Phenotype heterogeneity is classically infl uenced by genetic, epigenetic and environmental factors but the gut microbiome is also thought to contribute to this phenomenon. Here we studied a population of 50 isogenic C57BL/6J mice developing a range of dietary-induced diabetes, obesity and anxiety sub-phenotypes after 20 days of high-fat feeding. We then profi led the urinary metabolome of these mice by untargeted 1H Nuclear Magnetic Resonance spectroscopy before high-fat-diet feeding and sequentially afterwards. Multivariate statistical models were constructed to predict phenotype heterogeneity and disease sub-phenotypes (lean non diabetics, lean diabetics, obese diabetics) from using baseline metabolic profi les. Among the markers, we found that excretion of gut microbial detoxifi cation metabolites such as phenylacetylglycine (PAG), hippurate, dimethylamine and trimethylamine-N-oxide (TMAO) was strongly predictive of obesity and disease sub-phenotypes, insulin resistance, body/organ weights as well as anxiety/activity outcomes. Genome-wide adipose tissue gene expression profi ling tissue suggests that TMAO excretion correlates with reduced lipogenesis and insulin action. We then set up cellular investigations to further document the causative role of these microbial metabolites. For instance, we tested whether TMAO, PAG, hippurate affect cellular phenotypes associated with obesity and diabetes, through inhibition of in vitro cell differentiation, glucose uptake and lipid accumulation in adipocytes. This work describes new mechanistic detail that underpins the role of the microbiome in cardiovascular disease and also suggests future possibilities for disease prognosis based on microbial signatures and new therapeutic interventional approaches.