Cytochrome P450 (CYP) enzymes play a major role in drug metabolism. Many of these enzymes exhibit a huge inter-individual variation in activity, not only attributable to the individual's genotype. Such environmental influences are hardly measurable, let alone their outcome predictable. The actual phenotype is usually determined through administration of well-studied CYP probe substrates – an approach that is obviously unsuitable for forensic investigations, particularly postmortem. Therefore, we aimed at finding alternative biomarkers of the CYP phenotype. Application of metabolomics techniques in theory should reveal endogenous metabolites processed by CYP isoenzymes, which in turn could be indicative for a person's phenotype for particular CYP isoenzymes (e.g. 2D6). In a standard metabolomics approach, features (unidentified metabolites) could be compared between different known CYP genotype/phenotypes. While this approach is most promising, a controlled study needs to be conducted including genotyping and probe-substrate phenotyping. Instead of a new phenotyping study, we aimed to explore the intended approach with a previous crossover study with a typical CYP2D6 substrate (MDMA) and a known inhibitor of CYP2D6 (bupropion) for the general suitability to find endogenous correlates of CYP2D6 activity.

Plasma samples of a controlled MDMA/bupropion crossover study with four experimental test sessions (pretreatment – study day as follows: placebo-placebo, bupropion – placebo, placebo-MDMA, and bupropion-MDMA, n=16 each), initially intended to study pharmacological questions, were reused. Plasma from timepoint 9 from the placebo-placebo and bupropion-placebo session were worked-up by simple protein precipitation and measured with a previously published untargeted metabolomics method (Boxler. Drug Test Anal 2019;11(5):678–696). Briefly, LC-qTOF-MS (Sciex 6600) with reversed-phase (RP) and HILIC chromatography was used in both positive and negative ESI mode. Untargeted data was analysed using msDial 4.8 and a pre-release version of SIRIUS (5.4.1) for structure elucidation and comparison to common databases. Feature peak areas were correlated (spearman) with the ratio of the area under the curve (AUC) of MDMA metabolites/AUC MDMA as determined as the best CYP2D6 correlate in a former targeted analysis (Steuer. Plos One 2016;11:e0150955). Further filter criteria were: fold change between groups of>2 or<0.5 and mean S/N>150.

Out of 24858 features found in RP positive mode, by correlation analysis, 27 features with a spearman correlation greater than 0.55 and with assigned MS/MS and 109 features without MS/MS information were found, respectively. As expected, 3 of the features could be unambiguously identified to be bupropion and its known metabolites. Further 10 features indicated a chlorine cluster, not expected endogenously, which pointed to further fragments/metabolites or artifacts of bupropion. The remaining 14 features could be considered as promising biomarker candidates not related to bupropion and as such as tentative indicators of the respective CYP2D6 phenotype.

Using the pharmacokinetic parameters of MDMA, acting as a CYP2D6 substrate, with and without CYP2D6 inhibition by bupropion and corresponding metabolomics data, we could show, that the metabolome approach is able to find potential biomarkers for actual CYP2D6 status. As unfortunately typical for untargeted metabolomics, final identification of features still remains the bottleneck and requires further investigation. In addition, the applicability of the found biomarkers of course needs to be tested in a larger independent cohort. Ideally, a ratio of features that are substrate and product of the enzyme could be formed to increase the predictive power of this approach.