The Expression Level of CB1 and CB2 Receptors Determines Their Efficacy at Inducing Apoptosis in Astrocytomas

Background Cannabinoids represent unique compounds for treating tumors, including astrocytomas. Whether CB1 and CB2 receptors mediate this therapeutic effect is unclear. Principal Findings We generated astrocytoma subclones that express set levels of CB1 and CB2, and found that cannabinoids induce apoptosis only in cells expressing low levels of receptors that couple to ERK1/2. In contrast, cannabinoids do not induce apoptosis in cells expressing high levels of receptors because these now also couple to the prosurvival signal AKT. Remarkably, cannabinoids applied at high concentration induce apoptosis in all subclones independently of CB1, CB2 and AKT, but still through a mechanism involving ERK1/2. Significance The high expression level of CB1 and CB2 receptors commonly found in malignant astrocytomas precludes the use of cannabinoids as therapeutics, unless AKT is concomitantly inhibited, or cannabinoids are applied at concentrations that bypass CB1 and CB2 receptors, yet still activate ERK1/2.


Introduction
Cannabinoids produce the majority of their biological effects by activating two receptors: CB 1 and CB 2 . In healthy brain, CB 1 is expressed by neurons, astrocytes and neural progenitor cells, whereas CB 2 is only expressed by a small population of neurons located in the brain stem [1,2]. Activation of CB 1 and CB 2 receptors regulates fundamental physiological processes, including neurotransmission efficacy, astrocytic function, and the fate and proliferation rate of neural progenitor cells [3,4]. Thus, because these receptors regulate such fundamental physiological processes, extensive effort has been invested toward understanding how cannabinoid compounds activate CB 1 and CB 2 receptors, as well as regulate the signal transduction mechanisms they couple to, with the goal of developing novel therapeutics to treat various neuropathologies [5,6].
One of the most promising therapeutic uses of cannabinoids is linked to their ability to induce apoptosis in tumors, including in astrocytoma subclones grown either in vitro, subcutaneously in immunodeficient mice, or intracranially in rats [7,8]. In fact, these data were so compelling that one clinical trial has already been carried-out, in which patients afflicted with recurrent astrocytomas (grade IV) were treated with cannabinoids stereotactically injected directly into the mass of the malignant tumors [9]. While there are clear results for the use of cannabinoids to treat astrocytomas, the requirement of cannabinoid receptors in mediating this therapeutic effect is still unproven [10]. In fact, evidence suggests that high concentrations of cannabinoids may induce biological effects, including the induction of apoptosis in cells, independently of CB 1 and CB 2 receptors [7,11,12,13]. Thus whether CB 1 and CB 2 receptors, or yet another cannabinoid-like receptor, mediate the therapeutic effects of cannabinoids toward astrocytomas is still unclear.
Other unanswered questions are: Why does CB 1 and CB 2 expression in astrocyomas gradually increase as a function of cancerous transformation? [14,15] Does this increase in receptor expression modify the coupling of CB 1 and CB 2 to signal transduction pathways in astrocytomas? Does this increase in receptor expression modify the efficacy of cannabinoids at inducing apoptosis in astrocytomas? Indeed, while it is known that CB 1 and CB 2 couple to ERK1/2, AKT, JNK, p38 and GSK-3b [8,16,17,18,19], and that both ERK1/2 and AKT mediate the induction of apoptosis by cannabinoids in astrocytomas [8,11,18,20], nothing is known about how changes in CB 1 and CB 2 expression will modify the coupling of these receptors to each kinase, and whether the efficacy with which cannabinoids induce apoptosis is also affected. Answering these questions is important for the following reasons. Most of the studies reporting cannabinoid receptors coupling to specific kinases were performed in heterologous expression systems and the expression level of these receptors both surpass those reached by endogenouslyexpressed receptors and favor promiscuous coupling to effector proteins. One particularly controversial point stems from the ability of cannabinoids to induce apoptosis in astrocytomas when CB 1 and CB 2 receptors also couple to the pro-survival kinase AKT. Therefore, it is important to increase our understanding of how changes in the expression level of CB 1 and CB 2 will affect the coupling of these receptors to specific kinases, as well as the efficacy of cannabinoids at killing astrocytomas.
