CYP2C9, CYP3A and CYP2C19 metabolize Δ9-tetrahydrocannabinol to multiple metabolites but metabolism is affected by human liver fatty acid binding protein (FABP1)

Abstract

Δ9-tetrahydrocannabinol (THC) is the psychoactive constituent of cannabis. It is cleared predominantly via metabolism. Metabolism to 11-OH-THC by cytochrome P450 (CYP) 2C9 has been proposed as the main clearance pathway of THC, with the estimated fraction metabolized (f_m) about 70%. The remaining clearance pathways are not well established, and it is unknown how THC is eliminated in individuals with reduced CYP2C9 activity. The goal of this study was to systematically identify the CYP enzymes contributing to THC clearance and characterize the metabolites formed. Further, this study aimed to characterize the impact of liver fatty acid binding protein (FABP1) on THC metabolism by human CYPs. THC was metabolized to at least four different metabolites including 11-OH-THC in human liver microsomes (HLMs) and with recombinant CYPs. 11-OH-THC was formed by recombinant CYP2C9 (K_m,u = 0.77 nM, k_cat = 12 min⁻¹ and by recombinant CYP2C19 (K_m,u = 2.2 nM, k_cat = 14 min⁻¹). The other three major metabolites were likely hydroxylations in the cyclohexenyl ring and were formed mainly by recombinant CYP3A4/5 (K_m,u > 10 nM). HLM experiments confirmed the contributions of CYP2C9, CYP2C19 and CYP3A to THC metabolism. The presence of FABP1 and THC binding to FABP1 altered THC metabolism by recombinant CYPs and HLMs in an enzyme and metabolite specific manner. This suggests that FABP1 may interact with CYP enzymes and alter the f_m by CYPs towards THC metabolism. In conclusion, this study is the first to systematically establish the metabolic profile of THC by human CYPs and characterize how FABP1 binding alters CYP mediated THC metabolism.

1. Introduction

Cannabis use is at an all-time high in the U.S. with more than 48 million people reportedly having consumed cannabis at least once in 2019 [1]. In 2021, 43% of young adults (19–30 years) reported using cannabis and 11% reported daily use [2]. Δ−9-tetrahydrocannabinol (THC) is the major psychoactive component found in cannabis. It is also the main circulating bioactive cannabinoid found in humans after smoking cannabis. The primary circulating metabolite of THC found in humans is 11-hydroxy-THC (11-OH-THC), but several dozen different metabolites of THC have been identified [3–5]. Following cannabis consumption via edibles the relative exposure to 11-OH-THC is greater than that observed after smoking, largely due to the high hepatic clearance of THC and first pass formation of 11-OH-THC [6,7]. 11-OH-THC is pharmacologically active [8,9] and has been shown to be formed by cytochrome P450 2C9 (CYP2C9) in vitro [3,10–12]. Other CYP enzymes do, however, also form 11-OH-THC [3,11]. 11-OH-THC is further metabolized by CYP mediated oxidation and by alcohol and aldehyde dehydrogenase enzymes to 11-COOH-THC [11,13,14], the main circulating metabolite of THC.

THC pharmacokinetics in humans is altered in individuals with CYP2C9 polymorphisms [15]. Carriers of CYP2C9*3/*3 retain about 7% of the CYP2C9 activity when compared to CYP2C9*1/*1 carriers [16]. After oral dosing, THC exposure (AUC) is 3-fold greater in CYP2C9*3/*3 carriers compared to CYP2C9*1/*1 carriers [15]. Based on these data, CYP2C9 contributes to ~ 70% of THC clearance (f_m = 0.7). At present, it is unclear what other enzymes are responsible for the remaining ~ 30% of THC clearance in CYP2C9*1/*1 individuals. The main elimination pathways of THC in individuals who are carriers of CYP2C9 polymorphisms or in individuals who are taking CYP2C9 inhibitors are also unknown. The AUC of THC was increased by ~ 80% when co-administered with ketoconazole [17], an inhibitor of CYP3A4, suggesting that CYP3A4 may contribute to THC clearance. In vitro, 8β-OH-THC and 9α,10α-THC-epoxide were identified as THC metabolites formed in human liver microsomes (HLMs) when very high concentrations of THC were used [3,18,19]. CYP3A4 appeared to be responsible for the formation 8β-OH-THC and 9α,10α-THC-epoxide [3]. 8β-OH-THC and the di-hydroxylated metabolite 8β,11-diOH-THC were also detected in plasma from cannabis users [20] supporting a role of CYP3A4 in THC metabolism.

The relative importance of the enzymes and metabolites identified in the previous studies may be confounded by the concentrations of THC used. For example, CYP3A4 contribution was detected at 64 and 130 μM THC [3,19], concentrations that greatly exceed plasma THC maximum concentration (C_max) even in heavy cannabis users [21]. For 11-OH-THC formation the apparent K_ms in HLMs and with recombinant CYP2C9 were measured as 0.8–5.2 μM [12]. The unbound K_m (K_m,u) for 11-OH-formation in HLM was estimated as 8 nM and for THC depletion as 7 nM [13]. The observed circulating concentrations of THC in plasma vary widely depending on route of administration [15,22] and cannabis usage patterns [21]. If several enzymes with different K_m values contribute to THC clearance, the relative contributions of these enzymes to THC clearance and metabolite formation in cannabis users may vary depending on circulating THC concentrations and usage. Quantitative determination of the kinetics (K_m) of THC metabolism by individual CYPs is needed to accurately assess the contribution of CYPs in the context of the wide range of THC C_max values observed in vivo.

A challenge in studying THC metabolism in vitro is the high lipophilicity of THC (logP = 6.97) [23] and the “stickiness” of THC that results in extensive nonspecific binding to plastics and other labware. To address these issues, bovine serum albumin (BSA) was previously added to in vitro incubations to prevent nonspecific binding of THC to plastics [11]. Using this approach and via measurement of THC depletion at 500 nM nominal concentration, the majority (91%) of THC metabolic clearance in pooled HLMs was assigned to CYP2C9 with the remaining 9% assigned to CYP2D6 (f_m = 0.09) [11,13]. At 500 nM THC, recombinant CYP1A1, CYP1A2, CYP2C19 and CYP3A4 were also found to efficiently deplete THC, but based on inhibition studies these enzymes were not considered to be important for THC clearance in human liver [11]. At present, it is unclear how the incorporation of 0.2% BSA (which is not present in humans) into incubations altered the relative contribution of specific CYPs to THC clearance. Albumin has previously been shown to have enzyme specific effects on CYP mediated metabolism [24], and to increase the activity of CYP2C9 in HLMs [25,26]. Whether albumin has a similar effect on CYP3A4, CYP2C19, CYP2D6 or CYP1A is unclear. In addition, albumin effects are unlikely to be present in human hepatocytes in vivo as albumin is mainly a plasma protein and although expressed in the liver, it is secreted to circulation via vesicle trafficking and not released to liver cytosol. As such, effects of BSA may skew the relative contribution assessment of individual CYPs to THC clearance.

Liver fatty acid binding protein (FABP1) is an intracellular binding protein that binds and solubilizes endogenous lipids including fatty acids [27,28]. FABP1 is highly expressed in the liver cytosol with concentrations ranging from 0.7 mM to 1 mM [27,29]. FABP1 modulates the liver distribution and metabolism of fatty acids [27,28]. Binding to FABP1 was recently shown to also alter the metabolism of diclofenac by CYP2C9 [30]. THC has been shown to bind to FABP1 [31,32], and FABP1 also appears to alter THC disposition in vivo in mice [31]. FABP1 knockout mice had decreased rates of THC metabolism suggesting that FABP1 may be facilitating THC clearance [31]. We hypothesized that in the human liver FABP1 may affect the efficiency of THC metabolism and alter the relative importance of individual CYP enzymes to THC clearance. To test this hypothesis we first identified the major THC metabolites formed in human liver microsomes and determined the contribution of specific CYP enzymes to the formation of THC metabolites. We then characterized the binding of major cannabinoids to human FABP1 and established the effect of FABP1 on THC metabolism by human liver CYPs.

2. Materials and methods

2.1. Chemicals and reagents

Kanamycin, Trizma base (Tris), sodium chloride, sodium phosphate, potassium phosphate, protease inhibitor tablets, benzonase, butanol, thrombin, Rosetta 2 E. coli, Coomassie Brilliant Blue R, 11-(Dansylamino)undecanoic acid (DAUDA), β-Nicotinamide-adenine dinucleotide phosphate (NADPH) tetrasodium salt, tienilic acid, (+)-N-3-benzylnirvanol, CYP3cide, tolbutamide, 4-hydroxytolbutamide, midazolam, 1-hydroxymidazolam, and 1-hydroxymidazolam-d₄ were purchased from Millipore-Sigma (St. Louis, MO). (S)-mephenytoin, (±)4-hydroxymephenytoin, and (±)4-hydroxymephenytoin-d₃, and ethanol solutions (1 mg/mL) of anandamide (AEA) and 2-arachidonyl glycerol (2AG) were purchased from Cayman Chemical Company (Ann Arbor, MI). (−)-Δ⁹-THC (1 mg/ml), (−)-Δ⁹-THC-d₃ (0.1 mg/ml), (±)-11-OH-THC (1 mg/mL), (±)-11-OH-THC-d₃ (0.1 mg/mL), and cannabidiol (CBD, 1 mg/mL) were purchased as methanol solutions from Millipore-Sigma (St. Louis, MO). Tryptone, yeast extract, isopropyl β-d-1-thiogalactopyranoside (IPTG), phenylmethyl sulfonyl fluoride (PMSF), dithiothreitol (DTT), imidazole, sodium dodecyl sulfate (SDS), Pierce bicinchoninic acid (BCA), protein assay, low melt agarose, magnetic silica beads, high-performance liquid chromatography (HPLC) and mass spectrometry (Optima) grade acetonitrile, methanol, water and formic acid were purchased from Thermo Fisher Scientific (Waltham, MA). Lipidex-5000 suspension was purchased from Perkin Elmer Inc (Waltham, MA, USA). Glycine and Mini-PROTEAN TGX protein gels were purchased from Bio-Rad (Hercules, CA).

2.2. Recombinant enzymes and human liver tissue

Recombinant CYP1A2, CYP2A6, CYP2B6, CYP2C8, CYP2C9, CYP2C19, CYP2J2, CYP3A4, CYP3A5, CYP3A7 and CYP4F3A Supersomes co-expressed with cytochrome P450 reductase and cytochrome b5 and CYP2D6 co-expressed with cytochrome P450 reductase were purchased from BD Gentest (Franklin Lakes, NJ). Human liver microsomes (HLMs) pooled from 50 different donors were purchased from Gibco (Cat. HMMCPL, Thermo Fisher Scientific, Waltham, MA). Human liver microsomes from individual donors were obtained from the University of Washington (UW) and St. Jude (SJ) human liver banks. Twenty-two human livers were included from the University of Washington human liver bank and one from SJ human liver bank. For the HLMs from the UW liver bank, donors’ ages ranged from 7 to 64 years. The demographics of these donors comprised 7 females and 15 males, with 20 being Caucasian, 1 Asian, and 1 Black or African American. The demographic information for the HLM from the SJ liver bank was not available. Of all donors, fourteen were CYP2C9 *1/*1, three were CYP2C9 *1/*2, and six were CYP2C9 *1/*3. Seven donors were CYP3A5 expressors (*1/*3 genotype) and sixteen non expressors (*3/*3 genotype). For CYP2C19, five donors were *1/*1, six *1/*17, seven *1/*2, two *17/*17, and three *1/*17 genotype.

