Affinity and selectivity in CBR binding studies

To determine the affinity and selectivity of the 18 substances, we performed [³H]-CP55940 displacement assays using membrane fractions of CHO cells expressing recombinant human CB₂R and CB₁R, in two independent laboratories. In addition, mouse brain and spleen were used as source of mouse CB₁R and CB₂R, respectively.

Using the Pearson correlation analysis, we found a statistically significant correlation between the binding affinities between the different labs (Pearson coefficient: 0.9304 (hCB₁R), 0.6648 (hCB₂R) and 0.7720 (mCB₂R), see Supplementary Fig. 2). Figure 1 depicts the selectivity of the ligands for the CB₂R versus CB₁R. We found that HU210>CP55940, WIN55212-2>Δ⁹-THC were the highest affinity non-selective human CBR ligands. Conversely, HU308, HU910 and JWH133 were the most selective human CB₂R ligands (Supplementary Table 2), possessing 278-, 166- and 153-fold higher respective affinities for CB₂R than for CB₁R. Notably, JWH015 and Gp-1a were less than 30-fold selective for CB₂R. Importantly, the binding selectivity of the ligands for mouse CB₂R over mouse CB₁R appeared to be greatly reduced for all ligands (<100 fold), except AM630 and SR144528, which are actually more selective on mCB₂R than on hCB₂R. The most selective agonists on mCB₂R were (rac)-AM1241 (66-fold), JWH133 (40-fold) and Gp-1a (20-fold). As expected, AEA and 2-AG, the endogenous ligands of CB₁R and CB₂R, were non-selective and showed moderate binding affinities towards both receptors (pKi ~7).

Open in a separate window

Figure 1

CB₂R selectivity of cannabinoid reference ligands on mouse and human CBRs.

CB₂R selectivity for all cannabinoid receptor reference ligands are presented as the difference in mean pKi values between CB₂R and CB₁R for both the human (black bars) and mouse (white bars) orthologues. From left to right: ligands with decreasing CB₂R selectivity (from HU308 to Gp-1a), nonselective ligands (from HU210 to CP55940), ligands with CB₁R selectivity (AM251 and SR141716A). pKi values are obtained from three independent experiments performed in duplicate. Error bars shown in the figure represent the s.e. of the mean.

Activity and selectivity of CBR signalling pathways

To determine the functional activity and selectivity (towards CB₂R over CB₁R) of the ligands we performed five different assays (GTPγS, cAMP, β-AR, pERK and GIRK) on both human CB₂R and human CB₁R (Supplementary Tables 3–7). All ligands were tested on cAMP signalling on both mouse CBRs and HU910, HU308 and JWH133 were tested on G-protein activation and β-arrestin recruitment on mCB₂R, to determine interspecies behaviour of the ligands. Efficacy of the ligands is normalized to the effect produced by CP55940 (10μM) in all assays; however, it should be noted that efficacy is relative by definition, and is dependent on the reference ligand used as well as the assay conditions.

For both human and mouse CB₂R, the potency of the ligands correlated with their binding affinity in most assays, except for β-AR and GIRK signalling (Supplementary Fig. 3). Graphs showing the pEC50 values of the reference ligands for all assays are shown in Fig. 2a–d.

Open in a separate window

Figure 2

Radarplots of potencies of cannabinoid receptor reference ligands across different assays.

(a) non-selective CBR agonists, (b) selective CB₂R agonists, (c) antagonists, (d) endocannabinoids. These plots are a visualization of the differences in potency across different signalling pathways. All potency values are the mean of three independent experiments performed in duplicate.

CP55940 and HU308 behaved as potent full agonists at hCB₂R in the GTPγS assay (Supplementary Table 3), while WIN55212-2 acted as a partial agonist. HU910 behaved as a partial CB₂R agonist as well, but was, together with HU308 and JWH133, the most selective for CB₂R in this assay (185- and 193-fold, respectively). Of note, JWH133 was considered functionally inactive on hCB₁R, because its maximal effect was only 20% at 10μM. On mCB₂R, both HU308 and JWH133 were full agonists, but HU910 remained partially active. The potency of all three ligands was similar for human and mouse receptors.

In contrast to previous reports^27,28, Gp-1a acted as an inverse agonist on CB₂R, but was inactive at CB₁R. Both THC and the endocannabinoids AEA and 2-AG acted as partial agonists on both receptors with similar potency.

In the cAMP assay (Supplementary Table 4) all CB₂R agonists displayed higher selectivity (>1,000-fold) and higher efficacy, than in the GTPγS-assay, reflecting substantial signal amplification in this pathway. Only (rac)-AM1241 remained a partial agonist in the cAMP assay. Upon comparison of the efficacy of the ligands between species, in general it appears that many ligands on the mouse CBRs are partial agonists on cAMP, in contrast to the human CBRs. This difference in efficacy might be a result of a difference in CBR expression levels. Differences in expression levels may also account for the interspecies differences displayed by (rac)-AM1241, which was previously reported as a protean agonist (a protean agonist shows differences in signalling due to differences in experimental conditions, whereas a true biased agonist has signalling preference due to conformational changes of the receptor)²⁹. HU910 was found to bind with similar affinity to both mouse CBRs, but was inactive on mCB₁R in the cAMP assay. HU308, JWH015 and (rac)-AM1241 were the most selective agonists for hCB₂R in this assay, whereas CP55940, WIN55212-2 and HU210 displayed the highest potency. HU910, HU308 and JWH133 were the most selective on the mCB₂R. Of note, AEA and 2-AG were relatively weak partial agonists, especially on human and mouse CB₁R (pEC₅₀ <5.2 and Emax <70%).

All ligands modulated β-AR recruitment (Supplementary Table 5) to the membrane in CHO cells expressing human CB₂R or CB₁R. CP55940 was the most potent ligand in this assay, followed by WIN55212-2. CP55940 acted as full agonist at both receptors, but WIN55212-2 displayed partial agonism at CB₂R, as in the GTPγS assay. The other agonists, including the ligands JWH133, JWH015, HU308, HU910 (Emax 50–70%) and the endocannabinoids (Emax 40–80%), only partially recruited β-AR. The most selective CB₂R agonists were HU308, JWH133 and HU910, which were found to be 282-, 275- and 274-fold more potent for CB₂R than CB₁R. Of note, JWH133, HU308 and HU910 were all significantly less potent on mCB₂R in β-AR recruitment.

WIN55212-2 was a full agonist in the pERK assay (Supplementary Table 6) and demonstrated 86-fold selectivity for the human CB₂R, whereas CP55940 lacked selectivity in this assay. HU308 and JWH133 were potent and selective CB₂R full agonists, whereas Δ⁹-THC and (Rac)-AM1241 acted as partial agonists on the pERK signalling cascade. Interestingly, HU910, AEA and 2-AG had low potency in this assay (pEC50<5.5), but HU910 and 2-AG acted as full agonists at high concentrations.

Most ligands appeared to be less potent and less CB₂R-selective in the GIRK assay (Supplementary Table 7). For example, neither JWH133 nor THC activated the GIRK pathway at all. JWH015 was the most selective agonist in this assay, followed by HU308. WIN55212-2 and CP55940 activated the GIRK channels with the highest potency, but as expected, both were highly potent and efficacious at CB₁R as well.

The high variability in potency and efficacy that the CB₂R agonists displayed across the different signalling pathways strongly suggests biased signalling. To quantify this ligand bias towards distinct signal transduction pathways, we performed operational analysis based on van der Westhuizen et al. (Supplementary Tables 8–11) (ref. ³⁰). This analysis is based on the operational model of Black and Leff³¹, which calculates signal transduction strength on a given pathway, taking into account (a) the maximal effect of the system used, (b) the agonist concentration, (c) the agonist's maximum efficacy, (d) the ligand affinity for the receptor and (e) the transducer slope. In order to eliminate system and observation bias, such as the level of amplification between signalling pathways or assay sensitivity, we used CP55940 as a reference compound, because it was the only compound that behaved as a full agonist with comparable potency in all assays, except for the cAMP assay. The ΔΔlogR values resulting from this operational analysis are graphically shown in Fig. 3. The operational analysis on hCB₂R revealed that THC showed statistically significant bias towards pERK signalling compared to β-arrestin and GTPγS. In addition, THC did not activate GIRK, indicative of high bias against this pathway. (rac)-AM1241 was biased towards β-arrestin coupling and pERK signalling compared to GIRK channel activation. JWH133 was moderately biased towards β-arrestin compared to GIRK, whereas both WIN55212-2 and JWH015 showed preference for GIRK compared to cAMP signalling. AEA showed preference for pERK and GIRK signalling compared to cAMP, whereas 2-AG was significantly biased towards GIRK compared to G-protein signalling. Upon comparison between β-AR coupling and cAMP signalling, all ligands appear to be significantly biased. This observation is, however, confounded by the fact that CP55940, which is used as the reference ligand, has an exceptionally high potency in the cAMP assay compared to the other signal transduction pathways, and might be in fact biased itself towards the cAMP pathway. Of note, HU910 and HU308 were well-balanced ligands without significant bias towards any signal transduction pathway on hCB₂R. Of note, HU910, HU308 and JWH133 were significantly biased towards G-protein signalling over β-AR coupling and cAMP signalling on the mCB₂R highlighting a potentially important species difference.

