Establishing a cell-based screening workflow for determining the efficiency of CYP2C9 metabolism: moving towards the use of breath volatiles in personalised medicine

The use of volatile biomarkers in exhaled breath as predictors to individual drug response would advance the field of personalised medicine by providing direct information on enzyme activity. This would result in enormous benefits, both for patients and for the healthcare sector. Non-invasive breath tests would also gain a high acceptance by patients. Towards this goal, differences in metabolism resulting from extensive polymorphisms in a major group of drug-metabolizing enzymes, the cytochrome P450 (CYP) family, need to be determined and quantified. CYP2C9 is responsible for metabolising many crucial drugs (e.g., diclofenac) and food ingredients (e.g., limonene). In this paper, we provide a proof-of-concept study that illustrates the in vitro bioconversion of diclofenac in recombinant HEK293T cells overexpressing CYP2C9 to 4ʹ-hydroxydiclofenac. This in vitro approach is a necessary and important first step in the development of breath tests to determine and monitor metabolic processes in the human body. By focusing on the metabolic conversion of diclofenac, we have been able to establish a workflow using a cell-based system for CYP2C9 activity. Furthermore, we illustrate how the bioconversion of diclofenac is limited in the presence of limonene, which is another CYP2C9 metabolising substrate. We show that increasing limonene levels continuously reduce the production of 4ʹ-hydroxydiclofenac. Michaelis-Menten kinetics were performed for the diclofenac 4ʹ-hydroxylation with and without limonene, giving a kinetic constant of the reaction, K M, of 103 µM and 94.1 µM, respectively, and a maximum reaction rate, V max, of 46.8 pmol min−1 106 cells−1 and 56.0 pmol min−1 106 cells−1 with and without the inhibitor, respectively, suggesting a non-competitive or mixed inhibition type. The half-maximal inhibitory concentration value (IC50) for the inhibition of the formation of 4ʹ-hydroxydiclofenace by limonene is determined to be 1413 µM.