2.3. Incubation conditions to identify THC metabolites formed in HLMs and identification of the recombinant cytochrome P450 enzymes that metabolize THC

To determine the metabolite profile of THC in human liver and to systematically characterize the CYP enzymes that metabolize THC, THC was incubated with pooled HLMs and a panel of individual recombinant CYPs (CYP1A2, CYP2A6, CYP2B6, CYP2C8, CYP2C9, CYP2C19, CYP2D6, CYP2J2, CYP3A4, CYP3A5, CYP3A7 and CYP4F3A). For metabolite identification, THC-d₃ was also incubated with HLMs and each of the recombinant CYPs using identical methods as THC. Pooled HLMs (0.1 mg/mL) or recombinant CYP enzymes (10 pmol/mL) were preincubated with 10 μM THC for 5 min at 37 °C in 180 μL incubation buffer (100 mM potassium phosphate, pH 7.4) in 1.7 mL Eppendorf tubes. Reactions were initiated with addition of 20 μL NADPH (1 mM final concentration) and incubated at 37 °C for 15 min. The reactions were quenched with the addition of 600 μL of acetonitrile with 1% formic acid and 10 nM of 11-OH-THC-d₃ as an internal standard. The samples were then centrifuged at 18,000 g for 20 min at 4 °C and 200 μL of the supernatant was transferred to a mass spectrometry vial with glass inserts. Samples were analyzed by LC-MS/MS as described in Section 2.4. The relative formation of metabolites was determined as the ratio of the metabolite peak area to the peak area of the 11-OH-THC-d₃ internal standard. Metabolite peaks below a signal to noise (S/N) of 5 were not quantified. An M4 background peak (S/N ~ 10) was observed in no NADPH and substrate only controls in incubations with HLMs and recombinant CYPs. Therefore, only NADPH dependently formed peaks with S/N > 50 were considered quantifiable for M4 analysis. All metabolite identification incubations were repeated by a second individual on a separate day.

2.4. Quantification of THC metabolites by UHPLC-MS/MS

The concentrations of THC and THC metabolites were analyzed using an AB Sciex API6500 QTRAP mass spectrometer (Framingham, MA) coupled to an Agilent 1290 Infinity II Ultra-High-Performance Liquid Chromatography system (UHPLC, Santa Clara, CA). THC and its metabolites were separated using a Kinetex Evo C18 column (2.1 × 100 mm, 2.6 μM, Phenomenex, Torrance, CA) with the column oven set to 50 °C. Mobile phases A and B were composed of water and acetonitrile, respectively, both containing 0.1% formic acid. A gradient elution at a flow rate of 0.25 mL/min was used as follows: mobile phase A was kept at 60% and B at 40% for the first 5 min, then B was increased to 100 % by 7 min, returned to initial conditions by 8 min, and the column was re-equilibrated to initial conditions for an additional 4 min. THC, THC-d_3, THC metabolites and 11-OH-THC-d₃ were monitored in positive ion mode with electrospray ionization and the MS parameters were set as follows: IS 5500 V, TEM 550 °C, CUR 35 p.s.i., GS1 70, GS2 70, CAD medium, DP 35 and 60 (THC metabolites and THC, respectively), EP 4 and 8 (THC metabolites and THC, respectively), CE 33 and 29 (THC metabolites and THC, respectively), CXP 10 V. Two multiple reaction monitoring (MRM) transitions (331 > 193 m/z and 331 > 201 m/z) were used to monitor THC metabolites as previously described for 11-OH-THC [33]. Each metabolite was quantified using the more sensitive transition of the two for the given metabolite. For 11-OH-THC and for M4 metabolite 331 > 193 m/z was used for quantification. The 331 > 201 m/z transition was used to quantify the M1, M2, M2a and M3 metabolites. The MRM transitions used for 11-OH-THC-d₃, THC and THC-d₃ were 334 > 196 m/z, 315 > 193 m/z and 318 > 196 m/z, respectively [33]. A signal to noise cutoff of 5 was used as the lower limit of quantification and the linearity of detector response for M2, M3 and M4 metabolites was verified via analysis of a dilution series of an incubation sample containing high concentrations of the three metabolites.

To identify the oxidation products of THC formed in HLMs and CYP incubations and to gain structural information about the formed metabolites, enhanced product ion (EPI) scans in positive ion mode were performed. The MS/MS spectra of the [M + H]⁺ ions of the oxidation products, 331 m/z for THC incubations and 334 m/z for THC-d₃ in cubations, were collected within the mass range of 50–340 m/z. EPI scan rate was set at 10,000 Da/s, dynamic fill time was used for the linear ion trap, Q1 was set at unit resolution, and Q3 entry barrier was set at 8 V. LC-MS/MS settings and parameters were identical as described above except for the collision energy (CE), which was varied stepwise between 10 and 50 V in increments of 10 V to explore fragmentation patterns. The CE of 40 V was chosen for detailed spectral interpretation due to the fragmentation efficiency of the parent ions 331 m/z and 334 m/z observed, while still maintaining some detection of the parent ion in the MS/MS spectrum.

2.5. Characterization of CYP2C9, CYP3A and CYP2C19 activity in the human liver microsome panel

The activities of CYP2C9, CYP3A4/5 and CYP2C19 were characterized in the panel of 23 HLMs from individual donors. Tolbutamide (TBU), midazolam (MDZ) and (S)-mephenytoin (MPH) were chosen as probe substrates for CYP2C9, CYP3A4/5 and CYP2C19 activity, respectively. All HLM experiments were performed in duplicate at 0.1 mg/mL microsomal protein concentration in incubation buffer in 96-well plates and repeated on two separate days. After a 5-minute preincubation at 37 °C in the presence of substrate, reactions were initiated with NADPH (1 mM final concentration) and allowed to proceed at 37 °C in a water bath with a total reaction volume of 100 μL. The substrate concentrations used (10 μM for TBU, 1 μM for MDZ, and 6 μM for MPH) were > 80% lower than the K_ms towards the target CYP [34,35]. The incubations were allowed to proceed for 30 min [TBU], 4 min [MDZ], and 30 min [MPH] at 37 °C before quenching with an equal volume of acetonitrile containing 1% formic acid and 10 nM internal standard as listed below. The samples were then centrifuged at 3,600 g for 40 min at 4 °C and 100 μL of the supernatant was transferred to a 96 well MS plate. Samples were analyzed via LC-MS/MS as described below to quantify the formation of 4-OH-TBU from TBU, 1-OH-MDZ from MDZ, and 4-OH-MPH from MPH for CYP2C9, CYP3A and CYP2C19 specific activity, respectively.

The concentrations of 4-OH-MPH and 1-OH-MDZ were analyzed using AB Sciex API5500 and the concentrations of 4-OH-TBU were analyzed using AB Sciex API6500 QTRAP mass spectrometer (Framingham, MA) both coupled to an Agilent 1290 Infinity II Ultra-High-Performance Liquid Chromatography system (UHPLC, Santa Clara, CA). 4-OH-MPH-d₃ was used as the internal standard for 4-OH-TBU and 4-OH-MPH quantification, whereas 1-OH-MDZ-d₄ was used as the internal standard for 1-OH-MDZ quantification. 4-OH-TBU, 1-OH-MDZ and 4-OH-MPH were separated using a Zorbax Eclipse Plus C18 column (2.1 × 50 mm, 5 μM, Agilent Technologies, Santa Clara, CA) with the column oven set to 35 °C. Mobile phases A and B consisted of water and acetonitrile, respectively, both containing 0.1 % formic acid. A gradient elution at a flow rate of 0.4 mL/min was used as follows: mobile phase A was kept at 95% and B at 5% for the first 0.1 min, then B was increased to 70% and 95% by 3.5 and 4.0 min respectively, returned to initial conditions by 5 min, and the column was re-equilibrated in initial conditions for an additional 1.5 min. 4-OH-TBU, 1-OH-MDZ, 1-OHMDZ-d₄, 4-OH-MPH, and 4-OH-MPH-d₃ were monitored in positive ion mode with electrospray ionization and the MS parameters were set as follows: IS 5000 V, TEM 600 °C, CUR 35 p.s.i., GS1 50, GS2 50, CAD medium, DP 21, 90, 80, 80, and 90 (4-OH-TBU, 4-OH-MPH-d₃, 1-OHMDZ, 1-OH-MDZ-d₄, and 4-OH-MPH, respectively), EP 14, 5, 10, 10, and 5 (4-OH-TBU, 4-OH-MPH-d₃, 1-OH-MDZ, 1-OH-MDZ-d₄, and 4-OH-MPH, respectively), CE 24, 27, 37, 35, and 27 (4-OH-TBU, 4-OH-MPH-d₃, 1-OH-MDZ, 1-OH-MDZ-d₄, and 4-OH-MPH, respectively), CXP 15, 4, 14, 14, and 4 V (4-OH-TBU, 4-OH-MPH-d₃, 1-OH-MDZ, 1-OH-MDZ-d₄, and 4-OH-MPH, respectively). Five different MRM transitions (287 > 171 m/z, 238 > 150 m/z, 342 > 203 m/z, 346 > 203 m/z, and 235 > 150 m/z) were used to monitor 4-OH-TBU, 4-OH-MPH-d₃, 1-OH-MDZ, 1-OH-MDZ-d₄, and 4-OH-MPH, respectively. Signal-to-noise (S/N) cutoffs of 3 and 5 were used as the lower limit of detection (LLOD) and lower limit of quantification (LLOQ). For signals that were below LLOQ but above LLOD, LLOQ/2 was used as the quantification value for data presentation and statistical analyses. The linearity of detector response for all the metabolites was verified. All analytes were quantified based on a standard curve ranging from 1.56 nM to 200 nM for 4-OH-TBU and 4-OH-MPH quantification, and from 1.56 to 400 nM for 1-OH-MDZ quantification.

2.6. Characterization of THC metabolism in the panel of human liver microsomes

Formation of THC metabolites was measured in the panel of HLMs from 23 individual liver donors described above. THC (2 μM) was preincubated with 0.1 mg/mL HLM for 5 min at 37 °C in incubation buffer prior to initiating the reactions with 1 mM NADPH (200 μL total reaction volume). Reactions were incubated for 2 min at 37 °C and quenched with 600 μL of acetonitrile containing % formic acid and 10 nM 11-OH-THC-d₃. Quenched reactions were centrifuged and analyzed by LC-MS/MS as described in Section 2.4. A small (S/N ~ 5) peak was observed at M4 retention time in no NADPH and no enzyme incubation controls. The area of this peak was integrated and subtracted from the enzymatic formation of M4 in THC incubation samples. All incubations with HLMs were performed under conditions within time linearity for the formation of all THC metabolites. All incubations were repeated by different individuals on two separate days and done in technical duplicates on each day.

2.7. Correlation of THC metabolism with probe substrates

To evaluate the relative importance of CYP2C9, CYP3A and CYP2C19 in the formation of the individual THC metabolites the correlation between individual THC metabolite formation and the formation of the probe metabolite was determined. Linear regression and statistical analyses were performed using R statistical computing software (version 4.3.1; R Core Team 2023). Package tidyverse was used for data processing, ggplot2, ggeasy, ggpubr, scatterplot3d, and car for preparing figures, and Rmarkdown for text. Correlations with p-values <0.01 were considered statistically significant.

