Flipping the GPCR Switch: Structure-Based Development of Selective Cannabinoid Receptor 2 Inverse Agonists

Abstract

We report a blueprint for the rational design of G protein coupled receptor (GPCR) ligands with a tailored functional response. The present study discloses the structure-based design of cannabinoid receptor type 2 (CB₂R) selective inverse agonists (S)-1 and (R)-1, which were derived from privileged agonist HU-308 by introduction of a phenyl group at the gem-dimethylheptyl side chain. Epimer (R)-1 exhibits high affinity for CB₂R with K_d = 39.1 nM and serves as a platform for the synthesis of a wide variety of probes. Notably, for the first time these fluorescent probes retain their inverse agonist functionality, high affinity, and selectivity for CB₂R independent of linker and fluorophore substitution. Ligands (S)-1, (R)-1, and their derivatives act as inverse agonists in CB₂R-mediated cAMP as well as G protein recruitment assays and do not trigger β-arrestin–receptor association. Furthermore, no receptor activation was detected in live cell ERK_1/2 phosphorylation and Ca²⁺-release assays. Confocal fluorescence imaging experiments with (R)-7 (Alexa488) and (R)-9 (Alexa647) probes employing BV-2 microglial cells visualized CB₂R expressed at endogenous levels. Finally, molecular dynamics simulations corroborate the initial docking data in which inverse agonists restrict movement of toggle switch Trp258^6.48 and thereby stabilize CB₂R in its inactive state.

Short abstract

We report a generalizable strategy for structure-based agonist-to-inverse-agonist functional transformation and probe development by ligand modification that modulates the GPCR toggle switch of CB₂R.

Introduction

The endocannabinoid system is present in all vertebrates and comprises endogenous ligands, enzymes mediating ligand metabolism, transporters, and the two prominent cannabinoid receptors type 1 and type 2 (CB₁R and CB₂R).^1,2 Exploitation of the therapeutic potential of CB₂R has primarily focused on receptor activation with agonists and showed promise to ameliorate a plethora of diseases, such as autoimmune³ and metabolic disorders,^4,5 chronic pain,⁶ and multiple sclerosis.⁷ By contrast, CB₂R antagonists and inverse agonists remain vastly underexplored despite encouraging results in models of arthritis⁸ and neuroinflammation.⁹⁻¹¹ Notably, CB₂R antagonist TT-816 is currently being investigated in a phase II clinical trial as an immune checkpoint inhibitor for the treatment of solid tumors.¹²

Despite the current considerable endeavors to deliver selective CB₂R therapeutics,¹²⁻¹⁹ to date there are no such drugs available on the market. Poor understanding of CB₂R localization, expression, and signaling on the molecular level are key factors responsible for this absence.²⁰ Elucidation of CB₂R pharmacology has been hampered by the insufficient specificity of monoclonal antibodies²¹⁻²⁴ and the scarcity of reliable chemical probes.²⁵ Although some potent, selective, and validated fluorescent probes have been reported,²⁶⁻²⁸ these function as agonists that disturb cellular homeostasis by triggering downstream signaling and β-arrestin association, followed by agonist-mediated receptor internalization.^29,30 These limitations may be addressed by implementation of inverse agonist based fluoroprobes that do not prompt receptor endocytosis. Additionally, inverse agonists tend to possess greater affinity for receptors in the more populous inactive G protein coupled receptor (GPCR) conformation yielding improved specificity and signal-to-noise ratio of probes compared to agonists.³¹

Historically, development of high-affinity, selective fluorescent CB₂R inverse agonists has proven arduous. In the cases reported, fluorophore conjugation completely ablated³² or materially reduced³³ affinity. In one example, a study of a series of agonists led to the identification of a specific linker–fluorophore construct endowing inverse agonism in a cAMP assay.³⁴ Development of a potent, selective, and versatile CB₂R-targeting inverse agonist scaffold that can be conjugated to a variety of fluorophores and functionalities remains an unmet challenge.

Since its discovery in 1999,³⁵ CB₂R-selective agonist HU-308 (Scheme 1) has enjoyed privileged status for the study of CB₂R pharmacology.³⁶ HU-308 has been extensively applied to unravel effects of CB₂R activation in animal models of pain,³⁷ osteoporosis,³⁸ Parkinson’s disease,³⁹ and amyotrophic lateral sclerosis⁴⁰ and is currently investigated in phase I clinical trials for mitigation of inflammation.¹⁹ The pharmacophore embedded in HU-308 has served in the development of photoswitchable,⁴¹ fluorescent,^28,42 and ligand-directed covalent probes.²⁶ On the basis of our prior work with HU-308, we focused on this scaffold with the intent of transforming its functional profile from agonist to inverse agonist with minimal structural modification. In this respect, Schapira and Jones have independently discussed the conceptual benefits of working with a set of molecules closely related in structure to enable in-depth understanding of receptor pharmacology.^43,44

Scheme 1. Novel Structure-Based Design of HU-308-Derived CB₂R-Selective Inverse Agonists That Actuate Trp258^6.48 Toggle Switch.

