BACKGROUND: Asthma is a chronic respiratory disease with significant heterogeneity in its clinical presentation and pathobiology. There is need for improved understanding of respiratory lipid metabolism in asthma patients and its relation to observable clinical features. OBJECTIVE: To perform a comprehensive, prospective, cross-sectional analysis of the lipid composition of induced sputum supernatant obtained from asthma patients with a range of disease severities, as well as healthy controls. METHODS: Induced sputum supernatant was collected from 211 asthmatic adults and 41 healthy individuals enrolled in the U-BIOPRED study. Sputum lipidomes were characterised by semi-quantitative shotgun mass spectrometry, and clustered using topological data analysis to identify lipid phenotypes. RESULTS: Shotgun lipidomics of induced sputum supernatant revealed a spectrum of nine molecular phenotypes, highlighting not just significant differences between the sputum lipidomes of asthmatics and healthy controls, but within the asthmatic population as well. Matching clinical, pathobiological, proteomic and transcriptomic data informed on the underlying disease processes. Sputum lipid phenotypes with higher levels of non-endogenous, cell-derived lipids were associated with significantly worse asthma severity, worse lung function, and elevated granulocyte counts. CONCLUSION: We propose a novel mechanism of increased lipid loading in the epithelial lining fluid of asthmatics, resulting from the secretion of extracellular vesicles by granulocytic inflammatory cells, which could reduce the ability of pulmonary surfactant to lower surface tension in asthmatic small airways, as well as compromise its role as an immune regulator. CLINICAL IMPLICATION: Immunomodulation of extracellular vesicle secretion in the lungs may provide a novel therapeutic target for severe asthma.