Correlations between the specific activity of each CYP (formation of probe metabolite) and 11-OH-THC formation velocity or M2, M3, or M4 relative formation were initially tested using simple linear regression. By both data visualization and based on spearman correlation, CYP2C9 and CYP3A specific activities correlated with each other (rho = 0.8053 and p-value = 0.000005). This co-linearity confounds interpretation of individual correlations between CYP specific activity and the metabolite formation using simple linear regression. Thus, multiple regressions were performed using the F-test to detect which CYP isoform activity can best explain the variability seen in the THC metabolite formation across all HLM donors. Variation inflation factors (VIF) for CYP2C9, CYP3A, and CYP2C19 were 4.1, 4.1, and 1.2, signifying moderate correlations (VIF < 5) between CYP2C9 and CYP3A and reasonable inclusions of CYP2C9, CYP3A, and CYP2C19 activities in the multiple regression model [36]. Subsequently, CYP2C9, CYP3A, and CYP2C19 activities were adjusted as covariates to determine if any of them were significantly correlated (p-value < 0.01) with 11-OH-THC, M2, M3, or M4 formation. To further establish the importance of CYP enzymes that may play a role in THC metabolism and drive inter-individual variability in THC clearance, stepwise linear regressions were performed. One CYP activity was incorporated as a predictor at a time into the linear regression and analysis of variance (ANOVA) was used to compare the calculated F-statistic to the critical value. A p-value < 0.01 was used to determine whether inclusion of different CYP predictors improved model fit significantly.

2.8. Impact of selective CYP inhibitors on THC metabolism in human liver microsomes from individual donors

Selective inhibitors were used to determine the contribution of CYP2C9, CYP2C19 and CYP3A4 to the formation of the main THC metabolites detected in HLMs. Individual HLMs from donors with known CYP3A5 genotype were included with n = 2 livers with CYP3A5 expression and n = 4 that had no CYP3A5. The time dependent (irreversible) inhibitors used were tienilic acid for CYP2C9 and CYP3cide for CYP3A4 [37,38]. (+)-N-3-benzyl-nirvanol (benzylnirvanol) was used as a selective reversible inhibitor for CYP2C19 [39]. CYP inactivation by irreversible inhibitors was achieved by pre-incubating 1 mg/mL total HLM protein with tienilic acid (50 μM) for 30 min or CYP3cide (0.5 μM) for 10 min in the presence of 1 mM NADPH in incubation buffer. Ethanol for tienilic acid or methanol for CYP3cide were added instead of inhibitors in pre-incubations for no inhibitor (vehicle) controls. After preincubation, the HLM solution was diluted 40-fold (0.025 mg/mL final HLM protein) in incubation buffer containing 1 mM NADPH in a 1.7 mL Eppendorf tube. THC (2 μM) was then added immediately to initiate reactions (final volume of 200 μL). Reactions were allowed to proceed for 2 min prior to quenching with 600 μL of acetonitrile containing 1% formic acid and 10 nM 11-OH-THC-d₃. Inhibition experiments with benzylnirvanol were performed by pre-incubating benzylnirvanol (5 μM) with THC (2 μM) and 0.025 mg/mL total HLM protein in 180 μL incubation buffer in 1.7 mL Eppendorf tubes. Methanol was added instead of benzylnirvanol for no inhibitor (vehicle) controls. Reactions were initiated with 1 mM NADPH (200 μL final volume), allowed to proceed for 2 min and quenched with 600 μL of acetonitrile containing 1% formic acid and 10 nM 11-OH-THC-d_3. Quenched samples were centrifuged and prepared for LC-MS/MS analysis as described in Section 2.4. The % activity remaining was determined by normalizing the relative formation of THC metabolites in the presence of inhibitors to the formation in the absence of inhibitors. Incubations for the six HLMs from different donors were done in duplicate.

2.9. Kinetics of THC metabolite formation in human liver microsomes

The kinetics of 11-OH-THC, M2, M3, and M4 metabolite formation was determined in pooled HLMs with 0.1 mg/mL total microsomal protein using nominal THC concentrations ranging from 0.05 to 5 μM. The M1 metabolite was detected but below the limit of quantification (signal to noise < 5) in these kinetic experiments. Incubations were done under conditions of time linearity in 200 μL volumes of incubation buffer in 1.7 mL Eppendorf tubes. Reactions were preincubated in a 37 °C shaking water bath for 5 min before reactions were initiated with 1 mM NADPH. The reactions were incubated for 2 min at 37 °C before quenching with 600 μL of acetonitrile containing 1% formic acid and 10 nM 11-OH-THC-d₃. Quenched reactions were centrifuged at 18,000 g for 20 min and the resulting supernatant was transferred to MS vials for LC-MS/MS analysis as described in Section 2.4. The incubations were repeated on three separate days by two different individuals and conducted as technical duplicates on each day. Free concentrations of THC were determined for all nominal THC concentrations used as described in Section 2.13. Metabolite formation velocities (absolute velocity for 11-OH-THC formation and relative product formation based on peak area ratio for M2, M3, and M4) were plotted against free THC concentrations to determine the unbound K_m (K_m,u) and k_cat. The Michaelis-Menten equation was fit to the 11-OH-THC, M2 and M3 formation data using GraphPad Prism 10. No saturation was observed for M4 formation and hence only the linear portion of metabolite formation data was analyzed. A linear regression model was fit to M4 formation data in GraphPad Prism 10. K_m,u and k_cat values for 11-OH-THC, M2 and M3 are reported as means ± S.D. from experiments done on three separate days.

2.10. Expression and purification of human recombinant FABP1

Hexa-histidine-tagged (his-tag) human FABP1 was expressed in Rosetta 2 E. coli (Novagen, Madison, WI) and was purified according to a previously published protocol [30]. Briefly, his-tagged FABP1 was purified using a HisTrap HP affinity column (GE Healthcare, Chicago, IL). The his-tag was cleaved by thrombin followed by gel filtration with a Superdex 75 (Marlborough, MA) equilibrated with gel filtration buffer (10 mM potassium phosphate pH 7.4, 150 mM KCl). Delipidation was performed on the cleaved FABP1 using multiple rounds of butanol treatment and Lipidex-5000 (Perkin Elmer Inc., Waltham, MA, USA) as previously described [30]. The concentration of purified FABP1 was quantified via bicinchoninic acid (BCA) assay (Pierce, Waltham, MA). The final purified FABP1 was stored on ice in gel filtration buffer containing 0.5 mM DTT.

2.11. Characterization of cannabinoid binding to human FABP1 via DAUDA displacement assay

Binding of cannabinoids to FABP1 was determined using fluorescence displacement assay with DAUDA according to previously published protocols [30]. In brief, a solution of 0.5 μM DAUDA pre-bound with 0.3 μM FABP1 was prepared in 2 mL of incubation buffer in a 4 mL quartz cuvette. THC, 11-OH-THC, cannabidiol (CBD), 2-archidonyl glycerol (2-AG) or anandamide (AEA) were titrated into the solution of prebound FABP1 and DAUDA and the fluorescence emission throughout the course of the titration was monitored from 400 to 700 nm with an excitation wavelength of 335 nm. Fluorescence spectra were collected on a Cary Eclipse fluorescence spectrophotometer (Agilent, Santa Clara, CA) with the scan rate set to medium (600 nm/min) and the photomultiplier tube voltage set to high (800 V). The specific fluorescence of DAUDA bound with FABP1 from titration spectra was determined using singular value decomposition (SVD) as previously described [30]. EC₅₀ values and binding affinities (K_d) of cannabinoids with FABP1 were determined from the SVD data as previously described using COPASI [30]. EC₅₀ and K_d values are reported as means (±standard deviation) from replicate experiments done on three separate days and with at least two different batches of purified protein.

2.12. THC incubations with recombinant CYP2C9, CYP3A4 and CYP2C19 in the presence and absence of FABP1

The kinetics of 11-OH-THC formation by CYP2C9, M2, M3 and M4 metabolite formation by CYP3A4, and 11-OH-THC and M3 formation by CYP2C19 were determined in the presence and absence of FABP1. All incubations were done under conditions of time linearity. The total protein concentrations in the individual Supersomes were quantified via BCA assay as the expression level of each CYP is different when normalized to the total amount of microsomal protein. For experiments with CYP2C9 in the absence of FABP1, 1 nM CYP2C9 (0.0016 mg total microsomal protein/mL) was preincubated with seven different nominal concentrations of THC ranging from 10 to 640 nM for 5 min at 37 °C in 180 μL incubation buffer in 1.7 mL Eppendorf tubes. Methanol solutions of THC were added to the incubation mixtures and the total solvent volume did not exceed 1 % in any of the incubations. Reactions were then initiated with 1 mM NADPH (final concentration) in a final volume of 200 μL. A control reaction with no NADPH was included in all experiments. Reactions were quenched after 1 min incubation at 37 °C with the addition of three incubation volumes (3:1 ratio of organic to aqueous) of acetonitrile containing 1% formic acid and 10 nM 11-OH-THC-d₃. The samples were centrifuged at 18,000 g for 20 min and the resulting supernatant was transferred to MS vials for analysis as described in Section 2.3. CYP2C19 incubations were performed similarly using 1 nM of CYP2C19 (0.013 mg total microsomal protein/mL) with an incubation time of 2 min and nominal THC concentrations ranging from 0.02 to 1.0 μM. CYP3A4 incubations were also performed as described for CYP2C9 with 0.625 nM CYP3A4 (0.004 mg total microsomal protein/mL) and with seven THC concentrations ranging from 0.156 to 3.75 μM. Incubations with FABP1 were performed similarly as described above except that 20 μM FABP1 was added to the solutions containing CYP and THC before pre-incubation to achieve binding equilibrium. Reactions were then initiated with NADPH. Kinetic experiments with and without FABP1 were performed as matched pairs. All incubations were conducted as duplicates per experiment and all experiments repeated on at least two separate days.

Eadie-Hofstee plots were constructed for all kinetic experiments and based on the Eadie-Hofstee plots the Michaelis-Menten model was fit to the 11-OH-THC formation data by CYP2C9 and CYP2C19, M3 formation data by CYP2C19, and the M2 and M3 formation data by CYP3A4 using GraphPad Prism 10. Based on the concave Eadie-Hofstee plot of M4 formation by CYP3A4, the allosteric sigmoidal model (equation (1) was fit to the data of M4 formation by CYP3A4 in GraphPad Prism 10:

$$\mathit{\text{relative }}M4\mathit{\text{ formation}}=\frac{Ma{x}_{M4}\times {\left[S\right]}^h}{{K_{0.5}}^h+{\left[S\right]}^h}$$ (1)

In equation (1), Max_M4 is the maximum value for the relative formation of M4 in kinetic experiments, [S] is the nominal THC concentration, K_0.5 is the concentration of THC as half of Max_M4, and h is the hill slope. The unbound K_m (K_m,u) and k_cat values were determined by fitting the Michaelis-Menten model to the metabolite formation data plotted with free concentrations of THC determined as described in Section 2.13. K_m,u and k_cat values are reported as means ± S.D. from replicate experiments done on separate days. Differences in the K_m,u and k_cat values for 11-OH-THC and M3 formation by CYP2C19 in the presence and absence of FABP1 were evaluated using Graphpad Prism using the nonlinear regression comparison and extra sum of squares with significant differences assessed by the F-test as previously described [30]. p < 0.05 in all three experiments collectively was considered statistically significant.

