We report the direct metabolism of CBN by cytochrome P450s and identify its metabolites using LC-MS/MS. The primary metabolite was identified as CBN-11-OH, along with other metabolites, including CBN-1′–OH and CBN-quinone. Using LC-MS/MS, we identified the peak of the parent CBN molecule eluting at 7.84 min. The mass fragmentation patterns of CBN (Figure A,B) are identical to the previously reported pattern. The primary metabolism products of CBN included the CBN-11-OH (Figure C,D) and CBN-1′–OH. We also detected the formation of CBN-11-OH from the metabolism of CBN by purified CYP2C9 and CYP3A4 (Figure E–H). We subjected CBN-11-OH to human liver microsomes to analyze secondary metabolism (Figure S1). We also used Bio Transformer 3.0 software which helped to predict possible biotransformed metabolites from CBN-11-OH and CBN-1′–OH (Figure S1). Most of the products involve hydroxylation (single or at multiple sites) around the fused benzene ring or at different positions in the aliphatic tail. The predicted products of identified m/z peaks eluting at 4.65 min, which correspond to the m/z of a single hydroxylation (343.19) with a single hydrogen adduct [M + H]+ and also an m/z (358.20), which corresponds to hydroxylation at two different sites [M + H – H]+ (Figure S3A,B).

CBN is partially metabolized in Phase 1 by CYPs and is further subjected to Phase II enzymes such as UDP glucuronosyltransferase (UGTs). Glucuronidation of these metabolites makes them highly hydrophilic (polar), subsequently leading to their elimination through urine. We report the glucuronidated products of CBN and CBN-11-OH (Figure S2). Our results indicate the formation of glucuronidated products of CBN with a peak at m/z 488.35 corresponding to mass with a hydrogen adduct [M + 2H]+ and a retention time of 5.09 min. We also identified the CBN-11-OH glucuronidation, with a retention time of 8.94 min and a peak at m/z 503.35 corresponding to mass with a hydrogen adduct [M + H] ⁺ (Figure S3C,D).

After we identified the major metabolites of CBN upon direct metabolism by purified CYPs and HLMs, we developed a targeted mass spectrometry method for the quantification of these metabolites using synthesized and commercially purchased standards and measured the rate of formation of the metabolites. The details are provided in the Experimental Section below under quantitation of CBN metabolites by LC/MS-MS.

Using this targeted metabolomics method, we measured CBN and its metabolites generated during in vivo metabolism, where we fed CBN to female mice. We also measured the same metabolites during CBN metabolism by HLMs. We used two routes of CBN administration, intraperitoneal (i.p) and intravenous (i.v) and extracted plasma at two different time points, 0.5 and 2 h after treatment. Our results show that the amount of CBN available through the intravenous route of administration is 3.0 times higher than that available through intraperitoneal administration (Figure ). The primary metabolite observed from CBN metabolism is 11-hydroxy CBN. CBN-1′–OH and Cannabinol p-quinone (CBN-PQ) are also formed in both routes of administration. CBN-11-OH formation is ∼ 89 times higher than CBN-1′–OH and ∼55 times higher than CBN-quinone after 0.5 h in the IV route of administration. In the i.p. route of administration, CBN-11-OH formation is ∼82-fold higher than CBN-1′–OH and ∼90-fold higher than CBN-quinone after the same time. Between 0.5 and 2 h after administration with CBN, there is a ∼ 3-fold reduction in the availability of CBN through IV and a 3-fold reduction in CBN-11-OH. CBN is reduced by 3-fold in the IP route, and CBN-11-OH is reduced by 2-fold. CBN-PQ is another metabolite that forms second to CBN-11-OH in IV-mediated administration, and there is a 9-fold reduction in the levels between 0.5 and 2 h. CBN-1′–OH is reduced by 3-fold in the same duration.

Upon comparison of CBN metabolism in plasma samples extracted from CBN-fed mice with HLM-mediated metabolism of CBN at selected concentrations, we report a similar trend in the formation of metabolites. Both the data show that CBN-11-OH is the primary product being formed, followed by CBN-1′–OH. Another interesting observation was the extremely high metabolism rates for CBN by the HLM samples, with rates of metabolite formation reaching as high as 3.4 × 10⁴ pmol/min/nmol for CBN-11-OH. Compared to this, CBN-1′–OH and CBN-quinone rates were lower. CBN-1′–OH rates across different CBN concentrations were 9–9.4 × 10³ pmol/min/nmol, and quinone was further down by approximately three times the value of CBN-1′–OH. The human liver microsome-mediated CBN metabolism data at different concentrations of CBN also show a concentration-dependent increase in CBN-1′–OH and CBN-quinone. Although they were much lower in concentration than CBN-11-OH, this also corroborates our observation of the mice’s plasma metabolism.

We employed targeted mass spectrometry to elucidate the mechanism and rate of formation of specific CBN metabolites. We measured the rate of formation of CBN-1′–OH, CBN-11-OH and CBN-quinone using three purified CYP450s, 2C9 and 2D6 in Nanodiscs and pooled HLM. In these experiments, we used purified Cytochrome P450 reductase (CPR) reconstituted with an 8:2 POPC–POPS lipid mixture and purified CYP2C9. The CBN metabolism by CYP2D6 was done in a nanodisc. The metabolism experiments were performed in the concentration range of 5–80 μM as these reflect physiological concentrations and are within the limit of CBN solubility.

