Molecular Basis of Cannabinoid CB1 Receptor Coupling to the G Protein Heterotrimer Gαiβγ

Background: The molecular basis of CB1 coupling to its cognate G protein is unknown. Results: Using an approach combining mutagenesis and molecular dynamics simulations, we identified CB1 residues critical for G protein signaling. Conclusion: Tight interactions between CB1 and the C-terminal helix α5 of Gαi are crucial for G protein signaling. Significance: This is the first reported molecular description of CB1 receptor coupling at the receptor-Gi interface. The cannabinoid (CB1) receptor is a member of the rhodopsin-like G protein-coupled receptor superfamily. The human CB1 receptor, which is among the most expressed receptors in the brain, has been implicated in several disease states, including drug addiction, anxiety, depression, obesity, and chronic pain. Different classes of CB1 agonists evoke signaling pathways through the activation of specific subtypes of G proteins. The molecular basis of CB1 receptor coupling to its cognate G protein is unknown. As a first step toward understanding CB1 receptor-mediated G protein signaling, we have constructed a ternary complex structural model of the CB1 receptor and Gi heterotrimer (CB1-Gi), guided by the x-ray structure of β2-adrenergic receptor (β2AR) in complex with Gs (β2AR-Gs), through 824-ns duration molecular dynamics simulations in a fully hydrated 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine bilayer environment. We identified a group of residues at the juxtamembrane regions of the intracellular loops 2 and 3 (IC2 and IC3) of the CB1 receptor, including Ile-2183.54, Tyr-224IC2, Asp-3386.30, Arg-3406.32, Leu-3416.33, and Thr-3446.36, as potential key contacts with the extreme C-terminal helix α5 of Gαi. Ala mutations of these residues at the receptor-Gi interface resulted in little G protein coupling activity, consistent with the present model of the CB1-Gi complex, which suggests tight interactions between CB1 and the extreme C-terminal helix α5 of Gαi. The model also suggests that unique conformational changes in the extreme C-terminal helix α5 of Gα play a crucial role in the receptor-mediated G protein activation.

class GPCR is the largest with ϳ680 members (2). They are known to be some of the most important drug targets (3). A GPCR is coupled to the cognate heterotrimeric G␣␤␥ protein, the G␣ subunit of which is composed of two domains, a Ras-like nucleotidebinding domain (G␣Ras) that interacts not only with the receptor but also with the G␤ subunit and an ␣-helical domain (G␣AH) that covers the bound nucleotide on G␣Ras (4). When a GPCR is activated by an appropriate signal, it binds the G protein and catalyzes the release of guanosine diphosphate (GDP) from the nucleotidebinding pocket. The nucleotide-free G␣ binds guanosine triphosphate (GTP), and the resulting G␣-GTP dissociates from the receptor and G␤␥. The freed G protein subunits regulate adenylate cyclase, ion channels, phospholipase C, or the guanine nucleotide exchange factor RhoGEF activity (5). The G␣ protein hydrolyzes the GTP to GDP and reassociates with the ␤␥ dimer and the receptor to complete a cycle of G protein activation (6). In parallel with G protein signaling, a portion of activated receptors is simultaneously phosphorylated by GPCR kinase and binds to ␤-arrestin, initiating desensitization, internalization, and ␤-arrestin signaling (7,8).
Due to the nature of their structural dependence on intact cell membranes, elucidation of the function of integral membrane GPCRs has been hampered by the lack of high quality structural data. Recent publication of the x-ray crystal structures of some GPCRs has helped greatly to remedy this situation. With early x-ray structures of GPCRs in the inactive state, the overall transmembrane (TM) topology indicative of common structure and functional elements of GPCRs (9) was found to be conserved (for a review, see Ref. 10). With more recent x-ray GPCR structures in the active state, we now have a general view of GPCR activation, where an intracellular (IC) opening of the sixth TM helix (TM6) of the activated receptor is important for G protein interaction (for a review, see Ref. 11 and references therein). Moreover, the x-ray structure of the active ␤ 2 -adrenergic receptor (␤ 2 AR) in complex with the nucleotide-free active state of G s (␤ 2 AR-G s ) (12) allows us to have a general view of the molecular assembly of the GPCR-G protein complex, providing critical structural insight into how GPCRs interact with G proteins during the G protein activation cycle. Compared with the active state of the ␤ 2 AR alone (13), the receptor structure in the ␤ 2 AR-G s complex shows little change in the TM bundle but some noticeable changes in the IC region. These are an additional outward movement (ϳ3 Å) of the IC end of TM6 and a helical extension of TM5 toward the IC face in response to the coupled G protein. This provides a tight interaction especially with the extreme C terminus of G␣ at the end of helix ␣ 5 , the segment important for the receptor-mediated G protein activation (6,14). One striking molecular feature in the x-ray structure of the ␤ 2 AR-G s complex is that G␣AH is completely displaced from G␣Ras and moves in a direction toward the N terminus of G␣ near the IC membrane surface. Such displacement was predicted from the x-ray structure of G␣ (4) and experimentally confirmed by site-directed spin labeling and deuterium exchange mass spectroscopy studies (15). GPCR-mediated GDP-GTP exchange, where GDP release is the rate-determining step (16), is the key event in the G protein activation cycle. Despite the x-ray structure of the ␤ 2 AR-G s complex (12), it remains uncertain how the tightly bound GDP is released from the inactive G protein as a result of the binding of the activated receptor.
The human CB1 (brain cannabinoid) receptor is associated with neurons in the brain, spinal cord, and the peripheral nervous system (17), where it primarily activates the pertussis toxin-sensitive inhibitory G i/o protein and is among the most expressed receptors in the brain (18). It has been implicated in several disease states, including drug addiction, anxiety, depression, obesity, and chronic pain (19). CB1 exhibits modest basal activity in the absence of ligand, indicative of receptor-G protein coupling independent of agonist (20). Multiple chemical classes bind to the CB1 receptor and activate signal transduction pathways (21) in an agonist-specific manner (22). The molecular basis of CB1 receptor coupling to its cognate G protein is unknown. Thus, in the present study, we have constructed a ternary complex structural model of the CB1 receptor in complex with the GDP-bound inactive state of G i heterotrimer (CB1-G i ) as a first step toward understanding CB1 receptor-mediated G protein signaling. Because G␣ s and G␣ i have 42% overall sequence identity (23) and the CB1 receptor and ␤ 2 AR belong to the same rhodopsin family GPCRs, we started with a model of the CB1-G i complex resembling the x-ray structure of the ␤ 2 AR-G s complex (12) and determined the specific orientation of the two proteins (i.e. CB1 and G i ) through an extensive molecular dynamics (MD) simulation in a fully hydrated 1-palmitoyl-2oleoyl-sn-glycero-3-phosphocholine (POPC) membrane environment. By site-directed mutagenesis studies, we identified a group of residues at the juxtamembrane region of the IC loops 2 and 3 (IC2 and IC3) of the CB1 receptor, including Ile-218 3.54 , Tyr-224 IC2 , Asp-338 6.30 , Arg-340 6.32 , Leu-341 6.33 , and Thr-344 6.36 , as potential key contacts with the extreme C-terminal helix ␣ 5 of G␣ i . 3,4 Ala mutations of these residues at the receptor-G i interface resulted in little G protein coupling activity. The present model of the CB1-G i complex reveals that within the CB1-G i interface, there exist strong intermolecular interactions between IC2 and IC3 of the CB1 receptor and the G␣ i subunit. Overall, the results of the present studies are in agreement with recent experimental findings suggesting a crucial role of the extreme C-terminal helix ␣ 5 of G␣ in the receptor-mediated G protein activation.

CB1 Plasmid Construction, Expression, and Membrane
Preparation-To evaluate the importance of the residues identified from molecular modeling in G protein coupling, we produced nine mutant receptors, I218 3.54 A, H219 IC2 A, P221 IC2 A, Y224 IC2 A, D338 6.30 A, I339 6.31 A, R340 6.32 A, L341 6.33 A, and T344 6.36 A. All mutants were generated by site-directed mutagenesis (QuikChange, Stratagene, La Jolla, CA) using the human CB1 cDNA cloned into pcDNA3.1 as a template and confirmed by DNA sequencing. HEK293 cells were grown in DMEM supplemented with 10% fetal bovine serum and 3.5 mg/ml glucose at 37°C in 5% CO 2 and seeded at 1 million cells/ 100-mm dish on the day prior to transfection using the calcium phosphate precipitation method (24). 24 h post-transfection, membranes of transfected cells expressing the wild-type or mutant receptors were prepared using nitrogen cavitation as described previously (25).
