GPR55 ligands promote receptor coupling to multiple signalling pathways

Cell culture and transfection

HEK293-AD cells (hereafter termed HEK293; Invitrogen) were maintained in Dulbecco's modified Eagle's medium DMEM-F12 supplemented with 10% calf serum at 37°C and 5% CO₂. For studies involving reporter gene and the dynamic mass redistribution (DMR) assay, cells were cultured in DMEM supplemented with 10% calf serum and 1% sodium pyruvate. Cell lines stably expressing 3xHA-GPR55 (GPR55-HEK293; see Henstridge et al., 2009) were maintained in G418 at 500 µg·mL^–1. For experimental studies cells were plated onto poly-D-lysine-coated coverslips (10 µg·mL^–1) and serum-starved for 12–48 h. Transient transfections were carried out using Lipofectamine 2000 according to the manufacturers instructions.

Ca²⁺ microfluorimetry

A digital epifluorescence imaging system (Perkin-Elmer, Emeryville, CA, USA) mounted on an Olympus BX50WI microscope was used to measure changes in intracellular Ca²⁺ concentrations, [Ca²⁺]_i. Cells were incubated with the Ca²⁺ sensitive dye Fura-2-AM (6 µM; 40–60 min at 25°C) in HEPES-buffered saline (HBS; in mM: NaCl 135, HEPES 10, KCl 5, CaCl₂ 1.8, MgCl₂ 1.0, D-glucose 25, pH 7.4). Ratiometric images were collected at 5 s intervals with data derived from individual cells. The effects of ligands were quantified by measuring the mean change in fluorescence ratio. All n values represent data (number of cells) obtained from a minimum of three separate experiments derived from different passages or transfections.

Immunoblot analysis

Samples were separated on 12% polyacrylamide gels and transferred to nitrocellulose membranes. Following treatment with blocking solution (5% milk), membranes were incubated with polyclonal rabbit antibodies detecting ERK1/2, cAMP response element binding protein (CREB) or phoshorylated ERK1/2 (pERK1/2), at a dilution of 1:1000. The monoclonal antibody against pCREB was used at a dilution of 1:1000 and HRP-conjugated secondary antibodies at a dilution of 1:10 000. Detection was performed using enhanced chemoluminescence reagents. Densitometry analysis of bands was carried out using Photoshop software (Adobe Systems Incorporated, San Jose, CA, USA). Absolute band intensity values were generated by multiplying mean intensity levels above background by the pixel size of the band. Absolute intensity values of phosphorylated protein bands (pERK or pCREB) were then divided by those for total protein (tERK or tCREB) to generate a ratio. All ratio values were then normalized to that observed with 1 µM LPI, run in the same gel.

Reporter gene assays

GPR55-HEK293 cells (20 000 cells per well; DMEM, 10% FCS) were seeded in 96-well plates coated with 1% poly-D-lysine. Cells were transiently transfected with pNFAT-luciferase (Luc; 200 ng.well^–1) or pNF-κB-Luc (50 ng.well^–1) reporter plasmids (PathDetect cis-Reporters; Stratagene) using Lipofectamine 2000 (Invitrogen). Twenty four h later, cells were incubated with increasing amounts of LPI, AM251, AM281 and SR141716A for 6 h in serum-free OptiMEM medium at 37°C. Luciferase activity was visualized by using the Steadylite Plus Kit (Packard Instrument Company, Meriden, CT, USA) as previously described (Waldhoer et al., 2002; Henstridge et al., 2009). Luminescence was measured in a TopCounter (Top Count NXT; Packard) for 5 s. Luminescence values are given as relative light units (RLU).