The past two decades witnessed the exponential rise of reported GPCR structures; hence, structure-based ligand design is at present ideally positioned to capitalize on the ongoing revolution.⁴⁵ Since more than a third of all approved drugs exert their action by GPCR modulation, it is vital to comprehensively investigate and understand receptor pharmacology with functionally orthogonal chemical probes.⁴⁶ GPCRs of the most populous class A family are distinguished by high homology of the CWxP motif. In particular, the toggle switch of CWxP that modulates receptor activation, Trp258^6.48, is conserved within 78% of nonolfactory GPCRs.⁴⁷ Examination of the X-ray structure of CB₂R in its inactive conformation revealed a secondary binding pocket that hosts Trp258^6.48. Further investigation by in silico docking suggested that addition of a substituent at C(2′) of HU-308 might constrain Trp258^6.48 and hence modulate CB₂R activation (Scheme 1).

We report novel inverse agonists that demonstrate avid binding at CB₂R with excellent selectivity over the closely related CB₁R. The compounds were profiled for their functional response in a comprehensive panel of in vitro (β-arrestin and G protein recruitment) as well as cellular (cAMP, ERK_1/2 phosphorylation, Ca²⁺ signaling) assays. Remarkably, none of the probes activate CB₂R-mediated signaling in any of the tested pathways. Fluorescent probes demonstrated excellent specificity and visualized CB₂R expressed at endogenous levels in live-cell confocal microscopy experiments. Finally, molecular dynamics simulations investigated structural determinants that prevent receptor activation upon ligand binding and corroborate movement restriction of Trp258^6.48. The workflow and key considerations described herein may be used to successfully drive future structure-based switch of functionality involving ligands and proteins beyond HU-308 and CB₂R.

Results and Discussion

In Silico Probe Design

We have investigated the recently published active (PDB 8GUS, Figure 1A)⁴⁸ and inactive (PDB 5ZTY, Figure 1B)⁴⁹ conformations of CB₂R crystallized with agonist HU-308 and antagonist/inverse agonist AM10257, respectively. Close examination of the two receptor conformations revealed that AM10257 reaches into a secondary binding pocket that features a highly conserved CWxP motif in class A GPCRs⁵⁰ and moreover hosts Trp258^6.48, the recently designated single residue toggle switch of CB₂R activation.⁵¹

Figure 1.

Comparison of active (A, PDB 8GUS, ligand HU-308)⁴⁸ and inactive (B, PDB 5ZTY, ligand AM10257)⁴⁹ CB₂R conformations. (C) Docking study of HU-308-derived putative inverse agonist (R)-1 in the inactive CB₂R conformation (PDB 5ZTY). (R)-1 reaches into the secondary pocket occupied by the toggle switch responsible for CB₂R activation, Trp258^6.48 (orange), and shares binding interactions virtually identical with those of AM10257.

Comparison of the two receptor conformations combined with in silico docking suggested that a phenyl substituent introduced α to the gem-dimethyl group of HU-308 might occupy the same lipophilic subpocket as the phenyl of AM10257. The phenyl substitution creates a new C(2′) stereocenter at the pendent side chain; accordingly the explicit (S) and (R) designations preceding compound labels denote its absolute configuration. Additionally, structural features were incorporated that proved critical in our prior works to bestow excellent pharmacological profiles, yielding ligands (S)-1 and (R)-1 (Figure 2).^26,28,42 Namely, terminal azide was inserted and the allylic alcohol was substituted by an amine to allow facile, stable conjugation to fluorophores and confer improved affinity and selectivity for CB₂R. The novel putative HU-308-derived inverse agonist (R)-1 showed binding interactions virtually identical with those of AM10257 in the inactive CB₂R conformation (Figure 1C). In particular, the C(2′) phenyl group of (R)-1 oriented toward Trp258^6.48 and attained a favorable edge-to-face π-interaction similar to the phenyl of AM10257. The nearly identical interactions are essential as we have hypothesized that the impediment of the upward movement of Trp258^6.48 may effectively prevent receptor activation.

Figure 2.

Design of inverse agonists (S)-1 and (R)-1.

Synthesis

Access to (S)-1 and (R)-1 that feature a phenyl group at the homobenzylic position of cannabinoid scaffolds α to a sterically demanding gem-dimethyl group is synthetically challenging and unprecedented. Prior structure–activity relationship studies on cannabinoid ligands focused almost exclusively on the easily accessible benzylic or ω-position of the pendent side chain.⁵² To the best of our knowledge, there is only a single report of substitution at the homobenzylic position with a methyl group in a structure of Δ⁸-THC that lacks the sterically congesting gem-dimethyl motif.⁵³

