Urine metabolomics analysis based on ultra performance liquid chromatography-high resolution mass spectrometry combined with osmolality calibration sample concentration variability

Abstract

Urine is an important source of biomolecular information for metabolomic studies. However, the acquisition of high-quality metabolomic datasets or reliable biomarkers from urine is difficult owing to the large variations in the concentrations of endogenous metabolites in the biofluid, which are caused by diverse factors such as water consumption, drugs, and diseases. Thus, normalization or calibration is essential in urine metabolomics for eliminating such deviations. The urine osmolality ( Π), which is a direct measure of the total urinary solute concentration and is not affected by circadian rhythms, diet, gender, and age, is often considered the gold standard for estimation of the urine concentration. In this study, a pre-data acquisition calibration strategy based on osmolality was investigated for its feasibility to overcome sample concentration variability. Before data acquisition, the product of the osmolalityxinjection volume of all samples was set to be equivalent through the uses of a customized injection volume or dilution. After ultra performance liquid chromatography-high resolution mass spectrometry (UPLC-HRMS) analysis of the sample, the raw dataset was normalized to the total ion abundance or total useful MS signals ( MSTUS) to achieve further calibration. The osmolality of each urine sample was determined with a freezing-point depression osmometer. For the instrumental analysis, a Vanquish UPLC system coupled to a Q-Exactive Plus HRMS device was used for metabolite analysis and accurate mass measurement. Full-scan mass spectra were acquired in the range of m/z 60-900, and the MS/MS experiments were conducted in “Top5” data-depend- ent mode. A Waters UPLC column ( 100 mmx2. 1 mm, 1. 8 μm) was used for chromatography separation. The raw data were imported into Progenesis QI software for peak picking, alignment, deconvolution, and normalization. SIMCA-P software was used for the principal component analysis ( PCA) and orthogonal partial least-squares discrimination analysis ( OPLS-DA). This strategy was first applied to sequentially diluted urine samples, where three frequently used normalization methods were compared. In the identical injection volume experiment, the points were scattered and showed relevant distribution according to the dilution multiple in the plot of PCA scores. There was little improvement after normalization to either the total ion abundance or MSTUS. In the customized injection volume experiment, the urine samples derived from the same source showed ideal clustering. With total ion abundance and MSTUS normalization, the dataset was further improved in the PCA model fitting and prediction. As a result, there were more peaks with a peak area RSD of <30%, which indicated better parallelism. The diluted urine solutions had higher Spearman ’ s coefficient values with their sample source than those without calibration, which suggested less intra-group differences. The strategy was further validated using data from a metabolomic study of children with congenital hydronephrosis and healthy controls. As a concentration estimator, osmolality showed better linear correlation with the mass signal and was less influenced by physiological or pathological factors, thus obtaining broader application and more accurate results than creatinine. The concentration variability was effectively eliminated after customized dilution calibration and showed a more obvious clustering effect in the PCA score plot. The OPLS-DA-based statistical model used to identify discriminate metabolites was improved, with less chance of overfitting. In conclusion, the calibration strategy based on osmolality combined with total ion abundance or MSTUS normalization significantly overcame the problem of urine concentration variability, eliminated intra-group differences, and possessed better parallelism, thus giving better clustering effects in PCA or OPLS-DA and higher reliability of the statistical model. The results of this study provide guidance and a reference for future metabolomic studies on urine.