Novel Insights into CB₁ Cannabinoid Receptor Signaling: A Key Interaction Identified between the Extracellular-3 Loop and Transmembrane Helix 2

Abstract

Activation of the cannabinoid CB₁ receptor (CB₁) is modulated by aspartate residue D2.63¹⁷⁶ in transmembrane helix (TMH) 2. Interestingly, D2.63 does not affect the affinity for ligand binding at the CB₁ receptor. Studies in class A G protein-coupled receptors have suggested an ionic interaction between residues of TMH2 and 7. In this report, modeling studies identified residue K373 in the extracellular-3 (EC-3) loop in charged interactions with D2.63. We investigated this possibility by performing reciprocal mutations and biochemical studies. D2.63¹⁷⁶A, K373A, D2.63¹⁷⁶A-K373A, and the reciprocal mutant with the interacting residues juxtaposed D2.63¹⁷⁶K-K373D were characterized using radioligand binding and guanosine 5′-3-O-(thio)triphosphate functional assays. None of the mutations resulted in a significant change in the binding affinity of N-(piperidiny-1-yl)-5-(4-chlorophenyl)-1-(2,4-dichloro-phenyl)-4-methyl-1H-pyrazole-3-carboxamide hydrochloride (SR141716A) or (−)-3cis -[2-hydroxyl-4-(1,1-dimethyl-heptyl)phenyl]-trans-4-[3-hydroxyl-propyl] cyclohexan-1-ol (CP55,940). Modeling studies indicated that binding-site interactions and energies of interaction for CP55,940 were similar between wild-type and mutant receptors. However, the signaling of CP55,940, and (R)-(+)-[2,3-dihydro-5-methyl-3-[(4-morpholinyl)methyl]-pyrrolo[1,2,3-de]-1,4-benzoxazin-6-yl](1-naphthalenyl)-methanone mesylate (WIN55,212-2) was impaired at the D2.63¹⁷⁶A-K373A and the single-alanine mutants. In contrast, the reciprocal D2.63¹⁷⁶K-K373D mutant regained function for both CP55,940 and WIN55,212-2. Computational results indicate that the D2.63¹⁷⁶-K373 ionic interaction strongly influences the conformation(s) of the EC-3 loop, providing a structure-based rationale for the importance of the EC-3 loop to signal transduction in CB₁. The putative ionic interaction results in the EC-3 loop pulling over the top (extracellular side) of the receptor; this EC-3 loop conformation may serve protective and mechanistic roles. These results suggest that the ionic interaction between D2.63¹⁷⁶ and K373 is important for CB₁ signal transduction.

Introduction

The cannabinoid CB₁ receptor (CB₁), a member of the class A rhodopsin-like family of G protein-coupled receptors (GPCRs) (see Fig. 1), is found primarily in the central nervous system (CNS) and is important in the regulation of neuronal activity. In addition, there is evidence that the CB₁ receptor is expressed in peripheral tissues (albeit to a lesser extent), including the adrenal gland, bone marrow, heart, lung, and prostate (Howlett et al., 2002). The CB₁ receptor, a Gi/o coupled GPCR binds five structurally diverse classes of ligands; these include the endocannabinoids (typified by anandamide and 2-arachidonoylglycerol), the classic and nonclassic cannabinoids (typified by δ-9-tetrahydrocannabinol and CP55,940 [(−)-cis-3-[2-hydroxyl-4-(1,1-dimethylheptyl)phenyl]-trans-4-[3-hydroxyl-propyl] cyclohexan-1-ol], respectively), the aminoalkylindoles (typified by WIN55,212-2 [(R)-(+)-[2,3-dihydro-5-methyl-3-[(4-morpholinyl)methyl]-pyrrolo[1,2,3-de]-1,4-benzoxazin-6-yl](1-naphthalenyl)methanone mesylate]), and the diarylpyrazole antagonists/inverse agonists [typified by SR141716A (N-(piperidiny-1-yl)-5-(4-chlorophenyl)-1-(2,4-dichlorophenyl)-4-methyl-1H-pyrazole-3-carboxamide hydrochloride)] (see Fig. 2) (Picone et al., 2002).

