Development and application of multiple isotopes-assisted untargeted metabolomics in human health and disease

Global metabolite profiling, also known as untargeted metabolomics, is constantly used in the qualitative and quantitative assessment of a wide range of metabolites in human metabolomics research, where scientists are able to monitor changes in metabolite concentrations in human biofluids and tissues affected by complex clinical diseases including rheumatoid arthritis, Alzheimer’s, Parkinson’s and variety of cancers, to name a few. These metabolite profiling studies are regularly applied in human clinical samples in order to detect potential biomarkers, assist in drug discovery, monitor disease onset and its progress and many other bioanalytical areas. The application of the LC-MS analytical tool has been most widely used in comprehensive metabolic studies, due to its high throughput, soft ionisation and good metabolite coverage, in comparison to other analytical instruments such as NMR, FT-IR or Raman. LC-MS based metabolomics studies are currently facing challenges due to non-linear responses derived from matrix effects and biological variations within samples, which can result in biased quantitative analysis affecting true measurement of metabolite levels and their biological relevance. To overcome these issues, stable multiple 13C isotopically labelled metabolites acting as internal standards can be employed in metabolomics studies to reduce metabolome data variation and improve its accuracy, also known as a normalisation technique. The main limitation of this normalisation technique in global metabolite profiling studies, however, is a lack of available 13C labelled standards to cover a wide range of metabolites, as well as their high cost to produce. To solve this problem, uniformly (U) 13C labelled bacterial organisms, such as E. coli or Spirulina, can be used instead, acting as a source of multiple labelled internal standards. Currently, in metabolomics research, there is a lack of established validated 13C normalisation techniques that can be applied to a wide range of metabolites in mammalian samples using a global metabolite profiling method. In this thesis the proposed LC-MS method, involving a (U)-13C labelled bacterial organism as a source of internal standards for normalisation, has been developed and applied to a range of untargeted clinical studies to demonstrate the effectiveness of normalisation and improvement in data accuracy to answer biological questions. E. coli and Spirulina extracts were analysed using LC-MS-based metabolite profiling to select the appropriate source of internal standards. In E. coli samples, around 780 putative metabolites were detected with high peak signal response, compared to approximately 600 putative metabolites in the Spirulina bacterium with poor peak signal response. E. coli appeared to have more metabolites in common with human biofluid or tissue metabolomes than Spirulina, fully confirming E. coli to be a suitable bacterial organism to use for internal standards. The chosen (U)-13C labelled E. coli showed a large proportion of metabolites labelled with 13C isotope (77%), with only 23% of the metabolome unlabelled. To validate the proposed normalisation method, (U)-13C labelled E. coli was applied in human urine, human brain tissue and mouse plasma samples. 13C-labelled E. coli extract was added to the extraction solvent (methanol) and mixed with the samples of interest in a 1:1 ratio. In all three studies, percentage RSD of peak height intensities was calculated for detected metabolites, along with constructed PCA and OPLS-DA plots, to assess the efficiency of normalisation. In human urine and brain studies, approximately 70% of identified metabolites in each group had their percentage RSD reduced, while in the mouse plasma study the result was observed to be even better with 90% of metabolites successfully normalised. When compared to other normalisation techniques such as MSTUS, TIC and creatinine, the 13C normalisation has shown better results with percentage RSD range being less variable. In all three validation studies, PCA and OPLS-DA score plots showed clearer separation between sample groups with their replicates clustered and high Q2 scores in normalised metabolomes, compared to non-normalised. Finally, the normalisation technique has helped to detect more statistically significant metabolites in all three studies, compared to non-normalised datasets. This method has shown to be fully validated, reducing metabolite variation significantly and improving the accuracy of data by detecting a high number of statistically significant metabolites. A fully validated LC-MS method with proposed normalisation technique has been applied in two clinical investigation studies, to obtain a highly accurate metabolome from clinical samples and answer the main biological questions set up by the studies: mainly searching for potential biomarkers and the effect of a disease on metabolic pathways. One clinical study investigated the effect of a fatty meal (diabetes condition) on the human urine and plasma metabolome of healthy volunteers, while the other study performed metabolomics analysis on human low-grade glioma affected brain tissue. In both studies, normalised metabolome data have helped to detect a number of statistically significant metabolites which were observed to affect certain metabolic pathways involved in the investigated diseases, showing potential for being biomarkers. Overall, the proposed normalisation technique using multiple 13C labelled internal standards, with the assistance of a bacterial organism as their source and a powerful analytical LC-MS method, has achieved great results in reducing metabolite data variation and improving data accuracy of a wide range of metabolites in mammalian biofluid and tissue samples, analysed by untargeted metabolomics, and has shown great promise in the search for potential biomarkers in future clinical untargeted studies.