Open in a separate window

Figure 3

Heatmap of ΔΔlogR values resulting from the operational analysis.

The operational analysis was performed using the procedure described in the Supplementary Information, with data obtained from three independent experiments performed in duplicate. Dark blue indicates bias towards the first pathway, while dark red indicates bias towards the second pathway. White boxes are not determined. (a) The operational analysis on hCB₂R was done with all agonists, except HU210, which was tested in only two functional assays due to legal restrictions. Because of error propagation in the analysis, only a few were found to be statistically significant (P<0.05). (b) The operational analysis on mCB₂R was done with the three most selective and most broadly active compounds on hCB₂R; HU308, HU910 and JWH133. Here, all three ligands showed statistically significant bias towards G-protein signalling. Statistics was performed with one-way ANOVA Holm-Sidak's multiple comparisons test using Graphpad Prism 6.

Off-target activity in the endocannabinoid system

To rule out any indirect effects of the ligands on CB₁R and CB₂R, we investigated off-target activity on endocannabinoid-regulating enzymes, as well as their effects on AEA reuptake inhibition. None of the ligands showed any off-target activity on a panel of serine hydrolases, determined in a competitive activity-based protein profiling (ABBP) assay in mouse brain proteome up to a concentration of 10μM (Supplementary Fig. 4A). In addition, none of the ligands showed any significant effect at a concentration of 10μM when tested on NAPE-PLD, DAGL and MAGL-overexpressing cells (Supplementary Table 12). The compounds were also tested on FAAH activity using U937 cell homogenate at a concentration of 5μM. Only AM251 and Gp-1a showed partial inhibition of FAAH activity (~30–40%, Supplementary Fig. 4F).

In addition, none of the agonists had significant activity on ABHD6-, ABHD12- and COX2 activity up to a concentration of 5μM for COX2 and 10μM for the ABHDs (Supplementary Fig. 4B–E). In contrast, the antagonists SR141716A, AM251 and Gp-1a, which are the structurally most similar ligands in this panel of ligands, inhibited ABHD12 in the micromolar range with IC₅₀ values of 6.1, 1.6 and 0.8μM, respectively (Supplementary Table 12). AM251 was the only compound of the ligands tested that had high efficacy on GPR55 in β-arrestin recruitment (82±9%), albeit with low potency (pEC₅₀=5.49±0.09, see Supplementary Table 13).

AEA reuptake inhibition was determined in three different human cell lines: monocyte-like U937 cells, mast cell-like HMC-1 cells and keratinocyte-like HaCaT cells. In U937 and HaCaT cells, some of the ligands possessed micromolar potency, including AM251, SR141716A, Gp-1a, HU308 and HU910 (Supplementary Table 14). In HMC-1 cells, which lack FAAH expression³², all tested ligands, except SR141716A, were weakly active or inactive at a concentration of 5μM, which indicates a potential role of FAAH in the inhibition of AEA uptake in U937 and HaCaT cells. In agreement with this, the most active AEA reuptake inhibitors AM251 and Gp-1a partially inhibited FAAH at 5μM in U937 cell homogenate (Supplementary Fig. 4F). Of note, SR144528 was inactive up to a concentration of 10μM in all cell lines, whereas SR141716A showed FAAH-independent micromolar effects on AEA reuptake in all cells.

TRP-channels

As AEA and 2-AG activate some TRP channels, these channels may be regarded as ionotropic cannabinoid receptors. Here, we tested our ligands on six different TRP channels (TRPV1–4, TRPA1 and TRPM8). We found that most TRP channels were activated by one or more ligands, apart from TRPA1 that was activated by all of them (Supplementary Table 15). HU308 was the most selective agonist that activated only TRPV1 and TRPA1 in the high micromolar range, whereas HU910 activated TRPV3 with submicromolar potency (0.12±0.05μM). However, to date, TRPV3-related effects and toxicity (hypothermia and reduced blood pressure) have not been observed after in vivo administration of HU910 (ref. ³³).

JWH133 only activated TRPA1. Although its efficacy was fairly high (76.8±3.8% activation), its potency was low (8.5±2.3μM). Whereas SR144528 did not target any of the TRP channels, AM630 was a full agonist at TRPA1 (118±2%). Of note, SR141716A targeted three out of six TRP channels tested and, remarkably, nanomolar potency was observed at TRPM8.

Off-target panel (CEREP panel)

The CEREP panel served to screen off-target activity on 64 proteins, which are associated with common adverse side effects in humans. JWH133 was identified as the most selective ligand with no off-targets detected in this panel. The summary of all off-targets is shown in Supplementary Table 16, in Supplementary Table 20 all CEREP data are shown. In contrast, CP55940 was the most non-selective ligand of which we detected 17 off-targets with more than 50% inhibition at 10μM. HU910 and HU308 hit nine and four off-targets in this panel, respectively. However, of the nine off-targets of HU910 in this panel only the dopamine uptake reporter displayed an IC₅₀ of <10μM (IC₅₀=1.40μM). Therefore, these off-targets are not likely to be physiologically relevant at 10μM. The CB₂R-selective antagonist SR144528 had only two off-targets. Of note, the adenosine A₃ receptor was the most common off-target (Supplementary Table 16). The physiological relevance of this observation is currently unclear, although it has previously been published that the endocannabinoid 2-AG has allosteric activity on this receptor³⁴.

In vitro DMPK and pharmacokinetics

The high overall lipophilicity (see above) may strongly influence other ADME properties. Metabolic stability in human and mouse microsomes and hepatocytes is low for many of the ligands as indicated by their high in vitro clearances (Supplementary Table 17). Some compounds, such as AEA, suffered from high chemical instability even in dimethylsulfoxide (DMSO) stock solution. For most ligands, except Gp-1a, microsomal clearances seem to over-predict the corresponding values in both human and mouse hepatocytes. The fraction unbound (Fu) was either very low or not measured for many compounds due to their very high lipophilicity. None of the molecules is a strong human or mouse P-gp substrate. In combination with their low polar surface areas, the ligands are, therefore, likely to reach the brain.

We determined the primary pharmacokinetic parameters of the compounds (Supplementary Table 18). A mixture containing 15% DMSO and 85% PEG400 was used as a vehicle to dissolve the ligands at 1–2mgkg⁻¹ for intravenous (i.v). administration. In vivo clearances in mice were very high and span a range from the lowest value of 0.17lh⁻¹kg⁻¹ for HU910 up to 6.9lh⁻¹kg⁻¹ for (S)-AM1241. The volume of distribution was high (>3lkg⁻¹) for (S)-AM1241, AM630, Gp-1a and JWH015, moderate (1–3lkg⁻¹) for AM251 and JWH133 and low (<1lkg⁻¹) for HU910 and SR144528. This resulted in the longest half-life for HU910 (7h), whereas JWH133 had a half-life of only 1h. As reasonable C0 values, extrapolated from Cmax values, can be reached for all compounds (see Supplementary Table 18), the rather short in vivo half-lives raise the possibility that published mouse in vivo efficacy data using these molecules might be rather C0 than AUC driven. Future experiments using accurate concentration/effects relationships might answer this question.

Following oral administration using aqueous microsuspensions in rodents, absorption was strongly influenced by the physico-chemical properties of the compounds, formulation and feed conditions of the animals (Supplementary Table 19). HU910 was suspended in ethanol/cremophor EL/0.9% NaCl (5/5/90% w/w) whereas HU308 and JWH133 were suspended in an aqueous gelatine/NaCl vehicle (7.5/0.62% w/w). Maximal plasma concentration peaked ~1h after administration with Cmax ranging from 201 to 2070, ngml⁻¹. Half-lifes were comparable to i.v. administration. Bioavailability was not calculated as these were separate experiments. Taken together these data suggest a wide variety of application format for in vivo experiments if care is taken on the formulation aspects. Determination of the plasma concentrations and pharmacokinetic behaviour seems however warranted.

