Type 1 Cannabinoid Receptor Ligands Display Functional Selectivity in a Cell Culture Model of Striatal Medium Spiny Projection Neurons*

Background: To understand the differential response to cannabinoids, we examined the functional selectivity of type 1 cannabinoid receptor (CB1) agonists in a cell model of striatal neurons. Results: 2-Arachidonylglycerol, Δ9-tetrahydrocannabinol, and CP55,940 were arrestin2-selective; endocannabinoids and WIN55,212-2 activated Gαi/o, Gβγ, and Gαq; and cannabidiol activated Gαs independent of CB1. Conclusion: Cannabinoids displayed functional selectivity. Significance: CB1 functional selectivity may be exploited to maximize therapeutic efficacy. Modulation of type 1 cannabinoid receptor (CB1) activity has been touted as a potential means of treating addiction, anxiety, depression, and neurodegeneration. Different agonists of CB1 are known to evoke varied responses in vivo. Functional selectivity is the ligand-specific activation of certain signal transduction pathways at a receptor that can signal through multiple pathways. To understand cannabinoid-specific functional selectivity, different groups have examined the effect of individual cannabinoids on various signaling pathways in heterologous expression systems. In the current study, we compared the functional selectivity of six cannabinoids, including two endocannabinoids (2-arachidonyl glycerol (2-AG) and anandamide (AEA)), two synthetic cannabinoids (WIN55,212-2 and CP55,940), and two phytocannabinoids (cannabidiol (CBD) and Δ9-tetrahydrocannabinol (THC)) on arrestin2-, Gαi/o-, Gβγ-, Gαs-, and Gαq-mediated intracellular signaling in the mouse STHdhQ7/Q7 cell culture model of striatal medium spiny projection neurons that endogenously express CB1. In this system, 2-AG, THC, and CP55,940 were more potent mediators of arrestin2 recruitment than other cannabinoids tested. 2-AG, AEA, and WIN55,212-2, enhanced Gαi/o and Gβγ signaling, with 2-AG and AEA treatment leading to increased total CB1 levels. 2-AG, AEA, THC, and WIN55,212-2 also activated Gαq-dependent pathways. CP55,940 and CBD both signaled through Gαs. CP55,940, but not CBD, activated downstream Gαs pathways via CB1 targets. THC and CP55,940 promoted CB1 internalization and decreased CB1 protein levels over an 18-h period. These data demonstrate that individual cannabinoids display functional selectivity at CB1 leading to activation of distinct signaling pathways. To effectively match cannabinoids with therapeutic goals, these compounds must be screened for their signaling bias.

Modulation of type 1 cannabinoid receptor (CB 1 ) activity has been touted as a potential means of treating addiction, anxiety, depression, and neurodegeneration. Different agonists of CB 1 are known to evoke varied responses in vivo. Functional selectivity is the ligand-specific activation of certain signal transduction pathways at a receptor that can signal through multiple pathways. To understand cannabinoid-specific functional selectivity, different groups have examined the effect of individual cannabinoids on various signaling pathways in heterologous expression systems. In the current study, we compared the functional selectivity of six cannabinoids, including two endocannabinoids (2-arachidonyl glycerol (2-AG) and anandamide (AEA)), two synthetic cannabinoids (WIN55,212-2 and CP55,940), and two phytocannabinoids (cannabidiol (CBD) and ⌬ 9 -tetrahydrocannabinol (THC)) on arrestin2-, G␣ i/o -, G␤␥-, G␣ s -, and G␣ q -mediated intracellular signaling in the mouse STHdh Q7/Q7 cell culture model of striatal medium spiny projection neurons that endogenously express CB 1 . In this system, 2-AG, THC, and CP55,940 were more potent mediators of arres-tin2 recruitment than other cannabinoids tested. 2-AG, AEA, and WIN55,212-2, enhanced G␣ i/o and G␤␥ signaling, with 2-AG and AEA treatment leading to increased total CB 1 levels. 2-AG, AEA, THC, and WIN55,212-2 also activated G␣ q -dependent pathways. CP55,940 and CBD both signaled through G␣ s . CP55,940, but not CBD, activated downstream G␣ s pathways via CB 1 targets. THC and CP55,940 promoted CB 1 internaliza-tion and decreased CB 1 protein levels over an 18-h period. These data demonstrate that individual cannabinoids display functional selectivity at CB 1 leading to activation of distinct signaling pathways. To effectively match cannabinoids with therapeutic goals, these compounds must be screened for their signaling bias.
Cannabinoids mediate their effects through several receptors, including the type 1 cannabinoid receptor (CB 1 ), which has been studied intensively for its neuromodulatory activity. Many cannabinoids, including 2-AG, AEA, and THC, induce analgesic responses, and their use for chronic and acute pain conditions such as arthritis and migraine is being actively explored (2)(3)(4). Cannabinoids evoke hypolocomotive responses via CB 1 and may be useful in the treatment of movement disorders such as tremor, ataxia, Tourette syndrome, Parkinson disease, and Huntington disease (5). CBD, acting independently of CB 1 , has been shown to have therapeutic potential as an anti-epileptic and anti-inflammatory agent (6 -8). Modulation of CB 1 activity in the central nervous system and periphery also affects appetite and glucose and fat metabolism (9). Cannabinoids may play a therapeutic role in the management of metabolic syndrome, diabetes, and lipodystrophies (9). Additionally, it is important to understand how psychoactive cannabinoids, such as THC, affect neuronal activity via CB 1 and other effectors within the context of substance abuse and addiction (1, 10 -12).
Cannabinoids differ in their affinity for CB 1 and their potency and efficacy of action via CB 1 (1,2,13). The classical view of CB 1 activation was that a correlation exists between binding affinity at CB 1 and the potencies of cannabinoids to induce the in vivo tetrad responses of anti-nociception, hypoactivity, hypothermia, and catalepsy (2,(13)(14)(15). Because of this correlation, individual cannabinoids were expected to be similarly potent in all four tetrad responses (2). However, this is not the case for many cannabinoids. THC and WIN, for example, are more potent inducers of hypolocomotion than of catalepsy or hypothermia (16,17). Similarly, AEA and THC differ in their potencies for anti-nociception and hypolocomotion in the ICR strain of mice (18) and their ability to evoke tolerance and dependence in fatty acid amide hydrolase (FAAH) knock-out mice (19). Long et al. (20) and Schlosburg et al. (21) observed that selective blockade of AEA or 2-AG catabolism results in sustained analgesia or disruption of analgesia and cross-tolerance to other CB 1 agonists, respectively. CBD, unlike other cannabinoids, does not evoke the tetrad responses (8). CBD demonstrates low affinity for CB 1 , and the in vivo effects of CBD, including its anti-inflammatory properties, appear to be CB 1independent (6 -8, 12). Differences in the potency and efficacy of cannabinoids to evoke various responses in vivo may be exploited in the application of these compounds as therapies. Although these distinctions may result from pharmacokinetic differences, it is also possible that in vivo responses to cannabinoids may be mediated through the different effects of individual cannabinoids.
