Development and evaluation of new strategies for the general unknown toxicological screening in biosamples using high resolution mass spectrometry

Introduction: The screening for Pharmaco/Toxicologically Relevant Compounds (PTRC) in biosamples has benefited a lot from MS techniques. The so-called library search approach has enabled the development of effective identification methods based on comparison of unknown and reference spectra. However, a downside of this approach is the limited number of reference mass spectra, particularly in the case of LC-MS where in-house/commercial databases typically include not more than one thousand compounds. High resolution MS (HRMS) enables the identification of a molecular formula (MF) through the accurate measurement of mass and isotopic pattern. However, the identification of an unknown compound starting from MF requires additional tools: (a) a database associating MFs to compound names, and (b) a way to discriminate between isomers. Aims: To evaluate the ability of combined novel “metabolomic”/chemometric approach to reduce the list of candidate isomers. Methods: Urine/blood/hair samples collected from real positive cases were submitted to a screening procedure using ESI-MSTOF (positive ion mode) combined with either capillary electrophoresis or reversed phase LC (RPLC). Detected peaks were searched against a PTRC database (ca. 50.500 compounds and phase I and II metabolites) consisting of a subset of PubChem Compound. In order to discriminate between compounds with identical MF first a filter based on a “metabolomic” approach was applied. Starting from the mass of the unknown compound, defects/increments corresponding to pre-defined biotransformations (e.g. demethylation, hydroxylation, glucuronidation, etc.) were calculated and the corresponding mass chromatograms were extracted from the total ion current (TIC) in order to search for metabolite peaks. For each candidate in the retrieved list, the number of different functional groups in the molecule (N,O,S-methyls, hydroxyls, acetyls, etc.) was automatedly calculated using E-Dragon software (Talete srl, Milan, Italy). Then, the presence of metabolites in the TIC was matched with functional groups data in order to exclude candidates whose structure was not compatible with observed biotransformations (e.g. loss of methyl from a structure not bearing methyls, glucuronidation on a structure not bearing any site susceptible to conjugation). A further filter was then applied based on a mathematic model correlating RPLC relative retention time (ISTD nalorphine) with a number of parameters estimated for each candidate compound starting from the Simplified Molecular Input Line Entry Specification (SMILES), including the predicted octanol/water partition coefficient (LogP). Results: The procedure was tested on 121 compounds detected in real positive samples, including drugs of abuse (e.g. cocaine, opiates, MDMA), anticonvulsants (e.g. gabapentin, carbamazepine), benzodiazepines (e.g. flurazepam), antidepressants (e.g. citalopram, trazodone, fluoxetine, amitriptyline, venlafaxine), phenothiazines (e.g. chlorpromazine, promazine, pericyazine), antipsychotics (e.g. amisulpride), antihistamines (e.g. cetirizine), beta-blocker (e.g. bisoprolol), anti-retroviral agents (e.g. emtricitabine, tenofovir), acetyl-cholinesterase inhibitors (e.g. rivastigmine), histamine H2-receptor antagonists (e.g. ranitidine), and their phase I metabolites. Overall, the mean list length (MLL) of compounds was 6.71 ± 4.66 (median 6, range 1-28) before the application of the metabolomic approach and was shortened to 3.94 ± 3.07 (median 3, range 1-17) after. For RPLC-HRMS data the MLL was shorted from 6.02 ± 3.49 (median 6, range 2-21) to 3.42 ± 3.03 (median 3, range 1-17) after the metabolomic filter and to 3.09 ± 2.03 (median 2, range 1-9) after the chemometric approach. The application of both filters allowed a reduction of the MLL to 2.14 ± 1.63 (median 2, range 1-9). Conclusion: HRMS allows a much broader search for PTRC than other screening approaches. The combined metabolomic/chemometric approach significantly reduces the list of candidate isomers.