Cannabidiol is a negative allosteric modulator of the cannabinoid CB₁ receptor

Cell culture

HEK 293A cells were from the American Type Culture Collection (ATCC, Manaassas, VI, USA). Cells were maintained at 37°C, 5% CO₂ in DMEM supplemented with 10% FBS and 10⁴ U∙mL⁻¹ Pen/Strep. STHdh^Q7/Q7 cells are derived from the conditionally immortalized striatal progenitor cells of embryonic day 14 C57BL/6J mice (Coriell Institute, Camden, NJ, USA) (Trettel et al., 2000). Cells were maintained at 33°C, 5% CO₂ in DMEM supplemented with 10% FBS, 2 mM l-glutamine, 10⁴ U∙mL⁻¹ Pen/Strep and 400 µg∙mL⁻¹ geneticin. Cells were serum-deprived for 24 h prior to experiments to promote differentiation (Trettel et al., 2000; Laprairie et al., 2013, 2014a,2014b).

Plasmids and transfection

Human CB₁, CB_1A, CB_1B receptors and arrestin2 (β-arrestin1) were cloned and expressed as either green fluorescent protein² (GFP²) or Renilla luciferase II (Rluc) fusion proteins. CB₁-GFP² and arrestin2-Rluc were generated using the pGFP²-N3 and pcDNA3.1 plasmids (PerkinElmer, Waltham, MA, USA) as described previously (Hudson et al., 2010; Laprairie et al., 2014a). The GFP²-Rluc fusion construct and Rluc plasmids have been previously described (Laprairie et al., 2014a).

The human CB₁ receptor was mutated at two cysteine residues (Cys⁹⁸ and Cys¹⁰⁷). Mutagenesis was conducted as described previously (Laprairie et al. 2013) with the cysteine residues being mutated to alanine (C98A and C107A) or to serine (C98S and C107S) using the CB₁-GFP² fusion plasmid and the following forward and reverse primers: CB₁^C98A-GFP² forward 5′-AACATCCAGGCTGGGGAGAACT-3′, reverse 5′-AGTTCTCCCCAGCCTGGATGTT-3′; and CB₁^C107A-GFP² forward 5′-GACATAGAGGCTTTCATGGTC-3′, reverse 5′-GACCATGAAAGCCTCTATGTC-3′; CB₁^C98S-GFP² forward 5′-AACATCCAGTCTGGGGAGAACT-3′, reverse 5′-AGTTCTCCCCAGACTGGATGTT-3′; and CB₁^C107S-GFP² forward 5′-GACATAGAGTCTTTCATGGTC-3′, reverse 5′-GACCATGAAAGACTCTATGTC-3′.

Mutagenesis was confirmed by sequencing (GeneWiz, Camden, NJ, USA).

Cells were grown in six-well plates and transfected with 200 ng of the Rluc fusion plasmid and 400 ng of the GFP² fusion plasmid according to previously described protocols (Laprairie et al., 2014a) using Lipofectamine 2000® according to the manufacturer's instructions (Invitrogen, Burlington, ON, Canada). Transfected cells were maintained for 48 h prior to experimentation.

BRET²

Interactions between CB₁ and arrestin2 were quantified via BRET² according to previously described methods (Laprairie et al., 2014a). BRET efficiency (BRET_Eff) was determined as previously described (James et al., 2006; Laprairie et al., 2014a) such that Rluc alone was used to calculate BRET_MIN and the Rluc-GFP² fusion protein was used to calculate BRET_MAX.