Distinct agonists appear to modulate the signaling specificity of CB 1 through the coupling of different G proteins (2,22). CB 1 agonist-selective coupling to G␣ i , G␣ s , and G␣ q has been demonstrated in cell lines overexpressing CB 1 treated with WIN, CP, and other synthetic cannabinoids (23)(24)(25). The potency of AEA, CP, WIN, and other cannabinoids to stimulate [ 35 S]GTP␥S has been evaluated in rat cerebellar membranes (10) and N1E-115 cells overexpressing CB 1 (14). In these and subsequent studies, WIN and CP were found to be full agonists of G␣ i/o , whereas AEA and, to a lesser extent, THC were partial agonists (2,(13)(14)(15). It is thought that WIN and CP stabilize functionally different active conformations of CB 1 resulting in a differential interaction and activation of G proteins (26,27). In silico modeling of CB 1 -cannabinoid interactions suggests that each cannabinoid interacts with a different subset of residues on the third and fourth transmembrane helices of CB 1 (28 -30). Based on these data, Varga et al. (29) proposed that ligandspecific changes in CB 1 conformation may enhance the binding of different G proteins (e.g. G␣ i/o versus G␣ s ) or arrestins, which would in turn facilitate the activation of different signaling pathways downstream of CB 1 . Glass and Northup (22) used Sf9 cell membrane preparations containing CB 1 and various G proteins to differentiate the G␣ i -and G␣ o -mediated effects of CB 1. In their study, the synthetic cannabinoid HU210, WIN, and AEA were full agonists of G␣ i , whereas THC acted as a partial agonist, and WIN, AEA, and THC were all partial agonists of G␣ o relative to HU210 (22). Similar to Varga et al. (29), Glass and Northup (22) concluded that distinct agonists induce unique receptor conformations resulting in ligand-specific CB 1 -dependent G protein signaling. The data presented in their studies suggest that the pharmacological activity of cannabinoids acting through G proteins depends on their affinity for CB 1 as well as the signaling bias of specific cannabinoids.
Beyond G proteins, the recruitment of arrestin1 and -2 to CB 1 has also been examined (31)(32)(33). These studies report that CB 1 interacts weakly with arrestin2, which facilitates internalization upon stimulation with WIN or CP in HEK cells, AtT20 immortalized mouse anterior pituitary cells, and U2OS human osteosarcoma cells stably expressing CB 1 (31)(32)(33). WIN and CP have been shown to be differentially efficacious activators of tyrosine hydroxylase transcription, ERK1/2 phosphorylation, and JNK activation in neuroblastoma cells (14,34,35). Although these observations were not related to agonist-specific coupling, the authors (14,34,35) suggest that the differences between WIN and CP support functional selectivity of cannabinoids at CB 1 . Other cannabinoids, such as CBD, have been shown to have some CB 1 modulatory activity but act largely via CB 1 -independent effectors (6 -8). To complicate matters, the functional selectivity of cannabinoid ligands may be cell type-specific because reports of efficacy have varied across model systems and tissues (2). Therefore, individual cannabinoids may stabilize specific CB 1 receptor conformations, resulting in a cell-and tissue-specific response (32,33). This is interesting because it may be possible for cannabinoids to be designed that bias receptor signaling toward desirable effects and away from undesirable ones.
In this study, we sought to characterize the ligand bias of several cannabinoid ligands in an in vitro model of neurons that express CB 1 . To directly compare cannabinoid ligand bias, the downstream functional selectivity of two compounds from three classes of cannabinoids was examined in the STHdh Q7/Q7 cell culture model of striatal medium spiny projection neurons. This cell culture model was chosen to characterize cannabinoid ligand bias because these cells model the major output of the indirect motor pathway of the striatum where CB 1 levels are highest relative to other regions of the brain (36,37). STHdh Q7/Q7 cells endogenously express CB 1 and FAAH (36,37), as well as the dopamine D 2 receptor enkephalin and other markers of striatal neurons, making this in vitro model system ideally suited to studying cannabinoid signaling in a physiologically relevant context. The endocannabinoids AEA and 2-AG, the phytocannabinoids CBD and THC, and the synthetic cannabinoids WIN and CP were compared for their ability to activate arrestin2 (␤-arrestin1)-, G␣ i/o -, G␣ s -, and G␣ q -dependent pathways in STHdh Q7/Q7 cells. Based on the existing in vitro and in vivo data for cannabinoid ligands, we hypothesized that endocannabinoids, phytocannabinoids, and synthetic cannabinoids would differentially bias CB 1 -dependent signaling. chased from Sigma-Aldrich (Oakville, ON). Pertussis and cholera toxins (PTx and CTx) were purchased from Sigma-Aldrich. The G␤␥ modulator gallein was purchased from EMD Millipore (Billerica, MA). Cannabinoids and gallein were dissolved in dimethyl sulfoxide (final concentration of 0.1% in assay media for all assays) and added directly to the media at the concentrations and times indicated. No effects of vehicle alone were observed compared with assay media alone. PTx and CTx were dissolved in dH 2 O (50 ng/ml) and added directly to the media 24 h prior to cannabinoid treatment. Pretreatment of cells with PTx and CTx inhibits G␣ i/o and G␣ s , respectively (38). In the case of CTx, this occurs via down-regulation of G␣ s following ADP-ribosylation (38,39). All experiments included a vehicle treatment control.
STHdh Q7/Q7 cells are a cell line derived from the conditionally immortalized striatal progenitor cells of embryonic day 14 C57BlJ/6 mice (Coriell Institute, Camden, NJ) (36). Cells were grown at 33°C, 5% CO 2 in DMEM supplemented with 10% FBS, 2 mM L-glutamine, 10 4 units/ml penicillin-streptomycin, and 400 g/ml Geneticin. Twenty-four hours of serum deprivation promotes the differentiation of STHdh Q7/Q7 cells into an adult neuron-like phenotype characterized by increased neurite outgrowth, GABA release, and increased expression of CB 1 , dopamine D 2 receptors, preproenkephalin, and dopamine and cAMP-related phosphoprotein 32 kDa (DARPP-32), typical of mature medium spiny projection neurons of the indirect motor pathway of the striatum (36,37,40). These cells are ideally suited for the characterization of cannabinoid ligand bias in vitro because they model a neuronal cell type that expresses CB 1 at high levels compared with other cell types in the central nervous system. The striatum is a major site of action of centrally acting cannabinoid-based therapies (41,42).