NFκB-(p65)-EGFP translocation assay

GPR55-HEK293 cells were grown on coverslips to 70% confluency in DMEM supplemented with 10% FCS. Cells were transfected with 3 µg of the phosphorylated form of enhanced green fluorescent protein (EGFP)-p65 plasmid (Waldeck-Weiermair et al., 2008) and 12 h post transfection, cells were serum-starved in OPTIMEM for another 12 h. Cells were treated with the indicated concentrations of ligands for 15 min and the experiment was terminated by fixing cells with 3.7% formaldehyde. Fixed cells were washed twice with TBS (NaCl 135 mM, Tris-HCl 25 mM, CaCl₂ 1 mM, KCl 2.5 mM) and wet-mounted onto microscopy slides with Vectashield containing a DAPI dye to visualize nuclei (Roche). Images were taken with an Olympus inverted fluorescence microscope equipped with a Hamamatsu Orca CCD camera.

DMR assays

A beta version of the Corning^® Epic^® System (Corning, NY, USA) was used consisting of a temperature-control unit, an optical detection unit and an on-board robotic liquid handling device. Briefly, each well in the 384-well Epic^® microplate contains a resonant waveguide grating biosensor. The system measures changes in the local index of refraction upon mass redistribution within the cell monolayer grown on the biosensor. Ligand-induced DMR in living cells is induced upon receptor stimulation and is manifested as a shift in the wavelength of light that is reflected from the sensor. The magnitude of this wavelength shift in picometres (pm) is proportional to the amount of DMR. Notably, GPCR activation is translated into signalling pathway-specific optical signatures. Further technical details and the assay principle have been provided earlier (Fang et al., 2006; 2007; Schröder et al., 2009). Forty eight h before the assay, naïve HEK293 cells or GPR55-HEK293 cells were seeded at a density of 7500 cells per well in 384-well Epic^® sensor microplates with 30 µL growth medium (DMEM, 10% FCS) and cultured for 24 h (37°C, 5% CO₂) to obtain confluent monolayers. For starvation, cell culture medium was removed by aspiration and replaced by 30 µL of starvation medium [Hank's buffered salt solution (HBSS) with 20 mM HEPES, pH 7.15] in which cells were then cultured for another 24 h (37°C, 5% CO₂). Previous to the assay, cells were washed once with assay buffer [HBSS with 20 mM HEPES and 0.1% bovine serum albumin, fatty acid-free (BSAfaf), pH 7.15] and incubated in 30 µL per well of assay-buffer in the Epic^® reader at 28°C. Hereafter, the sensor plate was scanned and a baseline optical signature was recorded. Ten µL of test compound dissolved in assay buffer was then transferred from the compound plate into the sensor plate and DMR responses were monitored for at least 3600 s.

Receptor endocytosis

GPR55-HEK293 cells grown on coverslips were incubated with a rabbit anti-HA antibody (1:1000) for 30–45 min at 37°C (in HBS) to label surface receptors (‘antibody feeding’). Subsequently, cells were treated with ligand for 60 min at 37°C, then fixed in 4% paraformaldehyde for a further 15 min. Pre-labelled receptors remaining at the cell surface were detected with an Alexa488-conjugated anti-rabbit secondary antibody (1:500; 30 min). Subsequently, cells were permeabilized with 0.2% Triton X-100 (15 min) and internalized receptors were visualized with Cy3 or Cy5-conjugated anti-rabbit secondary antibodies. Images obtained with Cy5 were pseudocoloured red using Zeiss Lasersharp software to aid visualization.

Data analysis

Dose–response curves and EC₅₀ values were determined by nonlinear regression using Prism 4.02 (GraphPad Software Inc., CA, USA). For analysis of DMR data, the area under curve (AUC) values of DMR signals between the 1200 s and 3600 s time points with mean and standard error of the mean (SEM) were used to calculate agonist activities. AUC values were transformed into relative AUC to give equivalent baseline optical recordings for all dose–response curves from one assay plate. Data were then normalized and expressed as percent of maximum activation induced by a saturating concentration of LPI which was set 100%. The assays were performed in triplicate and figures are representative data of two to three independent experiments.