Radioligand Binding Assay-Saturation binding assays were performed as described previously (26). Briefly, ϳ5 g of membranes were incubated at 30°C for 60 min with [ 3 H]CP55940 (147.9 Ci/mmol; PerkinElmer Life Sciences) or [ 3 H]SR141716A (43 Ci/mmol; PerkinElmer Life Sciences) in a total volume of 200 or 500 l of TME buffer (25 mM Tris-HCl, 5 mM MgCl 2 , and 1 mM EDTA, pH 7.4) containing 0.1% fatty acid-free bovine serum albumin. At least nine radiolabeled-ligand concentrations were used to determine K d values of the receptors. Nonspecific binding was determined in the presence of 1 M unlabeled ligand. Reactions were terminated by filtration with a Brandel cell harvester through Whatman GF/C filter paper (Brandel Inc., Gaithersburg, MD) followed by four washes with ice-cold TME buffer to remove unbound radioactivity. Radioactivity was measured by liquid scintillation counting.
GTP␥S Binding Assay-Approximately 5 g of membrane preparations from HEK293 cells expressing CB1 receptors were incubated for 60 min at 30°C in a total volume of 500 l of GTP␥S binding assay buffer (50 mM Tris-HCl, pH 7.4, 3 mM MgCl 2 , 0.2 mM EGTA, and 100 mM NaCl) with unlabeled CP55940 (at least nine different concentrations were used, ranging between 100 pM and 100 M), 0.1 nM [ 35 S]GTP␥S (1250 Ci/mmol; PerkinElmer Life Sciences), 10 M GDP, and 0.1% (w/v) BSA. The basal GTP␥S binding was measured in the absence of ligand. Nonspecific binding was determined with 10 M unlabeled GTP␥S (Sigma). The reaction was terminated by rapid filtration to separate the membrane-bound fraction from the free fraction through Whatman GF/C filters. The radioactivity of the membrane-bound [ 35 S]GTP␥S trapped in the filters was determined by liquid scintillation counting.
Data Analysis-All ligand binding assays and GTP␥S binding assays were carried out in duplicate. Data are presented as the mean Ϯ S.E. value except for EC 50 values, which are the median with the corresponding 95% confidence limits, from at least three independent experiments. For the saturation radioligand binding assay, the K d and B max values were calculated by nonlinear regression using GraphPad Prism software (GraphPad Software Inc., San Diego, CA). EC 50 values for the GTP␥S binding assays were calculated using a sigmoidal dose-response relationship. Statistical significance between the wild-type and mutant receptors was assessed by analysis of variance, followed by Bonferroni's post hoc test. p values of Ͻ0.05 were considered to be statistically significant.
Structural Model of the CB1 Receptor-We chose the previously published active state model of the CB1 receptor (27) as a starting structure of CB1. In this CB1 receptor model, Glu-133 1.49 was neutralized, whereas Asp-163 2.50 remained charged, because the charged Asp-163 2.50 appears to be required for receptor activation in many diffusible ligand GPCRs (28). Similarly, the side chain of Asp-213 3.49 of the CB1 receptor was neutralized, because the neutralization of the corresponding residue in ␤ 2 AR and rhodopsin is crucial for receptor activation and G protein interaction (29,30). All other ionizable residues were in their ionization state at physiological conditions. For the N-terminal end, the first 10 residues from Asn-112 1.28 at the start of TM1 were retained. For the C-terminal end, the palmitoyl moiety (PAL) coordinated to Cys-415 was included. An acetyl group and an N-methyl group were attached to the N terminus and the C terminus, respectively, of the CB1 receptor. Briefly, the active state model of the CB1 receptor (27) was obtained by a smooth conversion of the receptor from an early stage of the cannabinoid agonist (Ϫ)-11hydroxydimethylheptyl-⌬ 8 -tetrahydrocannabinol (HU210)bound state (31) to the active form using some of the key molecular features of the active state of GPCRs (13,(32)(33)(34). We created five sets of distances between four residues at similar positions at the IC half of TM3, TM5, TM6, and TM7 in the x-ray structure of the active state for ␤ 2 AR (13) and performed the distance-constrained MD simulations. To avoid any abrupt changes in the structure, we obtained the targeted distances over five short distance-constrained MD simulations (3 ns each) using adaptive biasing force (35). When the targeted distances were reached, we gradually released the applied force (k ϭ 100) by scaling down the force by 80% over 20 short simulations (ϳ5 ns each). We finally performed the simulation without any constraint for over 150 ns until the protein structure was fully converged.
Considering that IC3 of active GPCRs appears to be highly structured, forming parallel TM5 and TM6 at the IC end (13,36), we refined the resulting active state model of the CB1 receptor to build a highly helical IC3 structure using the secondary structural information from an NMR study of the peptide (Lys-300 to Thr-344) corresponding to IC3 of the CB1 receptor (37). We chose one of the snapshots of the CB1 receptor model near the end of the simulation, where the helical structure of TM5 was extended to Lys-315 IC3 of the N-terminal region of IC3 with a kink at Ala-301 IC3 , and performed a constrained MD simulation by applying torsional constraints of an ideal helix ( ϭ Ϫ57°and ϭ Ϫ47°) for (i) Lys-300 IC3 -His-304 IC3 to correct the helical gap at A301 IC3 , (ii) Ser-316 IC3 and Ile-317 IC3 to extend TM5, and (iii) Gln-334 IC3 -Ile-339 6.31 to extend TM6. During 25 ns of the simulation, the initially applied constraints (k ϭ 100) were gradually released. We continued the simulation without constraint for ϳ80 ns. The final snapshot from the simulation was used as the starting structure for the construction of a model of the CB1-G i complex. All of the above mentioned simulations of the receptor embedded in a fully hydrated POPC lipid bilayer were performed at 310 K in the constant pressure (NPT) ensemble. The detailed setup of the simulations is described under "Simulations." Structural Model of G i -We chose the x-ray structure of the GDP-bound inactive state of G␣ i ␤␥ (Protein Data Bank code 1GP2) (38) as a starting structure and made the following protein modifications on the G␣ and G␥ proteins. For G␣ i , (i) we manually added the missing residues Cys-3 G␣i and Thr-4 G␣i to the N-terminal end residue Leu-5 G␣i in preparation for lipid modification, because it is known that the N-terminal Cys-3 G␣i is myristoylated or palmitoylated in the lipid bilayer (39); (ii) we replaced Ala-98 G␣i by a Ser residue to generate the human G␣ i sequence; and (iii) we also added the extreme C-terminal segment (Lys-349 G␣i -Phe-354 G␣i ), which was missing in the x-ray structure (38), using the coordinates of the corresponding segment (Lys-345 G␣t -Phe-350 G␣t ) in the x-ray structure of the G t C-terminal peptide (G t CT) (Protein Data Bank code 3DQB) (36). For G␥, we added the missing C-terminal segment (i.e. Arg-62 G␥ -Leu-71 G␥ ) using the coordinates from the x-ray structure of G␤ 1 ␥ 2 (Protein Data Bank code 1OMW) (40). After completion of the protein modifications, the extreme N-terminal and C-terminal end residues of each protein were capped as follows. (i) For G␣, the N terminus was acetylated, whereas the C terminus was retained as a free carboxylate form due to its importance in interacting with the receptor (6). (ii) For G␤, an acetyl group was attached to the N terminus, whereas an N-methyl group was attached to the C terminus. (iii) For G␥, an acetyl group was attached to the N terminus, whereas a methoxy group was attached to the C terminus in preparation for lipid modification. The resulting G i protein was solvated using the SOLVATE plug-in in VMD (41) and put into a water box of ϳ100 ϫ 120 ϫ 110 Å 3 . After removing solvent water molecules within 1 Å of the crystal water molecules, we carried out an energy minimization of 2,500 iterations, followed by a short MD simulation of ϳ3 ns, during which the backbone atoms of the G protein were fixed. The GDP-bound G i protein was then extracted from the water box and used as the starting structure for the construction of a model of the CB1-G i complex (see below).
