Vladyslav Dovhalyuk: Analytical Strategies for Host-Microbiome Interaction Metabolomics in Clinically Relevant Samples: From Detection Bias to Biological Interpretation

Abstract

The human metabolome reflects contributions from host metabolism, microbial activity, environmental factors, and analytical methodology. Metabolites provide a direct readout of biochemical processes and their alterations under different physiological and pathological conditions. Liquid chromatography-mass spectrometry (LC-MS)-based metabolomics enables the detection and comparison of these molecules across complex biological systems.

In this thesis, global LC-MS metabolomics was applied to investigate methodological and biological determinants of metabolite profiles across controlled experimental systems and human studies. The effects of sample preparation and storage on the fecal metabolome were systematically evaluated. Metabolite recovery depended on the combination of extraction strategy and storage conditions rather than a single optimal method. Double liquid extraction (DLE) using DMSO-containing solvents improved recovery of small polar metabolites, including amino acids and peptide derivatives. Methanol-based extraction methods performed well under fresh sample conditions, while specific combinations, including methanol extraction of lyophilized samples and prolonged methanol storage, reduced detection of nucleobase- and vitamin-related metabolites. Fecal metabolomics was applied to autoimmune and chronic pancreatitis. Metabolic differences were observed across metabolite classes, including amino acids, peptide derivatives, and microbiota-associated metabolites, indicating shifts in gut-associated metabolic processes.

The analysis was extended to feces, plasma, and urine in individuals at high risk of developing pancreatic cancer. Metabolic alterations were not uniformly reflected across matrices. In urine samples, broad variation with limited consistency were observed, while plasma had less changes, and fecal samples exhibited the most structured alterations across diverse metabolite classes. Metabolomics analysis was further applied to a controlled co-culture system of commensal vaginal bacteria with pathogens. Metabolite increases depended on specific bacterial combinations and were not resolved by global multivariate analysis alone. Pairwise contrast analysis between co-cultures and corresponding mono-culture controls identified interaction-dependent metabolites, including proline-containing dipeptides and heterocyclic compounds such as β-carboline derivatives and urocanic acid, which demonstrated bioactivity by modulating the growth of Lactobacillus crispatus, Gardnerella vaginalis, Prevotella bivia, Candida albicans, Escherichia coli, and Staphylococcus aureus.

In summary, the results in this thesis from metabolomics analyses of different biological sample types define methodological constraints affecting metabolite detection and characterize context-dependent metabolic patterns across biological systems, biospecimens, and microbial interactions.