The individual human cytochrome P450s were expressed and purified as discussed in the Experimental Section below. Our results show that CYP2C9-mediated metabolism of CBN led to the formation of CBN-1′–OH and CBN-quinone, and CBN-11-OH. We fit the total product rate which exhibited Michaelis–Menten kinetics. Interestingly, CBN-11-OH production, indicates atypical P450 kinetics (Figure B). Previous reports of phytocannabinoids show both Michaelis–Menten and atypical kinetics. , Among the main metabolites, CBN-11-OH shows the highest rate of formation. CBN-1′–OH shows a Vmax of 12.1 pmol/min/nmol, which is lower than CBN-11-OH, showing a Vmax of 34.2 pmol/min/nmol. Also worth noting is that although the formation rate of CBN-11-OH is higher than CBN-1′–OH, the Km of CBN-1′–OH (2.3 μM) is much lower than CBN-11-OH (51.6 μM) (Figure E) The estimated value of Vmax for CBN-quinone is 4.4 pmol/min/nmol, and Km is 2.5 μM. Our targeted metabolomics using CYP2D6 and CBN also indicated that CBN-11-OH is the major product compared to CYP2C9. We also detected the formation of CBN-quinone, although it was also at a lower rate than CBN-11-OH. Our results with CYP2D6-mediated metabolism of CBN showed that CBN-11-OH and CBN-quinone show different kinetics of formation. CBN-11-OH shows a Vmax of 6.6 pmol/min/nmol, which is higher than CBN-quinone, which shows a Vmax of 5.6 pmol/min/nmol (Figure A–E). The respective Km value for CBN-11-OH is 14.84 μM, and for CBN-quinone, it is undertermined. Surprisingly, we did not detect the formation of CBN-1′–OH with CYP2D6. We do not report significant metabolism by CYP3A4 (Figure ).

Microglial cells are macrophages in the central and peripheral nervous system that act as the first line of defense against pathogens. They respond to inflammation, cell and tissue damage, and injury to the brain, which includes the secretion of cytokines and chemokines. We have used the BV2 microglial cell line to assess the various pro (IL-6, NO and LDH) and anti-inflammatory (IL-10 and Arginase1) markers under lipopolysaccharide (LPS) stimulation in the presence and absence of CBN and its metabolites. The cell titer-blue (CTB) cell viability assay showed that cells are viable under the treatment with the molecules. LDH cytotoxicity assay showed that LDH production was reduced in cells treated with CBN and its metabolites in cells treated with LPS (Figure S7A,B). The concentration range used in the studies aligns with previous literature, which has shown that the cannabinoids are in low micromolar range levels in human plasma, 0.5 h after administration. ,, We assessed the pro-inflammatory and anti-inflammatory markers in LPS-stimulated microglial cells. In cells treated with CBN and CBN-11-OH, there is a slightly lower production of NO (pro-inflammatory marker) at 5 μM (Figure A,B). For pro-inflammatory interleukin IL-6, at a concentration of 5 μM, CBN and CBN-11-OH show a reduction in IL-6 levels. CBN-1′–OH also showed a decrease in IL-6 production (Figure C,D). In cells treated with CBN and its metabolites, there is a concentration-dependent increase in IL-10 levels in CBN and CBN-11-OH-treated cells, which is reversed in the case of cells treated with CBN-1′–OH. Still, the level of IL-10 is not higher than that of LPS-stimulated cells without any molecules (Figure E,F). Arginase 1 is an enzyme expressed in immunosuppressive environments and is reported to reduce NO levels and subsequent pro-inflammatory interleukins like IL-6. We show that there is a slight increase in Arginase1 levels when cells were treated with CBN and CBN-1′–OH, although it showed mixed results when cells were treated with CBN-11-OH (Figure S7C).

To investigate the modulation of cannabinoid receptors CB1 and CB2 in response to CBN and its metabolites CBN-1′–OH, CBN-11-OH, and CBN-quinone, we used a cell-based G protein nano BRET assay. We first employed this assay in agonist mode. Transfected cells were stimulated with 10 μM of each compound, and ΔBRET ratios were calculated as an index of receptor activation in the presence of representative Gα proteins to obtain a fingerprint of G protein activation for each receptor (Figure A,B). These experiments showed a partial agonist effect of CBN-1′–OH at the CB1 receptor when activating Gαo. Later, we obtained concentration–response curves for Gαo activation in response to the application of CBN and each of its metabolites for both CB1 and CB2 receptors (Figure C). The full agonist CP-55940 was used as a positive control. This data confirmed partial agonism of CBN-1′–OH on the CB1 receptor, detectable only at high concentrations of the ligand. The partial agonism behavior of Δ9-THC has been previously shown in our work using docking and simulation.

