Cannabis constituents interact at the drug efflux pump BCRP to markedly increase plasma cannabidiolic acid concentrations

Cannabis is a complex mixture of hundreds of bioactive molecules. This provides the potential for pharmacological interactions between cannabis constituents, a phenomenon referred to as “the entourage effect” by the medicinal cannabis community. We hypothesize that pharmacokinetic interactions between cannabis constituents could substantially alter systemic cannabinoid concentrations. To address this hypothesis we compared pharmacokinetic parameters of cannabinoids administered orally in a cannabis extract to those administered as individual cannabinoids at equivalent doses in mice. Astonishingly, plasma cannabidiolic acid (CBDA) concentrations were 14-times higher following administration in the cannabis extract than when administered as a single molecule. In vitro transwell assays identified CBDA as a substrate of the drug efflux transporter breast cancer resistance protein (BCRP), and that cannabigerol and Δ9-tetrahydrocannabinol inhibited the BCRP-mediated transport of CBDA. Such a cannabinoid-cannabinoid interaction at BCRP transporters located in the intestine would inhibit efflux of CBDA, thus resulting in increased plasma concentrations. Our results suggest that cannabis extracts provide a natural vehicle to substantially enhance plasma CBDA concentrations. Moreover, CBDA might have a more significant contribution to the pharmacological effects of orally administered cannabis extracts than previously thought.

www.nature.com/scientificreports/ Cannabis-based products have gained widespread media attention over the last decade due to artisanal CBD-dominant extracts being reported to have remarkable anticonvulsant effects in children with intractable epilepsies 10 . More recently there has been a "CBD craze", with a substantial increase in demand for cannabisbased products which are perceived to treat a myriad of health conditions 11 . These products, which contain a multitude of cannabinoids, are administered at much lower doses than purified forms of CBD and Δ 9 -THC that have been shown to be effective in clinical trials 10,11 . Consistent with the "entourage effect" hypothesis it has been suggested that pharmacodynamic interactions between phytochemicals in cannabis occur due to a concerted action at an individual drug target or via activating complementary pathways. However, an "entourage effect" could also arise from pharmacokinetic interactions between components in medicinal cannabis, whereby the absorption, distribution, metabolism and excretion of the cannabinoids are affected. Indeed, pharmacokinetic interactions have been observed between cannabinoids with co-administration resulting in increased cannabinoid concentrations in tissues and blood 8,12,13 .
In the present study we aimed to explore the potential for pharmacokinetic interactions between cannabinoids within a full-spectrum cannabis extract administered orally. Oral administration is an increasingly preferred mode of delivery of cannabis oils and is the dominant mode of delivery for childhood epilepsy patients 10,11,14 . We compared the pharmacokinetic parameters of cannabinoids administered as an extract to those when administered as an individual compound at equivalent doses.

Results
The pharmacokinetic profiles of various cannabinoids administered in a full-spectrum cannabis extract differ substantially from cannabinoids administered as single molecules at equivalent doses. The cannabinoid profile of the full-spectrum cannabis extract was diverse, containing the greatest quantities of cannabidiolic acid (CBDA), Δ 9 -tetrahydrocannabinolic acid (Δ 9 -THCA), CBD and Δ 9 -THC (Fig. 1a). To infer whether compound-compound interactions alter the pharmacokinetic profile of the cannabinoids in the full-spectrum extract, we compared the profiles of the cannabinoids administered in a full-spectrum extract to those of the cannabinoids administered as individual components (Fig. 1). The full-spectrum extract was administered orally as a single dose and plasma cannabinoid concentrations were quantified. CBC, cannabidivarin (CBDV), cannabigerol (CBG), cannabinol (CBN) and Δ 9 -tetrahydrocannabivarin (Δ 9 -THCV) were not detected in plasma following oral administration of the full-spectrum extract so no further pharmacokinetic characterization of these cannabinoids was conducted.
Six cannabinoids were detected in plasma following oral administration of the full-spectrum extract: CBD, CBDA, cannabidivarinic acid (CBDVA), cannabigerolic acid (CBGA), Δ 9 -THC and Δ 9 -THCA. Each was then administered orally as an individual compound at an equivalent dose to that found in the full-spectrum extract.