Fig. 1.

Helix net representation of the hCB₁ receptor sequence. The most highly conserved residue position in each transmembrane helix across class A GPCRs is highlighted in bold. The amino acids mutated in this study are highlighted in red.

Fig. 2.

Compounds evaluated in this study.

Considering its fundamental role in the CNS, it is not surprising that the CB₁ receptor has been reported to mitigate numerous pathologies, including Alzheimer’s disease, cancer, obesity, and pain (Pertwee, 2009). Unfortunately, many attempts at harnessing the therapeutic potential of the CB₁ receptor have failed due to unacceptable CNS-related side effects, such as euphoria, depression, and suicidal fixation (Christopoulou and Kiortsis, 2011). Clearly, a better understanding of the CB₁ receptor’s signal transduction mechanism(s) at a molecular level would be useful in realizing this receptor’s therapeutic potential.

Traditionally, the high degree of sequence homology of amino acid residues from transmembrane helices (TMHs) of different GPCRs has led to the identification of conserved residues, which have been shown to be crucial for receptor function using biochemical studies (Tao and Abood 1998). In addition, charged interactions between amino acid residues from different TMH domains have been shown to be essential for either ligand binding or receptor function (Zhou et al., 1994; Sealfon et al., 1995; Xu et al., 1999). Residues from the extracellular (EC) loops demonstrate low sequence homology (Peeters et al., 2011b) and were initially thought to connect the TMH domains rather than to have a direct role in receptor functioning.

However, recent studies have demonstrated the critical role of the EC loops to ligand binding and receptor signaling. Mutation studies have demonstrated that the first EC loop (EC-1 loop) is important to the activation of the adenosine A_2B receptor (Peeters et al., 2011a). The second EC loop (EC-2 loop) has been shown to be important in ligand binding and activation at the V_1a vasopressin receptor (Conner et al., 2007), to be important to helix movement in rhodopsin (Ahuja et al., 2009), and to be involved in the binding of allosteric modulators at the M₂ acetylcholine receptor (Avlani et al., 2007). Less is known about the third EC loop (EC-3 loop); however, a key salt bridge between the EC-3 and EC-2 loops has been observed to influence ligand binding and receptor activation at the β₂-adrenergic receptor (Bokoch et al., 2010).

The EC-1 and EC-2 loops of the CB₁ receptor (Murphy and Kendall, 2003; Ahn et al., 2009a; Bertalovitz et al., 2010) have been better characterized than its EC-3 loop. EC-3 loop modeling studies reported here suggest that the EC-3 loop residue K373 may form a functionally-important ionic interaction with a transmembrane residue, D2.63¹⁷⁶. Our previous D2.63¹⁷⁶ mutation studies have demonstrated that the negative charge of D2.63¹⁷⁶ is critical for agonist efficacy but not ligand binding at the CB₁ receptor (Kapur et al., 2008). We hypothesized that this functional requirement (of a negatively charged residue at 2.63¹⁷⁶) may be due to this residue’s participation in an ionic interaction with K373 that is necessary for signal transduction. To test this hypothesis, three mutations that would disrupt this putative interaction, D2.63¹⁷⁶A, K373A, and D2.63¹⁷⁶A-K373A, and a charge-reversal mutation D2.63¹⁷⁶K-K373D that would restore the interaction were evaluated for their impact on ligand binding and agonist efficacy. The binding affinities for CP55,940 and SR141716A were not significantly affected by any of the mutations. However, the efficacy of CP55,940 and WIN55,212-2 was markedly reduced by the alanine-substitution mutations, while the charge-reversal mutation led to partial rescue of wild-type (WT) levels of efficacy. Computational results indicate that the D2.63¹⁷⁶-K373 ionic interaction strongly influences the conformation(s) of the EC-3 loop, providing a structure-based rationale for the importance of the EC-3 loop to signal transduction in CB₁. Specifically, the putative ionic interaction results in the EC-3 loop pulling over the top (EC side) of the receptor; this EC-3 loop conformation may serve protective and mechanistic roles.