Statistical analyses were performed using analysis of variance (anova) for comparisons between multiple groups, followed by Bonferroni's post hoc analysis (using GraphPad Prism and Mirocal Origin software). P values <0.05 were considered to be significant.

Materials

Cell culture reagents, Alexa Fluor 488-tagged secondary antibodies and Lipofectamine 2000 were from Invitrogen (Paisley, UK). Rabbit anti-HA antibody was obtained from Abcam (Cambridge, UK). Cy3 and Cy5-tagged secondary antibodies were obtained from Jackson Immuno Research (West Grove, PA, USA). Monoclonal HA.11 antibody was obtained from Covance (Berkeley, CA, USA). Antibodies against pERK were obtained from Cell Signaling Technology (Danvers, MA, USA) and pCREB from Millipore (Upstate, Billerica, MA, USA). Horseradish peroxidase-conjugated secondary antibodies were purchased from Thermo Scientific (Pierce, Rockford, IL, USA). The pNFAT-Luc and pNF-kB-Luc reporter plasmids were obtained from Stratagene (La Jolla, CA, USA). SR141716A was a gift from Sanofi-Synthélabo Recherche (Montpellier, France). AM281 and AM251 were obtained from Tocris Cookson (Avonmouth, UK). L-α-lysophosphatidylinositol mixture (LPI; 58% C₁₆, 42% C₁₈), G418, poly-D-lysine, foetal calf serum (FCS), BSAfaf, HEPES and all other chemicals were obtained from Sigma-Aldrich (Dorset, UK or Taufkirchen, Germany).

Effects of GPR55 ligands on Ca²⁺ signalling

We and others have shown that the activation of GPR55 by various ligands can promote the release of Ca²⁺ from intracellular stores (Oka et al., 2007; Lauckner et al., 2008; Henstridge et al., 2009). Here we have extended upon these findings to evaluate the effects of a series of well-described cannabinoid antagonists/inverse agonists on Ca²⁺ signalling in GPR55-HEK293 cells. Consistent with our previous work (Henstridge et al., 2009), a brief 5 min exposure of GPR55-HEK293 cells (Figure 1A) to LPI (300 nM, panel a) and AM251 (3 µM, panel b) resulted in the induction of oscillatory Ca²⁺ transients. This effect was concentration-dependent (Figure 1B), with EC₅₀ values for LPI and AM251, shown in Table 1). Similar effects were observed with SR141716A (Figure 1A, panel c and 1B; Table 1), a close structural analogue of AM251 (see Figure S1). The related diarylpyrazole ligand AM281 was also able to promote Ca²⁺ signalling; however, it was much less potent, with responses observed over the range of 10–30 µM (Figure 1A, panel d and 1B; Table 1). No effects of any ligands on intracellular Ca²⁺ levels were observed in control HEK293 at the concentrations tested.

As we have recently shown that the GPR55 agonist LPI can induce NFAT-activity in a concentration-dependent manner (Henstridge et al., 2009), we also compared the ability of the cannabinoid ligands to promote NFAT activation. LPI (Figure 1C) induced NFAT-luciferase activity and AM251 (Figure 1C) was as potent and efficacious as LPI (Table 1). High concentrations of SR141716A (>3 µM) appeared to suppress NFAT activation (Figure 1C), preventing an accurate determination of EC₅₀ and E_max values. However, AM281 (Figure 1C) was clearly the least potent compound tested in activating NFAT (Table 1).

Activation of ERK1/2 MAP-kinase

Although GPR55 activation has been suggested to promote ERK1/2 signalling, this is controversial, having only been observed with LPI and not with cannabinoid ligands (Oka et al., 2007; Lauckner et al., 2008; Kapur et al., 2009). We therefore profiled the ability of various GPR55 ligands to activate ERK1/2 in GPR55-HEK293 and control HEK293 cells. In GPR55-HEK293 cells, LPI induced a marked, concentration-dependent increase in ERK1/2 phosphorylation (Figure 2A), with a peak response observed after 25 min stimulation (1 µM LPI; data not shown). No significant response was observed in control HEK293 cells at this time point (Figure 2A, HEK). In GPR55-HEK293 cells, threshold pERK responses were observed at 30 nM of LPI, with peak activation apparent at 300 nM LPI (Figure 2D) and, overall, with similar potency as already observed for Ca²⁺ signalling (Table 1).