On-cell™ and In-cell™ Western

On-cell Western analyses were completed as described previously (Laprairie et al., 2014a) using primary antibody directed against N-CB₁ receptors (1:500; Cayman Chemical Company, Ann Arbor, MI, USA; Cat No. 101500). All experiments measuring CB₁ receptors included an N-CB₁ blocking peptide control (1:500; Cayman Chemical Company), which was incubated with N-CB₁ antibody (1:500). Immunofluorescence observed with the N-CB₁ blocking peptide was subtracted from all experimental replicates. In-cell Western analyses were conducted as described previously (Laprairie et al., 2014a). Primary antibody solutions were as follows: N-CB₁ (1:500), pERK1/2(Tyr205/185) (1:200), ERK1/2 (1:200), pPLCβ3(S537) (1:500), PLCβ3 (1:1000) or β-actin (1:2000) (all from Santa Cruz Biotechnology, Santa Cruz, CA, USA). Secondary antibody solutions were as follows: IR^CW700dye or IR^CW800dye (1:500; Rockland Immunochemicals, Gilbertsville, PA, USA). Quantification was completed using the Odyssey Imaging system and software (v. 3.0; Li-Cor, Lincoln, NE, USA).

Data analysis and curve fitting

Data are presented as the mean ± the SEM or mean and 95% confidence interval, as indicated, from at least four independent experiments. All data analysis and curve fitting were carried out using GraphPad (v. 5.0) (Prism; La Jolla, CA). Concentration–response curves (CRCs) were fit with the nonlinear regression with variable slope (four parameters), Gaddum/Schild EC₅₀ shift model or operational model of allosterism (equation 1) (Keov et al., 2011) and are shown in each figure according to the best-fit model as determined by R² value (GraphPad Prism v. 5.0). Pharmacological statistics were obtained from nonlinear regression models as indicated in figures and tables. Global curve fitting of allosterism data was carried out using the following operational model (Keov et al., 2011; Smith et al., 2011; Hudson et al., 2014):

(1)

where E is the measured response, A and B are the orthosteric and allosteric ligand concentrations, respectively; E_max is the maximum system response; α is a measure of the allosteric co-operativity on ligand binding; β is a measure of the allosteric effect on efficacy; K_A and K_B are estimates of the binding of the orthosteric and allosteric ligands, respectively; n represents the Hill slope; and τ_A and τ_B represent the abilities of the orthosteric and allosteric ligands to directly activate the receptor (Smith et al., 2011). To fit experimental data to this equation, E_max and n were constrained to 1.0 and 1.0, respectively, which allowed for estimates of α, β, K_A, K_B, τ_A and τ_B.

Relative receptor activity (RA) was calculated according to equation 2 (Christopoulos and Kenakin, 2002):

(2)

where E_max % is the E_max of the CRC in the presence of a given concentration of CBD, EC₅₀ is the EC₅₀ (μM) in the presence of a given concentration of CBD, E_{max Agonist Alone} % is the E_max in the absence of CBD and EC_{50 Agonist Alone} is the EC₅₀ (μM) in the absence of CBD. Statistical analyses were one-way or two-way ANOVA, as indicated, using GraphPad. Post hoc analyses were performed using Dunnett's multiple comparisons as well as Bonferroni's or Tukey's test, as indicated. Homogeneity of variance was confirmed using Bartlett's test. The level of significance was set to P < 0.001 or P < 0.01, as indicated. To improve the readability of the data, figures have been laid out where possible such that data from HEK 293A cells appear above data from STHdh^Q7/Q7 cells and data for O-2050 appear before data for CBD (Figures 2-4).

Materials

2-AG, CBD and O-2050 were purchased from Tocris Bioscience (Bristol, UK). THC was purchased from Sigma-Aldrich (Oakville, ON, Canada). Stock solutions were made up in ethanol (THC) or DMSO (2-AG, CBD and O-2050, AM251) and diluted to final solvent concentrations of 0.1%.