Bioluminescence Resonance Energy Transfer 2 (BRET 2 )-Direct interactions between CB 1 and arrestin2 were quantified via BRET 2 (46). Cells were grown in a 6-well plate and transfected with the indicated GFP 2 and Rluc constructs. Forty-eight hours post-transfection, the cells were washed twice with cold 0.1 M PBS and suspended in 90 l of 0.1 M PBS supplemented with glucose (1 mg/ml), benzamidine (10 mg/ml), leupeptin (5 mg/ml), and a trypsin inhibitor (5 mg/ml). Cells were dispensed into white 96-well plates and treated as indicated (PerkinElmer Life Sciences). Coelenterazine 400a substrate (50 M; Biotium, Hayward, CA) was added, and light emissions were measured at 405 nm (Rluc) and 510 nm (GFP 2 ) using a Luminoskan Ascent plate reader (Thermo Scientific, Waltham, MA) with an integration time of 10 s and a photomultiplier tube voltage of 1200 V. BRET efficiency (BRET eff ) was determined using previously described methods (47) such that Rluc alone was used to calculate BRET min and the Rluc-GFP 2 fusion protein was used to calculate BRET max .
Fluorescence Resonance Energy Transfer (FRET)-Receptor dimerization was visually assessed via FRET according to the methods of Wu et al. (48). Cells were grown in a 6-well plate and transfected with the indicated GFP 2 and RFP constructs. Fortyeight hours post-transfection cells were moved to coverslips and grown for an additional 24 h. Cells were treated as indicated and visualized on a Zeiss 510 upright laser scanning microscope with 20ϫ and 63ϫ objective lenses. Images were captured using Zen Image Capture 2009 edition (Carl Zeiss Canada). The following excitation/emission filters were used to directly visualize fluorescence: for GFP 2 , 492 nm/510 nm; and for RFP, 543 nm/565 nm. For FRET, GFP 2 was excited 488 nm, separated by a 488/564 dichromic mirror, with emitted fluorescence detected between 502 and 651 nm (48). To measure the endogenous association between CB 1 and arrestin, paraformaldehyde-fixed cells were used for the immunocytochemical detection of CB 1 with a C-terminal CB 1 primary antibody (1:500; catalog No. 10006590, Cayman Chemical Co., Ann Arbor, MI) and Alexa Fluor 488 secondary antibody (donor) and detection of arrestin2 with an arrestin1/2 primary antibody (1:250; Santa Cruz Biotechnology, Santa Cruz, CA) and Cy 3 secondary antibody (acceptor), as described by Knowles et al. (49). Cells were grown on coverslips and treated as indicated. Cells were washed with 0.1 M PBS, fixed with 4% paraformaldehyde, and washed three times with 0.1 M PBS for 5 min each. Cells were incubated with blocking solution (0.1 M PBS and 5% normal goat serum in dH 2 O) for 1 h at room temperature. Cells were incubated with primary antibody solutions directed against C-CB 1 (1:500) and arrestin1/2 (1:250) diluted in antibody dilution buffer (0.1 M PBS, 1% (w/v) BSA, in dH 2 O) overnight at 4°C. Cells were washed three times with 0.1 M PBS for 5 min each. Cells were incubated in Alexa Fluor 488 (1:500) and Cy 3 (1:500) (Rockland Immunochemicals, Gilbertsville, PA) for 1 h at room temperature. Cells were then washed three times with 0.1 M PBS for 5 min each. Microscopy and FRET were then conducted using the same methodology described for FRET in transfected cells. The specificity of the C-terminal CB 1 and arrestin1/2 primary antibodies was confirmed using blocking peptide controls (1:500) (Cayman Chemical Co. and Santa Cruz Biotechnology). FRET efficiency was calculated in ImageJ by dividing the average pixel intensity at 565 nm for any given image by the intensity at 522 nm for that image after background subtraction. FRET was represented visually by mapping a pseudo-color lookup table (16 colors, ImageJ) onto the resulting image (48).
In-and On-cell TM Western Analyses and Immunocytochemistry-For On-cell TM Western analyses, cells were fixed for 10 min at room temperature with 4% paraformaldehyde and washed three times with 0.1 M PBS for 5 min each. Cells were incubated with blocking solution (0.1 M PBS and 5% normal goat serum in dH 2 O) for 1 h at room temperature. Cells were incubated with primary antibody solutions directed against N-CB 1 (1:500; catalog No. 101500, Cayman Chemical Co.) diluted in antibody dilution buffer (0.1 M PBS and 1% (w/v) BSA in dH 2 O) overnight at 4°C. Cells were washed three times with 0.1 M PBS for 5 min each. Cells were incubated in IR CW800 dye (1:500; Rockland Immunochemicals) for 1 h at room temperature. Cells were then washed three times with 0.1 M PBS for 5 min each. Cells were allowed to air-dry overnight. On-cell TM data were then collected using the Odyssey imaging system and software (version 3.0; Li-Cor, Lincoln, NE). These data represent the fraction of CB 1 detected on the plasma membrane. The same cells were then used to quantify total CB 1 protein levels using the In-cell TM Western technique. The Oncell TM CB 1 levels were divided by the In-cell TM (total) CB 1 levels to determine the fraction plasma membrane CB 1 . In-cell TM Western analyses and immunocytochemistry were conducted as described above except that 0.3% Triton X-100 was added to the blocking and antibody dilution solutions. Primary antibody solutions were: N-CB 1  were chosen because phosphorylation at these sites demonstrates activation of the ERK, PI3K/Akt, CREB, and G␣ q pathways, respectively. Secondary antibody solutions were IR CW700 and IR CW800 dyes (1:500; Rockland Immunochemicals). Incell TM Western analyses were then conducted using the Odyssey imaging system and software (version 3.0; Li-Cor). All experiments measuring CB 1 included an N-CB 1 blocking peptide (1:500) control, which was incubated with N-CB 1 antibody (1:500). Immunofluorescence observed with the N-CB 1 blocking peptide was subtracted from all experimental replicates.
Quantitative Reverse Transcriptase PCR-RNA was harvested from cells using the TRIzol (Invitrogen) extraction method according to the manufacturer's instruction. Reverse transcription reactions were carried out with SuperScript III reverse transcriptase (ϩRT; Invitrogen) or without (ϪRT) as a negative control in subsequent PCR experiments according to the manufacturer's instructions. Two micrograms of RNA was used per RT reaction. qRT-PCR was conducted using the Light-Cycler system and software (version 3.0; Roche Applied Science). Reactions were composed of a primer-specific concen-tration of MgCl 2 (Table 1), forward and reverse primers at 0.5 M each (Table 1), 2 l of LightCycler FastStart SYBR Green I reaction mix, and 1 l of cDNA to a final volume of 20 l with dH 2 O (Roche Applied Science). The PCR program was as follows: 95°C for 10 min, 50 cycles of 95°C 10 s, a primer-specific annealing temperature (Table 1) for 5 s, and 72°C for 10 s. Experiments always included sample-matched ϪRT controls, a no-sample dH 2 O control, and a standard control containing product-specific cDNA of a known concentration. cDNA abundance was calculated by comparing the cycle number at which a sample entered the logarithmic phase of amplification (crossing point) with a standard curve generated by amplification of cDNA samples of known concentration (LightCycler software, version 4.1; Roche Applied Science). qRT-PCR data were normalized to the expression of ␤-actin (50).
