4.4. Untargeted Metabolomics with LC-MS

The blood serum samples were frozen at −80 °C until analysis. A Phree filter was used to remove proteins and phospholipids (Phenomenex Inc., Torrance, CA, USA). Untargeted analysis was performed on an ultra-performance liquid chromatography system (UltiMate 3000, ThermoFisher Scientific, Waltham, MA, USA) with a qTOF mass spectrometer (maXis 4G+, Bruker Daltonik GmbH, Bremen, Germany). The samples from each participant were run together, and their injection order was randomized. The injection order of the 37 participants was also randomized. The measurements were all performed in the positive mode in the range 75 to 1500 m/z. The stability of the MS signals was tested with a control sample (QC), which was a mixture of all serum samples. Contamination of the system was controlled by the injection of ultrafiltered water. The identification of the components was performed with the National Institute of Standards and Technology database (NIST v14), the Human Metabolome Database [69], the MassBank of North America (MoNA) [70], and Metlin [71]. The mass accuracy was set to 5 ppm. The identity of the components was verified using collision-induced dissociation (CID) with a range from 5 to 70 eV collision energy.

The following levels of identification were used [15]. Level 1: The compound was identified based on a difference in retention time below <20%, and the m/z difference below 10 ppm was compared with the chemical reference standard. Level 2: The tentative annotation was performed by spectral similarity with public/commercial databases with a pectral matching with an m/z difference below 10 ppm, but the reference substance was not available. Level 3: The spectral similarity (m/z difference below 10 ppm) was checked with public/commercial spectral libraries. The compound can only be attributed to a compound class. Level 4: The retention time and m/z signal were available, but identification was not possible.