2.13. Determination of THC unbound fraction in incubations with recombinant CYPs and in human liver microsomes

To determine the unbound concentrations of THC in incubations with recombinant CYPs and with HLMs, magnetic silica beads (MGSBs, G-Biosciences, St. Louis, MO) were used to separate microsomal protein [40] from free THC in solution. Initial experiments without microsomal protein present were done to verify that THC did not bind in a nonspecific manner to MGSBs. To measure non-specific binding of THC, 500 nM THC was incubated separately with 100 μL of MGSB in 0.5 mL incubation buffer in 1.7 mL Eppendorf tubes for 30 min at 37 °C. After 30 min, 100 μL of the mixture containing MGSBs with THC was collected as the total THC present sample (this was necessary as THC binds extensively to plastic surfaces), then the MGSBs were separated from solution using a DynaMag-2 Magnet (Thermo Fisher Scientific, Waltham, MA) and the supernatant was collected for measurement of free THC in solution. 300 μL of acetonitrile containing 1% formic acid and 30 nM of THC-d₃ internal standard were added to 100 μL of the total and supernatant samples containing THC. The samples were centrifuged at 18,000 g for 20 min and the supernatant was transferred to MS vials for analysis. THC concentrations in the samples were measured via LC-MS/MS using the same method as described in Section 2.4.

To determine the free concentrations of THC in incubations with recombinant CYPs and HLMs, 50 μL of MGSBs (5.25 × 10⁹ beads) [40] were used with recombinant CYPs and 200 μL of MGSBs (21 × 10⁹ beads) were used for HLMs. All of these experiments were conducted using identical Eppendorf tubes, protein concentrations and buffers as described for incubations to maintain constant partitioning of THC to nonspecific plastic surfaces. Before the experiment, the beads were washed 3 times with 1 mL of assay buffer and pre-equilibrated with either recombinant CYP or HLM at the above described protein concentrations on ice for 30 min in 0.2 mL of incubation buffer for recombinant CYPs and 0.5 mL of incubation buffer for HLMs in 1.7 mL Eppendorf tubes. THC was then added to the mixture of MGSBs containing CYP or HLM protein at THC concentrations corresponding to each of the nominal concentrations used for kinetic experiments. Samples were then incubated for an additional 30 min in a shaking water bath at 37 °C, then removed from the water bath and cooled at room temperature for 5 min. After cooling, 75 μL of the mixture containing the MGSBs, and recombinant CYPs with THC were collected as the total THC sample. 100 μL of the mixture containing the MGSBs and HLMs with THC were collected as the total THC sample for HLMs. The MGSBs were then separated from solution using a DynaMag-2 Magnet and 75 μL of supernatant were collected as the free THC sample for CYP samples and 100 μL were collected for HLM samples. Three sample volumes of acetonitrile containing 1% formic acid and internal standard were added to the total and supernatant samples. The samples were centrifuged and prepared for analysis by LC-MS/MS as described above. THC concentrations were determined via LC-MS/MS as described in Section 2.4.

The unbound fraction (f_u) in microsomes was calculated as the ratio of the concentration of drug measured in supernatant (C_free) to the concentration of total drug measured prior to magnetic separation (C_total):

$$f_u=\frac{C_{\mathit{\text{free}}}}{C_{\mathit{\text{total}}}}$$ (2)

The amount of THC bound to microsomes (A_microsome) was calculated as the difference between the total amount of drug measured in solution in the presence of MGSBs and the amount of drug measured in the supernatant (amount of drug in the supernatant = concentration in supernatant measured*volume of supernatant):

$$A_{\mathit{\text{microsome}}}=A_{\mathit{\text{total}}}-A_{\mathit{\text{supernatant}}}$$ (3)

For the experiments containing FABP1, the free and FABP1 bound THC could not be feasibly separated and nonspecific binding of THC to nickel beads prevented use of alternative methods to determine free concentrations of THC in the presence of FABP1. Hence, the concentrations of unbound THC in the presence of 20 μM FABP1 were calculated from Equation (4) assuming that the thermodynamic equilibrium between unbound THC in solution and THC bound to microsomes is unaffected by the presence of FABP1:

$$\frac{A_{\mathit{\text{free}}}}{A_{\mathit{\text{microsome}}}}=\frac{A_{\mathit{\text{free}},\mathit{\text{FABP}}1}}{A_{\mathit{\text{microsome}},\mathit{\text{FABP}}1}}$$ (4)

In equation (4), A_free is the amount of unbound THC measured in the absence of FABP1, A_microsome is the amount of THC bound to microsomes in the absence of FABP1, A_free,FABP1 is the amount of unbound THC in the presence of FABP1 and A_{microsome,FABP1} is the amount of THC bound to microsomes as measured in the presence of FABP1. This equation is based on the fact that due to the excessive nonspecific binding of THC to plastics the FABP1 effectively diminishes the amount of THC nonspecifically bound to plastics without affecting the ratio between THC bound to microsomes and free THC in solution at equilibrium. The volumes of microsomes and supernatant solution were assumed to be constant between the experiments. A_free,FABP1 was calculated based on experimentally determined values of A_free, A_microsome and A_{microsome,FABP1}. The unbound concentration of THC in the presence of FABP1 was then calculated as the A_free,FABP1/V_incubation. The f_u of THC in the presence of FABP1 was calculated as the ratio of unbound THC to total THC measured prior to magnetic separation. K_m,u and k_cat values were determined by fitting appropriate kinetic models (Michaelis-Menten or sigmoidal) to metabolite formation data plotted as a function of free concentrations of THC. Similarly, unbound CL_int (CL_int,u) values were determined by fitting a linear regression to metabolite formation data plotted with free concentrations of THC for metabolites for which K_m,u and k_cat (or maximum relative formation) could not be determined.

2.14. Impact of FABP1 on THC metabolism in human liver microsomes from individual donors

The impact of FABP1 on THC metabolism was further studied in HLMs from individual donors. THC (2 μM) was preincubated with 0.025 mg/mL HLM protein from 3 individual liver donors in the presence or absence of 20 μM FABP1 in 180 μL of incubation buffer. Reactions were initiated with 1 mM NADPH (200 μL final volume), allowed to proceed for 2 min, quenched, centrifuged, and prepared for LC-MS/MS analysis as described in Section 2.3. Incubations for all donors were done in technical duplicate on two separate days. Free concentrations of THC in incubations with and without FABP1 were determined as described in Section 2.13.

3. Results

3.1. 11-OH-THC, M2, M3 and M4 are the major metabolites of THC formed in HLMs and by recombinant CYPs

The major metabolites of THC were identified in pooled HLMs and their formation by individual CYP enzymes defined using recombinant CYPs (Fig. 1). 11-OH-THC along with three mono-oxidation products (+16 Da) labeled as M2, M3 and M4 were the 4 major metabolites detected in incubations of THC with pooled HLMs. Another oxidation product (M1) eluting before M2 was also observed in pooled HLMs but this metabolite was minor compared to the other four metabolites. The M2 and M4 metabolites appeared more abundant in the incubations in pooled HLMs than 11-OH-THC. This is likely due to the high THC concentration used in these experiments (10 μM) as these metabolites became less abundant than 11-OH-THC when THC concentration was decreased (data not shown). The change in metabolite ratios with substrate concentration suggested that M2 and M4 are formed by a different enzyme than 11-OH-THC in HLMs and that the K_m towards the enzyme forming M2 and M4 is much higher than that towards the predominant CYP forming 11-OH-THC.

Fig. 1. Characterization of THC metabolism by human liver microsomes (HLMs) and recombinant cytochrome P450 (CYP) enzymes.

Pooled HLMs and recombinant CYPs were incubated with THC (10 μM) and the formation of single oxidation products of THC (+16 Da of THC parent mass) was monitored using LC-MS/MS. THC concentration of 10 μM was chosen to ensure that all potential metabolites formed could be detected. (A) The formation of THC metabolites in incubations with HLMs and recombinant CYPs is shown for a representative experiment. Relative formation is reported as the absolute peak area for each metabolite as normalized to the internal standard 11OH-THC-d₃ (B) Selected ion chromatograms (dark blue 331 > 193 m/z, gold 331 > 201 m/z) for detection of THC metabolites M1, M2, M3, 11-OH-THC and M4 identified in pooled HLMs. (C) Selected ion chromatograms (dark blue 331 > 193 m/z, gold 331 > 201 m/z) depicting the metabolite formation by recombinant cytochrome P450 enzymes that metabolized THC. For each panel, chromatograms are scaled to the largest peak observed in the sample.

The identity of the CYPs forming the individual metabolites detected in pooled HLMs was evaluated using recombinant CYPs. With the exception of CYP2A6 and CYP4F3A, all other CYPs tested metabolized THC forming several different THC metabolites (Fig. 1C). Based on incubations with recombinant CYPs, CYP1A2, CYP2B6, CYP2C8, CYP2C9, CYP2C19, CYP2D6 and CYP2J2 formed 11-OH-THC (Fig. 1C). However, the velocity of 11-OH-THC formation (10–12 pmol min⁻¹pmol CYP⁻¹) by CYP2C9 and CYP2C19 was ~3-fold higher than that observed with CYP2D6 and >10-fold higher than that by other CYPs (0.06–0.9 pmol min⁻¹pmol CYP⁻¹) suggesting these other enzymes do not significantly contribute to 11-OH-THC formation. The M2 metabolite was observed in incubations with CYP3A4, CYP3A5, and the fetal liver CYP3A isoform, CYP3A7. M3 was formed by CYP1A2, CYP2C8, CYP2C19, CYP2D6, CYP2J2, CYP3A4, CYP3A5, and CYP3A7. Based on the absolute peak area ratios per pmol CYP, CYP2C19 and CYP3A5 formed M3 with the greatest velocity followed by CYP3A4 (35% of CYP3A5 velocity). The M3 formation velocity was >70% lower with the other enzymes that formed M3 than with CYP3A4 at this substrate concentration. M4 was solely formed by CYP3A4, CYP3A5 and CYP3A7 (Fig. 1C).

M1 formation was observed in incubations with CYP1A2, CYP2D6, CYP2C8, CYP2J2 and CYP3A5 with M1 appearing as the most abundant metabolite in CYP2D6 and CYP2J2 incubations based on peak areas. However, in comparison to the abundance of the metabolites observed by recombinant CYPs overall, the peak areas for M1 with these enzymes were only ~5% of those observed for 11-OH-THC or M2 and M4 with CYP2C9, CYP2C19 and CYP3A. An oxidation product coeluting with M2 was detected in CYP1A2, CYP2C8, CYP2D6 and CYP2J2 incubations, but this metabolite was subsequently found to have a different fragmentation pattern than M2 (Fig. 2) and was considered a different metabolite (M2a).

Fig. 2. MS/MS spectra and proposed fragmentation of THC metabolites formed from THC and THC-d₃ in human liver microsome (HLMs) and recombinant cytochrome P450 (CYP) incubations.

(A) Enhanced product ion (MS/MS) spectra are shown for the six specific metabolites detected. The representative spectra are from incubations with CYP2D6 (M1), CYP3A4 (M2), CYP1A2 (M2a) CYP2C19 (M3), CYP2C9 (11-OH-THC) and CYP3A4 (M4). (B) Proposed fragmentation and fragment structures for THC and THC-d₃ metabolites hydroxylated on the aliphatic side chain using 3′OH-THC as an example. (C) Proposed fragmentation and fragment structures for THC and THC-d₃ metabolites hydroxylated in the cyclohexenyl C-ring, using 11-OH-THC as an example. Masses in the spectra that correspond to the proposed fragment structures are labeled in red. Asterisks indicate the location of deuterium (d₃) labels on the metabolites and MS/MS fragments.