Introduction
A significant number of patients cannot optimally benefit from available treatments because of individual variations in drug metabolism. Adverse drug reactions are responsible for drug failure and are a major leading cause of death [1].
An important cause for the 'non-responder' phenotype lies in patient-specific genomic variations of the drug metabolizing cytochrome P450 (CYP) enzymes. These CYP enzymes catalyse reactions of the phase I metabolism-including hydrolysis, reduction, and oxidation-yielding more polar and watersoluble metabolites [2,3]. CYP2C9 is one of the most abundant CYP2C isoforms in the human liver and is known to have a highly polymorphic nature [4]. The CYP2C9 enzyme contributes about 20% to the total hepatic P450 protein and is the most relevant CYP enzyme, besides CYP3A4 and CYP2D6, with regards to the number of therapeutic agents oxidised. CYP2C9 accounts for approximately 15% of all clinically used drugs that undergo biotransformation by cytochrome P450 enzymes. A number of drugs metabolised by CYP2C9 belong to the substance class of the non-steroidal anti-inflammatory drugs, which include ibuprofen, naproxen and meloxicam, and anti-coagulants such as warfarin and antiestrogens [5]. Owing to the high polymorphism prevalence of CYP2C9, inter-individual variability in the expression of CYP2C9 protein and activity affects the efficacy and safety of drug treatments [6,7].
Genome-based CYP profiling is critically limited by the highly polymorphic nature of the genes, which makes it difficult to predict the metabolic activities on the encoded proteins in a patient. Factors such as age, polypharmacy, health status, lifestyle, etc, also contribute to patient specific phenotypes [8]. It would therefore be highly desirable to have point-of-care tests at hand, which permit a fast determination of enzyme activity.
A test that detects key volatile metabolites in exhaled breath could potentially be used to identify a CYP's phenotype quickly and non-invasively. A limited number of breath tests, which are based on isotopically labelled probes, are in current use for such clinical diagnosis [9,10]. In these tests, the respective labelled precursor compounds, such as 13 C-pantoprazol for CYP2C19 or 13 C-methacetin for CYP1A2, are metabolised resulting in increases of 13 CO 2 in the exhaled breath, which is measured using infrared spectroscopy [11]. However, the high amount of the isotopically labelled substrates that need to be ingested (several hundred mg) makes the breath test expensive.
A major goal of the work presented here is to establish a workflow for testing potential CYP2C9 substrates that are not isotopically labelled and that give rise, through metabolic processes, to unique and hence unambiguous biomarkers, e.g., volatiles in exhaled breath. The requirements for precursor substrates for use in breath tests are that they (i) are non-toxic, (ii) show high specificity for the CYP isoform of interest, (iii) are efficiently taken up by the organism, and (iv) result in unique volatile metabolic biomarkers in breath. We are undertaking in silico studies to determine appropriate substrates that provide unique volatiles resulting from CYP metabolic processes, which we are currently investigating. These will be the subject of a later paper. In addition, the volatile metabolites in exhaled breath should be accessible for detection using clinical friendly analytical platforms in real or near to real-time with a high chemical specificity and selectivity. Analytical instruments such as proton transfer reaction/selective reagent ion-time-of-flight mass spectrometry and fast gas chromatography-ion mobility spectrometry have been demonstrated to be useful for such applications [12][13][14].
Diclofenac (C 14 H 11 Cl 2 NO 2 ) satisfies most of the above requirements, with the exception for the formation of volatile metabolites, whereas the substrate limonene (C 10 H 16 ) meets all the above-mentioned criteria. Although the metabolite of diclofenac is nonvolatile, the choice of diclofenac to establish a workflow was decided by its well-documented CYP2C9 metabolic pathway, i.e., a substrate that provides a unique metabolic volatile that can be seen in exhaled breath is still be determined. That is the subject for the next stage of our research programme, which we are currently undertaking. CYP2C9 is responsible for the metabolism of diclofenac, yielding 4 ′hydroxydiclofenac as the major metabolite [15,16]. Limonene is metabolised in the liver by two different enzymes, CYP2C9 and CYP2C19, yielding carveol and perillyl alcohol, respectively, as its major products [17].
In this work, we have undertaken in vitro investigations to determine the influence of limonene on diclofenac bioconversion. To achieve this, human embryonic kidney 293 cells (HEK293 cells) as well as microsomes isolated from these cells were utilised to investigate the bioconversion of diclofenac with or without limonene being present. HEK293 cells do not express CYP2C9. However, genetically modified HEK293T cells expressing CYP isoforms, such as CYP2C9, allow the investigation of metabolic experiments based on the specific enzymatic reaction.
Investigations on the impact of different growth media were performed to optimize sample preparation. For efficient cell growth and viability of HEK293 cells, DMEM with 10% FBS was routinely used. However, the use of this percentage of FBS turned it into a complex matrix for ultra-highperformance liquid chromatography/mass spectrometry (UHPLC/MS) analysis. For this reason, DMEM with a reduced amount of FBS, 0.3% instead of 10%, or PBS medium was added to the cells for incubation prior to analysis. After 24 h, cells were washed with 0.1 M PBS at room temperature. DMEM with the two different serum levels or PBS, all containing 100 µM diclofenac, was then added to the cells, and incubated at 37 • C. The progress of bioconversion was evaluated at set time points in the time range from 0 h to 24 h. All further experiments were carried out with 10% FBS.
Key parameters, including the concentration of substrate resulting in half-maximal enzyme activity (Michaelis-Menten kinetic constant of the reaction, K M ), the maximum reaction rate catalysed by the enzyme (V max ) and the half-maximal inhibitory concentration (IC 50 ), determine the inhibition type. Therefore, the recombinant HEK293T cells were incubated with 25-300 µM diclofenac in the absence or presence of 500 µM limonene. In order to account for losses of the sample during incubation and for better comparability, different diclofenac and 4 ′hydroxydiclofenac concentrations were incubated in DMEM with 10% FBS, but without cells as samples for the calibration curves. All plates were incubated at 37 • C and 5% CO 2 for 4 h.
To investigate the effect of limonene on diclofenac metabolism by CYP2C9, a final concentration of 100-1000 µM limonene (dissolved in DMSO and diluted in DMEM with 100 µM diclofenac to a final DMSO concentration of maximal 1%) was added to the cells. HEK293 cells were also treated with medium plus 1% DMSO to examine any impact of the solvent. The IC 50 value for the inhibition of diclofenac by limonene was determined by incubation of the cells with 125-3000 µM limonene and 100 µM diclofenac. Samples were incubated at 37 • C and 5% CO 2 for 4 h.
We further integrated the trypan blue exclusion assay into the workflow to investigate the effects of diclofenac on cell viability. After removing the supernatants, cells were trypsinised and collected in 1 ml 10% DMEM. Cell suspensions were mixed 1:1 with trypan blue dye and viable and dead cells were counted using a Neubauer cell counting chamber.

