Understanding Functional Residues of the Cannabinoid CB₁ Receptor for Drug Discovery

Abstract

The brain cannabinoid (CB₁) receptor that mediates numerous physiological processes in response to marijuana and other psychoactive compounds is a G protein coupled receptor (GPCR) and shares common structural features with many rhodopsin class GPCRs. For the rational development of therapeutic agents targeting the CB₁ receptor, understanding of the ligand-specific CB₁ receptor interactions responsible for unique G protein signals is crucial. For a more than a decade, a combination of mutagenesis and computational modeling approaches has been successfully employed to study the ligand-specific CB₁ receptor interactions. In this review, after a brief discussion about recent advances in understanding of some structural and functional features of GPCRs commonly applicable to the CB₁ receptor, the CB₁ receptor functional residues reported from mutational studies are divided into three different types, ligand binding (B), receptor stabilization (S) and receptor activation (A) residues, to delineate the nature of the binding pockets of anandamide, CP55940, WIN55212-2 and SR141716A and to describe the molecular events of the ligand-specific CB₁ receptor activation from ligand binding to G protein signaling. Taken these CB₁ receptor functional residues, some of which are unique to the CB₁ receptor, together with the biophysical knowledge accumulated for the GPCR active state, it is possible to propose the early stages of the CB₁ receptor activation process that not only provide some insights into understanding molecular mechanisms of receptor activation but also are applicable for identifying new therapeutic agents by applying the validated structure-based approaches, such as virtual high throughput screening (HTS) and fragment-based approach (FBA).

Introduction

G-protein coupled receptors (GPCRs), composed of seven transmembrane (TM) spanning helices (H1-H7) interconnected by three intracellular loops (I1-I3) and three extracellular loops (E1-E3) [1,2], are known to be among the most important drug targets [3,4] (for reviews, see [5-7]). For the rational development of therapeutic agents targeting a specific GPCR, it is crucial to understand the ligand-receptor interactions that determine the effectiveness of the ligand for modulating receptor activity toward unique G protein signals. If the ligand binding affinity is defined by how strongly a ligand binds to a receptor and the receptor efficacy is defined by how efficiently a receptor activates the coupled G protein, a ligand as a drug candidate should exhibit high binding affinity and more importantly desirable binding efficacy. Thus, it should be at the center of the rational development of GPCR drugs to understand the ligand-initiated receptor conformational change responsible for GPCR signaling [8,9].

Understanding of binding efficacy in GPCRs is challenging due to the conformational complexity of GPCRs [10,11]. Even in the absence of ligands GPCRs exhibit basal activity, suggesting that GPCRs are in motion with inherent conformational flexibility. Conformational equilibrium of a GPCR between the inactive state and the active state can be modified by ligand binding [11]. By definition, agonists activate the receptor and produce signaling activity, inverse agonists stabilize the receptor and inhibit basal activity, and antagonists block the receptor and produce no activity: Binding of an agonist or a partial agonist would lower the energy barrier from the inactive state to the active state and/or stabilize the active state, shifting the equilibrium toward the active state, while binding of an inverse agonist would enhance the energy barrier and/or stabilize the inactive state, shifting the equilibrium toward the inactive state. Not only ligand binding but also coupling to cognate G proteins would modify the equilibrium between the inactive state and the active state [11]. In addition to ligands or G proteins that contribute to receptor stability, altering receptor residues involved in receptor activation (e.g., constitutively active mutations (CAMs) that increase the basal activity of the receptor in the absence of a ligand (for reviews, see [12-14]) is an alternative way to modify the equilibrium toward the active state.

It should be noted that structurally diverse ligand classes of a GPCR can induce the ligand-specific conformational changes in the receptor, determining different receptor states that are capable of activating specific subtypes of cognate G protein, as proposed in the protein ensemble theory describing proteins as a collection of conformational states [15]. In fact, evidence from many biophysical studies supports the existence of multiple, ligand-specific conformational states of GPCRs [16-19]. In addition to ligand-specific receptor conformations, it has been indicated that the agonist bound GPCRs can have multiple, state-specific conformations corresponding to different signaling states [20,21]. For example, it is known that two distinct forms of the photoactivated rhodopsin, MII and MIIa/MIIb have been reported [22-24] and that the agonist-bound β₂ adrenergic receptor (β₂AR) exists in at least two different conformational states [25,26].

To date, the X-ray structures available include rhodopsin [27-37], β₁AR [38], β₂AR [39-41], and the adenosine A_2A receptor (AA_2AR) [42]. In spite of the overall same topology, these X-ray structures show several interesting differences in local regions [43], as illustrated in the comparison of β₂AR with rhodopsin: 1) H1 of rhodopsin is packed closely to the helical bundle because of a kink in the middle of the helix, while H1 of β₂AR is loosely packed to the helical bundle without the equivalent kink; 2) The N-terminal of rhodopsin is ordered with β-sheets stacked on the top of E2 covering the ligand binding pocket, while that of β₂AR is disordered; 3) The binding pocket in rhodopsin with a β-sheet alignment of E2 is totally blocked, while that in β₂AR with an α-helical segment within E2 is only partially blocked for easy solvent or ligand access [40]; 4) The X-ray structure of rhodopsin [46] represents the inactive state, while that of β₂AR-T4L fusion protein [40] represents an active-like state [44,45].

It appears that a receptor conformational change suitable for G protein activation does not require a dramatic change in the whole receptor but a change in the cytoplasmic side. For example, the X-ray structure of β₂AR-T4L fusion protein [40], an active-like state, shows little difference from the X-ray structure of rhodopsin (inactive state) [46]. Similarly, compared with the inactive form of rhodopsin [46], the photoactivated state metarhodopsin II (MII) [32] showed surprisingly little change, except an increased disorder in I3. This result suggests that photoactivation results in I3 flexibility. Evidence not only from the X-ray structures of GPCRs but also from many biophysical studies supports the conformational change in the intracellular side [47-51], including the breakage of the salt bridge between R^3.50 and E(D)^6.30 of the receptor, proposed as the ionic lock [52], that maintains the inactive state of the rhodopsin class GPCRs mainly through H6 stabilization but needs to be cleaved upon receptor activation. Elling et al. [50] proposed a global toggle switch activation model, where upon activation the receptor changes conformation such that the intracellular segments of H6/H7 move outward and the highly conserved TM Pro in the middle of H6 and H7 act as the pivot for the vertical seesaw movements. (In this review, a numbering system similar to Ballesteros-Weinstein system [53] is used for all the amino acids. For example, R135^3.50 represents Arg135 with the highest conservation, indicated by the number 50, in the TM helix 3, indicated by the number 3. Similarly, V173^E2 represents Val 173 of the second extracellular loop.)

All the X-ray structures are the antagonist or inverse agonist bound forms: for rhodopsin, retinal (inverse agonist) [28]; for β₂AR, carazolol (partial inverse agonist) [40]; for β₁AR, cyanopindolol (antagonist) [38]; and for AA_2AR, ZM241385 (antagonist) [43]. In considering the roles of the antagonist in blocking the receptor and the inverse agonist in stabilizing the inactive form of the receptor, it would be natural to have the X-ray structures of antagonist or inverse agonist bound GPCRs compared to those with agonist bound GPCRs, which is short-lived for G protein signaling. It has been argued that the X-ray structures of GPCRs are unsuitable for screening partial or full agonists [54-56]. Thus, if the screening of novel compounds is directed by using a receptor structure that is from its complex with an antagonist or inverse agonist, one may end up with antagonists or inverse agonists rather than with agonists. For example, Kolb et al., [57] used the X-ray structure of β₂AR to dock approximately 1 million commercially available small molecules and identified some compounds with relatively high affinities, most of which showed inverse agonist activity, suggesting that the receptor conformation of the reported X-ray structure of β₂AR is adapted to mostly recognize the inverse agonist.

