Unless otherwise noted, drugs were dissolved in sterile dimethyl sulfoxide (DMSO) (To avoid water absorption into the solvent, the sterile filtered DMSO was stored in small aliquots at -20°C.) The stock solutions were split into small aliquots and a maximum of 2 freeze-thaw cycles were allowed. In all experiments, the final concentration of DMSO was limited to 0.15%. (For the origin of chemicals, final concentrations and for further methodological details see Supplementary Table I ).

GT1-7 cells (25) were from Sigma Aldrich, Neuro 2a and HEK 293 cells were from American Type Culture Collection (ATCC). All cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM) containing 10% foetal bovine serum and supplemented with 100 μg/ml streptomycin and 100 U/ml penicillin (“complete DMEM”). (1.5 g/L NaHCO₃ formulations were used exclusively to achieve pH 7.4 at 5% CO₂ and 37°C.) Cells were plated into poly-L-lysine-coated 6-well or 12-well plates at a density of 6-7.5-10×10⁴/cm² (GT1-7), 3-4×10⁴/cm² (Neuro 2a) or 2-3×10⁴/cm² (HEK 293) on day 1. GT1-7 cells were electroporated (1350 V, 30 ms; 1x) on the day of plating with the Neon Transfection system (Invitrogen) following the manufacturer’s protocol or transfected in Ultra-MEM with Lipofectamine 2000 (0.3 µL/cm²) for 6 h on day 2 and subsequently on day 3 with Lipofectamine LTX + Plus Reagent (0.5 µL/cm² Lipofectamine LTX + 0.67 µL/µg total DNA Plus Reagent) in complete DMEM. HEK 293 cells were transfected on day 2 in Ultra-MEM with Lipofectamine 2000 (0.2-0.3 µL/cm²) or in complete DMEM with Lipofectamine LTX + Plus Reagent (0.6 µL/cm² Lipofectamine LTX + 0.5 µL/µg total DNA Plus Reagent). Neuro 2a cells were transfected in complete DMEM using Lipofectamine LTX + Plus Reagent (0.6 µL/cm² Lipofectamine LTX + 0.5 µL/µg total DNA Plus Reagent). Transfection of siRNA into Neuro 2a cells was performed as follows: on day 2 cells were treated with siRNA in Ultra-MEM for 6 h in the presence of Lipofectamine RNAiMAX (0.5 µL/cm²). Then, on day 3, cells were treated again with siRNA in complete DMEM for 6 h with Lipofectamine RNAiMax. Unless otherwise indicated, siRNA concentration was 20 nM and construct DNA amount was 0.05-0.07 µg/cm²/construct. Serum deprivation was performed by changing the complete medium to empty DMEM 14 h (GT1-7) or 4 h (Neuro 2a and HEK293) prior to experimentation. One hour prior to stimulation, medium was changed to DMEM + HEPES (Pan-Biotech) and cell stimulation was carried out in this incubation medium. In most of the experiments, 6-well plates and 12-well plates were snap frozen with liquid nitrogen and stored at -80°C until analysis. Passages numbers 4–30 were used. At all steps, bicarbonate-containing media were equilibrated at 37°C and 5% CO₂ for at least 6 h before application.

Human wild-type CB₁R and Δ64 CB₁R [lacking the N-terminal 1-64 amino acids (22)] were expressed in pcDNA3.1 vectors (driven by the early-immediate CMV promoter for high-level expression; sequence of the CMV promoter: CGCCCCATTGACGCAAATGGGCGGTAGGCGTGTACGGTGGGAGGTCTATATAAGCAGAGCTGGTTTAGTGAACCGTCAGATC . For mild heterologous expression a pEYFP-N1 backbone was modified by omitting YFP and changing the promoter to the human herpes simplex virus thymidine kinase (TK) promoter (in order to gain moderate expression of the transgene). Sequence of the thymidine kinase promoter was: ATGACACAAACCCCGCCCAGCGTCTTGTCATTGGCGAATTCGAACACGCAGATGCAGTCGGGGCGGCGCGGTCCCAGGTCCACTTCGCATATTAAGGTGACGCGTGTGGCCTCGAACACCGAGCGACCCTGCAGCGACCCGCTTAA. The cloning was performed with the help of the following enzymes: HindIII, AgeI, BamHI (Thermo Fisher Scientific). For fluorescent labelling, human wild-type and Δ64 CB₁Rs were inserted into pEGFP-N1, or promoter modified (CMV to TK) pEYFP-N1 or promoter modified pEGFP expression vectors. Control (non-silencing) dsRNA sequences were designed based on the “C9-11” method (26). For dsRNA sequences, please refer to Supplementary Table 2 .