Sample preparation
After given incubation times, cell supernatants were collected in fresh 1.5 ml Eppendorf tubes, diluted 1:1 with ACN and then centrifuged at 10 000 × g for 10 min to remove any floating cells or debris. The samples were 100-fold diluted with DMEM with 10% FBS/ACN in a 1:1 ratio. This has the additional advantage of protein precipitation if any cell residues are present. Polytetraflon ® (PTFE) syringe filters with 0.22 µm pore size (Agilent Technologies, Santa Clara, CA, USA) were used to remove any solid parts from the samples. Diclofenac and 4 ′ -hydroxydiclofenac solutions for calibrations were incubated also in DMEM with 10% FBS simultaneously to the cell experiments and treated in the same way as the cell samples to account for any effects from the incubation or possible influences of the matrix that might lead to a loss of substances. The samples were then analysed using an UHPLC/MS (see section 2.5).

Treatment of microsomes with diclofenac
Microsome preparation from HEK293 and HEK293T-CYP2C9 cells was carried out using Merck's Microsome Isolation Kit according to the manufacturer's protocol [19], with the following modifications: the cell suspension was homogenized with 20 strokes using the dounce homogenizer and the last centrifugation step at 20 000 × g and 4 • C was performed for a total of 60 min. The extracted microsomes were immediately used or were frozen at −80 • C for later use. To the microsomes (20 µg in 100 µl final volume), 50 µM diclofenac, a NADPH regeneration system and 0.1 M PBS (pH 7.4) were added. Incubation was set at 37 • C and 400 rpm in a thermocycler for 0.5 h, 1 h, 1.5 h, or 2 h. The reaction was quenched with 50 µl cold ACN on ice to precipitate proteins, followed by centrifugation at 10 000 × g for 10 min. The supernatant was then transferred to a fresh vial and diluted 1:25 in 50% ACN for further analysis with the UHPLC/MS.
In addition to the above, we investigated the influence of different microsome concentrations on the metabolism of the substrate through the CYP2C9 enzyme. The diclofenac concentration (50 µM) as well as the incubation times were kept constant, whereas the amount of microsomes were varied by adding 5, 10 or 20 µl of 1 mg ml −1 microsomes to a total volume of 100 µl.

LC/MS analysis of diclofenac and 4 ′ -hydroxydiclofenac
Both diclofenac and 4 ′ -hydroxydiclofenac were measured by use of a reversed-phase UHPLC/MS. For this study, a Vanquish UHPLC system (Thermo Fisher Scientific, Waltham, MA, USA) with a binary pump was used. Aliquots of cell supernatants were filtered by a PTFE syringe filter as described above, and were directly injected into a ZORBAX Eclipse XDB-C18 (Agilent, Santa Clara, CA, USA) column (2.1 × 100 mm, 1.8 µm) protected by a guard column ZORBAX Eclipse Plus C18 (Agilent, Santa Clara, CA, USA) (2.1 × 5 mm, 1.8 µm) at 40 • C. The flow rate was set to 0.25 ml min −1 with water and ACN, with both containing 0.1% formic acid as the mobile phase. Chromatographic separation was achieved using the following gradients: 0-5 min 0%-90% ACN; 5-7 min 90% ACN; 7-7.2 min 90%-0% ACN and 7.2-9.2 min 0% ACN for column washing and equilibration. The retention times of diclofenac and 4 ′ -hydroxydiclofenac were 5.90 min and 5.25 min, respectively. The temperature of the sample tray was set to 6 • C and the sample volume injected was 1 µl.
All mass spectrometric measurements were performed with an Orbitrap Q-Exactive plus Vanquish Flex system (Thermo Fisher Scientific, Waltham, MA, USA) operating in positive ionization mode using a Heated Electro Spray Ionization source with the spray voltage set to 3.5 kV. Nitrogen was used as sheath, auxiliary and sweep gases, with flows set to maximise the ion signals. The capillary and auxiliary heater temperatures were set to 250 • C and 350 • C, respectively. The mass spectrometer scan range was set to 50-500 amu, and a scan rate of 1000 amu s −1 was applied. The mass resolution (m/∆m) of the instrument is 35 000.
In order to quantify diclofenac (C 14 H 11 Cl 2 NO 2 ) and 4 ′ -hydroxydiclofenac (C 14 H 11 Cl 2 NO 3 ), the two natural isotopes of Cl ( 35 Cl and 37 Cl) were taken into account when using the protonated parents. Namely, we recorded the intensities of m/z 296.02396, 297.0273 and 298.02101 (with an intensity distribution of 100: 15.8: 66.3) for diclofenac, and m/z 312.01888, 313.02223 and 314.01593 (with an intensity distribution of 100: 15.1: 64.8) for 4 ′hydroxydiclofenac. Limonene was not measured directly via UHPLC/MS, but its inhibitory effect on diclofenac conversion was quantified.

