Materials.

(−)-Cotinine, (±)-cotinine-D₃, trans-cinnamaldehyde, 2-methoxycinnamaldehyde, and reduced glutathione were obtained from Sigma-Aldrich. S-(−)-nicotine and NADPH tetrasodium salt were purchased from Alfa Aesar and EMD Chemicals, respectively. Pooled (N = 50 livers; 28 male and 22 female) human liver microsomes (HLMs) were obtained from Gibco. Human liver cytosol and NADPH solution A and solution B were obtained from Corning Life Sciences. High performance liquid chromatography columns were obtained from Phenomenex and Waters, as indicated below.

Expression and Purification of CYP2A6 and Rat Cytochrome P450 Reductase.

CYP2A6 was expressed and purified as previously reported, with some slight modifications (Stephens et al., 2012; Chan et al., 2016). Spectral ligand binding assays, metabolite identification, apoprotein, and heme analysis experiments were conducted using modified recombinant human CYP2A6, which had an N-terminal transmembrane sequence truncation (∆2–23) and a C-terminal His₄-tag. The recombinant protein was expressed from pKK2A6dH in DH5α Escherichia coli with a 72-hour induction time in δ-amino levulinic acid–supplemented Terrific broth media. The protein was purified using Ni affinity chromatography, followed by carboxymethyl Sepharose cation exchange chromatography (Stephens et al., 2012). The detergent included in purification buffers was Brij L23. P450 content was determined from the carbon monoxide difference spectrum (Omura and Sato, 1962, 1964; Guengerich et al., 2009). Rat P450 reductase was expressed and purified as previously reported (Shen et al., 1989).

Molecular Dynamic Studies of CYP2A6-CA and CYP2A6-MCA Interactions.

The three-dimensional structures of the ligand molecules CA and MCA were built, minimized, and geometry optimized in Molecular Operating Environment (MOE) (2019) from Chemical Computing Group (https://www.chemcomp.com/Products.htm). Charge calculation and atom types were assigned by the CHARMM General Force Field small-molecule parameterization module.

Molecular docking and dynamics studies were conducted using the published X-ray crystal structure of coumarin-bound CYP2A6 (PDB: 1Z10) (Yano et al., 2005); resolution 1.9 Å). Missing transmembrane helix residues 1–29 were obtained as the FASTA sequence from UniProt (P11509-1) and constructed through homology modeling with MODELLER 9.19 (Šali and Blundell, 1993; Webb and Sali, 2016) using the intact X-ray crystal structure of the transmembrane domain of CYP51 [PDB: 5EQB (Monk et al., 2014)]. Of 100 models successfully generated by the software, the top-scoring model with the most favorable discrete optimized protein energy score was chosen for simulation. The fully constructed protein was prepared using the “Structure Preparation” module in MOE (missing atoms, residues, and H atoms were added; protonation states of residues were assigned using Protonate 3D within MOE) and then inserted and oriented within the 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine membrane using the Orientations of Proteins in Membranes web server (Lomize et al., 2012). The reoriented structure was measured to have a heme tilt angle and a transmembrane helix angle of 67.495° and 35.876° versus the bilayer normal (z-axis), respectively.

As guidance for docking CA and MCA to CYP2A6, 28 known CYP2A6 substrates or inhibitors with reported K_i values and binding modes were docked to identify the optimal combination of parameter settings within MOE for placement and refinement scoring (i.e., by reproducing the greatest number of crystal poses) for an induced-fit docking process. This pool included reversible inhibitors coumarin (PDB: 1Z10) and pilocarpine (PDB: 3T3Q), as well as irreversible inhibitors methoxsalen (PDB: 1Z11), aldrithiol (4,4′-dipyridyl disulfide, PDB: 2FDY), N,N-dimethyl(5-(pyridin-3-yl)furan-2-yl)methanamine (PDB: 2FDU), 5-(pyridin-3-yl)furan-2-yl)methanamine (PDB: 2FDW), and N-methyl(5-(pyridin-3-yl)furan-2-yl)methanamine (PDB: 2FDV). MMFF94X optimized force field for small molecules was chosen for docking within MOE. Usage of the London dG scoring function, the Triangle Matcher placement method, and Generalized-Born Volume Integral/Weighted Surface Area dG refinement scoring methods was found to reproduce the best poses for the studied molecules, as determined based on overlay root-mean-square deviation (RMSD) and adoption of the expected binding modes. Both CA and MCA were docked using the above set of parameters, and docked structures were used as the starting points for running all-atom MD simulations. In addition, a structure-based pharmacophore model (Fig. 2) was generated using the superpositioned crystal structures to obtain essential functional features. This model was used for selecting the best poses in addition to the docking scores.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Structure-based pharmacophore model generated from the superposition of published X-ray crystal structures of CYP2A6 bound to both reversible and irreversible inhibitors. The majority of the ligands had an H-bond acceptor (part of an aromatic ring) that made H-bond interaction with the amido -NH group of the binding site residue N297. Several phenylalanine residues, including F107, F118, F209, and F480, form an aromatic cage that makes π stacking interactions with the aromatic ring of the ligands. The distal end of ligands is in close proximity to the heme iron atom, coordinating through a lone pair of electrons carried by an electronegative atom such as N.