CB₁ receptor internalization and kinetic experiments

We had previously observed that CBD reduced CB₁ receptor internalization in STHdh^Q7/Q7 cells (Laprairie et al., 2014a). Here, we sought to determine how CBD affected the kinetics of CB₁ receptor internalization and arrestin2 recruitment in STHdh^Q7/Q7 cells. The fraction of CB₁ receptors at the plasma membrane was concentration-dependently decreased by THC (Figure 1A) and 2-AG in STHdh^Q7/Q7 cells (Figure 1B). The efficacy and potency of THC and 2-AG in inducing internalization of CB₁ receptors were reduced by increasing concentrations of CBD (Figure 1A and B). BRET² between arrestin2-Rluc and CB₁-GFP² was measured every 10 s for 4 min in STHdh^Q7/Q7 cells treated with 1 μM THC (Figure 1C) or 2-AG (Figure 1D). Increasing concentrations of CBD decreased the rate of association between arrestin2 and CB₁ receptors over 4 min (Figure 1E) and decreased maximal BRET_Eff observed at 10 min (Figure 1C–E). The fraction of CB₁ receptors at the plasma membrane was also reduced in STHdh^Q7/Q7 cells treated with 1 μM THC (Figure 1F) or 2-AG (Figure 1G) over 60 min. CBD alone increased the fraction of CB₁ receptors at the membrane (Figure 1F–H). The rates of internalization and the maximum fraction internalized, for CB₁ receptors, were reduced by increasing concentrations of CBD (Figure 1F–H). Similarly, Cawston et al. (2013) observed that the rate of arrestin recruitment to CB₁ receptors was reduced by the allosteric modulator ORG27569. Therefore, CBD delayed the interactions between CB₁ receptors and arrestin2 and increased the pool of receptors present at the plasma membrane at submicromolar concentrations, which is similar to the actions of the previously described CB₁ receptor NAM, ORG27569 (Cawston et al., 2013).

CB₁ receptor-arrestin2 BRET² experiments

2-AG and THC enhance the interaction between CB₁ receptors and arrestin2, as indicated by BRET² in STHdh^Q7/Q7 cells (Laprairie et al., 2014a). Here, we used HEK 293A cells as a heterologous expression system for CB₁ receptors and arrestin2 to determine whether CBD acted as a NAM of these receptors. Treatment of HEK 293A cells with 0.01–5 μM THC or 2-AG for 30 min produced a concentration-dependent increase in BRET_Eff between arrestin2-Rluc and CB₁-GFP² (Figure 2A–D). The CB₁ receptor antagonist O-2050 (0.01–5.00 μM) produced a concentration-dependent rightward shift in the THC and 2-AG CRCs that were best fit using the Gaddum/Schild EC₅₀ nonlinear regression model indicative of competitive antagonism (Figure 2A and B). CBD (0.01–5 μM) treatment produced a concentration-dependent rightward and downward shift in the THC and 2-AG CRCs that were best fit using the operational model of allosterism (equation 1 and Figure 2C and D). The rightward shift in EC₅₀ was significant at 1.00 and 0.50 μM CBD for THC- and 2-AG-treated cells respectively (Table 1). The decrease in E_max was significant at 0.1 and 0.5 μM for THC- and 2-AG-treated cells respectively (Table 1). The Hill coefficient (n) was less than 1 at 0.1 and 0.5 μM for THC- and 2-AG-treated cells respectively (Table 1). Relative RA (estimated using equation 2) was significantly reduced at 0.01 μM for THC- and 2-AG-treated cells (Table 1). Schild analyses of these data demonstrated that while O-2050 behaved as a competitive antagonist, inhibition of BRET_Eff by CBD was nonlinear for THC- and 2-AG-treated HEK 293A cells (Figure 2E and Table 2). These data demonstrated that CBD behaved as a NAM of THC- and 2-AG-mediated arrestin2 recruitment to CB₁ receptors in the HEK 293A heterologous expression system.