Synthetic procedures general remarks

All reactions were performed using oven or flame-dried glassware and dry solvents. Reagents were purchased from Sigma Aldrich, Acros and Merck and used without further purification unless noted otherwise. All moisture sensitive reactions were performed under an argon atmosphere. Traces of water were removed from starting compounds by co-evaporation with toluene. ¹H- and ¹³C-NMR spectra were recorded on a Bruker AV 400MHz spectrometer at 400.2 (¹H) and 100.6 (¹³C) MHz. Chemical shift values are reported in ppm with tetramethylsilane or solvent resonance as the internal standard (CDCl₃ δ 7.26 for ¹H, δ 77.0 for ¹³C, CD₃OD: δ 3.31 for ¹H). Data are reported as follows: chemical shifts (δ), multiplicity (s=singlet, d=doublet, dd=double doublet, td=triple doublet, t=triplet, q=quartet, quinted=quint, br=broad, m=multiplet), coupling constants J (Hz), and integration. HPLC purification was performed on a preparative LC-MS system (Agilent 1200serie) with an Agilent 6130 Quadruple MS detector. High-resolution mass spectra were recorded on a Thermo Scientific LTQ Orbitrap XL. Flash chromatography was performed using SiliCycle silica gel type SiliaFlash P60 (230–400 mesh). TLC analysis was performed on Merck silica gel 60/Kieselguhr F254, 0.25mm.

Chemical syntheses are described in detail in the Supplementary Methods.

Cell culture and membrane preparation

CHOK1hCB₁_bgal, CHOK1hCB₂_bgal and CHOK1mCB₂_bgal cells (source: DiscoveRx, Fremont, CA, USA) were cultured in Ham's F12 Nutrient Mixture supplemented with 10% fetal calf serum, 1mM glutamine, 50μgml⁻¹ penicillin, 50μgml⁻¹ streptomycin, 300mgml⁻¹ hygromycin and 800μgml⁻¹ geneticin in a humidified atmosphere at 37°C and 5% CO₂. Cells were subcultured twice a week at a ratio of 1:20 on 10-cm ø plates by trypsinization. For membrane preparation the cells were subcultured 1:10 and transferred to large 15-cm diameter plates. For membrane preparation, the cells were detached by scraping them into 5ml phosphate-buffered saline (PBS), collected and centrifuged at 1,000g for 5min. Pellets derived from 30 plates were added together and resuspended in 20ml ice-cold buffer (50mM Tris-HCl, 5mM MgCl₂, pH 7.4). An UltraThurrax homogenizer was used to homogenize the cell suspension. Membranes and the cytosolic fraction were separated by ultracentrifugation (31.000rpm, with a Ti-70 rotor in a Beckham Coulter Ultracentrifuge) at 4°C for 20min. The supernatant was discarded and the pellet was resuspended in 10ml of the same buffer and the homogenization and centrifugation steps were repeated. Supernatant was discarded and the pellet was resuspended in 5ml buffer. Aliquots of 200μl were frozen at −80°C until further use. Protein concentration was determined using the BCA method⁴⁴.

[³H]CP55940 displacement assay

[³H]CP55940 displacement assays were used for the determination of affinity (K_i) values of ligands for the cannabinoid CB₁ and CB₂ receptors. Membrane aliquots containing 5μg (CHOK1hCB₁_bgal) or 1μg (CHOK1hCB₂_bgal) of membrane protein in 100μl assay buffer (50mM Tris-HCl, 5mM MgCl₂, 0.1% bovine serum albumin (BSA), pH 7.4) were incubated at 30°C for 1h, in presence of 3.5nM [³H]CP55940 (CHOK1hCB₁_bgal) or 1.5nM [³H]CP55940 (CHOK1hCB₂_bgal). Non-specific binding was determined in the presence of 10μM AM630 (CHOK1hCB₂_bgal) or 10μM SR141716A (CHOK1hCB₁_bgal). When AEA or 2-AG were used as ligands, phenylmethylsulphonyl fluoride was used as a serine protease inhibitor. Incubation was terminated by rapid filtration performed on GF/C filters (Whatman International, Maidstone, UK), presoaked for 30min with 0.25% PEI, using a Brandel harvester (Brandel, Gaithersburg, MD, USA). Filter-bound radioactivity was determined by scintillation spectrometry using a Tri-Carb 2900 TR liquid scintillation counter (Perkin Elmer, Boston, MA, USA). For mouse experiments, brain (for CB₁) and spleen (for CB₂) were resuspended in 2mM Tris-EDTA, 320mM sucrose, 5mM MgCl₂ (pH 7.4), then homogenized in a Potter homogenizer and centrifuged three times at 1000g (10min). The supernatants were centrifuged at 18000g (30min), and the pellets were resuspended in assay buffer (50mMTris-HCl, 2mMTris-EDTA, 3mM MgCl₂, pH 7.4). These membrane fractions were used in rapid filtration assays with 400pM of [³H]CP55,940. In all binding experiments, nonspecific binding was determined in the presence of 1μM ‘cold' agonist. When AEA or 2-AG were used as ligands, the FAAH inhibitor URB597 or the MAGL inhibitor JZL184, respectively, were included in the assay buffer⁴⁵. Brain and spleen tissue was collected from male C57BL/6J mice, 8 months old (supplier: Charles River Via Indipendenza 11 Calco (Lecco) 23885 Italy), at the Santa Lucia Foundation within the project: ‘Ruolo del Sistema Endocannabinoide nei Processi Neurodegenerativi in modelli animali di Alzheime' by Mauro Maccarrone. Ethical approval was given by the Ministry of Health, n. 47/2014/PR (deadline 17 November 2019).

[³⁵S]GTPγS assay

G-protein activation by the receptor is measured by the binding of radiolabelled GTP, [³⁵]GTPγS, to the receptor⁴⁶. Five micrograms of homogenized CHOK1CB_bgal membranes in 20μl assay buffer (50mM Tris-HCl buffer (pH 7.4), 5mM MgCl₂, 150mM NaCl, 1mM EDTA, 0.05% BSA and 1mM DTT, freshly prepared every day) were pretreated with 5μg saponin and 1μM GDP. To determine the pEC₅₀ and Emax values of cannabinoid reference ligands, the membranes were incubated with various concentrations of the ligands of interest. The basal level of GTP binding was measured in untreated membrane samples, and the maximal level of GTP binding was measured by treatment of the membranes with 10μM CP55940. All samples were preincubated for 30min at rt, followed by addition of [³⁵S]GTPγS (0.3nM). The total value of each sample was 100μl. The samples were incubated for 90min at 25°C on a shaking platform. Incubations were terminated by rapid vacuum filtration to separate the bound and free radioligand through Whatman GF/B filters (Perkin Elmer, Groningen, The Netherlands) using a Filtermate 96-well harvester (Perkin Elmer, Groningen, The Netherlands). Filters were subsequently washed ten times with ice-cold wash buffer containing 50mM Tris HCl, pH 7.4 and 5mM MgCl₂. Microscint scintillation fluid (25μl, Perkin Elmer, Groningen, The Netherlands) was added to each filter. After 3h, the filter-bound radioactivity was determined by scintillation spectrometry using a 1450 Microbeta Wallac Trilux scintillation counter (Perkin Elmer).

PathHunter β-arrestin recruitment assay

The assay was performed using the PathHunter hCB₁_bgal, hCB₂_bgal, hGPR55_bgal or mCB₂_bgal CHOK1 β-arrestin recruitment assay kit (source: DiscoveRx, Fremont, CA, USA), according to the manufacturer's protocol⁴⁷. Briefly, cells were seeded at a density of 5,000 cells per well of solid white walled 384-well plates (Perkin Elmer, MA, USA) in 20μl cell culture medium and incubated overnight (16–18h) in a humidified atmosphere at 37°C and 5% CO₂. In the endocannabinoid assay, a concentration of 50μM of phenylmethylsulphonyl fluoride was freshly added to each well and the cells were incubated for 30min in a humidified atmosphere at 37°C and 5% CO₂. In the agonistic assay, the cells were stimulated with 10μM of each agonist (single point assay) or 11 increasing concentrations of each agonist and incubated for 90min in a humidified atmosphere at 37°C and 5% CO₂. In the antagonistic assays, the cells were exposed to 10μM of each antagonist (single point assay) or 11 increasing concentrations of each antagonist and incubated for 30min in a humidified atmosphere at 37°C and 5% CO₂, followed by the addition of the EC80 concentration of CP55940 (25nM for CHOK1hCB1_bgal and 46nM for CHOK1hCB2_bgal). The cells were incubated for 90min in a humidified atmosphere at 37°C and 5% CO₂. Compounds in DMSO stocksolutions were added using a HP D300 Digital Dispenser (Tecan, Männedorf, Switzerland). Endocannabinoids 2-AG and anandamide (stocksolutions in acetonitrile) were added manually. The final concentration of organic solvent per assay point was ≤0.1%. β-Galactosidase enzyme activity was determined by using the PathHunter Detection mixture, according to the kit's protocol⁴⁷. Detection mixture, 12μl per well, was added and the plate was incubated for 1h in the dark at room temperature. Chemiluminescence, indicated as relative light unit, was measured on an EnVision multilabel plate reader (Perkin Elmer, MA, USA).