Statistical Analyses-These were conducted by one-or twoway analysis of variance (ANOVA), as indicated, using GraphPad (version 5.0, Prism). Post-hoc analyses were performed using 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 Ͻ 0.001, Ͻ 0.01, or Ͻ 0.05, as indicated, and all results are reported as the mean Ϯ S.E. from at least four independent experiments. To improve the readability of the data, many of the figures are subdivided as endocannabinoids (AEA and 2-AG), phytocannabinoids (CBD and THC), and synthetic cannabinoids (CP and WIN).

Interactions between CB 1 and Arrestin2 (␤-arrestin1) Are
Ligand-specific-Initially we wanted to determine whether the interaction between CB 1 and arrestin2 differed among CB 1 agonists. To do this, BRET eff was measured between CB 1 -GFP 2 and arrestin2-Rluc. Arrestin2 was chosen because it is endogenously expressed by STHdh Q7/Q7 cells (Fig. 1A). The amount of donor and acceptor plasmid used and the ratio of donor to acceptor plasmids were optimized using a BRET saturation curve at 400 ng of CB 1 -GFP 2 to 200 ng of arrestin2-Rluc (2:1; Fig. 1B). Basal BRET eff between CB 1 -GFP 2 and arrestin2-Rluc was ϳ0.2 and greater than BRET eff between HERG-GFP 2 and arrestin2-Rluc (Fig. 1B). We also verified that BRET eff was independent of time and plasmid expression level for CB 1 -GFP 2 and arrestin2-Rluc (Fig. 1C). Cells were treated with 1 M AEA, 2-AG, CBD, THC, CP, or WIN or with 500 nM O-2050, for 0 -30 min (Fig. 2, A-C). Treatment with AEA, 2-AG, THC, CP, or WIN increased BRET eff within 10 min compared with vehicle treatment, and BRET eff was stable over 30 min (Fig. 2, A-C). CBD and O-2050 treatment did not change BRET eff relative to vehicle treatment. In addition, BRET eff between CB 1 -GFP 2 and arrestin2-Rluc was greater in cells treated with 2-AG compared with AEA by 15 min, with THC compared with CBD by 10 min, and with CP compared with WIN by 5 min (Fig. 2, A-C). The ligand-specific differences in CB 1 -arrestin2 association were further analyzed by measuring BRET eff in cells treated with 0.01-5.00 M AEA, 2-AG, CBD, THC, CP, or WIN in the absence or presence of 500 nM O-2050 for 10 ( Fig. 2, D-F) or 30 min (Fig. 2, G-I). At 10 min, BRET eff between CB 1 -GFP 2 and arrestin2-Rluc was not different in AEA-and 2-AG-treated cells (Fig. 2D), whereas THC and CP were more potent and efficacious ligands than CBD and WIN, respectively (Fig. 2, E and F). At 30 min, 2-AG was a more efficacious ligand than AEA (Fig. 2G). As observed at 10 min, THC and CP were more potent and efficacious ligands than CBD and WIN, respectively (Fig. 2, H and I). AEA-, 2-AG, THC-, WIN-, and CP-mediated recruitment of arrestin2 to CB 1 was inhibited by co-treatment of cells with the CB 1 antagonist O-2050, as demonstrated by a significant rightward shift in the BRET eff dose-response curves (Fig. 2, D-I). The EC 50 , Hill slope, and E max values generated from these dose-response relationships were also compared ( Table 2). THC was more potent than AEA and WIN at 10 min and more potent than AEA, 2-AG, and WIN at 30 min ( Table  2). The flat dose-response relationship observed with CBD demonstrates that this ligand had very little effect on the interaction between CB 1 -GFP 2 and arrestin2-Rluc because the basal BRET eff is not significantly different from the E max (Fig. 2, E and H). 2-AG (30 min), THC (30 min), and CP (10 and 30 min) were more efficacious ligands and CBD (10 and 30 min) less efficacious than AEA for BRET eff between CB 1 -GFP 2 and arrestin2-Rluc ( Table 2). The BRET eff E max values were greater at 30 min than at 10 min when cells were treated with 2-AG or THC (Table 2). No statistically significant changes in the Hill slope were observed. Based on these data, we concluded that 1) with the exception of CBD, each ligand promoted interactions between CB 1 and arrestin2; 2) 2-AG, THC, and CP displayed higher maxima than the other cannabinoid ligands tested for enhancing CB 1 -arrestin2 interactions; and 3) in the assay, THC and CP were more potent than the other cannabinoid ligands tested for enhancing CB 1 -arrestin2 interactions. A final concentration of 1 M was used for all subsequent experiments because this dose consistently produced a response that approximated the E max observed for BRET eff for all cannabinoids tested. Because BRET assays quantify the level of interaction between two proteins but do not provide data on the localization of protein complexes, FRET analyses were conducted to determine the localization CB 1 and arrestin2 complexes within STHdh Q7/Q7 cells in the presence of the cannabinoids studied. FRET was used to study the interaction between CB 1 -GFP 2 and arrestin2-RFP or endogenous CB 1 and arrestin2 detected via fluorescent antibodies. A photobleaching experiment was conducted as a control for FRET (48). Cells were transfected with CB 1 -GFP 2 and arrestin2-RFP. As expected, direct excitation of RFP at 543 nm for 5 min eliminated the fluorescent signal at 565 nm in a small, cytoplasmic region of interest, and the GFP 2 signal in that area was enhanced, whereas the RFP and GFP 2 signals in a non-photobleached region of interest were unchanged (Fig. 3, A and B) (48). The specificity of the anti-CB 1 and anti-arrestin1/2 antibodies was analyzed via immunohistochemistry in the absence and presence of CB 1 -and arrestin1/2 antibody-blocking peptides (Fig. 3C). Fluorescence intensity was ϳ60-fold greater than in the absence of blocking peptide for both CB 1 and arrestin1/2 antibodies (Fig. 3D) FRET was qualitatively higher in cells treated with all cannabinoids tested (1 M, 30 min) except CBD, indicating that interactions between CB 1 and arrestin2 had increased in transfected cells overexpressing CB 1 -GFP 2 and arrestin2-RFP (Fig. 4A) and cells endogenously expressing CB 1 and arrestin2 (Fig. 4B). Quantification of total FRET for cells expressing CB 1 -GFP 2 and arres-tin2-RFP revealed that FRET was greater in cells treated with 1 M AEA, 2-AG, THC, CP, or WIN for 30 min than in cells treated with vehicle (Fig. 4C, dotted line). Similarly, total FRET FIGURE 1. Optimization of BRET 2 between arrestin2 (␤-arrestin1) and CB 1 in STHdh Q7/Q7 cells. A, representative image demonstrating that arrestin2, and not arrestin3, is expressed in STHdh Q7/Q7 cells. PCR products were amplified via qRT-PCR and resolved on agarose gels. B, cells were transfected with varying amounts of donor (CB 1 -Rluc, HERG-Rluc, and arrestin2-Rluc) and acceptor (CB 1 -GFP 2 , HERG-GFP 2 , and arrestin-GFP 2 ) plasmids, and BRET eff was determined. The amount of DNA transfected was kept constant (800 ng) by the addition of pcDNA3.1. The non-interacting HERG receptor and Rluc alone were used as negative controls. GFP 2 -Rluc was used as a positive control. *, p Ͻ 0.001 compared with HERG-Rluc, arrestin2-GFP 2 , arrestin2-Rluc, and HERG-GFP 2 as determined via two-way ANOVA followed by Bonferroni's post-hoc test. n ϭ 6. C, cells were transfected with arrestin2-Rluc and CB 1 -GFP 2 or arrestin2-Rluc and HERG-GFP 2 plasmids, and BRET eff was measured (time 0 h ϭ 48 h post-transfection). *, p Ͻ 0.001 compared with arrestin2-Rluc and HERG-GFP 2 as determined via one-way ANOVA followed by Tukey's post-hoc test; n ϭ 6.