The synthesis of (S)-1/(R)-1 commenced with 10, which was prepared by methylation of 3,5-dimethoxyphenylacetonitrile and subsequent treatment with phenyl lithium (Scheme 2).^54,55 Introduction of the alkyl side chain to ketone 10 proved a formidable challenge due to steric hindrance. Initial attempts using established phosphonium ylide routes yielded no reaction even at elevated temperatures.^56,57 An extensive screening of Grignard reagents either yielded no reaction or afforded exclusively Grignard reduction product 11. Finally, using a modified procedure for the preparation of alkyl lithiums by Punzalan,⁵⁸ we employed, for the first time, 5-chloropentyl lithium to forge the tertiary alcohol 12 in 83% yield.⁵⁹ Subsequent Chugaev elimination of the benzylic alcohol under mild conditions yielded 13 in 95% yield. BBr₃-mediated demethylation followed by high-pressure hydrogenation over Pd/C afforded (S)-14/(R)-14 as a racemic mixture in 94% yield over two steps. The synthesis was continued with the racemate to rapidly access material for initial pharmacological evaluation. To this end, Friedel–Crafts allylation with verbenol derivative 15 followed by treatment with (MeO)₂SO₂ furnished methylated epimeric mixture (S)-16/(R)-16 in 50% yield over the two steps. Subsequent substitution of the primary alkyl chloride with NaN₃ and hydrazine-mediated phthalimide deprotection revealed allylic amines (S)-1/(R)-1 in 73% yield. Finally, the synthesis was concluded by functionalization of the diastereomeric mixture (S)-1/(R)-1 with DY-480XL or 4-pentynoic acid to yield mixtures of epimers (S)-2/(R)-2 and (S)-3/(R)-3, respectively.

Scheme 2. Synthesis of Novel CB₂R-Selective HU-308-Derived Inverse Agonists.

Reagents and conditions: (a) 1-chloro-5-iodopentane, t-BuLi, n-pentane, Et₂O, −78 °C to rt, 83%; (b) KHMDS, CS₂, THF, −78 °C to rt and then MeI, 40 °C, 95%; (c) BBr₃, CH₂Cl₂, 0 °C, 97%; (d) Pd/C, H₂, EtOAc, rt, 97%; (e) semipreparative SFC, (S)-14, 25%, >99% ee, (R)-14, 20%, 96% ee; (f) 15, pTsOH·H₂O, CH₂Cl₂, rt, 64–71%; (g) (MeO)₂SO₂, K₂CO₃, acetone, rt, 78–82%; (h) NaN₃, DMF, 50 °C, 88–96%; (i) N₂H₄·H₂O, (E)/(Z)-crotyl alcohol, EtOH, 75 °C, 76–94%; (j) for conjugation conditions and details, see Supporting Information; (k) H₂O, microwave irradiation, 150 °C, 99%; (l) 4-nitrobenzoyl chloride, NEt₃, DMAP, CH₂Cl₂, rt, 85%.

We then investigated access to epimers (S)-1 and (R)-1 separately. Following introduction of verbenol fragment 15, screening of conditions to separate the resulting epimers ((S)-16/(R)-16) by silica gel chromatography, HPLC, and supercritical fluid chromatography (SFC) proved unsuccessful. Gratifyingly, we found that enantiomers (S)-14/(R)-14 could be separated by semipreparative SFC using a chiral stationary phase to yield (S)-14 and (R)-14 in >99% ee and 96% ee, respectively. Resorcinols (S)-14 and (R)-14 were then functionalized to yield enantio- and diastereomerically pure (S)-1–(S)-4 and (R)-1–(R)-9. To assign the absolute configuration at the C(2′) stereocenter, (S)-14 was converted to a p-nitrobenzoate (S)-17, whose structure was elucidated by X-ray crystallography.

Pharmacological Profiling

Saturation Binding Assays

We assessed whether the phenyl substitution in (S)-1, (R)-1, and their derivatives (S)-2–(S)-4 and (R)-2–(R)-9 impedes interaction with CB₂R. To this end, time-resolved Förster resonance energy transfer (TR-FRET) binding assay was employed to determine the affinities of the new probes at room temperature.²⁸ HEK293 membrane preparations of SNAP-Lumi4-Tb labeled hCB₂R were incubated with a fluorescent probe in the presence or absence of a validated inverse agonist, SR-144,528,³⁶ to determine its binding parameters. Gratifyingly, the epimeric mixture (S)-2/(R)-2 demonstrated good affinity for CB₂R (K_d = 67.9 nM), suggesting that the C(2′) functionalization was well tolerated and validated the in silico guided design.

Encouraged by the promising result, we studied the impact of configuration at the C(2′) stereocenter on the pharmacological properties by examining each epimer individually. A 12-fold greater binding affinity for CB₂R was shown by (R)-1 (K_d = 39.1 nM) in comparison to (S)-1 (K_d = 476 nM). Functionalization of (S)-1 and (R)-1 with 4-pentynoic acid was well tolerated, and the resulting compounds, (S)-3 and (R)-3, retained the stereoisomeric preference with K_d = 2.10 and 0.42 nM, respectively. Conjugation of (S)-1 and (R)-1 with DY-480XL and N-NBD yielded probes (S)-2 and (R)-2 and (S)-4 and (R)-4, respectively. Fluoroprobes (S)-2 and (S)-4 displayed inferior CB₂R affinity (K_d = 162 and 158 nM, respectively) compared to the excellent binding potencies of (R)-2 and (R)-4 (K_d = 10.2 and 12.3 nM, respectively). Furthermore, strong agreement was observed between CB₂R K_d and K_i values obtained by independent TR-FRET and radioligand binding assays for (R)-2 (K_d = 10.2 nM and K_i = 8.26 nM) and (R)-3 (K_d = 0.42 nM and K_i = 0.66 nM). Collectively, the results further validate the TR-FRET assay and imply that the orthosteric binding pocket of CB₂R shows preference for the R-epimer of the parent compound and its derivatives.