pERK assay

Assays to determine pERK stimulation were performed in HEK cells stably expressing 3HA-tagged human CB₁ receptors and Flp-in HEK cells stably expressing 1HA-3TCS-tagged hCB₂ receptors (3TCS refers to three ‘thrombin cleavage sites', which are not relevant for these experiments). HEK-hCB₁ cells were created by transfecting Human Embryonic Kidney 293 (HEK) cells (ATCC #CRL-1573) with human CB₁ (hCB₁) chimerized with three hemeagglutinin (HA) tags at the amino terminus. A single HA tag, three thrombin cleavage sites and a single ‘L' residue (for cloning purposes) were chimerized at the N terminus of hCB₂ (Missouri S&T cDNA Resource Center, www.cdna.org, CNR0200000) by overlap-extension PCR and restriction digest/ligation from DNA oligonucleotides. The subsequent hCB2 construct was cloned into pcDNA5/FRT (Life Technologies #V601020) via the NheI and XhoI sites using standard molecular biology techniques and sequence verified. The human Flp-In-293 cell line (Invitrogen #R750-07) was stably transfected with the human hCB2 construct for targeted integration into the FRT site and maintained in 50μg hygromycin Bml⁻¹ (ref. ⁴⁸). Cells were grown in high glucose DMEM, supplemented with 10% fetal bovine serum (FBS) and were cultured under standard conditions. Briefly, cells from semi-confluent T75 flasks were seeded in 100μl per well complete medium, in 96-well plates pre-treated with poly-D-lysine. Approximately 24h after seeding, cells were 60% confluent and complete medium was removed and replaced with 50μl per well serum-free DMEM containing 1mgml⁻¹ BSA, for 18h to reduce the pERK background. Drugs were made up in DMEM+1mgml⁻¹ BSA at 2 × concentrations and cells were stimulated by dispensing 50μl per well in a 37°C waterbath for exactly 5min. At the conclusion of stimulation, plates were moved to an ice bed and well contents were rapidly aspirated. Thirty microliters per well of lysis buffer was immediately dispensed, and plates were placed on shakers for 10min prior to proceeding to detection. Lysis buffer and detection reagents were from a PerkinElmer AlphaScreen SureFire pERK (Thr202/Tyr204) assay kit, and were utilized according to the manufacturer's instructions. In order to reduce use of detection reagents, lysates were first transferred to 1/2 area, white 96-well plates (PerkinElmer). Detection of hCB₁-mediated pERK responses was performed using the manufacturer's standard detection protocol, whereas hCB₂ pERK responses were only detected utilizing the high sensitivity detection protocol. Plates were read on an EnSpire plate reader (PerkinElmer).

GIRK assay

Activation of native GIRK channels in mouse AtT20 neuroblastoma cells stably transfected with HA-tagged hCB₁ or hCB₂ receptors (source: cells were generated by transfecting WT AtT20 with HA-tagged hCB₁ or hCB₂ receptor. The cDNA clones for the human CB₁ and CB₂ receptor with 3 × N-terminal HA tags were obtained from the Missouri S&T cDNA Resource Center (www.cdna.org)) was measured as a change in membrane potential using membrane potential-sensitive dye (blue, Molecular Devices) in a FlexStation three-plate reader^49,50. Briefly, AtT20 cells from an 80–90% confluent 75mm² flask were resuspended in L-15 medium supplemented with 1% FBS, 100U penicillin and 100μg streptomycin ml⁻¹ and plated in 96-well black-walled plates (90μl per well). Cells were incubated overnight in humidified room air at 37°C. Membrane potential dye was dissolved in low K HEPES buffered saline (90μl) and added to the plate an hour before assay. Fluorescence was measured every 2s (λexcitation=530nm, λemission=565nm). Assays were carried out at 37°C, drugs were added in volumes of 20μl after at least 2min of baseline recording. Changes in fluorescence were expressed as a percentage of predrug values, after subtraction of the small changes produced by solvent (0.1% DMSO, 0.1% BSA) alone. For antagonist experiments, drugs were preincubated for at least 5min before CP55940 (30nM) was added. For each drug concentration, the peak change in fluorescence was normalized to the change produced a maximally effective concentration of CP55–940 (1μM). Data are expressed as the mean±s.e.m. of N independent experiments, each performed in duplicate.

Mouse cAMP assay

The effect of ligands on the forskolin-stimulated accumulation of cAMP was determined with the LANCE ULTRA cAMP kit (Perkin-Elmer Life Sciences, Boston, MA, USA) (ref. ⁵¹). Brain (for CB₁) and spleen (for CB₂) were resuspended in 50mM of Tris-HCl (pH 7.4), then were homogenized in a Potter homogenizer and centrifuged at 1000g for 10min. The supernatants were incubated for 30min with 1-methyl-3-isobutylxanthine (IBMX), then forskolin was added in the presence or absence of ligands. The supernatants were incubated for 30min in ATP Regeneration Buffer (50mM HEPES, pH 7.4, 10mM phosphocreatine, 10 unitsml⁻¹ creatine phosphokinase, 10μM GTP, 200μM ATP, 10mM MgCl₂, 250μM IBMX), and the reaction was stopped by adding lysis buffer. Time-resolved fluorescence was measured with a Victor V Multilabel counter (Perkin-Elmer Life Sciences, Boston, MA, USA). Brain and spleen tissue was collected from male C57BL/6J mice, 8 months old (supplier: Charles River Via Indipendenza 11 Calco (Lecco) 23885 Italy), at the Santa Lucia Foundation within the project: ‘Ruolo del Sistema Endocannabinoide nei Processi Neurodegenerativi in modelli animali di Alzheime' by Mauro Maccarrone. Ethical approval was given by the Ministry of Health, n. 47/2014/PR (deadline 17 November 2019).

Human cAMP assay

cAMP assays were performed with CHO cells stably expressing human CB₁ or human CB₂ receptors⁵² (source: DiscoveRx, Fremont, CA, USA) using the cAMP-Nano-TRF detection kit (Roche Diagnostics, Penzberg, Germany). Cells were seeded 17–24h prior to the experiment at a density of 3 × 10⁴ cells per well in a black 96-well plate with flat clearbottom (Corning, Wiesbaden, Germany) and incubated in 5% CO₂ at 37°C in a humidified incubator. The growth medium was exchanged with Krebs Ringer bicarbonate buffer with 1mmoll⁻¹ 3-isobutyl-1-methylxanthine (IBMX), 0.1% fatty acid-free BSA and incubated at 30°C for 60min. Agonist was added to a final assay volume of 100μl and the mixture was incubated for 30min at 30°C. The assay was stopped by the addition of 50μl 3 × lysis reagent and shaken for 2h at room temperature. The time-resolved energy transfer was measured using an LF502 Nanoscan FLT (IOM, Berlin, Germany), equipped with a laser as excitation source. cAMP content was determined from the function of a standard curve spanning from 10 to 0.13nmoll⁻¹ cAMP.

Data analysis of functional assays

All experimental data were analysed using the nonlinear regression curve fitting program GraphPad Prism 6.0 (GraphPad Software, Inc., San Diego, CA, USA). From displacement assays, pIC₅₀ values were obtained by non-linear regression analysis of the displacement curves. The obtained pIC₅₀ values were converted into pK_i values using the Cheng Prusoff equation to determine the affinity of the ligands (K_D: 0.33 (CB₂R), 0.10 (CB₁R)) (ref. ⁵³). β-Arrestin recruitment and GTPγS curves were analysed by the nonlinear regression option ‘log (agonist or inhibitor) versus response-variable slope' to obtain potency, inhibitory potency or efficacy values of agonists and inverse agonists (EC₅₀, IC₅₀ or E_max, respectively). Basal activity of the cells is set at 0%. For the β-arrestin recruitment assay, all data points were corrected for any background (for example, background luminescence). For the analysis of antagonists measured in competition with CP55940, the nonlinear regression option ‘log (inhibitor) versus response' was chosen. The response of agonists per sample is normalized to the effect of 10μM CP55940 and the response of antagonists is normalized to effect of the EC₈₀ of CP55940. For the GTPγS assay, agonistic effect is normalized to the effect of 10μM CP55940. Data shown are the mean±s.e.m. of at least three independent experiments, each performed in duplicate, unless stated otherwise.