Construction of a Model of the CB1-G i Complex-The GDPbound G i protein was docked to the HU210-bound CB1 receptor embedded in a fully hydrated POPC bilayer using the following two sequential steps. In Step 1, the x-ray structure of the ␤ 2 AR-G s complex (12), as a docking template, was superimposed onto the CB1 receptor with respect to the sequencealigned TM helical backbone atoms, using the highly conserved TM residues as published previously (42). In Step 2, the GDPbound G␣ i was then fitted to the G s of ␤ 2 AR-G s superposed on the CB1 receptor from Step 1 with respect to the backbone atoms of the commonly conserved secondary structures in the G␣Ras domain. We used SuperPose (43) for superposition. A PAL was added to Cys-3 G␣i of Ga i N terminus (Ga i )NT and a geranylgeranyl moiety was added to Cys-68 G␥ of G␥CT (39), such that both lipid moieties were positioned just below the IC membrane layer. Any water molecule and POPC molecule within 1.5 Å of the newly added G i and the following lipid modifications was removed from the system of the CB1-G i complex model. Sodium chloride molecules were used to ionize (0.15 M) and neutralize the system to satisfy electrostatic calculations. A total of ϳ219,600 atoms, including two proteins (CB1 and G i ), two bound ligands (HU210 and GDP), ϳ51,450 water molecules, ϳ360 POPC molecules, 73 Na ϩ , and 72 Cl Ϫ , resulted in a system of the CB1-G i complex in a simulation box of ϳ90 ϫ 130 ϫ 190 Å 3 (Fig. 1A). At the end of this stage, the area per lipid in the system was ϳ60 Å 2 . Because it has been demonstrated that sodium chloride contributes to a compression of POPC membranes by ϳ10% (44,45), this value appeared to be in agreement with the experimentally measured values for a saltfree POPC bilayer in the range of 63-68 Å 2 of the liquid crystalline phase (46,47), which is the most biologically relevant phase (48).
The system of the CB1-G i complex model in a fully hydrated POPC bilayer environment was carefully equilibrated using a sequence of steps. In Step 1, to relax unfavorable steric conflict in the side chains at the interface of the CB1 receptor and the G protein, the system of the CB1-G i complex was initially subjected to an energy minimization of 2,500 iterations and followed by a simulation at 310 K for a duration of 6 ns only for the regions of His-219 IC2 -Thr-229 IC2 and Thr-313 IC3 -Asp-333 IC3 of the CB1 receptor and the G␣ i C-terminal residues Ile-344 G␣i -Phe-354 G␣i , whereas the protein backbone atoms of the rest of the proteins were constrained. The G␥ C-terminal residues Phe-66 G␥ -Cys-68 G␥ were also allowed to move freely. In Step 2, we removed the backbone constraints for the CB1 receptor but retained the backbone constraints for the G protein. We newly added a group of distance and torsion constraints to approximately maintain some key intermolecular interactions at the receptor-G␣ i C-terminal interface, as determined in the x-ray structure of the ␤ 2 AR-G s complex (12), which appeared commonly applicable to the CB1-G complex. These constrained distances between mostly conserved residues of the CB1 receptor and G␣ i (Fig. 1B) included Arg-307 5.71 -Asp-341 G␣i , Lys-225 IC2 -Asn-347 G␣i , Arg-214 3.50 -Cys-351 G␣i (backbone), and Thr-344 6.36 -Gly-352 G␣i (backbone). We applied additional intramolecular distance constraints: Arg-214 3.50 -Tyr-294 5.58 , Tyr-224 IC2 -Asp-213 3.49 , and Ser-152 2.39 -Asp-213 3.49 for the CB1 receptor and Glu-318 G␣i -Lys-345 G␣i , Glu-28 G␣i -Arg-32 G␣i , and Asn-347 G␣i -Asp-350 G␣i for the G protein. We also applied torsional constraints for Ile-344 G␣i -Phe-354 G␣i to maintain the backbone conformation as found in the x-ray structure of the ␤ 2 AR-G s complex (12). The system was simulated for 25 ns. In Step 3, we removed the backbone constraints for the G protein but retained the distances and torsional constraints for an additional 52 ns of the simulation, during which the initially applied constraints (k ϭ 40) were gradually released. In Step 4, we removed all of the constraints and continued the simulation for an additional 824 ns. In this final step, the entire system was free to equilibrate, and the lateral area of the box was kept fixed while the z dimension of the box was allowed to move freely in the NPT ensemble. The structures taken every 400 ps of the simulation were used for the analysis.
To validate the present model of the CB1-G i complex, a second model with slightly changed orientation of G i relative to CB1 was constructed using the CB1-G i complex in the initial stage of the 824 ns simulation by the protein-protein docking program ZDOCK (version 3.0.2) (49). To locate the extreme C-terminal region of G␣ in close proximity to the IC region of the CB1 receptor, the G␣ residues Lys-192 G␣i -Leu-194 G␣i , Ser-293 G␣i -Glu-318 G␣i , and Asp-341 G␣i -Phe-354 G␣i and the CB1 residues Asp-213 3.49 -Lys-226 IC2 , Ile-297 5.61 -Leu-345 6.37 , and Arg-400 7.56 -Asp-403 H8 at the CB1/G␣ interface were used as the contact residues. In this model, GDP was removed from the nucleotide binding site to obtain an intermediate state, where the CB1-G i interactions are expected to be highly dynamic (50). This model was simulated for 300 ns.
Simulations-All simulations were performed using the NAMD simulation package (version 2.7 Linux-x86_64) (51), initially using CHARMM22 force field parameters for the proteins (52) and CHARMM27 force field parameters for the lipids (53) and later switching to CHARMM36 force field parameters for proteins with the / angle cross-term map correction (52,54,55) and lipids (56) and the TIP3P water model (57). The topology definitions and the parameters for the palmitoylated Cys, including the parameters around the bond connecting the Cys residues and the carbonyl carbon of PAL, as used in the literature (58), were found in the NAMD Parameter Topology Repository site. To describe geranylgeranylated Cys-68 G␥ of G␥ in the MD simulations using the CHARMM force field, we determined missing parameters, including the angle parameter for S-CH 2 -CH(ϭC) and the torsional parameters for S-CH 2 -C-H(ϭC) using MeSCH 2 CHϭC(Me) 2 obtained by ab initio MP2/ 6 -31G* level calculations. The temperature was maintained at 310 K through the use of Langevin dynamics (59) with a damping coefficient of 1/ps. The pressure was maintained at 1 atm by using the Nosé-Hoover method (60) with the modifications as described in the NAMD user's guide. The van der Waals interactions were switched at 10 Å and zero smoothly at 12 Å. Electrostatic interactions were treated using the particle mesh Ewald method (61). A pair list for calculating the van der Waals and electrostatic interactions was set to 13.5 Å and updated every 10 steps. A multiple time-stepping integration scheme, the impulse-based Verlet-I reversible reference system propagation algorithm method (62), was used to efficiently compute full electrostatics. The time step size for integration of each step of the simulation was 1 fs. Fig. 1A shows the resulting model of the CB1-G i complex at the end of an 824-ns simulation. To measure the structural stability of the present model of the CB1-G i complex, the RMSDs have been calculated over the performed simulation. As shown in Fig. 2A, the RMSD values of the CB1-G i complex are moderately increased for the initial ϳ250 ns of the simulation, due to the positional adjustment of the individual proteins to relieve unfavorable steric contacts present as a result of imposing a structure similar to the ␤ 2 AR-G s complex at the beginning of the simulation. Increased RMSD values at the end of the simulation relative to those at the start of the simulation indicate noticeable changes in orientation of the individual proteins. Little change in the RMSD values after 725 ns of the simulation, however, suggests that the protein-protein interactions in the CB1-G i complex become optimized at the interface. After 400 ns of the simulation, the RMSD values of G␤ are somewhat increased due to the displacement of its N-terminal helix ␣ N resulting from the flexibility of the segment (Cys-25 G␤ -Thr-29 G␤ ) that links helix ␣ N to helix ␣ 2 . At around 600 ns of the simulation, the RMSD values of the CB1-G i complex are noticeably increased, mainly due to the increased RMSD values of  FIGURE 1. A, simulation system of the CB1-G i complex (the CB1 receptor in red, G␣ in green, G␤ in blue, and G␥ in orange). PALs, which are covalently bonded to Cys-415 of the CB1 receptor and Cys-3 G␣i of G␣ i , and the geranylgeranyl moiety (GER), which is covalently bonded to Cys-68 G␥ of G␥, are represented by green sticks. Lipids and water molecules are represented as lines, whereas ions (Na ϩ in yellow and Cl Ϫ in cyan) are represented as spheres. Lipid and water hydrogen atoms are omitted for clarity. The system at 824 ns of the simulation is shown. B, sequence alignment by T-COFFEE (95) of CB1, ␤ 2 AR, and rhodopsin at the G protein interface (i) and of human G␣ i1 , G␣ s , and G␣ t (ii). The TM helical boundaries for the CB1 receptor (light red) are from the present study, whereas the TM helical boundaries for ␤ 2 AR and rhodopsin are from the respective x-ray structures (63,89). Highly conserved residues in the rhodopsin family of GPCRs reported by Baldwin et al. (82) are in boldface type. The helical structural information of G␣ i1 , G␣ s , and G␣ t is from the present study, the x-ray structure of G␣ s (Protein Data Bank code 3SN6) is from Ref. 12, and the x-ray structure of G␣ t (Protein Data Bank code 1GOT) is from Ref. 96. The color code for helices is: CB1 (light red), ␤ 2 AR (cyan), rhodopsin (gray), G␣ i1 (green), G␣ s (yellow); and G␣ t (light gray). Conservation of the aligned sequence by ClustalW (97) is represented by consensus symbols: *, identical residues; :, conserved substitutions; and dot, semiconserved substitutions. G␣ i . It appears that the RMSD values of G␣ i are closely associated with the tilt angle change in helix ␣ 5 of G␣ (Fig. 2B). This suggests that the conformation of helix ␣ 5 influences the orientation of the CB1 receptor relative to G␣ i and determines the degree of the interaction with the receptor (see "Tilt in Helix ␣ 5 of G␣ i "). High RMSD values for HU210 (Ͼ2 Å) are caused by the rotameric angle change in the ligand's C3 side chain which led to its end carbon being positioned between Val-204 3.40 and Val-282 5.46 (see under "Molecular Features of the Active CB1 Receptor in the CB1-G i Complex"), whereas low RMSD values for GDP (Ͻ1 Å) are due to little change in GDP tightly bound to the nucleotide-binding pocket.

