Standards of drugs (Table 1) and their metabolites: quetiapine, 7-hydroxyquetiapine, norquetiapine, clozapine, clozapine-N-oxide, N-desmethylclozapine, aripiprazole, dehydroaripiprazole, risperidone, 9-hydroxyrisperidone, citalopram, N-desmethylcitalopram, venlafaxine, O-desmethylvenlafaxine, vilazodone, M10 metabolite, vortioxetine, levomepromazine, perazine, olanzapine, and N-desmethylolanzapine were obtained from Sigma-Aldrich Chemie (Steinheim, Germany). Water (99.9%), methanol, acetonitrile, ammonium acetate, ammonium formate, ammonia, and formic acid (all LC-MS purity) were supplied by Sigma-Aldrich Chemie (Steinheim, Germany). Isopropanol (99.9%) was purchased from J.T. Baker (Deventer, Netherlands). Water was obtained from Milli-Q deionized water (MiliporeIntertech, Bedford, USA). Methanol pure p. a. (POCH S.A., Gliwice, Poland) was used to clean the EC system. GSH (LC-MS grade, Sigma-Aldrich Chemie, Steinheim, Germany) was used for phase II conjugation reactions. Human liver enzymes of the microsomal fraction (HLMs) (Sigma-Aldrich Chemie, Steinheim, Germany), nicotinamide adenine dinucleotide β-phosphate (β-NADPH) tetrasodium salt (Sigma-Aldrich Chemie, Steinheim, Germany), magnesium chloride hexahydrate (MgCl₂×6H₂O), potassium chloride (KCl), potassium diphosphate (KH₂PO₄), ethylenediaminetetraacetic acid (EDTA) (Sigma-Aldrich Chemie, Steinheim, Germany) were used for in vitro approaches.

Putative metabolites of the target pharmaceuticals were predicted using selected compounds’ SMILES strings.

(QUE: N\1 = C(\c3c(Sc2c/1cccc2)cccc3)N4CCN(CCOCCO)CC4;

Through the free online software GLORYx (de Bruyn Kops et al., 2021) and Biotransformer 3.0 (Djoumbou-Feunang et al., 2019). Briefly, GLORYx integrates machine learning-based site of metabolism (SoM) prediction with biotransformation reaction rule sets to predict and classify the putative structures of metabolites that could be formed by phase I and/or phase II metabolism. Biotransformer 3.0 is an open-access software tool that supports the rapid, accurate, and comprehensive prediction of small molecules’ metabolism in mammals and environmental microorganisms. ChemDraw software generated all metabolite structures (PerkinElmer, Inc., Waltham, MA, USA).

Electrochemical experiments were performed with a ROXY™ system (Antec, Alphen aan den Rijn, Netherlands) containing a potentiostat, an oven compartment, and an infusion pump. A ReactorCell™ (Antec, Alphen aan den Rijn, Netherlands) was employed for the electrochemical transformation. The ROXY™ system was controlled via the Dialogue™ software (Antec, Alphen aan den Rijn, Netherlands). Oxidation products were directly transferred to the electrospray ionization (ESI) source of the mass spectrometer (Agilent, Waldbronn, Germany), which was operated in positive ion mode. The MassHunter software (Agilent, Waldbronn, Germany) was used to controlthe mass spectrometer and data analysis. The experiments consisted of two stages: electrochemical oxidation for phase I-like transformation and the addition of GSH for phase II-like conjunction. To carry out the first stage of experiments, 5 and 10 μM solutions of the drug in the selected mobile phase were prepared. At a flow rate of 10 μL/min, the solution of the target compound was passed through the electrochemical cell heated to 36,7 °C with selected electrode by a syringe pump. The electrochemical cell was coupled online and offline to the mass spectrometer. The recorded datawere obtained by increasing the working electrode potential linearly from 0 to 2000 mV (for glassy carbon (GC), gold (Au), and platinum (Pt) electrodes) or from 0 to 3,000 mV (for the BDD electrode) at a rate of 10 mV/s (Supplementary Table S1). For phase II-like conjugation reactions, the cell effluent was mixed with a GSH-solution in the same mobile phase as the analyzed drug in a reactor coil. The effluent of the reaction coil was transferred online and offline to the mass spectrometer.