We then set out to investigate whether the excellent CB₂R affinities of (R)-1–(R)-4 are impacted by linker and fluorophore substitution. To this end, probes (R)-5–(R)-9 were prepared that feature a variety of linker lengths and fluorophores, spanning a wide range of size, lipophilicity, and membrane permeability. When tested by TR-FRET at 37 °C, fluorescein, tetramethylrhodamine (TAMRA), and Alexa488 bearing probes (R)-5, (R)-6, and (R)-7 all emerged as high affinity binders for CB₂R with excellent K_d values of 30.3, 2.78, and 24.9 nM, respectively (Table 1). BODIPY 576/589 conjugate (R)-8 showed good binding potency with CB₂R, K_d = 44.7 nM. Particularly remarkable was the retention of high affinity displayed by probe (R)-9 (K_d = 25.9 nM) functionalized with Alexa647. These results illustrate substantial improvement over previous work with agonists where functionalization with the highly polar Alexa488 and sterically demanding Alexa647 led to 64-fold and 611-fold drops in affinity, respectively.²⁸ Importantly, fluoroprobes (R)-5, (R)-6, (R)-7, and (R)-9 emit robust fluorescence signals with exquisite specific binding windows when tested at the physiologically relevant temperature, 37 °C (see Figure 3 and Figure S1).

Table 1. TR-FRET-Based Profiling of Binding Affinity^a.

		Kd [nM]
probe	dye	CB2R	CB1R	Kd ratio (CB1R/CB2R)
(R)-2	DY-480XL	18.9	1740	92
(R)-5	fluorescein	30.3	1280	42
(R)-6	TAMRA	2.78	396	142
(R)-7	Alexa488	24.9	3300	133
(R)-9	Alexa647	25.9	7050	272

^aSaturation binding data (K_d) were determined in a TR-FRET assay at 37 °C with membrane preparations from either hCB₂R-HEK293 or hCB₁R-HEK293 cells. Data shown as a mean, N = 3.

Figure 3.

TR-FRET-based saturation binding profile of (R)-7 (Alexa488) at CB₂R determined at 37 °C. Nonspecific binding was determined in the presence of SR-144,528 (10 μM). Data shown as a mean ± SEM, N = 3.

The binding selectivity of fluorescent probes was tested against the closely related CB₁R in a saturation binding assay at 37 °C using membrane preparations derived from HEK293 cells expressing hCB₁R (Table 1). Fluoroprobes (R)-2, (R)-5, (R)-6, and (R)-7 displayed 42–142-fold selectivity for CB₂R over CB₁R. Notably, (R)-9 demonstrated an exceptional 272-fold preference for CB₂R over CB₁R. Collectively, the excellent affinity and selectivity of a range of physicochemically distinct substituents and fluorophores highlight the versatility of novel platform ligand (R)-1.

Kinetic Binding TR-FRET Assay

We were intrigued by the performances of our probes in the saturation binding assay and leveraged TR-FRET to study ligand binding kinetics at a physiologically relevant temperature, 37 °C (Table 2). The results suggest that all compounds, except (S)-1, possess dramatically slower receptor dissociation rates, k_off, in comparison to the control inverse agonist, SR-144,528. Therefore, high CB₂R affinity of the probes once bound stems from slow receptor dissociation rates. The probes are thus endowed with long receptor residence times, τ, an attribute that has been argued particularly important for GPCRs⁶⁰ as a better suited determinant, compared to K_d, of ligand–protein interactions in living systems.^61,62 Importantly, excellent agreement was found between K_d values obtained in saturation and kinetic binding experiments.

Table 2. TR-FRET-Based Kinetic Profiling at CB₂R^a.

probe	kon [106 M–1 min–1]	koff [10–2 min–1]	τ [min]	kinetic Kd [nM]
(S)-1	9.39	115	0.87	122
(R)-1	12.0	13.9	7.19	11.6
(S)-3	44.3	4.96	20.2	1.12
(R)-3	88.7	2.29	43.7	0.26
(R)-6	3.60	1.08	92.6	3.00
(R)-7	1.49	2.37	42.2	15.9
(R)-9	1.75	1.71	58.5	9.77
SR-144,528	240	122	0.82	5.08

^aKinetic K_d data were measured at 37 °C in a TR-FRET assay using hCB₂R-HEK293 membrane preparations. Data shown as a mean, N = 3.

Functional Profiling: cAMP, G Protein Recruitment, and β-Arrestin

HU-308 is a potent full agonist at CB₂R in the [³⁵S]-GTPγS assay, triggers inhibition of cAMP production, promotes recruitment of β-arrestin, stimulates ERK_1/2 phosphorylation, and facilitates release of Ca²⁺ from intracellular stores.^36,41,63 Compounding evidence indicates that distinct CB₂R agonists favor discrete receptor conformations, leading to preferential activation of one specific signaling pathway over another, a phenomenon known as biased agonism.^36,64−66 Accordingly, we have dedicated substantial efforts to comprehensively profile the pharmacological responses elicited by the new probes across known CB₂R signaling pathways.