Our results have for the first time identified an interaction between the residues from TMH2-EC3, suggesting the proximity of these two domains and their role in modulating CB₁ signal transduction.

Materials and Methods

Materials

[³H]CP55,940 (160-180 Ci/mmol) and [³⁵S]GTPγS (guanosine 5′-3-O-(thio)triphosphate; 1250 Ci/mmol) were purchased from PerkinElmer (Boston, MA). WIN55,212-2, CP55,940, and SR141716A were obtained from Tocris (Ellisville, MI). The Pfu Turbo DNA polymerase for mutagenesis experiments was from Stratagene (La Jolla, CA). All other reagents were obtained from Sigma-Aldrich (St. Louis, MO) or other standard sources. The CB₁ antibody was kindly provided by Ken Mackie (Indiana University).

Amino Acid Numbering

The numbering scheme suggested by Ballesteros and Weinstein (1995) was employed here. In this system, the most highly conserved residue in each TMH is assigned a locant of 0.50. This number is preceded by the TMH number and followed by the absolute sequence number in superscript. All other residues in a TMH are numbered relative to this residue. The sequence numbers used are human CB₁ sequence numbers unless otherwise noted (Bramblett et al., 1995).

Mutagenesis and Cell Culture

The D2.63¹⁷⁶A, K373A, D2.63¹⁷⁶A-K373A, and D2.63¹⁷⁶K-K373D mutants of the human CB₁ in the vector pcDNA3 were constructed using the QuikChange site-directed mutagenesis kit (Stratagene). The mutagenic oligonucleotides used were between 27 and 33 base pairs long. Restriction endonuclease digestion and DNA sequencing subsequently confirmed the presence of the mutation. Stably transfected human embryonic kidney (HEK)-293 cell lines were created by transfection with WT or mutant CB₁-pcDNA3 cDNA by the lipofectamine reagent (Invitrogen, Carlsbad, CA) and selected in growth medium containing geneticin (1 mg/ml), as previously described elsewhere (McAllister et al., 2003).

Radioligand Binding and GTPγS Binding Assay

Protein membrane preparations harvested from stably transfected HEK293 cells were prepared and assayed as previously described elsewhere (Kapur et al., 2007). In brief, binding assays (saturation and competition binding assays) were initiated by the addition of 50 μg membrane protein to glass tubes pretreated with siliconizing fluid (Pierce, Rockford, IL; to reduce nonspecific binding) containing [³H]SR141716A, and an appropriate volume of binding buffer A (50 mM Tris-Base, 1 mM EDTA, 3 mM MgCl₂, and 5 mg/ml bovine serum albumin, pH7.4) to bring the final volume to 500 μl. Nonspecific binding was determined in presence of excess (1 μM) unlabeled SR141716A. Reactants were allowed to reach equilibrium (~1 hour). Subsequently, free and bound radioligand were separated by vacuum filtration through Whatman GF-C filters, and the radioactivity retained on the filters was quantified by a liquid scintillation counter.

The K_d (equilibrium dissociation constant) and B_max (maximal binding) values were determined by analyzing the saturation binding data by nonlinear regression and fitted to a one-site binding model using GraphPad Prism 4.0 software (GraphPad, San Diego, CA). The displacement log IC₅₀ values were determined by nonlinear regression and fitting the data to one-site competition and then were converted to K_i (inhibitory constant) values using the Cheng and Prusoff method (Cheng and Prusoff, 1973) with the use of GraphPad Prism.