Operating parameters of the triple quadrupole mass spectrometer with an ESI interface operating in positive ion mode for the analyzed drugs and their metabolites, biological samples, and in vitro tests were carried out on LC-ESI-MS/MS 8050 (Shimadzu, Tokyo, Japan). LC measurements were carried out using an ACE C18 column (150 mm × 4.6 mm) with mobile phase A (water with 0.1% formic acid and 2 mM ammonium formate) and mobile phase B (acetonitrile with 0.1% formic acid and 2 mM ammonium formate) at a flow rate of 0.4 mL/min. Table 2 shows the relevant mass spectrometric operating parameters. The conditions were determined using a central composite design (CCD) (Supplementary Table S2; Supplementary Figure S1).

Whole blood samples (peripheral venous blood) were collected from patients. Sample collection was performed by doctors of the II Department of Psychiatry of the Ludwik Rydygier Provincial Polyclinical Hospital in Torun. The study was approved by the Nicolaus Copernicus University Bioethics Committee in Torun (decision no. 477/2021). All study participants were informed verbally and in writing about the purpose of the study and completed a questionnaire and a declaration of voluntary consent to participate in the study. Raw blood samples (2 mL) were collected from adult patients treated with selected psychotropic drugs (aripiprazole, clozapine, quetiapine, venlafaxine, vortioxetine) after oral administration of the drug (minimum period of 2 weeks). A chiral drugs in a pure enantiomeric form were applied. It was significant because enantiomers can have different biological effects, therapeutically beneficial, potentially harmful, or inactive. It was crucial for optimizing therapeutic efficacy, minimizing adverse effects, and advancing drug development and personalized medicine. The pharmacokinetics of antidepressants is stereoselective and predominantly favors one enantiomer. The use of pure enantiomers offers (i) better specificity than the racemates in terms of certain pharmacological actions, (ii) enhanced clinical indications, and (iii) optimized pharmacokinetics. Therefore, controlling the stereoselectivity in the pharmacokinetics of antidepressive drugs is of critical importance in dealing with depression and psychiatric conditions. Whole blood samples were collected in tubes. Then, after centrifugation of the morphotic elements, the plasma was frozen (−80 °C) and thawed only immediately before performing the analytical procedure to evaluate in vivo metabolism.

MEPS was used to reach this objective.

Plasma-fortified samples were used to select the extraction parameters. A summary of the tested MEPS sorbents is shown in Supplementary Table S3.

In vitro drug incubations were carried out in the microsomal fraction of liver cells to obtain and analyze metabolites. Supplementary Table S4, shows the composition of the microsomal fractions used in this study and their activity. The drug incubation (10 μL of a 2 mM solution of the respective drug in the phosphate saline buffer) with human liver microsomes (CYP450; 10 μL of 20 mg/mL) was performed in 35 μL phosphate buffer (BR) with the addition of 10 μL NADPH (20 mM). All the samples were incubated at 37 °C for 0, 15, 30, 60, 90 or 120 min and shaken during the incubation. The reaction was terminated by the immersion of the samples in an ice bath and the addition of 65 μL ice-cold ACN. After that, all samples were centrifuged for 5–15 min at 5000rpm. Then, the supernatant was analyzed by LC- MS/MS.

The method was validated according to the ICH guidelines. The following validation characteristics were addressed: linearity, detection and quantitation limits, accuracy, precision, robustness, selectivity and stability (Supplementary Tables S5–S7).

The influence of the electrode type, pH, and type of buffer on the EC analysis of psychotropic drugs was discussed, and the most optimal conditions were determined.

During the analysis, four types of electrodes were tested: platinum (Pt), gold (Au), glassy carbon (GC), and boron-doped diamond electrode (MD) consisting of an ultra-thin crystalline diamond layer deposited on top of a silicon substrate. The criterion for choosing a working electrode was the amount and intensity of the signals on the mass spectrum. Figure 2 shows an example of MS spectra obtained for olanzapine using four different working electrodes.

Analyzing the obtained MS spectra of the tested compounds, it can be concluded that the highest number of signals is possible when using a boron-doped diamond MD electrode. The efficiency of this working electrode is probably related to the possibility of conducting electrochemical processes in a wider potential range (0–3000 mV).