Astonishingly, the plasma CBDA concentrations that were observed following administration of the fullspectrum cannabis extract were substantially higher than that observed when CBDA was administered as a single molecule at an equivalent dose (Fig. 1b). Accordingly, the plasma C max value of CBDA within the full-spectrum extract (47 ± 4 µg/mL) was substantially higher than the C max value achieved as a single molecule (6 ± 1 µg/mL). Moreover, total exposure of CBDA as determined by AUC values when administered in a full-spectrum extract was nearly 14 × the exposure that was observed following its administration as an individual compound (Fig. 1b, Table 1). Conversely, the plasma concentrations of CBD, CBDVA, CBGA, Δ 9 -THC and Δ 9 -THCA following administration of the full-spectrum extract were substantially lower when they were administered as single molecules at equivalent doses; the total plasma exposure of each was nearly 2-4 × lower following administration in the full-spectrum extract to when administered individually ( Fig. 1c-g, Table 1).
Absorption of the cannabinoids into plasma following oral administration of the full-spectrum extract was slow with t max values of 45-60 min (Table 1). While CBD, CBDA, CBGA and Δ 9 -THC were all maximally absorbed (t max ) by 60 min when administered as a full-spectrum extract, plasma t max values were delayed (90-120 min) when each were administered individually ( Fig. 1, Table 1). In contrast, absorption of CBDVA (t max 15 min) and Δ 9 -THCA (t max 30 min) was more rapid as individual cannabinoids compared to within the full-spectrum extract (t max values 60 and 45 min, respectively).
Overall, the differing pharmacokinetic parameters for the cannabinoids when administered in a full-spectrum extract compared to as individual compounds indicate pharmacokinetic interactions might be occurring between cannabinoids within the full-spectrum extract.
CBDA, CBD, CBDVA, CBG and Δ 9 -THC are BCRP substrates. Drug transporters, including ATPbinding cassette (ABC) transporters, facilitate the movement of substrates across biological membranes and transporter-mediated interactions within the intestinal lumen can profoundly affect oral bioavailability of coadministered drugs. The best characterised ABC transporters, P-glycoprotein and breast cancer resistance protein (BCRP), are located on the apical surface of epithelial cells in the intestine and extrude substrates back into the intestinal lumen, thereby limiting systemic absorption. Cannabinoids are both substrates and/or inhibitors of ABC transporters so we aimed to examine whether the converging action of the cannabinoids on these transporters might provide a mechanism for the pharmacokinetic interaction observed here [15][16][17][18][19] . www.nature.com/scientificreports/ We first aimed to determine whether the cannabinoids found in the full-spectrum cannabis extract were substrates of P-glycoprotein and BCRP by using MDCK cells expressing human P-glycoprotein or BCRP in vitro. Transwell assays were conducted to assess bidirectional transport of the cannabinoids detected in plasma following administration of the full-spectrum cannabis extract (CBD, CBDA, CBDVA, CBGA, Δ 9 -THC and Δ 9 -THCA), including the respective neutral compounds (CBDV and CBG) across wildtype, BCRP and P-glycoprotein MDCK cell monolayers. Permeability in the basolateral to apical (B > A) and apical to basolateral (A > B) directions were determined for each of the transporters and compared to respective permeabilities in the wildtype control cells. BCRP and P-glycoprotein preferentially transport substrates in the B > A direction.
CBD, CBDA and CBDVA were BCRP substrates, as the cell permeabilities of these compounds in the B > A direction was significantly greater in the BCRP overexpressing cells than in wildtype cells (CBD, p = 0.0105; CBDA, p = 0.0002; CBDVA, p = 0.0028) without impacting the A > B direction ( Fig. 2a-c, Table 2). Moreover, the efflux ratios (r) calculated for each cannabinoid further support the characterization of CBD, CBDA and CBDVA as BCRP substrates, as they exceeded the generally accepted transport ratio threshold of 1.5 for BCRP substrates 20,21 . Additionally, the BCRP inhibitor elacridar (10 µM) significantly inhibited the transport of these three cannabinoid substrates (CBD, p = 0.0249; CBDA, p < 0.0001; CBDVA, p < 0.0001; Fig. 2a-c).