Structural Convergence of the CB1-G i Complex-
The second model became structurally stable after 200 ns of the simulation, as indicated by the RMSD values maintained at ϳ4 Å without any noticeable change. As shown in Fig. 3A, the initial orientation of G i relative to CB1 in the second model is different from that of the present model of the CB1-G i complex. However, the final orientation of G i in the second model becomes quite similar to that in the present model of the CB1-G i complex, supporting the validity of the orientation of G i relative to CB1 in the present model of the CB1-G i complex. It appears that the initial orientation of G i in the second model is better than that in the present CB1-G i complex in achieving the converged orientation.
Comparison of the CB1-G i Complex with the Known X-ray Structures-To evaluate the present model of the CB1-G i assembly, we compared it with the x-ray structure of the ␤ 2 AR-G s complex (12) (Fig. 3B). The overall orientation of G␣Ras relative to the receptor in the CB1-G i complex is maintained as in the ␤ 2 AR-G s complex. Interestingly, the position of the extreme C-terminal helix ␣ 5 (Thr-329 G␣i -Phe-354 G␣i ) of G␣ in the CB1-G i complex is tilted away from TM6 compared with the corresponding helix of G s (Fig. 3B). We also compared the present model of the CB1-G i complex with the x-ray structure of the activated rhodopsin (metarhodopsin II) in complex with the C-terminal peptide of G t (metarhodopsin II-G t CT) (Protein Data Bank code 3PQR) (63). The position of the extreme C-terminal helix ␣ 5 of G␣ in the CB1-G i complex is remarkably similar to that of the corresponding G t CT peptide (Ile-340 G␣t -Phe-350 G␣t ) (Fig. 3C), suggesting that the orientation of the G␣ subunit relative to the receptor in the CB1-G i complex is similar to that of the corresponding G t subunit in the metarhodopsin II-G t complex. The outward movement of TM6 is small in the CB1-G i complex and in the metarhodopsin II-G t CT complex (Fig. 3C), compared with the ␤ 2 AR-G s complex (Fig. 3B), suggesting that the position of the extreme C-terminal helix ␣ 5 of G␣ is sensitive to the receptor conformation (see "Tilt in Helix ␣ 5 of G␣ i "). Further comparisons with the x-ray structures of G␣ proteins reveal that helix ␣ 5 of G␣ in the CB1-G i complex is aligned slightly better with helix ␣ 5 in the GTP␥S-bound active state of G␣ s (64) than with helix ␣ 5 in the nucleotide-free ␤ 2 AR-G s complex, suggesting that the conformation of helix ␣ 5 is associated with nucleotide binding to G␣Ras (12). Interestingly, a close examination reveals that helix ␣ 5 in the GTP␥S-bound active state G␣ s (64) is tilted in the middle of the helix at Asp-378 G␣s and Cys-379 G␣s , the residues equivalent to Ala-338 G␣i and Val-339 G␣i of G␣ i tilted in the present CB1-G i complex at the middle stage of the simulation (Fig. 2B).
Contact Number Analysis of the CB1-G i Complex-As shown in Fig. 2C, the number of contacts of the CB1 receptor to G i is in suggesting key roles of IC2 and IC3 of the receptor in G i coupling. Similarly, the number of contacts of G i to the CB1 receptor is mainly with G␣ i (ϳ84), low with G␤ (ϳ18), and none with G␥. The number of contacts between the receptor and G i is temporarily increased for the initial ϳ250 ns of the simulation and then decreased to a value slightly higher than the initial value. Temporal increases in the contact number between the CB1 receptor and G i at the early stage (100 -250 ns) of the simulation (Fig. 2C), which are closely related to the observed temporal increases in RMSD values of the CB1-G i complex ( Fig.  2A), are mainly attributed to the temporarily increased contacts between the CB1 receptor and G␤ (Fig. 2C). Because the main candidate regions of the receptor in contact with G␤ would be IC1 and H8, it appears that IC1 and H8 try to maximize the local contacts to G␤ in response to the docked G protein at the early stage of the simulation, but these contacts become few as the   3. A, the orientation of G i (green surface) relative to CB1 (red ribbon) in the present CB1-G i complex in comparison with the orientation of G i (magenta surface) relative to CB1 (pink ribbon) in the second model, viewed from the extracellular side, at the initial and final stages of the simulations. We superimposed these complex models with respect to the backbone C␣ atoms of the TM helical residues of CB1. The extracellular half of CB1 and the G␤␥ subunits of G i are omitted for clarity. B, comparisons of the present CB1-G i complex structure (CB1 in red, G␣ in green, G␤ in blue, and G␥ in orange) with the x-ray structure of the ␤ 2 AR-G s complex (12) (␤ 2 AR in cyan and G s in yellow). C, comparisons of the present CB1-G i complex structure complex (CB1 in red, G␣ in green, G␤ in blue, and G␥ in orange) with the x-ray structure of metarhodopsin in complex with G t CT (63) (metarhodopsin in gray and G t CT in light gray). In B and C, we superimposed these structures with respect to the backbone C␣ atoms of the TM helical residues of the receptors whenever the receptor was available. The structures on the right are the receptor-G protein interfaces marked by the blue dotted area viewed from the IC side by rotating 90°about the x axis. D, a sequence of conformational changes in helix ␣ 5 of G␣ i influenced directly by IC2 and indirectly by IC3 of the CB1 receptor during the present simulation of the CB1-G i complex. Solid arrows indicate the conformational changes of IC2 that affect the conformation of the C-terminal half of helix ␣ 5 and helix ␣ N of G␣ i , whereas dotted arrows indicate the conformational changes of IC3 that affect the conformation of the C-terminal half of helix ␣ 5 and helix ␣ 4 of G␣ i . Color coding for snapshots of CB1-G i at different times of the simulation is: 50 ns (blue), 400 ns (red), 600 ns (orange), and 800 ns (green). E, superposition of the present model of the inactive state of the CB1-G i complex (CB1 in red and G␣ in green) on the active state of the ␤ 2 AR-G s complex (12) (␤ 2 AR in cyan and G s in yellow). The cyan arrows indicate the conformational changes in the CB1 receptor required for achieving the active state of CB1-G i , as observed in the ␤ 2 AR-G s complex, by rotating G␣ i Ras as indicated by the yellow arrows. TM1, TM2, and TM7 of the receptors are omitted for clarity.
interactions between IC2 and IC3 of the receptor and G␣ i are fully established after 300 ns of the simulation. The number of contacts between IC2 and IC3 of the receptor and G i is little changed, suggesting that the interactions between IC2 and IC3 of the receptor and G i are important for the CB1-G i complex. A transient decrease in the contact number between the CB1 receptor and G␣ i right after 600 ns of the simulation resulted from the formation of an IC3 helical segment (Glu-323 IC3 -Ala-335 IC3 ) that moves toward the IC membrane surface (Fig. 3).
