Abstract

Background/Objectives: Lactic acid bacteria (LAB) are essential drivers of food fermentation, yet systematic metabolic comparisons across different LAB strains remain underexplored. This study characterized and contrasted the metabolic fingerprints of Lactiplantibacillus plantarum and Lacticaseibacillus rhamnosus in vegetable fermentation and subsequently evaluated whether the same untargeted metabolomics pipeline could be applied to rapid foodborne pathogen detection. Methods: A multi-platform untargeted metabolomics strategy integrating GC-MS and UPLC-Q-TOF-MS was applied to profile four experimental conditions: non-fermented control, L. plantarum monoculture, L. rhamnosus monoculture, and mixed-culture fermentation. Multivariate statistical tools (PCA and PLS-DA) were used to identify differential metabolites and perturbed pathways. The identical analytical workflow was then applied to beef samples artificially contaminated with Escherichia coli O157:H7, Salmonella enterica, or Listeria monocytogenes. Results: Across all samples, 847 metabolites were annotated, of which 312 showed significant abundance changes upon fermentation. PLS-DA delivered robust group discrimination (R2X = 0.89, R2Y = 0.95, Q2 = 0.91). Organic acids (32.0%) and amino acids (24.0%) dominated the metabolic landscape, with lactic acid, acetic acid, diacetyl, and acetoin as the most elevated compounds. KEGG analysis highlighted glycolysis/gluconeogenesis, pyruvate metabolism, and branched-chain amino acid degradation as the most heavily rewired pathways. When the same pipeline was applied to pathogen detection, it yielded AUC values of 0.89, 0.87, and 0.88 for E. coli O157:H7, S. enterica, and L. monocytogenes, respectively, with detection times of 18 h, 24 h, and 30 h. Conclusions: This work delivers a side-by-side metabolic atlas of two prominent LAB species and demonstrates the technical portability of an untargeted metabolomics pipeline from a fermentation model system to a pathogen detection scenario. The identified biomarker panels warrant further validation in diverse food matrices, and translation to routine monitoring will require matrix-specific model training and validation.

IPC Classification

Keywords