During the analysis, four types of buffer (concentration 10 mM) with different pH were tested: ammonium formate buffer at pH 3 (A), ammonium acetate at pH 7(B), ammonium acetate at pH 5 (C), and ammonium acetate at pH 9 (D). The criterion for selecting the appropriate buffer and pH was the amount and intensity of signals in the mass spectra for selected psychotropic drugs. Figure 3 shows an example of MS spectra obtained for clonazepam in four different MD electrode solutions.

Analyzing the mass spectra of the studied compounds (Figure 3), it can be concluded that the highest number and intensity of signals correspond to electrochemical processes using the mobile phase at pH values of 5 and 7. The ammonium acetate buffer was selected for further study due to the highest intensity of signals.

EC-MS data was collected for the eleven psychotropic drugs studied in this work. Psychotropic drug intensity decreases with the increasing potential. This suggests a degradation of the parent compound. At the same time, an increase in signal intensity is observed for individual electrochemical transformation products.

An exemplary mass voltammogram recorded on a BDD electrode in a potential range from 0 to 3000 mV is shown in Figure 4 for quetiapine.

The most efficient formation of detectable electrochemical transformation products was observed in a potential range from 1800 to 2,200 mV. At higher potentials, the intensity electrochemical transformation products decreases. Similar effects occurred for other psychotropic drugs. Based on potential-dependent signal intensities and MS/MS spectra, structures were proposed for electrochemical transformation products (see Table 3).

The structural formulas presented in the article are exemplary structures of the obtained and identified electrochemical products. A nuclear magnetic resonance (NMR) spectroscopic analysis should be performed to determine the exact structures of the obtained products. The path of the proposed electrochemical reactions including reactions characteristic of the first and second phases of biotransformation are shown in Supplementary Figures S3–S7. The chemical structures were proposed based on EC-MS/MS experiments and information available in literature. Four proposed electrochemical products were identified for the electrochemical simulation of vortioxetine (Supplementary Figure S3). The electrochemical product M1 observed for m/z = 315 indicates a hydroxylation reaction probably at the methylene group site; M2 for m/z = 329 is a continuation of the response for M1 and the formation of a carboxyl group at the methylene group site; M3 and M4 are the typical attachment and detachment of two hydrogen atoms that are present in virtually every electrochemical simulation. Four proposed electrochemical products were identified for the electrochemical simulation of aripiprazole (Supplementary Figure S4). The electrochemical product M1 and M4 are typical attachment and detachment reactions of two hydrogen atoms (m/z = 446 and m/z = 450). In contrast, M2 characteristic for m/z = 464, is a hydroxylation reaction of the aromatic ring; M3 for m/z = 432 is the result of detachment of an oxygen atom from the cyclic ring. Five proposed electrochemical products were identified for the electrochemical simulation of citalopram (Supplementary Figure S5). The electrochemical product M1 is probably an additional reaction of an oxygen atom on a nitrogen atom; M2 for m/z = 311 indicates a demethylation reaction at the nitrogen atom. Attaching an oxygen atom to a nitrogen atom, M3 and M4 are the typical attachment and detachment of two hydrogen atoms (m/z = 327 and m/z = 323)—the product M5 for m/z = 297 results from two demethylation reactions at the nitrogen atom. Five proposed electrochemical products were identified for the electrochemical simulation of venlafaxine (Supplementary Figure S6). The electrochemical product M1 for m/z = 262 is probably a dehydroxylation reaction at the cyclic ring site; M2 and M4 are typical attachment and detachment reactions of two hydrogen atoms (m/z = 280 and m/z = 276); M3 for m/z = 294 is probably an oxygen atom attachment reaction at the nitrogen atom. The product M5 for m/z = 264 results from demethylation at the oxygen atom. Eight proposed electrochemical products were identified for the electrochemical simulation of vilazodone (Supplementary Figure S7). Electrochemical products M1, M3, and M5-7 are hydroxylation reactions of the attachment of one or more OH groups at different sites of vilazodone. Product M2 is the result of dehydrogenation, while M4 is probably the attachment of two oxygen atoms to the cyclic ring of the piperazine derivative. The m/z = 747 ion for the electrochemical product M8 indicates the addition of glutathione to vilazodone.