Next, we investigated the effects of AM251 and SR141716A on ERK1/2 signalling (Figure 2B–C). Interestingly, both cannabinoid ligands promoted ERK phosphorylation, with a similar efficacy to that observed with LPI. Both AM251 and SR141716A were slightly more potent than observed with the Ca²⁺ signalling assay (Figure 2D; Table 1). AM281 was clearly less potent at ERK activation than the other diarylpyrazole ligands, with effects only observed with concentrations >3 µM (Figure 2D). No ERK activation was observed in control HEK293 cells following treatment with any of these ligands (Figure 2A–C, HEK).

Activation of CREB transcription factor via GPR55

The CREB transcription factor activates the transcription of target genes in response to a diverse array of stimuli, including hormones and growth factors, that activate not only protein kinase A (PKA), but also a variety of other kinases such as ERK MAP-kinases and Ca²⁺/calmodulin-dependent protein kinases (CaMKs) (see Shaywitz and Greenberg, 1999). As LPI (Oka et al., 2007) and the cannabinoid antagonists/inverse agonists AM251, AM281 and SR141716A were able to activate the ERK MAP-kinase pathway (see Figure 2), we tested whether CREB, being a downstream target of ERK-MAP kinases and Ca²⁺ signalling, can also be activated by these ligands.

We investigated whether LPI, AM251, AM281 and SR141716A were capable of activating CREB using immunoblot detection of phospho-CREB (pCREB) (Figure 3). Treatment of GPR55-HEK293 cells with LPI induced a marked, concentration-dependent increase in CREB phosphorylation (Figure 3A), with a peak response observed after 25 min stimulation (1 µM LPI; data not shown). In GPR55-HEK293 cells, threshold pCREB responses were observed at 10 nM, with peak activation apparent in the range 0.3–1 µM LPI (Figure 3D; Table 1). Next, we investigated the effects of AM251 and SR141716A on CREB signalling (Figure 3B and 3C). Both these ligands exhibited a higher efficacy than LPI, with peak responses approximately 150–175% of the LPI response (Figure 3D). As with the pERK activation, both AM251 and SR141716A were more potent at inducing pCREB than triggering Ca²⁺ transients (Table 1). AM281 was also less potent at CREB activation than the other diarylpyrazole ligands, with effects observed with concentrations >3 µM (Figure 3D; Table 1). No CREB activation was observed in control HEK293 cells following treatment with these ligands (Figure 3A–C, HEK).

Activation of NF-κB transcription factor via GPR55

Recently, we have shown that the endogenous cannabinoid ligand anandamide can activate the transcription factor NF-κB in endothelial cells (Waldeck-Weiermair et al., 2008); however, we did not determine whether this was a GPR55-mediated effect. In addition, NF-κB transcription can be regulated by ERK MAP-kinases (Delfino and Walker, 1999). Hence, we tested whether LPI, AM251, AM281 and SR141716A can activate NF-κB transcription factor and nuclear translocation in GPR55-HEK293 cells.

As Figure 4A shows, LPI was as potent (Table 1) and efficacious in inducing NF-κB luciferase activity as NFAT activation (compare with Figure 1C). High concentrations of AM251 and SR141716A (>3 µM) appeared to suppress NF-κB activity, preventing accurate determinations of EC₅₀ and E_max values. However, AM281 was clearly less potent (Figure 4A; Table 1).