The NAM properties of CBD on CB₁-arrestin2 interactions were confirmed in the STHdh^Q7/Q7 cell culture model of medium spiny projection neurons. As in HEK 293A cells, O-2050 treatment produced a concentration-dependent rightward shift in the THC and 2-AG CRCs that were best fit using the Gaddum/Schild EC₅₀ nonlinear regression model indicative of competitive antagonism (Figure 2F and G), and CBD treatment produced a concentration-dependent rightward and downward shift in the THC and 2-AG CRCs that were best fit using the operational model of allosterism (Figure 2H and I) in STHdh^Q7/Q7 cells. The rightward shift in EC₅₀ was significant at 0.5 μM CBD for THC- and 2-AG-treated cells (Table 1). The decrease in E_max was significant at 1 and 5 μM for THC- and 2-AG-treated cells respectively (Table 1). The Hill coefficient (n) was less than 1 at 0.5 and 5 μM for THC- and 2-AG-treated cells respectively (Table 1). Relative RA (equation 2) was significantly reduced at 0.1 μM for both THC- and 2-AG-treated cells (Table 1). The Schild regression for these data demonstrated that O-2050 modelled competitive antagonism for THC- and 2-AG-treated STHdh^Q7/Q7 cells (greater slope and R²) (Figure 2J and Table 2). CBD alone displayed weak partial agonist activity in this assay at concentrations >2 μM (Supporting Information Fig. S1). Taken together, these data indicate that CBD behaved as a NAM of THC- and 2-AG-mediated arrestin2 recruitment to CB₁ receptors at concentrations below its reported affinity to these receptors in a cell culture model endogenously expressing CB₁ receptors (Pertwee, 2008).

CB₁ receptor-mediated phosphorylation of PLCβ3

THC and 2-AG treatment both result in a concentration-dependent increase in PLCβ3 phosphorylation in HEK 293A cells (Figure 3A–D) and STHdh^Q7/Q7 cells (Laprairie et al., 2014a; Figure 3F–I). O-2050 treatment resulted in a concentration-dependent rightward shift in the THC and 2-AG CRCs (Figure 3A, B, F and G), while CBD treatment resulted in a rightward and downward shift in the THC and 2-AG CRCs, in both cell lines (Figure 3C, D, H and I). O-2050 CRCs were best fit with the Gaddum/Schild EC₅₀ model, while CBD CRCs were best fit with the operational model of allosterism. The rightward shift in EC₅₀ was significant at 0.5 μM CBD for THC- and 2-AG-treated HEK 293A cells (Table 3) and 0.5 and 1 μM CBD for THC- and 2-AG-treated STHdh^Q7/Q7 cells respectively (Table 3). The decrease in E_max was significant at 1 and 0.5 μM for HEK 293A and STHdh^Q7/Q7 cells respectively (Table 3). The Hill coefficient (n) was less than 0.5 and 1 μM for THC- and 2-AG-treated in both HEK 293A and STHdh^Q7/Q7 cells (Tables 1 and 3). Relative RA was significantly reduced at 0.1 μM for THC- and 2-AG-treated HEK 293A and STHdh^Q7/Q7 cells (Table 3). The Schild regression for these data demonstrated that O-2050 modelled competitive antagonism for THC- and 2-AG-treated STHdh^Q7/Q7 cells, while CBD did not (greater slope and R²) (Figure 3E and J and Table 2). As with arrestin2 recruitment, CBD alone was a weak partial agonist at concentrations >2 μM (Supporting Information Fig. S1). In the presence of 2-AG or THC, CBD was a NAM of PLCβ3 phosphorylation in HEK 293A cells overexpressing CB₁ receptors and STHdh^Q7/Q7 cells endogenously expressing these receptors.

