Investigation of pH-dependent ¹H NMR urine metabolite profiles for diagnosis of obesity-related disordering: Physiology and Biochemistry

Background: Human urine is highly favorable for ¹H NMR metabolomics analyses of obesity-related diseases, such as non-alcoholic fatty liver, type 2 diabetes, and hyperlipidemia (HL), due to its non-invasiveness and ease of large-scale collection. However, the wide range of intrinsic urine pH (5.5–8.5) results in inevitably chemical shift and signal intensity modulations in the ¹H NMR spectra. For patients where acidic urine pH is closely linked to obesity-related disease phenotypes, the pH-dependent modulations complicate the spectral analysis and deteriorate quantifications of urine metabolites. Methods: We characterized human urine metabolites by NMR at intrinsic urine pH, across urine pH 4.5 to 9.5, to account for pH-dependent modulations. A pH-dependent chemical shift database for quantifiable urine metabolites was generated and integrated into a “pH intelligence” program developed for quantifications of pH-dependent modulations at various pH. The ¹H NMR spectra of urines collected from patients with Ob-HL and healthy controls were compared to uncover potential metabolic biomarkers of Ob-HL disease. Results: Three urine metabolites were unveiled by pH-dependent NMR approach, i.e., TMAO, glycine, and pyruvic acid, with VIP score >1.0 and significant q-value < 0.05, that represent as potential biomarkers for discriminating Ob-HL from healthy controls. Further ROC-AUC analyses revealed that TMAO alone achieved the highest diagnostic accuracy (AUC 0.902), surpassed to that obtained by neutralizing pH approach (AUC 0.549) and enabled better recovering potential urine metabolites from the Ob-HL disease phenotypes. Conclusions: We concluded that ¹H NMR-derived urine metabolite profile represents a snapshot that can reveal the physiological condition of humans in either a healthy or diseased state under intrinsic urine pH. We demonstrated a systematic analysis of pH-dependent modulations on the human urine metabolite signals and further developed software for quantification of urine metabolite profiles with high accuracy, enabling the uncovering of potential metabolite biomarkers in clinical diagnosis applications. (Figure presented.)