G protein‐coupled receptors (GPCRs) regulate a vast amount of cellular processes, ^[1] thus, they form one of the most important pharmaceutical drug‐target class (475 drugs in the market that represent ∼34 % of all drugs approved by the US Food and Drug Administration). ^[2] However, these drugs target only 108 unique GPCRs that are <15 % of the ∼800 genes (or ∼30 % of the ∼360 non‐olfactory GPCRs). ^[2] One of the reasons is that the orthosteric binding site for a particular endogenous ligand is often highly conserved across a GPCR subfamily, thus making it difficult to achieve high selectivity for specific receptor subtypes. ^[3] Novel approaches to overcome this problem include the discovery of bitopic ligands that bind the orthosteric site as well as a less conserved site within the same receptor unit.[ ⁴ , ⁵ , ⁶ ] This type of complementary cavity is often located at the entrance of the orthosteric binding site, as identified in ligand binding pathway simulations, which have been named extracellular vestibule ^[7] or entrance, ^[8] or secondary ^[9] or metastable ^[10] binding site, or exosite. ^[11] In this work, we will name this cavity as receptor vestibule or exosite. Bitopic ligands that target the orthosteric site and the receptor vestibule improve selectivity,[ ¹¹ , ¹² ] off‐rates and signaling bias,[ ⁴ , ¹³ , ¹⁴ ] maintaining bioavailability and brain penetration properties in mice. ^[15] Other type of comparable ligands are designed to simultaneously bind two orthosteric sites of a (homo/hetero) GPCR dimer. ^[16] These type of ligands have been recently reviewed. ^[17]

While most GPCRs recognize polar ligands, GPCRs for lipid mediators are activated by hormone‐like signaling molecules derived from lipid species, which possess long hydrophobic moieties. This subfamily is mostly composed of the sphingosine‐1‐phosphate (S1P), lysophosphatidic acid (LPA) and cannabinoid (CB₁R and CB₂R) receptors. ^[18] In the crystal structures of these receptors ^[19] the extracellular N‐terminus and extracellular loop 2 folds over the ligand binding pocket blocking the access to the orthosteric binding cavity from the extracellular environment. These structures together with binding pathway simulations suggest that binding of lipid‐like ligands to a lipid GPCR occurs through a narrow channel between transmembrane helices (TMs) 1 and 7 that connects the orthosteric binding site to the lipid bilayer. ^[20] Notably, the access to the ligand pocket of the MT1 melatonin receptor that binds polar ligands (serotonin‐derived compounds) is also via the lipid bilayer. ^[21] Thus, the design of bitopic ligands for lipid GPCRs is more challenging than for other GPCRs that fully expose the binding site to the extracellular environment, due to the narrow channel linking the binding site and the lipid bilayer.

In the present communication, we have designed bitopic ligands for the cannabinoid CB₂R. We have selected CB₂R, instead of CB₁R, due to its lack of adverse psychotropic effects along with its wide therapeutic application in pathologies such as cancer, neuroinflammation and pain. ^[22] Bivalent ligands have already been reported for CB₁R[ ²³ , ²⁴ , ²⁵ ] and CB₂R. ^[26] These ligands were reported before the release of crystal structures, thus, their binding characteristics remain unclear.[ ²⁷ , ²⁸ ] Here, we have used the recently released structure of CB₂R in its active ^[29] G_i‐bound conformation to identify the binding mode of the designed ligands.

The design of bitopic ligands requires the selection of a moiety able to bind the orthosteric site (pharmacophore). In this regard, we have selected chromenopyrazole derivatives A and B (Figure 1), which have been previously identified as CB₂R orthosteric agonists. ^[30] Then, a second pharmacophore unit needs to be developed for the vestibule or exosite. This is challenging because this additional cavity has not been properly characterized yet for most GPCRs. We have taken advantage of the simulated binding process of a lipid inhibitor to the S1P1 receptor. ^[20] The process consists of the diffusion of the ligand through the bilayer leaflet to contact the vestibule at the top of TM 7 (the rate‐limiting step), subsequently moving from this lipid‐facing vestibule to the orthosteric binding cavity through the channel between TMs 1 and 7. We propose that lipid GPCRs are capable to recognize orthosteric ligands at the vestibule of the receptor. Thus, we have also selected the chromenopyrazole moiety as the second pharmacophore so that the designed bitopic ligands are symmetrical (it contains two copies of the same pharmacophore). In addition, an appropriate length spacer to cover the distance between both pharmacophores is required. Importantly, this approach has been recently supported in the model of the bitopic ligand CTL01‐05‐B‐A05, a symmetrical agomelatine molecule linked by an ethoxyethane spacer, which binds both the orthosteric binding site and the exosite of the MT1 melatonin receptor. ^[21]

Chromenopyrazole A, in its most stable conformation (Figure S1), was docked into the orthosteric and vestibule sites of CB₂R, and its stability was assessed by molecular dynamic (MD) simulations (Figure S2). Results showed that pharmacophore units remain highly stable at the orthosteric site and moderately stable at the exosite. Visual inspection of the models shows that the −CH₃ group of the methoxy moieties of both pharmacophores are suitable attachment points to link the spacer moiety. Linker lengths from six to sixteen methylene units were chosen for the bivalent molecules.