CBG and Δ 9 -THC were also weak substrates of BCRP (Fig. 2d,e Table 2). BCRP permeabilities in the B > A direction were significantly greater than those of wildtype cells (CBG, p = 0.0002 and Δ 9 -THC, p = 0.0183; Fig. 2, Table 2). Because permeabilities in the A > B direction for cells expressing BCRP were also significantly greater than for wildtype cells (CBG, p = 0.0018 and Δ 9 -THC, p = 0.0125) the transport ratios for CBG (1.1 ± 0.3) and Δ 9 -THC (1.4 ± 0.5) were below the threshold for BCRP substrates. However, because BCRP-mediated transport of both CBG and Δ 9 -THC was significantly inhibited by elacridar (CBG, p = 0.0118 and Δ 9 -THC, p = 0.0361) these cannabinoids were deemed weak BCRP substrates (Fig. 2d,e). The B > A directional permeability of CBDV in cells expressing BCRP was significantly greater that that of wildtype cells (p = 0.0001) suggesting CBDV might be a BCRP substrate; however, CBDV transport was not inhibited by elacridar so CBDV was not considered a substrate of BCRP (Table 2).
CBD, CBDA, CBDVA, CBG and Δ 9 -THC were not P-glycoprotein substrates, with transport ratios < 2.5, which is the accepted threshold for P-glycoprotein substrates ( Table 2) 20 . CBGA, CBDV and Δ 9 -THCA were not substrates of either BCRP or P-glycoprotein. Transport ratios for CBGA and Δ 9 -THCA could not be calculated in wildtype or BCRP-expressing cells since rates of transport were not significantly different from zero. While transport of CBGA and Δ 9 -THCA was achieved in cells expressing P-glycoprotein, it was minimal and transport ratios were 0.7 ± 0.5 and 0.5 ± 0.4, respectively (Table 2). CBG and Δ 9 -THC inhibit BCRP-mediated transport of CBDA. Since CBDA was identified as a BCRP substrate, it is possible that cannabinoids within the full-spectrum extract inhibited BCRP-mediated efflux of CBDA in the intestinal lumen, which would enhance plasma CBDA exposure following oral dosing with the full-spectrum extract. Hence, we investigated whether the cannabinoids identified as BCRP substrates (CBD, CBDVA, CBG and Δ 9 -THC) inhibited BCRP-mediated transport of CBDA, as substrates may competitively inhibit the transport of other substrates. Rates of CBDA transport in both the B > A and A > B directions were significantly inhibited by 10 µM CBG and Δ 9 -THC (B > A: p = 0.0015 and p = 0.0131, respectively; A > B: p < 0.0001 and p = 0.0007, respectively) resulting in lower transport ratios (Fig. 3, Table 3). Neither CBD nor CBDVA affected CBDA transport via BCRP. CBDA permeability was also examined in the presence of a mixture of all four cannabinoids and a lower transport ratio was observed; however, the mixture only significantly increased A > B permeability (p < 0.0001) (Fig. 3f, Table 3).

CBD but not the other phytocannabinoids modestly inhibited BCRP-mediated transport of prazosin.
We also assessed substrate specificity of inhibition by assessing whether the BCRP substrates CBDA, CBG, Δ 9 -THC, CBD and CBDVA (10 µM) similarly inhibited transport of the established BCRP substrate prazosin. Elacridar, the positive control BCRP inhibitor significantly reduced the transport ratio of prazosin compared to vehicle (B > A, p = 0.0006; A > B, p < 0.0001; Fig. 4a,b; Table 4). Interestingly, CBD was the only cannabinoid to significantly reduce the permeability of prazosin in the B > A direction (p = 0.0284) so was identified as an inhibitor of BCRP (Fig. 4, Table 4).