Next, we tested whether these ligands could induce nuclear translocation of p65, the functional subunit of NF-κB, whose nuclear translocation upon stimulation can be monitored to report activation of this pathway, using EGFP-labelled p65 (Waldeck-Weiermair et al., 2008). While p65-EGFP was predominantly localized to the cytosol and excluded from the nucleus in cells treated with vehicle (DMSO) for 15 min (Figure 4B), cells treated with 1 µM LPI, 3 µM AM251, 3 µM of SR141716A or 10 µM AM281 all showed a redistribution of p65-EGFP (green) into the cell nucleus (overlay with DAPI, blue; Figure 4B). Cells treated with lower concentrations of LPI, AM251, AM281 or SR141716A did show sporadic p65-EGFP translocation to the nucleus, but to a lesser extent (data not shown).

Screening for GPR55 agonists using the label-free DMR assay

Given the difficulties associated with establishing efficient screening assays for GPR55, we evaluated an alternative – label-free and non-invasive – methodology based on optical detection of dynamic changes in cellular density following receptor activation (Fang et al., 2007; Schröder et al., 2009). Here, we employed the novel resonant wave guide grating biosensor technology (Corning Epic^®) to resolve the real-time signalling capacity of GPR55 upon activation with the set of compounds identified in the assays above. In brief (see Schröder et al., 2009), activation of GPCRs is known to cause a translocation of multiple signalling molecules upon receptor stimulation; thus, the DMR of intracellular content is monitored as an optical signal. Typically, these signals, generated upon ligand addition, reflect the G protein coupling profile of the respective receptors, i.e. ‘optical signatures’ and are distinct for Gα_i, Gα_s, etc.-coupled receptors (Fang et al., 2007; Schröder et al., 2009).

The optical signals for increasing concentrations of LPI (0.01 nM to 10 µM) were recorded in GPR55-HEK293 cells (Figure 5A), and control HEK293 cells (Figure 5B). A concentration–response curve for LPI using the AUC values of the optical signal traces, illustrated in Figure 5A, is shown in Figure 5C and revealed an EC₅₀ value in the low nanomolar range for LPI, thus appearing about 5-fold more potent than when measuring the Ca²⁺ assay (compare with Figure 1B; Table 1).

Next, we tested the capacity of AM251, AM281 and SR141716A to elicit optical signals from GPR55-HEK293 (Figure 5C) and control HEK293 (Figure 5D) cells and calculated their respective potencies and efficacies (Table 1). All three ligands were approximately two log units less potent than LPI in this assay (Figure 5C), reflecting a rank order similar to those found for all other assays employed so far. However, using the Epic^® real-time optical measurements, all compounds were potent, being equivalent to the most sensitive signalling assays readouts for detecting GPR55-mediated responses. In summary, this assay approach proved extremely suitable for non-invasively screening compounds on this Gα₁₃-coupled GPCR.

Agonist-induced GPR55 receptor internalization

Lastly, we evaluated whether our candidate ligands were capable of inducing GPR55 internalization, using a modified antibody feeding protocol that distinguishes between cell surface and intracellular receptors (Coutts et al., 2001). Following antibody prelabelling of live GPR55-HEK293 cells for 30 min, cells were treated with the compounds and subsequently processed for cell surface and intracellular HA-immunolabelling (Figure 6). Analysis of HA-immunoreactivity using confocal microscopy showed that, under control conditions GPR55 was predominantly located on the cell surface (Figure 6, DMSO, green; Henstridge et al., 2009). However, following treatment with LPI (1 µM), SR141716A (3 µM) and AM251 (3 µM) for 60 min, a pronounced redistribution of GPR55 into intracellular vesicles was observed (red). Interestingly, only a modest effect on GPR55 internalization was seen with AM281 (30 µM).