In silico prediction of metabolism was performed on GloryX and Biotransformer3.0 free ware. Moreover, a robust approach to ranking the predicted metabolites is attained by using the SoM (sites of the metabolism) probabilities predicted by the FAME 3 (Supplementary Table S8) machine learning models to score the predicted metabolites. Predicted derivatives with an assigned score were considered probable or minority and less likely. Overall, Phase I and Phase II metabolites were predicted. Major Phase I metabolites included hydroxylated and oxidated derivatives. Major Phase II metabolites included glutathione conjugated, sulfo conjugated, and methylated derivatives.

Incubating the microsomal fraction of liver cells without NADPH at 37 °C for a specified time did not affect the amount of the compound in the reaction mixture. The results indicate that the enzymatic assays should be carried out with NADPH cofactor. The degree of conversion of psychotropic drugs depended on the concentration and incubation time of NADPH—the results of the experiments illustrated for clozapine are shown in Supplementary Figure S8. In the graph, the substrate concentration decreased linearly with the progress of the reaction. No signals from other substances were observed during the analyses when the drug was incubated without the cofactor. The results were also confirmed for the other psychotropic medications.

In the study, we determined the chemical structures of the resulting metabolic biotransformation products of the studied drugs. Mass spectra were recorded for each analyte tested (substrates) and potential metabolites (products) using the HPLC-ESI/MS technique. The Mass Hunter Metabolite ID computer program was helpful during the analysis of the results. This analytical tool compared the analyte signals with the control samples by comparing their retention times, MS spectra, and area fields. MS spectra and [M + H]⁺ fragmentation spectra for the proposed metabolites were recorded for the available metabolite standards of selected psychotropic drugs. This was not possible for the other identified potential metabolites due to the lack of commercially available standards or their high price. All the commercially available standards of metabolites discussed in the manuscript are the predominant metabolites for each drug (pharmacologically active metabolites).

Using pre-optimized mass spectrometer analysis conditions, quetiapine (QUE as a model compound) was identified as [M + H]+ for m/z = 384. Fragmentation of QUE and other drugs was performed to obtain data to interpret MS spectra for potential metabolites of selected drugs. Incubation on the example QUE along with liver enzymes of the microsomal fraction in the presence of NADPH as a cofactor for enzymatic reactions led to the formation of five potential metabolites (S/N˃5), (Figure 5).

The analysis of the MS chromatogram below indicates that five (M1-M5) metabolites of phase I of the biotransformation reaction were formed. The potential metabolites were numbered M1-M5 based on their retention times. The chemical structures proposed for them were confirmed through parallel detection of 2 MS/MS fragmentation transitions for the selected molecule (potential metabolite).

Using a mass spectrometer, the various potential metabolites were identified and confirmed based on the proposed fragmentation mechanism (Supplementary Figure S9; Supplementary Figure S10). The specific fragments were identified on the characteristic ions, indicating the relevant functional groups in the molecule. Mass ions with characteristic m/z values were observed on MS spectra of the potential metabolites of microsomal fraction enzymes (Table 4). After incubation, the potential metabolites were formed due to the phase I biotransformation. Table 4 characterizes the retention times, chemical structures, m/z ratio, elemental composition, and a proposed biochemical reaction depending on the selected psychotropic drug. Interpreting the collected mass spectra (MS) of the peaks of the obtained metabolites allows us to conclude that the other studied psychotropic drugs are susceptible to enzymatic transformations in the presence of selected cosubstrates.

Taking everything into consideration human microsomes are a suitable model for studying and predicting the metabolic transformations of the analytes that may occur in the patient’s body. All proposed phase I reactions of metabolism are typical of processes catalyzed by the isoenzymes of the cytochrome P450 group.

The isolation of the tested compounds was carried out using different MEPS sorbents and the eVol^® automatic electronic syringe. Depending on the filling used, different efficiency, selectivity, and reproducibility of extraction concerning selected psychotropic drugs and their metabolites were obtained.

Analyzing the obtained results, it can be concluded that the use of the C18 sorption for the isolation of selected analytes from contaminated plasma samples allowed for higher recoveries (average recovery 94.35% ± 1.72%) than in the case of the use of other sorption beds, while maintaining reasonable repeatability determinations (average CV value was approx. 3%). An additional advantage was the possibility of using the C18 sorption bed in a wide pH range (3–9) of the tested samples. A comparison of the obtained recoveries using selected MEPS sorption beds is presented in Supplementary Table S9.