CB₁ receptor-mediated phosphorylation of ERK1/2

2-AG treatment results in the phosphorylation of ERK1/2 in STHdh^Q7/Q7 cells, while THC does not (Laprairie et al., 2014a). 2-AG treatment produced a concentration-dependent increase in ERK1/2 phosphorylation in both HEK 293A and STHdh^Q7/Q7 cells (Figure 4A, B, D and E). O-2050 treatment resulted in a concentration-dependent rightward shift in the 2-AG CRCs (Figure 4A and D), while CBD treatment resulted in a rightward and downward shift in the 2-AG CRCs, in both cell lines (Figure 4B and E). O-2050 CRCs were best fit with the Gaddum/Schild EC₅₀ model, while CBD CRCs were best fit with the operational model of allosterism. The rightward shift in EC₅₀ was significant at 0.5 and 1 μM CBD for HEK 293A and STHdh^Q7/Q7 cells respectively (Table 4). The decrease in E_max was significant at 5 and 1 μM for HEK 293A and STHdh^Q7/Q7 cells respectively (Table 4). The Hill coefficient (n) was less than 1 at 0.01 and 0.1 μM CBD for HEK 293A and STHdh^Q7/Q7 cells respectively (Table 4). Relative RA was significantly reduced at 0.01 and 0.1 μM for 2-AG-treated HEK 293A and STHdh^Q7/Q7 cells respectively (Table 4). The Schild regression for these data demonstrated that O-2050 modelled competitive antagonism in HEK293A (Figure 3C) and STHdh^Q7/Q7 (Figure 4F) cells, whereas CBD did not (greater slope and R²) (Table 2). CBD was a NAM of 2-AG-mediated ERK1/2 phosphorylation at CB₁ receptors overexpressed in HEK 293A cells and endogenously expressed in STHdh^Q7/Q7 cells, at concentrations lower than those reported for CB₁ receptor agonist activity (Mechoulam et al., 2007; McPartland et al., 2014) (Supporting Information Fig. S1). Therefore, CBD behaved as a NAM in these cell lines for arrestin2 recruitment, PLCβ3 and ERK1/2 phosphorylation.

Operational modelling of allosterism

While O-2050 acted as a competitive orthosteric antagonist, CBD acted as a NAM in arrestin2, PLCβ3 and ERK1/2 assays. Global curve fitting of data to the operational model of allosterism was used to assess the NAM activity of CBD. Data were fit to this model by constraining E_max and n (Hill slope) to 1.0 and 1.0, respectively. In this way, the allosteric co-operativity coefficient for ligand binding (α) was found to be less than 1.0 (0.37), with no significant difference between cell lines, orthosteric ligands or assays (Table 5) indicating that CBD acted as a NAM to reduce the binding of THC and 2-AG. CBD also reduced the efficacy of the orthosteric ligand because β (co-operativity coefficient for ligand efficacy) was consistently less than 1 (0.44). Based on the estimated value of orthosteric ligand affinity (K_A) and the ability of the orthosteric ligand to activate CB₁ receptors (τ_A), 2-AG (241 nM) and THC (97 nM) were able to directly activate CB₁ receptors within a concentration range similar to that published earlier (see Pertwee, 2008). CBD did not display agonist activity, as shown by the estimate of τ_B, but exhibited a greater estimated affinity (304 nM) for CB₁ receptors (K_B) than would be predicted for the orthosteric site (see Pertwee, 2008). β and αβ can be used to assess ligand bias (functional selectivity) for allosteric modulators (Keov et al., 2011). No differences in β and αβ were observed in HEK 293A cells in all assays (Table 5). In STHdh^Q7/Q7 cells, β and αβ were reduced in PLCβ3 assays compared with arrestin2 recruitment and ERK assays, indicating that CBD was a functionally selective inhibitor of arrestin2 and ERK1/2 pathways (Table 5). Overall, CBD was a NAM of orthosteric ligand binding and efficacy at CB₁ receptors.

Negative allosteric modulation of antagonist binding

If CBD reduced the binding of orthosteric agonists to CB₁ receptors , as predicted by the operational model of allosterism, then CBD should also reduce the binding of CB₁ receptor inverse agonists and antagonists. In order to test this hypothesis, STHdh^Q7/Q7 cells were treated with the CB₁ receptor inverse agonist AM251 (Pertwee, 2005), and CBD and ERK phosphorylation was measured (Figure 5A). CBD treatment resulted in a rightward and upward shift in the AM251 CRC (Figure 5A). CBD CRCs were best fit with the operational model of allosterism. To further test our hypothesis, STHdh^Q7/Q7 cells were treated with 2-AG and 500 nM O-2050, 500 nM CBD or 500 nM O-2050 and 500 nM CBD (Figure 5B). Treatment of STHdh^Q7/Q7 cells with 2-AG, O-2050 and CBD produced a CRC that was shifted right and down relative to 2-AG alone and left relative to 2-AG and O-2050, indicating that CBD had reduced the competitive antagonist activity of O-2050 and reduced the efficacy of 2-AG (Figure 5B). Therefore, CBD was a NAM of orthosteric ligand binding as demonstrated by the reduced potency and efficacy of the CB₁ receptor inverse agonist AM251 and the antagonist O-2050.