To elucidate the identity of M1–M4 metabolites of THC, MS/MS spectra were collected for each metabolite formed in HLM and recombinant enzyme incubations of THC and THC-d₃ (Fig. 2). The deuterium labels in THC-d₃ are in the terminal methyl group in the aliphatic side chain and hence the retention of the deuterium labels in MS/MS fragments provides information for assessment of hydroxylation sites and fragmentation patterns. Several fragments were considered informative to define the oxidation sites in THC. The presence of the 191 m/z fragment observed in a metabolite formed from THC, with a corresponding fragment of 194 m/z in a metabolite formed from THC-d₃, is characteristic for a hydroxylation in the aliphatic side chain of THC (Fig. 2B). Hence M1 and M2a were considered oxidations in the aliphatic side chain (Fig. 3). In contrast the 193 m/z fragment in the metabolites formed in THC incubations, that corresponds to the 196 m/z fragment in the metabolites formed in THC-d₃ incubations, is indicative of an intact aliphatic side chain and lack of oxidation in the aromatic A-ring (Fig. 2C, Fig. 3). The presence of the 257 m/z fragment in the metabolites formed in THC incubations, corresponding to the 260 m/z fragment in the metabolites formed in THC-d₃ incubations, corroborates the lack of hydroxylation in the side chain or in the aromatic A-ring (Fig. 2C). Metabolites containing these fragments were considered hydroxylation products in the cyclohexenyl C- ring based on the observed fragmentation patterns. The proposed structures of the major fragments observed in the MS/MS spectra and the predicted fragmentation patterns are shown in Fig. 2B and 2C.

Fig. 3. Proposed structures and sites of oxidation in THC by human liver cytochrome P450 (CYP) enzymes and main CYPs that form the individual metabolites.

The location of putative hydroxylation sites are indicated by asterisk for the primary THC metabolites formed in HLMs. M2, M3 and M4 metabolites are hydroxylated in the cyclohexenyl C-ring. M1 and M2a metabolites are hydroxylated at the aliphatic side chain. The major CYPs that contribute to the formation of each metabolite are listed. Figure created with Biorender.com.

Previously, the hydroxylation product formed by CYP2J2 was proposed to be 1′OH-THC [41]. The main metabolite formed by CYP2J2 here was M1 which may be the 1′OH-THC. However, the fragmentation observed for M1 was inconclusive for determining the hydroxylation site in the aliphatic side chain. Previous studies have also proposed 3′OH-THC formation in human liver [18]. The fragmentation observed here for M2a is consistent with 3′OH-THC but similar to M1. The 191 m/z, 257 m/z and 242 m/z fragments in the MS/MS spectra of M1 and M2a formed from THC and the 194 m/z, 257 m/z and 242 m/z fragments in the MS/MS spectra of M1 and M2a formed from THC-d₃ (Fig. 2) can all be products of hydroxylation in any of the carbons in the aliphatic side chain (Fig. 3). Metabolomic studies have in fact proposed hydroxylation on all the carbons in the side chain [4]. Further studies are needed to fully elucidate the exact positions of THC hydroxylation on the aliphatic side chain.

M2, M3 and M4 were tentatively identified as hydroxylations in the C-ring. The presence of the 271 m/z and 274 m/z fragments in M2, M3 and M4 spectra collected from the THC and THC-d₃ incubations (Fig. 2), respectively, suggested that the gem-dimethyl groups in the B-ring were not hydroxylated. These metabolites may be 8-OH-THC and 7-OH-THC (Fig. 3). In previous studies two 8-OH-THC metabolites, 8α- and 8β-OH-THC were identified [42] and 8α- and 8β-OH-THC were shown to be formed by CYP3A4 [3]. If present, the fragmentation pattern of 8α- and 8β-OH-THC would be expected to be close to identical. Indeed, M2 and M3 mass spectra have the same characteristic fragmentation patterns (Fig. 2) and both are formed by CYP3A4. These data suggest that M2 and M3 are likely 8α- and 8β-OH-THC. M4 may represent 7-OH-THC (Fig. 3) as the fragmentation pattern of M4 was different from M2 and M3 but still characteristic of a hydroxylation product based on the loss of water from the parent ion. The 7-OH-metabolite has been previously reported for Δ8-THC [43]. The current data do not exclude the formation of additional hydroxylation products that have not been previously identified. Notably, M3 was also formed by other CYPs not just CYP3A4 showing that more CYPs metabolize THC than previously assumed [3,18].

None of the metabolites detected showed spectral evidence of a presence of the 9α,10α-THC-epoxide although previous reports have suggested that 9α,10α-THC-epoxide is formed by CYP3A4 [3,18]. This may be due to the much lower substrate concentrations used in this study than in the previous experiments [3,18,19]. An epoxide metabolite would be expected to fragment differently to the hydroxylation products, and the epoxide would be expected to be present in the MS/MS fragments containing the C-ring (Fig. 2). The loss of water and the presence of the 271 m/z and 274 m/z in M4 MS/MS spectrum suggest this metabolite is a hydroxylation product rather than an epoxide, even though its late retention time would support a less polar epoxide rather than hydroxylation. The expected metabolic sites of the major THC metabolites formed by human liver CYPs are illustrated in Fig. 3. Based on MS/MS spectral analyses of HLM incubations, M2a was absent in the HLM incubations even though it was detected in recombinant CYP incubations.

3.2. THC metabolite formation correlates strongly with CYP2C9, CYP3A4 and CYP2C19 activity in human liver microsomes from individual donors

The metabolite formation data in incubations of THC with recombinant CYPs showed that specific CYPs form distinct THC metabolites. To explore the formation of these metabolites and the interindividual variability in THC metabolism, THC metabolism was characterized in a panel of HLMs from individual liver donors with varying CYP2C9, CYP2C19 and CYP3A activity (Fig. 4). The formation of 11-OH-THC varied by 10.5-fold across the 23 individual HLMs tested. The formation of M2, M3, and M4 was more variable than 11-OH-THC formation between donors and varied 129-fold, 40-fold, and 112-fold, respectively (Fig. 4). M2, M3 and M4 formation was detected in all 23 HLM donors. The ratio of M2/M3 varied between livers (mean ± S.D.: 3.7 ± 1.8; range: 0.9 – 6.3) suggesting different enzymes contribute to M2 and M3 formation. In contrast M2/M4 ratio was relatively constant between donors (mean ± S.D.: 3.0 ± 0.6; range: 2.1–4.6) suggesting M2 and M4 are formed by the same enzyme(s). M1 was not detected in any of the incubations in individual HLM donors.

Fig. 4. Characterization of THC metabolite formation and interindividual variability in THC metabolism in a panel of individual human livers.

The formation of (A) 11-OH-THC, (B) M2, (C) M3 and (D) M4 metabolites in HLMs from 23 individual donors is shown in incubations with THC (2 μM). Nominal THC concentration of 2 μM was chosen to capture formation of all four metabolites at a THC concentration reflecting C_max concentrations in heavy cannabis users. Gold bars indicate livers that express CYP3A5 while blue bars are HLMs without CYP3A5 expression. The mean value of replicate experiments each done in duplicate is shown for each donor. Asterisks indicate metabolite formation below limit of quantification for at least one technical replicate within the data set.

Based on the data of metabolite formation in individual HLMs together with the information of recombinant CYPs that form the specific metabolites, we hypothesized that 11-OH-THC is formed in HLMs by CYP2C9 and CYP2C19, M2 and M4 are formed by CYP3A4 and CYP3A5, and M3 is formed by CYP2C19 and CYP3A enzymes. CYP2D6 was not expected to contribute to THC metabolite formation in HLMs even though recombinant CYP2D6 formed M3 and 11-OH-THC. M1 is the most abundant metabolite formed by recombinant CYP2D6, but M1 was not detected in the individual HLM incubations suggesting minimal contribution of CYP2D6 to THC metabolism.

The hypothesis of CYP2C9, CYP2C19 and CYP3A4 contribution to THC metabolite formation was first tested via correlation analysis between the THC metabolite formation and the formation of specific CYP probe metabolites 4-OH-tolbutamide (4-OH-TBU; CYP2C9), 1-OH-midazolam (1-OH-MDZ; CYP3A) and 4-OH-mephenytoin (4-OH-MPH; CYP2C19). Correlation between the formation of THC metabolites and the activity of CYP2C9, CYP3A and CYP2C19 for individual HLMs was examined using simple linear regression (Fig. 5). 11-OH-THC formation correlated significantly with 4-OH-TBU and 1-OH-MDZ but not with 4-OH-MPH formation (Table 1). The correlation with 1-OH-MDZ formation was surprising as 11-OH-THC formation was not observed with recombinant CYP3A4 (Fig. 1). We hypothesized that this may be due to a correlation between CYP2C9 and CYP3A4 activity. Indeed, co-linearity of CYP2C9 and CYP3A activity was detected with p-value < 0.01. This co-linearity also impacted correlation analyses between M2, M3, and M4 formation and 4-OH-TBU and 1-OH-MDZ formation. Formation of all three THC metabolites correlated with CYP2C9 and CYP3A4 activity (Table 1) although recombinant CYP2C9 did not form any of these three metabolites. Only M3 formation correlated significantly with 4-OH-MPH formation (Table 1). To address the co-linearity, multiple linear regression was used to assess which CYP isoform(s) likely explain metabolite formation in the dataset (Table 1). In the multiple linear regression analysis, the formation of 11-OH-THC strongly correlated with CYP2C9 activity (p-value < 0.01). CYP3A activity correlated with M2, M3, and M4 formation (p-value < 0.01) and CYP2C19 activity correlated with M3 formation (p-value < 0.01) (Table 1).

Fig. 5. Correlation of THC metabolite formation with cytochrome P450 (CYP) probe metabolism in the human liver microsome (HLM) panel.

Panels A through F show simple linear regression between metabolite formation and CYP specific activity in HLM donors (n = 23). Panel G shows simple linear regression between CYP2C9 and CYP3A specific activities in HLM donors (n = 23). Correlations between (A) velocity of 11-OH-THC formation and CYP2C9 specific activity (4-OH-TBU formation rate), (B) M2 relative formation and CYP3A specific activity (1-OH-MDZ formation rate), (C) M4 relative formation and CYP3A specific activity, (D) M3 relative formation and CYP3A specific activity and (E) M3 relative formation and CYP2C19 specific activity (4-OH-MPH formation rate) are shown. (F) Lack of correlation between M2 relative formation and CYP2C19 specific activity is depicted. (G) Correlation between CYP2C9 and CYP3A specific activity is shown. (H) Multiple linear regression of M3 relative formation with CYP3A and CYP2C19 specific activities, both as predictors, is shown plotted from two different angles (left panel and right panel). The correlations were significant (p-value < 0.01) between metabolite formation and CYP isoform specific activity in panels A-E and H. CYP2C9 and CYP3A specific activities were also found to be correlated (p-value < 0.01).

Table 1.

Correlations between cytochrome P450 (CYP) specific activities and THC metabolite formation in human liver microsomes (HLMs) from the panel of individual liver donors (n = 23). Absolute p-values for the regression models are shown. R-squared (R²) is reported for the simple linear regression models.

Model used	Activity	11-OH-THC formation	M2 relative formation	M3 relative formation	M4 relative formation
Simple linear regression	CYP2C9	p = 0.000000301	p = 0.000000347	p = 0.0000197	p = 0.000000578
		R2 = 0.721	R2 = 0.7172	R2 = 0.5881	R2 = 0.7034
	CYP3A	p = 0.000113	p = 0.000000000000305	p = 0.0000569	p = 0.0000000000000595
		R2 = 0.5159	R2 = 0.9242	R2 = 0.5456	R2 = 0.9351
	CYP2C19	p = 0.253	p = 0.68555	p = 0.005432	p = 0.577067
		R2 = 0.06175	R2 = 0.007963	R2 = 0.3138	R2 = 0.01505
Multiple linear regression	CYP2C9 + CYP3A + CYP2C19	2C9: p = 0.00493	2C9: p = 0.4196	2C9: p = 0.344	2C9: p = 0.5315
		3A: p = 0.60454	3A: p = 0.00000127	3A: p = 0.000000222	3A: p = 0.000000311
		2C19: p = 0.11172	2C19: p = 0.9654	2C19: p = 0.0000000000909	2C19: p = 0.6904

To further confirm the statistical rigor of the correlation analyses in establishing the relevant CYPs forming the THC metabolites, stepwise linear regression for predictor selection in multiple linear regression models was performed (Table 2). The model fit for M3 formation improved significantly when CYP2C19 activity was included in addition to CYP3A activity, suggesting contribution of both CYP3A and CYP2C19 towards M3 formation (p-value < 0.01). No significant improvement with any other combinations of the three CYPs analyzed was detected for other metabolites in the ANOVA test. Therefore, the correlation analysis suggests that 11-OH-THC formation is predominantly by CYP2C9, M2 and M4 formation is by CYP3A, and M3 formation is dependent on CYP3A and CYP2C19 activities.