between Alexa Fluor 488-conjugated antibodies (CB 1 ) and Cy 3conjugated antibodies (arrestin2) was greater in cells treated with 1 M AEA, 2-AG, THC, CP, or WIN for 30 min than in cells treated with vehicle (Fig. 4C, solid line). Total FRET between Alexa Fluor 488 and Cy 3 was reduced in cells treated with 1 M CBD for 30 min relative to vehicle-treated cells (Fig.  4C). We also observed that total FRET was greater in cells treated with 2-AG (Alexa Fluor 488 and Cy 3 only), THC, or CP and less in cells treated with CBD than in cells treated with AEA ( Fig. 4C). At the plasma membrane, FRET was greater in cells treated with AEA, 2-AG, or WIN (Alexa Fluor 488 and Cy 3 only) and less in cells treated with CBD (Alexa Fluor 488 and Cy 3 only) or CP compared with vehicle-treated cells (Fig. 4D).
Moreover, FRET was reduced in cells treated with CBD, WIN (CB 1 -GFP 2 and arrestin2-RFP only), or CP relative to AEAtreated cells (Fig. 4D). Within the cytoplasm, FRET was greater in cells treated with AEA (CB 1 -GFP 2 and arrestin2-RFP only), 2-AG, THC, WIN (CB 1 -GFP 2 and arrestin2-RFP only), or CP and less in cells treated with CBD (Alexa Fluor 488 and Cy 3 only) than in vehicle-treated cells (Fig. 4E). FRET within the cytoplasm was also greater in 2-AG (CB 1 -GFP 2 and arrestin2-RFP only)-, THC-, and CP-treated cells and less in CBD-treated cells (Alexa Fluor 488 and Cy 3 only) than in AEA-treated cells (Fig. 4E). A further comparison of FRET between the plasma membrane (Fig. 4D) and cytoplasm (Fig. 4E) demonstrates although 2-AG, THC, and CP all enhanced arrestin2 recruit- ment to CB 1 to a greater extent than other ligands tested, THC and CP biased CB 1 -arrestin2 complexes toward internalization to a greater extent than 2-AG.
Analysis of FRET at 10 min revealed no significant difference between 10 and 30 min of treatment with any cannabinoid tested (data not shown). Quantification of FRET in the nucleus   and dendrites of cells revealed no difference among treatments (data not shown). The cell diameter, cell area, projection length, and projection number were not different between treatment groups (n ϭ 50; data not shown). Overall, these data demonstrate that THC and CP appear to bias CB 1 toward arrestin2mediated internalization to a greater degree than the other cannabinoid ligands tested. Cannabinoid Ligands Biased Intracellular Signaling-Because we had observed ligand-specific differences in CB 1 -arrestin2 interactions, we wanted to determine whether intracellular signaling differed among cannabinoids. Treatment with AEA or 2-AG for 10 min resulted in a PTx-and O-2050sensitive increase in ERK phosphorylation compared with vehicle (Fig. 5, A and B). By 30 min, AEA-mediated ERK phosphorylation was not detectable, whereas O-2050-sensitive ERK phosphorylation persisted in 2-AG-treated cells and was no longer PTx-sensitive compared with vehicle-treated cells or with treatment at 10 min (Fig. 5, A and B). AEA and 2-AG treatment did not change the levels of CREB phosphorylation (Fig. 6, A and B). AEA and 2-AG treatment did increase O-2050 and PTx-sensitive Akt phosphorylation at 10 and 30 min com-pared with vehicle treatment (Fig. 7, A and B). AEA and 2-AG also increased the CB 1 -and G␣ q -dependent phosphorylation of PLC␤3 at 10 min compared with vehicle treatment (Fig. 8, A  and B). Treatment with CBD did not change ERK, Akt, or PLC␤3 phosphorylation but did increase CTx-sensitive CREB phosphorylation at 30 min compared with vehicle treatment and compared to treatment for 10 min with CBD (Figs. 5C, 6C, 7C, and 8C). CBD-mediated CREB phosphorylation was CB 1independent because it was not inhibited by O-2050. Therefore, CBD may enhance CREB activation via other cannabinoid receptors, GPCRs, or GPCR-independent mechanisms (7,12). Like 2-AG, THC increased CB 1 -and G␣ q -dependent phosphorylation of PLC␤3 at 10 min and CB 1 -dependent, PTx-insensitive ERK phosphorylation at 30 min compared with vehicle treatment and treatment at 10 min, and it did not alter CREB phosphorylation (Figs. 5D, 6D, 7D, and 8D). Unlike 2-AG, THC treatment did not increase ERK phosphorylation at 10 min or Akt phosphorylation at 10 and 30 min (Figs. 5D, 6D, 7D, and 8D). WIN and CP treatment for 10 min resulted in a PTx-and O-2050-sensitive increase in ERK phosphorylation (Fig. 5, E and F) and CB 1 -and G␣ q -dependent phosphorylation of PLC␤3 at 10 min (Fig. 8, E and F), relative to vehicle treatment. As with AEA, ERK phosphorylation was not detected in cells treated with WIN for 30 min (Fig. 5E). CP treatment for 30 min resulted in CB 1 -dependent, PTx-insensitive ERK phosphorylation compared with vehicle treatment, as observed with 2-AG and THC (Fig. 5F). WIN treatment did not alter CREB phosphorylation, but CP treatment for 30 min did increase O-2050and CTx-sensitive CREB phosphorylation relative to vehicle treatment and 10-min treatment with CP (Fig. 6, E and F). CPdependent CREB phosphorylation was less than CBD-dependent CREB phosphorylation (Fig. 6, C and F). Both WIN and CP treatment for 10 and 30 min increased Akt phosphorylation compared with vehicle treatment but was less than either AEA or 2-AG (Fig. 7, A, B, E, and F). Therefore, AEA, 2-AG, WIN, and CP treatment resulted in G␣ i/o -dependent transient ERK phosphorylation (51), G␣ q -dependent transient PLC␤3 phos-phorylation, and persistent Akt phosphorylation. THC treatment also resulted in G␣ q -dependent transient PLC␤3 phosphorylation. Further, treatment with 2-AG, THC, and CP, the ligands that enhanced BRET eff and FRET between CB 1 and arrestin2 more than the other cannabinoids tested, resulted in persistent (30 min) G␣ i/o -independent ERK phosphorylation. Finally, CBD and CP treatment enhanced G␣ s -mediated CREB phosphorylation, although CBD did so independent of CB 1 . AEA, 2-AG, THC, WIN, and CP increased the phosphorylation of ERK, CREB, Akt, or PLC␤3 via CB 1 because these effects were blocked by the CB 1 -selective antagonist O-2050 (1). Therefore, the functional selectivity between the cannabinoids tested here is the result of ligand bias at CB 1 receptors.