The GTPγS assay was initiated by the addition of 20 μg of membrane protein into silanized glass tubes containing 0.1 nM [³⁵S]GTPγS, 10 μM GDP in GTPγS binding buffer (50 mM Tris-HCl, 100 mM NaCl, 3 mM MgCl₂, 0.2 mM EGTA, and 0.1% bovine serum albumin, pH7.4). Nonspecific binding was assessed in the presence of 20 μM unlabeled GTPγS. Free and bound radioligand were separated, and bound radioactivity was quantified as described previously. Nonlinear regression of log concentration values versus the percentage effect fitted to sigmoidal dose-response was used to obtain estimates of agonist concentrations that elicit half the maximal response (EC₅₀) and maximal response (E_max).

Statistical Analyses

Data are reported as the mean value of the replicates along with their 95% confidence limits (CL). The K_i and log EC₅₀ values in the mutant and WT CB₁ receptors were compared using one-way analysis of variance with Bonferroni multiple comparison post tests. P<.05 was considered statistically significant.

Molecular Modeling

Receptor Model Construction Protocol for Loop Calculations.

Wild-type CB₁ activated (R*) receptor model construction.

Using interactive computer graphics, extracellular (EC-1 F180–S185, EC-2 G254–E273, and EC-3 G369–K376) and intracellular loops (IC-1 R145–R150, IC-2 P221–T229, and IC-3 S303–M336) were manually added to our previously constructed TMH bundle model of the CB₁ R* (active state) receptor, with CP55,940 docked in its global minimum energy conformation (Kapur et al., 2007). The program Modeler was then used to refine loop structures (Sali and Blundell, 1993; Fiser et al., 2000). Because of their close spatial proximity, the conformations of all three EC loops were calculated together followed by calculation of the three IC loop conformations. Chosen loop conformations were those that produced a low value of the Modeler objective function. The loops were minimized in three stages (stages 1 to 3, as described later). Next, portions of the N and C termini were added, and conformations of each were refined in Modeler. The termini were minimized using stages 4 to 5 of the minimization protocol.

N terminus.

The first 89 residues of the N terminus were truncated, based on results from the Chin laboratory (Andersson et al., 2003) which showed that CP55,950 has WT binding affinity and efficacy at the N-terminal truncated CB₁, whereas the receptor has better cell surface expression than WT. X-ray crystal structures of class A GPCRs with lipid-derived endogenous ligands show that the N terminus occludes the binding pocket. In the crystal structure of rhodopsin (Li et al., 2004), the N terminus is positioned centrally, occluding the EC side of the bundle (i.e., the retinal plug). This general placement of the N terminus is also observed in the crystal structure of the sphingosine 1-phosphate receptor (Hanson et al., 2012). Because CB₁ also has a lipid-derived endogenous ligand, a truncated N-terminal conformation (positioned centrally over the EC side of the receptor) was chosen.

EC-2 loop.

One of the significant sequence divergences between rhodopsin and CB₁ is in the EC-2 loop region. This loop in CB₁ is shorter than in rhodopsin and is missing the conserved disulfide bridge between the cysteine in the EC-2 loop and C3.25 in TMH3 of rhodopsin. Instead, there is a Cys at the extracellular end of TMH4 in CB₁ and a Cys near the middle of the EC-2 loop that experiments suggest may form a disulfide bridge (Fay et al., 2005). Consequently, the position of the EC-2 loop with respect to the binding site crevice in CB₁ around TMHs 3, 4, and 5 is likely to be quite different from that in rhodopsin. Therefore, this loop was modeled with an internal C257–C264 disulfide bridge based upon mutation results from the Farrens laboratory (Fay et al., 2005), which show that these two cysteines are required for high-level expression and receptor function.

