Untargeted 1H-NMR metabolomics of bronchoalveolar lavage fluids to identify biomarkers for early diagnosis of lower respiratory tract infections

Abstract

Lower respiratory tract infections (LRTIs) remain a leading cause of death worldwide. Their complex clinical presentation underscores the need for accurate, timely laboratory diagnosis; however, reliable diagnostic biomarkers are still lacking. We performed untargeted 1 H-NMR metabolomics of bronchoalveolar lavage fluids (BALFs) to identify potential diagnostic biomarkers. BALFs from patients with infectious (≥ 10 4 CFUs/mL, LRTI group) and non-infectious (< 10 4 CFUs/mL, control group) lower respiratory tract diseases were analyzed using conventional microbiological diagnosis in parallel with NMR-based metabolomics. The most frequent pathogens identified were Enterobacterales (~ 26%), Haemophilus influenzae (18%), Staphylococcus aureus (10%), and Moraxella catarrhalis (10%). Macroscopic BALF features (e.g. color, viscosity) were found unreliable indicators of infection, whereas nitrite concentrations above ~ 20 µM were detected exclusively in the LRTI group. Multivariate analysis of 1 H-NMR metabolic data showed significant discrimination between both groups. This distinction was primarily driven by elevated levels of branched-chain amino acids (valine, leucine, isoleucine) and their catabolic by-products (isovalerate, 2-hydroxy-2-methylbutyric acid), although additional metabolites also contributed to the observed variance. Further refinement by focusing on samples from patients with Haemophilus influenzae infection led to complete differentiation of both groups, with improved model fit, although no species-specific markers were identified across pathogens. Overall, we concluded that Gram-negative species were the predominant pathogens causing LRTIs in our patient population (79%). Nitrite levels and branched-chain amino acids emerged as promising diagnostic biomarkers for distinguishing infectious from non-infectious lower respiratory tract diseases, warranting validation in larger multicenter cohorts.