Data analysis for bias calculations

The data used for the operation analysis was the data of at least three independent experiments on each assay, all normalized to the effect of 10μM CP55940. The analysis used was based on van der Westhuizen et al.³⁰ For more details about the discovery of the operational model and its mathematical background, see Black and Leff and Kenakin et al.^31,54,55 See Supplementary Methods for a full step-by-step procedure.

AEA reuptake inhibition in HaCaT cells

The uptake of [³H]AEA was measured in intact HaCaT cells (a kind gift of Prof N.E. Fusenig (German Cancer Research Center, Heidelberg, Germany), that were incubated in PBS at 37°C with a mixture of AEA [arachidonyl-5,6,8,9,11,12,14,15-³H] (200Cimmol⁻¹) and cold AEA (at a final concentration of 400nM) for 15min (ref. ⁵⁶). Control experiments were carried out also at 4°C and in the presence of the selective AEA reuptake inhibitor OMDM-1 (10μM). The effect of different ligands on AEA reuptake was tested by adding each substance directly to the incubation medium.

AEA reuptake inhibition in HMC-1 cells and U937 cells

Compound screening and IC50 determinations for AEA cellular uptake in U937 cells and HMC-1 cells (source: U937 cells were purchased from ATCC, Manassas, VA, USA, HMC-1 cells were a gift of Prof S. Ständer, University of Münster with the permission of the Mayo Foundation, USA) was performed in 96-well format using AquaSil silanized glass vials (Chromacol 1.1-MTV) (refs ^{32, 57}). First, required amounts of U937 cells were centrifuged at 100g for 5min and resuspended in RPMI (37°C) to a final concentration of 2 × 10⁶ cellsml⁻¹. Then, 250μl of cell suspension (0.5 × 10⁶ cells per sample) were transferred into the glass vials. After addition of 5μl vehicle (DMSO) or test compounds dissolved in DMSO at indicated final concentrations the cells were incubated at 37°C for 15min. After pre-incubation a mixture of 0.5nM [ethanolamine-1-³H]-AEA, (60Cimmol⁻¹) and 99.5nM of cold AEA (final 100nM) was added and the samples were incubated at 37°C for another 15min. In the FAAH-deficient HMC-1 cells, [arachidonoyl-5,6,8,9,11,12,14,15-³H]-AEA (200Cimmol⁻¹) was used in order to achieve a better signal-to-background ratio, due to the lower rate of AEA uptake. The reaction was stopped by rapid filtration over UniFilter-96 GF/C filters (Perkin Elmer) pre-soaked with PBS 1% BSA. Cells on filters were washed three times with 100μl ice-cold PBS buffer containing 1% fatty acid free BSA. After drying, 45μl MicroScint 20 scintillation cocktail (Perkin Elmer, Waltham, MA, USA) was added to the wells and the plate was sealed. Radioactivity was measured by liquid scintillation counting on a Perkin Elmer Wallac Trilux MicroBeta 1450 during 2min. Samples were measured in triplicates in n=3 independent experiments except the screening run (n=1). Each run was validated using the positive controls OMDM-2 and UCM707 at a concentration of 10μM reaching 59.7±6.6% (n=7) and 71.5±8.0% (n=7) respectively.

NAPE-PLD inhibition

Full length human cDNA NAPE-PLD was obtained from Natsuo Ueda⁵⁸ and cloned into mammalian expression vector pcDNA3.1, containing a C-terminal Flag-tag and genes for ampicillin and neomycin resistance. All plasmids were grown in XL-10 Z-competent cells and prepped (Maxi Prep, Qiagen). Sequence analysis for the confirmation of the sequences was performed at the Leiden Genome Technology Centre.

HEK293T cells (source: Dutch Cancer Institute) were cultured at 37°C and 7% CO₂ in DMEM with glutamax, penicillin (100μgml⁻¹), streptomycin (100μgml⁻¹) and 10% New Born Calf Serum iron supplemented (Hyclone SH30072.03). Cells were passaged twice a week to appropriate confluence. Twenty-four hours before transfection 10⁷ cells were seeded on a 15cm dish. Two hours before transfection, the medium was refreshed. Transfection is performed with PEI in a ratio of 3:1 with human NAPE-PLD or Mock pcDNA3.1Neo, 20μg per dish. Medium is refreshed after 24h and cells are harvested after 72h. Cell suspensions are centrifuged at 1,000g for 10min, supernatant removed and pellets frozen at −80°C until further use.

Cell pellets are re-suspended in lysis buffer 1: 20mM Hepes, 2mM DTT, 0.25M sucrose, 1mM MgCl₂, 2.5Uml⁻¹ benzonase and incubated 30min on ice. The cytosolic fraction (supernatant) is separated from the membranes by ultra-centrifugation (32,000rpm for 30min 100,000g). The pellet is resuspended in buffer 2: 20mM Hepes, 2mM DTT (membrane fraction). All samples are stored at −80°C. Enzyme concentrations are determined using a Bradford assay⁵⁹.

The membrane protein fraction from transient overexpression of NAPE-PLD in HEK293T cells was diluted to 0.4mgml⁻¹ in assay buffer: 50mM Tris-HCl (pH 7.5), 0.02% Triton X-100, 150mM NaCl (ref. ⁶⁰). The substrate PED6 (Invitrogen) 10mM stock was consecutively diluted in DMSO (25 × ) and in assay buffer (10 × ). Relevant concentrations of compounds are prepared in DMSO. The assay is performed in a dark Greiner 96-well plate, end volume 100μl. The compound or DMSO is incubated with membrane protein lysate (final concentration 0.04mgml⁻¹) for 30min at 37°C. A sample without membrane protein lysate is incorporated for background subtraction. Then, substrate is added (final concentration 1μM) and the measurement is started immediately on a TECAN infinite M1000 pro at 37°C (excitation 485nm, emission 535nm), scanning every 2min for 1h.

Cell culture and membrane preparation for DAGL, MAGL assay

Cell culture and membrane preparation were performed as previously described⁴³. HEK293T cells (source: Dutch Cancer Institute) were grown in DMEM with stable glutamine and phenolred (PAA), 10% New Born Calf serum, penicillin and streptomycin. Cells were passaged every 2–3 days by resuspending in medium and seeding them to appropriate confluence. Membranes were prepared from transiently transfected HEK293T cells. One day prior to transfection 10⁷ cells were seeded in a 15cm petri dish. Cells were transfected by the addition of a 3:1 mixture of polyethyleneimine (60μg) and plasmid DNA (20μg) in 2ml serum-free medium. The medium was refreshed after 24h, and after 72h the cells were harvested by suspending them in 20ml medium. The suspension was centrifuged for 10min at 1000, rpm, and the supernatant was removed. The cell pellet was stored at −80°C until use. Cell pellets were thawed on ice and suspended in lysis buffer A (20mM HEPES, 2mM DTT, 0.25M sucrose, 1mM MgCl₂, 25Uμl⁻¹ Benzonase). The suspension was homogenized by polytrone (3 × 7s) and incubated for 30min on ice. The suspension was subjected to ultracentrifugation (100,000 × g, 30min, 4°C, Beckman Coulter, Type Ti70 rotor) to yield the cytosolic fraction in the supernatant and the membrane fraction as a pellet. The pellet was resuspended in lysis buffer B (20mM HEPES, 2mM DTT). The protein concentration was determined with Quick Start Bradford assay (Biorad). The protein fractions were diluted to a total protein concentration of 1mgml⁻¹ and stored in small aliquots at −80°C until use.

Biochemical DAGLα activity assay

hDAGL-α activity was measured by the extent of the hydrolysis of para-nitrophenylbutyrate (PNP-butyrate) by membrane preparations from HEK293T cells (source: Dutch Cancer Institute) transiently transfected with hDAGL-α (ref. ⁴³). Percentage of inhibition of reference ligands was determined in comparison with an untreated control. Reactions were performed in 50mM pH 7.2 HEPES buffer with 0.05μgμl⁻¹ final protein concentration hDAGL-α transfected protein.