Table 2.

Stepwise linear regression analysis of cytochrome P450 (CYP) activities as measured by specific metabolite formation from probe substrates and THC metabolite formation. Y ~ X indicates the linear regression with Y as a response variable and X as the predictor. Added predictor as indicated was included to the linear regression one at a time to determine whether it improves the goodness of fit. The absolute p-value from the analysis of variance (ANOVA) test is shown.

Linear Regression	Added Predictor	P-value
11-OH-THC ~ CYP2C9	CYP3A	0.949
	CYP2C19	0.122
M2 ~ CYP3A	CYP2C9	0.369
	CYP2C19	0.728
M3 ~ CYP3A	CYP2C9	0.0721
	CYP2C19	0.00000000000880
M3 ~ CYP3A + CYP2C19	CYP2C9	0.344
M4 ~ CYP3A	CYP2C9	0.594
	CYP2C19	0.844

3.3. Contribution of CYPs to the formation of 11-OH-THC, M2, M3 and M4 varies between individual donors

In vivo data suggest that the majority of THC metabolism is mediated by CYP2C9 [15]. The metabolite formation data from recombinant CYPs and correlation analysis data in HLMs suggest that CYP2C19 and CYP3A may also contribute to THC metabolism. Therefore, the contribution of CYP2C9, CYP2C19 and CYP3A4 to the formation of THC metabolites in HLMs from six different donors was determined using selective inhibitors of CYP2C9 (tienilic acid), CYP3A4 (CYP3cide) and CYP2C19 ((+)-N-3-benzylnirvanol) (Fig. 6). To assess the relative role of CYP3A5 in THC metabolism, two CYP3A5 expressing donors were included in the inhibition studies. CYP3cide only inhibits CYP3A4 and not CYP3A5 and hence % inhibition by CYP3cide can be interpreted as the fractional formation of the metabolite by CYP3A4 specifically.

Fig. 6. Inhibition of THC metabolism in human liver microsomes (HLMs) from six individual liver donors.

The formation of (A) 11-OH-THC, (B) M2, (C) M3 and (D) M4 in HLMs in the presence of selective inhibitors is shown as the % activity remaining compared to no inhibitor controls. Inhibition by tienilic acid (CYP2C9), CYP3cide (CYP3A4) and (+)-N-3-benzyl-nirvanol (CYP2C19) are shown in gold, green and blue, respectively. (E) Specific activity of CYP2C9, CYP3A and CYP2C19 enzymes in the six livers as measured using the indicated probe substrates. The order of the six livers from left to right for each dataset is HL-105, HL-109, HL-114, HL-134, HL-108, and HL-158. Shaded bars represent livers with CYP3A5 expression. Metabolite formation below limit of quantification is indicated by asterisks. (F) CYP2C9, CYP3A5 and CYP2C19 genotypes for the six individual livers included.

Tienilic acid inhibited 65–78% of 11-OH-THC formation in the six donor HLMs (Fig. 6A) supporting the predominant role of CYP2C9 in 11-OH-THC formation. CYP3cide had little to no impact on 11-OH-THC formation (0–12% inhibition). Benzylnirvanol inhibited 11-OH-THC formation by up to ~17% in the individual donors. The greatest percent inhibition by benzylnirvanol was observed in donors with high CYP2C19 activity. The tienilic acid inhibition was also the weakest in these donors with high CYP2C19 activity (HL-105 and HL-114) supporting a role of CYP2C19 in THC clearance and 11-OH-THC formation in individuals with either low CYP2C9 activity/expression or high CYP2C19 activity.

Tienilic acid had no effect on M2, M3 and M4 formation (Fig. 6B–D), a finding consistent with the lack of formation of these metabolites by recombinant CYP2C9. Inactivation of CYP3A4 by CYP3cide inhibited the majority of M2 (57–90%) and M4 formation (63–96%) in all six donors supporting the major role of CYP3A4 in formation of these THC metabolites. The extent of inhibition of M4 formation by CYP3cide was greater in HLMs with no CYP3A5 expression compared to HLMs with CYP3A5 expression (Fig. 6D) consistent with specific inhibition of CYP3A4 by CYP3cide and contribution of CYP3A5 to M4 formation when expressed.

Benzylnirvanol inhibited M3 formation by up to 67% but did not inhibit M2 and M4 formation (Fig. 6C). In one donor who had low CYP2C19 activity, high CYP3A activity and did not express CYP3A5 (HL-109), benzylnirvanol had no impact on M3 formation and CYP3cide inhibited nearly 100% of M3 formation in this donor. In the other three HLMs that had no CYP3A5 expression, benzylnirvanol inhibited 67%, 66% and 22% of M3 formation and inhibition by CYP3cide accounted for the remaining M3 formation in the same livers (28, 39, and 81 %). These data suggest that CYP3A4 and CYP2C19 are the only contributors to M3 formation in the absence of CYP3A5 expression. In contrast, total M3 formation in livers expressing CYP3A5 could not be completely accounted for by CYP3cide and benzylnirvanol inhibition suggesting that CYP3A5 also contributes to M3 formation when expressed.

Taken together, the inhibitor studies are consistent with CYP2C9 being responsible for forming 11-OH-THC, CYP3A contributing to the majority of M2 and M4 formation, and CYP3A and CYP2C19 both contributing to M3 formation. These data also show that the contributions by CYP2C9, CYP3A4, CYP3A5 and CYP2C19 toward THC metabolism vary considerably between individuals based on specific CYP expression/activity. The genetic background of individuals relating to these enzymes may have an impact on the metabolism and clearance of THC.

The relative contribution of individual enzymes to substrate metabolism can be dependent on the substrate concentration if multiple enzymes contribute to the formation of the metabolite, and if the K_m values for the enzymes are sufficiently different. The contribution of multiple enzymes can also be apparent in biphasic Eadie-Hofstee plots. To further explore the kinetics of THC metabolism in HLMs and to determine the potential dependence of the relative contribution of individual enzymes on THC concentration, the formation kinetics of 11-OH-THC, M2, M3 and M4 were characterized in pooled HLMs (Fig. 7, Table 3). The unbound fraction (f_u) of THC measured for each nominal concentration of THC used in kinetic experiments was 0.015 ± 0.001 and did not change within the range of THC concentrations used. The 11-OH-THC formation had a linear Eadie-Hofstee plot and a K_m,u of 1.4 ± 0.2 nM (Fig. 7A). This result is consistent with the inhibition data in HLMs from individual donors showing that CYP2C9 is likely the major contributor to 11-OH-THC formation.

Fig. 7. Kinetics of 11-OH-THC, M2, M3 and M4 formation in pooled human liver microsomes (HLMs).

Formation of (A) 11-OH-THC, (B) M2, (C) M3 and (D) M4 metabolites in HLMs as a function of free THC concentration is plotted. The Michaelis-Menten equation (11-OH-THC, M2 and M3 formation data, A-C, solid lines) or linear regression model (M4 formation data D, solid lines) was fit to the data and model fits are shown as solid lines. Eadie-Hofstee plots are shown as insets. Replicate experiments performed on separate days are shown in dark blue, green and gold symbols and lines.

Table 3.

Kinetic parameter estimates of THC metabolite formation in pooled human liver microsomes. The data are reported as means ± standard deviation from experiments done on 3 separate days. N.D. Not determined due to lack of saturation of the metabolite formation.

	11-OH-THC	M2	M3	M4
Km,unbound (nM)	1.4 ± 0.2	> 45	> 45	> 45
Vmax (pmol/min/mg HLM)	521 ± 53	N.D.	N.D.	N.D.

The formation kinetics for M2, M3 and M4 were different from each other suggesting different CYPs contribute to the formation of these metabolites in the pooled HLMs. The K_m,u values for M2, M3 and M4 formation were >10-fold higher than the K_m,u for 11-OH-THC. The K_m,u values for M2 and M3 were >45 nM. Eadie-Hofstee plots were non-linear for M2 and M3 formation (Fig. 7B and 7C) suggesting that multiple enzymes contribute to M2 and M3 formation in HLMs. These results are consistent with the incomplete inhibition of M2 formation even in livers that did not express CYP3A5, and the predicted contribution of CYP3A and CYP2C19 to M3 formation.

M4 formation was linear within the range of THC concentrations tested suggesting that the K_m,u for M4 formation is ≫ 45 nM (Fig. 7D). As CYP3cide inhibited nearly all of M4 formation in all the individual livers, the K_m,u of M4 formation likely reflects the K_m,u of THC with CYP3A4. The kinetic data in HLMs suggest that the K_m,u for CYP3A4 is greater than the K_m,u of CYP2C9 and CYP2C19, and that CYP3A4 contribution to THC metabolism will be greater with higher circulating concentrations of THC in vivo.

3.4. THC and 11-OH-THC bind to human liver fatty acid binding protein (FABP1)

Previous reports have shown that THC binds to the liver fatty acid binding protein (FABP1) and a crystal structure of THC bound to FABP1 was reported (PDB: 6MP4) [31] illustrating the binding characteristics of THC with FABP1. FABP1 also alters diclofenac metabolism by recombinant CYP2C9 [30]. Hence, we hypothesized that binding of cannabinoids to FABP1 may alter their metabolism by CYPs. To test this hypothesis recombinant FABP1 was expressed and purified, and binding of the key cannabinoids THC, 11-OH-THC, CBD, 2AG and AEA to FABP1 was tested prior to testing the impact of FABP1 on THC metabolism by CYPs.

None of the cannabinoids tested fully diminished the specific fluorescence of DAUDA bound to FABP1 when cannabinoid binding was saturated in DAUDA displacement assays (E_min = 34–77%) (Fig. 8, Table 4). In addition, the fluorescence spectra of DAUDA-FABP1 in the presence of all cannabinoids tested were blue-shifted relative to the spectrum of DAUDA in solution (Fig. 8). These results suggest that cannabinoids do not fully displace DAUDA from FABP1 but rather that cannabinoids form ternary complexes with FABP1 and DAUDA. Yet, based on the prior crystal structure [31], THC can also bind to FABP1 in the absence of an additional fatty acid as a THC-FABP1 complex. Hence, the binding orientation and affinity of THC to FABP1 may differ in the presence and absence of DAUDA or endogenous fatty acids. Ternary DAUDA-FABP1-drug complexes were previously reported for other drug ligands of human FABP1 [30]. Assuming formation of DAUDA-FABP1-cannabinoid ternary complex, the binding affinities (K_d) of cannabinoids in the DAUDA displacement assays were determined by fitting a ternary complex binding model to the fluorescence data. All cannabinoids bound to FABP1 with sub-micromolar binding affinities (Table 4). Given the high expression of FABP1 in the liver (0.7–1 mM) [27,29], these data suggest that the pharmacologically active cannabinoids THC, 11-OH-THC and CBD are bound to FABP1 in the liver in vivo. These findings also suggest that THC, 11-OH-THC and CBD may form ternary complexes with endocannabinoids or other fatty acids and that FABP1 may impact endocannabinoid signaling and metabolism.