Next, we tested the ability of each compound to act as antagonists at CB1 and CB2 receptors. For this, we pretreated transfected cells with 1 μM of orthosteric full agonist CP-55940, we recorded the signal for 100s, then we applied increasing concentrations of each CBN metabolite and calculated the BRET ratio for an additional 100s. The concentration–response curves obtained for the CB1 receptor showed a competitive inhibition of signal, indicating that some of the CBN metabolites bind at the orthosteric site. We found a significant but partial inhibition of CB1 activation when we applied 50 μM CBN and a complete ablation of the BRET signal with 50 μM CBN-11-OH (Figure D). Similarly, we obtained concentration–response curves showing significant inhibition of the CB2 receptor by application of all the CBN compounds at 50 μM (Figure E). The curves indicate partial inhibition by CBN and CBN-quinone, complete inhibition by CBN-11-OH, and an inverse agonist effect by CBN-1′–OH at CB2, reducing the signal below the basal constitutive activity of the receptor (Figure E).

CBN is found in extracts of the plant and is the stable degradation product of Δ⁹-THC following air oxidation. Therefore, it is expected that CBN will be present in increased amounts following the aging or heating of and/or Δ⁹-THC. Despite its apparent abundance, there is limited knowledge about the structure, abundance and physiologic consequences of its metabolite(s) in the nervous system. , Importantly, a major metabolite of CBN is CBN-11-OH, reported to have a differential binding affinity to the cannabinoid receptor CB₁R than the parent compound CBN. Little is known about the physiology of CBN-11-OH or its potential action on sensory neurons. ,

While CBN has been studied for its ability to activate/inactivate sensory neurons in the pain pathway , and members of the TRP family of channels in heterologous systems (11, 32)­Physiological studies of its metabolite, CBN-11-OH, are largely absent. Given our findings that CBN-11-OH is the primary metabolite of CBN, we performed the physiologic characterization of CBN-11-OH-induced activation. [Ca²⁺]_i responses of 38.9 ± 4.3% (n = 56), which did not differ from the peak values directed by CBN-11-OH, 31.6 ± 2.8% (n = 29; ns, not significant) (Figure C). In contrast (Figure D), CBN-11-OH activated a surprisingly lower percentage of DRG neurons (5.4 ± 1.1%) as compared with CBN (29.3 ± 4.9%) (p < 0.0001). Moreover, we found that CBN-11-OH induced calcium responses more rapidly than those observed with CBN (Figure E), with CBN-11-OH directing less time to peak response (22.8 ± 2.2 s) compared with CBN (40.7 ± 4.8 s).

Unless otherwise noted, all chemicals were purchased from commercial suppliers and used as received. Acetonitrile (MeCN, HPLC grade), dichloromethane (CH₂Cl₂, HPLC grade), dichloroethane (DCE), benzene (ACS grade), and methanol (ACS grade) were used as solvents without further purification. 4-dimethylaminopyridine (DMAP), triethylamine (Et₃N), acetic anhydride (Ac₂O), selenium dioxide (SeO₂), azobis­(isobutyronitrile) (AIBN), and silver acetate (AgOAc) were purchased from commercial sources and used without further purification. Cannabinol (CBN) was purchased from commercial sources and further purified by flash column chromatography (15:1 hexane: EtOAc). N-bromosuccinimide was purchased from commercial sources and further purified by recrystallization in water. Reaction temperatures correspond to the external temperature of the reaction vessel unless otherwise noted. Analytical thin-layer chromatography (TLC) was performed on Merck silica gel 60 F254 aluminum sheets. Visualization was accomplished with UV light and/or potassium permanganate (KMnO₄). Retention factor (R _f) values reported were measured using a 10 × 2 cm TLC plate in a developing chamber containing the solvent system described. Silicycle SiliaFlash P60 (SiO2, 40–63 μm particle size, 230–400 mesh) was used for flash column chromatography. ¹H NMR spectra were obtained at 500 MHz and ¹³C NMR were obtained at 126 MHz. NMR spectra were recorded using a Bruker Avance III 500 MHz spectrometer equipped with BB CryoProbe or Varian/Agilent VNMRS 750 MHz Narrow Bore and were referenced to residual chloroform (7.26 ppm, ¹H) and solvent chloroform-d (77.16, ¹³C). Chemical shifts are reported in parts per million (ppm), and multiplicities are indicated as s (singlet), d (doublet), t (triplet), q (quartet), p (pentet), m (multiplet), and br (broad). Coupling constants, J, are reported in Hertz. The University of Illinois Mass Spectrometry Laboratory performed mass spectrometry (MS). Electron Impact (EI+) spectra were performed at 70 eV using methane as the carrier gas, with a time-of-flight (TOF) mass analyzer. Electrospray ionization (ES+) spectra were performed using a time-of-flight (TOF) mass analyzer. Data are reported in the form of m/z. Infrared (IR) spectra were measured neat on a PerkinElmer Spectrum Two FT-IR ATR spectrometer. Peaks are reported in cm^–1 with indicated relative intensities: s (strong, 0–33% T); m (medium, 34–66% T); w (weak, 67–100% T); and br (broad). Both CBN-1′–OH and CBN p-quinone, which were synthesized, were validated for their purity using HPLC analysis. Both of these compounds are reported to be more than 95% pure, as shown in our HPLC traces in the Supporting Information (SI). CBN and CBN-11-OH of purity greater than 98% were purchased from Cayman for other experiments.