To guide selection of an appropriate EC-2 loop conformation, we used mutation results from the Kendall laboratory (Ahn et al., 2009a; Bertalovitz et al., 2010), which demonstrate that mutation of EC-2 loop residue F268 to a tryptophan severely damages the binding affinity and efficacy of CP55,940 but has no significant effect on the binding affinity of SR141716A. Thus, an EC-2 loop conformation was chosen that placed F268 in close proximity to CP55,940. A F268W mutant bundle was constructed to verify that this mutation resulted in significant steric overlaps with CP55,940 in our model but not with SR141716A (Supplemental Fig. 1).

EC-3 loop.

The EC loops were refined by use of Modeler in two stages. In the first stage, no harmonic distance constraints were used. This calculation was performed to examine the general conformational space of the EC-3 loop. The EC-3 loop conformation with the lowest objective function placed the EC-3 loop over the top of the receptor; in addition, the putative ionic interaction between D2.63¹⁷⁶ and K373 had formed. In the second stage, a 3.0 kcal/mol harmonic distance constraint was placed between the EC-3 loop residue K373 and D2.63¹⁷⁶. Specifically, the distance between the OD1 atom of D2.63¹⁷⁶ and the NZ atom of K373 was constrained to 3.0 ± 2.0 Å. This second calculation was performed to obtain a focused conformational sampling of the EC-3 loop conformation with the lowest objective function (obtained in the first stage of the calculation).

IC-3 loop.

The CB₁ IC-3 loop is much longer than the corresponding sequence in rhodopsin. Nuclear magnetic resonance experiments have been performed on a peptide fragment composed of the CB₁ sequence span from the IC end of TMH5 to the IC end of TMH6 in micelles (Ulfers et al., 2002). This study suggested that part of the IC-3 loop is α helical. This region occurs after the IC end of TMH5 [K5.64³⁰⁰] and consists of a short α-helical segment from A301 to R307, followed by an elbow region (R307–I309) and an α-helical segment (Q310–S316) up to an III sequence (I317–I319) in IC-3. Based on these results, we replaced the initial Modeler-built IC-3 loop with this α-helix-elbow-α-helix region, and then the rest of IC-3 loop (I317–P332) was rebuilt and optimized using Modeler.

C terminus.

A C-terminal fragment S414–G417, which contains a putative palmitoylation site at Cys⁴¹⁵ (Fay et al., 2005), was added to the model and C415 was palmitoylated. C-terminal truncation experiments from the Mackie laboratory (Jin et al., 1999) have shown that CB₁ (with truncation at C417) signals normally in the presence of agonists. With the exception of helix 8, the C terminus is largely unstructured, though recent work on an isolated C-terminal peptide suggests the existence of an additional C-terminal helix, helix 9 (Ahn et al., 2009b). However, recent results from the Mackie laboratory (Straiker et al., 2012) reinforce that the functional significance of the C terminus pertains to desensitization and receptor internalization—not necessarily to receptor signaling by heterotrimeric G proteins. Therefore, we modeled the truncated C terminus as unstructured.

Receptor Model Energy Minimization Protocol.

The energy of the ligand/CB₁ R* complex, including loop regions and N and C termini, was minimized using the OPLS 2005 force field in Macromodel 9.9 (Schrödinger Inc., Portland, OR). An 8.0-Å extended nonbonded cutoff (updated every 10 steps), a 20.0-Å electrostatic cutoff, and a 4.0-Å hydrogen bond cutoff were used in each stage of the calculation.

The minimization was performed in five stages. In the first stage of the calculation, the ligand and TMH bundle were frozen, but the loops were allowed to relax. The generalized born/surface area continuum solvation model for water as implemented in Macromodel was used. This stage of the calculation consisted a of Polak–Ribier conjugate gradient minimization in 1000-step increments until the bundle reached the 0.05 kJ/mol gradient. Because mutation results from the Kendall laboratory (Ahn et al., 2009a; Bertalovitz et al., 2010) demonstrate that mutation of EC-2 loop residue F268 to a tryptophan severely damages the binding affinity and efficacy of CP55,940, the terminal side-chain hydrogen of F268 (atom name: HZ) was frozen in place. Freezing this hydrogen allowed F268 the most conformational freedom, allowing the side-chain to pivot about HZ while requiring F268 to stay in close proximity to CP55,940. The second stage of the calculation was performed exactly as the first stage, except that HZ of F268 was unfrozen.