Biochemical MAGL activity assay

MAGL activity was measured by the extent of the production of glycerol from 2-AG hydrolysis by MAGL-overexpressing membrane preparations from transiently transfected HEK293T cells (source: Dutch Cancer Institute)⁶¹. Percentage of inhibition of reference ligands was determined in comparison with an untreated control. Standard assay conditions are as follows: 0.2Uml⁻¹ glycerol kinase (GK), glycerol-3-phosphate oxidase (GPO) and horseradish peroxidase (HRP), 0.125mM ATP, 10μM AmplifuRed, 5% DMSO, 25μM 2-AG and 0.5% acetonitrile in a total volume of 200μl. Fluorescence was measured in 5min intervals for 60min.

FAAH activity

FAAH activity was assessed using U937 cell homogenate (source: U937 cells were purchased from ATCC, Manassas, VA, USA). Tested compounds (at the concentrations of 1 and 5μM) or solvent (≤1% of the final volume) were pre-incubated at 37°C for 30min with 100μg of cell homogenate diluted in assay buffer (10mM Tris-HCl and 1mM EDTA, pH 7.6 and 1% w/v fatty acid-free BSA). Successively, a mixture of 99.5nMnM of AEA and 0.5nM of [ethanolamine-1-3H]-AEA (40–60Cimmol⁻¹) was added to the homogenate (final concentration of 100nM) and incubated for 10min at 37°C under shaking. After the incubation time, 2 volumes of an ice-cold methanol: chloroform mixture 1:1 (v/v) were added to each sample, vigorously vortexed and centrifuged at 10,000g for 10min at 4°C to separate aqueous and organic phases. The aqueous phases were collected and the radioactivity associated with the hydrolysis product [3H]-ethanolamine was measured after addition of 3ml of scintillation cocktail using the Tri-Carb 2100 TR scintillation counter. Data were collected from three independent experiments performed in triplicate and results were expressed as FAAH activity, relative to that in vehicle-treated samples (=100%).

COX2 activity

The COX2 activity was assessed using a COX fluorescent inhibitor screening assay kit from Cayman chemicals. The assay was performed in a final volume of 50μl in black 384-well non-binding microplates. Tested compounds (at the screening concentration of 5μM, 2.5μl) or solvent were incubated for 15min at 37°C with the following solution: 37.5μl of assay buffer (Tris-HCl, 100mM, pH=8), 2.5μl of heme (final concentration of 1μM), 2.5μl of COX-2 enzyme and 2.5μl of 10-acetyl-3,7-dihydroxyphenoxazine (ADHP) fluorometric substrate (final concentration of 30μM). The reaction was started by addition of arachidonic acid (AA) or 2-AG (2.5μl, final concentration of 10μM). Fluorescence signals were measured after 5min of incubation with a TECAN FARCyte Ultra (Ex. 535nm, Em. 580nm). The COX inhibitor DuP-697 (1μM) was used as a positive control. Spontaneous non-enzymatic cleavage of ADHP was quantified in samples where COX2 was replaced by assay buffer. The non-specific value was then subtracted from the measured fluorescence. Data were collected from three independent experiments performed in triplicate and results were expressed as COX2 activity, relative to that in vehicle-treated samples (=100%).

ABHD6 and ABHD12 activity

HEK293T cells were obtained from American Type Culture Collection (ATCC, Manassas, VA, USA). Lipofectamine 2000 was obtained from Life Technologies (Carlsbad, CA, USA) and geneticin from InvivoGen. Cell culture medium, FBS, and cell culture supplements were from Invitrogen (Carlsbad, CA, USA). pCMV6-AC-hABHD6 and pCMV6-XL4-hABHD12 were a kind gift from Prof Jarmo T. Laitinen. hABHD12 sequence was subsequently inserted in a pcDNA3.1 plasmid in order to generate a stable cell line. The effect of CB₂R ligands on ABHD6 and ABHD12 hydrolytic activity was determined using cell homogenate from hABHD6 and hABHD12 overexpressing HEK293T cells (40μg total protein per condition). Samples were pre-incubated with the compounds (10μM in the screening assay and concentrations in the range 1nM—50μM for the dose-response curves) in Tris 1mM, EDTA 10mM (pH=7.6) buffer containing 0.1%(w/v) BSA. DMSO was used as vehicle control and WWL70 10μM or THL 20μM as positive controls. [³H]-2-OG (10μM final concentration) was added and after incubation of 5min at 37°C, the reaction was stopped by the addition of 400μl of ice-cold CHCl₃:MeOH (1:1). The samples were vortex and centrifuged (16,100g 10min 4°C). Aliquots (200μl) of the aqueous phase were assayed for tritium content by liquid scintillation spectroscopy. The values obtained for each measurement were corrected for non-specific hydrolysis (non-ABHD6 mediated) by subtracting the signal obtained in the hydrolysis of [³H]- 2-OG by non-transfected HEK293T.

Activity-based protein profiling in mouse brain tissue

Mouse tissue was isolated according to guidelines approved by the ethical committee of Leiden University (DEC#13191) (ref. ⁴³). Animal care, tissue isolation and the following experimental procedures was all performed at the lab of Mario van der Stelt, Leiden University. Mouse tissues were homogenized in pH 7.2 lysis buffer (20mM Hepes, 2mM DTT, 1mM MgCl₂, 25Uml⁻¹ Benzonase) and incubated for 5min on ice, followed by low speed spin (2,500g, 3min, 4°C) to remove debris. The suspension was subjected to ultracentrifugation (100,000g, 45min, 4°C, Beckman Coulter, Type Ti70 rotor) to separate the cytosolic fraction from the membrane fraction. The pellet was resuspended in storage buffer (20mM Hepes, 2mM DTT). The total protein concentration was determined with Quick Start Bradford assay (Biorad) or Qubit protein assay (Invitrogen). Tissue proteome (2.5mgml⁻¹) was incubated with vehicle (DMSO) or inhibitor in 0.5μl DMSO (10μM final concentration) for 30min at rt, and subsequently incubated with broad spectrum ABPs MB064 (250nM final concentration) or TAMRA-FP (500nM final concentration) for 15min at rt (ref. ⁴³). The reactions were quenched with 10μl standard 3 × SDS-PAGE sample buffer. The samples were directly loaded (7μl=~12μg) and resolved on SDS page gel (10% acrylamide). The gels were visualized and scanned with a Bio-Rad Universal Hood III (Bio-Rad Laboratories B.V.) using Cy3/TAMRA settings (excitation wavelength 532nm, emission wavelength 580nm).

Off-target activity on TRP channels

HEK293 (human embryonic kidney) cells stably over-expressing recombinant human TRPV1 or rat TRPA1, TRPV2, TRPV3, TRPV4 or TRPM8 (source: human embryonic kidney cells were purchased by DSMZ (Germany), TRPV1-HEK-293 cells were a kind gift from John Davis, GlaxoSmithKline, Harlow, UK, the plasmid for TRPV2, as well as TRPV3-HEK-293 and TRPV4-HEK-293 were a kind gift from HB Bradshaw Indiana University, the plasmid containing TRPA1 was a kind gift from Sven-Eric Jordt then at Department of Cellular and Molecular Pharmacology University of California, San Francisco, California, USA and now at Department of Anesthesiology, Duke University School of Medicine, Durham, NC, USA, TRPM8-HEK-293 was a gift from Mario van der Stelt) were grown on 100mm diameter Petri dishes as mono-layers in minimum essential medium supplemented with non-essential amino acids, 10% FBS, 2mM glutamine, and maintained at 5% CO₂ at 37°C. Quantitative real-time analysis was carried out to measure TRP gene over-expression in transfected-cells (data not shown). On the day of the experiment, cells were loaded with the methyl ester Fluo-4 AM in minimum essential medium (4μM in DMSO containing 0.02% Pluronic F-127, Invitrogen), kept in the dark at room temperature for 1h, washed twice with Tyrode's buffer (145mM NaCl, 2.5mM KCl, 1.5mM CaCl₂, 1.2mM MgCl₂, 10mM D-glucose, and 10mM HEPES, pH 7.4), resuspended in the same buffer and transferred (about 100,000 cells) to the quartz cuvette of the spectrofluorimeter (Perkin-Elmer LS50B equipped with PTP-1 Fluorescence Peltier System; PerkinElmer Life and Analytical Sciences, Waltham, MA, USA) under continuous stirring. The effects on intracellular Ca²⁺ concentration ([Ca²⁺]_i) before and after the addition of various concentrations of test compounds was measured by cell fluorescence (λ_EX=488nm, λ_EM=516nm) at 25°C. The effects of compounds were normalized against the response to ionomycin (4μM) in each experiment. The increases in fluorescence in wild-type HEK293 cells (that is, not transfected with any construct) were used as baseline and subtracted from the values obtained from transfected cells. Efficacy was defined as the maximum response elicited by the compounds tested and was determined by comparing their effect with the analogous effect observed with 4μM ionomycin (Cayman), while the potency of the compounds (EC₅₀) was determined as the concentration required to produce half-maximal increases in [Ca²⁺]_i. Curve fitting (sigmoidal dose-response variable slope) and parameter estimation were performed with GraphPad Prism (GraphPad Software Inc., San Diego, CA, USA).