One of the canonical signaling pathways of CB₂R involves association with Gα_i/o proteins, which elicit reversible inhibition of adenylyl cyclase resulting in a decrease of cellular cAMP levels and suppression of protein kinase A activity.⁶⁷ Consequently, we have investigated the change in cAMP levels upon probe addition using homogeneous time-resolved fluorescence (HTRF) cAMP assay (Table 3).

Table 3. Functional Characterization in a CB₂R cAMP Assay^a.

probe	pEC50	Emax [%]
(S)-1	5.57	–44
(R)-1	6.95	–44
(S)-2	6.47	–37
(R)-2	7.15	–31
(S)-3	7.39	–49
(R)-3	7.48	–55
(S)-4b	4.98	+44
(R)-4	5.91	–40
ago-3	8.47	+112

^aPotency (pEC₅₀) and E_max data were obtained in a cAMP HTRF assay using hCB₂R-CHO cells. Data were normalized to agonist CP-55,940 response (100%) and basal level (0%), unless noted otherwise.

^bData were normalized to the response of inverse agonist AM10257 (0%) and basal level (100%). Data shown as a mean, N = 3.

All compounds behaved as inverse agonists, with efficacy (E_max) ranging between −31 and −55%. Both epimers of the parent amine ligand inhibited cAMP production with the (R)-1 stereoisomer demonstrating greater potency (pEC₅₀ = 6.95) than (S)-1 (pEC₅₀ = 5.57). DY-480XL and alkyne functionalized probes, (R)-2 (pEC₅₀ = 7.15) and (R)-3 (pEC₅₀ = 7.48), were favored over (S)-2 (pEC₅₀ = 6.47) and (S)-3 (pEC₅₀ = 7.39) with respect to potency, albeit to a lesser degree. Interestingly, the N-NBD probes (S)-4 and (R)-4 behaved as inverse agonists only at high concentration (pEC₅₀ = 4.98 and 5.91, respectively). As a control, we prepared HU-308-derived agonist probe ago-3 from ago-1 that features the same scaffold as (S)-3 and (R)-3 except that it lacks the C(2′) phenyl substituent (eq 1). In the cAMP assay ago-3 displayed potent receptor activation (pEC₅₀ = 8.47, E_max = 112%). These results provide direct experimental evidence as to the critical role of the phenyl substituent in facilitating the switch in ligand functionality from agonist to inverse agonist.

1

To complement the functional response elicited in the cAMP assay, we tested whether the probes trigger association of Gα_i protein with CB₂R using our recently reported bioluminescence resonance energy transfer (BRET) Gi-CASE assay.⁶⁸ Membrane preparations harvested from hCB₂R-HEK293 T-REx cells that genetically incorporate fluorescent NanoLuciferase donor and Venus acceptor proteins to the Gα and Gγ subunits, respectively, were incubated with a probe, and the change in BRET signal was detected. Agonist binding triggers CB₂R activation and dissociation of the Gα and Gβγ subunits resulting in BRET signal reduction. Conversely, inverse agonists elicit increase in BRET intensity by stabilization of inactive CB₂R conformation and G protein accumulation beyond the basal level. Compounds (S)-1, (R)-1, (S)-3, and (R)-3 were selected as representatives to circumvent interference among fluorophores in the BRET assay as previously reported.³⁴ The results indicate that all tested compounds behave as potent inverse agonists with respect to G protein recruitment at CB₂R (Figure 4 and Table 4).

Figure 4.

BRET-based Gi-CASE membrane assay to characterize G protein recruitment at CB₂R. Efficacy, E_max, of the compounds is shown as a mean ± SEM, N = 3–4.

Table 4. Functional Characterization of G Protein Recruitment at CB₂R in a BRET Gi-CASE Assay^a.

probe	pEC50	Emax [%]
(S)-1	6.72	–21
(R)-1	6.80	–28
(S)-3	7.51	–30
(R)-3	7.22	–25
SR-144,528	8.23	–45

^aPotency (pEC₅₀) and E_max data were obtained in a Gi-CASE BRET-based assay using membrane preparations from hCB₂R-HEK293 T-REx cells. Data were normalized to agonist HU-210 response (100%) and basal level (0%). Data are shown as mean, N = 3–4.

Alkyne functionalized probes (S)-3 and (R)-3 (pEC₅₀ = 7.51 and 7.22, respectively) have shown superior potency in comparison to free amines (S)-1 and (R)-1 (pEC₅₀ = 6.72 and 6.80, respectively). Control agonist HU-210 and inverse agonist SR-144,528 demonstrated potency consistent with previously reported [³⁵S]-GTPγS binding assay values, further validating the experimental results (pEC₅₀ = 8.83 and 8.23, respectively).^36,69 With respect to efficacy (E_max), probes (S)-1, (R)-1, (S)-3, and (R)-3 induced functional responses between −21 and −30%. Remarkably, comparison of the effects elicited by (S)-1 in the cAMP and Gi-CASE assays (pEC₅₀ = 5.57 and 6.72, respectively) suggests 14-fold increased potency of G protein recruitment over adenylyl cyclase inhibition, a striking bias within a CB₂R–Gα_i-mediated pathway.

