Methodical aspects of urinary metabolomics by liquid chromatography-mass spectrometry

With recent instrumental improvements, untargeted high-resolution liquid chromatography mass spectrometry (LC-MS) has been increasingly applied to the analysis of complex biological matrices such as urine. The latter is a preferred matrix in large scale metabolomics studies, because it can be obtained non-invasively in sufficient quantities. However, the widely varying concentrations of urine metabolites, mostly due to different fluid intake, pose analytical challenges, such as ion suppression, detector saturation, column overload or failure to detect low abundant metabolites. In this doctoral thesis, both targeted and untargeted analysis of urine by LC-MS was investigated to address some of these challenges. Further, the question was addressed if untargeted fingerprinting of urine specimens is adequate for the quantitative analysis of these specimens. First, five different dilution and normalization strategies of spot urinary speci-mens were compared. Specimens were adjusted either to a uniform creatinine concentration or osmolality without any further normalization of the acquired data, or they were uniformly diluted and post-acquisition normalized to creatinine, osmolality, or sum of all integrals. Spot urine specimens from both an apparently healthy cohort and two different cohorts suffering from chronic kidney disease (CKD) were investigated to test the effect of the various strategies on sample classification. The advantages of pre-acquisition dilution of urine to a fixed creatinine concentration in the acquisition of metabolite fingerprints by LC coupled to a high resolution time-of-flight mass spectrometer (LC-HRTOFMS) were demonstrated. These included the absence of significant correlations of missing values with either the original urine creatinine concentration (r = 0.216, p = 0.133) or osmolality (r = - 0.074, p = 0.608). Moreover, pre-acquisition dilution to a uniform creatinine concentration was the only method to correctly assign urine specimens of controls (national cohort, NC) and CKD patients to their respective clusters in hierarchical cluster analysis, while the other dilution and normalization methods yielded indistinct clusters. Therefore, time and labor associated with pre-acquisition dilution to a uniform creatinine concentration are more than justified. Next, a targeted liquid chromatography triple quadrupole mass spectrometry (LC-QqQMS) method for the quantification of urine metabolites associated with CKD was developed. Some of these metabolites had been identified during the testing of different pre-acquisition dilution and normalization strategies. The dynamic range of the method showed adequate sensitivity for all urine metabolites investigated. The lower limits of quantification (LLOQs) ranged between 0.04 nM to 11 nM except for creatinine (CRE), creatine (CRT), hypoxanthine (HX) and xanthurenic acid (XA), which featured LLOQs of 65.9 nM, 700 nM, 87.8 nM, and 22.0 nM, respectively. Spike-in experiments yielded recoveries of 86% to 121%. The method was applied successfully to the determination CRE, C-mannosyltryptophan (CMT) and pseudouridine (PSU) in urine (Sekula et al. 2017, Scientific Reports 7(1): 17400) and to CRE and CRT in plasma (Wallmeier et al. 2017, Journal of Proteome Research 16(4): 1784-1796). Finally, targeted and untargeted measurements of urine specimens were compared with respect to the quantitative performance of untargeted measurements. Combining both targeted and untargeted analysis of urine in a single analysis would constitute a tremendous saving of time and resources. First, calibration curves of selected metabolites were compared, measured in both full scan and MRM mode on a Bruker Maxis Impact quadrupole time-of-flight and a Sciex 4000 QTRAP mass spectrometer, respectively. The LLOQs on the former instrument ranged from 5.49 nM to 1730 nM, whereas the corresponding range on the latter instrument was 1.37 nM to 65.9 nM. The linear ranges were also narrower for the untargeted than the targeted method. Reproducibility over the linear range compared well to the targeted approach. All relative standard deviation (RSD) values for the selected metabolites were below 10%. Next, targeted and untargeted measurements of 200 CKD and 100 non-CKD urine specimens were compared to examine the quantitative performance of the untargeted platform. For the CKD samples, targeted and untargeted measurements of PSU, XA, CMT and tryptophan (TRP) yielded Spearman’s rank correlation coefficients ranging from 0.67 to 0.87, while the corresponding coefficients for urine specimens from apparently healthy individuals ranged from 0.73 to 0.95. PSU yielded consistently the weakest correlation coefficient, most likely due to matrix effects as a consequence of its low retention and, thus, co-elution with many other analytes. It is concluded, that the correlation of quantitative data gained by targeted and non-targeted analysis depends on the instruments and methods used as well as the metabolites and their abundance