Thus, it seems apparent that the receptor structure of the agonist-bound form, at least, is necessary for drug design of agonists, though without knowing the protein structure of the high-affinity state GPCRs in complex with a G-protein [58,59] the picture is incomplete to fully understand the receptor in its active state to which an agonist preferentially binds. However, the agonist bound GPCR structures, are extremely difficult to obtain [56]. Theoretically it is possible to obtain the receptor in its active state by starting from the inactive state of the receptor along the reaction path allowed for this conversion. However, the current state of the art MD simulations reported in the microsecond time scale [60,61] are not sufficient to provide the molecular details of the active state of a GPCR that would form in the millisecond scale [44]. In addition, this path, largely unknown, contains multiple binding motifs for diverse ligands and multiple active conformers of the receptor. In this regard, recent studies [62-71] reported possible early events along this path from the inactive state toward the active state.

The Cannabinoid Receptors Belong to Gpcrs

Brain CB₁ cannabinoid receptors [72] are GPCRs and belong to the rhodopsin-like subfamily [2]. The CB₁ receptor is coupled to G-proteins for signal transduction pathways that inhibit adenylyl cyclase activity and regulate ion channels [72-76]. As shown in Fig. (1), the sequence alignment of the CB₁ receptor with some GPCRs whose X-ray structures are available indicates that they share common features, including the seven TM helices and the highly conserved functional motifs, suggesting that the CB₁ receptor has a similar molecular mechanism for receptor activation as for other rhodopsin class GPCRs. Based upon the phylogenetic study by Joost and Methner [77], where 241 human GPCRs were divided into 19 subgroups (A1 through A19), it was shown that the CB₁ receptor belonged to subgroup A13 next to subgroup A17 from which AA_2AR and βARs were branched and that these receptors altogether belonged to a larger cluster different from subgroup A16 to which rhodopsin belonged. Similarly, Fredriksson et al. [78] performed phylogenetic analyses of 342 unique functional non-olfactory human GPCRs to obtain five main families, including glutamate, rhodopsin (α, β, γ, and δ groups), adhesion, frizzled/taste2, and secretin. Further, the α-group of the rhodopsin family was classified into several distinct clusters, including the prostaglandin cluster, the amine receptor cluster, the opsin receptor cluster, the melatonin receptor cluster and the MECA (melanocortin, endothelial differentiation, cannabinoid and adenosine binding) receptor cluster to which cannabinoid receptors belong. Interestingly, both phylogenetic analyses [77,78] revealed that some of the closest receptors to the cannabinoid receptors are the melanocortin (MC) receptors. Both receptors, commonly lacking the conserved disulfide linkage between H3 and E2, common to most rhodopsin class GPCRs [79,80] and important for receptor structure and activity [81], exhibit a high degree of basal activity [82,83], suggesting a role of the flexibility of E2 in basal activity. Interestingly, the MC-4 receptor (MC4R), which exhibits the highest sequence homology to the CB receptors (approximately 80 %) among the MC receptors, is known to be associated with 6 % of early onset obesity [84], indicating their common roles in regulating obesity. In addition to the traditional cannabinoid receptors by which agonist activation is mediated through Gi/o protein [85], a novel putative cannabinoid receptor GPR55 has recently been characterized [86,87]. According to the GPCR classification by Fredriksson et al. [78], GPR55 belonged to the δ-group of the rhodopsin family within the purine receptor cluster, which is different from the α-group to which the cannabinoid receptors belong.

Fig. (1).

Sequence alignment by T-COFFEE (V7.71, mode: expresso) (http://www.tcoffee.org) [211] of CB₁ and CB₂ with the GPCRs whose structures have been determined by the X-ray crystallography, including AA_2AR (PDB code: 3EML) [42], β₁AR (PDB code: 2VT4) [38], β₂AR (PDB code: 2RH1) [40], and rhodopsin (PDB code: 1U19) [46]. The TM helical boundaries for the CB₁ and CB₂ receptors are from the respective homology models [136,182], while the TM helical boundaries for AA_2AR, β₁AR, β₂AR, and rhodopsin are from the respective X-ray structures. Conservancy of the aligned sequence by CLUSTALW (http://www.ebi.ac.uk/Tools/clustalw/) [212] is represented by consensus symbols: “*” for identical residues; “:” for conserved substitutions; and “.” for semi-conserved substitutions. Highly conserved residues in the rhodopsin family of GPCRs reported by Baldwin et al (1997) are in bold. Color codes for TM helices: H1 (in red); H2 (in orange); H3 (in yellow); H4 (in green); H5 (in cyan); H6 (in blue); H7 (in purple); and H8 (in dark green).

Three Types of Cannabinoid Receptor Functional Residues

Mutagenesis data for the cannabinoid receptors have been accumulated for more than a decade. The mutagenesis analysis, often combined with the computational modeling analysis, provides invaluable insights into characterizing the role of the CB₁ receptor residues in ligand binding and receptor activity. However, it is not always straightforward to interpret the effects of the mutated residue due to the conformational complexity inherent to GPCRs [11] as described earlier. This is especially true for the case of the cannabinoid receptors that exhibit a high level of basal activity [82,88-90]. Thus, Picone et al. [91] writes, “Introduction of a point mutation within a receptor binding site has the potential of altering ligand binding properties. However, it could also confer undesirable and frequently unpredictable consequences on the global conformation of the receptor that may affect ligand binding in a manner unrelated to binding site structure.” In the same context, Beukers and Ijzerman [92] describe, “By the technique of site-directed mutagenesis by which point mutations are introduced, the role of specific residues in receptor structure and function can be easily studied. However, careful analysis of mutagenesis data is required because the replacement of an amino acid can result in a local effect as a result of a gain or loss of interaction with neighboring residues or in a global effect such as an alteration of protein folding or protein stability”.

In this review, I divide the CB₁ receptor functional residues into three different types, ligand binding (B), receptor stabilization (S) and receptor activation (A) residues, according to their roles as suggested from the mutational studies. Effects of the mutations of these residues are schematically shown in Fig. (2). It is interesting to note that all these residues can contribute to receptor activation. It should be noted that it is often difficult to define the function of a residue as one specific type over the others due to the fact that their roles are often closely associated with each other.

Fig. (2).

The functional residues of the CB₁ receptor are divided into three different types, receptor stabilization (S) (in red circle), ligand binding (B) (in blue circle) and receptor activation (A) (in yellow circle). The partial overlaps of the circles indicate those residues with more than one function. Among the residues that affect ligand binding, the type B_d residues directly contact with the ligand, while the types B_iS and B_iA residues indirectly alter the ligand binding site geometry (see text). Color code of mutational effects: upon the receptor structure (in magenta); upon ligand binding (in cyan); and signaling (in orange).

The type B_d residues form the binding pocket and directly contact with the ligand (Fig. (2)). Thus, mutations of the type B_d residues alter binding site structure and affect ligand binding by disrupting specific ligand interactions. This type of residues are useful not only for describing the binding pockets distinct to the individual classes of ligands but also for delineating the early stages of receptor activation (see below).

The type S residues stabilize the receptor structure [93] (Fig. (2)). Mutations of this type of residues may cause the failure of receptor expression due to misfolding and abolish the ligand binding affinity and receptor activation. An example of the type S residue is illustrated by W^4.64 (i.e., the Trp residue at the 4.64 position [53]) of the cannabinoid receptors that is known to be critical for ligand binding and signaling [94]. It was reported that the mutation of W255^4.64 of the CB₁ receptor resulted in the failure of receptor expression [95,96], suggesting that W^4.64 is important for receptor structure. Combining these results with the finding that the highly conserved W^4.64 in many GPCRs is known to be important for receptor folding [97], W255^4.64 of the CB₁ receptor is a type S residue important for receptor structure.

If a type S residue affects ligand binding by indirectly modifying the binding pocket topology, it can be further classified into B_iS residues important for the binding of inverse agonists that stabilize the receptor. Thus, mutations of the type B_iS residues result in alteration in ligand binding and impair ligand binding and receptor activation (Fig. (2)). In fact, many residues reported to be important for ligand binding belong to this type of residues (see below).

The type A residues play an important role in receptor activation by being directly involved in receptor activation (Fig. (2)). Thus, mutations of the type A residues directly alter G protein signaling. Examples of the type A residues are seen in CAMs which decrease the inverse agonist binding and increase basal activity [13,98]. Another example is seen from those residues associated with the D(E)RY motif known to be crucial for G-protein activation [99], by the formation of the ionic lock [52], without participating in ligand binding. Thus, Song and his colleagues [100] reported that D^6.30N mutation of the CB receptors maintained the ligand binding affinity but exhibited the greatly reduced signaling activity, suggesting that these residues are the type A residues.