In the current manuscript, we have used three Cytochrome P450s, CYP3A4, CYP2C9, and CYP2D6. We expressed and purified both 2C9 and 3A4 as described previously The CYP2D6 construct was a gracious donation of Eric Johnson. The construct of CYP2D6 was also in the pcWori vector with modified and truncated N-terminal residues of MAKKTSSKGKL. We mention the procedures briefly here.

The CYP3A4 gene was in an NF-14 construct in the pcWori+ vector with ampicillin resistance and a C-terminal polyhistidine tag (Guengerich construct). The primary growth was done in Luria Bertini (LB) broth supplemented with Ampicillin (100 μg/mL). The secondary growth of the DH5alpha cells with the CYP3A4 gene was done in a total of 3 L in Terrific Broth media. The growth medium was supplemented with 0.1 M KPi buffer, 1 mM thiamine, ampicillin, and trace minerals. After the O.D. reached 0.4, we supplemented the growth media with arabinose and δ-ALA. The final expression was induced by adding 1 mM IPTG once the O.D. had reached between 0.8 and 1.0 in all the flasks. The protein growth was done for 48 h, and the cells were harvested by spinning at 8000 rpm for 15 min at 4 °C and stored at −80 °C. At the start of the purification process, the pellet was dissolved in a lysis buffer containing 20% glycerol (v/v), 0.1 M KPi, 500 mM Sodium Acetate and 5 mM and Beta mercaptoethanol (βME). We used 1 mM PMSF to inhibit proteases and sonicated the dissolved suspension on ice for a minimum of 4 cycles (60 s ON and 60 s OFF). The amplitude was set at 60%. The sonicated lysate was then subjected to ultracentrifugation at 30,000 rpm at 4 °C for 37 min. This step isolated the membrane fraction pellet, which was further dissolved in a second lysis buffer supplemented freshly with 1% cholate as a detergent. Once homogenized, the suspension was again subjected to sonication and ultracentrifuged using the same parameters mentioned above. Finally, the supernatant was loaded onto a Ni-NTA column and purified using a washing step supplemented with 15 mM imidazole, and a subsequent elution was done using 300 mM imidazole. The eluted protein was then further concentrated using 30 kDa amicons and buffer exchanged, removing imidazole.

CYP2C9 was also purified by similar methods with some modifications. Briefly, the protein growth and expression were done at 29 °C, 190 rpm for 48 h, and the cells were harvested at 4000g for 15 min. The pellet was resuspended in a suspension buffer (300 mL) consisting of 20% glycerol (v/v), 0.02 M KPi, 10 mM and β-mercaptoethanol (βME) and 1 mM PMSF. Once the pellet was completely resuspended, Lysozyme (0.3 mg/mL) was added to the buffer and further stirred for another 30 min at 4 °C. Spheroplasts were centrifuged at 5460 g for 30 min in the next step. The spheroplasts were stored at −80 °C before further processing. The spheroplasts were resuspended in 150 mL of buffer consisting of 500 KPi, 20% (v/v) Glycerol, 10 mM BME, 0.5 mM PMSF, Deoxyribonuclease (10 μg/mL), Ribonuclease A (10 μg/mL) and 0.1 mM Ibuprofen followed by sonication (40 s ON and 60 s OFF) for 6 cycles. The sonicated lysate was stirred at 4 °C for 90 min and then ultracentrifuged at 26000 rpm for 1 h. The composition of the column buffer used for equilibrating the Ni-NTA column and subsequent purification steps was 100 mM Potassium Phosphate, 20% (v/v) glycerol, 0.5% CHAPS and 1 mM BME. The washing step was done using the column buffer with the addition of 100 mM NaCl, 0.5 mM PMSF and 5 mM Histidine. The final elution was done using 10 mM KPi, 20% (v/v) glycerol, 0.5% CHAPS, one mM BME, 100 mM NaCl, 0.5 mM PMSF, and 50 mM Histidine (30 mL).

CYP2D6 was also purified by similar methods with some modifications. Briefly, the protein growth and expression were done at 30 °C, 190 rpm for 48 h, and the cells were harvested at 8000 rpm for 15 min. The pellet was resuspended in a Lysozyme buffer (300 mL) consisting of Tris HCl, pH 8, Sucrose (0.25 M), EDTA (0.25 mM) and freshly added Lysozyme (0.02 mg/mL). We pelted the spheroplast at 4000 rpm for 30 min. The spheroplasts were then resuspended in another buffer. The buffer composition included Glycerol (20%), 0.5 M KPi, 6 mM MgCl2, 5 mM βME, 0.2 mM PMSF, 1% Cholate and 0.1 mM Thioridazine followed by sonication (40 s ON and 40 s OFF) for 6 cycles. The sonicated lysate was then ultracentrifuged at 30000 rpm for 45 min. The composition of the column buffer used for equilibrating the Ni-NTA column and subsequent purification steps was 100 mM Potassium Phosphate, 20% (v/v) glycerol, 0.1% Cholate, 0.5 mM βME and 0.05 mM Thioridazine. The first washing step was done using the column buffer, adding 5 mM ATP, 10 mM MgCl2 and 150 mM KCl to remove the Gro complex. The second washing step used the column buffer base with 0.04 M l-histidine. The final elution was done using 0.1 M l-Histidine and 1 M NaCl in an elution volume of 100 mL.