Here we sought to address the following two questions: Are the therapeutic effects of cannabinoids toward astrocytomas truly mediated by CB 1 and CB 2 receptors? Does the increase in cannabinoid receptor expression known to occur in malignant astrocytomas determine the efficacy of cannabinoids at killing these tumors? To answer these questions, we generated astrocytoma subclones that express low, medium and high levels of either CB 1 or CB 2 , and studied the ability of the cannabinoid agonist CP-55,940 to regulate the activity of various kinases and to induce apoptosis in these tumor cells.

Astrocytoma Subclones Stably Expressing Set Levels of Either CB 1 or CB 2 Receptors
To determine whether the expression level of CB 1 or CB 2 receptors dictate cannabinoid-mediated responses in astrocytomas, we sought to use heterologous gene expression technology to drive the expression of these receptors to precise levels in an astrocytoma cell line that normally lacks these receptors. Specifically, we found that the murine delayed brain tumor (DBT) cell line expresses neither CB 1 nor CB 2 receptors (Table 1). Thus we transfected these cells with IRES constructs containing the gene that encodes either CB 1 or CB 2 receptors in tandem with the gene encoding eGFP, and used FACS to select for the cells that had stably incorporated the construct (see Experimental Procedures). Using this approach, we obtained 27 CB 1 -expressing DBT subclones and 49 CB 2 -expressing DBT subclones. We then randomly selected 24 subclones (twelve CB 1 and twelve CB 2 ) and performed qPCR to find out the relative expression level of CB 1 and CB 2 receptor mRNA in each single cell subclone (Table S1). Based on these data, we then focused our experiments on three CB 1 subclones and three CB 2 subclones that expressed relatively low, medium and high levels of receptor mRNA (Table 1). Please note that all CB 1 -expressing DBT cells were still devoid of CB 2 receptor mRNA, and vice versa. Quantification of CB 1 and CB 2 receptor proteins by radioligand binding showed a correlation between the relative level of mRNA encoding each receptor and the corresponding protein expression level as expressed by the B max value (Table 1). Here it is worth noting that the expression level of CB 1 and CB 2 receptors in the CB 1 -mid and the CB 2 -mid subclones, respectively, lie well within the range of their levels when endogenously expressed by various cell lines and tissues [21,22]. Thus, we generated DBT subclones that stably express set levels of either CB 1 or CB 2 receptors, with expression levels that lie within the range of what is found for cells endogenously expressing these receptors.

Independent of Expression Levels, CB 1 and CB 2 Receptors Similarly Regulate ERK1/2 Signaling
Because previous studies performed on astrocytoma cell lines have implicated ERK1/2 signaling in the therapeutic actions of cannabinoids [23], we sought to determine whether different expression levels of CB receptors would affect their coupling to this signal transduction pathway. To do so, we treated each subclone with CP-55,940 (1 mM, maximally efficacious concentration [17,24]) and measured ERK1/2 phosphorylation by Western blot analysis. We found that CP-55,940 induced a significant increase in ERK1/2 phosphorylation above vehicletreated controls in all six subclones ( Figure 1a). All these responses were receptor-mediated since they were significantly attenuated by either the CB 1 antagonist SR141716A, or the CB 2 antagonist SR144528, respectively ( Figure 1a). Furthermore, CP-55,940 did not affect ERK1/2 phosphorylation in wild-type DBT cells ( Figure S1a). The magnitude and kinetics of CP-55,940-induced ERK1/2 phosphorylation mediated in each subclone were remarkably similar despite a 10-fold difference in receptor expression level between the CB 1 -low and the CB 1 -high subclones, and between the CB 2 -low and the CB 2 -high subclones (Figure 1b,c). The only differences that we noted between CB 1 -and CB 2 -mediated regulation of ERK1/2 phosphorylation were that CP-55,940 induced a slightly slower and sustained increase in ERK1/2 phosphorylation in CB 1expressing subclones (peak at 5 min returning to basal by 20-30 min) compared to CB 2 -expressing subclones (peak at 3 min returning to basal by 10-20 min) (Figure 1b,c). These results show that CB 1 and CB 2 receptors heterologously expressed in DBT cells are functional for they regulate ERK1/2 phosphorylation. They also show that the efficacy of this regulation is largely independent of receptor expression levels and receptor subtype.