If the type A residues indirectly affect ligand binding through modifying the binding pocket topology, they can be further classified into B_iA (Fig. (2)). An example of the type B_iA residue was illustrated by Song and Feng [101] in their site-directed mutagenesis study showing that Y209^5.58A mutation of the peripheral cannabinoid (CB₂) receptor [102] reduced the binding affinity of 2,3-dihydro-5-methyl-3-[(4-morpholinyl)methyl]pyrrolo[1,2,3-de]-1,4-benzoxazin-6-yl] (1-naphthyl)methanone (WIN55212-2), (6aR,10aR)-9-(Hydroxymethyl)-6,6-dimethyl-3-(2-methyloctan-2-yl)-6a,7,10,10a-tetrahydrobenzo[c]chromen-1-ol (HU210) and N-arachidonoylethanolamine (anandamide) by 5- to 8-fold but abolished signaling activity. Although this study was done on the CB₂ receptor, considering that Y^5.58 is important for breaking the ionic lock [103] and located deep down the core and not involved directly in ligand binding, it is likely that the conserved Y294^5.58 of the CB₁ receptor is a type B_iA residue for cannabinoid agonists.

It was shown that the L341^6.33A/A342^6.34L double mutation of the CB₁ receptor exhibited decreased (1R,3R,4R)-3-[2-hydroxy-4-(1,1-dimethylheptyl)phenyl]-4-(3-hydroxypropyl)cyclo-hexan-1-ol (CP55940) binding by 4-fold an partial constitutive activation (not G_i but G_s) [159] and that D130^3.49A, R131^3.50A and A244^6.34E mutations of the CB₂ receptor abolished agonist-induced G-protein activation with retaining the binding affinity of anandamide, HU210 and WIN55212-2 only by R131^3.50A mutation [104]. Considering all these residues are conserved in both the CB₁ and CB₂ receptors (Fig. (1)), it is likely that R131^3.50 is a type A residue without interfering in ligand binding, while D213^3.49 and A342^6.34 of the CB₁ receptor are B_iA residues for anandamide, CP55940 and WIN55212-2.

It was shown that T210^3.46I and L207^3.43A mutations of the CB1 receptor increased the binding affinity of methanandamide, a synthetic analog of anandamide known to be metabolically stable, by 6-fold and 3-fold, respectively [105,106]. It was shown that these mutations increased CP55940 binding affinity by 3-fold, decreased WIN55212-2 binding by 3-fold and decreased N-(piperidin-1-yl)-5-(4-chlorophenyl)-1-(2,4-dichlorophenyl)-4-methyl-1H-pyrazole-3-carboxamide (SR141716A) binding by about 30-fold, suggesting that these residues are CAMs. With the location of these residues near the DRY motif far away from the binding site, these residues are less likely to directly interact with the ligand but play an important role in regulating the receptor conformation in association with G protein activation. Thus, it is likely that they are B_iA residues for anandamide, CP55940, WIN55212-2 and SR141716A.

CB₁ Receptor Residues Important for Anandamide Binding

It was shown from a recent study combining computational and mutational analyses that anandamide exhibited an approximately 13-fold decrease in binding affinity by the Y275^5.39F mutation of the CB₁ receptor and that the Y275^5.39I mutation abolished ligand binding and receptor signaling [107]. These results suggest that Y275^5.39 is important for anandamide binding, possibly through H-bonding. The authors concluded that the Y275^5.39I mutation altered ligand interactions within the binding site. Thus, it is likely that Y275^5.39 of the CB₁ receptor is a B_d residue for anandamide. However, it is possible that the Y275^5.39I mutation may cause a global effect [92] resulting in a significant modification in the ligand binding pocket. In support, it was reported that Y275^5.39 of the CB₁ receptor and Y191^5.39 of the CB₂ receptor were not direct ligand contact sites [108,109] and that the Y275^5.39S and Y275^5.39A mutations resulted in failure of receptor expression [110].

No mutations of F200^3.36A, W279^5.43A and W356^6.48A, but F189^3.25A moderately (approximately 6-fold) decreased anandamide binding affinity [95]. Illustrated by an anandamide docking model developed from these obtained mutation data, the C5-C6 double bond of the ligand interacted with F189^3.25 via aromatic/π interaction [95]. These results indicate that F189^3.25 is a B_d residue for anandamide.

CB₁ Receptor Residues Important for CP55940 Binding

McAllister et al. [107] showed that CP55940 binding to the CB₁ receptor was retained by the Y^5.39F mutation but abolished by the Y^5.39I mutation, suggesting that the aromaticity of the residue at the 5.39 position is crucial for CP55940 binding. The authors concluded that altered interactions caused by the Y^{5 39}I mutation resulted in alteration in ligand binding. Thus, it is likely that Y275^5.39 of the CB₁ receptor is a B_d residue for CP55940.

It was shown that the Ala mutations of the E1 residues (H181, R182, K183 and D184) and the H3 proximal residues (V188^3.24 and F189^3.25) resulted in reduced binding affinities of CP55940, but not of SR141716A [111]. With decreases in CP55940 binding less than expected by direct ligand contact residues, the results suggest that the effects are not due to direct contact with the ligand but due to loop structural alteration that indirectly affects ligand binding through modifying the binding pocket topology. Thus, these residues on the extacellular side of the receptor are B_iS residues for CP55940. Interestingly, for the binding affinity of CP55940 by the F189^3.25A mutation of the CB₁ receptor, Murphy and Kendall [111] reported an approximately 60-fold decrease, but McAllister et al. [112] reported a 3-fold decrease.

It has been shown from a combined study of affinity labeling, site-directed mutagenesis, and ligand docking studies [91] that a classical cannabinoid derivative having the reactive electrophilic isothiocyanate moiety at the end of the C3 side chain formed a covalent bond to the nucleophilic C355^6.47 and retained the binding affinities to the C355^6.47A and C355^6.47S mutants but had a reduced affinity to the C355^6.47L mutant. No change in WIN55212-2 binding affinity for any of mutant receptors was shown. These results suggest that C355^6.47 of the CB₁ receptor is part of the binding site of the classical and non-classical cannabinoids, distinct from that of WIN55212-2, and is involved in receptor activation as part of the highly conserved functional CWxP motif [112,113]. Thus, C355^6.47 of the CB₁ receptor is a B_d residue for CP55940.

In contrast to the profound effect the S383^7.39A mutation of the CB₁ receptor produced upon CP55940 binding, the binding of WIN55212-2 and SR141716A was unchanged by this mutation [114]. Rhee observed that the S285^7.39A mutation of the CB₂ receptor resulted in dropping HU243 binding approximately 13-fold [94], suggesting that the conserved S^7.39 of the CB receptors are involved in CP55940 binding. Demonstrating by computational modeling studies that the S383^7.39A mutation reduced the H7 kink, Kapur et al. [114] proposed that S383^7.39 induces a bend in the extracellular side of H7 required for CP55940 binding. It is interesting to note that the mutation of the equivalent residue A292^7.39 of rhodopsin resulted in a constitutively active receptor [115]. Taken together, it is likely that S383^7.39 is a B_iA residue for CP55940.

In a very recent Ala scanning mutagenesis study, Kendall and her colleagues [96] comprehensively analyzed the E2 of the CB₁ receptor and demonstrated that the Ala mutations of both the N-terminal and C-terminal residues abolished CP55940 binding and G protein signaling. Interestingly, Ala mutations of P269^E2, H270^E2 and I271^E2 were insensitive for SR141716A binding. As shown from the accompanied homology molecular model, these hydrophobic residues were deeply inserted into the core and formed part of the ligand binding site, suggesting that they are CP55940-specific residues. Together, W255^E2 and N256^E2 are S residues, while F268^E2, P269^E2, H270^E2 and I271^E2 are B_d residues for CP55940.