All MD simulations were performed using the NAMD package (Phillips et al., 2005) under isothermal/isobaric NPT (number of particles; pressure constant; temperature constant) ensemble with periodic boundary conditions. The CHARMM36 force field (Huang and MacKerell, 2013), CHARMM General Force Field (Vanommeslaeghe et al., 2010, 2012; Vanommeslaeghe and MacKerell, 2012), and transferable intermolecular potential with 3 points (TIP3P) water model (Jorgensen et al., 1983) were employed for the protein and membrane, including monovalent ions, ligands, and solvent molecules, respectively. The CHARMM-GUI (Lee et al., 2016) membrane builder plug-in was used to construct the model membrane, which consisted of 90 and 105 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine lipids and 10 cholesterol molecules in the upper and lower leaflets, respectively. The CYP2A6-ligand complex oriented bilayer (Supplemental Fig. 1) was solvated with a minimum water thickness of 22.5 Å on each side.

The net charge of the membrane-CYP2A6-ligand complex systems was kept neutral (with 0.15 M KCl, i.e., 52 K⁺ and 60 Cl⁻ ions) and subjected to a multistage minimization and equilibration protocol. In the first stage, the solvent molecules and ions were briefly minimized (2500 steps of steepest descent minimization followed by another 2500 steps of conjugate gradient method) to remove any bad contacts with the complexes. The rest of the system was position-restrained using a force constant of 100 kcal/(mol × Å²). This step was followed by another 5000 steps of minimization of the whole solvated complex. The readjustment of membrane and solvent molecules to the potential field of the protein-ligand complex was achieved by subsequent heating and equilibration steps. The system was heated to a final temperature of 310 K in three steps; Langevin dynamics (Feller et al., 1995) was used in all stages to control the temperature using a collision frequency of 1.0 picoseconds⁻¹. The final solvent equilibration step was performed for 50 picoseconds under NPT conditions as above. The SHAKE algorithm (Ryckaert et al., 1977) was used to constrain bonds involving hydrogen, allowing a time step of 2 femtoseconds. Lennard-Jones and electrostatic interactions were calculated explicitly within a cutoff of 1.2 nm, and long-range electrostatic interactions were calculated by particle mesh Ewald summation (Ulrich et al., 1995). For both CA and MCA, the production simulation was carried out for 150 nanoseconds, and the trajectory files were written for every 5000 steps, resulting in 15,000 frames for each molecule.

The resulting trajectories were analyzed for potential H-bond interactions as well as the distance between ligand atoms and the heme iron atom using Hbonds plug-in Visual Molecular Dynamics software with the following criteria for the H-bonds: the distance between the donor atom (D) and the acceptor atom (A) is less than the cutoff distance of 3.5 Å, and the angle D-H-A is less than the cutoff angle of 40°.

Spectral Binding Studies with MCA and Purified CYP2A6.

Samples of 1 µM purified rCYP2A6 in 50 mM potassium phosphate buffer (pH = 7.4) were titrated with aliquots of MCA dissolved in DMSO. The final DMSO concentration in the titrations was kept below 1%. During the titration, solvent-only additions of equal volume (that is, equal to the volume added to the sample cuvette containing MCA) were added to a reference cuvette containing an identical protein sample. Difference spectra were recorded on a Shimadzu UV-visible 1601 spectrophotometer, and the ∆absorbance between the minimum and maximum was measured. Equilibrium dissociation constants (K_S) were determined from nonlinear least-squares fits to two equations: one-site specific ligand binding and specific binding with Hill slope using GraphPad Prism 8 (La Jolla, CA). Fits were compared using Akaike’s Information Criterion (AICc). The experiment was repeated on three different days (N = 3), and each individual experiment was analyzed independently for comparison of model fits, as well as K_S and Hill coefficient determination.