Molecular Features of the Active CB1 Receptor in the CB1-G i Complex-To confirm the active state of the CB1 receptor in the CB1-G i complex, we monitored the breaking of the ionic lock Arg-214 3.50 /Asp-338 6.30 along with inward movements of Tyr-294 5.58 and Tyr-397 7.53 (Fig. 4A), key features of the GPCR active state (13,(65)(66)(67)(68). The ionic lock Arg-214 3.50 /Asp-338 6.30 stays broken (C␣-C␣ distance, ϳ15 Å) throughout the simulation, similarly as those in the x-ray structures of metarhodopsin II (63) and ␤ 2 AR (12). Whereas Arg 3.50 is conserved in ␤ 2 AR as part of the DRY motif, the residue corresponding to Lys-345 G␣i of G␣ i is Arg-385 G␣s in G␣ s . It is shown that Arg-214 3.50 released from Asp-338 6.30 is confined to the region created by Thr-210 3.36 and Tyr-294 5.58 of the CB1 receptor and Cys-351 G␣i and Leu-353 G␣i of G␣ i . Arg-214 3.50 maintains a tight interaction with Tyr-294 5.58 through a hydrogen bond, as indicated by the distance between the side chain nitrogen of Arg-214 3.50 and the side chain oxygen of Tyr-294 5.58 (Ͻ3 Å) (Fig.  4A). In contrast, Arg-214 3.50 forms a loose hydrogen bond with Tyr-397 7.53 , as indicated by fluctuations in the distance between Arg-214 3.50 and Tyr-397 7.53 (Fig. 4A). A close examination reveals that when Tyr-397 7.53 stays away from Arg-214 3.50 , it is deeply inserted into the TM core and appears to play a key role in establishing an extensive water-mediated hydrogen-bonding network in the IC half of the TM core formed by TM2/TM3/TM7 residues mainly of the (S/N)LAXAD and NPXXY motifs. Assuming that the deeply inserted Tyr-397 7.53 represents the receptor in a fully activated state, as proposed in a recent study (68), Tyr-397 7.53 appears to be an important molecular switch for the formation of a water channel that possibly stabilizes the active state of the receptor (69).
Substitution of Thr-210 3.46 of the CB1 receptor with Ile and Ala generates constitutively active and inactive receptor mutants, respectively (25). A recent study proposed that mutations of Thr-210 3.46 greatly alter the pattern of the salt bridges at the IC side of the receptor and the TM helical packing, leading to stabilization or destabilization of the wild type receptor activity (70). Thr-210 3.46 in the present model of the CB1-G i complex appears to play an important role in stabilizing Arg-  by Tyr-294 5.58 in the present receptor model. Thus, it appears that the bulky Ile in place of Thr-210 3.46 causes an unfavorable interaction that facilitates an outward movement of the IC end of TM6, leading to the breakage of the ionic lock.
Mutation of the highly conserved Asp-163 2.50 of the (S/N)LAXAD motif in the CB1 receptor to an Asn or a Glu exhibited significant disruption of G protein coupling (72,73). In the present study, these highly conserved residues, Asp-163 2.50 and Asn-393 7.49 , appear to play key roles in the formation of the water channel, possibly stabilizing the active state of the receptor (69). The distance between the side chain oxygen of Asp-163 2.50 and the side chain nitrogen of Asn-393 7.49 is Ͻ3 Å occasionally but ϳ6 Å most times during the simulation. It is possible that individual residues are important for the formation of the water channel, but a tight interaction between these residues is not required for G protein coupling. In support, Roche et al. (73) showed that the D163 2.50 N/N393 7.49 D mutation behaved similarly to the D163 2.50 N mutant, indicating that Asp-163 2.50 and Asn-393 7.49 do not interact cooperatively for G protein coupling. Interestingly, the corresponding Asp-79 2.50 and Asn-322 7.49 of ␤ 2 AR form a hydrogen bond according to the x-ray structure of the ␤ 2 AR-G s complex (12).
Throughout the simulation, the 1 rotameric angle (the rotation around the C␣-C␤ bond) of Trp-356 6.48 remained as trans (Fig. 4B), as was proposed to be preferred for the active state GPCRs (74,75). It is interesting to see that the 1 angle (the rotation around the C3-C1Ј bond) in the C3 alkyl chain of the ligand changes from trans to gauche at 100 ns of the simulation (Fig. 4B), causing increases in the RMSDs. This only slightly affects the 2 rotameric angle (the rotation around the C␤-C␥ bond) in Trp-356 6.48 but moves HU210 deep into the TM core, leading the C3 alkyl chain moving deep inside the core toward TM5. A group of water molecules, which stabilizes the extracellular half of the TM3/TM5/TM6 interhelical region around the Trp-356 6.48 toggle switch in the early stage of the activated state (31), is now completely eliminated, possibly due to the replacement of the water molecules by the ligand C3 alkyl chain. Its driving force is interaction with a hydrophobic pocket created by Val-204 3.40 on TM3 and Val-282 5.46 and Leu-286 5.50 on TM5, as demonstrated by the close distances between the C3 alkyl chain of the ligand and Val-204 3.40 and Val-282 5.46 (Fig.  4C). The resulting hydrophobic interaction between the C3 alkyl chain of the ligand and TM5 appears to be important for transferring the molecular signal of the ligand. As shown in Fig.  4C, occupying a similar region in the TM core as HU210, the agonist BI-167107 interacts tightly through hydrogen bonding with a group of Ser residues on TM5 of ␤ 2 AR, which induces a bulge in the middle of TM5. In contrast, hydrophobic interactions between HU210 and Val-282 5.46 and Leu-286 5.50 on TM5 of the CB1 receptor appear not strong enough to induce such a bulge. It should be noted that the highly conserved Pro 5.50 in many GPCRs, equivalent to Leu-286 5.50 in the CB1 receptor, has been proposed to play a key role in receptor activation (76).
Structural Properties of the CB1-G i Complex-The current model of the CB1-G i complex reveals that the N-and C-terminal regions, including helix ␣ N , loop ␣ N -␤ 1 , and helix ␣ 5 , and other regions, including loop ␤ 2 -␤ 3 , helix ␣ 4 , loop ␣ 4 -␤ 6 , and strand ␤ 6 , of the G␣ subunit, form major contacts with the receptor. Thus, these segments appear to be crucial for transferring the molecular signal from the activated receptor to the nucleotide-binding pocket. In support, most of these segments have been demonstrated to be important for GPCR-G protein interactions leading to GDP release (6,(77)(78)(79). Because our simulation analyses indicate that IC2 and IC3 play key roles in G i coupling, detailed structural analyses of these regions in contact with G i are described as follows. According to the present model of the CB1-G i complex, IC2 of the CB1 receptor, which spans 11 residues (His-219 IC2 -Thr-229 IC2 ) and contains one helical segment (Ala-223 IC2 -Ile-227 IC2 ), forms one of the major contacts to G i (Fig. 2C). A hydrophobic patch, formed by Pro-221 IC2 , Leu-222 IC2 , and the aliphatic hydrocarbon portions of Arg-220 IC2 and Lys-225 IC2 , is positioned toward G␣ i , providing critical contacts mainly to the N-and C-terminal segments, including helix ␣ N (Arg-32 G␣i ), loop ␤ 2 -␤ 3 (Lys-192 G␣i , Asp-193 G␣i , and Leu-194 G␣i ), and helix ␣ 5 (Thr-340 G␣i , Ile-343 G␣i , and Ile-344 G␣i ). Leu-222IC2, the last residue of a consensus DRYXX(V/I)XXPL motif in many GPCRs, which has been proposed to be crucial for G protein coupling (80), is at the center of the hydrophobic patch. It has been reported that L222 IC2 A or L222 IC2 P mutations in the CB1 receptor exhibited coupling preference to G i over G s , whereas the L222 IC2 F mutation exhibited coupling preference to G s over G i (81). Notably, Ile-218 3.54 forms a tight hydrophobic cluster, not only with Ile-297 5.61 , Lys-300 5.64 , and Leu-341 6.33 of the receptor but also with Ile-344 G␣i , Leu-348 G␣i , and Leu-353 G␣i of G␣ i (Fig. 5A). Considering that both Ile-218 3.54 and Ile-297 5.61 are highly conserved residues in the rhodopsin family of GPCRs (82) and that the corresponding residues in all subtypes of G␣ proteins to Leu-348 G␣i and Leu-353 G␣i are highly conserved (Fig. 1B), it appears that Ile-218 3.54 plays a crucial role in receptor-G protein coupling. Of interest, Tyr-224 IC2 forms two hydrogen bonds to Asp-350 G␣i , which appear to be tight, as indicated by the close distances maintained during the MD simulation (Fig. 4D).
