Abstract Peptide-centric machine learning enhanced (PCML) data-independent acquisition tandem mass spectrometry (LC-MS/MS-DIA) effectively matches low abundance MS fragmentation spectra to in silico predicted peptide spectra derived from libraries of protein sequences pertinent to the biological circumstances of the experiment. We analyzed sputum from infected patients using a LC-MS/MS-DIA/PCML workflow that captured host and pathogen proteomes at an unparalleled depth. This workflow represents a leap forward in low abundance pathogen identification along with insights into host-pathogen crosstalk in the airways. The PCML approach utilized neural network algorithms from a published method termed DIA-NN. We identified up to 6,800 proteins in total and 1,530 microbial proteins in a single LC-MS analysis. Conventional spectral library searches yielded only up to 1,600 and 50 such protein identifications, respectively. Coefficients of variation for protein quantities among biological replicate sputum samples using PCML were lower than 0.12. The data from two patient specimens showed that Pseudomonas aeruginosa and Stenotrophomonas maltophilia had infected their airways. Inferred from relative protein abundances, P. aeruginosa responded to the inflamed airway milieu by expressing energy generation systems used at low oxygen tension and a type 6 secretion system known to perturb functions of host and other microbial cells. The protein Smlt2713 of S. maltophilia , highly abundant in sputum from one patient, was previously found to induce expression of human interleukin-10, suggesting a role in immune system suppression in the airways. We conclude that the LC-MS/MS-DIA/PCML workflow allows unprecedented insights into host-microbial crosstalk from the analysis of respiratory clinical samples. It is a tool for translational medicine investigating pharmacodynamic responses to therapy with novel antimicrobial drugs.

Crosstalk of pathogens with human immune system in airway mucus profiled via machine learning-enhanced data-independent acquisition mass spectrometry