The Expression Level of CB 1 and CB 2 Receptors Dictates Their Ability to Regulate AKT, JNK and p38 Signaling CB 1 and CB 2 receptors have been shown to regulate a plethora of kinases, including AKT, JNK, p38 and GSK-3b [8,16,17,18,19]. Thus we sought to assess if CB 1 and CB 2 heterologously expressed in DBT cells also regulate these four kinases. To do so, we treated all DBT subclones with CP-55,940 and assessed the phosphorylation state of these kinases using a Luminex multiplexed immunoassay system. With regard to AKT, three points are noteworthy. First, the CP-55,940-induced increase in AKT phosphorylation correlated with receptor expression levels ( Figure 2a). Specifically, CP-55,940 did not significantly increase AKT phosphorylation in the CB 1 -low subclone, which expresses the lowest level of receptor of all six subclones that we have selected; it induced a small, but significant increase in AKT phosphorylation in CB 2 -low and CB 2 -mid, which express slightly more receptors than the CB 1 -low subclone; and it induced the most pronounced increase in AKT phosphorylation in CB 1 -mid, CB 1 -high and CB 2 -high subclones (see Table 1 for B max values). Second, all responses were mediated by cannabinoid receptors since they were antagonized by either SR141716A or SR144528, respectively, and CP-55,940 did not increase AKT in wild type DBT cells (Figures 2a and S1b). Third, there was a clear difference in the kinetics of AKT activation linked to either CB 1 or CB 2 receptors. Specifically, while the response obtained in the CB 1 -high subclone peaked at 3 min and returned to the basal level at 5 min, the response measured in the CB 2 -high clone remained above 250% of basal during the entire 30 min of the incubation period (Figure 2b,c).
With regard to JNK and p38, CP-55,940 induced a small (150-170% of basal) but significant increase in the phosphorylation state of these kinases only in the CB 2 -high clone that also lasted throughout the 30 min incubation period ( Figure S2a,b). With regard to GSK-3b, CP-55,940 did not significantly affect the phosphorylation state of this kinase in any of the subclones throughout the 30 min incubation time period (data not shown).
Taken together, these results suggest that the different expression levels of CB 1 and CB 2 receptors determine the efficacy with which cannabinoids regulate AKT, JNK and p38 signaling. They also suggest that CB 1 and CB 2 receptors differentially couple to these kinases.

Only Activation of CB 1 Receptors Expressed at Low Levels in DBT Cells Leads to Apoptosis: Involvement of ERK1/2
Several studies suggest that activation of CB 1 or CB 2 is sufficient to induce apoptosis in astrocytomas [8,14], whereas other studies suggest that only high concentrations of cannabinoids kill astrocytomas and furthermore, independently of cannabinoid receptor activation [7,11,12,13]. We directly tested these possibilities by treating the DBT subclones, as well as wild-type DBT cells, with CP-55,940 at either 1 mM (receptor-mediated) or 10 mM (receptor-independent). To assess for cell viability, we used WST-1, an indicator of mitochondrial activity. We found that 1 mM CP-55,940 induced ,50% cell death in the CB 1 -low DBT subclone, whereas 10 mM CP-55,940 induced 35-90% cell death in wild type DBT cells and all six subclones ( Figure 3a). This result suggests that only the activation of CB 1 receptors expressed at low levels in DBT cells leads to significant receptor-dependent killing of these cells. Two additional sets of results support this conclusion. First, SR141617A blocked the 1 mM CP-55,940-induced killing of the CB 1 -low DBT cells, whereas this antagonist did not affect the 10 mM CP-55,940-induced cell death of wild type DBT cells ( Figure 3b). Second, CP-55,940 killed CB 1 -low DBT cells with an EC 50 of 0.56 mM, whereas this compound induced the cell death of wild type DBT cells with an EC 50 of 5 mM (Figure 3c). Interestingly, when comparing the effect of 1 mM CP-55,940 on CB 1 -low DBT cells to the effect of 10 mM CP-55,940 on wild-type DBT, we found that the 10 mM CP-55,940-induced killing of wildtype DBT (receptor-independent) is more rapid than the CB 1mediated cell death (Figure 3d). We then determined whether the CB 1 -dependent and the cannabinoid receptor-independent killing of DBT cells were apoptotic