P450-Glo assay
We measured CYP2C9 activity in intact HEK293 cells as well as in microsome preparations using Promega's P450-Glo Assay, according to the manufacturer's protocol (Promega Corporation, USA).

Statistical analyses
Values shown in graphs (see results later) represent a mean value ± one standard deviation of at least two individual experiments, each with duplicate or triplicate measurements in separate wells of the culture plate or test tube. Limits of detection and limits of quantification were determined according to the German industry norm DIN32645: 2008-11.
Data visualization was undertaken using the OriginLab © (Origin Pro 2022) software. GraphPad Prism 5.0 (GraphPad Software Inc., San Diego, CA, USA) was used to calculate Michaelis-Menten enzyme kinetics, including V max and K m , to perform linear regression analyses for the Lineweaver-Burk plot and to determine IC 50 values. To compare groups, one-way ANOVA with Tukey's multiple comparison test was performed. An F-test was used to determine statistical differences between V max and K m . Results were considered to be statistically significant if p < 0.05.

Impact of different growth media on diclofenac bioconversion
Investigations of cell viability showed that in the time range of up to 4 h, neither the growth medium nor the presence of diclofenac had any significant negative impact on cell viability (see figure 1(a)). Investigating the bioconversion of diclofenac in different media showed no difference between 0.3% serum in DMEM and PBS. The results showed a statistically significant higher yield of 4 ′ -hydroxydiclofenac using the recommended DMEM with 10% FBS, (see figure 1(b)), and therefore all further experiments were performed using this medium. To account for ion suppression and matrix effects, which originate from the growth media, samples for calibrations were also incubated in this growth medium under the identical conditions as used for the cells.

Bioconversion of diclofenac to 4 ′ -hydroxydiclofenac by CYP2C9 in HEK293T cells
Experiments with HEK293T cells overexpressing CYP2C9 enzymes incubated with 100 µM diclofenac at set time points (over a period of 0 h-24 h) revealed that as a result of bioconversion the concentration of the metabolite 4 ′ -hydroxydiclofenac increased as the substrate concentration decreased with time (see figure 2). After 24 h, a final 4 ′ -hydroxydiclofenac concentration of 34 ± 1 µmol l −1 and a final diclofenac concentration of 63 ± 1 µmol l −1 were measured.
Analysis of HEK293 cell supernatant displayed no 4 ′ -hydroxydiclofenac signal at any given time. This provides proof that the bioconversion of diclofenac is a result of the expressed CYP2C9 enzymes and does not originate from other cellular or technical mechanisms appearing during sample preparation or measurements.

Bioconversion of diclofenac to 4 ′ -hydroxydiclofenac via CYP2C9 in microsomes
Similar results to the above have been obtained with microsomes isolated from these cells. The yield of 4 ′hydroxydiclofenac increases with increasing microsome amount (see figure 3(a)). This indicates that the amount of CYP2C9 enzymes has a major impact on the bioconversion of the substrate added.
Regarding the analytical performance of the UHPLC/MS, the use of microsome amounts of 5 µg or less resulted in signals for 4 ′ -hydroxydiclofenac  near to the detection limit of 0.395 µM and below the quantification limit of 1.44 µM. By doubling the microsome quantity used from 10 to 20 µg, the amount of the metabolite is, within experimental error, also doubled from approximately 1.7 ± 0.5 µM to 4.0 ± 1.1 µM. The small discrepancy in the measurements are attributed to the use of small volumes associated with the isolation of the microsomes, which very likely caused slightly different amounts of the enzymes being present in different batches.
Alternative reactions being responsible for the formation of 4 ′ -hydroxydiclofenac can be eliminated owing to the absence of signals from the negative control samples (microsomes isolated from HEK293 cells).
In order to determine the maximal conversion rate, a constant amount of microsomes (20 µg) were incubated with 50 µM diclofenac for different times between 30 and 120 min. Figure 3(b) shows the concentration of 4 ′ -hydroxydiclofenac formed at different time points after analysis of the microsome supernatant via UHPLC/MS. The 4 ′ -hydroxydiclofenac concentration is found to increase with increasing incubation time, and has a maximum recorded value at the 90 min time point.
We determined the relative CYP2C9 enzyme activity with the Promega's P450-Glo Assay. We found a pronounced and dose-dependent increase in the luminescence signals of the microsomes derived from CYP2C9 expressing cells compared to the microsomes from the HEK293 cells (see figure 4).