Antagonist/desensitizing behaviour was evaluated by adding the test compounds in the quartz cuvette 5min before stimulation of cells with agonists. In the case of human TRPV1-expressing HEK293 cells the agonist used was capsaicin (0.1μM, in the case of SR141716A 10nM was also used), which was able of elevating intracellular Ca²⁺ with a potency of EC₅₀=5.3±0.4nM and efficacy=78.6±0.6%.

For TRPV2, the rat TRPV2-HEK293 cells exhibited a sharp increase in [Ca²⁺]_i upon application of lysophosphatidylcholine (LPC) 3μM. The concentration for half-maximal activation was 3.40±0.02μM and efficacy was 91.7±0.5%.

In the case of TRPV3, rat TRPV3-expressing HEK-293 cells were first sensitized with the non-selective agonist 2-aminoethoxydiphenyl borate (100μM). Antagonist/desensitizing behaviour was evaluated against thymol (100μM), which showed an efficacy of 34.7±0.2% and a potency of EC₅₀=84.1±1.6μM.

In the case of rat TRPV4-expressing HEK-293 cells the agonist used was 4α-phorbol 12,13-didecanoate (4α-PDD) (1μM), which was able of elevating intracellular Ca²⁺ with a potency of EC₅₀=0.46±0.07μM, and an efficacy of 51.9±1.7%.

In the case of rat TRPM8-expressing HEK-293 cells, antagonist/desensitizing behaviour was evaluated against icilin at 0.25μM and 0.10μM. For icilin, efficacy was 75.1±1.1 and potency EC₅₀=0.11±0.01μM.

In the case of HEK-293 cells stably over-expressing recombinant rat TRPA1, the effects of TRPA1 agonists are expressed as a percentage of the effect obtained with 100μM allyl isothiocyanate, which showed a potency of EC₅₀=1.41±0.04μM and an efficacy of 65.9±0.5.

The effect on [Ca²⁺]_i exerted by agonist alone was taken as 100%. Data are expressed as the concentration exerting a half-maximal inhibition of agonist-induced [Ca²⁺]_i elevation (IC₅₀), which was calculated again using GraphPad. All determinations were performed at least in triplicate. Statistical analysis of the data was performed by analysis of variance at each point using ANOVA followed by the Bonferroni's test.

CEREP panel

Pharmacological profiles of the test compounds were generated. Data shown give the percentage of inhibition for binding assays and the percentage of inhibition for enzyme and cell-based assays at a test concentration of 10μM. The assays were performed at CEREP according to standard procedures described under http://www.cerep.fr/cerep/users/index.asp. For the first compounds being evaluated (WIN55212-2, CP55940, SR141716A, JWH133, HU308, Gp-1a and HU910) standard profiles as suggested by CEREP were applied. For the other compounds reduced panels concentrating on most predictive antitargets and omitting rarely hit sites were used.

Microsomal clearance

For human or mice, pooled commercially available microsome preparations from liver tissues are used (source: pooled human microsomes (BD UltraPool HLM 150,Lot 38289) and pooled male mouse microsomes (C57BL/6J, Lot 4339006) were purchased from Corning Incorporated (Woburn, USA))⁶². For human, ultra-pooled (150 mixed gender donors) liver microsomes are purchased to account for the biological variance in vivo. For the microsome incubations, 96 deep well plates are applied, which are incubated at 37°C on a TECAN (Tecan Group Ltd, Switzerland) equipped with Te-Shake shakers and a warming device (Tecan Group Ltd, Switzerland). The incubation buffer is 0.1M phosphate buffer at pH 7.4. The NADPH regenerating system consists of 30mM glucose-6-phosphate disodium salt hydrate; 10mM NADP; 30mM MgCl₂ × 6 H₂O and 5mgml⁻¹ glucose-6-phosphate dehydrogenase (Roche Diagnostics) in 0.1M potassium phosphate buffer pH 7.4.

Incubations of a test compound at 1μM in microsome incubations of 0.5mgml⁻¹ plus cofactor NADPH are performed in 96-well plates at 37°C. After 1, 3, 6, 9, 15, 25, 35 and 45min 40μl incubation solutions are transferred and quenched with 3:1 (v/v) acetonitrile containing internal standards. Samples are then cooled and centrifuged before analysis by LC-MS/MS.

Log peak area ratios (test compound peak area/internal standard peak area) are plotted against incubation time using a linear fit. The calculated slope is used to determine the intrinsic clearance: Clint (μlmin⁻¹ per mg protein)=−slope (min−1) × 1,000/[protein concentration (mgml)⁻¹]. Data are obtained from single experiments measured with multiple time-points.

Hepatocyte clearance

For animals, hepatocyte suspension cultures are either freshly prepared by liver perfusion studies or prepared from cryopreserved hepatocyte batches⁶³. For human, commercially available, pooled (5–20 donors), cryopreserved human hepatocytes from non-transplantable liver tissues are mainly used (source: primary, pooled human cryopreserved hepatocytes (Lot ECO) from nontransplantable liver tissues and pooled C57BL6 mouse hepatocytes (Lot PJJ) were purchased from BioreclamationIVT (NY, USA)). For the suspension cultures, Nunc U96 PP-0.5ml (Nunc Natural, 267245) plates are used, which are incubated in a Thermo Forma incubator from Fischer Scientific (Wohlen, Switzerland) equipped with shakers from Variomag Teleshake shakers (Sterico, Wangen, Switzerland) for maintaining cell dispersion. The cell culture medium is William's media supplemented with Glutamine, antibiotics, insulin, dexamethasone and 10% FCS.

Incubations of a test compound at 1μM test concentration in suspension cultures of 1 Mio cellsml⁻¹ (~1mgml⁻¹ protein concentration) are performed in 96-well plates and shaked at 900rpm for up to 2h in a 5% CO₂ atmosphere and 37°C. After 3, 6, 10, 20, 40, 60 and 120min 100μl cell suspension in each well is quenched with 200μl methanol containing an internal standard. Samples are then cooled and centrifuged before analysis by LC-MS/MS.

Log peak area ratios (test compound peak area/internal standard peak area) or concentrations are plotted against incubation time and a linear fit made to the data with emphasis upon the initial rate of compound disappearance. The slope of the fit is then used to calculate the intrinsic clearance: Clint (μlmin⁻¹1⁻¹ × 106 cells)=−slope (min−1) × 1,000/[1 × 106 cells]. Data are obtained from single experiments measured with multiple time-points.

Plasma protein binding

Pooled and frozen plasma from selected species were obtained from commercial suppliers (The pooled and frozen plasma from human (HMPLEDTA, Lot BRH1060627) and mouse (MSEPLEDTA3-C57, Lot MSE196204) were obtained from BioreclamationIVT (NY, USA).)^64,65. The Teflon equilibrium dialysis plate (96-well, 150μl, half-cell capacity) and cellulose membranes (12–14kDa molecular weight cutoff) were purchased from HT-Dialysis (Gales Ferry, CT, USA). Both biological matrix and phosphate buffer pH are adjusted to 7.4 on the day of the experiment. The reference substance is diazepam.

The determination of unbound compound is performed using a 96-well format equilibrium dialysis device with a molecular weight cut-off membrane of 12-14kDa. The equilibrium dialysis device itself is made of Teflon to minimize non-specific binding of the test substance. Compounds are tested in cassettes of 2–5 with an initial total concentration of 1000, nM, one of the cassette compound being the positive control diazepam.

Equal volumes of matrix samples containing substances and blank dialysis buffer (Soerensen buffer at pH 7.4) are loaded into the opposite compartments of each well. The dialysis block is sealed and kept for 5h at a temperature of 37°C and 5% CO₂ environment in an incubator. After this time, equilibrium will have been reached for the majority of small molecule compounds with a molecular weight of <600. The seal is then removed and matrix and buffer from each dialysis is prepared for analysis by LC-MS/MS. All protein binding determinations are performed in triplicates. The integrity of membranes is tested in the HTDialysis device by determining the unbound fraction values for the positive control diazepam in each well.

At equilibrium, the unbound drug concentration in the biological matrix compartment of the equilibrium dialysis apparatus is the same as the concentration of the compound in the buffer compartment. Thus, the percent unbound fraction (fu) can be calculated by determining the compound concentrations in the buffer and matrix compartments after dialysis as follows: %fu=100 × buffer conc after dialysis/matrix conc after dialysis. The device recovery is checked by measuring the compound concentrations in the matrix before dialysis and calculating the percent recovery (mass balance). The recovery must be within 80-120% for data acceptance.