According to the present model of the CB1-G i complex, IC3 forms the largest contact with G i (Fig. 2C). As TM5 and TM6 of the CB1 receptor are extended to include a few N-and C-terminal IC3 residues, the extent of IC3 of the CB1 receptor is reduced to the sequence spanning 21 residues (Ile-309 IC3 -Arg-336 IC3 ), among which the C-terminal residues Glu-323 IC3 -Ala-335 IC3 form a helical segment near the start of TM6. The helical content of IC3 of the CB1 receptor in the present model is similar to that of the IC3 peptide determined by NMR (37), although the NMR-determined helix (Ile-309 5.72 -Ser-316 IC3 ) right next to the N-terminal helix is not seen in the present IC3 structure, possibly because of the presence of the G protein which affects the IC3 conformation. Extensive intramolecular hydrophobic interactions within IC3 lead to two distinct hydrophobic clusters, including the first one near the TM region (Ile-297 5.61 , Leu-298 5.62 , Ala-301 5.65 , Leu-341 6.33 , Ala-342 6.34 , Leu-345 6.37 , Val-346 6.38 , and Leu-349 6.41 ) and the second one in the middle of IC3 (Ala-305 5.69 , Val-306 5.70 , Ile-308 5.72 , Ile-309 IC3 , Arg-331 IC3 , Pro-332 IC3 , Ala-335 IC3 , and Ile-339 6.31 ). In particular, the highly conserved Ile-297 5.61 , part of the first hydrophobic cluster, is in close proximity not only to Arg-214 3.50 of the DRY motif, Ile-218 3.54 , and Leu-345 6.37 but also to Leu-353 G␣i , suggesting an important role of Ile-2975.61 in receptor-G cou-pling (Fig. 5A). Similar interactions by the corresponding Val-222 5.61 in ␤ 2 AR are shown in the x-ray structure of the ␤ 2 AR-G s complex (12). Arg-311 IC3 forms two hydrogen bonds to Gln-304 G␣i of (Fig. 4D). Compared with the ionic lock residue Glu-247 6.30 of rhodopsin, which is released from Arg-135 3.50 and forms an intramolecular salt bridge to Lys-231 5.66 (63), the corresponding ionic lock residue of the CB1 receptor Asp-338 6.30 is released from Arg-214 3.50 (Fig. 4A) and forms two intermolecular salt bridges to Lys-345 G␣i of G␣ i (Fig. 5B). These interactions appear to be tight, as indicated by the close distances maintained during the MD simulation (Fig. 4E). The BBXB or BBXXB motif, where B represents a basic residue and X represents a non-basic residue, at the IC3 C terminus of the G i -coupled receptors has been proposed to be necessary for G protein activation (83). Arg-340 6.32 , the second residue of the LRLAK sequence in the CB1 receptor, positioned toward the receptor-G␣ i interface, secures the G␣ i C-terminal carboxylate of Phe-354 G␣i (Figs. 4E and 5B), suggesting that a basic residue at this position is necessary for tight interactions with G␣ i . It is quite interesting to see that Thr-344 6.36 and Arg-400 7.56 , located near the IC membrane surface, appear to help securely lock the extreme C-terminal carboxylate of Phe-354 of G␣ i (Figs. 4E and 5B). Arg-340 6.32 and Arg-400 7.56 of the CB1 receptor each form two hydrogen bonds to the C-terminal carboxylate of Phe-354 G␣i , whereas Thr-3446.36 forms a hydrogen bond with the backbone carbonyl oxygen atom of Leu-353 G␣i . The side chain alkyl moiety of Thr-3446.36 also forms close contacts with the side chain alkyl moieties of Arg-214 3.50 , Leu-341 6.33 , Leu-345 6.37 , Ile-348 6.40 , and Arg-400 7.56 of the receptor and Leu-353 G␣i (Fig. 5C). The importance of Arg-400 7.56 of the CB1 receptor in G protein coupling has been tested by the H8 (Arg-400 to Glu-416) peptide (84). The removal of Arg-400 7.56 from the peptide resulted in a 6-fold reduction in G i affinity and an almost complete loss of the inhibitory effect on adenylyl cyclase activity.
Tilt in Helix ␣ 5 of G␣ i -According to the present MD simulation of the CB1-G i complex, the conformation of helix ␣ 5 of G␣ i was altered as follows: a straight form was observed in the Only the side chains without hydrogen atoms of these residues (in stick representations) are represented for clarity. C, hydrophobic interactions of Thr-344 6.36 with Arg-214 3.50 , Leu-341 6.33 , Leu-345 6.37 , Ile-348 6.40 , and Arg-400 7.56 of the receptor and hydrogen-bonding interaction with the backbone carbonyl oxygen atom Leu-353 G␣i . D, interactions of Ile-218 3.54 , Tyr-224 IC2 , Asp-338 6.30 , Arg-340 6.32 , Leu-341 6.33 , and Thr-344 6.36 of the CB1 receptor with the extreme C terminus of G␣ i at the end of helix ␣ 5 , viewed from the start of the helix ␣ 5 to the receptor IC face. We have examined these receptor residues by site-directed mutagenesis studies. The receptor is shown in a red volume, whereas the G␣ i is shown in a green schematic. The extreme C-terminal helix ␣ 5 (Thr-329 G␣i -Phe-354 G␣i ) of G␣ in the CB1-G i complex is also shown in a helical wheel representation created by the helical wheel plotting program (available on the RZ Lab Web site), where hydrophobic residues are shown as diamonds, potentially negatively charged residues as triangles, and potentially positively charged residues as pentagons. Color coding for residues is as follows: from green for the most hydrophobic residue to yellow for the least hydrophobic residue and red for the hydrophilic residues.
initial stage (ϳ100 ns) of the simulation; a tilted form with the tilt angle ϳ20°during the next 500 ns of the simulation; a transient stage at ϳ600 ns of the simulation where the tilt angle of helix ␣ 5 of G␣ i reached its largest value (ϳ30°); and a straight form in the late stage of the simulation (Fig. 2B). To understand what causes these conformational changes in helix ␣ 5 of G␣ i during the simulation, we superimposed the snapshots at 50 ns (i.e. straight helix ␣ 5 ), 400 ns (i.e. tilted helix ␣ 5 ), 600 ns (i.e. highly tilted helix ␣ 5 ), and 800 ns (i.e. straight helix ␣ 5 ) of the simulation with respect to the TM helices of the receptor (Fig.  3D). The snapshot at 400 ns in comparison with that at 50 ns reveals that IC2 of the receptor alters its conformation and moves close to helix ␣ 5 of G␣ i . This causes the C-terminal half of helix ␣ 5 and the N terminus of G␣ i to move away from IC2, whereas the N-terminal half of helix ␣ 5 moves very little due to the minute effect the IC3 conformation has on G␣ i (helix ␣ 4 ). As a result, the C-terminal half of helix ␣ 5 becomes tilted toward TM6 (Fig. 3D).