in nature. Specifically, we treated CB 1 -low DBT cells with 1 mM CP-55,940 and wild type DBT cells with 10 mM CP-55,940, and assessed for propidium iodide and annexin V staining (to assess for cell permeability and apoptosis, respectively). Figure 4 shows that after five days, cell death by apoptosis was induced in 78% of the wild-type DBT cells treated with 10 mM CP-55,940 compared to 48% of the CB 1 -low DBT cells treated with 1 mM CP-55,940. Furthermore, in agreement with the data we have obtained with WST-1, 10 mM CP-55,940 killed more wild-type DBT cells by apoptosis (27%) 3 days after treatment compared to CB 1 -low DBT cells treated with 1 mM CP-55,940 (15%) (Figure 4). Together, these data show that both the CB 1 -dependent and the cannabinoid receptor-independent killing of DBT cells are due to apoptosis, but that these processes occur with slightly different time courses.
Because we found that 1 mM CP-55,940 increases ERK1/2 signaling in CB 1 -low DBT cells without affecting the AKT, JNK, p38 and GSKb signaling pathways, and one study suggested that ERK1/2 mediates the apoptotic properties of cannabinoids on astrocytomas [8], we then tested if inhibiting ERK signaling would affect the CB 1 -mediated and the cannabinoid receptor-independent killing of DBT cells. Thus, we pretreated CB 1 -low and wildtype DBT cells with the ERK kinase kinase inhibitor PD98059 before applying either 1 or 10 mM CP-55,940. We found that both responses were partially prevented by this inhibitor (Figure 3b). Thus both the CB 1 -dependent and the cannnabinoid receptorindependent induced apoptosis of DBT cells rely on ERK signaling.
Together, these results show that cannabinoid concentrations known to activate CB 1 and CB 2 receptors (here 1 mM) induce apoptosis in astrocytomas only when these cells express low levels of CB 1 receptors, and that this response is mediated through ERK signaling. Conversely, high concentrations of cannabinoids known to bypass cannabinoid receptors (here 10 mM) induce apoptosis in all DBT subclones, a mechanism that also depends on ERK signaling.

CB 1 -and CB 2 -Induced Apoptosis of DBT Cells Is Uncovered When AKT Is Inhibited
Because increased AKT signaling promotes cell survival [25], we reasoned that activation of this pathway by CB 1 and CB 2 receptors expressed at mid and high levels might explain why 1 mM CP-55,940 was unable to induce apoptosis in DBT subclones. To test this hypothesis, we treated the six subclones of DBT cells with 1 mM CP-55,940 in combination with Inhibitor X, a cell-permeable inhibitor of AKT [26]. Using the WST-1 assay, we found that while this inhibitor had no effect by itself on cell viability, it uncovered the ability of 1 mM CP-55,940 to kill all the DBT subclones that were normally resistant, with the most pronounced response occurring in CB 2 -low DBT cells (Figure 5a). This latter response was due to apoptosis, since 76% of the cells stained positively for annexin V when measured five days after coapplication of 1 mM CP-55,940 and Inhibitor X (Figure 5b). These results suggest that the coupling of CB 1 and CB 2 receptors to the AKT pathway (when these receptors are expressed at mid and high levels) eliminates the ability of cannabinoids to induce apoptosis in DBT cells.

Discussion
While many laboratories have reported that cannabinoids induce apoptosis in various tumor subtypes, including astrocyto- mas, the requirement for CB 1 and CB 2 in mediating this therapeutic effect has not yet been demonstrated. This point is especially relevant when considering that many of these studies used high concentrations of cannabinoids known to bypass CB 1 and CB 2 receptor activation. Here we demonstrate that cannabinoid receptors and ERK1/2 indeed mediate this therapeutic effect, but only when these compounds are applied at submicromolar concentrations and the expression level of these receptors remains low. Conversely, high concentrations of cannabinoids kill all astrocytoma subclones independently of CB 1 , CB 2 and AKT, yet still through a mechanisms involving ERK1/2. We also show that increased expression of CB 1 and CB 2 receptor allows for their coupling to additional kinases, especially AKT, the result of which eliminates the ability of cannabinoids to induce apoptosis even though these receptors still couple to ERK1/2.