Mutagenesis of the CB₁ receptor

The CB₁ receptor splice variants CB_1A and CB_1B differ in the first 89 amino acids of the N-terminus, relative to CB₁. We compared the allosteric activity of CBD in STHdh^Q7/Q7 cells expressing CB₁, CB_1A and CB_1B receptors using BRET². BRET_Eff did not differ between CB₁-GFP²-, CB_1A-GFP²- and CB_1B-GFP²-expressing cells treated with 0.01–5 μM THC or 2-AG ± 0.5 μM O-2050 or 5 μM CBD (Supporting Information Fig. S2A and B). Therefore, the site of the negative allosteric activity of the CB₁ receptor was not contained within the amino acids 1–89 that differ between CB₁, CB_1A and CB_1B receptors but was associated with the conserved residues common to all three variants (Bagher et al., 2013; Fay and Farrens, 2013).

Fay and Farrens (2013) previously reported that Cys⁹⁸ and Cys¹⁰⁷ in the extracellular N-terminus of CB₁ receptors contribute to the allosteric activity of ORG27569 and PSNCBAM-1 by the formation of a disulfide bridge (Fay and Farrens, 2013). We hypothesized that these residues might similarly influence the allosteric activity of CBD. We wanted to determine whether it was the polarity of Cys⁹⁸ and Cys¹⁰⁷ or the formation of a disulfide bridge that contributed to allosteric activity. Each of these residues was individually mutagenized to Ala or Ser in the CB₁-GFP² plasmid (CB₁^WT-GFP², CB₁^C98A-GFP², CB₁^C107A-GFP², CB₁^C98S-GFP² and CB₁^C107S-GFP²) and transfected with arrestin2-Rluc into STHdh^Q7/Q7 cells. Treatment of CB₁^WT-, CB₁^C98A-, CB₁^C107A-, CB₁^C98S- or CB₁^C107S-expressing cells with 0.01–5 μM THC or 2-AG alone resulted in a response that did not differ between CB₁ receptor mutants or between THC and 2-AG treatments (Figure 6A and B). Further, the competitive antagonist activity of 0.5 μM O-2050 was not different in CB₁ receptor mutant-expressing cells treated with 0.01–5 μM THC or 2-AG (Supporting Information Fig. S2C and D). Together, these data indicated that mutation of Cys⁹⁸ or Cys¹⁰⁷ did not alter the response of CB₁ receptors to orthosteric ligands. Treatment of CB₁^WT-expressing cells with 0.01– μM THC or 2-AG and 5 μM CBD resulted in a rightward and downward shift in the BRET_Eff CRCs (Figure 6A and B). Similarly, treatment of CB₁^C98A-expressing or CB₁^C107A-expressing cells with 0.01–5 μM THC or 2-AG and 5.00 μM CBD resulted in a rightward and downward shift in the BRET_Eff CRCs compared with vehicle treatment (Table 6). The magnitude of the rightward and downward shift was less pronounced in CB₁^C98A- and CB₁^C107A-expressing cells compared with CB₁^WT-, CB₁^C98S- and CB₁^C107S-expressing cells treated with CBD (Table 6 and Figure 6A and B). The presence of a polar Ser or Cys at position 98 or 107 was sufficient to restore the wild-type response to CBD. Therefore, the allosteric activity of CBD at CB₁ receptors depended in part on the presence of polar residues at positions 98 and 107, independent of a disulfide bridge. Additional residues common to CB₁, CB_1A and CB_1B receptors may also contribute to the allosteric effect of CBD (Figure 6C).