Bivalent chromenopyrazoles and their monovalent analogues were synthesized, starting from chromenopyrazoles 4 and 5 (Scheme 1). ^[31] Preparation of 9‐alkoxychromenopyrazoles (6–17) was achieved in high yields by deprotonation of the hydroxyl group with sodium hydride, followed by rapid addition of an excess of the appropriate 1‐bromoalkane. Bivalent compounds 18–29 were achieved by alkylation of the corresponding chromenopyrazoles with 0.5 equivalents of the desired dibromoalkanes. Different bases were tested, and finally cesium carbonate was selected and used under inert atmosphere. Thus, the desired bivalent compounds (18–29) were achieved in low to moderate yields.

In vitro binding affinities of the alleged bitopic ligands 18–29 (Table 1) and the corresponding monovalent counterparts 6–17 (Table S1) were obtained from [³H]CP‐55,940 competition‐binding assays using membrane fractions of the human CB₁R and CB₂R, respectively, expressed in HEK‐293T cells. None of the monovalent chromenopyrazoles exhibits affinity towards any of the cannabinoid receptors. Addition of the second pharmacophore makes bivalent ligands capable of binding CB₂R in a selective manner. Optimal spacer length for N1‐ and N2‐ethyl derivatives is from 10 (n=8) to 14 (n=12) methylene units.

Compounds with CB₂R affinity constants in the low micromolar range (22, 24–27) were selected for functional evaluation by measuring their effect on forskolin‐induced cAMP levels in HEK‐293 cells expressing hCB₂R (Figure 2 A). Dose‐response experiments demonstrated that these compounds are able to inhibit cAMP accumulation as efficiently as CP55,940 but with a slight drop of potency (27: pEC₅₀=7.6 vs. CP55,940: pEC₅₀=8.2) (Table 2), while their monovalent counterparts are inactive (Figure S3). None of the compounds displayed an effect in non‐transfected HEK‐293T cells, confirming that the results are fully mediated by CB₂R (Figure S4).

In order to assess the binding mode of homobivalent chromenopyrazoles 22, 25 and 27 (10, 12 and 14 methylene units), we docked their most favorable conformation (Figure S1) into CB₂R in its active state in such a manner that both pharmacophoric units bind into the orthosteric and vestibule sites (Figure 3 A). Unbiased MD simulations show that these chain lengths can simultaneously bind both sites (Figure 3, S5). The methylene spacer expands toward the lipid‐facing vestibule interacting with the hydrophobic side chains of Phe91^2.61, Ala282^7.36, Met286^7.40, and Val36^1.35. In the vestibule, the chromenopyrazole moiety forms aromatic–aromatic interactions with Phe283^7.37. In the simulations Phe283^7.37 adopts the trans conformation that opens the channel between TMs 1 and 7 and permits the ligand to reach the membrane, in contrast to the gauche ⁺ conformation observed in the crystal structure of CB₂R[ ²⁹ , ³² ] that closes the channel. The heptyl chain is accommodated in a hydrophobic cavity of TMs 1 and 7, facing the membrane, which is formed by Leu39^1.38, Cys40^1.39, Met286^7.40, Leu43^1.39, Ile290^7.44, Met293^7.47 and Leu46^1.45. In the case of the N1‐ethyl derivative the lone pair in the N2 atom forms a hydrogen bond with Gln32^1.31. A detailed description of these interactions is shown in Figure S6.

Figure 3 B shows the mesh surface formed by Val36^1.35 and Ala282^7.36. Clearly, these TMs side chains, which are located midway between the orthosteric site and the receptor vestibule, delimit the channel between TMs 1 and 7. Thus, in order to validate the proposed binding mode of bitopic ligands, we mutated the side chains of Val36^1.35 (Figure 2 B) and Ala282^7.36 (Figure 2 C) to the much larger Met side chain. As expected, these mutations do not influence the function of the orthosteric agonist CP55,490, but clearly impair signaling of bitopic ligands 22 and 27 (Table 2) by occupying the volume of the channel. Notably, these results contrast with other MD simulations, suggesting the entrance of the ligands from the extracellular environment.[ ³³ , ³⁴ ]

In summary, we present herein the discovery of CB₂R homobivalent bitopic ligands. These compounds were designed as symmetrical bivalent compounds using the previously reported chromenopyrazole scaffold as potential pharmacophore for both units linked by a methylene spacer. Binding and functional cAMP studies revealed their ability to selectively activate CB₂R versus CB₁R. The longer Ile^1.35 in CB₁R than Val^1.35 in CB₂R seems responsible for the observed selectivity (Figure S7). MD simulations and site‐directed mutagenesis studies show that these bitopic ligands bind into the orthosteric site and in a vestibule/exosite located at the ligands entry/egress channel that connects the orthosteric site with the lipid bilayer membrane. Whether these bitopic ligands at CB₂R show beneficial therapeutic application needs further investigation such as structure–activity relationship studies to improve potency.