GPR55 signalling and pharmacology

In this study, we demonstrated that agonist interactions with GPR55 promoted the activation of a range of signalling pathways, including Ca²⁺ release, NFAT-, CREB- and NF-κB-transcription, ERK1/2 phosphorylation, Epic^® traces and receptor internalization. It is becoming increasingly clear that the coupling efficiency of a GPCR to downstream effector systems can be selectively modulated by distinct ligands, a phenomenon sometimes referred to as ligand biased signalling (Kenakin, 2007; Galandrin et al., 2008). Although with the different GPR55 assays employed there were some notable differences in ligand efficacy, the rank order of potency for the ligands was consistent throughout this study. Thus, while there may be some evidence for agonist bias with GPR55, it has only a relatively modest impact on the observed pharmacology. The most marked difference was observed with the CREB assay, where LPI was clearly much less efficaceous than AM251 or SR141716A. An increasing number of studies have shown ERK MAP kinase activation with GPR55, both in recombinant systems (Oka et al., 2007; Kapur et al., 2009) and also in cells that express wild-type GPR55, including microglia (Pietr et al., 2009) and osteoclasts (Whyte et al., 2009). However, in some cases this is only observed with LPI and not with cannabinoid ligands (Oka et al., 2007; Kapur et al., 2009). This discrepancy may be due to a lack of concentration–response profiling with cannabinoid ligands in the different assays, as we find cannabinoids to have a similar efficacy as LPI for ERK activation, although they are less potent. In some of the reporter gene studies with SR141716A (NF-κB and NFAT assays) and AM251 (NF-κB assay) responses were suppressed at high ligand concentrations (>3 µM) preventing accurate determination of EC₅₀ and E_max values. This may reflect non-specific effects and/or toxicity in these long-duration assays.

Cannabinoid activity at GPR55: implications for CB₁ receptor antagonists

The widely used diarylpyrazole CB₁ receptor antagonists/inverse agonists AM251 and SR141716A elicited a significant agonist response via GPR55 at concentrations typically encountered in experimental studies of CB₁ receptor function. These data are consistent with previous studies reporting G protein activation with AM251 (Henstridge et al., 2009) and recently with beta-arrestin (Kapur et al., 2009; Yin et al., 2009) and pGL3-CRE-MRE-SRE-luc reporter assays (Yin et al., 2009) for both AM251 and SR141716A. The related compound, AM281 was a less potent agonist at GPR55, although its ability to bind to and antagonize the CB₁ receptor is comparable with that of the other diarylpyrazole ligands (Lan et al., 1999). This suggests that the characteristics of the carboxamide group at position 3 of the diarylpyrazole backbone may be more important for ligand binding to GPR55 than for binding to the CB₁ receptor, within this chemical series.

Although all the diarylpyrazole compounds induce GPR55 activity, it was generally observed that much higher concentrations than are typically used for CB₁ receptor antagonsim [5–15 nM (Pertwee, 2005; Ryberg et al., 2007)] were required to elicit an effect. Thus, acute, physiological effects of low nM concentrations of these agents are likely to reflect CB₁ receptor blockade. However, a role for GPR55 activation needs to be considered where micromolar concentrations of SR14171A and AM251 are used, a situation that is encountered with many functional/physiological studies of cannabinoid action. In addition, it is possible that long-term treatment with SR141716A (Rimonabant^®) may have some impact on GPR55 function, especially if it accumulates in fatty tissues, such as the brain, and may therefore contribute to some of the biological activity and/or side-effects observed with this compound in humans.

Which GPR55 assay is best?

This study demonstrates that GPR55 mediates the activation of a plethora of downstream signalling pathways. However, evidence for agonist bias in their activation means that the pharmacology of GPR55 is influenced by the assay used to assess receptor function, and in some cases this could miss potential ligands. We therefore evaluated a novel label-free assay that should circumvent these problems, by allowing an integrated readout of GPR55 function. Indeed, using this approach, we were able to detect agonist activity for all GPR55 ligands tested. GPR55 pharmacology has also been assessed previously using beta-arrestin assays, and produced similar agonist profiles to the present study (Kapur et al., 2009; Yin et al., 2009). However, LPI was notably less potent with these assays, giving an EC₅₀ values in the low µM range compared with the 9 nM value obtained from the Epic^® real-time optical measurements.