In the third stage, the loops were frozen but the ligand and the side chains of the TMHs were allowed to optimize. The minimization consisted of a conjugate gradient minimization using a distance-dependent dielectric, performed in 1000-step increments until the bundle reached the 0.05 kJ/mol gradient. Because a previously minimized TMH bundle was used as the starting structure in constructing this model (Kapur et al., 2007), the backbone atoms of the transmembrane helices were frozen to prevent the bundle from over packing.

In the fourth stage, the N and C termini were minimized using the same protocol used in second stage. In this stage, only the termini were minimized. In the fifth stage, the TMH bundle was minimized again, exactly as described in the third stage.

Mutant CB₁ activated (R*) receptor models construction and minimization.

Four mutant bundles were constructed: K373A, D2.63¹⁷⁶A, D2.63¹⁷⁶A-K373A, and D2.63¹⁷⁶K-K373D. These mutant models were constructed using the final WT CB₁ R* model (Kapur et al., 2007) as the starting structure, using interactive computer graphics to perform the appropriate mutations. The N terminus was temporarily removed to prevent it from biasing the mutant loop refinement. Modeler was used (as before) to refine the EC-1 and EC-3 mutant loop structures. No harmonic distance constraint was used for the alanine-substitution mutants (however, the same distance constraint used for WT was also used for D2.63¹⁷⁶K-K373D).

The WT conformations of the EC-2 loop, the IC loops, and the termini were preserved. As with the WT CB₁ R*, the chosen loop configurations for the mutant bundles were those that produced a low value of the Modeler objective function; the loops were minimized using stages 1 to 3 of the minimization protocol (see above). Next, the N terminus was reattached to the mutant bundles. The termini were minimized using stages 4 to 5 of the minimization protocol (see the earlier description).

Assessment of Pairwise Interaction and Total Energies

Interaction energies between CP55,940 and the WT, K373A, D2.63¹⁷⁶A, D2.63¹⁷⁶A-K373A, or D2.63¹⁷⁶K-K373D receptors were calculated using Macromodel (Schrodinger). After defining the atoms of CP55,940 as one group (group 1) and the atoms corresponding to a residue that lines the binding site in the final ligand/CB1 R* complex as another group (group 2), Macromodel was used to output the pairwise interaction energy (coulombic and van der Waals) for a given pair of atoms. The pairs corresponding to group 1 (ligand) and group 2 (residue of interest) were then summed to yield the interaction energy between the ligand and the receptor.

Results

The binding of [³H]SR141716A to WT and mutant receptors stably expressed in HEK 293 cells was measured to generate an estimate of the K_d and B_max values. Similar cell surface expression of WT and mutant cell lines was verified by immunofluorescence staining (unpublished data).

Radioligand Binding Assays

Saturation binding analysis of [³H]SR141716A at the D2.63¹⁷⁶A, K373A, D2.63¹⁷⁶A-K373A, and D2.63¹⁷⁶K-K373D mutations displayed K_d (CL) values of 4.2 (0.1–9.8) nM, 1.7 (0.2–3.5) nM, 4.4 (0.1–9.1) nM, and 3.5 (0.1–24) nM, respectively (see Table 1). The K_d for the WT hCB₁ receptor was 2.2 (0.4–3.9) nM. The K_d values for the mutants versus WT were not statistically significantly different. Similarly, the B_max values for each cell line demonstrated that expression levels of these receptors between the different cell lines were comparable. The cell lines D2.63¹⁷⁶A, K373A, D2.63¹⁷⁶A-373A, and D2.63¹⁷⁶K-K373D respective B_max (CL) values were 2.3 (1.0–3.5) pmol/mg, 1.8 (0.1–3.7) pmol/mg, 2.7 (1.7–3.7) pmol/mg, and 0.7 (0.1–2.4) pmol/mg. The WT CB₁ cell line displayed a B_max of 2.4 (1.9–2.9) pmol/mg.