Once the CYP proteins were purified, we aimed to incorporate them into lipid nanodiscs. The composition of the lipid nanodiscs included 100% POPC with purified membrane scaffold protein (MSP). We optimized the lipid-to-MSP ratio to 1:140 for the preparation of the nanodiscs. The method is described briefly here. The commercially available lipids were solubilized in CHCl₃ to the required mol % and stored at −80 °C. To prepare the discs, we aliquoted out the required amount in 18 mm x100 mm test tubes and dried the lipids down consecutively under a stream of nitrogen. Once dried, the test tubes were stored in a desiccator overnight in a vacuum. The following morning, we solubilized the lipids using 200 mM cholate solution, warming the tubes in warm water. This step was followed by the sonication of the tubes for 15 min. After the sonication, we added everything, including MSP, glycerol, water and 0.1 M KPi and incubated the mixture for 45 min in a gentle rocker at 4 °C. Following the incubation, we added the calculated amount of CYP 30 and let it incubate for 45 min more. The Biobeads were added after the incubation and gently rocked at 4 °C for 7 h for CYP2C9 and CYP3A4. The incubation with biobeads was close to 12 h with CYP2D6. The more prolonged incubation with the Biobeads ensured the complete removal of the detergents. Once the incubation was over, we washed the Biobeads with 0.1 M KPi and filtered the mixture by spinning at 3000 rpm for 5 min. The eluent was concentrated down to the required amounts. The final concentrated nanodisc preparation was purified using size-exclusion chromatography using the AKTA Go FPLC with a Superdex 200 10/300 GL column in 0.1 M KPi, pH 7.4. The methods and the lipid ratio have been used for other CYPs.

The expression and purification of CYP450 reductase (CPR) were done using the previously described method.

We carried out concentration-dependent metabolism of CBN in the presence of 2 CYPs: 2C9 and 2D6. For the experiment, we used an enzyme concentration of 0.5 μM, reconstituted in the lipid system of POPC: POPS (8:2 v/v), and CPR at 0.6 μM dissolved in 0.1 M potassium phosphate buffer for CYP2C9. The substrate CBN was used in a range of 5–80 μM. The reaction mixture was incubated at 37 °C for 10 min, and then NADPH was added (1 mM), allowing the reaction to proceed for an additional 30 min. We quenched the reactions using a mixture of Hexane: Ethyl Acetate (8:2 v/v) and used the same composition to extract the CBN metabolites. We did a total extraction for three cycles. Following the extractions, the extractions were dried under a continuous stream of nitrogen, and the residual products were finally dissolved in 90% ethanol. We used CYP2D6 in a nanodisc for the metabolism of CBN. Since the protein was in a nanodisc, we did not add the lipid mixture separately, but the rest of the components were the same. We carried out the same protocol for the metabolism of CBN using 1 μg of human liver microsome, CPR (0.6 μM) and NADPH (1 mM).

Initially dissolved in ethanol, CBN was combined with a volume of Tween 80, constituting 6% of the total solution volume for injection. Ethanol was eliminated via vacuum evaporation, and saline was added to prepare a 4 mg/mL solution. CBN (20 mg/kg) or vehicle was administered to C57BL/6 mice intravenously (IV) through the retro-orbital sinus or intraperitoneally (IP) at the lower right quadrant of the abdomen. Animals were euthanized at 30- or 120 min postinjection. Blood samples, collected via cardiac puncture into heparin tubes, underwent cold centrifugation at 4500 rpm for 15 min to isolate plasma. Plasma and whole liver samples were flash-frozen at −80 °C for subsequent analysis. Our decision to use female mice for this experiment was based on the availability of age-matched mice in the colony at the time of the experiment.

Studies utilizing mice were performed under a protocol approved by the University of California, San Francisco’s Institutional Animal Care and Use Committee (IACUC) and in accordance with the ACS Ethical Guidelines for Publication of Chemical Research (Ethics Approval Number: AN199674-00C and AN200700-00).

Mice were anaesthetized with a 3% inhalant dose of isoflurane until their respiratory rate decreased to 50% of baseline (i.e., 40–60 breaths/min) and the righting reflex was abolished. Mice were subsequently removed from the induction chamber and laid on their side. Using the index finger and thumb of the nondominant hand, the skin above and below the eye was retracted until a slight protrusion from the socket was noted. A 29G1/2 needle was inserted, bevel down, into the medial canthus at an approximate 45-degree angle until slight resistance was noted. Volume was slowly infused at a rate of 15uL/second. The needle was subsequently removed from the eye socket, and ophthalmic ointment was added.

Animals were gently restrained by scruff, and the injection site at the lower right quadrant of the abdomen was identified. The needle was inserted at a 60-degree angle to the body, and the plunger was lightly withdrawn to ensure the needle was not in contact with viscera or organs before dispensing volume.