CB₁ Receptor Residues Important for WIN55212-2 Binding

From a mutagenesis study [116], it was shown that the CB₁ D163^2.50N mutation reduced WIN55212-2 binding by approximately 400-fold, but not for other cannabinoid ligands. It was also shown that the mutated receptor exhibited greatly attenuated inhibition of cyclic AMP production by both CP55940 and WIN55212-2. These results strongly suggest that D163^2.50, a highly conserved residue in many GPCRs, is important in the CB₁ receptor for WIN55212-2 binding and receptor activation. As the authors discussed, the D163^2.50N mutation may cause the receptor to alter the binding pocket uniquely defined for WIN55212-2. Thus, it is likely that D163^2.50, without directly interacting with the ligand, is a B_iA residue for WIN55212-2. However, it should be noted that Mackie and his colleagues, in contrast, reported that D164^2.50 in the rat CB₁ receptor had no effect on ligand binding or inhibition of cyclic AMP production but did alter signaling via G_i by WIN55212-2 [117].

It was shown from the CB₁ chimeric receptor formed by the replacement of H3 with that of the CB₂ receptor [118] that H3 residues were important for WIN55212-2 binding and that the G195^3.31S mutated CB₁ receptor exhibited an enhanced WIN55212-2 binding by about 5-fold. The authors proposed that the enhanced WIN55212-2 binding affinity was attributed to H-bonding between the carboxyl oxygen of the ligand and S^3.31 of the mutated receptor, based upon the report that AAI analogs lacking the carboxyl oxygen displayed a 7- to 10-fold decrease in binding affinity [119]. Thus, it is likely that G195^3.31 of the CB₁ receptor is a B_d residue for WIN55212-2.

It was shown from a study combining molecular modeling and site-directed mutagenesis approaches [108] that the V282^5.46F mutation of the CB₁ receptor enhanced the WIN55212-2 binding affinity by 12-fold; this mutation did not change the binding affinity of cannabinoid compounds and anandamide, suggesting that V282^5.46 is a B_d residue for WIN55212-2 binding [95].

McAllister et al. [107] showed that WIN55212-2 binding was retained by the Y275^5.39F mutation of the CB₁ receptor but abolished by the Y275^5.39 I mutation, suggesting that the aromaticity of the residue at the 5.39 position is crucial for WIN55212-2 binding. Thus, it is likely that Y275^5.39 of the CB₁ receptor is a B_d residue for WIN55212-2.

It was shown that the F200^3.36A and W279^5.43A mutations of the CB₁ receptor reduced the binding of WIN55212-2 by 9-fold and 16-fold, respectively, but did not affect CP55940 binding [95], leading to proposing the importance of an aromatic cluster within the CB₁ receptor aromatic microdomain formed by H3-H4-H5-H6 for WIN55212-2. Similarly, the F200^3.36A mutation of CB₁ showed a moderate decrease in WIN55212-2 binding but no change in CP55940 binding [120]. Thus, these residues are B_d residues for WIN55212-2.

CB₁ Receptor Residues Important for SR141716A Binding

It was shown that the W279^5.43A or W356^6.48A mutation of the CB₁ receptor reduced significantly the binding of SR141716A [95], suggesting the important of these residues for the binding of SR141716A. Thus, these residues are type B_d residues for SR141716A.

It was shown that the mutation of E2 Cys residues C257^E2 and C264^E2 to Ala abolished the binding affinity of SR141716A [121], implying a significant conformational change, possibly due to breaking the intra-loop disulfide bond and a modification of the ligand binding pocket. In the same study, it was also shown that the introduction of a bulky group on C386^7.42 inhibited SR141716A binding [121], suggesting that C386^7.42 directly interacts with SR141716A. Thus, C257 and C264 are B_iS residues, while C386^7.42 is a B_d residue for SR141716A.

A Role of K192^3.28 in Ligand Binding

From site-directed mutagenesis studies of the human CB₁ receptor [122,123], it was shown that the K192^3.28A mutation resulted in a complete loss of binding and a significant reduction in receptor activity for HU210, CP55940 and anandamide, but retained binding and receptor activity for WIN55212-2. It was shown that a similar binding affinity decrease occurred (17-fold) for SR141716A in the K192^3.28A mutant as was seen (14-fold) for a SR141716 analog lacking the C3 carboxamide oxygen binding in the wild type receptor [124,125]. Thus, the current understanding of ligand binding to the CB₁ receptor is heavily dependent upon the generally accepted role of K192^3.28 as one of the key residues that directly interact with CP55940 [122,126] and SR141716A [124,125] but not with WIN55212-2. Thus, most of the computational docking studies of the CB₁ receptor [91,95,114,128-133] employed K192^3.28 as the primary interaction site for the phenolic OH of the cannabinoid ligands through H-bonding.

However, as opposed to the generally accepted hypothesis that K192^3.28 residue directly contacts with the ligand (i.e., as a B_d residue), there have been a number of findings indicative of a role of K192^3.28 in indirectly modifying the binding pocket geometry (i.e., as a B_iS residue). First, the effect of the K192^3.28A mutation of the CB₁ receptor on ligand binding was, in fact, universal to all the ligands: anandamide and CP55244 were very sensitive and exhibited no binding, SR141716A binding was moderately sensitive to exhibit a 17-fold decrease [124], and WIN55212-2 was less sensitive and exhibited about 2-fold decrease [122,123]. This can be interpreted as follows: K192^3.28A mutation leads to an indirect modification of the ligand binding pocket geometry such that the binding of anandamide and CP55244 is more severely disrupted than the binding of WIN55212-2. Second, it was observed from the mutation study by Chin et al. [123] that CP55940 showed no difference in binding affinity for the wild type and the K192^3.28R CB₁ receptor. This finding suggests that K192^3.28 is not a B_d residue for CP55940, for if K192^3.28 directly interacts with CP55940 the binding should be less favored by the bulkier replacement. Third, the removal of the phenolic OH of CP55940 showed only a 35-fold drop in ligand binding but a 200-fold decrease in receptor activity [134]. This drop in ligand binding is much lower than expected [135] for deletion of the direct H-bond between the charged N of K192^3.28 and the phenolic OH of the ligand. Fourth, according to a recent CB₁ receptor homology model embedded in a lipid bilayer [136], it was shown that K192^3.28, as the charged form at physiological conditions, was snorkeling [137-139] by stretching out its long side chain N atom for the formation of a strong salt bridge with D184^E1 near the polar membrane surface, while continuing to surround the side chain hydrocarbon moiety with several hydrophobic residues from H2 and H3 inside the lipophilic membrane. It is conceivable, however, that K192^3.28 can directly interact with the ligand if the ligand contains a suitable functional group, such as a carboxylate or phenoxy moiety, or a π-electron rich moiety (i.e., an aromatic ring) that satisfies the positively charged N side chain of K192^3.28 in the hydrophobic lipid environment by forming a H-bond or π-cation interaction [140,141] by the replacement of the existing salt bridge with D184^E1.

Taken altogether, it is possible that K192^3.28 can play roles in ligand binding not by directly contacting with the ligand but by indirectly modifying the binding pocket geometry. With its role remaining to be further confirmed for a better understanding of the CB₁ receptor-ligand interaction, which is crucial not only for studying receptor activation but also for developing CB₁ receptor structure-based drugs, K192^3.28 is considered in this review as a B_iS residue for the CB₁ receptor ligands.

Emerging Binding Sites for Structurally Diverse Cannabinoid Ligands

The functional residues of the CB₁ receptor involved in ligand binding (B_d,B_iS and B_iA) are listed in Table 1. Based upon the CB₁ receptor residues defined as the types B_d that are directly involved in contacting with the ligand, the binding pockets of anandamide, CP55940, WIN55212-2 and SR141716A can be approximately delineated. Among these residues identified as key binding site residues, it is interesting to note that F189^3.25 [95] is uniquely defined for anandamide, F268^E2-I271^E2 [96] and C355^6.47 [91] for CP55940, G195^3.31 [118] and V282^5.46 [108] for WIN55212-2, and W356^6.48 and C386^7.42 [121] for SR141716A (Table 1 and Fig. (3)). These results indicate that, with some overlap in common structural features for CB₁ receptor binding, each class of structurally diverse ligands exhibits unique interactions with the receptor [142] for ligand-specific effects on signaling activity. Considering the fact that an agonist preferentially binds to the high-affinity state receptor (i.e., an active state receptor), where the rigid-body movements occur mainly in H6 and H7 according to the global toggle switch model [50], the B_d residues on H6 and H7 of the inactive receptor are not included as initial contacts for agonist binding but become available when the receptor is fully activated by the inward movements of H6 and H7 [50,103]. Thus, it is likely that C355^6.47 [91] is not an initial contact residue of the receptor in the inactive state with CP55940 (Fig. (3B)). In contrast, the B_d residues on H6 and H7 for SR141716A (e.g., W356^6.48 [95] and C386^7.42 [121]), positioned toward the binding core without rigid-body movements of H6 and H7, are directly involved in the ligand binding (Fig. (3D)) and stabilize the inactive receptor. In this regard, the competitive ligand binding between the CB₁ receptor agonists and SR141716A should be viewed in terms of not only competing with the ligand binding pocket but also competing with shifting the equilibrium between the inactive state and the active state.