The cultured cells were suspended in 4°C complete lysis buffer (v.i.) in a volume of 150-200 µL/9.5 cm² growth area. After a 30 min incubation period on ice, insoluble material was removed by centrifugation at ~20.000 g (4°C for 10 min) and protein concentration was determined with BCA Assay (Thermo Fisher Scientific). The complete lysis buffer was based on a modified RIPA buffer containing: 150 mM NaCl, 1% sodium deoxycholate, 1% Triton X-100, 0.1% SDS, 1 mM EGTA 1 mM EDTA and 20 mM Tris-HCl (pH 7.35 at 4°C). This modified RIPA buffer was supplemented with 1 mM sodium orthovanadate, 1 mM phenylmethylsulfonyl fluoride, aprotinin, protease inhibitor cocktail, phosphatase inhibitor cocktail 1 and 2 (all 1:100). After the addition of 4x Laemmli sample buffer (Bio-Rad), proteins were separated in reducing mini or midi format Tris-glycine polyacrylamide gradient gels (4-15%, Bio-Rad). (Final concentration of β-mercaptoethanol in the samples was set to 5%). Alternatively, 2x Laemmli sample buffer with 10% of β-mercaptoethanol was used to suspend the cultured cells, so that the samples were directly loaded to the Tris-glycine polyacrylamide gradient gels. Proteins were then blotted onto nitrocellulose membranes using the Transblot-cell semi-dry transfer system with the fitting transfer packs (Bio-Rad).

Membranes were blocked with Tris buffered saline + 0.1% Tween-20 (TBST) supplemented with 5% milk and incubated overnight at 4°C with primary antibodies in TBST + 5% BSA + 0.1% sodium azide (for the list of antibodies, please see Supplementary Table 3 ). Horseradish peroxidase conjugated secondary antibodies (PerkinElmer) were diluted in TBST + 5% milk and incubated at room temperature for 1 h. Luminescence was measured with Azure 600 (Azure Biosystems, Dublin, CA, USA) chemiluminescence imaging device. West Pico Plus (ThermoFisher) or custom-made solution (100 mM Tris-Cl, 0.2 mM p-coumaric acid, 1.1 mM luminol, 2.6 mM H₂O₂, pH 8.5) was used as substrate for the peroxidase.

For densitometry, captured images were background subtracted with the apt module of Image J (NIH). The integrated density of individual protein bands was measured also with the Image J software and the ratio of phosphorylated to total protein (p/t ratio) was regarded as the degree of activation. Representative images were brightness and contrast adjusted with the appropriate modules of Image J (NIH) and blots were aligned in Adobe Illustrator. (Since chemiluminescent images have 16-bit depth, when needed, both high and low contrast versions of the same raw image are presented to provide a better representation. Please also note that western blots images comparing phosphorylated and total protein amounts of ERK1/2 and p38 MAPKs were exposed to identical conditions throughout the entire immunoblot process including image capture and brightness/contract adjustments. Therefore, phospho- and total blot images of different CB₁R constructs within the same cell type are directly comparable.) Unless otherwise specified, data were normalized to the vehicle treated or the lowest agonist concentration treated TK-CB₁R group.

For the assessment of CB₁R degradation, cells were treated with various protease and proteasome inhibitors or vehicle for 8 h in serum-free DMEM on day 4. For further details on pharmacons, please see Supplementary Table 1 . To evaluate the intact-to-cleaved receptor relationship, the ratio of integrated densities between 50 to 100 kDa to that below 40 kDa was calculated.