200 μL plasma samples were collected and used for extraction using 1 mL of Hexane: Ethyl acetate (8:2 v/v) following a similar protocol described above.

Samples were analyzed using the Sciex QTrap 6500+ system (Framingham, MA) with Waters Acquity I- class plus UPLC. Software Analyst 1.7.1 was used for data acquisition.

The LC separation was performed on a Waters Acquity UPLC BEH column (2.1 × 100 mm, 1.7 μm) with mobile phase A (0.1% formic acid in water) and mobile phase B (0.1% formic acid in acetonitrile). The flow rate was 0.5 mL/min unless noted otherwise. The linear gradient was as follows: 0–0.5 min, 70%A; 2 min, 20%A; 4–4.2 min, 0%A; 4.21–5.2 min, 0%A (0.8 mL/min); 5.3–6.9 min, 70%A. The autosampler was set at 10 °C, and the column was kept at 50 °C. The injection volume was 5 μL.

Mass spectra were acquired under positive electrospray ionization (ESI) with the ion spray voltage of 4500 V. The source temperature was 500 °C. The curtain gas, ion source gas 1, and ion source gas 2 were 32, 50, and 70 psi, respectively. Multiple reaction monitoring (MRM) was used for quantitation with the MRM transition of CBN (m/z 311.2 → m/z 269.1), CBN-1′–OH (m/z 327.2 → m/z 253.1), CBN-11-OH (m/z 327.2 → m/z 128.1), CBN-p-quinone (m/z 325.2 → m/z 283.1) and EPEA-d4 as internal standard (m/z 350.3 → m/z 66.1).

Samples were analyzed with the Sciex QTrap 6500+ system (Framingham, MA) with Waters Acquity I-class plus UPLC. Software Analyst 1.7.1 was used for data acquisition.

The LC separation was performed on a Waters Acquity UPLC BEH column (2.1 × 100 mm, 1.7 μM) with mobile phase A (0.1% formic acid in water) and mobile phase B (0.1% formic acid in acetonitrile). The flow rate was 0.5 mL/min unless noted otherwise. The linear gradient was as follows: 0–0.5 min, 70%A; 2 min, 20%A; 4–4.2 min, 0%A; 4.21–5.2 min, 0%A (0.8 mL/min); 5.3–6.9 min, 70%A. The autosampler was set at 10 °C, and the column was kept at 50 °C. The injection volume was 5 μL.

Mass spectra were acquired under positive electrospray ionization (ESI) with the ion spray voltage of 4500 V. The source temperature was 500 °C. The curtain gas, ion source gas 1, and ion source gas 2 were 32, 50, and 70 psi, respectively. Multiple reaction monitoring (MRM) was used for quantitation, as described in transitions CBN (m/z 311.2 → m/z 205.1), CBN-1′–OH (m/z 327.2 → m/z 253.1), CBN-11-OH (m/z 327.2 → m/z 291.1), CBN-p-quinone (m/z 325.2 → m/z 283.1) and CBC-d9 as internal standard (m/z 324.2→ m/z 202.1).

All data were analyzed using the software Skyline (daily version 22.2.1.278). It includes raw data import, peak integration, and a linear regression fit with 1/×2 weighting for the calibration curves. QCs and standards have good signals and linear responses in the calibrated concentration ranges.

The untargeted mass spectrometry to determine the primary metabolite products were done using an Acquity UPLC BEH with a C18 column (100 mm × 2.1 mm, 1.7 μm). Mobile A phase consists of water and 0.1% formic acid, and mobile phase B consist of 20% acetonitrile (ACN), 80% isopropyl alcohol (IPA) and 0.1% formic acid. The column temperature was 60 °C, and the injection volume was 2 μL. The scan range for mass spectrometry was 150–2000 m/z. The analysis of raw files was done using the Thermo Fisher Xcalibur software.

The initial goal was to investigate the relative cytotoxicity of the molecules. To do this, we performed a Lactate Dehydrogenase assay (LDH) assay, monitoring the LDH release into the medium upon treatment of the metabolites at a concentration range between 1 and 10 μM. The assay was performed using the LDH Assay kit from Cayman Chemicals. The BV2 microglial cells were cultured in DMEM supplemented with FBS and 1× P/S. 50,000 cells were plated in a 96-well plate and grown for 24 h. After 24 h, the cells were treated with various molecules in triplicate across the specified concentration range. The assay was performed 24 h after the cells were treated with the molecules.

The BV2 microglial cells were subjected to the treatment of LPS both in the absence and presence of CBN and its selected metabolites. The seeding density for this assay was 1 × 10⁵ cells/well. The cells were initially treated with the molecules for 4 h before being stimulated with LPS (25 ng/mL). The NO assay was performed 24 h after the treatment with LPS. The NO assay was performed according to the manufacturer’s protocol, and NO levels were quantified using nitrate standard curves. For the assay, 50 μL of cell media was used in a 96-well plate, and the absorbance was recorded at 540 nm. All measurements were performed in n = 5.

Following the manufacturer’s protocol, the same supernatant was used to assess LDH levels using the LDH assay kit.