Table 1.

Key residues of the CB₁ receptor involved in ligand binding (i.e., B_d, B_iS and B_iA residues) of anandamide, CP55940, WIN55212-2 and SR141716A

Ligand	Residue	Position	Residue type	Reference
anandamide	F1893.25	H3	Bd	[95]
	K1923.28	H3	BiS	[122]
	L2073.43	H3	BiA	[106]
	T2103.46	H3	BiA	[105]
	D2133.49,a)	H3	BiA	[104]
	Y2755.39	H5	Bd	[107]
	Y2945.58,a)	H5	BiA	[101]
	A3426.34,a)	H6	BiA	[104]
CP55940	H181E1	E1	BiS	[111]
	R182E1	E1	BiS	[111]
	D184E1	E1	BiS	[111]
	V1883.24	H3	BiS	[111]
	F1893.25	H3	BiS	[95,111]
	K1923.28	H3	BiS	[122,123]
	L2073.43	H3	BiA	[106]
	T2103.46	H3	BiA	[105]
	D2133.49,a)	H3	BiA	[104]
	F268E2	E2	Bd	[96]
	P269E2	E2	Bd	[96]
	H270E2	E2	Bd	[96]
	I271E2	E2	Bd	[96]
	Y2755.39	H5	Bd	[107]
	Y2945.58,a)	H5	BiA	[104]
	A3426.34,a)	H6	BiA	[104,159]
	C3556.47	H6	Bd	[91]
	S3837.39	H7	BiA	[114]
WIN55212-2	D1632.50	H2	BiA	[116]
	K1923.28	H3	BiS	[122,123]
	G1953.31	H3	Bd	[118]
	F2003.36	H3	Bd	[95,120]
	L2073.43	H3	BiA	[106]
	T2103.46	H3	BiA	[105]
	D2133.49,a)	H3	BiA	[104]
	Y2755.39	H5	Bd	[95]
	W2795.43	H5	Bd	[95]
	V2825.46	H5	Bd	[108]
	Y2945.58,a)	H5	BiA	[101]
	A3426.34,a)	H6	BiA	[104]
SR141716A	K1923.28	H3	BiS	[124]
	F2003.36	H3	Bd	[95]
	L2073.43	H3	BiA	[106]
	T2103.46	H3	BiA	[105]
	C257E2	E2	BiS	[121]
	C264E2	E2	BiS	[121]
	W2795.43	H5	Bd	[95]
	W3566.48	H6	Bd	[95]
	C3867.42	H7	Bd	[121]

^a)inferred from the results of the CB₂ receptor.

Fig. (3).

Extracellular top view of the key initial CB₁ receptor binding contacts (by ellipsoids), using the recently developed homology model of the CB₁ receptor (apo) in the inactive state [136], of (A) anandamide (in cyan), (B) CP55940 (in green), (C) WIN55212-2 (in blue) and (D) SR141716A (in red) are delineated by the type B_d residues (in stick). The type B_iS and B_iA residues are also represented in transparent stick. At the center of the figure, the side views of the ligand binding sites (by rectangles) of anandamide (in cyan), CP55940 (in green), WIN55212-2 (in blue) and SR141716A (in red) are depicted by the type B_d residues (in stick). Only the TM helical domain (in black ribbon) and E1/E2/E3 (in gray ribbon) is shown and other segments are omitted for clarity. For residues, only side chains are shown. W356^6.48 (in orange) of the proposed toggle switch [112,165] is also represented. Considering the fact that an agonist preferentially binds to the active state receptor, where the rigid-body movements occur in H6 and H7 [50,103] (by arrows), the B_d residues on H6 or H7 are not included as initial contacts for agonist binding. For example, C355^6.47 [91] & S383^7.39 [114] (in green dotted circles), B_d residue for CP55940, become fully engaged in ligand binding only when the receptor is fully activated. In contrast, B_d residues for SR141716A, including those in H6 and H7 (in red dotted circles), are fully engaged in ligand binding without a significant change in the receptor conformation. It appears that WIN55212-2 interaction with F200^3.36 disrupts W356^6.48, leading to the H6 displacement, while SR141716A interaction with F200^3.36 protects W356^6.48 from the H6 displacement. In this regard, the competitive ligand binding between the CB₁ receptor agonists and SR141716A should be viewed in terms of not only competing with the ligand binding pocket but also competing with shifting the equilibrium between the inactive state and the active state.

Only a few residues, including F189^3.25 [95] and Y275^5.39 [107,110], are known to directly interact with anandamide. Taken the H2/H3/H6/H7 region proposed to be important for anandamide binding [95] together with the exclusion of H6 and H7 due to their involvement in agonist binding at the late stage of receptor activation, the key initial contacts for anandamide binding would be in the H2-H3-H5 region (Fig. (3A)).

For CP55940, F268^E2/P269^E2/H270^E2/I271^E2 [96], Y275^5.39 [107] and C355^6.47 [91] are directly contacting with the ligand. A very recent substituted-cysteine accessibility method (SCAM) [143] study on H6 of the CB₂ receptor [100] showed that V261^6.51, L262^6.52, L264^6.54, M265^6.55, L269^6.59 and T272^6.62 were on the solvent-accessible surface of the binding site crevice of CP55940, suggesting that the equivalent residues in the CB₁ receptor are accessible to CP55940. Taken the region formed by H3-H5-H6-H7 proposed to be important for CP55940 binding [114,144] together with the exclusion of H6-H7 due to their involvement in agonist binding at the late stage of receptor activation, the key initial contacts for CP55940 binding would be in the H3-E2-H5 region (Fig. (3B)).

For WIN55212-2, G195^3.31 [118], F200^3.36 [95,120], Y275^5.39 [107], W279^5.43 [95] and V282^5.46 [108] are the residues in direct contact. Taken the proposed H3-H4-H5-H6 region for WIN55212-2 binding [95,145] together with the exclusion of H6/H7 due to their involvement in agonist binding at the late stage of receptor activation, the key initial contacts for WIN55212-2 binding would be in the H3-H4-E2-H5 region (Fig. (3C)).

For SR141716A, F200^3.36 [95], W279^5.43 [95], W356^6.48 [95] and C386^7.42 [121] are the residues in direct contact. It has been suggested that E1 and H3 (proximal to the extracellular side) [111] and E2 [96,110] are not sensitive to SR141716A binding. Taken the proposed H3-H4-H5-H6 region for SR141716A binding [95] together with the inclusion of H6/H7, the key contacts for SR141716A binding would be in the H3-H5-H6-H7 region (Fig. (3D)). It is interesting to note that the binding site of SR141716A is located at W356^6.48 [95], in the region deep in the receptor core and relatively distal to the extracellular side (see Fig. (3)), possibly protecting the receptor from activation.

