Exhaled breath metabolomics reveals a pathogen-specific response in a rat pneumonia model for two human pathogenic bacteria: a proof-of-concept study.

INTRODUCTION: Volatile organic compounds (VOCs) in breath can reflect host and pathogen metabolism and might be used to diagnose pneumonia. We hypothesized that rats with Streptococcus pneumoniae ( SP) or Pseudomonas aeruginosa ( PA) pneumonia can be discriminated from uninfected controls by thermal desorption - gas chromatography - mass-spectrometry (TD-GC-MS) and selected ion flow tube - mass spectrometry (SIFT-MS) of exhaled breath.

METHODS: Male adult rats ( n=50) received an intra-tracheal inoculation of 1) 200 µL saline, 2) 1x107 colony forming units (CFU) of SP or 3) 1x107 CFU of PA. 24 hours later the rats were anaesthetized, tracheotomized and mechanically ventilated. Exhaled breath was analyzed via TD-GC-MS and SIFT-MS. Area under the receiver operating characteristic curves (AUROCCs) and correct classification rate (CCRs) were calculated after leave-one-out cross-validation of sparse partial least squares-discriminant analysis (sPLS-DA).

RESULTS: Analysis of GC-MS data showed an AUROCC (95% CI) of 0.85 (0.73 - 0.96) and CCR of 94.6% for infected vs. non-infected animals, AUROCC 0.98 (0.94 - 1) and CCR of 99.9% for SP vs. PA, 0.92 (0.83 - 1.00) and CCR of 98.1% for SP vs. controls and 0.97 (0.92 - 1.00) and CCR of 99.9% for PA vs. controls. For these comparisons the SIFT-MS data showed AUROCCs of 0.54, 0.89, 0.63 and 0.79, respectively.

DISCUSSION: Exhaled breath analysis discriminated between respiratory infection and no infection, but with even better accuracy between specific pathogens. Future clinical studies should not only focus on the presence of respiratory infection, but also on the discrimination between specific pathogens.