The snapshot at 600 ns in comparison with the snapshot at 400 ns reveals that IC2 moves closer to helix ␣ 5 of G␣ i , which causes the C-terminal half of helix ␣ 5 and the N terminus of G␣ i to move further away from IC2, whereas an altered IC3 conformation pushes G␣ i (helix ␣ 4 ) away. This causes the N-terminal half of helix ␣ 5 to move toward IC2 (Fig. 3D). As a result, the tilt angle of helix ␣ 5 of G␣ i becomes larger (ϳ30°). At 600 ns of the simulation, IC2 moves slightly away from helix ␣ 5 of G␣ i , which causes the C-terminal half of helix ␣ 5 and the N terminus of G␣ i to move toward IC2. In contrast, IC3 undergoes a significant conformational change in which the C-terminal helical segment (Glu-323 IC3 -Ala-335 IC3 ) moves close to the IC membrane surface. This allows G␣ i (helix ␣ 4 ) to be relaxed and the N-terminal half of helix ␣ 5 of G␣ i to move away from IC2. As a result, helix ␣ 5 of G␣ i becomes straight (Fig. 3D). Overall, the position of helix ␣ 5 of G␣ i is sensitive to the conformations of IC2 and IC3 of the CB1 receptor. It appears that IC2 of the CB1 receptor directly affects both the C-terminal side of helix ␣ 5 and the N terminus of G␣ i , whereas IC3 of the CB1 receptor indirectly affects the N-terminal side of helix ␣ 5 through interactions with helix ␣ 4 of G␣ i .  6.36 A of the CB1 Receptor-The results from the simulation of the CB1-G i complex were examined by site-directed mutagenesis studies. We selected a group of residues predicted to be crucial in receptor-G protein coupling. I218 3.54 , P221 IC2 , Y224 IC2 , D338 6.30 , R340 6.32 , L341 6.32 , and T344 6.36 of the CB1 receptor were of particular interest due to their tight interactions with the extreme C-terminal helix ␣ 5 , which forms the key contact surface for the activated receptor (85). We also selected a couple of residues, including His-219 IC2 and Ile-339 6.31 , predicted to be little involved in receptor-G protein coupling. Thus, we assessed the impact of these mutations on ligand binding properties by saturation binding experiments using the CB1 agonist [ 3 H]CP55940 and the inverse agonist [ 3 H]SR141716A. As shown in Table 1  H]SR141716A. These ligand binding data (with the exception of Y224 IC2 A mutant) suggest that all of the mutant receptors have levels of expression comparable with that of the wild-type and retain the ligand binding affinity of the wild-type CB1. To directly evaluate the importance of these nine residues in G protein coupling, [ 35 S]GTP␥S binding assays were performed for the I218 3.54 A, H219 IC2 A, P221 IC2 A, Y224 IC2 A, D338 6.30 A, I339 6.31 A, R340 6.32 A, L341 6.32 A, and T344 6.36 A receptors. This assay monitors the level of G protein coupling activity by determining the extent of binding of the nonhydrolyzable

Impact of CB1 amino acid mutation on ligand binding and G protein coupling
Data are the mean Ϯ S.E., except for EC 50 values, which are the median and corresponding 95% confidence limits, from at least three independent experiments performed in duplicate. ND, no specific binding detected. Asterisks indicate statistically significant differences in the GTP␥S binding parameters (basal and E max levels) from wild type (p Ͻ 0.05) using analysis of variance followed by Bonferroni's post hoc test: *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. No significant differences were observed in ligand binding parameters of any of the receptors tested, except for the Y224 IC2 A receptor, from wild type.

DISCUSSION
We experimentally evaluated the importance of key residues for G protein interaction. Our mutational study of the I218 3.54 A, H219 IC2 A, D338 6.30 A, R340 6.32 A, L341 6.33 A, and T344 6.36 A receptors indicates that these residues play key roles in CB1-mediated G protein coupling (Table 1 and Fig. 6), consistent with the present model of the CB1-G i complex. Tyr-224 IC2 may also be involved in G protein coupling, but because the Y224 IC2 A receptor did not bind the agonist CP55940, we cannot discriminate between a trafficking problem and a ligand binding defect. All of these receptor residues except for His-219 IC2 form an interface with residues that are on different sides of the extreme C-terminal helix ␣ 5 (Thr-329 G␣i -Phe-354 G␣i ) (Fig. 5D). Considering that the extreme C-terminal helix ␣ 5 of G␣ i forms the key contact surface for the activated receptor (85), Ala mutation of these residues in the receptor appears to eliminate crucial contacts with G␣ and thereby greatly impair G protein coupling activity. That is what we observed in the present study (Table 1 and Fig. 6). Given that a shift in binding affinity for CP55940 (an agonist) or SR141716A (an inverse agonist) is expected if G protein coupling is altered, we are surprised to see that these mutant receptors bind the ligands like the wild-type receptor despite significantly disrupted G i coupling activity (Table 1 and Fig. 6). Because these mutant receptors retain the wild-type ligand binding profile, it is not likely that they disrupt the receptor conformation such that the ligand-binding pocket is unfavorably modified. Instead, as in the wild-type receptor, the agonist-bound mutant receptors show the characteristic features of the activated receptor, including the breaking of the ionic lock and the outward movement of the IC end of TM6. Because the CB1 receptor binds G i through multiple contacts not only with the C-terminal segment but also with other parts (helix ␣ N , loop ␤ 1 -␣ 1 (P-loop), helix ␣ 1 , and strand ␤ 6 ), the activated mutant receptors can still bind some G i but much less due to the loss of a few key contacts for helix ␣ 5 . We cannot rule out the possibility that the mutant receptors bind G s preferentially because its activity is not adequately detected by the [ 35 S]GTP␥S assay, which is much more sensitive to G i than G s . However, according to the present CB1-G i complex model, Arg-385 G␣s , Leu-388 G␣s , and Leu-393 G␣s of G s , the corresponding residues that interact with Asp-338 6.30 , Arg-340 6.32 , Leu-341 6.33 , and Thr-344 6.36 , are highly conserved, suggesting that these mutant CB1 receptors may impair G protein coupling equally in G i and G s .
The finding that His-219 IC2 also plays a key role in CB1mediated G protein coupling (Table 1 and Fig. 6) is unexpected because this residue maintains little contact with G i in the present model of the CB1-G i complex. A close examination, however, reveals that H219I IC2 at the junction of TM3 and IC2 forms a hydrophobic cluster with Ile-216 3.52 , Ala-223 IC2 , and the aliphatic hydrocarbon portion of Arg-220 IC2 , suggesting its indirect role in CB1-mediated G protein coupling. Our observation that the P221 IC2 A receptor does not affect CB1-mediated G protein coupling to a large extent (Table 1 and Fig. 6) is also unexpected because this residue forms close contacts with Leu-194 G␣i , Thr-340 G␣i , Ile-343 G␣i , and Asn-347 G␣i on the C-terminal helix ␣ 5 based on the present model of the CB1-G i complex. This observation is also supported by the x-ray structure of the ␤ 2 AR-G s complex (12), wherein the equivalent Pro-138 IC2 of ␤ 2 AR forms close contacts with Ile-383 G␣s and Gln-384 G␣s on the C-terminal helix ␣ 5 . Thus, an Ala mutation of Pro-221 IC2 would be expected to greatly destabilize the helical conformation of IC2 of the receptor (88). This disrupts CB1mediated G protein coupling as a result of interference with the CB1-G i interaction. Given the conservation of this position shared by Pro and Ala residues in class A GPCRs, the Ala mutation of Pro-221 IC2 in the CB1 receptor appears to be a permissive mutation, retaining G protein coupling (Table 1). In support of this observation, the Ala mutation of the equivalent Pro-131 IC2 of the M1 muscarinic acetylcholine receptor was found to little impair G protein coupling (80). Given that IC2 of ␤ 2 AR forms an L-shaped non-helical conformation in the inactive state (89) but a helical conformation in the active state (12,13), it appears that the helical conformation of IC2 is maintained in the P221 IC2 A mutant receptor due to the propensity of IC2 to form a helix in the activated receptor upon association with G protein coupling.
