross-validation and lambda. Min; (B) A coefficient profile plot was produced against the log(lambda) sequence; (C) Performance results of the seven metabolites screened as assessed by ROC curves. The larger the AUC, the redder the bar. (D) Box plots of the seven metabolites screened. Statistical significance by two-tailed Student’s t-tests: ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001; ^∗∗∗∗p < 0.0001.