CYP2A6-Mediated Metabolism of 2-Methoxycinnamaldehyde and Metabolite Identification by LC-MS.

Purified rCYP2A6 (1 µM) was reconstituted with rat cytochrome P450 reductase (1 µM) and incubation buffer (100 mM potassium phosphate buffer; pH = 7.4) in microcentrifuge tubes. The incubations were preheated at 37°C for 5 minutes after the addition of MCA (200 µM) and reduced glutathione (5 mM) to trap epoxide metabolites that could form. A 100 mM solution of MCA (in DMSO) was diluted in incubation buffer to make a 2 mM working solution, which was further diluted to the final concentration used in the incubations. The working stock of reduced glutathione was made in incubation buffer. Incubations (100 µl final volume) were initiated with NADPH (1 mM), heated at 37°C for 1 hour in a water bath, and terminated with an equal volume of acetonitrile:methanol (1:1) and centrifuged. The supernatants were removed for LC-MS analysis, and those samples were analyzed using an AB Sciex 5600 TripleTOFTM system and mass defect filtering. Details of the method and data analysis are described in Supplemental Material 4.

Time-Dependent Inhibition of Nicotine Metabolism by CA and MCA in Human Liver Microsomes and Cytosol and LC-MS/MS Analysis of Nicotine and Cotinine.

Time-dependent inhibition of CYP2A6 was tested using a standard two-step approach using pooled (N = 50) human liver microsomes (Yadav et al., 2018) with slight modifications. Briefly, 12 concentrations of either CA (0–1000 μM) or MCA (0–250 μM) were incubated at 37°C with a 5 mg/ml suspension of HLM in 100 mM potassium phosphate buffer (pH = 7.4) as a primary incubation. After 7 minutes of warming the primary incubation, the reaction was initiated by addition of NADPH regenerating system (final concentration 1.3 mM NADP⁺, 3.3 mM glucose-6-phosphate, 0.4 U/ml glucose-6-phosphate dehydrogenase, and 3.3 mM magnesium chloride). The NADPH regenerating system was used rather than NADPH salt to avoid the inhibition of P450 reductase at later time points by NADP⁺ accumulation. At specific time points, an aliquot (10 μl) of the primary incubation was added to 190 μl of secondary incubation. The secondary incubation consisted of nicotine (75 μM), NADPH (1 mM), and human liver cytosol (1 mg/ml). The conditions for the secondary incubation were developed based on previous studies of nicotine metabolism (Nakajima et al., 1996; Messina et al., 1997; Yamazaki et al., 1999; Le Gal et al., 2003). Since nicotine is metabolized by cytochrome P450 to nicotine-Δ^1′(5′) -iminium ion, which is further metabolized to cotinine by aldehyde oxidase (Murphy, 1973; Brandänge and Lindblom, 1979; Hukkanen et al., 2005);, cytosol was added to the secondary incubation to facilitate cotinine formation. The primary incubation was run for 0–90 minutes, and the secondary incubation was allowed to run for 10 minutes, followed by quenching with ice-cold acidified acetonitrile containing cotinine-D₃ as the internal standard. Each assay was conducted in triplicate. After centrifugation at 2000 rpm for 10 minutes, the supernatant was analyzed for cotinine by LC-MS/MS.

LC-MS/MS Analysis of Cotinine.

Calibration curves were prepared in 0.25 mg/ml HLM in phosphate buffer (pH = 7.4) spiked with analyte standards, followed by precipitation with acetonitrile containing the internal standard cotinine-D₃. The supernatant was analyzed with LC-MS/MS. For chromatographic separation of cotinine, a Phenomenex Synergi 4-µm Polar-RP 80-Å column (50 × 2 mm) was used. The sample volume was 10 µl. An AB Sciex API 5500 LC-MS/MS system was used to analyze samples in positive ion mode using the following multiple reaction monitoring transitions: 176.979–80.016 m/z for cotinine and 179.985–80.022 m/z for cotinine-D₃. LC-MS solvents consisted of 10 mM ammonium formate in water (pH = 4.0) as aqueous mobile phase (A) and 0.1% formic acid in acetonitrile as an organic mobile phase (B). The flow rate was 1 ml/min, with a gradient elution programmed from 5% to 95% B in 0.20 minutes maintained at 90% until 0.4 minutes and returned to initial condition at 0.8 minutes and maintained until 1.1 minutes. The total run time was 1.1 minutes, and the retention time was 0.61 minutes.