For BRET measurements, Neuro 2a or HEK 293 cells were transfected in suspension using Lipofectamine 2000 (Invitrogen; 0.5 μl/well) and plated on white poly-L-lysine coated 96-well plates in 50.000 cells/well density. The DNA amounts were 0.175 μg Epac-BRET sensor/well (27, 28) and 0.25 μg CB₁ receptor construct/well. Before the BRET measurement, cells were serum starved for 3 hours. Experiments were performed on adherent cells 24 hours after the transfection using a Varioskan Flash multimode plate reader (Thermo Scientific, Waltham, MA). Prior to stimulation, the medium was changed to a modified Krebs-Ringer buffer containing 120 mM NaCl, 4.7 mM KCl, 1.2 mM CaCl₂, 0.7 mM MgSO4, 10 mM glucose, and Na-HEPES 10 mM, pH 7.4, experiments were carried out at 37°C. The BRET measurements were started by adding cell-permeable coelenterazine h (Regis Technologies, Morton Grove, IL) to the wells at a final concentration of 5 μM. The luminescence intensities were recorded at 530 nm and 480 nm using filters (0.5 s/well). Since within the Epac-BRET sensor the intramolecular BRET decreases with the increase of [cAMP] (27), the 485 to 530 nm emission intensity ratio (BRET ratio) was regarded as a measure of [cAMP] (as opposed to the conventional 530:480 nm ratio). BRET ratios were normalized to the average of baseline (i.e. prior to the addition of any drug or vehicle).

Cells expressing various GFP-tagged CB₁R constructs were imaged on a spinning disk confocal imaging setup at room temperature (~26°C) using a Nikon Eclipse Ti2 microscope equipped with a CFI SR HP Plan Apochromat Lambda S 100XC silicon immersion objective lens, a Yokogawa CSU-W1 Spinning Disk unit, a Photometrics Prime BSI sCMOS camera and an Omicron LightHUB+ diode laser light engine. EGFP fluorescence was excited using the 488 laser line and emission was collected using a 525/20 bandpass filter (Chroma). Up to 70 Z-stack slices with a resolution of 0.3 µm were acquired per field of view with the NIS-Elements software (Nikon). Plasma membrane and cytosolic CB₁R fluorescence in non-processed raw images were determined at approx. the bottom 1/3 in the z-axis (i.e. closer to the cell-coverslip interface) along a profile running through the cell. The first peak above the cell-free background was interpreted as plasmalemmal fluorescence whereas cytosolic florescence was defined as the intensity ‘below’ the plasma membrane, at a position exactly 1 µm towards the cell centre. For the purpose of demonstration, background subtraction and brightness/contrast adjustments were performed with Image J (NIH) on representative images.

Means + s.e.m. or ± s.e.m. are shown, unless indicated otherwise. Data were obtained from at least 3 independent experiments or specified otherwise. In some experiments, minimal and maximal values have been uniformly excluded in all groups based on the ROUT method. For calculating significance of differences, one and two-way parametric or non-parametric ANOVA and post-hoc tests were applied, as appropriate. Concentration-response curves were fitted using the 3-parameter log[agonist] – response equation [(Y=Bottom + (Top-Bottom)/(1 + 10^(LogEC50-X))]. Data were analysed with Microsoft Excel (Microsoft), Image J (NIH) and GraphPad Prism 5 (GraphPad Software Inc.) software.

In order to study the effect of expression level and the long N-terminal tail on CB₁ receptor abundance in heterologous systems, we cloned the full-length and the N-terminally truncated (Δ64) human CB₁R into vectors that use either the conventional, strong early-immediate CMV (cytomegalovirus) or the weak HSV (herpes simplex virus) thymidine kinase (TK) promoter for transcription initiation ( Figure 1A ). We transfected the constructs into 3 different cell types, namely into HEK 293 cells, into undifferentiated Neuro 2a neuroblasts (29) and into highly-differentiated GT1-7 neurons (30). Neuro 2a cells express functional CB₁Rs ( Supplementary Figure 1A and Supplementary Figure 2A ), therefore, non-transfected Neuro 2a cells served as endogenous CB₁R controls. As expected, TK promoter-driven CB₁R expression was about 1-1.5 order of magnitude lower than that provided by the conventional CMV promoter ( Figures 1B, C ). In all 3 cell types, shortening the long N-terminal tail (22) by 64 amino acids also enhanced receptor expression by approx. 2 to 5-fold independently of the promoter ( Figures 1B, C ). Most importantly, however, the expression of the full-length TK-CB₁R was comparable to that of endogenous receptors ( Figures 1B, C ) suggesting that this construct may be sufficient to provide close-to-physiological receptor levels in heterologous systems.