TABLE 1.

Radioligand binding properties of wild-type and mutant cell lines

The K_d and B_max values were determined from saturation binding experiments using [³H]SR141716A on HEK293 cell membrane preparations stably transfected with the wild-type or mutant hCB₁ receptor. Data represent the mean and corresponding S.E.M. of at least three independent experiments performed in triplicate. No statistically significant difference was observed between the wild-type and mutant binding properties as determined by a two-tailed Student’s t test.

Radioligand	Cell Line	Kd (nM)	95% CL	Bmax	95% CL
				pmol/mg
[3H]SR141716A	WT hCB1	2.2	(0.4–3.9)	2.4	(1.9–2.9)
	D2.63176A	4.2	(0.1–9.8)	2.3	(1.0–3.5)
	K373A	1.7	(0.2–3.5)	1.8	(0.1–3.7)
	D2.63176A-K373A	4.4	(0.1–9.1)	2.7	(1.7–3.7)
	D2.63176K-K373D	3.5	(0.1–24)	0.7	(0.1–2.4)

Competitive Binding Assays

We investigated the binding affinity of the bicyclic cannabinoid agonist CP55,940 to displace [³H]SR141716A bound to the WT and mutant hCB₁ receptors. The K_i values between WT and mutant receptors overlapped and were not statistically significantly different (see Fig. 3; Table 2). The K_i (CL) values for WT, D2.63¹⁷⁶A, K373A, D2.63¹⁷⁶A-K373A, and D2.63¹⁷⁶K-K373D were 17 (5.3–53) nM, 4.9 (0.6–43) nM, 17 (3.3–85) nM, 5.1 (0.63–42) nM, and 15 (3.0–69) nM, respectively.

Fig. 3.

Competitive displacement of [³H]SR141716A. CP55,940 was used as the displacing compound. [³H]SR141716A binding in membranes prepared from HEK293 cells stably transfected with wild-type, D2.63¹⁷⁶A, K373A, or D2.63¹⁷⁶A-K373A mutant CB₁ receptors. Each data point represents the mean ± S.E.M. of at least three independent experiments performed in triplicate.

TABLE 2.

The effects of amino acid mutations of recombinant hCB₁ receptors on the displacement of [³H]SR141716A by CP55,940

Data represent the mean and corresponding 95% confidence limits of at least three independent experiments performed in triplicate. The K_i value of CP55,940 at the mutant receptors was not statistically significantly different from wild-type CB₁ receptors using a two-tailed Student’s t test.

[3H]SR141716A	CP 55,940 (Ki) + 95% CL
WT	17 nM (5.3–53)
D2.63176A	4.9 nM (0.6–43)
K373A	17 nM (3.3–85)
D2.63176A-K373A	5.1 nM (0.6–42)
D2.63176K-K373D	15 nM (3.0–69)

Agonist Stimulated GTPγS Binding

We used [³⁵S]GTPγS binding to measure the stimulation of WT and mutant cannabinoid receptors upon stimulation with different classes of cannabinoid ligands (see Fig. 4; Table 3). The EC₅₀ and E_max values were generated for WT and mutant receptor activation in the presence of CP55,940 and WIN55,212-2. The EC₅₀ values of CP55,940 and WIN55,212-2 at WT CB₁ were 12.6 nM and 36.7 nM, respectively. The D2.63¹⁷⁶A mutation statistically significantly increased EC₅₀ values for CP55,940 and WIN55,212-2 to 67 nM (5.3-fold) and 231 nM (6.3-fold), respectively, and the maximum agonist responsiveness was lower. The K373A mutation resulted in similar effects on the EC₅₀ and E_max values. The K373A mutant generated a statistically significant increase in EC₅₀ values from WT for CP55,940 and WIN55,212-2 to 70 nM (5.6-fold) and 274 nM (7.5-fold), respectively.