Fig. 8. Binding of cannabinoids to FABP1.

Top panels show fluorescence emission spectra of titrations with THC, 11-OH-THC, cannabidiol (CBD) and the endogenous cannabinoids anandamide (AEA) and 2-arachidonoylglycerol (2AG) from DAUDA displacement assays. The top dark blue spectra represent the spectra of DAUDA bound with FABP1 (DAUDA-FABP1) in the absence of a cannabinoid and each subsequent spectrum represents increasing concentrations of cannabinoids to saturation (pink spectra). The shaded gray spectrum represents the spectrum of DAUDA in solution in the absence of FABP1. Bottom panels show the change in the specific fluorescence of DAUDA-FABP1 based on singular value decomposition analysis of titration spectra. Dark blue, gold, and green open circles show replicate experiments done on three separate days and the corresponding solid lines show COPASI fits of a ternary complex binding model to the data.

Table 4.

The binding affinities of the tested cannabinoids with FABP1. Ligand binding was determined via DAUDA displacement assay. The K_d value was determined from the COPASI fit of the ternary binding model to the DAUDA displacement data while EC₅₀ and E_min were determined from fit of a three parameter dose response curve to the data. All parameter estimates are reported as means ± standard deviation from three replicate experiments performed on separate days. The ligands tested included THC, 11-OH-THC, cannabidiol (CBD) and the endogenous cannabinoids anandamide (AEA) and 2-arachidonoylglycerol (2AG).

	Kd (μM)	EC50 (μM)	Emin (% fluorescence remaining)
THC	0.30 ± 0.34	0.22 ± 0.13	52 ± 4.0
11-OH-THC	0.86 ± 0.15	1.5 ± 0.23	23 ± 3.8
CBD	0.57 ± 0.006	0.92 ± 0.07	46 ± 6.1
2AG	0.06 ± 0.04	0.18 ± 0.02	65 ± 1.3
AEA	0.19 ± 0.12	0.31 ± 0.14	66 ± 5.7

3.5. FABP1 alters the kinetics of THC metabolite formation by CYP2C9, CYP2C19 and CYP3A4

To determine if THC binding to FABP1 impacts metabolism of THC by CYP enzymes, THC metabolite formation kinetics by recombinant CYP2C9, CYP2C19 and CYP3A4 was characterized in the presence and absence of 20 μM FABP1 (Fig. 9). Free concentrations of THC were determined with and without FABP1 present using magnetic silica beads as described in Materials and Methods. K_m,u and k_cat values determined based on measured free THC concentrations in the incubations are summarized in Table 5. In the absence of FABP1, the f_u of THC was 0.42 ± 0.04 with CYP2C9, 0.15 ± 0.02 with CYP3A4 and 0.07 ± 0.01 with CYP2C19 and unchanged across THC concentrations used (Fig. 9G–I). However, f_u decreased with increasing concentrations of total microsomal protein. The microsomal protein content in the incubations was 0.0016 mg/mL for CYP2C9, 0.004 mg/mL for CYP3A4 and 0.013 mg/mL for CYP2C19. As such the highest f_u was observed with the CYP with the highest recombinant CYP expression level per mg microsomal protein and lowest total microsomal protein content in the incubations.

Fig. 9. Kinetics of THC metabolism by recombinant cytochrome P450 (CYP) enzymes CYP2C9, CYP2C19 and CYP3A4 in the presence and absence of FABP1.

The velocities of 11-OH-THC formation by (A) CYP2C9 and (B) CYP2C19 are plotted as a function of free THC concentration ([THC_u]). The relative rates of (C) M3 formation by CYP2C19, (D) M2 formation by CYP3A4, (E) M3 formation by CYP3A4 and (F) M4 formation by CYP3A4, are plotted as a function of free THC concentrations. The formation of metabolites in the absence and presence of FABP1 is indicated by solid and open circles, respectively. The solid lines show the Michaelis-Menten model fit to the 11-OH-THC, M2 and M3 formation data in the absence of FABP1 (A-E), the allosteric model fit to the M4 formation data in the absence of FABP1 (F), the Michaelis-Menten model fit to the 11-OH-THC and M3 formation by CYP2C19 data in the presence of FABP1 (B, C), a linear regression model fit to the 11-OH-THC and M2 and M4 formation data for CYP2C9 and CYP3A4, respectively (A, D, F). Replicate experiments performed on separate days are shown in dark blue, green and gold. The unbound fraction of THC (G-I) in the absence (closed circles) and presence (open circles) of 20 μM FABP1 is shown for all nominal THC concentrations used in kinetic experiments with CYP2C9 (G), CYP3A4 (H) and CYP2C19 (I).

Table 5.

Kinetic parameter estimates of THC metabolite formation by recombinant cytochrome P450 (CYP) enzymes CYP2C9, CYP2C19 and CYP3A4 in the presence and absence of 20 μM FABP1. All the data except M4 formation are reported for the fits of the Michaelis Menten model to the metabolite formation data as a function of unbound THC concentration. The allosteric model was fit for M4 formation and the K_0.5,u and hill slope (h) are reported. N.D. Not determined due to lack of detection of the metabolite in the presence of FABP1 (M3, CYP3A4), or lack of saturation of the metabolite formation in the presence of FABP1.

	Km,u −FABP1 (nM)	Km,u + FABP1 (nM)	kcat −FABP1 (min−1)	kcat + FABP1 (min−1)
11-OH-THC
CYP2C9	0.77 ± 0.16	> 20	12 ± 0.87	N.D
CYP2C19	2.2 ± 0.87	5.2 ± 0.64	14 ± 0.68	8.0 ± 2.0
M2
CYP3A4	29 ± 19	> 20	N.D	N.D
M3
CYP2C19	2.2 ± 0.85	4.7 ± 0.72	N.D	N.D
CYP3A4	18 ± 13	N.D.	N.D.	N.D
M4
CYP3A4	43 ± 25 (h = 1.4 ± 0.5)	>20	N.D	N.D

In the absence of FABP1 the K_m,u for 11-OH-THC formation by CYP2C9 (0.8 nM) was lower than the K_m,u for 11-OH-THC (2.2 nM) and M3 (2.2 nM) formation by CYP2C19. The k_cat values for 11-OH-THC formation by CYP2C9 and CYP2C19 were comparable (12 and 14 min⁻¹, respectively) (Table 5). Surprisingly, M4 formation by CYP3A4 was allosteric while the formation of M2 and M3 was not (Fig. 9). The K_m,u values for M2 and M3 formation, and the K0.5,u for M4 formation by CYP3A4 (29 ± 19, 18 ± 13 and 43 ± 25 nM, respectively) were over 8-fold greater than K_m,u values determined for either CYP2C9 or CYP2C19. The k_cat for M2, M3 and M4 formation by any of the CYPs could not be determined due to the lack of available reference standards for these metabolites.

FABP1 had a considerable effect on THC metabolism by all three CYPs although the impact differed between the three CYPs. The f_u values for THC in the presence of FABP1 were 0.08 ± 0.02 with CYP2C9, 0.02 ± 0.01 with CYP3A4 and 0.01 ± 0.01 with CYP2C19 and lower than those measured (0.07–0.42) in the absence of FABP1 (Fig. 9G–I). This is consistent with THC binding to FABP1. The f_u was not THC concentration dependent across the THC concentrations used. The K_m,u and k_cat could not be estimated for THC metabolite formation by CYP2C9 and CYP3A4 in the presence of FABP1 due to the considerable decrease in metabolite formation and apparent increase in K_m (Fig. 9A and 9D–F). FABP1 had no significant effect on the K_m,u of 11-OH-THC (p = 0.38, 0.089, 0.045 for three replicate experiments done on separate days) or M3 (p = 0.64, 0.061, 0.15 for three replicate experiments done on separate days) formation by CYP2C19 (Table 5) nor on the k_cat of 11-OH-THC or M3 formation by CYP2C19 (p-values 0.08–0.49).

3.6. FABP1 decreases THC metabolite formation in human liver microsomes and may alter relative contributions of CYP enzymes to THC metabolism

Data with recombinant CYPs suggest that FABP1 impacts THC metabolite formation by CYP2C9, CYP2C19 and CYP3A4 and that the effects are enzyme specific. We hypothesized that the effects of FABP1 on THC metabolism in recombinant systems will translate to altered THC metabolite formation in HLMs when FABP1 is present. To determine how FABP1 affects THC metabolism by CYP2C9, CYP2C19 and CYP3A4, incubations were done with HLMs from three donors (non-CYP3A5 expressors) in the presence and absence of 20 μM FABP1 (Fig. 10). The formation of M2 and M4 metabolites was decreased by 65–75% and 76–85%, respectively, in the presence of FABP1. This is consistent with the significant decrease in the formation of these metabolites by recombinant CYP3A4 in the presence of FABP1. The decrease in the formation of the M3 metabolite was smaller 49–68%. This was surprising as with the recombinant CYP3A4 M3 formation was completely abolished in the presence of FABP1. This suggests that in the presence of FABP1 CYP2C19 is the predominant enzyme forming M3 as CYP2C19 still formed this metabolite in the presence of FABP1. Surprisingly, there was little to no effect on the formation of 11-OH-THC in the three donor HLMs (0–9%) in the presence of FABP1. To determine if free drug hypothesis could explain these results, free concentrations of THC were measured in the incubations in the presence and absence of FABP1. Based on these measurements, FABP1 binding decreased the free concentrations of THC in incubations with FABP1 by 86 % compared to incubations in the absence of FABP1 (2 vs 14 nM free THC in the presence and absence of FABP1, respectively) confirming THC binding to FABP1 in HLM incubations.

Fig. 10. The effect of FABP1 on THC metabolite formation in human liver microsomes (HLMs).

The panels show (A) 11-OH-THC, (B) M2, (C) M3 and (D) M4 metabolite formation in the absence (dark blue) and presence (gold) of 20 μM FABP1 in HLMs from three individual donors with no CYP3A5 expression.

4. Discussion

THC is cleared in humans mainly by metabolism with less than 1 % of dose recovered in urine and feces after iv administration of THC [44]. The metabolic clearance of THC is high with the hepatic extraction ratio around 0.8 despite the high plasma protein binding (f_u = 0.01) and low blood to plasma ratio (0.67) [45]. The high metabolic clearance leads to poor oral bioavailability of THC (<20%) [46]. Yet, the exact contribution of CYPs to THC metabolism in the liver and in the gut mucosa upon first pass remains uncertain. Based on studies of THC AUC in individuals with different CYP2C9 genotypes and in the presence and absence of ketoconazole [17], CYP2C9 and CYP3A4 contribute to THC metabolism in vivo. Surprisingly, no CYP3A4 contribution to THC depletion in in vitro metabolic studies was detected [11] and instead, in HLMs, CYP2C9 was estimated to contribute to >90% of THC clearance with a minor contribution from CYP2D6 [13]. This is despite previous identification of metabolites of THC formed by CYP3A4 [3,18] and depletion of THC by recombinant CYP3A4, CYP2C19 and CYP1A1 [11]. Taken together, these discrepancies highlight remaining gaps in the understanding of THC metabolism by specific CYPs and the need for systematic evaluation of the metabolites formed by CYPs in the human liver and intestines.