Integrity of E2 of the CB₁ Receptor Critical for Ligand Binding and Receptor Activation

The second extracellular loop (E2) connecting H4 and H5 appears to play an important role in receptor stabilization, ligand binding and receptor activation [16,17,79,146-148]. It is shown from the GPCR X-ray structures that E2 residues form part of the binding site for the ligand. It has been suggested that different E2 residues are involved in different functions [149,150], possibly with distinct conformations [16,19]. It has been suggested from several studies [96,110,121,145,151] that E2 of the cannabinoid receptors are important for ligand binding and receptor activity. It is known that E2 Cys residues C257^E2 and C264^E2 of the CB₁ receptor are required for receptor stabilization [110,121]. It was shown from an early study by Martin and his colleagues [152] that CP55940 binding was inhibited competitively by sulfhydryl blocking agents but not by a disulfide reducing agent, suggesting that at least one reactive sulfhydryl group exists at the CP55940 binding site and that a disulfide bond, whose reduction impacts on CP55940 binding, exists away from the CP55940 binding site. Among the Cys residues conserved in the cannabinoid receptors, C355^6.47 would be the best candidate to provide the reactive sulfhydryl group with the following reasons: First, C355^6.47 is known to form part of the CP55940 binding site [91]; Second, it was shown from a mutational analysis [121] that the CB₁ receptor mutations of C257^E2 and C264^E2 did not affect CP55940 binding; Third, it was shown that sulfhydryl blocking of C386^7.42 did not affect CP55940 binding [121]. Accordingly, it can be argued that although the CB₁ receptor lacks the inter-disulfide linkage between H3 and E2, two E2 Cys residues form an intra-loop disulfide bond [110,121,145] important for CP55940 binding [152] as well as receptor stabilization [110,121].

It has been shown that the C-terminal of E2 of many GPCRs is important for ligand binding and receptor function [146,150,153]. It is also shown from the GPCR X-ray structures of rhodopsin [46] and β₂AR [40,154] that an aromatic residue of the Cys-X-X-X-Ar motif, where Cys is tied to another Cys from H3, conserved in many GPCRs [153], plays a role in receptor stabilization and ligand binding. The importance of the C-terminal region of E2 shown by many GPCRs has also been demonstrated in the CB₁ receptor [96]. Thus, it was shown from a recent Ala scanning mutagenesis of E2 of the CB₁ receptor [96] that the C-terminal residues F268^E2-I271^E2 were critical for CP55940 but not for SR141716A binding, suggesting that the C-terminal end of E2 plays an important role not only in ligand binding but also in receptor activation.

It has been suggested that E2 plays an important role in receptor activation, primarily by coupling to the TM helical domain [8,155,156]. It appears that E2 rearrangement triggered by the ligand is coupled to the TM helical domain, especially the segments of H5 [156] and H7 [8] to achieve receptor activation. For the CB₁ receptor, it was suggested from a recent study [96] that the C-terminal end of E2 of the CB₁ receptor played an important role in receptor activation, presumably through the coupling to H5 similar to rhodopsin [156]. It is possible that the B_d residues in the extracellular side of the H5 through H7 region [95,144], including Y275^5.39 [107], W279^5.43 [95], V282^5.46 [108] and C355^6.47 [91] of the CB₁ receptor (Table 1), would be involved in providing the necessary movement in the TM helical region for receptor activation.

Emerging CB₁ Receptor Activation Mechanism Involving E2

Activation of a GPCR is a multistep process [11,157] that is initiated by binding of an agonist to the inactive state of the receptor. Ahuja and Smith [103] proposed a mechanism of GPCR activation by multiple sequential switches, including a switch of E2 displacement, a switch of the H5 motion coupled to E2, a rotamer toggle switch of the CWxP motif on H6, and an internal switch of the NPxxY motif on H7, the final switch that eventually breaks the ionic lock between R^3.50 and E^6.30 on the intracellular sides of H3 and H6 to relay the molecular signal to the coupled G protein. Thus, according to this mechanism, the molecular signal from the extracellular side passes through the receptor in the following order: E2 -> H5 -> H6 (the rotamer toggle switch) -> H7 (the internal switch) -> H3/H6 (the ionic lock). As shown in Fig. (4A), most of these residues involved in receptor activation (i.e., A and B_iA residues) identified from the mutational studies of the CB₁ receptor are, in fact, part of the functional motifs conserved in many GPCRs [158] (see Fig. (1)). R214^3.50 [104], D338^6.30 [100], L341^6.33/A342^6.34 [159] and T210^3.46/L207^3.43 [105,106] are associated with the DRY motif; and D163^2.50 [116,160] of the SLAxAD motif. Although C355^6.47 [112,113] and W356^6.48 [95] of the CWxP motif are classified as B_d residues, they also play an important role in receptor activation. It was shown in a recent CB₁ receptor helical bundle model [136] that most of the functional motifs were involved in maintaining the inactive state of the CB₁ receptor, suggesting that disruption of these functional motifs is necessary in order to achieve a conformational change of the receptor necessary for G-protein signaling [49,161].

Fig. (4).

(A) Functional motifs, represented by space-filling, including the SLAxAD motif (in yellow) on H2, the DRY motif (in red) on H3, the CWxP motif (in orange) on H6, and NPxxY motif (in purple) on H7 of the CB₁ receptor are represented along with the CB₁ receptor types A and B_iA residues (in red stick) important for G protein signaling. It appears that disruption of these functional motifs is necessary for receptor activation [49,161]. W356^6.48 (in orange) of the proposed toggle switch [112,165] is also represented. It is noted that all the A and B_iA residues are located in the interface of H6 and its surrounding helices H2/H3/H5/H7, suggesting that the rigid-body movement of H6 is important for receptor activation (see text). (B) Molecular constraints of W356^6.48 in the inactive state of the CB₁ receptor [136], including the aromatic stacking interactions (in magenta triangle) between W356^6.48 and F200^3.36/F170^2.57, a direct H-bond network (in black dotted lines) by W356^6.48 and C386^7.42/N389^7.45/D163/N393, a water-mediated H-bond network (in black dotted lines) by C355^6.47/L359^6.51/C382^7.38/C386^7.42. The water molecule coordinated to the CWxP motif was quite stable during 105 ns duration of molecular dynamics (MD) simulations in a 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) bilayer [136]. The water molecules coordinated to the CWxP and the SLAxAD motifs are represented by space-filling. Color coding: C, cyan; O, red; N, blue; and S, yellow. H6 and H7 are represented in black ribbon, while other helices are represented in gray ribbon. Hydrogen atoms are omitted for clarity.

It has been proposed that upon receptor activation an outward movement of the extracellular side of H5, coupled to an outward displacement of E2 [156], releases the molecular constraints on W^6.48 [49], while an inward movement of the intracellular H5 replaces E^6.30 by Y^5.58, disrupting the ionic lock [103]. It has also been proposed that an outward movement of the intracellular side of H6, as a result of the disruption of the rotamer toggle switch [52], is one of the key steps in receptor activation [103]. The X-ray structure of rhodopsin [28] reveals that W265^6.48 is stabilized by aromatic stacking with F261^6.44 and Y268^6.51, highly homologous residues in rhodopsin class GPCRs [162], and by the water-mediated H-bond interaction with N302^7.49 of the NPxxY motif [163]. It has been suggested that the breakage of the H-bond between W265^6.48 and N302^7.49 allows M257^6.40 to move away and Y306^7.53 to come in contact with H6, contributing to breaking the ionic lock [164]. Thus, it appears to be crucial to see how W356^6.48 is stabilized in the CB₁ receptor inactive state for understating the mechanism of the CB₁ receptor activation. The CB₁ receptor lacks aromatic residues at the 6.44 and 6.51 positions, and consequently is expected to have unique interaction patterns for W356^6.48. As shown in Fig. (4B), it was revealed from a recent homology model of the CB₁ receptor in the inactive state [136] that W356^6.48 formed aromatic stacking with F200^3.36 and F170^2.57 and was involved in an extensive H-bond network by direct H-bonds with C386^7.42/N389^7.45. It was also shown that W356^6.48 was conserved by a water-mediated H-bond network by C355^6.47/L359^6.51/C382^7.38/C386^7.42, located just above W356^6.48 (Fig. (4B)). The importance of the aromatic stacking between W356^6.48 and F200^3.36 is supported by the proposed rotamer toggle switch by W356^6.48/F200^3.36 of the CB₁ receptor [165; 112], similar to W286^6.48/F290^6.52 of β₂AR [113] and also by the finding that the F200A mutation of the CB₁ receptor resulted in higher constitutive activity compared with the wild type receptor [120]. Thus, alternative ways to release the molecular constraints on W356^6.48 of the CB₁ receptor to achieve CB₁ receptor activation include: 1) the breakage of aromatic stacking between W356^6.48 and F200^3.36/F170^2.57; and 2) the breakage of a water-mediated H-bond network by C355^6.47/L359^6.51/C382^7.38/C386^7.42.