One of the most unexpected and interesting findings of our study is that the coupling of CB 1 and CB 2 receptors to AKT, JNK and p38 is dictated by receptor expression levels, whereas their coupling to ERK1/2 is not. What may be the molecular mechanism underlying a gradual coupling to AKT, JNK and p38, while preserving a steady coupling to ERK1/2? Reports on the molecular mechanisms linking cannabinoid receptors to various kinases suggest that these receptors activate AKT through G-proteins, while activating ERK1/2 through EGF receptor transactivation [27]. Based on these reports, a parsimonious interpretation of our results would be that the rate limiting step controlling the efficacy of cannabinoids at stimulating AKT lies in the cannabinoid receptors themselves, whereas the rate limiting step controlling the efficacy of cannabinoids at stimulating ERK lies in EGF receptors, the expression of which is likely to remain constant in the DBT subclones that we generated. Our results may have broader implications when considering the therapeutic efficacy of cannabinoids toward other neuropathologies. For example, epileptic seizures and cerebral ischemia lead to a 3-4 fold increase in the expression of neuronal CB 1 receptors, a response thought to constitute a protective mechanism [28,29],  whereas Huntington disease leads to a 50-70% decrease in the expression of neuronal CB 1 receptors, a decrease thought to participate in the pathogenesis of this neurodegenerative disease [30,31]. Based on the results that we report here, one would predict that up-or down-regulation of neuronal CB 1 receptor expression should not affect the efficacy of cannabinoids at stimulating ERK signaling, whereas it might dictate their ability to modulate the activity of the prosurvival kinase AKT. Thus our study is the first to suggest that changes in cannabinoid receptor expression levels will affect only certain types of signal transductions pathways, several of which are clearly involved in the control of cell survival and death. This result has implications when considering the efficacy of cannabinoids at treating astrocytomas of increasing malignancy, but also when considering other neuropathologies associated with changes in cannabinoid receptor expression.
An additional noteworthy result was the differential coupling of CB 1 and CB 2 receptors to AKT. Indeed, cells expressing similar ranges of CB 1 and CB 2 receptors (CB 1 -mid to CB 1 -high being between 0.7 and 1.3 pmol/mg, and CB 2 -mid to CB 2 -high being between 0.6 and 2.3 pmol/mg) exhibited a clearly different coupling to this prosurvival kinase. How might the same agonist acting at CB 1 and CB 2 differentially stimulate this kinase? CB 1 and CB 2 exhibit significant disparities in their amino acid sequence, particularly in their intracellular domains which interact with G proteins and effector proteins. Indeed, upon their molecular identification, it was evident that while both receptors exhibit similar pharmacological profiles, they only exhibited an overall 44% amino acid identity [32]. We preformed a sequence alignment between mouse CB 1 and mouse CB 2 , and highlighted in red the sequence motifs located in the intracellular loops that may account for the differential interactions with G proteins and found that these regions exhibit a much diminished 29% similarity ( Figure S3). In agreement with this reduced amino acid similarity in intracellular sequences, a recent study showed that the binding of the same agonist to either CB 1 or CB 2 receptors leads to differential activation of G-proteins [33]. Another possible explanation of how the same agonist engaging either CB 1 or CB 2 receptors might differentially activate a signal transduction mechanism is based on studies that were obtained with ''ligandassisted protein structure'' (LAPS) analysis [34,35]. Here the positioning of the tricyclic ring of a specific cannabinoid agonist, AM841 (a close analogue of CP-55,940), within the binding pockets of either CB 1 or CB 2 receptors has been shown to vary considerably, from being parallel to the transmembrane helices in CB 1 to being perpendicular to the transmembrane helices in CB 2 [34,35]. Since agonists engaging the binding pocket of a GPCR will favor a specific R* activation conformation, the same agonist differentially engaging the binding pocket of either CB 1 or CB 2 receptors will likely lead to distinct R* conformations for these receptors, and thus differential G protein coupling. Hence, both sequence disparities in the intracellular loops and the precise positioning of a cannabinoid ligand within the binding pocket of CB 1 and CB 2 receptors could account for the differential activation of a specific kinase.
