Proteomic characteristics and diagnostic potential of exhaled breath particles in patients with COVID-19

SARS-CoV-2 has been shown to predominantly infect the airways and the respiratory tract and too often have an unpredictable and different pathologic pattern compared to other respiratory diseases. Current clinical diagnostical tools in pulmonary medicine expose patients to harmful radiation, are too unspecific or even invasive. Proteomic analysis of exhaled breath particles (EBPs) in contrast, are non-invasive, sample directly from the pathological source and presents as a novel explorative and diagnostical tool.

Patients with PCR-verified COVID-19 infection (COV-POS, n = 20), and patients with respiratory symptoms but with > 2 negative polymerase chain reaction (PCR) tests (COV-NEG, n = 16) and healthy controls (HCO, n = 12) were prospectively recruited. EBPs were collected using a "particles in exhaled air" (PExA 2.0) device. Particle per exhaled volume (PEV) and size distribution profiles were compared. Proteins were analyzed using liquid chromatography-mass spectrometry. A random forest machine learning classification model was then trained and validated on EBP data achieving an accuracy of 0.92.

Significant increases in PEV and changes in size distribution profiles of EBPs was seen in COV-POS and COV-NEG compared to healthy controls. We achieved a deep proteome profiling of EBP across the three groups with proteins involved in immune activation, acute phase response, cell adhesion, blood coagulation, and known components of the respiratory tract lining fluid, among others. We demonstrated promising results for the use of an integrated EBP biomarker panel together with particle concentration for diagnosis of COVID-19 as well as a robust method for protein identification in EBPs.

Our results demonstrate the promising potential for the use of EBP fingerprints in biomarker discovery and for diagnosing pulmonary diseases, rapidly and non-invasively with minimal patient discomfort.