The instability of CB₁R in heterologous expression systems has been primarily attributed to the receptor’s long N-terminal tail that complicates receptor positioning across the ER membrane and thus triggers rapid proteasomal degradation (22). Indeed, in harmony with previous reports (22), N-terminal ablation eliminated the sensitivity of the CMV-CB₁R construct to the proteasome inhibitor Mg-132 ( Supplementary Figure 1C ) indicating that the full-length CMV-CB₁R is in fact prone to ubiquitination and subsequent proteasomal degradation. Similar observations were made in HEK 293 cells as well (data not shown). However, no significant proteasomal degradation was observed with the low-expression TK-CB₁R variants ( Supplementary Figure 1D ). Instead, these low expression receptors displayed sensitivity to chloroquine ( Supplementary Figure 1D ), an inhibitor of endolysosomal degradation, similarly to that observed for the endogenous CB₁ receptor ( Supplementary Figure 1B ). Interestingly, despite the substantial proteasomal degradation of the CMV-CB₁ receptor, caused most likely by abnormal folding (22), none of the constructs induced measurable ER unfolded protein response, as assessed by the phosphorylation of eukaryotic initiation factor 2 subunit α (eIF2α) ( Supplementary Figure 1E ). Thus, weak promoter-driven, low-level expression of the full-length CB₁R precludes drastic proteasomal degradation without the need for truncation of the receptor. Additionally, weakly expressed CB₁R variants retain endolysosomal processing resembling endogenous CB₁Rs in this respect.

Next, we assessed whether changes in expression level translates into different signalling behaviour of the CB₁R. To this end, we first monitored the phosphorylation of ERK1/2 (p42/44 MAPK) by western blotting. HEK 293, Neuro 2a and GT1-7 cells were stimulated with increasing concentrations of the CB₁R specific anandamide analogue arachidonyl-2-chloroethylamide (ACEA) (31), which brings about ERK 1/2 activation in a CB₁R dependent manner ( Supplementary Figure 2A ). In non-transfected HEK and GT1-7 cells, ACEA failed to induce ERK1/2 or p38 MAPK phosphorylation (data not shown). In Neuro 2a cells, the full length TK-CB₁R variant produced a stimulus-response curve that practically overlapped with that produced by endogenous CB₁ receptors in non-transfected cells ( Figures 2A, B ). On the other hand, CMV promoter-driven high expression CB₁R variants increased the basal phosphorylation of ERK1/2 and tended to shift the dose-response curve to the left. Congruent ERK 1/2 activation data were obtained in HEK 293 cells (but not in GT1-7 neurons) ( Figure 2A, B ). In contrast, ACEA-induced activation of p38 MAPK, another downstream target of CB₁Rs that may be recruited independently of ERK1/2 (32, 33), proved to be much less sensitive to CB₁R abundance or N-terminal Δ64 truncation than observed for ERK1/2 phosphorylation ( Supplementary Figures 2B, C ).

Confocal microscopy in Neuro 2a and HEK 293 cells revealed that the ability to increase basal ERK1/2 activity corresponds well to the cell surface expression of CB₁R constructs, with CMV-CB₁R variants having around an order of magnitude higher plasmalemmal abundance than their TK-CB₁R counterparts ( Figures 3A, B ). Intriguingly, whereas Δ64 truncation improved plasma membrane targeting of CB₁Rs under high-expression conditions, the same ablation did not enhance cell surface localization of weakly expressing TK-CB₁R variants. These data together imply that CB₁R-activated parallel MAPK signalling pathways display different sensitivity to supraphysiological CB₁R expression, and that weak-expression recombinant CB₁ receptor variants mimic the behaviour of the endogenous receptors more closely than conventional high-expression CB₁R constructs.

The coupling preference of CB₁Rs may be swayed from G_i/o proteins towards the G_s pathway by several factors (34–36) including the expression level of the receptor itself (15). Therefore, we tested the effect of ACEA and the CB₁R full agonist WIN 55,212-2 on cytosolic cAMP in Neuro 2a neuroblasts and HEK 293 cells expressing various CB₁R clones. In Neuro 2a cells expressing endogenous CB₁Rs only, both ligands reduced the basal and forskolin-stimulated cAMP concentrations signifying G_i/o coupling. This G_i/o preference was clearly retained in Neuro 2a and HEK 293 cells transfected with either the TK-CB₁R or the TK-Δ64-CB₁R clones but was lost or shifted towards G_s when CB₁R expression was driven by the CMV promoter ( Figure 4 ). Thus, as seen with the other cannabinoid signalling pathways and receptor degradation, the weakly expressed full-length CB₁Rs satisfyingly resemble the signalling behaviour of endogenous receptors.