Detection of hair metabolome changes in cocaine users using untargeted metabolomics

Abstract

The aim of this study was to investigate changes in the metabolome of the hair of cocaine users. Studying changes in the hair metabolome can help toxicologists to find robust markers of drug intake that are not affected by long-term perturbations such as circadian variations. Furthermore, the use of an untargeted method allowed us to search not only endogenous compounds but also possible new cocaine metabolites. An untargeted metabolomics method was applied to hair samples using liquid chromatography coupled to high-resolution time-of-flight mass spectrometry (HPLC-HRMS). Hair samples were obtained from 153 cases analyzed previously at the Center for Forensic Hair Analytics and divided in three groups according to their cocaine content (negative cohort: no cocaine detected, n = 52; below cut-off cohort: between 100 pg/mg and 500 pg/mg (the cut-off for determining a positive result in hair), n = 52; and true positive cohort: > 500 pg/mg, n = 49). Samples were decontaminated by sequential washes with DCM, acetone, deionized water and acetone and left to dry overnight. Then, the hair was cut in fine pieces and 30 mg were transferred to Eppendorf tubes and milled for 10 minutes at 30 Hz. Extraction was performed by adding 1 mL of ACN/H2O 2:8 (v/v) and ultrasonicating for 16 h. After centrifugation and evaporation of the supernatant, samples were reconstituted in 250 μL ACN/H2O 2:8 (v/v), filtered and transferred into vials in two 80 μL aliquots, pooling the rest into a QC solution. The analysis was performed using a reverse phase column [Waters XSelect HSST RP-C18 column (150 mm × 2.1 mm, 2.5 μm) in positive electrospray ionization mode]. Untargeted data processing, statistical analysis and tentative identification were performed using MSDial, R studio (version 4.1.3), and Sirius (v. 5.4.1), respectively. The true positive and the negative cohort groups were selected for a first statistical assessment. Comparing the mean areas of each feature in both groups allowed the selection of 1387 features with a fold change between the groups higher than 2-fold. A Mann–Whitney–Wilcoxon test was then performed for these selected features and 133 were found to differ significantly between groups ( P < 0.05). Identification was attempted using the Sirius software, identifying cocaine, benzoylecgonine, ecgonidine, methylecgonidine, norcocaine, the adulterant levamisole and 3 additional features with common fragments of cocaine or benzoylecgonine. Using the commercially available databases (Bio Database, COCONUT, HMDB, MeSH, NORMAN, PubChem, PubMed, YMDB, Zinc Bio…) did not provide any unambiguous identifications of endogenous metabolites, but pointed towards discerning the different metabolite classes of other features, finding amino acids and their derivatives, purine and pyrimidine derivatives, carbohydrates, peptides and lipids, for another 100 features. The identification of cocaine and its metabolites proves the general suitability of the described hair metabolomics workflow to identify differing compounds between cocaine positive and negative samples, since those are indeed the principal differing compounds between the positive and negative groups. Moreover, more features of different classes were significantly different between both groups. Approximately 65% of the features increased in concentration in cocaine positive samples while 35% decreased in concentration. Further investigation should be performed to identify them and study the applicability to discerning cocaine use via hair metabolome. The use of untargeted metabolomics analysis allowed the detection of different cocaine metabolites and possible endogenous biomarkers of cocaine use.