Among the best studied G protein independent signaling pathways of CB₂R is the β-arrestin cascade. β-Arrestins bind activated CB₂R following receptor phosphorylation, block further G protein mediated signaling, and destine the receptor for internalization.³⁰ Representative compounds were profiled for β-arrestin recruitment in a BRET assay where an increase of the BRET ratio corresponds to recruitment of β-arrestin. Baseline BRET signal was retained by (S)-1, (R)-1, (S)-3, and (R)-3 (Figure 5). In contrast, control agonists HU-308 and HU-210 showed expected recruitment of β-arrestin to CB₂R as indicated by increases in BRET intensity (pEC₅₀ = 8.37 and 10.0, respectively). These results imply that (S)-1, (R)-1, (S)-3, and (R)-3 do not activate CB₂R toward β-arrestin recruitment.

Figure 5.

BRET-based assay to characterize β-arrestin recruitment at CB₂R. L = ligand. Data are a representative of N = 3.

Phosphorylation of ERK

Activation of CB₂R is associated with downstream stimulation of mitogen-activated protein kinases, such as ERK_1/2, mediated via either Gβγ or β-arrestins.^70,71 We have tested representative high-affinity fluorescent probe (R)-2 for CB₂R-mediated phosphorylation of endogenous ERK_1/2 in a CB₂R inducible breast cancer HCC1954 cell line using the AlphaScreen SureFire phospho-ERK assay (Figure 6). Expression of CB₂R was optionally induced with doxycycline (DOX), and after 24 h the cells were incubated with a vehicle (0.1% DMSO), CB₂R selective agonist JWH133⁷² (1 μM), or (R)-2 (1 μM) for 30 min. Following cell lysis, lysates were incubated with a mixture containing donor and acceptor beads for 2 h at room temperature and the luminescence emission signal was measured.

Figure 6.

Live cell AlphaScreen SureFire phospho-ERK assay with CB₂R inducible breast cancer HCC1954 cell line. Cells were optionally induced with doxycycline (DOX) for 24 h to stimulate expression of CB₂R followed by incubation with a vehicle (0.1% DMSO), agonist JWH133⁷² (1 μM), or (R)-2 (1 μM) for 30 min. Statistical significance was examined by one-way ANOVA followed by Tukey’s post hoc test. ns = nonsignificant, ∗∗, p < 0.01; ∗∗∗, p < 0.001. Data are an average of three independent biological replicates.

In the absence of CB₂R expression inducer (DOX), the phosphorylation levels of ERK_1/2 remained the same for cells treated with a vehicle, agonist JWH133, and (R)-2. In cells induced to express CB₂R with DOX (1 μg/mL), JWH133 effectively stimulated ERK_1/2 phosphorylation mediated by CB₂R activation, in agreement with previously reported findings.⁷³ Addition of (R)-2 had no effect on the level of phosphorylated ERK_1/2, which remained the same as for a vehicle. Data of the phospho-ERK cellular assay imply that (R)-2 does not induce phosphorylation of ERK_1/2 by either CB₂R-mediated Gβγ or β-arrestin signaling (or by non-CB₂R-mediated pathways).

Ca²⁺ Signaling

Upon activation of CB₂R, Ca²⁺ is often released from intracellular reservoirs.⁷⁴⁻⁷⁶ Our previous work reported that HU-308 and its photoswitchable derivative, azo-HU-308, increase intracellular Ca²⁺ in the mouse AtT-20 cell line.⁴¹ Naturally, we were intrigued to investigate the response elicited by our probes. The epimeric mixture (S)-3/(R)-3 was chosen to avoid fluorophore interference with the Fluo-4AM Ca²⁺ dye and test both diastereomers simultaneously.

Live AtT-20 cells overexpressing rat CB₂R [AtT-20(rCB₂R)] were treated with Fluo-4AM Ca²⁺ dye and imaged by confocal microscopy (Figure 7). Addition of (S)-3/(R)-3 (20 μM) did not elicit increase in Fluo-4AM fluorescence, whereas subsequent addition of agonist HU-308 (20 μM) triggered a robust fluorescence spike. Ionomycin was added at the end of the experiment to saturate Ca²⁺ levels. This result indicates that neither (S)-3 nor (R)-3 induces Ca²⁺ release via CB₂R activation and suggests that the probes can be displaced by HU-308.

Figure 7.

Live cell fluorescent Ca²⁺ imaging in rat CB₂R overexpressing AtT-20 cells [AtT-20(rCB₂R)] loaded with Fluo-4AM (2 μM). After initial equilibration, (S)-3/(R)-3 (20 μM) was added, followed by HU-308 (20 μM) and ionomycin (10 μM). Shown are the average responses of 200 cells (A, left), individual traces of five representative cells (A, right), and representative fluorescence images from different time points (B). Averaged data plotted as mean ± SEM, T = 4.