The effect of the five BCRP substrates on P-glycoprotein function was also examined using digoxin as a substrate (Fig. 5, Table 4). Interestingly, CBDA was not a substrate of P-glycoprotein but was an inhibitor as it was the only cannabinoid to significantly reduce the permeability of digoxin in the B > A direction (p = 0.0481). www.nature.com/scientificreports/ www.nature.com/scientificreports/

Discussion
Here we provide evidence for pharmacokinetic interactions between cannabinoids within a full-spectrum cannabis extract. The pharmacokinetic profiles of the cannabinoids when administered in a cannabis extract were markedly different to those when delivered as individual compounds at equivalent doses. Notably, CBDA plasma concentrations were substantially increased, with the total CBDA plasma exposure being 14-fold higher when administered in a cannabis extract than when administered as a single molecule. Conversely, the peak plasma concentrations of the other cannabinoids (e.g. CBD, Δ 9 -THC and Δ 9 -THCA) were considerably lower. The dramatic increase in plasma CBDA exposure likely results, at least in part, from cannabinoid-cannabinoid interactions at the ABC transporter BCRP in the intestinal lumen. We found that CBDA was a substrate of BCRP and that its transport was inhibited by Δ 9 -THC and CBG in vitro. Since CBDA is a BCRP substrate, systemic absorption of orally administered CBDA would be limited by BCRP transporters located in the apical membrane of the intestine. CBG or Δ 9 -THC inhibiting BCRP-mediated CBDA efflux back into the intestinal lumen would result in increased plasma concentrations of CBDA (Fig. 3g).
Our results here provide a potential mechanism explaining the high plasma CBDA concentrations observed following oral dosing of cannabis oils in a human study 5 . This study measured concentrations of CBD and CBDA in biological fluids of healthy individuals treated with an oral cannabis decoction and oil. Serum CBDA concentrations were approximately 20-30 times higher than serum CBD concentrations despite the products containing only 3-6 times the amount of CBDA compared to CBD 5 . A future human pharmacokinetic study could be conducted to test whether our results in mice translate to humans by utilising a similar study design. That is, plasma CBDA concentrations could be compared following oral administration of a CBD dominant cannabis oil versus a purified CBDA oil.
CBD-dominant cannabis-based nutraceutical oils are increasingly being used worldwide with users suggesting that they are effective in treating numerous ailments 11 . Our data suggest that CBDA might have a more significant contribution to the pharmacological effects of these nutraceutical products than previously thought. Given the emerging preclinical evidence that CBDA has anxiolytic, anti-inflammatory, anticonvulsant and antiemetic properties, it is plausible that CBDA might contribute to any medicinal properties of these cannabis-based products [22][23][24][25] . However, future placebo-controlled randomized trials are required to examine whether CBDA has therapeutic effects in humans. Artisanal cannabis oils are being used to treat seizures in children with intractable epilepsies. However, as yet, there has been no satisfactory explanation for how these oils exert anticonvulsant effects since the CBD doses administered in these oils are substantially lower than those reported to be effective in reducing seizures in clinical trials 10,26,27 . Given that CBDA is anticonvulsant in the Scn1a +/− mouse model of Dravet syndrome, the present results suggest CBDA might contribute to the anticonvulsant effects of orally administered artisanal cannabis extracts 24 . Conversely, CBDA might contribute to the adverse effects of these oils, as its safety profile is not well understood.
In vitro transwell assays were used to determine whether ten cannabinoids were substrates of human BCRP and P-glycoprotein. CBD, CBDA, CBDVA, CBG and Δ 9 -THC were identified as substrates of BCRP. The identification of Δ 9 -THC as a BCRP substrate confirms a prior study in mice showing Δ 9 -THC was a Bcrp1 substrate 15 . That study also reported that Δ 9 -THC was a P-glycoprotein substrate (Mdr1a/Mdr1b) in mice, which is inconsistent with the present findings showing Δ 9 -THC is not a human P-glycoprotein substrate 15 . Moreover, CBD was not a substrate of mouse Bcrp1 but was demonstrated to be a human BCRP substrate here 28 . These inconsistencies suggest some caution when comparing mouse and human data on ABC transporter substrates. The identification Table 2. Permeabilities of cannabinoids in wildtype, BCRP and P-glycoprotein MDCK cells. P Permeability calculations (× 10 −5 cm/s). n.d. not determined; slope of concentration-time curve was not significantly different from zero. *p < 0.05, **p < 0.005, ***p < 0.0005 compared to corresponding wildtype condition. a p < 0.05, b p < 0.0001 compared to without inhibitor.

