Universal, untargeted detection of bacteria in human tissues using spatial metabolomics

Abstract

Fast and reliable identification of bacteria directly in clinical samples is the holy grail of microbiological diagnostics. Current approaches predominantly require time-consuming bacterial isolation and enrichment procedures, delaying stratified treatment. Here, we describe a biomarker-based, hierarchical strategy that utilises conserved bacterial small molecular metabolites and lipids for direct mass spectrometric detection of bacteria in clinical samples. Marker discovery was carried out by creating a metabolic library of 232 bacterial species and mining this library for markers showing specificity at various taxonomic levels. We have found 359 bacterial taxon-specific markers (TSMs) via a combination of univariate and multivariate statistical analysis methods. We applied the concept of TSMs to the in situ detection of bacteria directly from complex clinical specimens, using healthy and cancerous colorectal tissues as well as faecal samples. To show the method independent nature of these TSMs, samsamples were analysed using chemical histology as well as traditional metabolomics approaches. TSMs were found in >90% of samples tested, suggesting the general applicability of this novel workflow to detect bacterial presence with standard MS-based analytical methods.