Multivariate data analysis of metabolomic multi-tissue samples

Sammanfattning: Multi-tissue metabolomics involves characterisation of the metabolome of several tissue types. The metabolome consists of small chemical entities of low molecular weight called metabolites, which are constantly produced and interchanged through a vast variety of biochemical reactions occurring throughout living organisms. Metabolome alterations can be attributed to genetics, environment, and diseases. We used gas chromatography timeof-flight mass spectrometry (GC TOF-MS) to characterise the metabolome of mouse organ samples: gut, kidney, liver, muscle, pancreas and plasma. Samples were obtained from wild-type mice and mice carrying a mutation in the hepatocyte nuclear factor 1b (HNF1b) gene, referred to as MODY5/RCAD (for maturity onset diabetes of the young 5/renal cysts and diabetes syndrome) mice. MODY is a class of hereditary diabetes mellitus, and MODY5 is caused by mutations in HNF1B, resulting in a wide range of manifestations, including renal diseases, kidney and genitourinary malformation, and elevation of liver enzymes. Today, MODY5 in humans is diagnosed using genetic tests, and varying referral rates and manifestations have resulted in misdiagnosis. Our main focus was therefore to increase understanding of the metabolism associated with MODY5/RCAD by studying the metabolic profiles of individual organs and plasma (Paper I) from MODY5/RCAD mutant and wildtype mice. The mouse model displayed an overall metabolic pattern consistent with the presumed outcome of the mutation in humans, making the MODY5/RCAD model suitable for studies of HNF1B-associated diseases. An understanding of metabolite origin would be beneficial for understanding the plasma profile associated with MODY5/RCAD. We used hierarchical modelling to provide an understanding of metabolite origin by detecting how metabolites from the organs contributed to the plasma metabolic profile (Paper II). Both specific and overall organ metabolite contributions to the plasma metabolic profile were studied. Further exploration of the dataset involved study of its innate variation using joint and unique multiblock analysis (JUMBA; Paper III). In addition, we explored the effects of improper sample handling for metabolomic multi-tissue data, and we studied the similarities and differences in the responses to thawing between organ tissues (Paper IV) and plasma samples (Paper V), thus identifying metabolic profiles that could indicate compromised samples. These profiles could be beneficial for large-scale collaborations that involve sample exposure to unsuitable conditions. Altogether, we have contributed to an increased understanding of the MODY5/RCAD multi-tissue metabolomic dataset and worked up protocols and strategies for how small datasets should be handled.