Sustained and G␣ i/o -independent ERK phosphorylation occurs via arrestin2 (47). We tested this possibility by treating cells overexpressing an arrestin2 dominant negative mutant (arrestin2 V53D) with 1 M 2-AG, THC, or CP with or without 50 ng/ml PTx (Fig. 9, A-C). We observed that PTx-insensitive ERK phosphorylation was sustained at each time point above the levels observed before drug treatment in cells treated with 2-AG (Fig. 9A) and for 12 h in cells treated with THC (Fig. 9B) or CP (Fig. 9C). However, the levels of phosphorylated ERK did not differ from basal levels (0 h) in cells expressing arrestin2 V53D. Based on these data, the sustained ERK phosphorylation observed with 2-AG, THC, and CP occurred via arrestin2-mediated signaling.
We also observed PTx-sensitive Akt phosphorylation in cells treated with AEA or 2-AG for 10 or 30 min. Akt phosphorylation is not commonly associated with the activation of G␣ i/omediated signaling (52) but does occur via G␤␥-dependent activation, which is typically associated with G␣ i/o (53). To determine whether this was occurring in our model system, cells were treated with 1 M AEA or 2-AG with or without 50 ng/ml PTx or 500 nM gallein, the G␤␥ inhibitor (Fig. 7, D and E). Co-treatment of AEA (Fig. 7D)-or 2-AG (Fig. 7E)-treated cells with PTx or gallein prevented Akt phosphorylation over the 18-h time period analyzed. Therefore, AEA-and 2-AG-dependent Akt phosphorylation was mediated by G␣ i/o and G␤␥.
The Functional Selectivity of Cannabinoid Ligands Altered the Expression and Localization of CB 1 Receptors-Arachidonyl-2Ј-chloroethylamide (ACEA), methanandamide, and AEA increased the steady-state levels of CB 1 mRNA and protein via Akt and NF-B (37). Akt activation was observed in AEA-, 2-AG-, WIN-, and CP-treated cells and not in THC-and CBDtreated cells. We hypothesized that this increase in CB 1 levels was unique to those cannabinoids that increased Akt phosphorylation. To test this hypothesis, cells were treated with 1 M AEA, 2-AG, CBD, THC, WIN, or CP with or without 50 ng/ml PTx or CTx or 500 nM O-2050 for 18 h, and CB 1 mRNA levels were quantified relative to ␤-actin (Fig. 10A). AEA and 2-AG, and to a lesser extent WIN, increased CB 1 mRNA levels relative to vehicle treatment, whereas CBD, THC, and CP treatment did not change CB 1 mRNA levels (Fig. 10A). The increase in CB 1 mRNA levels may have been less in WIN-treated cells and absent in CP-treated cells because Akt phosphorylation was lower in WIN-and CP-treated cells relative to AEA-and 2-AGtreated cells (Fig. 7), resulting in insufficient activation of this signaling pathway. In addition, the increase in CB 1 mRNA levels was blocked by treatment with O-2050 or PTx and therefore occurred through CB 1 and G␣ i/o (Fig. 10A). Therefore, AEA, 2-AG, and WIN treatment biased CB 1 signaling toward activation of G␣ i/o signaling, resulting in increased CB 1 mRNA levels.
CBD and CP treatment increased CREB phosphorylation (Fig. 6, C and F). Therefore, we wanted to know whether treatment with CBD or CP would increase preproenkephalin (ppENK) expression, which is known to be CREB-dependent (54,55). ppENK mRNA levels were quantified in cells treated with 1 M AEA, 2-AG, CBD, THC, WIN, or CP with or without 50 ng/ml PTx or CTx or 500 nM O-2050 for 18 h. AEA, 2-AG, and WIN treatment were associated with a CB 1 -dependent decrease in ppENK mRNA levels, whereas CBD treatment increased ppENK mRNA levels compared with vehicle treatment (Fig. 10B). The CBD-mediated increase in ppENK mRNA levels was CB 1 -independent, because it was not inhibited by O-2050 (Fig. 10B). CP treatment did not affect ppENK mRNA levels (Fig. 10B). CBD treatment resulted in greater CREB phosphorylation than CP treatment (Fig. 6, C and F). Therefore, CP treatment may have failed to increase ppENK mRNA levels because the magnitude of CREB phosphorylation was too low. Based on these data, AEA-, 2-AG-, and WIN-dependent activation of G␣ i/o through CB 1 inhibited CREB-mediated gene expression, whereas CBD-mediated, CB 1 -independent activation of G␣ s increased CREB-mediated gene expression.