Microsomal Partitioning of CA and MCA.

Nonspecific microsomal partitioning and human plasma protein binding were estimated using the model developed by Gao et al. (2008). The model was developed using molecular descriptors including calculated log P, calculated log D, pK_a, Kier connectivity, shape, and E-state indices, a subset of MOE descriptors, and a set of absorption, distribution, metabolism, and excretion (ADME) features. The statistical model was derived using the Cubist committee plus composite models, which is a tool for generating rule-based predictive models (Gao et al., 2008).

Model Fitting Using the Numerical Method and Estimate of TDI Parameters for CA and MCA.

Cotinine concentrations obtained from the in vitro TDI experiments were converted to log (PRA) plots and further evaluated for model development. The concave upward curvature is indicative of either quasi-irreversible or partial inactivation, as shown previously (Nagar et al., 2014). Using the numerical method (Korzekwa et al., 2014; Nagar et al., 2014), kinetic models were fit to the data, and kinetic parameters were estimated. The initial estimates of the rate constants were obtained from analyzing the data as detailed in previous publications (Korzekwa et al., 2014; Nagar et al., 2014; Barnaba et al., 2016; Yadav et al., 2018, 2020). Association rate constants were fixed at 270 μM⁻¹ min⁻¹ (Barnaba et al., 2016). For nicotine, the association (k₁) and dissociation rate constants (k₂) were fixed at 270 μM⁻¹ min⁻¹ and 4050 min⁻¹, respectively, based on the reported K_M of nicotine (Yamazaki et al., 1999; Le Gal et al., 2003). Lipid partitioning was also incorporated in the models to account for microsomal partitioning (Yadav et al., 2019). The association rate constant for lipid was set at 2000 μM min⁻¹, and the dissociation rate constant was calculated using the following equation (Nagar and Korzekwa, 2017):where k_on is the association rate constant, k_off is the dissociation rate constant, and f_u,mic is the unbound fraction in the microsomes.

As previously mentioned, curvature in the PRA plots indicates either MIC formation or partial inactivation (Nagar et al., 2014). Curved PRA plots are also observed in cases of inhibitor depletion (Yadav et al., 2020). Several models were developed to capture the inactivation kinetics of CA. Further, a competitive inhibition model was fit to 0- and 60-minute time point data to obtain an initial estimate for K_I. A difference in initial estimates of K_I from 0 versus 60 minutes was indicative of multiple binding. As shown previously (Barnaba et al., 2016), MIC formation is a complex multistep process involving the formation of Fe⁺³:carbene and Fe²⁺:carbene. Hence, enzyme inactivation was modeled with three types of rate constants: e.g., in Fig. 8A, k₆ and k₁₁ for Fe⁺³:carbene formation, k₇ for the reformation of active enzyme, and k₈ for Fe²⁺:carbene formation. The second type of quasi-irreversible model was also evaluated, which involved two steps denoted as k₆ and k₁₁ for Fe⁺²:carbene formation and k₇ for the reformation of active enzyme (Rodgers et al., 2018). Partial inactivation (PI) models were also tested, including partial inactivation with inhibitor depletion, double binding with partial inactivation, and the sequential formation of an intermediate leading to partial inactivation. The kinetic scheme for the simplest partial inactivation model is shown in Fig. 8C (Nagar et al., 2014).

Assuming rapid equilibrium, K_I values were estimated from ratios of association and dissociation rate constants. K_I obtained from the numerical method is the same as unbound K_I,u (K_I,u = K_I) since lipid partitioning was incorporated in the model. Net k_inact was calculated using the net rate constant concept (Cleland, 1975). For MCA, net k_inact was calculated by using the following equation, which was derived using the net rate constant concept for the EII-PI-M-IL model:whereWhereas for CA, k_inact was calculated by using the following equation for EII-MIC-M-IL:where

The replot method was also used to analyze the in vitro TDI data sets. The linear portion of the PRA plots was used to calculate the slopes (k_obs). The estimates of K_I and k_inact were obtained using the following replot equation:Although post hoc correction for lipid partitioning can lead to errors (Yadav et al., 2019), for comparison of replot and numerical method, unbound K_I was obtained by post hoc multiplying the K_I obtained from the above equation.