CBD behaves as a NAM of CB₁ receptors

In this study, we provide in vitro evidence for the non-competitive negative allosteric modulation of CB₁ receptors by CBD. CBD treatment resulted in negative co-operativity (α < 1) and reduced orthosteric ligand (THC and 2-AG) efficacy (β < 1) at concentrations lower than the predicted affinity of CBD for the orthosteric binding site at CB₁ receptors [304 nM (this study) versus >4 μM (see Pertwee, 2008)]. As a NAM of CB₁ receptor orthosteric ligand-dependent effects, CBD reduced both G protein-dependent signalling and arrestin2 recruitment, which explains both the diminished signalling and diminished BRET observed between CB₁-GFP² and arrestin2-Rluc. In contrast to the NAM activity of CBD and as shown previously, O-2050 acted as a competitive orthosteric antagonist of CB₁ receptors (Canals and Milligan, 2008; Higuchi et al., 2010; Hudson et al., 2010; Ferreira et al., 2012; Anderson et al., 2013; Laprairie et al., 2014b) rather than a partial agonist (Wiley et al., 2011, 2012). To directly test the hypothesis that a disulfide bridge between Cys⁹⁸ and Cys¹⁰⁷ regulated the activity of CB₁ receptor allosteric modulators, these residues were mutated to either Ala or Ser (Fay and Farrens, 2013). Mutation of these residues to Ala (nonpolar) decreased the NAM activity of CBD at CB₁ receptors but not the activity of THC, 2-AG, or O-2050. The NAM activity of CBD depended upon the presence of polar (Ser or Cys) residues at positions 98 and 107 in the CB₁ receptor, rather than a disulfide bridge, because replacement of either Cys residue with Ser did not change CBD NAM activity. These findings suggest that the N-terminal, extracellular residues Cys⁹⁸ and Cys¹⁰⁷ partly regulate either the allosteric activity of CBD at CB₁ receptorsdirectly or the communication between the allosteric and orthosteric sites of these receptors.

Allosteric modulators are probe-dependent; that is, the activity of the allosteric modulator depends on the orthosteric probe being used (Christopoulos and Kenakin, 2002). ORG27569 and PSNCBAM-1 both display probe dependence because they are more potent modulators of CP55,940 binding and CP55,940-mediated CB₁ receptor activation than WIN55,212-2 binding and WIN55,21-2-mediated CB₁ receptor activation (Baillie et al., 2013). 2-AG was chosen as an orthosteric probe in this study because it is the most abundant endocannabinoid in the brain, and therefore, 2-AG would be the predominant endogenous orthosteric ligand if exogenous CBD was administered (Sugiura et al., 1999). THC and CBD are the most abundant phytocannabinoids in marijuana and are used together in varying ratios both medicinally and recreationally in marijuana (Thomas et al., 2007). Therefore, THC was selected as an alternative orthosteric probe. In HEK 293A cells, CBD did not display probe-dependence (Table 2). In STHdh^Q7/Q7 cells, CBD was a more potent NAM of CB₁ receptor-dependent arrestin2 recruitment when THC was the orthosteric probe compared with 2-AG (Table 2). No probe-dependence was observed for PLCβ3 and ERK1/2 signalling. BRET was used in this study to directly measure the association of CB₁ receptors and arrestin2, which may be a more sensitive method for detecting probe dependence than In-cell Western assays that measured PLCβ3 or ERK1/2.