Increased CB 1 mRNA levels translated to increased CB 1 protein abundance, as determined by On-and In-cell TM Western analyses. Treatment with 1 M AEA or 2-AG resulted in increased CB 1 levels within 3 h compared with the 0 h measurement or with vehicle treatment within the time point, and this increase was still observed at 18 h (Fig. 11A). In contrast, treatment with 1 M THC or CP resulted in decreased CB 1 levels by 6 h (THC) and 12 h (CP) compared with the 0 h mea-surement or vehicle control (Fig. 11, B and C). Treatment with 1 M CBD or WIN did not change CB 1 protein levels (Fig. 11, B and C). CB 1 localization was also analyzed over an 18-h treatment period. The fraction of CB 1 receptors at the membrane of AEA-and 2-AG-treated cells was decreased between 0.5 and 3 h compared with the 0 h time point or vehicle-treated cells, which returned to basal levels by 6 h (Fig. 11D). A decrease in the fraction of CB 1 receptors at the membrane was also observed in THC-and WIN-treated cells between 1 and 12 h (THC) and at 1 h (WIN) compared with the 0 h time point or vehicle-treated cells, which returned to basal levels by 18 h (THC) and 3 h (WIN) (Fig. 11, E and F). Treatment with CBD increased the fraction of CB 1 receptors at the membrane between 3 and 18 h relative to the 0 h time point or vehicletreated cells (Fig. 11E). In contrast, treatment with CP resulted in a sustained decrease in the fraction of CB 1 receptors at the membrane beginning at 0.5 h and persisting to 18 h, as com-   11F). CB 1 receptor localization was also examined via confocal microscopy in cells expressing CB 1 -GFP 2 that were treated with 1 M AEA, THC, or CBD for 10 or 30 min or 1, 3, 6, 12, or 18 h. Similar to the observations made in Fig. 11D, CB 1 -GFP 2 localization shifted from the plasma membrane to the cytoplasm and back to the plasma membrane in cells treated with AEA for 18 h (Fig. 12). In contrast to the AEA-treated cells, CB 1 -GFP 2 was first internalized and subsequently degraded, as indicated by decreased fluorescence, in THC-treated cells (Fig. 12). In CBD-treated cells CB 1 -GFP 2 fluorescence at the plasma membrane gradually increased over the 18 h observation period (Fig. 12). Similarly, CB 1 -GFP 2 was localized to the plasma membrane in cells treated with 500 nM O-2050 for 3 h (Fig. 12). Therefore, although AEA and THC treatment affected CB 1 signaling and internalization, CBD did not affect CB 1 internalization and CBD-mediated signaling was CB 1 -independent.
In conclusion, the endocannabinoids, AEA and 2-AG, facilitated an increase in CB 1 mRNA and protein via G␣ i/o and G␤␥.  WIN also activated G␣ i/o and G␤␥ signaling but to a lesser extent than AEA and 2-AG. Treatment with the phytocannabinoid THC and the synthetic cannabinoid CP did not alter CB 1 mRNA levels but did lead to a decrease in CB 1 protein levels over the 18-h time period analyzed. CP also enhanced G␣ s signaling via CB 1 . CBD-mediated G␣ s signaling occurred independent of CB 1 as observed elsewhere (8).

CB 1 -mediated Intracellular Signaling Was Ligand-specific-
The goal of this study was to compare the CB 1 -mediated functional selectivity of six cannabinoids in a cell line that models striatal medium spiny projection neurons endogenously expressing CB 1 . Each ligand displayed functional selectivity for a subset of intracellular signaling pathways (see summary in Fig.   13). With the exception of CBD-dependent G␣ s signaling, this functional selectivity was CB 1 -dependent.
2-AG, THC, and CP enhanced the interaction between CB 1 and arrestin2 to a greater extent than other cannabinoids tested, suggesting a high degree of interaction between the population of CB 1 and arrestin2 molecules in the in vitro system following 30 min of treatment with these compounds. The relative BRET eff was a conservative estimate of the interaction between CB 1 and arrestin2, because endogenous CB 1 and arres-tin2 would have competed with their labeled counterparts in STHdh Q7/Q7 cells in the BRET assays. These observations differ from previous reports that CB 1 interacts weakly with arrestins in U2OS, CHO, and HEK cell heterologous expression systems treated with WIN or CP for 5 min (33) or 2 h (33). Previous studies also observed that recruitment of arrestins to CB 1 occurs over a wider range of ligand concentrations (1 ϫ 10 Ϫ10 -1 ϫ 10 Ϫ6 M) (32,33) than that observed here (1 ϫ 10 Ϫ8 -1 ϫ 10 Ϫ5 M). The variability between our results and previous reports may reflect differences between the functionality of CB 1 in STHdh Q7/Q7 cells and CB 1 overexpression in U2OS, CHO, or HEK cells (33). In addition, BRET 2 , used in this study, is a more sensitive assay for detecting protein-protein interactions compared with the Tango and PathHunter reporter assays used previously (32). Moreover, previous studies examined the recruitment of arrestin3 (␤-arrestin2) in HEK cells (33) and not arrestin2.
At 1 M, AEA, 2-AG, WIN, and CP biased CB 1 signaling toward G␣ i/o -mediated ERK phosphorylation to a greater degree than other cannabinoids tested. The consequence of this functional selectivity is that transient ERK signaling was enhanced by endocannabinoids compared with other cannabinoids tested, whereas sustained ERK signaling from 10 to 30  min through arrestin2 was enhanced by 2-AG, THC, and CP and not by AEA, CBD, or WIN. Other studies have also reported that transient ERK activation occurs via G␣ i/o in the N18TG2 mouse neuronal cell line (56) and in HEK 293 cells stably expressing CB 1 (51). In our studies CB 1 receptors recruited arrestin2, leading to sustained ERK signaling, whereas previous studies have found that sustained ERK signaling is receptor tyrosine kinase-dependent (56).
AEA and 2-AG activated Akt via G␣ i/o -and G␤␥-dependent pathways. This resulted in increased CB 1 mRNA and protein levels. Although WIN and CP treatment also resulted in G␣ i/odependent ERK phosphorylation and G␤␥-dependent Akt phosphorylation, the magnitude of G␤␥ activation was less following WIN and CP treatment than that observed following treatment with AEA and 2-AG. The net result was that WIN and CP treatment did not lead to significantly increased CB 1 mRNA and protein levels.
In addition to the activation of G␣ i/o , AEA, 2-AG, WIN, THC, and CP enhanced transient G␣ q -coupled PLC␤3 phosphorylation. Few studies have examined direct coupling of CB 1 to G␣ q (25,57). Coupling of CB 1 to G␣ q has been reported in HEK 293 cells stably expressing CB 1 (25) and human trabecular meshwork cells (58). In these studies, the authors observed transient, G␣ q -dependent Ca 2ϩ efflux following stimulation of CB 1 with WIN (25,57), which is an indirect measure of G␣ q coupling. In support of studies that have indirectly measured CB 1 coupling to G␣ q via Ca 2ϩ efflux, the work here measured PLC␤3 activation, which is a direct effect of G␣ q . Because G␣ q signaling may affect cellular function, future studies examining CB 1 signaling should determine whether CB 1 couples to G␣ q in other model systems.