Parameter errors for net rate constants were calculated with error propagation for individual rate constants. AICc (Akaike, 1974) and adjusted R² were used to compare different models for each data set. Model fitting was conducted with Mathematica 11.3.0.0 (Wolfram Research, Champagne, IL) using the NonlinearModelFit function to fit the model to the data with PrecisionGoal = 10, finite difference derivatives with an order of 4, and 1/Y weighting.

Herb-Drug Interaction Prediction Using In Vitro–In Vivo Extrapolation.

HDI predictions using K_I,u and k_inact obtained from the numerical method were performed using the following mechanistic static equations (https://www.fda.gov/media/134582/download). HDI predictions were performed assuming no substrate-dependent interactions by using nicotine and letrozole as CYP2A6 substrates. The estimated exposure of CA is between 8 and 275 mg (Friedman et al., 2000; Kirkham et al., 2009; Chan et al., 2016). Hence, a 275-mg dose was used to predict all the CA and MCA interactions with a dosing interval of 24 hours. For HDI predictions with MCA, K_Iu2 and the net k_inact calculated (as shown above) were used. CYP2A6 is not known to be present in gut; therefore, only the systemic interaction was predicted for oral dosing of nicotine and letrozole. The following equations were used to predict AUC changes for the probe substrate dosed through the oral route:where AUC_i and AUC are the areas under the plasma concentration–time curve of the probe substrate in the presence and absence of the inactivator, respectively; f_m,CYP2A6 is the fraction of nicotine metabolized by CYP2A6;[I]_h is the inactivator concentration in the hepatic portal vein (defined below); and k_deg,h is the degradation rate constant of hepatic CYP2A6 (Yang et al., 2008). K_iu is the unbound competitive inhibition constant. K_iu was obtained from the literature by assuming competitive inhibition (Chan et al., 2016). I_h,u was given by the following equations (Rostami-Hodjegan and Tucker, 2004; Obach et al., 2007):where I_max is the maximum concentration of the inactivator in plasma after oral dosing, D is the dose, F_a is the fraction absorbed, F_g is the fraction escaping gut metabolism, k_a is the first-order absorption rate constant, Q_h (1.5 l/h) is the hepatic blood flow, f_u,p is the unbound fraction in the plasma, f_u,g (assumed to be equal to 1) is the unbound fraction in the enterocyte, and BP is the blood-to-plasma partition ratio. Maximum unbound plasma concentration I_max,u was given by the following equationswhere τ is the dosing interval, and k is elimination rate constant, which is calculated by the following equation:where CL_s is the systemic clearance, and V_ss is the volume of distribution at steady state.

Average unbound systemic concentration I_sys,u is the unbound average systemic inactivator concentration:where F is the bioavailability.

Human pharmacokinetic parameters for CA were predicted from rats to humans. Rat clearance and volume of distribution were obtained from the literature (Yuan et al., 1992). Human clearance was predicted using the following equation (Tang et al., 2007):

Volume from rats was scaled to human by using the following method (Caldwell et al., 2004):

For MCA, in vitro CL_int was predicted using an internal in silico Cubist-based model (Wenzel et al., 2019). The predicted CL_int was then scaled using biologic scaling factors [microsomal protein per gram of liver = 45; liver weight = 1800 g (Davies and Morris, 1993) for a 70-kg human] assuming no extra hepatic clearance. In vivo CL for MCA was estimated using the well stirred model (Pang and Rowland, 1977) and the predicted in vitro CL_int. Volume of distribution for MCA was predicted using the Rodgers model (Rodgers and Rowland, 2006) using log P of 2.35, f_u,p of 0.261, and pk_a of 14.

Inactivator pharmacokinetic parameters used for calculation of in vivo plasma inactivator concentrations are listed in Table 1.

LC-MS Search for Heme and Protoporphryin IX Adducts with MCA and Metabolites of MCA.

Purified rCYP2A6 (1 µM) was reconstituted with rat cytochrome P450 reductase (1 µM) and incubation buffer (100 mM potassium phosphate buffer; pH 7.4) in microcentrifuge tubes. The P450:reductase ratio for heme adduct analysis and heme degradation was based on the range of ratios used in previous studies (Raner et al., 1997; Kuo et al., 1999; Lin et al., 2005; Foti et al., 2011). The incubations were preheated at 37°C for 5 minutes after the addition of 200 µM MCA, which was estimated to be a saturating concentration based on preliminary TDI experiments. Stock and working solutions of MCA were prepared as above for metabolite identification studies. Incubations were initiated with NADPH (1 mM final concentration) and heated at 37°C for 1 hour in the dark. The incubations were terminated with an equal volume of acetonitrile:methanol (1:1) with 1% formic acid, vortexed, and centrifuged. The supernatants were removed for LC-MS analysis, and those samples were analyzed using an AB Sciex 5600 TripleTOFTM system. Details of the method and analysis are described in Supplemental Material 5.