Fluorescence Confocal Microscopy in Live Cells

Having validated that the probes do not trigger CB₂R signaling at multiple downstream pathways, we employed (R)-7 and (R)-9 to visualize CB₂R by confocal fluorescence microscopy. Probes (R)-7 and (R)-9 were selected due to their bright, photostable, and extensively applied fluorophores Alexa488 and Alexa647. Additionally, the green- and red-shifted fluorescence spectra of (R)-7 and (R)-9 provide flexibility and potential for synergy with additional fluorescent proteins and small molecule dyes for multiplexed imaging studies.

First, AtT-20 cells stably expressing N-terminal SNAP-tagged human CB₂R [AtT-20(SNAP-hCB₂R)] were coincubated with (R)-7, Janelia Fluor SNAP-549i (JF549i), and Hoechst33342 to label CB₂R, SNAP-tags, and nuclei, respectively. Confocal microscopy revealed bright fluorescence of Alexa488 and JF549i delineating the plasma membranes of AtT-20 cells (Figure 8A). Analysis of the corresponding intensity plot showed virtually identical colocalization overlap between the Alexa488 ((R)-7) and JF549i (SNAP-hCB₂R) signals (Figure 8A).

Figure 8.

Confocal imaging of (R)-7 in live cell lines. (A) AtT-20(SNAP-hCB₂R) cells were labeled for 15 min with (R)-7 (Alexa488, 625 nM, green), SNAP-JF549i (JF549i, 500 nM, red), and Hoechst33342 (Hoechst, 1 μM, blue) to visualize CB₂R, SNAP-tags, and nuclei, respectively. Fluorescence intensity profiles across the white line for Alexa488, JF549i, and Hoechst33342 are shown on the right. (B) AtT-20(rCB₂R) cells (left) and AtT-20(WT) cells (right) were treated with (R)-7 (625 nM, green) and Hoechst33342 (1 μM, blue) for 15 min and imaged by confocal microscopy. (C) Live BV-2 microglial cells that endogenously express CB₂R were incubated with (R)-7 (2.5 μM, green) and Hoechst33342 (1 μM, blue) for 15 min and imaged by confocal microscopy.

Since many cannabinoid ligands tend to accumulate in plasma membranes due to their lipophilic nature, specificity of (R)-7 and (R)-9 for CB₂R was evaluated using AtT-20(rCB₂R) and AtT-20 wild-type (WT) cells, which do not express CB₂R.⁷⁷ AtT-20(rCB₂R) and AtT-20(WT) cells were incubated with (R)-7 and Hoechst33342 and imaged by confocal microscopy. A robust Alexa488 fluorescence signal was detected at the plasma membrane of AtT-20(rCB₂R) cells (Figure 8B, left). In stark contrast, AtT-20(WT) cells showed only minimal background fluorescence with no signal stemming from the cellular membrane (Figure 8B, right). These results confirm that (R)-7 specifically labels CB₂R at the plasma membrane.

Encouraged by the promising results with (R)-7 in cells overexpressing CB₂R, we proceeded to investigate the probes’ ability to detect CB₂R at endogenous expression levels. To this end, the murine derived BV-2 microglial cell line was selected due to its extensive use as a high fidelity, primary microglia culture model⁷⁸ that was applied in the study of neurodegeneration and neuroinflammation.⁷⁹⁻⁸¹ Importantly, BV-2 cells endogenously express CB₂R.^82,83 Following incubation of BV-2 cells with (R)-7 and Hoechst33342, an intense Alexa488 signal was observed at the plasma membrane across BV-2 cells (Figure 8C). These results confirm that (R)-7 can visualize CB₂R at endogenous expression levels. Importantly, when (R)-9 (Alexa647) was subjected to analogous experiments, it demonstrated equal specificity for CB₂R in AtT-20 cells (see Figure S2A,B) combined with strong signal intensity in the BV-2 microglial cell line (see Figure S2C). Finally, these data imply that the performances of (R)-7 and (R)-9 remain uncompromised by interspecies differences and the probes can be employed to investigate both human and murine orthologs of CB₂R.

Molecular Dynamics Simulations Unravel Pharmacophore Determinants of Receptor Activation

Molecular dynamics (MD) studies were performed in a membrane environment with inverse agonists (S)-3 and (R)-3 and their agonist counterpart ago-3 to contrast their interactions with CB₂R at a molecular level and elucidate their orthogonal functional profiles. To this end, the X-ray structure of CB₂R in an inactive state in complex with AM10257 (PDB 5ZTY) was selected as a starting point for 1 μs MD simulations to assess ligand stability and identify rearrangements within the binding site.

All three ligands adopt an L-shape conformation with the pendent alkyl chain hosted in a cleft formed by Phe183^ECL2, Tyr190^5.39, Trp194^5.43, and Thr114^3.33 (see Figure 9 and Figures S3 and S4). The resorcinol engages in π–π interactions with Phe183^ECL2, while the pinene core is surrounded by aromatic residues (Phe183^ECL2, Phe91^2.61, and Phe94^2.64). An rmsd plot of ago-3 following a best fit of protein backbone shows an initial rearrangement followed by a periodic “breathinglike” motion of the resorcinol and alkyl chain that oscillates between bent and flat conformations (see Figures S3 and S5). Furthermore, in accord with our earlier work,⁴² in the CB₂R–ago-3 complex the amide group of ago-3 forms a stable hydrogen bond with the carbonyl of Ser90^2.60.