Sorbents with octadecylsilane filling showed high extraction efficiency (recovery) and repeatability of the tested drugs. In contrast, the remaining sorption beds tested in the study had low extraction efficiency for most tested compounds. In addition, silica gel, modified with octadecyl groups (C18), is the most commonly used sorption material. Although the C18 sorbent is non-polar, it can be used for both polar and non-polar compounds. This is due to their long octadecyl chain, in which drugs, as non-polar compounds, get stuck and easily wash out with a polar solvent.

The next step was the selection of the solvent used for the elution of analytes from sorption. Depending on the type of eluting solvent used, various recoveries of analytes from plasma samples were obtained and related to their diversified polarity. In the case of methanol, the most minor average recovery of psychotropic drugs was obtained, which was about 50%. Modifying the solvent composition by adding water resulted in a significant increase in the average recovery of analytes to about 75%. Elution with buffer solutions at pH = 3, 5, and 9 did not improve extraction efficiency compared to methanol: water (50:50; v/v). The best results (mean recovery 98.16% ± 1.75%) were obtained using a mixture of acetonitrile: methanol: water (5:3:2; v/v/v) as the extraction medium. Supplementary Table S9, summarizes the results obtained after using various sorbents for extraction and an example solvent for the elution of analytes (acetonitrile: methanol: water (5:3:2; v/v/v)). The next step in selecting appropriate extraction conditions was to study the effect of the number of extraction cycles (1, 2, or 4). The best recovery for each analyte was obtained by performing two extraction cycles, where recoveries ranged between 91% and 99%.

In the study of biological samples, the microextraction technique on packed sorbent (MEPS) was used to isolate and enrich analytes from plasma samples. Each patient was treated with a different psychotropic substance. The identified metabolite in the plasma sample of a patient treated with aripiprazole (30 mg dose) was its primary metabolite, dehydroarypiprazole (DARI). The molecular ion of this compound is characteristic of m/z = 446. The MS fragmentation spectrum for DARI with a distinctive signal for m/z = 285 (with the highest intensity) is confirmed by the MS spectrum for the sample of the metabolite standard (Figure 6). In the case of a plasma sample from a patient treated with venlafaxine (225 mg dose), the MS fragmentation spectrum shows two signals for m/z = 246 (with the highest intensity) and m/z = 107, which are characteristic ions for its metabolite O-desmethylenlafaxine. The result is consistent with the analyses performed for the standard metabolite of this drug, DVEN (Figure 7). For a plasma sample of a patient treated with clozapine, the MS and MS fragmentation spectra in Figure 8 identified the primary two metabolites: clozapine N-oxide and N-desmethylclozapine. Figure 8A shows the two fragmentation ions m/z = 192 and m/z = 256 (with the highest intensity) characteristic of NoxCLO. In contrast, Figure 8B shows two fragmentation ions for m/z = 192 (with the highest intensity) and m/z = 227, characteristic of DCLO. Four of its metabolites were identified in the plasma sample of a patient treated with quetiapine (750 mg dose). Figure 9A shows the MS spectrum for 7-hydroxy quetiapine, as indicated by the molecular ion m/z = 400 and the characteristic signals for m/z = 269 (highest intensity) and m/z = 237 on the MS fragmentation spectrum. The result is consistent with the data obtained for the HQUE metabolite standard. Figure 9B shows a spectrum with a molecular ion for m/z = 400 and characteristic fragment ions for SoxQUE m/z = 221 (highest intensity), m/z = 253 (medium intensity) and m/z = 279. Figure 9C shows an MS spectrum for NQUE, as indicated by the presence of a molecular ion for m/z = 296 and characteristic fragment ions for m/z = 210 (highest intensity) and m/z = 183 (medium intensity). Figure 9D shows three fragmentation ions for m/z = 269, m/z = 237 (with medium intensity), and m/z = 226 (with highest intensity) characteristic of HDQUE.

For a patient treated with vortioxetine (20 mg dose), four metabolites were identified in a plasma sample. Figure 10 shows MS spectra with signals from the precursor ion of the metabolites: m/z = 315 for M8 (A), m/z = 329 for M0 (B), m/z = 391 for M11 (C), and m/z = 491 for M3 (D). These results are confirmed in the literature. Due to the lack of commercially available metabolite standards or their high price, MS fragmentation was not performed.

Based on the results, it can be concluded that the MEPS approach can be successfully adopted at the stage of biological sample preparation as a microextraction technique for isolating drug metabolites.