LC-MS/MS Analysis of Heme Degradation.

Incubations were performed in microcentrifuge tubes in a 37°C water bath. Purified rCYP2A6 (400 pmol) and rat P450 reductase (400 pmol) were combined and left at room temperature for 10 minutes. Buffer (100 mM potassium phosphate; pH = 7.4) and MCA (80 µM) were added and preincubated at 37°C for 5 minutes. The MCA concentration was selected for the purposes of comparison with CA from a previous study (Chan et al., 2016); also, it was estimated that 80 µM approached enzyme saturation based on preliminary TDI experiments. NADPH (1 mM) was added to initiate the incubations, and the samples were heated at 37°C for 5 minutes. Incubations were terminated with an equal volume of acetonitrile, mixed with a vortex mixer, and centrifuged at 11,000g for 5 minutes. Enzyme-only controls (without MCA or NADPH) and controls without NADPH were prepared similarly. The supernatant was removed and was injected (50 µl) onto a Phenomenex Kinetix C18 column (50 × 2.1 mm; 2.6 μm) using an AB Sciex Exion high performance liquid chromatograpy system interfaced with a 3500 triple quadrupole mass spectrometer. Heme was eluted using a gradient of 30% B to 95% B over 6 minutes (A = water/0.1% formic acid; B = acetonitrile/0.1% formic acid; 0.6 ml/min). Heme was detected by UV absorption at 405 nm and by ion intensity using multiple reaction monitoring at m/z = 616/557.2 (parent mass/transition state) in positive mode. Instrument parameters were as follows: ionspray voltage, 5500; declustering potential, 190 V; entrance potential, 10 V; collision energy, 54 eV; collision cell exit potential, 6 V; curtain gas, 30 psi; collision gas, 9 psi; ion source gas 1 and 2, 60 psi; temperature, 550°C. Nitrogen was used as the nebulizer, heater, and auxiliary gas. Heme loss was determined by comparing the change in peak area relative to the enzyme-only or enzyme and NADPH control areas, which were set to 100% of heme content. Statistical significance between individual data sets was determined using a two-tailed t test.

LC-MS/MS Analysis of the CYP2A6 Apoprotein.

rCYP2A6 (60 pmol) and reductase (600 pmol) were combined and preincubated at room temperature for 10 minutes. The P450:reductase ratio was based on previous LC-MS/MS analyses of P450 apoprotein adducts with time-dependent inhibitors (Lin et al., 2005; Baer et al., 2007; Harrelson et al., 2012). Preliminary experiments showed that the detection and deconvolution of the protein envelope for rCYP2A6 was substantially better than reductase, so the relative amount of reductase used here was higher than the previous reports to improve the detection of reductase. Buffer (100 mM potassium phosphate, pH = 7.4) and MCA or CA (200 µM) were added and preincubated at 37°C for 5 minutes. NADPH (1 mM) was added, and the samples were heated at 37°C for 20 minutes. A positive control containing 80 µM 8-methoxypsoralen (8-MOP) was prepared in a similar fashion. All samples were compared with negative controls containing no inhibitor or no NADPH. A portion (10 µl) of the incubation was analyzed on the aforementioned AB Sciex 3500 triple quadrupole LC-MS/MS system using a POROS 10 R1 column (2.1 mm internal diameter × 100 mm; Thermoscientific). Protein was eluted using a gradient of 35% B to 55% B over 4 minutes (A = water/0.1% formic acid; B = acetonitrile/0.1% formic acid; 1.2 ml/min). Protein envelopes were detected using a Q1 scan in positive mode from 1000 to 2000 m/z. Instrument parameters were as follows: ionspray voltage, 4500; declustering potential, 170 V; entrance potential, 3 V; curtain gas, 30 psi; ion source gas 1 and 2, 60 psi; temperature, 25°C. Nitrogen was used as the nebulizer, heater, and auxiliary gas. After the spectra were obtained, deconvolution of the ion envelopes was undertaken by averaging the peak area corresponding to the retention time of CYP2A6 using the Biotool Kit plug-in in the AB Sciex Peakview software suite.