Wildtype
BCRP P-glycoprotein www.nature.com/scientificreports/ of CBDA as an ABC transporter substrate is somewhat consistent with previous work where its brain-plasma ratio was significantly increased in a Tween-based vehicle compared to a vegetable oil vehicle 24 . Since non-ionic surfactants such as Tween80 are known to inhibit ABC transporters, inhibition of CBDA efflux by BCRP at the blood-brain barrier by Tween80 is a possible mechanism for the increased brain permeability 29 . We also determined whether the cannabinoids were inhibitors of BCRP and P-glycoprotein. Previous in vitro and ex vivo studies reported that CBD, Δ 9 -THC and cannabis-based products inhibit BCRP [17][18][19] . Consistent with these previous studies, CBD inhibited BCRP-mediated transport of prazosin. Surprisingly, CBD did not inhibit BCRP-mediated transport of CBDA; whereas, Δ 9 -THC and CBG inhibited CBDA but not prazosin transport by BCRP. These results further reinforce the importance of considering substrate specificity when Cannabinoids were tested at 10 µM. CBG and Δ 9 -THC significantly inhibit (red shading) transport of CBDA. Data are expressed as means ± SEM, with n = 4 per time point. Curves represent fits to a linear regression and transport efflux ratios (r) are listed (*p < 0.05, ***p < 0.0005, ****p < 0.0001 compared to vehicle; Extra sumof-squares F test). (g) Schematic of CBDA efflux by BCRP located in the intestinal lumen when administered alone (left panel) or as a full-spectrum cannabis extract where its efflux is inhibited by CBG and Δ 9 -THC (right panel). Schematic created using BioRender.com. www.nature.com/scientificreports/ Table 3. BCRP permeabilities of CBDA in the presence of cannabinoids. P Permeability calculations (× 10 −5 cm/s). *p < 0.05, ***p < 0.0005, ****p < 0.0001 compared to vehicle.  Curves represent fits to a linear regression and transport efflux ratios (r) are listed (*p < 0.05, **p < 0.005, ***p < 0.0005 compared to vehicle; Extra sum-of-squares F test). www.nature.com/scientificreports/ evaluating potential transporter-mediated drug-drug interactions (DDIs), as substrate-specific inhibition is a common observation for BCRP and P-glycoprotein 21,[30][31][32][33] . Binding sites and affinities of both the substrate and the inhibitor contributed to substrate-dependent interactions, especially with the ABC transporters for which multiple binding sites have been proposed 30,[33][34][35] . The multiple drug binding sites of P-glycoprotein may explain why CBDA inhibited digoxin transport via P-glycoprotein but was not itself a substrate. Several P-glycoprotein inhibitors with allosteric mechanisms of action such as reduced substrate affinity, decreased ATPase activity, conformational changes that prevent substrate translocation and reduced rates of dissociation have been identified [36][37][38][39] . Future studies could explore whether CBDA is a non-competitive inhibitor of P-glycoprotein and its mechanism of transporter inhibition. Digoxin, a P-glycoprotein substrate with a narrow therapeutic window, has been implicated in several transporter-mediated DDIs. Cardiac and gastrointestinal toxicity has been reported for digoxin when co-administered with P-glycoprotein inhibitors such as quinidine [40][41][42] . Since DDIs occurring at ABC transporters can have serious clinical consequences, an integral part of the drug development process for new candidates is to evaluate whether they are substrates, inhibitors and/or inducers of ABC transporters as required by drug regulatory agencies 43 . Here, only four of the cannabinoids (CBD, CBDA, CBG and Δ 9 -THC) inhibited ABC transporter function, suggesting the likelihood of transporter-mediated DDIs by cannabinoids is low. However, before a definitive conclusion on DDI liability can be made, additional distinct probe substrates should be screened so any potential DDIs resulting from substrate specificity are not overlooked.

Inhibitor P (B > A) P (A > B) r
While DDIs occurring at ABC transporters can have serious adverse effects, inhibition of transporters can also be therapeutically advantageous. Many anticancer and antimicrobial drugs are substrates of ABC transporters and, therefore, have low bioavailability [44][45][46] . Rational drug design efforts have involved non-toxic BCRP and P-glycoprotein inhibitors, including excipients, to purposefully enhance oral absorption of substrates 36,[46][47][48][49][50] . A similar therapeutic advantage of cannabinoids improving the low bioavailability of co-administered therapeutic drugs that are ABC transporter substrates could be explored in future studies.