STHdh^Q7/Q7 cells express several effector proteins that CBD has been shown to modulate, including CB₁, 5HT_1A, GPR55, μ-opioid receptors, PPARγ and FAAH, suggesting that CBD could have acted independently of CB₁ receptors (Trettel et al., 2000; Lee et al., 2007; Laprairie et al., 2014a). However, the NAM activity of CBD was also observed in HEK 293A cells that heterologously express CB₁ receptors but do not express 5HT_1A, GPR55 and μ-opioid receptors, demonstrating that these effectors did not alter the actions of CBD (Ryberg et al., 2007). HEK 293A cells do express PPARγ, but modulation of this nuclear receptor would not affect arrestin and G protein assays used over the duration of these experiments. Importantly, the NAM activity of CBD at CB₁ receptors was dependent on the cannabinoid agonists 2-AG and THC, suggesting that CBD was acting at these receptors. FAAH inhibition would have enhanced, not diminished, cannabinoid efficacy, which was not observed here. Therefore, the NAM activity of CBD at CB₁ receptors documented in this study adds to the mechanisms of action through which chronic CBD mediates its effects in vivo.

No significant signalling bias was observed for CBD in HEK 293A cells because allosteric ligand efficacy (β) and co-operativity (αβ) were not different among arrestin, PLCβ3 and ERK1/2 assays (Table 5). In STHdh^Q7/Q7 cells, we observed that CBD was biased for PLCβ3 signalling compared with ERK signalling and arrestin2 recruitment as indicated by reduced β and αβ values (Table 5). Previous studies have reported that ORG27569 is also biased against ERK and arrestin signalling (Ahn et al., 2012, 2013; Baillie et al., 2013). The observation that CBD-dependent bias was observed in STHdh^Q7/Q7 cells compared with HEK 293A cells suggests that heterologous expression systems may under-represent ligand bias (Ahn et al., 2013; Baillie et al., 2013).

CBD compared with other NAMs of CB₁ receptors

Based on the functional effects of CBD on PLCβ3, ERK, arrestin2 recruitment and CB₁ internalization, CBD behaved like the well-characterized NAMs, ORG27569 and PSNCBAM-1 in vitro (Horswill et al., 2007; Cawston et al., 2013). At higher doses (>2 μM), CBD was able to enhance PLCβ3 and ERK phosphorylation and arrestin2 recruitment, as well as limit CB₁ receptor internalization, suggesting that CBD may behave as a weak partial agonist at high concentrations, as observed elsewhere (see Mechoulam et al., 2007; McPartland et al., 2014). In this study, the primary effect of CBD at CB₁ receptors was negative allosteric modulation at concentrations below 1 μM. The studies by Price et al. (2005) and Baillie et al. (2013) demonstrated that ORG27569 and PSNCBAM-1 paradoxically reduce orthosteric ligand efficacy and potency while increasing orthosteric ligand binding affinity and duration. It is thought that, in general, increased ligand binding results in rapid desensitization of receptors (Price et al., 2005; Ahn et al., 2013). In this study, we did not directly test receptor desensitization or duration of ligand binding. We did, however, estimate ligand co-operativity and found that CBD, unlike ORG27569 and PSNCBAM-1, displayed negative co-operativity for ligand binding (α < 1) (Price et al., 2005; Ahn et al., 2013). ORG27569 and PSNCBAM-1 increase the CB₁ receptor pool at the cell surface and in doing so may potentiate CB₁ receptor signalling (Cawston et al., 2013). In vivo, ORG27569 reduces food intake similar to the CB₁ receptor inverse agonist rimonabant (Gamage et al., 2014). However, the in vivo actions of ORG27569 are CB₁ receptor-independent, suggesting that the in vitro pharmacology of ORG27569 does not correlate with in vivo observations (Gamage et al., 2014). Like ORG27569, CBD may mediate a subset of its in vivo actions through non-CB₁ receptor targets (Campos et al., 2012). For example, the anxiolytic and antidepressant actions of CBD may be 5HT_1A receptor-dependent, while the antipsychotic activity of CBD may be mediated by TRPV1 channels (Bisogno et al., 2001; Russo et al., 2005; Ryberg et al., 2007; Campos et al., 2012). Regardless of whether CBD has alternative targets in vivo, the work shown here demonstrates that CBD can alter the activity of common endocannabinoids and phytocannabinoids at CB₁ receptors and this action is likely to be therapeutically important.