Impact of limonene on diclofenac bioconversion
Different substrates that are metabolised by CYP2C9 can be expected to influence the efficiency of bioconversion. To illustrate this, we carried out experiments to test the dependency of the 4 ′ -hydroxydiclofenac formation with increasing limonene concentrations (see figure 5), with the concentration of diclofenac held constant at 100 µM. Figure 5 clearly demonstrates that the metabolic competition of the two compounds (diclofenac and limonene) results in a reduction of 4 ′ -hydroxydiclofenac formation.

Michaelis-Menten kinetics of diclofenac conversion and inhibitory effect by limonene
The Based on the progression of the curves from the Michalis-Menten plot and the fitted data from the Lineweaver-Burk plot, the inhibition in the observed system was found to be between non-competitive inhibition type (same x-intercepts at y = 0 for both conditions) and mixed inhibition type. In line with this, no statistical difference is observed between the determined K m values. However, the V max value is statistically significantly lower (p = 0.01) following the addition of limonene. The IC 50 value determined for limonene is 1413 µM, and the curve obtained is shown in figure 7.
Our enzyme activity data are in good agreement with those reported in the literature for primary human hepatocytes and primary-like human hepatocytes [20,21] when incubated with 100 µM

Conclusion
Polymorphic variants and the chance for multiple binding regions within the CYP2C9 active site lead to the potential for inhibitory effects as well as targetdrug metabolism by a co-administered drug [22]. The need for multiple medications is common in the older population and is often associated with adverse drug reactions that can lead to increased lengths of stay in hospital and increased mortality [23]. Polypharmacy may be clinically appropriate, but in order to decrease the risk of adverse events, lack of drug efficiency, and potentially death, the effects of substances originating from diet and environment on drug metabolism need to be investigated. The identification of unique markers resulting from the CYP metabolism of an ingested substrate provide the underpinning knowledge needed to develop phenotypic tests based on blood or breath analyses to assess an individual's drug response.
In the development of a cell-based workflow system to characterize CYP activity, various factors need to be considered. These include the solubility of the substrate and the associated difficulties of the predominantly hydrophobic CYP2C9 substrates and their application to cell culture experiments. Since diclofenac is a water-insoluble drug, the use of the pure substance in cell culture is challenging, because the amount of organic solvent that can be used in cell experiments is limited. However, diclofenac is commercially available in a salt (sodium or potassium) form, which facilitates drug uptake. Although bioconversion of diclofenac is based on hydroxylation, and our further studies will be mainly focused on demethylation reactions, we have successfully shown in this proof-of-concept investigation that we can track the bioconversion of a substrate and the competitive metabolism with another compound present in a human cellular model.
Based on the successful conversion of diclofenac using the recombinant HEK293T-CYP2C9 cells, a workflow using a cell-based system for CYP2C9 activity has thus been established. Using this workflow, we have shown that the results obtained for the enzyme activity in this work is comparable to other published work for primary human hepatocytes and primarylike human hepatocytes. Our next step is to implement the workflow developed in this proof-of-concept study to investigate other substrate bioconversions involving not only CYP2C9, but also other CYP isoforms. The choice of substrates will be determined using drug databases, such as DrugBank [24], with a strong focus being placed on those precursors that yield volatile metabolites normally not present in human breath. We are aware that the road to implement a breath test using non-labelled precursors for CYP activity determination will be long and challenging, however the foundation for this road has been laid with the workflow presented in this paper.

Data availability statement
The data cannot be made publicly available upon publication because the cost of preparing, depositing and hosting the data would be prohibitive within the terms of this research project. The data that support the findings of this study are available upon reasonable request from the authors.