ELISA assays of IL-6, IL-10 (Item No. 88–7105), and Arginase 1 were performed using the same supernatant with which we had completed the NO and LDH assays. The culture media was collected, and the IL-6 (Thermo Scientific) and Arginase 1 (abcam) assays were performed using the specific monoclonal antibodies as per the manufacturer’s instructions.

We purchased CP-55,940 from Cayman Chemicals (Item No. 13241). Codon-optimized sequences encoding for CB1R, CB2R, GPR55, and GPR119 were a kind gift from Dr. Bryan Roth (University of North Carolina, NC) (Add gene, PRESTO-Tango Kit- #1000000068). GPCR coding sequences were subcloned by In-Fusion HD Cloning (Clontech) in a pcDNA3.1 vector, removing the V2-tail, TEV domain, and tTA sequences. All constructs were verified by Sanger sequencing.

We purchased HEK293T/17 cells (RRID: CVCL_1926) from ATCC. The cells were cultured in DMEM (Gibco, 10567-014) supplemented with 10% FBS dialyzed (Biowest, S181D), nonessential amino acids (Gibco, 11140-050), penicillin 100 units/mL and streptomycin 100 μg/mL (Gibco, 15140-122), and amphotericin B 250 μg/mL (Thermo Fisher, 15290-018) at 37 °C and 5% CO₂. The cells were routinely monitored for possible mycoplasma contamination. We seeded two million cells in each well of 6-well plates in medium without antibiotics for 4 h and then transfected with a 1:3 ratio of total DNA plasmids (2.5 μg) and polyethylenimine (PEI; 7.5 μL) (Polysciences, 23966). We kept a 4:1 ratio between transfected Gα proteins and Gβγ proteins. Accordingly, the following amount of each DNA plasmid was used for transfection: 13 ng of masGRK3CT-Nluc, 208 ng of β1-Venus (156-239), 208 ng of γ2-Venus (1-155), 833 ng for each Gα protein, 208 ng of receptor, 208 ng for chaperones RIC8B (for Gs) or RIC8A (for Gq, G11). The empty vector pcDNA3.1 was used to normalize the ratio of transfected plasmids. We transiently transfected cells and incubated them for 16 h. Then, we starved cells in Optimem (Gibco, #11-058-021) for 6 h before being tested.

We collected the serum-starved cells in 1.5 mL tubes and spun them at 500g for 5 min at room temperature. The cell pellet was resuspended in 250 μL of BRET buffer (PBS supplemented with 0.5 mM MgCl₂ and 0.1% glucose). Thirty μL cells were plated in each well of white 96-well microplates (Greiner Bio-One). The nanoluc substrate used in the experiments is NanoGlo (Promega, N1120), diluted 1:250 in BRET buffer, and 30 μL was applied to plated cells. The BRET measurements were obtained using a POLARstar Omega microplate reader (BMG Labtech). All the measurements were obtained at room temperature. The BRET signal was determined as the ratio of the light emitted by Gβ1γ2-venus (emission filter 535/30) to the light emitted by masGRK3CT-Nluc (475/30). For (Figure A,B) experiments, NanoBRET procedures were performed as previously. Briefly, all cells were plated and mixed with NanoGlo. Initial readings were used to establish a basal BRET ratio, and then 60 μL of 2x agonist was added. Emission intensity was recorded for 5 min. ΔBRET was obtained by subtracting the basal BRET ratio from the maximal BRET ratio obtained after vehicle/agonist application. The ΔBRET obtained on pcDNA3.1 (empty vector) transfected cells was subtracted from the receptor conditions.

For the antagonist mode (Figure C,D), 30 μL of cells were combined with 30 μL of NanoGlo (1:250) and read for a baseline of 2 min. Then, we added 60 μL of 2× agonist (CP 55,940) (final concentration 1 μM), and we recorded until the BRET ratio reached a plateau. Then, we applied 60 μL of vehicle, CBN, and the metabolites from 0.01 to 50 μM. Measurements continued for 100 s. ΔBRET is obtained, as previously mentioned, and normalized to 100% of the maximal BRET ratio obtained upon agonist application.

We obtained the full-length CYP2C9 structure from the AlphaFold Protein Structure Database and determined its membrane orientation using the Orientations of Proteins in Membranes (OPM). The OPM-oriented structure was loaded into the CHARMM-GUI Membrane to construct a membrane-bound system embedding CYP2C9 in a bilayer composed entirely of 1-palmitoyl-2-oleoylphosphatidylcholine (POPC), which represents the primary lipid type of the endoplasmic reticulum membrane. The protein termini were capped with an acetylated amino terminus (ACE) and an N-methylamide carboxy terminus (CT3). We integrated the missing heme group from the CYP2C9 crystal structure (PDB ID: 1R9O), along with the associated water molecule coordinating the heme, using the PSFGEN plugin of Visual Molecular Dynamics. Protonation states of key residues were estimated using pK _a calculations in Maestro’s PropKa tool and modeled via PSFGEN. The system was solvated with the SOLVATE plugin and neutralized with 0.15 M NaCl using VMD’s AUTOIONIZE plugin. Subsequently, the membrane-bound CYP2C9 system was simulated for 1000 ns, with further details on simulation parameters provided below.