When considering cannabinoids to treat astrocytomas, many have attempted to isolate the non-psychotropic therapeutic effects ascribed to CB 2 receptor activation, while eliminating the psychotropic and addictive properties ascribed to CB 1 receptor activation [14]. However, since cannabinoids have outstanding LD 50 values (mainly because cannabinoid receptors are barely expressed in brains regions that control vegetative functions) [36], the treatment of tumors with high concentrations of cannabinoids should not be overlooked. In fact, stereotaxic injection of high concentrations of cannabinoids will eradicate a significant portion of C6 astrocytomas inoculated into rodent brains without affecting healthy surrounding tissue or inducing overt adverse effects [8]. Since stereotaxic injection of chemotherapeutic compounds directly into human brain tumor masses constitutes a routine approach for neurosurgeons, high concentrations of cannabinoids can easily be delivered by this technique [9]. There are two additional advantages to delivering high concentrations of cannabinoids directly into the tumor mass. Malignant transformation of astrocytomas is associated with an increase in cannabinoid receptor expression [14,15] and we show here that this increase in expression precludes the use of low concentrations of cannabinoids known to activate these receptors. Conversely, local injection of high concentrations of cannabinoids will induce apoptosis in all astrocytoma subclones (independently of CB 1 and CB 2 receptor expression), which constitutes an asset when considering the phenotypic heterogeneity that astrocytomas adopt during malignant transformation [37,38]. Thus, our results suggest that high concentrations of cannabinoids constitute the preferred regimen for neurosurgeons to use when treating malignant astrocytomas with this class of compounds.
One outstanding question that is raised by our study is: what constitutes the molecular target that is engaged by high concentrations of cannabinoids, a target that also leads to ERK activation and induction of apoptosis in astrocytomas? One possibility is another GPCR, several of which have recently been shown to be engaged by high concentrations of cannabinoids, including CP-55,940 [39]. One obvious candidate is GPR55 [40], although this receptor can be ruled out in the context of our studies since we found no evidence for the presence of GPR55 mRNA in DBT cells (as assessed by RT-PCR and qPCR, data not shown). Thus, while the molecular identification of this target remains to be determined, its identification will most certainly raise a lot of interest for research aimed at developing novel therapeutic approaches to treat astrocytomas and other tumors.
In conclusion, our study provides the first evidence for the differential activation of AKT, p38 and JNK by a highly efficacious cannabinoid agonist, CP-55,940, engaging either CB 1 or CB 2 receptors expressed at different levels. This shift in coupling precludes the use of cannabinoids at submicromolar concentrations as pro-apoptotic therapeutics, unless AKT is concomitantly inhibited. Our results also suggest that high concentrations of cannabinoids are preferable for efficacious treatment of malignant astrocytomas, because these concentrations bypass CB 1 and CB 2 receptor activation and induce apoptosis in all astrocytoma cell subpopulations. Generation of DBT Subclones Expressing mCB 1 and mCB 2 mCB 1 and mCB 2 were PCR-cloned from an existing construct using the following primers, respectively: forward, 59-gcgaattc-atgaagtcgatcttagacggccttgc-39 and reverse, 59-gcggatcctcacagagcctcggcagacgtgtctg-39; forward, 59-gcgaattcatggagggatgccgggagacagaagtg-39 and reverse, 59-gcggatcctaggtggttttcacatcagcctctg-39. Please note that these receptors are not linked to any molecular tag, because such additions often affect the coupling of GCPRs to their signal transduction pathways. Amplicons were digested with EcoR I and BamH I, and ligated into the corresponding multiple cloning site of the pIRES2-EGFP vector (BD Biosciences Clontech, Mountain View, CA). The murine delayed brain tumor (DBT) cell line, which was originally developed using the Schmidt-Ruppin strain of Rous sarcoma virus inoculated into an adult CDF1 mouse brain [41], was expanded at 37uC with 5% CO 2 in 10 cm Falcon dishes (BD Biosciences, San Jose, CA) containing 10 ml of culture media: DMEM+GlutaMAX TM -I (Gibco, Carlsbad, CA) supplemented with fetal bovine serum (FBS, 10%), HEPES (10 mM) and penicillin/streptomycin (100 units/ml and 100 mg/ml, respectively). Once cells reached ,70% confluence, they were transfected with either pIRES2-EGFP-mCB 1 or pIRES2-EGFP-mCB 2 using SuperFect (QIAGEN, Valencia, CA) according to the manufacturer's instructions. Three days after transfection, the resulting positive cells (,5% of fluorescent green cells) were isolated using fluorescence assisted cell sorting (FACS, BD FACS Vantage SE) ( Figure S4). These enriched cells were put back into culture for an additional 7 days of selection and expanded in media containing G418 (10 ml per 10 cm culture dishes). Cells that survived were isolated again by FACS, and similarly reselected and expanded for three additional rounds to ensure for the stability of each clone. In the final round, we used FACS to isolate single eGFP-positive cells and seeded them into Costar 96-well plates (0.2 ml per well, changing the media every 3-5 days). Three to six weeks later, all single-cell subclones that had expanded were harvested. Please note that both the fluorescence emitted by these single-cell subclones and the receptor expression levels were stable for at least 20 passages, showing that the shuttled plasmid had been stably incorporated in these cells.

qPCR and Radioligand Binding
Cells (10 6 ) were expanded in 10 cm dishes using 10 ml culture media. Once they reached ,70% confluence, they were rinsed with PBS and kept for an additional 24 hrs in serum-deprived media, which is composed of DMEM+GlutaMAX TM -I supplemented with fatty acid free bovine serum albumin (0.1%, BSA, Sigma-Aldrich, St. Louis, MO) and insulin-transferin-sodium selenite liquid media (16, ITS, Sigma-Aldrich). For qPCR, media was removed and cells washed once with 10 ml PBS, then cells were scraped and total RNA isolated using RNeasy extraction kit (QIAGEN) according to the manufacturer's instructions. The following primer and TaqMan probe sequences were generated using Primer Express (Perkin-Elmer Applied Biosystems, Foster City, CA): mCB 1 forward, 59-cacaagcacgccaataacaca-39, reverse, 59-acagtgctcttgatgcagctttc-39, probe, 59-(gccagcatgcacagggccgcg)-39; mCB 2 forward, 59-gaacatggccgtgctctatattatc-39, reverse, 59gaacaggtacgagggctttctg-39, probe, 59-(ctgtcctcccggcggctccg)-39. Total RNA from each sample (2.5 ng) was reverse transcribed in the same 10 ml quantitative PCR reaction mix using the Brilliant Single-Step Quantitative RT-PCR Core Reagent Kit (Stratagene, La Jolla, CA). Combined reactions were carried out on a Stratagene Mx3000P Real-Time Detection System using the following protocol: 30 min single-strand synthesis reaction at 45uC; 40-cycle, two-step PCR amplification (15 sec at 95uC followed by 1 min at 56uC). Duplicate reactions with and without reverse transcriptase were run for each sample. For radioligand binding, media was removed and cells washed once with 10 ml PBS. Cells were scraped in 1 ml buffer (50 mM Tris-HCl, 3 mM MgCl 2 , 1 mM EDTA, pH 7.4) and pulse homogenized 10 sec using PRO200 tissue homogenizer (PRO Scientific, Oxford, CT). Crude membrane preparations were generated by centrifugation of cell homogenates for 20 min at .10,0006g. Cell pellets were resuspended in buffer containing BSA (0.1%, fatty acid free) and 50 mg protein added to siliconized glass tubes for assay. Concentrations of [ 3 H]CP-55,940 between 0.1 and 10 nM were used for saturation experiments, and 1 mM CP-55,940 was used to determine specific binding. After 1 hr incubation at 30uC, reactions were stopped by rapid filtration over GF/B glass fiber filters on a 24-well M-24T Brandel cell harvester (Brandel, Gaithersburg, MD) and washed 3 times with ice-cold binding buffer containing BSA. Radioactivity on filters was measured using scintillation counter.