Several limitations of the present study need to be considered. While not examined here, terpenoids and flavonoids inhibit ABC transporters and could contribute to the increased absorption of CBDA within the full-spectrum extract 51,52 . Moreover, interactions mediated by the cytochrome P450 (CYP450) family of drug metabolizing enzymes could also contribute to pharmacokinetic entourage effects in cannabis. CYP450s tend to be the most common source of pharmacokinetic interactions since competition between two drugs for the same metabolizing pathway can drastically affect metabolism and elimination parameters. While the metabolic pathways are still unknown for many of the cannabinoids, CYP450-mediated metabolism contributes extensively to the elimination of CBD and Δ 9 -THC 53 . The prolonged t 1/2 values observed for the cannabinoids when administered as a full-spectrum extract could be the consequence of interactions between the cannabinoids at the drug metabolizing CYP450 enzymes. Indeed, we recently reported that several phytocannabinoids found in the full-spectrum extract inhibited CYP450 enzymes including CYP3A4, CYP2C9, CYP1A2, CYP2B6 and CYP2C19 54 . While the CYP450 enzymes involved in the metabolism of CBDA have not been characterized, it www.nature.com/scientificreports/ is possible that cannabinoid inhibition of CYP450-mediated first-pass metabolism of CBDA could account for its increased plasma concentrations when administered as a full-spectrum extract compared to as an individual cannabinoid. In any case, the present observation of cannabinoid-CBDA interactions at the ABC transporter BCRP are very likely to contribute to the enhanced plasma CBDA concentrations that were observed following oral administration of a cannabis extract.

Conclusion
Many have been puzzled by the high bioavailability of CBDA in humans following the oral ingestion of CBDdominant cannabis-derived nutraceutical oils 5, 55-57 . Our results suggest that the oral administration of such cannabis extracts provides a natural vehicle to enhance plasma CBDA concentrations due to cannabinoidcannabinoid interactions at the drug efflux transporter BCRP. Taken together with emerging preclinical evidence that CBDA has anti-emetic, anxiolytic and anticonvulsant effects, the present results highlight that the contribution of CBDA to the pharmacological effects of hemp nutraceutical products warrants further inspection. Our results showing pharmacokinetic interactions between cannabinoids provides one mechanism for the much touted "entourage effect" of cannabis.
Analytical methods. Concentrations of cannabinoids in plasma samples were quantified as described previously 8,24,58 . Briefly, plasma samples were spiked with diazepam as an internal standard and protein precipitation was achieved by vortex-mixing with acetonitrile. The organic layer was isolated by centrifugation (4000 g for 10 min) and evaporated to dryness with N 2 . Samples were reconstituted in acetonitrile and 0.1% formic acid in water (1:3.3, v/v) for supported-liquid extraction (SLE) with methyl tert-butyl ether (MTBE) using Biotage Isolute SLE+ columns (Uppsala, SWE). Samples were evaporated to dryness with N 2 and reconstituted in acetonitrile and 0.1% formic acid in water (1:1, v/v) for analysis by LC-MS/MS as previously described 8 www.nature.com/scientificreports/ Bidirectional transport assays. Corning Transwell polycarbonate membrane cell culture inserts (0.4 µm, 6.5 and 12 mm; Corning Inc.; Corning, USA) were used for bidirectional transport assays. Briefly, 72 h prior to the transwell assay, cells (2.5 × 10 5 cells/well or 2.0 × 10 5 cells/well for 12 mm or 6.5 mm inserts, respectively) were plated. For substrate assays, cells were rinsed with PBS and vehicle (DMSO) or 10 µM inhibitor (loratadine, P-glycoprotein or elacridar, BCRP) in DMEM supplemented with 10% FBS was added to both the apical and basolateral chambers and incubated for 15 min at 37 °C in a humidified 5% CO 2 atmosphere. Media in the donor chamber was then replaced with that containing 10 µM of an individual cannabinoid in the presence of either vehicle or inhibitor and returned to 37 °C. Aliquots (25 or 50 µL) were removed from the accepter chamber at 60, 120, 180 and 240 min.