CBD treatment resulted in G␣ S -and CREB-dependent expression of ppENK (7,12). CBD signaling was independent of CB 1 , as demonstrated previously by the inability of the direct antagonist O-2050 to block agonist-dependent G␣ s signaling (8). Although CBD has a relatively low affinity for CB 1 (6), CBD has been shown to act as an agonist and antagonist at the type 2 cannabinoid receptor, an adenosine A 2A agonist, a 5HT 1A agonist, and a modulator of FAAH and monoacylglycerol lipase activity (MAGL) (6 -8, 58). STHdh Q7/Q7 cells express adenosine A 2A and 5HT 1A receptors and FAAH (36,37). The inability of O-2050 to block G␣ s signaling indicates that CBD acted at non-CB 1 targets in STHdh Q7/Q7 cells. In our assays, CBD treatment increased CB 1 levels at the plasma membrane but did not affect CB 1 -dependent signaling through G␣ i/o or G␣ q . CBD is being investigated for its utility as an anti-epileptic (8) and an anti-inflammatory (12) and for its neuromodulatory activities in vivo (7,8). CBD has a relatively safe side effect profile compared with THC and other cannabinoids (7,8). Our work suggests that CBD has little effect on CB 1 -dependent signaling. The fact that we observed that CBD selectively increased CREB-dependent gene expression may also have therapeutic potential in neurodegenerative diseases, where CREB-dependent gene expression is dysregulated (6,8,37).
Unlike CBD, CP-dependent CREB phosphorylation occurred via CB 1 . Although CB 1 does not typically signal through G␣ s , CP may have promoted a conformational change in the receptor that favored G␣ s binding. Alternatively, CP treatment may promote the dimerization of CB 1 with other GPCRs that signal through G␣ s . CB 1 is known to homodimerize with CB 1 and CB 1 splice variants (40) and heterodimerize with other receptors including the dopamine D 2 receptor (37, 43, 59). STHdh Q7/Q7 FIGURE 13. Cannabinoid ligands biased CB 1 -depending signal transduction. Each cannabinoid tested here biased CB 1 signaling toward different pathways. The endocannabinoids AEA and 2-AG promoted G␣ i/o -dependent ERK and Akt activation more effectively than other cannabinoids tested. THC and CP were the most efficacious ligands with regard to CB 1 -arrestin2 interactions, but 2-AG, WIN, and AEA also promoted interactions between CB 1 and arrestin2 above the level observed in vehicle-treated cells. CBD appeared to inhibit the internalization of CB 1 . CBD treatment enhanced G␣ s -dependent CREB phosphorylation independent of CB 1 , whereas CP-dependent CREB phosphorylation occurred through CB 1 . AEA, 2-AG, THC, WIN, and CP promoted G␣ q -dependent transient PLC␤3 activation. cells express dopamine D 2 receptors and CB 1 -D 2 dimerization may contribute to the actions of CP in these cells (59). Together, these data demonstrate that CB 1 is a receptor that couples to multiple G proteins within a single cell type (pleiotropic). CB 1 -mediated Signaling Had Immediate and Sustained Components-In addition to showing cannabinoid ligand bias, CB 1 signaling was also time-dependent. G␣ i/o and G␣ q signaling was transient, being detected at 10 min and returning to basal levels at 30 min. Transient CB 1 -dependent activation of G␣ i/o and G␣ q signaling has been observed elsewhere in HEK 293 cells stably expressing CB 1 (25,51). In contrast, G␣ s signaling was not detected before 30 min. The association between CB 1 and arrestin2 peaked shortly after 10 min for all ligands tested and remained high for 30 min relative to vehicle control, demonstrating that the pleiotropically coupled CB 1 receptor switches between G protein signaling and arrestin signaling within ϳ30 min of ligand administration and that this switch is ligand-specific. Over an 18-h treatment period, THC, CP, and 2-AG treatment resulted in CB 1 receptor internalization (beginning at 30 min), but only THC and CP treatment resulted in decreased CB 1 receptor protein levels (beginning at 12 h). This difference may be due to the higher affinity of THC and CP for CB 1 compared with 2-AG (60), which implies that 2-AG is a "fast-off" cannabinoid relative to THC or CP (30 -32). In vivo, 2-AG is ϳ1000 times more abundant than AEA (60 -62). 2-AG is likely to have a greater effect on the arrestin-mediated recycling of CB 1 between the membrane and intracellular space relative to AEA. Overall, the biased agonism displayed by the six cannabinoids tested here in an in vitro model of striatal neurons supports the hypothesis that individual ligands promote unique conformational changes in the CB 1 receptor leading to functionally divergent intracellular effects such as G protein-coupled signaling, arrestin recruitment, receptor trafficking, and gene expression (22,29,(31)(32)(33).
The Effect of Cannabinoids Is Brain Region-and Agonistspecific-We observed increased CB 1 mRNA and protein levels following 18 h of treatment with 2-AG or AEA in an in vitro cell culture model of striatal neurons. In vivo, CB 1 receptor binding does not differ between FAAH knock-out mice and wild-type littermates in the striatum, hippocampus, or cerebellum when treated with vehicle or AEA for 5 consecutive days (19). Further, CB 1 receptor binding decreases in these brain regions of FAAH knock-out mice treated with THC for 5 consecutive days (19). MAGL knock-out mice and mice treated for 6 days with the MAGL inhibitor JZL184 display decreased CB 1 receptor binding in the cortex, hippocampus, and periaqueductal gray but no difference in the striatum (21). In vivo then, the alteration of CB 1 level depends on the brain region, animal genotype, and duration of treatment, as well as the cannabinoid ligand (19 -21). Inhibition of MAGL (5 days), the principle regulator of 2-AG levels, results in functional antagonism of CB 1 , whereas inhibition of FAAH, the principle regulator of AEA levels, maintains CB 1 signaling (19 -21). Subchronic or chronic exposure to exogenous cannabinoids, such as THC, and high potency cannabinoids decreases CB 1 receptor binding (19 -21). The down-regulation and desensitization of CB 1 receptor following repeated THC or WIN treatment are more pronounced in the hippocampus compared with the striatum (63,64). CB 1 desensitization and arrestin3 (␤-arrestin2) recruitment also vary widely among brain regions (65). FAAH knock-out mice show less CB 1 down-regulation and desensitization following AEA treatment compared with THC-treated FAAH knock-out mice (19). CB 1 internalization is also promoted by WIN more than methanandamide in primary rat hippocampal neurons (66). THC-mediated desensitization is faster than WIN-, CP-, and 2-AG-mediated desensitization in HEK 293 cells stably expressing CB 1 and primary neuronal cultures (67). In contrast, AEA-mediated CB 1 desensitization is slower than that of WIN-, CP-, and 2-AG (67). Endogenous cannabinoids, specifically AEA, may have a different effect on CB 1 levels than other cannabinoids. We do not yet know whether there is a difference in the brain region-specific effect on CB 1 mRNA and protein levels under various treatment regimens in vivo.
There is the potential to exploit biased agonism at CB 1 . Effects such as receptor internalization via arrestin2, G␣ i/o -mediated increases in CB 1 , G␣ q -mediated modulation of Ca 2ϩ release, G␣ s -mediated CREB activation, and CB 1 protein down-regulation could be selected or avoided according to their usefulness in different disease states. This could lead to the development of therapeutics that avoid the psychoactive effects of cannabinoids and promote their neuroprotective effects.