It was also shown in a recent CB₁ receptor model [136] that N393^7.49 of the NPxxY motif formed direct H-bonds to N389^7.45 and D163^2.50 (Fig. (4B)). Thus, similar to rhodopsin, the breakage of the N389^7.45-mediated H-bond between W356^6.48 and N393^7.49 contributes to breaking the ionic lock in the CB₁ receptor. It has been proposed that L207^3.43, one of the highly conserved residues among many GPCRs [158], maintains the receptor in the inactive state by locking the movement of H6. In support of this, L207^3.43 of the CB₁ receptor has been reported to be a CAM residue [106]. It is possible that an outward movement of the intracellular side of H6 allows I348^6.40, similar to M257^6.40 in rhodopsin [164], to move away from L207^3.43, thereby causing N393^7.49 to come in contact with H6 (Fig. (5)). Taking into consideration the role of D163^2.50 in receptor activation [116], it is possible that the breakage of the interaction between N393^7.49 and D163^2.50, which is proposed as an activation switch in other GPCRs [166-168], would contribute to breaking the ionic lock. It also appears that upon receptor activation Y294^5.58 and Y397^7.52 of the CB₁ receptor, both highly conserved in many GPCRs, come closer and contribute to the breakage of the ionic lock (Fig. (5)).

Fig. (5).

The proposed molecular mechanism of the CB₁ receptor activation by CP55940. The CB₁ receptor model of the inactive state [136] is used. Propagation of the molecular signal from the extracellular side to the intracellular side of the receptor is depicted by a series of the sequential steps: ligand binding (in green) -> E2 coupling to H5/H7 (in yellow) -> the W356^6.48 rotameric trigger and the breakage of the H-bond between W356^6.48 and N389^7.45 (in orange) -> a series of internal switches, including the interaction between L207^3.43 and N393^7.49 (in purple) and the interaction between Y294^5.58 and Y397^7.52 (in pink) -> the breakage of the ionic lock between R214^3.50 and D338^6.30 (in red). The rigid-body movements in H6 and H7 [50,103] are represented by arrows.

Ligand-Specific CB₁ Receptor Activation

Supported by similar results from other GPCR systems [169], several studies on the CB₁ receptor [170,171] suggest that different CB₁ receptor ligand classes evoke selective G-protein signaling. Considering the concept that one receptor can couple to different signaling pathways by different ligands [10], the binding of structurally distinct CP55940 and WIN55212-2 contributes to the ligand-specific conformational change in the receptor [131]. In this regard, it is insightful to see that among the residues of the CB₁ receptor involved in the ligand-induced receptor activation (i.e., B_d residues), F268^E2-I271^E2 [96] and C355^6.47 [91] are unique to CP55940, while G195^3.31 [118], F200^3.36 [95,120], W279^5.43 [95] and V282^5.46 [108] are unique to WIN55212-2 (Table 1). Thus, these unique residues play important roles in inducing ligand-specific receptor activation. Considering the suggested important role of aromatic stacking in receptor activation [95,113,157,172], it is likely that F268^E2 [96] and F200^3.36 [95,120] are important for receptor activation by CP55940 and WIN55212-2, respectively. Thus, employing the known type B_d residues, albeit limited in number, it is possible to describe the ligand-specific interference with the CWxP motif as the key step of the CB₁ receptor activation mechanism [112,165]. Here, the ligand-specific CB₁ receptor activation is illustrated by comparing CP55940 with WIN55212-2.

For CP55940, the B_d residues include F268^E2/P269^E2/H270^E2/I271^E2 [96], Y275^5.39 [107] and C355^6.47 [91]. It has been suggested that the E2 C-terminal F268^E2 is crucial for CP55940 binding but not for that of SR141716A [96]. It has also been suggested that Y275^5.39 is crucial for receptor activation by CP55940 [107]. Thus, upon CP55940 binding, the aromatic stacking interaction between the ligand aromatic A-ring and the receptor residues, including F268^E2 [96] and Y275^5.39 [107], composes the key initial contacts, which allows the hydrophobic C3 side chain of CP55940, the steric trigger for receptor activation [173], to position deep in the core. It is likely that the interaction between the C3 side chain of the ligand and L359^6.51 on the solvent-accessible surface of the binding site crevice of CP55940 [100] contributes to the breakage of the water-mediated H-bonds (Fig. (4B)), which, in turn, interferes with the extracellular side of H7 and with the CWxP motif (Fig. (5)). As a result, the molecular constraints on W356^6.48 are released and W356^6.48 becomes free to rotate for receptor activation. It is possible that the coupling movements of H5/H7 [8,155,156] by the ligand binding to E2 [96] initially and then to the extracellular side of the H5 through H7 region [95,144] assist of the rigid-body rotation of H6 [52,174-179]. As shown in Fig. (3B), C355^6.47 [91], whose side chain is pointing toward H7 and located off the receptor core region, is less likely to be accessible to CP55940 in the receptor inactive state, but becomes available for ligand binding with the rigid-body rotation of H6 upon receptor activation. The findings that C355^6.47 [91] and S383^7.39 [114] of the CB₁ receptor were important for CP55940 but not for WIN55212-2 suggest that these residues play a role in CP55940-specific receptor activation. Overall, as shown in Fig. (5), the molecular signal by CP55940 binding from the extracellular side passes through the receptor in the following order: E2 -> H5/H7 -> H6 (the rotamer toggle switch) -> H7 (the internal switch) -> H6 (the ionic lock).

For WIN55212-2, the B_d residues include G195^3.31 [118], F200^3.36 [95,120], Y275^5.39 [107], W279^5.43 [95,120] and V282^5.46 [108]. Upon binding, WIN55212-2 interacts directly with Y275^5.39 [107] and W279^5.43 [95,120] within the proposed aromatic microdomain [95] which allows the aromatic naphthyl ring, the steric trigger of WIN55212-2 for receptor activation (Shim and Howlett, 2006), to bind to F200^3.36 of the proposed rotamer toggle switch of the CB₁ receptor [112,165]. It is likely that the interaction between the naphthyl ring of the ligand and F200^3.36 [95,120] contributes to the disruption of aromatic stacking between W356^6.48 and F200^3.36, interfering with the extracellular side of H3 and with the CWxP motif. As a result of the rotameric change in F200^3.36, the molecular constraints on W356^6.48 are released and W356^6.48 becomes free to rotate for receptor activation. Because the water-mediated H-bonds above W356^6.48 are in close proximity to F200^3.36 (Fig. (4B)), it is possible that the water-mediated H-bonds are broken by WIN55212-2 binding to F200^3.36. The findings that W279^5.43 [95,120] and V282^5.46 [108] of the CB₁ receptor were important for WIN55212-2 but not for CP55940 suggest that these residues play a role in WIN55212-2-specific receptor activation. Overall, the molecular signal by WIN55212-2 binding passes through the receptor in the following order: H5 -> H5/H3 -> H6 (the rotamer toggle switch) -> H7 (the internal switch) -> H6 (the ionic lock).

Interestingly, in order to disrupt the CWxP motif, CP55940 interferes with H7 and WIN55212-2 interferes with H3, through unique interactions with H5. Both paths of interference appear to be effective in disrupting the CWxP motif according to the global toggle switch model [50] where upon activation, W356^6.48 of the CWxP motif was displaced with the outward movement of the intracellular segment of H6. The observed ligand-specific G-protein signaling is likely to be attributed different ways to disrupt W356^6.48, through H5/H7 in CP55940 or through H5/H3 in WIN55212-2. It has been proposed that H5 plays an important role in agonist-specific conformational change [180]. Thus, it is tempting to speculate that different paths of disrupting the CWxP motif contribute to the ligand-specific receptor activation.