Figure 9.

Representative frames from molecular dynamics (MD) simulations of CB₂R (PDB 5ZTY) in complex with (S)-3 or (R)-3. Superimposition at level of protein backbone of (A) CB₂R X-ray structure with AM10257 (light gray) and (R)-3 MD complex (ligand in gold and protein in light yellow) and of (B) the two inverse agonist complexes (S)-3 (ligand in blue and protein light blue) and (R)-3 (ligand in gold and protein in light yellow).

In agreement with the in silico docking (Figure 1), the MD simulations suggest the C(2′) phenyl rings of (S)-3 and (R)-3 mirror that of AM10257 and engage in π–π contacts with Phe117^3.36 and Trp258^6.48 (see Figure 9A and Figure S6). In contrast to ago-3, no initial rearrangement in the binding poses of (S)-3 and (R)-3 was observed in their rmsd plots (see Figure S7). Ligands (S)-3 and (R)-3 share similar binding modes of the pinene-resorcinol core; however, significant differences are observed in the orientation adopted by gem-dimethyl groups and the C(2′) phenyl rings (see Figure 9B and Figure S8). In particular, the gem-dimethyl group of (S)-3 is rotated clockwise compared to that of (R)-3 and the C(2′) phenyl of (S)-3 is rotated toward Leu262^6.44, inducing a minor displacement of helix H6, while that of (R)-3 protrudes deeper toward Trp258^6.48. In both cases, the conformation of the Trp258^6.48 toggle switch is restricted and CB₂R is thus stabilized in its inactive state, in stark contrast to the interactions observed with the agonist complex.

Superimposition of CB₂R protein backbone employing the X-ray structure with AM10257 and the representative MD frames in complex with either (S)-3 (see Figure S6) or (R)-3 (Figure 9A) imply that (R)-3 more closely resembles the binding mode of AM10257 in the crystallized complex. In particular, the secondary binding pocket featuring Trp258^6.48 and the surrounding residues Phe117^3.36 and Leu262^6.44 are in excellent agreement between AM10257 and (R)-3.

Finally, the difference in free energy of binding (ΔΔG) was determined for (R)-3 and (S)-3 using molecular mechanics/Poisson–Boltzmann (generalized Born) surface area (MM/PB(GB)SA) calculations (see Table S1). The data imply that binding of (R)-3 is more stable by −0.76 kcal mol^–1 (MM/GBSA) and −0.40 kcal mol^–1 (MM/PBSA) in comparison to (S)-3. Notably, the calculated ΔΔG values are in agreement with the experimental difference of ΔΔG = −0.9 kcal mol^–1 for epimers (S)-3 and (R)-3.

Conclusion

This study describes the in silico guided, structure-based switch of functionality from agonist to inverse agonist of HU-308, a ligand extensively applied to unravel CB₂R pharmacology and currently investigated in clinical trials. The novel inverse agonist platform ligands (S)-1 and (R)-1 demonstrated high binding affinity for CB₂R and selectivity against CB₁R that was retained upon functionalization with a range of chemically distinct substituents and fluorophores. The functional response exerted on CB₂R by (S)-1, (R)-1, and their derivatives was evaluated by HTRF and BRET and implied an inverse agonist profile in cAMP as well as G protein recruitment assays, and no induction of β-arrestin–receptor association. Live cell experiments with (R)-2 and (S)-3/(R)-3 demonstrated that the probes do not activate CB₂R toward ERK_1/2 phosphorylation and Ca²⁺ signaling pathways, respectively. Fluorescence microscopy experiments with (R)-7 and (R)-9 in AtT-20 cells expressing human and rat CB₂R isoforms demonstrated excellent target specificity and species translatability. Treatment of the BV-2 microglial cell line with (R)-7 and (R)-9 allowed imaging of endogenous CB₂R in live cells. Finally, MD simulations with (S)-3, (R)-3, and ago-3 corroborate the critical role of the C(2′) phenyl substituent in conferring the functional profile by modulating the CB₂R toggle switch Trp258^6.48 of the CWxP motif.

More broadly, this work discloses the first ligand platform for CB₂R that retains its inverse agonist functional profile, affinity, and selectivity independent of its conjugation to a range of diverse functional groups. The probes introduce a long-sought-after complementarity to an agonist-dominated toolkit to study, elucidate and unlock the full therapeutic potential of CB₂R. Moreover, the platform ligands promise broad application and synergy with previously published work which awaited discovery of an inverse agonist.²⁶ The exponential rise in resolved structures of class A GPCRs, many of which are available in inactive, intermediate, and fully active states, has enabled unprecedented insight into the mechanism of receptor activation.^50,84,85 Thus, the strategy and experimental framework disclosed herein may aid in the structure-based design of agonists, antagonists, and inverse agonists for GPCRs beyond CB₂R.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acscentsci.3c01461.

• Experimental procedures and characterization data for all compounds (PDF)

The authors declare the following competing financial interest(s): M.K., R.C.S., B.K., W.G., U.G., and E.M.C. have filed a patent on CB2R selective modulators and fluorescent probes.