For inhibitor assays, cells were rinsed with PBS and 10 µM of an individual cannabinoid in DMEM supplemented with 10% FBS was added to both the apical and basolateral chambers and incubated for 15 min at 37 °C in a humidified 5% CO 2 atmosphere. Following the incubation, media in the donor chamber was replaced with 1 µM substrate (digoxin, P-glycoprotein or prazosin, BCRP) and its respective cannabinoid, returned to 37 °C and aliquots were removed from acceptor chamber as described above.
Concentrations of cannabinoids, digoxin or prazosin in the acceptor chamber were quantified using LC-MS/ MS. Samples were spiked with diazepam as an internal standard and then either 0.1% formic acid in water (cannabinoids and digoxin) or 0.5 M sodium hydroxide (prazosin) was added for SLE with MTBE (cannabinoid and digoxin) or ethyl acetate (prazosin). Samples were evaporated to dryness with N 2 and reconstituted in 1:1, v/v acetonitrile and 0.1% formic acid in water (cannabinoids), methanol and 0.1% formic acid in water (prazosin) or methanol and 0.1% formic acid in 10 mM ammonium acetate (digoxin) for analysis by LC-MS/ MS. Cannabinoids were analyzed as described above. The mass spectrometer was operated in positive electrospray ionization mode with multiple reaction monitoring (digoxin: 798. 35  Lucifer yellow permeability assay. At the completion of the transwell assay, a Lucifer yellow permeability assay was conducted to confirm monolayer integrity. High Potassium Hank's Balanced Salt Solution (HBSS) replaced the media in both chambers. Lucifer yellow (250 µM) was added to the apical chamber and cells were incubated at 37 °C for 60 min. A CLARIOstar plate reader (BMG Labtech; Offenburg, GER) was used to take fluorescence readings over 0.5 ms (excitation 485 nm, emission 535 nm) from samples taken from the basolateral chamber. Baseline fluorescence as measured from samples containing HBSS only was subtracted and fluorescence readings were normalized to those of 250 µM Lucifer yellow. Monolayers were considered intact if Lucifer yellow permeability was less than 5% 60 .

Data analysis.
Rates of substrate transport were determined by linear regression of concentration-time curves using GraphPad Prism. Apparent permeability (P) of substrate transport across MDCK cell monolayers were calculated for both the basolateral to apical (B > A) and apical to basolateral (A > B) directions as previously described using the following equation: V, volume of acceptor chamber (B > A: 0.5 mL and 0.2 mL and A > B: 1.5 mL and 0.6 mL for 12 mm and 6 mm inserts, respectively). C 0 , initial substrate concentration in the donor chamber (10 µM cannabinoids or 1 µM digoxin and prazosin). SA, monolayer growth surface area (1.12 cm 2 , 12 mm inserts; 0.33 cm 2 , 6.5 mm inserts) 21 . ΔC/Δt, slope calculated concentration-time curves. Transport efflux ratios (r) were calculated by dividing the apparent permeability calculated for the B > A direction by that calculated for the A > B direction. Transport ratios could not be calculated in instances when the slope for concentration-time curves in the A > B direction were not significantly different from zero.
Comparisons of curve fits for concentration-time curves between wildtype MDCK cells and MDCK cells expressing P-glycoprotein or BCRP were calculated using the Extra sum-of-squares F test to determine whether a cannabinoid was a substrate. Rates of substrate transport in the A > B direction were not different between wildtype and transporter-expressing cells for any cannabinoid, with the exception of CBG and Δ 9 -THC in cells expressing BCRP. Comparisons of curve fits in the B > A direction with p < 0.05 were considered significantly different and indicative of a substrate for the corresponding transporter. In order to determine whether cannabinoids were inhibitors, comparisons of curve fits for concentration-time curves between vehicle-treated and cannabinoid-treated cells were calculated using the Extra sum-of-squares F test. Comparisons of curve fits with p < 0.05 were considered significantly different and indicative the compound being an inhibitor.

Data availability
All relevant data are presented within the manuscript and are available from the corresponding author on reasonable request.