CB₁ Receptor Homology Models Suitable for Drug Design

Even before the X-ray structures of GPCRs became available, in their early pioneering works, Reggio and her colleagues [181] successfully determined the length and orientation of the membrane spanning 7 TM helices and the presence of H8 in the CB₁ receptor. After the X-ray structure of rhodopsin [27] became available, this X-ray structure was utilized as the template to construct cannabinoid receptor homology models [95,124,128-130,132,182-184]. Recent CB₁ receptor homology models [96,136] were constructed using the X-ray structure of β₂AR [40] as templates. In spite of the similarity in overall structural topology, homology receptor models using these X-ray structures as templates may result in structures which are locally quite different. Although the CB₁ receptor shows a low sequence homology to β₂AR and rhodopsin (43 % and 42 % identical to human β₂AR and bovine rhodopsin), homology models of the CB₁ receptor using these GPCRs as templates have been justified by reliable sequence alignment with these GPCRs according to the highly conserved amino acid residues and functional motifs within the TM region (Fig. (1)), which are conserved in greater than 90 % of all GPCRs [158]. Yuzlenko and Kieć-Kononowicz [185] tested the feasibility of the homology models of adenosine receptors using the X-ray structures of rhodopsin and β₂AR as templates and concluded that the β₂AR-based homology models were better than the rhodopsin-based homology models judging from the stability of the ligands inside the binding pockets. The feasibility of the homology models of the CB₁ receptor, using either rhodopsin or β₂AR as the template, remains to be seen. The CB₁ receptor is 44 %, 45 %, 43 %, 42 %, and 45 % identical to AA2R (human), β₁AR (turkey), β₂AR (human), rhodopsin (bovine), and rhodopsin (squid), respectively. This suggests that the X-ray structures of AA2R [42], β₁AR [38] and squid rhodopsin [186] would be better templates for constructing homology models of the CB₁ receptor.

Some of recent homology models of the CB₁ receptor [136,184,187] have taken advantage of increasing computational resources and were determined in a fully hydrated lipid bilayer to mimic the physiological environment. Such homology models of the CB₁ receptor provide a detailed understanding of its interaction with the lipid bilayer, including hydrophobic core and hydrophilic interfaces, and with water [188]. Some studies [95,132,184] also included CB₁ receptor activated state models by modifying their receptor homology models in the inactive state on the basis of the receptor activation information obtained from biophysical studies [52,174-179]. On one hand, the active state of a GPCR is the state that perfectly fits to an agonist; a homology model of the CB₁ receptor in its active state would better serve to understand the molecular interactions with an agonist. On the other hand, it is still very challenging to construct a computational model of a fully activated GPCR due to the reasons described earlier even with the large amount of accumulated biophysical data on the GPCR active state.

Structure-Based Drug Design in the CB₁ Receptor Models

The X-ray structures of GPCRs and high quality homology models have been used for identifying potential lead compounds by the popular GPCR drug-design approach of structure-based high throughput screening (HTS) of a large collection of commercially available compounds [189-196] (for recent reviews, see [54,197]). Similarly, HTS has been applied to the cannabinoid receptor area [129,198,199]. A new drug-design approach, fragment-based approach (FBA) [200-202] has been successfully applied to screen relatively few compounds (approximately 1,000 compounds) using NMR or X-ray crystallography and to identify lead-like fragments with relatively low binding affinity (in the μM range) positioned non-redundantly at the binding pocket that can be connected to derive novel compounds with relatively high binding affinity (in the nM range). Compared with HTS, which needs an exhaustive search of the chemistry space allowed for the library of larger molecules, the chemistry space is efficiently probed by screening collections of small ligand fragments [202]. FBA using NMR or X-ray crystallography cannot be applied widely to GPCRs owing to the difficulty in obtaining their X-ray structures, particularly for small fragments within the active site [56]. To overcome this problem, a similar approach to FBA, multiple copy simultaneous search (MCSS) [203] can be applied to homology modeled GPCRs. MCSS generates several thousand replicas of a given functional group and orients the functional group in a favorable way within the binding-site region of the receptor [203-205]. A recent study [206], where MCSS approach was applied against the homology models of the ARs constructed by using the X-ray structure of rhodopsin as a template, successfully identified a few potent antagonists selective to the α_1dAR.

It appears that if any homology model is constructed using a template of the X-ray GPCR in its apo state or inverse agonist bound type, the binding region needs be modified to accommodate various types of ligands [207-208]. For the purpose of rational drug design based upon a homology model of the CB₁ receptor, it appears that the early intermediate stages of the active state of the receptor bound to an agonist, resembling the inactive state of the receptor, are sufficient as alternative structures to the fully activated structure for the receptor-based agonist design [209]. In this case, it is crucial to identify the receptor residues involved in kinetically distinct steps in the receptor activation. For β₂AR, it has been proposed that the agonist binding of β₂AR occurs through at least three distinct steps [11,157]: 1) the first step involves an ionic/H-bond interaction between the protonated amine of the agonist and D113^3.32 and N312^7.39 of the receptor and aromatic stacking between the catechol ring of the ligand and F290^6.52 of the receptor. A combination of these interactions composes the key initial ligand contacts to a minimal low-affinity receptor binding site; 2) the second step involves H-bonding interactions between the catechol hydroxyl groups of the ligand and S203^5.42, S204^5.43, and S207^5.46 of the receptor [98]. These interactions, which properly position the catechol ring of the ligand and H5, are required for the receptor to achieve the fully active state [210]; and 3) the third step involves a rotamer toggle switch of W286^6.48/F290^6.52 [113], which modulates H6 movement about the P6.50 kink, and H-bond interaction between the side chain hydroxyl group of the ligand and N293^6.55 of the receptor. Accordingly, de Graaf and Rognan [209] modified the rotameric states of S204^5.43 and S207^5.46 of the X-ray of β₂AR and successfully screened selective full and partial agonists.

Thus, similar steps of agonist binding can be considered for the CB₁ receptor. Several issues need to be addressed: First, with a variety of ligands with distinct structures, the cannabinoid ligand binding pocket has not been well established. It is generally accepted that structurally diverse cannabinoid agonists bind to the binding sites uniquely defined for the individual classes of the ligand but with partial overlap [142] (see Fig. (3)). Second, in contrast to β₂AR agonists, classical and non-classical cannabinoid agonists do not contain the protonated amine, while WIN55212-2 contains the morpholino N that can be protonated. On the other hand, similar to β₂AR agonists, most cannabinoid agonists do contain at least one aromatic moiety. Thus, aromatic residues [95,96,107,120] in the extracellular side of the CB₁ receptor would serve as the key initial ligand contacts to a minimal low-affinity receptor binding site. Third, it is conceivable that several aromatic residues on H5 of the CB₁ receptor, including Y275^5.39 [107] for CP55940 and Y275^5.39 [107] and W279^5.43 [95] for WIN55212-2, perform a similar role as the agonist-contact residues S203^5.42, S204^5.43 and S207^5.46 [98] on H5 of β₂AR. These residues need to be repositioned with a rotameric change for a full activation of the receptor [210]. Fourth, the toggle switch W356^6.48/F200^3.36 [165] and the VxxI helical groove [127] just proximal to the CWxP sequence of H6 can be the key interaction sites for the agonist binding contacts that induce the agonist-initiated receptor micro-conformational change [131,173].

Based upon the results from CB₁ receptor mutagenesis, pharmacological and computational studies, it is possible to generate a CB₁ receptor model in an early active state by modifying the rotameric states of the binding site residues. Candidate modifications include C355^6.47 for CP55940 [91] and F200^3.36 [95,120]. These residues are known to be involved in receptor activation as well as ligand binding. Such early active state CB₁ receptor models are useful in screening CB₁ receptor selective full and partial agonists, as similarly demonstrated for β₂AR [209]. As one example of this strategy, Salo et al. [129] modified a homology model of the CB₂ receptor and docked HU210 in the binding pocket prior to database screening to identify novel agonist candidates.

Summary

The CB₁ receptor is a GPCR with conformational flexibility as indicated by its high basal activity. The CB₁ receptor functional residues reported from mutational studies are utilized to describe the molecular events of the ligand-specific CB₁ receptor activation from ligand binding to G protein signaling. For the CB₁ receptor, as shown in many GPCRs, the integrity of the extracellular loops in ligand binding and coupling to the TM helical domain, the interference of the CWxP motif and the breakage of the ionic lock are important for receptor activation. Yet, little is known about the molecular structure of the CB₁ receptor in the active state. But, it is possible to construct homology models of the CB₁ receptor corresponding to the early stage of its activation that are suitable for identifying new therapeutic agents by applying the validated structure-based approaches, such as virtual HTS and FBA.