Among the residues examined by the present mutagenesis studies, intermolecular interactions of Tyr-224 IC2 , Asp-338 6.30 , Arg-340 6.32 , and Thr-344 6.36 with helix ␣ 5 (Thr-329 G␣i -Phe-354 G␣i ) of G␣ i through hydrogen bond/charge are unexpectedly tight, as indicated by the close interresidual distances maintained during the simulation (Fig. 4, D and E), given that hydrogen bond/charge interactions are weakened in the presence of solvent (90). A detailed examination of the CB1-G i complex model reveals that most of these residues are partially buried by surrounding residues and that their accessibility to solvent is limited. Asp-338 6.30 is unique in that it forms the intramolecular ionic lock with Arg 3.50 . Thus, its Ala mutation would lead to the breaking of the ionic lock and thereby shifting the equilibrium toward the active state, suggesting that Asp-338 6.30 is involved in receptor activation. Interestingly, however, the D338 6.30 N mutation of the CB1 receptor resulted in a decrease in the maximum response of agonist-induced activation without affecting CP55940 binding affinity relative to the wild type (91). Consistent with this, we found that the D338 6.30 A mutation did not alter CP55940 binding affinity but caused an ϳ70% decrease in the E max value (Table 1). Thus, it is possible that the Asp-338 6.30 3 Ala mutation does not result in a fully active conformation of the receptor, because it cannot couple to G protein, supporting the present CB1-G i complex model. Among the residues examined by the present mutagenesis studies, Ile-218 3.54 , Leu-341 6.33 , and Thr-344 6.36 are unique in that they form tight hydrophobic interactions at the receptor-G␣ C-terminal interface (Fig. 5, A and C). Thus, it is possible that their Ala mutations would alter the receptor conformation, inducing a steric occlusion in the receptor-G␣ C-terminal binding interface rather than a direct contact. The significant decrease of coupling by I218 3.54 A and the complete loss of coupling by L341 6.33 A and T344 6.36 A, compared with other examined Ala mutations (Table 1 and Fig. 6), would support this idea.
The G protein cycle is thought to proceed via the following sequential events: (i) the GDP-bound heterotrimeric G␣␤␥ protein forms a complex with the activated receptor; (ii) the GDP release from the nucleotide-binding pocket leads to the nucleotide-free G␣␤␥-receptor complex; (iii) the GTP (or GTP␥S) is added to the "empty complex"; and (iv) the GTPbound G␣ is dissociated from the G␤␥ subunits and the receptor (6). Our finding that the position of the extreme C-terminal helix ␣ 5 of G␣ in the CB1-G i complex is remarkably similar to that of the corresponding G t CT peptide in the metarhodopsin II-G t CT complex (63) (Fig. 3C) suggests that our system of the CB1-G i complex is at the same stage as the metarhodopsin II-G t CT complex, which is in the initial stage of the complex formation between the activated receptor and the G protein.
Given that recent electron paramagnetic resonance spectroscopy (EPR) studies suggest that ϳ10 residues from the extreme C-terminal helix ␣ 5 of G␣ form the contact surface with the activated receptor and that a rigid body rotation/translation of helix ␣ 5 induced by the activated receptor is necessary for GDP release (85), it appears that helix ␣ 5 of G␣ plays a key role in the interaction of G␣ with the activated receptor. It is known that the extreme C-terminal helix ␣ 5 undergoes a conformational change upon the binding of the activated receptor (i.e. disordered when unbound to the receptor but highly ordered when bound to the receptor) (85,92). It appears that the conformational changes in helix ␣ 5 govern the degree of the interaction of G␣ with the receptor, as shown in the RMSD values of G␣ in close association with the tilt angle change in helix ␣ 5 of G␣ (Fig. 2B). The results of the present simulation also show that the conformation of helix ␣ 5 of G␣ i changes dynamically in response to conformational changes in IC2 and IC3 of the CB1 receptor (Fig. 3D).
The position of the G t CT peptide in the metarhodopsin II-G t CT complex (63) is tilted by ϳ30°away from TM6, compared with the position of the corresponding helix ␣ 5 of G␣Ras in the nucleotide-free active state of G s (␤ 2 AR-G s ) (12). Based on these observations, Rasmussen et al. (12) has proposed that the initially bound C-terminal helix ␣ 5 of G␣Ras of the ␤ 2 AR-G s complex requires a rotation of G␣Ras ϳ90°along the axis parallel to the membrane normal. As a result, the N terminus of G␣Ras aligned with the G t CT peptide (i.e. inactive state) becomes overlapped with the N terminus of G␣Ras in the ␤ 2 AR-G s complex (12) (i.e. active state). Superposition of the present model of the inactive state G i (CB1-G i ) on the active state of G s (␤ 2 AR-G s ) (12) reveals that IC2 and TM6/IC3 of ␤ 2 AR move away from the middle and the C-terminal end of helix ␣ 5 , respectively, compared with the CB1 receptor (Fig.  3E). On the other hand, both helix ␣ N and helix ␣ 5 of G␣ i Ras are in the same orientation but with slightly different tilt angles in these complexes (Fig. 3E). Thus, it appears that in order for the CB1-G i complex to achieve the active state of G i (i.e. GDP release) a rotation along the axis perpendicular to the membrane side is required, as shown in Fig. 3E. Such rotation, somewhat different from the proposed rotation around the axis parallel to the membrane normal (12), would lead to GDP release as a result of the disruption of (i) loop ␤ 1 -␣ 1 (P-loop), a contact region for the purine ring moiety of GDP, attached to helix ␣ N of G␣Ras, and (ii) loop ␤ 6 -␣ 5 (TCAT motif), a contact region for the ␤-phosphate moiety of GDP, attached to helix ␣ 5 of G␣Ras. A similar mechanism for receptor-mediated GDP release has recently been proposed (79).
Homology models of the CB1 receptor using the known x-ray structures of GPCRs as templates have been justified by sequence alignment with these GPCRs according to the highly conserved (Ͼ90%) TM residues (82). In this study, under our working hypothesis that the conformational change in G␣Ras is small during G protein activation, compared with that in G␣AH as seen in the x-ray structure of the ␤ 2 AR-G s complex (12), we started with a model of the CB1-G i complex, where G i represents the GDP-bound inactive state, using the x-ray structure of the ␤ 2 AR-G s complex, where G s represents the nucleotide-free active state (12), the only available x-ray structure of the GPCR-G complex, as the docking template and determined the specific orientation of the two proteins (i.e. the CB1 receptor and G i ) through an extensive MD simulation. To keep the CB1 receptor and G i in a reasonably close distance for favorable complex formation and at the same time to minimize a biased complex conformation, we applied a minimum set of interresidual distances between the CB1 receptor and the G protein only at the receptor-G␣ i C-terminal interface while allowing the rest of the intermolecular interactions to be freely adjusted during the simulation. At the end of the simulation, most of the applied interresidual distances of the receptor-G␣ i C-terminal interface were lost. In fact, the resulting CB1-G i complex appears to resemble the metarhodopsin II-G t CT complex (63) (Fig. 3C), which is the inactive GDP-bound G protein at an early stage of complex formation, rather than the ␤ 2 AR-G s complex, which is at the activated "empty complex" stage. This result suggests that although we started with a structure similar to the known ␤ 2 AR-G s complex (12), we ended with a structure unique to the CB1-G i complex. As discussed above, the position of the extreme C-terminal helix ␣ 5 of G␣ in the CB1-G i complex is distinct from the corresponding helix of G s (Fig. 3E), whereas there is little change in the G␣Ras structure of G␣ in these structures, as indicated by small RMSDs (Ͻ2.0 Å). Because the overall orientation of the G protein relative to the receptor would be similar regardless of the state of the G protein, it is reasonable to assume that the CB1-G i complex resembles the ␤ 2 AR-G s complex. Although the present CB1-G i complex model is of value for examining an initial complex formation between the activated receptor and the G protein at the receptor-G interface, further work on the present CB1-G i complex model is warranted, including (i) refinement of the CB1 receptor in consideration of the recently determined x-ray structure of the sphingosine 1-phosphate receptor 1 (S1P1) (93), which has higher sequence identity (32%) than ␤ 2 AR (26%) to CB1 for TM helices; (ii) enhanced sampling to obtain a statistically robust model of the CB1-G i complex; and (iii) a substantially long simulation for a full structural convergence. The refined model of the CB1-G i complex model would serve to provide some insight into the mechanism of how the activated receptor communicates with the coupled G protein, leading to the disruption of GDP that is tightly bound to the nucleotidebinding pocket on G␣ i , which is located approximately Ͼ30 Å from the receptor-G i interface.