Quantifying the Phosphorylation of ERK1/2, AKT, JNK, p38 and GSKb Cells were expanded in Costar 24-well plates (10 5 cells per well) in culture media (0.5 ml per well). Once they reached ,70% confluence, they were rinsed with PBS and kept for an additional 24 hrs in serum-deprived media, at which time point drugs or vehicle (DMSO, 0.1%) prepared in 50 ml serum-deprived media were directly added to each well for the corresponding number of minutes. We chose to directly administer the drugs to cell culture media instead of replacing the media to avoid stressinduced increases in basal kinase phosphorylation ( Figure S1). To quantify Erk phosphorylation, reactions were stopped by removing the media and adding NuPAGEH LDS sample buffer (300 ml, Invitrogen, Carlsbad, CA) supplemented with 2mercaptoethanol (1%). An aliquot of each sample (20 ml) was loaded onto NuPAGEH 4-12% Bis-Tris pre-cast gels (Inivtrogen) and electrophoresed according to the manufacturer's instructions. Gels were electrophoretically transferred to P-nitrocellulose membrane (Millipore, Billerica, MA) and blocked in LiCOR blocking buffer (LiCOR, Lincoln, NE) for 1 hr at 25uC. Membranes were then incubated in blocking buffer containing rabbit anti-phospho-p44/42 antibody (1:1000, Cell Signaling Technology, Inc., Danvers, MA) for 1 hr at 25uC. Membranes were washed 3 times with Tris (50 mM) buffered saline containing 0.1% Tween-20 (TBS-T). Membranes were then incubated in blocking buffer containing donkey anti-rabbit antibody conjugated to IRDye 800 (1:7500, Rockland Immunochemicals, Inc., Gilbertsville, PA) for 1 hr at 25uC. Membranes were washed 3 times with TBS-T. The fluorescence intensity of each band was quantified using the LiCOR Odyssey. To quantify AKT, JNK, p38 and GSKb phosphorylation, reactions were stopped by addition of NP40 cell lysis buffer (100 ml, Invitrogen). Substrate phosphorylation was quantified using a Biosource Multiplex Bead Immunoassay kit (Invitrogen) according to the manufacturer's instructions and using a Luminex 100 TM instrument (Luminex, Austin, TX). Note that the basal phosphorylation state of these kinases in each subclone did not change significantly when the cells were incubated over the 30 min time-period of the incubation with vehicle only ( Figure S1).
Quantification of Cell Viability and Apoptosis by WST-1, Propidium Iodine (PI) and Annexin V Cells were expanded in 24-well plates (10 5 cells per well) in culture media (0.5 ml per well). Once they reached ,70% confluence, they were rinsed with PBS and kept for an additional 24 hrs in serum-deprived media, at which time drugs or vehicle (DMSO, 0.1%) prepared in 50 ml serum-deprived media were directly added to each well. After 1, 3, or 5 days, cell viability was assessed using the Cell Proliferation Reagent WST-1 (Roche, Indianapolis, IN). Briefly, WST-1 reagent (50 ml) was added to each well for 4 hrs at 37uC with 5% CO 2 , and then WST-1 products were read at 450 nm using Packard SpectraCount TM . Apoptosis was assessed using PI and annexin V (Pacific Blue TM , Invitrogen). Briefly, cells were detached using trypsin, rinsed with 25uC PBS and resuspended to a density of 10 6 cells per ml of annexin-binding buffer (10 mM HEPES, 140 mM NaCl, 2.5 mM CaCl 2 , pH 7.4). PI (100 ng) and annexin V were added to 100 ml of cell suspension and incubated at 25uC for 30 min. Additional annexin-binding buffer was added (400 ml) and samples were sorted (10,000 events) on a LSR II flow cytometer (BD Biosciences, San Jose, CA) and the data analyzed using FlowJo (Tree Star, Inc., Ashland, OR).

Data Analysis
Statistics, B max , k d , and EC 50 values were calculated using Prism.