P-glycoprotein binding

P-glycoprotein (permeability-glycoprotein, abbreviated as ‘P-gp' also known as multidrug resistance protein 1 (MDR1)) is the most studied and best characterized drug transporter. The P-gp assay evaluates the ability of test compounds to be transported transcellularly as a P-gp substrate⁶⁶. The assay uses transfected LLC-PK1 cells (porcine kidney epithelial cells, obtained from the Netherlands Cancer Institute) expressing human or mouse P-gp, cultured on 96-well semi-permeable filter membrane plates, where they form a polarized monolayer with tight junctions, and act as a barrier between apical and basolateral compartments. P-gp is expressed in the apical-facing membrane of the monolayer (tightness confirmed using Lucifer yellow). For substrate testing the assay determines the unidirectional permeability (Papp) of a test compound by separately dosing to the apical (for A>B Papp) and basolateral (for B>A Papp) sides of the cell monolayer (that is, donor compartments) and measuring the movement of the compound into the respective receiver compartments over a 3h incubation at 37°C. The effect of P-gp is measured by expressing the efflux ratio (ER) of the unidirectional A>B and B>A Papp values. The mean permeability (A>B and B>A Papp) is determined in the absence of P-gp via addition of the selective inhibitor zosuquidar at a concentration of 1μM. The ER and mean Papp are then used to categorize compound properties regarding their degree of efflux and permeability.

PAMPA

PAMPA (Parallel Artificial Membrane Permeability Assay) is a method which determines the permeability of substances from a donor compartment, through a lipid-infused artificial membrane into an acceptor compartment. Read-out is a permeation coefficient Peff drug as well as test compound concentrations in donor, membrane and acceptor compartments⁶⁷.

A 96-well microtiter plate completely filled with aqueous buffer solutions (pH 7.4/ 6.5) is covered with a microtiter filterplate like a sandwich construction. The hydrophobic filter material (Durapore/Millipore; pore size 0.22–0.45μm) of the first 48 wells (sample) of the filterplate is impregnated with a 1–20% solution of lecithin in an organic solvent (dodecane, hexadecane, 1,9-decadiene). The filter surface of the remaining 48 wells (reference) is wetted with a small volume (4–5μl) of a 50% (v/v) methanol/buffer solution. Transport studies were started by the transfer of 100–200μl of a 250 or 500μM stock solution on top of the filterplate in the sample and in the reference section, respectively. In general 0.05M TRIS, pH 7.4, or 0.05M phosphate, pH 6.5, buffers were used. The maximum DMSO content of the stock solutions was 5%.

Kinetic solubility

The solubility of a test compound in phosphate buffer at pH 6.5 from evaporated DMSO compound stock solution is measured over time, resulting in the kinetic solubility of the compounds.

Triad assay in mice using cumulative dosing procedure

These animal experiments were performed in the lab of Prof Lichtman, Virginia Commonwealth University. Male C57BL6/J mice were purchased from Jackson Laboratories (Bar Harbor, ME, USA) and were between 10 and 12 weeks of age during treatment. The animal protocol for the triad assay was approved by the Institutional Animal Care and Use Committee at Virginia Commonwealth University in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals⁶⁸. After testing was completed, all mice were humanely euthanized via CO₂ asphyxia, followed by rapid cervical dislocation. All studies involving animals are reported in accordance with the ARRIVE guidelines for reporting experiments involving animals⁶⁹.

Each subject (23–29g) was assessed for catalepsy, antinociception and hypothermia as a measure of in vivo CB₁ receptor activity using a cumulative dosing procedure to examine the dose-effect relationship of each drug tested⁷⁰. Before dosing, baseline measurements of catalepsy, antinociception and hypothermia were taken and mice were injected with the vehicle consisting of 1 part ethanol, 1 part Alkumus-620 Emulphor and 18 parts 0.9% saline. At 30min after injection, mice were again assessed for catalepsy, antinociception and hypothermia, which required 10min to test all mice. This process was repeated using cumulative doses of HU210 (0.03–3.0mgkg⁻¹), HU308 (1–100mgkg⁻¹), HU910 (1–100mgkg⁻¹), and JWH133 (1–100mgkg⁻¹) in half log increments, such that mice were injected every 40min and tested 30min after each injection until the end of the test session. Separate groups of mice (n=8) were utilized for each of the four drugs. All injections were given via the intraperitoneal (i.p.) route of administration in an injection volume of 10μl per 1g body mass.

For the antagonism experiments, mice from cumulative dosing experiments for HU308, HU910 and JWH133 were utilized again (after a 1 week washout period), to assess whether HU910 (30mgkg⁻¹) would antagonize the cannabimimetic effects of an ED₈₄ dose of HU210. For comparison, a separate experiment was conducted using the CB₁ receptor antagonist/inverse agonist rimonabant (3mgkg⁻¹). Mice were evenly distributed with respect to drug history across treatment groups in each of these two experiments assessing antagonism. Baseline responses were assessed prior to injections. Subjects in each experiment received an i.p. injection of either vehicle or the appropriate receptor antagonist at 0min, a second i.p. injection of vehicle or HU210 (1.7mgkg⁻¹, which reflected the approximate ED₈₄ value in producing antinociception and hypothermia) at 40min. Mice were tested for catalepsy, antinociception and hypothermia 30min after each injection.

Data from the dose-response studies were analysed via one-way repeated measures ANOVA followed by Dunnett's post hoc test (versus vehicle). ED₅₀ and ED₈₄ values were determined via linear regression for drugs eliciting maximal effects for a given endpoint. Data from the antagonism experiments were analysed via two-way ANOVA in which the factors were antagonist versus vehicle and HU210 versus vehicle. Post hoc analysis was conducted using the Holm-Sidak test. P values of<0.05 were defined as significant.

In vivo pharmacokinetic studies

Male C57BL/6J mice were obtained from CRL-F (Charles River, France), male Wistar rats were supplied by RCC Füllinsdorf and male C57BL/6JRj mice were from Janvier (France). Animals were up to 12 weeks of age. All animal studies in this section were performed at Hoffmann-La Roche and approved by the Bundesamt für Lebensmittelsicherheit und Veterinärwesen (BLV) der Schweizerischen Eidgenossenschaft. Test compounds were formulated according to respective protocols either by dissolution (i.v.) or in a glass potter until homogeneity was achieved (p.o) (formulations for p.o. administration: JWH133 and HU308 were formulated as a microsuspension in Gelatine/NaCl (7.5/0.62%) in water and HU910 was formulated as a solution in ethanol/cremophor EL/NaCl 0.9% (5/5/90%)). Formulations were injected i.v. using a 30G needle in the lateral tail vein of mice using a volume of 50μl in the dose indicated. For p.o. applications animals were gavaged using a volume of 100μl in the dose indicated. At the following time points blood was drawn into EDTA: 0.08, 0.25, 0.5, 1, 2, 4, 7h (for p.o. the first time point was omitted). Six animals were used for each compound in vivo experiment. Animals were distributed randomly over the time course and at each time point, a volume of 100μl of blood was taken. Quantitative plasma measurement of the compound was performed by LC-MS/MS analysis. Pharmacokinetic analysis was performed using Phoenix WinNonlin 6.4 software using a non-compartmental approach consistent with the route of administration. For assessment of the exposure Cmax, Tmax and AUC were determined from the serum concentration profiles. Parameters (CL, versus, T1/2) were estimated using nominal sampling times relative to the start of each administration. CO was extrapolated from the first concentration measured following intravenous administration.

This paper is dedicated to Prof Dr Raphael Mechoulam, the founder of cannabinoid research. M.S., L.H.H. and M.v.d.S. were supported by an ECHO-STIP grant from the Dutch Research Council-Chemical Sciences. A.M. was supported by the RPF program. Partial support from Ministero dell'Istruzione, dell'Università e della Ricerca (PRIN 2010–2011 grant to M.M.) is gratefully acknowledged. P.P. was supported by the Intramural Research Program of NIH/NIAAA. Mark Rau (laboratory J.G.) is acknowledged for measuring AEA uptake inhibition. Pharmacokinetics in vivo laboratories, Drug Disposition and Safety laboratories. Pharmaceutical Sciences pRED and F. Hoffmann-La Roche Ltd, Basel, Switzerland are acknowledged for performing the in vitro DMPK and in vivo studies. T.W.G and A.H.L were supported by NIH grants P30 DA033934 and T32DA007027.