The primary metabolite of THC detected in vivo is 11-OH-THC [46,47]. 11-OH-THC has been shown to be formed by CYP2C9, CYP2C19, CYP1A2 and CYP2D6 [3,11]. The data shown here agrees with the prior work on 11-OH-THC formation (Fig. 1). However, our data also show that the intrinsic clearance of 11-OH-THC formation is ~ 3-fold higher by CYP2C9 (CL_int = 13.9 mL·min⁻¹·pmol P450⁻¹) than by CYP2C19 (CL_int = 4.7 mL·min⁻¹·pmol P450·¹), and that the formation rates of 11-OH-THC by CYP2D6 and CYP1A2 are minor compared to CYP2C9. Notably, CYP2C9 did not form any other oxidation products from THC except 11-OH-THC, while, in pooled HLMs, four other metabolites were formed (Fig. 1). This clearly indicates that multiple CYPs in the HLMs oxidize THC. The metabolite profile and metabolite ratios of THC also showed large variability across the 23 donor HLMs tested strongly suggesting that multiple enzymes are responsible for THC metabolism in human liver. Although CYP1A1 has been shown to deplete THC, it was not considered here as CYP1A1 is not expressed in the liver but is typically expressed mainly in the lungs, and no depletion of THC was detected in human lung microsomes [48].

The detection of the additional oxidation products of THC in HLMs is consistent with early studies that observed multiple oxidation products of THC using TLC and identified these metabolites as 11-OH-THC, 8α/β-OH-THC, 9α,10α-THC-epoxide and 3′OH-THC in HLMs from a single Japanese male donor [18]. More recently, 1′OH-THC instead of 3′OH-THC was reported as a THC metabolite formed by CYP2J2 [41]. The data shown here supports multiple metabolites (M1 and M2a) resulting from hydroxylation of the aliphatic side chain but the oxidation site in the side chain is inconclusive. The data here also suggest that M2 and M3 are 8α- and 8β-OH-THC and that M4 is likely 7-OH-THC. Notably, M3 was also efficiently formed by CYP2C19, a finding that has not been previously reported. Whether these metabolites identified are present in vivo in humans is currently unknown.

Significant correlation of CYP2C9 activity with 11-OH-THC formation and the >80% inhibition of 11-OH-THC formation by tienilic acid, a selective irreversible inhibitor of CYP2C9 are consistent with previous in vitro work showing that CYP2C9 is the major contributor to 11-OH-THC formation [3,10,11]. However, stepwise regression analysis showed a trend towards better correlation when CYP2C19 activity was added and (+)-N-3-benzylnirvanol inhibited portion of 11-OH-THC formation in livers in which CYP2C19 activity is high (Fig. 6). Hence CYP2C19 may be important in 11-OH-THC formation in some individuals and if CYP2C9 activity is impaired. The greater 11-OH-THC formation CL_int by recombinant CYP2C9 and higher expression level of CYP2C9 than CYP2C19 in human liver [49] explain the dominant role of CYP2C9 in 11-OH-THC formation.

The strong correlation of M2 and M4 formation in donor HLMs with CYP3A activity together with the inhibition of M2 and M4 formation by CYP3cide support a role for CYP3A4 in forming the cyclohexenyl hydroxylated metabolites M2 and M4. This is in agreement with previous findings of the role of CYP3A4 in THC metabolism [3,18]. The less efficient inhibition of M2 formation than M4 formation by CYP3cide in HLM donors that did not express CYP3A5 (Fig. 6) indicates possible other enzyme contributions to M2 formation. CYP3cide inhibited nearly 100% of M4 formation in donors without CYP3A5 expression consistent with this metabolite being CYP3A specific. However, CYP3A5, when expressed, is likely important in formation of the M2 and M4 metabolites of THC.

CYP2C19 likely plays an important role in M3 formation. The correlation of M3 formation was improved when both CYP3A and CYP2C19 activities were included as predictors in multiple linear regression analysis compared to CYP3A activity alone. The CYP2C19 inhibitor (+)-N-3-benzylnirvanol inhibited up to 67% of M3 formation in high CYP2C19 activity donors. To our knowledge, this is the first evidence reported for the contribution of CYP2C19 to the formation of THC metabolites other than 11-OH-THC.

The relative contribution of individual enzymes to THC clearance are likely concentration dependent. The K_m,u for CYP3A4 is >10-fold higher than the K_m,u for CYP2C9 and CYP2C19 suggesting that CYP3A4 will likely have greater contribution to THC metabolism at higher concentrations of THC. Overall, in heavy cannabis users who may have free THC plasma concentrations that saturate CYP2C9 (>2 nM) the contribution of CYP3A and CYP2C19 to THC clearance is expected to be higher than in individuals with lower THC concentrations. CYP3A enzymes are also expected to be particularly important following oral consumption of cannabis/THC due to the high expression of CYP3A in the gut. Similarly, these data predict that in individuals with CYP2C9 polymorphisms THC will be cleared mainly by CYP3A and CYP2C19. This is important in the context of drug-drug interactions as one may expect that CYP3A and CYP2C19 inhibitors will result in decreased clearance of THC. These data suggest that CYP3A4 contribution may also be a major metabolic pathway for individuals with high CYP3A activity such as following administration of CYP3A inducers. Indeed, this aligns with in vivo findings that co-administration of rifampicin, an inducer of CYP3A, decreases THC exposure following oral dosing of THC [17].

The finding of contributions of CYP3A4 and CYP2C19 to THC clearance is in contrast to previous in vitro work [11]. This discrepancy may be due to the presence of 0.2% BSA in previously reported incubations with HLMs to address non-specific binding of THC. BSA has been shown to increase CYP2C9 activity in HLMs [25,26] and to inhibit CYP2C19 [24]. Whether BSA also alters CYP3A4 activity in HLMs is not known. However, BSA may have artificially increased the contribution of CYP2C9 to THC metabolism and 11-hydroxylation in particular. Although human serum albumin may be present in hepatocytes in vivo [24], its impact on CYP mediated metabolism is likely minor in vivo as albumin expressed in the liver is secreted to circulation via vesicle trafficking and not released to liver cytosol. In contrast, FABP1 is likely to play an important role in modulating drug binding and metabolism in human liver due to its high expression (0.7–1 mM) in the liver.

THC and 11-OH-THC as well as the endocannabinoids AEA and 2AG were found to bind to FABP1 using fluorescence DAUDA displacement assay and singular value decomposition (SVD) analysis. The binding affinities (K_d) for THC and 11-OH-THC with human FABP1 (0.3 and 0.9 μM, respectively) determined here are lower than previously reported (2.9 and 7.1 μM, respectively) [31]. The differences in the K_d values are likely due to the more rigorous analysis of FABP1-DAUDA fluorescence by SVD and more appropriate model fitting based on ternary DAUDA-FABP1-THC complex formation than what was previously reported. Based on these data, THC and 11-OH-THC are likely bound to FABP1 in the liver in vivo. However, as FABP1 can bind multiple ligands simultaneously, it is possible that THC binds to FABP1 in the liver together with endocannabinoids or other fatty acids. How such ternary binding complexes alter THC and endocannabinoid metabolism requires further study.

FABP1 had a substantial effect on THC metabolism by recombinant CYP2C9, CYP2C19 and CYP3A4. Although K_m,u and k_cat values for 11-OH-THC, M2, M3 and M4 formation by different CYPs could not be completely characterized, the data here suggest that FABP1 decreases the intrinsic clearance of THC metabolism by recombinant CYPs. This appeared to be largely due to a decrease in the k_cat by each CYP although FABP1 binding also decreased free concentrations of THC in incubations with recombinant CYPs and in HLMs consistent with the tight binding of THC to FABP1. One explanation for the apparent CYP specific effects of FABP1 on THC metabolism is direct protein–protein interaction between FABP1 and CYPs. FABP1 has been shown to directly interact with other metabolic enzymes to modulate enzyme activity. However, direct protein–protein interactions between FABP1 and carnitine palmitoyl transferase I (CPTI) enhance CPTI activity toward acyl-CoA [50] rather than decrease the activity. The results here with CYP2C9 align with the finding that FABP1 decreases the k_cat of the metabolism of diclofenac by CYP2C9 [30]. However, the effects of metabolite formation kinetics could also be due to other FABP1-lipid and FABP1-protein interactions such as interactions with P450 reductase. The mechanisms of how FABP1 alters CYP mediated metabolism require further study. Based on the findings shown here it is likely that metabolism of other CYP3A4 and CYP2C19 substrates that bind to FABP1 will also be affected.

The impact of FABP1 on THC metabolism in recombinant CYPs translated to enzyme specific effects of FABP1 on THC metabolism in HLMs (Fig. 10). However, the impact of FABP1 on THC metabolism in HLMs could partially be explained by FABP1 functioning as a binding sink of THC and altering the free concentrations of THC in the incubations. The decrease in the formation of M2 and M4 metabolites in the presence of FABP1 corresponded with the decrease in free THC in the incubations. The impact of FABP1 on M3 formation was not as large as the impact on M2 and M4, consistent with the contribution of CYP2C19 toward M3 formation and the weaker effect of FABP1 on CYP2C19 than CYP3A4 activity. The lack of decrease in 11-OH-THC formation in the HLMs in the presence of FABP1 was unexpected and indicate that the effects of FABP1 on THC metabolism by recombinant CYP2C9 may not fully replicate in HLMs. Collectively, these data suggest that the presence of FABP1 alters CYP activity and modulates the relative contribution of individual CYPs to substrate clearance. FABP1 appeared to have the largest impact on CYP3A4 activity which may lead to an increase in the relative contribution of CYP2C9 and CYP2C19 to THC clearance in vivo. The translation of these findings to in vitro-to-in vivo extrapolations is warranted to determine whether FABP1 binding should be considered in clearance predictions. Future studies could include evaluation of how FABP1 alters the overall THC depletion in in vitro systems and how that can be used in hepatic clearance predictions.

In conclusion, these data show that THC metabolism in the human liver is more complex than suggested by analysis of 11-OH-THC formation alone. The major metabolites identified in HLM appear to be a result of oxidation of THC at multiple carbons in the cyclohexenyl C-ring. Multiple CYPs, including CYP2C9, CYP2C19 and CYP3A4 contribute to THC oxidation resulting in concentration and metabolite dependent contributions of individual enzymes to THC oxidation. THC binds tightly to FABP1 and FABP1 altered THC oxidation in recombinant systems and in HLMs. Taken together these data warrant further studies regarding the mechanisms by which FABP1 impacts CYP mediated metabolism in vitro and in vivo.

Abbreviations:

2-AG

2-archidonylglycerol

AEA

anandamide

AUC

area under the plasma concentration–time curve

BCA

bicinchoninic acid

CBD

cannabidiol

CE

collision energy

CL_int

intrinsic clearance

CPTI

carnitine palmitoyltransferase I

CYP

cytochrome P450

DAUDA

11-(Dansylamino) undecanoic acid

DTT

dithiothreitol

EPI

enhanced product ion

FABP1

liver fatty acid binding protein

f_m

fraction metabolize

f_u

unbound fraction

HLM

human liver microsome

k_cat

enzyme catalytic rate

K_d

equilibrium binding affinity

K_m

Michaelis constant

K_{m, u}

Michaelis constant unbound

UHPLC-MS/MS

ultra-high performance liquid chromatography tandem mass spectrometry

LLOD

lower limit of detection

LLOQ

lower limit of quantification

LogP

octanol–water partition ratio

MDZ

midazolam

MGSB

magnetic silica beads

MPH

(S)-mephenytoin

OH

hydroxy

SJ

Saint Jude’s

SVD

singular value decomposition

TBU

tolbutamide

THC

Δ−9-tetrahydrocannabinol

UW

University of Washington

VIF

variation inflation factor

Data availability

Data will be made available on request.