Metabolic phenotyping facilitates the analysis of low molecular weight compounds in complex biological samples, with resulting metabolite profiles providing a window on endogenous processes and xenobiotic exposures. Accurate characterisation of the xenobiotic component of the metabolome (the xenometabolome) is particularly valuable when metabolic phenotyping is used for epidemiological and clinical population studies where exposure of participants to xenobiotics is unknown or difficult to control/estimate. Additionally, as metabolic phenotyping has increasingly been incorporated into toxicology and drug metabolism research, phenotyping datasets may be exploited to study xenobiotic metabolism at the population level. This thesis describes novel analytical and data-driven strategies for broadening xenometabolome coverage to allow effective partitioning of endogenous and xenobiotic metabolome signatures.
The data driven strategy was multi-faceted, involving the generation of a reference database and the application of statistical methodologies. The database contains over 100 common xenobiotics profiles - generated using established liquid chromatography-mass-spectrometry methods – and provided the basis for an empirically derived screen for human urine and blood samples. The prevalence of these xenobiotics was explored in an exemplar phenotyping dataset (ALZ; n = 650; urine), with 31 xenobiotics detected in an initial screen. Statistical based methods were tailored to extract xenobiotic-related signatures and evaluated using drugs with well-characterised human metabolism.
To complement the data-driven strategies for xenometabolome coverage, a more analytical based strategy was additionally developed. A dispersive solid phase extraction sample preparation protocol for blood products was optimised, permitting efficient removal of lipids and proteins, with minimal effect on low molecular weight metabolites. The suitability and reproducibility of this method was evaluated in two independent blood sample sets (AZstudy12; n=171, MARS; n=285).
Finally, these analytical and statistical strategies were applied to two existing large-scale phenotyping study datasets: AIRWAVE (n = 3000 urine, n=3000 plasma samples) and ALZ (n= 650 urine, n= 449 serum) and used to explore both xenobiotic and endogenous responses to triclosan and polyethylene glycol exposure. Exposure to triclosan highlighted affected pathways relating to sulfation, whilst exposure to PEG highlighted a possible perturbation in the glutathione cycle.
The analytical and statistical strategies described in this thesis allow for a more comprehensive xenometabolome characterisation and have been used to uncover previously unreported relationships between xenobiotic and endogenous metabolism.