To characterize the putative binding modes of cannabinoids (CBN, THC) to CYP2C9, we performed molecular induced-fit docking using the Molecular Operating Environment (MOE) software (2024.06, Chemical Computing Group ULC, Montreal, QC). We used the CYP2C9 structure from PDB ID: 1R9O and completed missing segments (the N-terminal transmembrane helix, residues 38-VIGNI-42, and 214-QICNNFS-220) using MOE’s homology modeling tool. To guide this, we aligned a snapshot from the end of our membrane-bound CYP2C9 simulation with the PDB structure (1R9O). In the homology model tool, we selected 1R9O as the template structure, specified the membrane-bound CYP2C9 regions to override the missing regions, and set the environment to include the ligand and water molecules. These settings retained most of the protein from the 1R9O crystal structure while reconstructing only the missing sections using the AlphaFold model incorporated in our membrane-bound system.

Cannabinoid structures were obtained from the PubChem database and prepared in MOE using the ‘conformation import’ workflow to generate multiple low-energy conformations. These conformations were docked to the active site of CYP2C9, defined by the location of the crystal ligand, flurbiprofen in PDB: 1R9O. For induced-fit docking in MOE, we used the following settings: the placement method was ‘triangle matcher,’ the refinement method was ‘induced fit’ (targeting residues within 6 Å of the cannabinoid), and the scoring function was GBVI/WSA dG with the AMBER10 force field. We saved the top 20 docking poses for each cannabinoid. The GBVI/WSA dG scoring function was selected as it was the only one able to capture the crystal position of flurbiprofen accurately. A summary of the docking scores for each cannabinoid is presented in Table S1. Since the top 20 docked poses of each cannabinoid displayed minimal variation in docking scores (Table S1), we applied additional metrics to further explain the Results section to select the best pose for subsequent simulations. The hydrogen bond occupancy of CBN and THC is shown in Table S2.

The selected CYP2C9-cannabinoid complexes were embedded in a pure POPC membrane using the final frame of the previously simulated membrane-bound CYP2C9 AlphaFold model. The systems were then solvated with VMD’s SOLVATE plugin and neutralized to 0.15 M NaCl using the AUTOIONIZE plugin. Each membrane-bound CYP2C9 system, with its docked cannabinoid, was subsequently simulated for 1000 ns, with additional details of the simulation protocol provided below.

The interaction probability of residues with cannabinoids was calculated using a distance metric to define an interaction. Specifically, an interaction was defined as any heavy atom of a residue being within 4 Å of any heavy atom of a cannabinoid. The number of interacting residues was counted at each nanosecond of the simulation, and the total interactions for each residue were normalized by the total simulation time of 1000 ns. Thus, an interaction probability of 1.0, as shown in Figure S8E,F, indicates that the residue remained within 4 Å of the cannabinoid throughout the entire simulation. Interaction energies between key interacting residues and each cannabinoid were calculated throughout the simulations using the NAMD Energy plugin in VMD.

To estimate binding free energies, we employed the Prime MM-GBSA module in Schrödinger’s Maestro (Schrödinger Release 2024-4: Maestro, Schrödinger, LLC, New York, NY, 2024.), using the VSGB solvation model and OPLS4 force field. Residues within 4 Å of the ligand were treated as flexible, with the sampling method set to ‘minimize.’

The membrane-bound CYP2C9 AlphaFold model, along with each CYP2C9-cannabinoid complex, was simulated using NAMD3 , with CHARMM36m parameters for the protein, CHARMM36 parameters for lipids, and the TIP3P water model, applying a 2 fs time step. Cannabinoid parameters were obtained from the CHARMM General Force Field (CgenFF) Web server. Each system underwent energy minimization for 10,000 steps, followed by a four-step equilibration process with gradually relaxed restraints to facilitate the cannabinoids’ accommodation within the CYP2C9 active site.

During equilibration, lipid tails were initially allowed to melt while everything except waters and ions were restrained with a 1 kcal mol^–1Å^–2 force constant for 1 ns. Then, all lipid atoms, waters and ions were allowed to move while protein, heme, ligand, and heme-coordinated water oxygens were restrained with a 1 kcal mol¹ å² force constant for 2 ns. Subsequently, protein backbone, heme, and cannabinoid-heavy atoms were restrained at 0.5 kcal mol^–1Å^–2 for an additional 2 ns. Finally, Cα atoms and cannabinoid ring systems were restrained with force constants of 0.5 and 0.2 kcal mol^–1Å^–2, respectively, for 4 ns. Each equilibrated system was then subjected to a 1000 ns production simulation without restraints. The temperature was maintained at 310 K using the Langevin thermostat, and pressure at 1 bar using the Nosé-Hoover piston method. Long-range electrostatics were calculated with the particle mesh Ewald (PME) method using a 1 Å grid spacing, while nonbonded forces employed a 12 Å cutoff and a 10 Å switching distance. Simulations were visualized and analyzed using VMD.

All statistical analyses have been done using GraphPad Prism version 10.2.3 for Windows, GraphPad Software, Boston, Massachusetts, USA. The results are represented as mean ± SEM. We used the one-way ANOVA followed by Dunnett’s multiple comparison tests to calculate statistical significance.