Fig. 4.

Activation of wild-type and mutant receptors. (A) CP55,940. (B) WIN55,212-2. Concentration-effect curves were obtained from [³⁵S]GTPγS binding in HEK293 membrane preparations expressing wild-type or D2.63¹⁷⁶A, K373A, or D2.63¹⁷⁶A-K373A, D2.63¹⁷⁶K-K373D mutant CB₁ receptors. Each data point represents the mean ± S.E.M. of at least three independent experiments performed in triplicate.

TABLE 3.

Concentration-effect data for agonist stimulation of [³⁵S]GTPγS binding of WT and mutant receptors stably expressed in HEK293 cells

Data represent the mean of at least three independent experiments performed in triplicate. EC₅₀ values were determined from concentration-effect curves using GraphPad Prism software. The values in parentheses are 95% confidence intervals. Statistical analysis was performed by comparing the log EC₅₀ of the mutant receptor to the wild-type CB₁ receptors using a two-tailed Student’s t test to determine the level of significance.

	Drug	EC50 (95% CL)	Emax/Top (95% CL)	Mutant/WT EC50
WT	CP55,940	12.6 nM (3.6–43.7)	127% (109–145)	NA
	WIN55,212-2	36.7 nM (12.4–108.3)	95% (87–116)	NA
D2.63176A	CP55,940	67 nM (17–259)*	59% (48–71)*	5.3*
	WIN55,212-2	231 nM (27–1912)*	89% (60–118)	6.3*
K373A	CP55,940	70 nM (5.6–870)*	70% (44–100)	5.6*
	WIN55,212-2	274 nM (61–1230)*	83% (53–112)	7.5*
D2.63176A-K373A	CP55,940	39.8 nM (10–153)*	29% (23–35)*	3.2*
	WIN55,212-2	561 nM (60–5193)*	59% (38–81)*	15.3*
D2.63K-K373D	CP55,940	38 nM (6.9–209)	82% (73–93)	3
	WIN55,212-2	126 nM (26–607)	79% (63–95)	3.4

^*P < 0.05.

However, when the ionic interaction between D2.63¹⁷⁶ and K373 was disrupted by double-alanine substitutions, the receptor activity was severely reduced. The D2.63¹⁷⁶A-K373A mutant resulted in dramatic shifts of either or both the EC₅₀ and E_max values and for CP55,940 and WIN55,212-2 to 39.8 nM (E_{max =} 29%) and 561 nM (E_{max =} 59%), respectively. The largest shift observed was from WIN55,212-2, 15.3-fold above the WT value. In contrast, the charge-reversal mutant D2.63¹⁷⁶K-K373D displayed an EC₅₀ and E_max for WIN55,212-2 of 126 nM and 79%, respectively. Likewise, the D2.63¹⁷⁶K-K373D mutant EC₅₀ and E_max values for CP55,940 were 38 nM and 82%, respectively.

Modeling Studies

Modeler Results: EC Loop Conformations in the WT CB₁ R* and the D2.63¹⁷⁶A, K373A, D2.63¹⁷⁶A-K373A, and D2.63¹⁷⁶K-K373D Mutants.

As described in Materials and Methods, low-energy WT and mutant loop conformations were added to our previous CB₁ R* bundle (Kapur et al., 2007). Consistent with the experimental results, the WT model includes an ionic interaction between D2.63¹⁷⁶ and EC-3 loop residue K373 (see Fig. 5). This ionic interaction causes the EC-3 loop to pull across the top (EC side) of the receptor. Clearly, this specific ionic interaction cannot form in the D2.63¹⁷⁶A, K373A, or the D2.63¹⁷⁶A-K373A mutant. By not forming this ionic interaction, the EC-3 loops of the mutant receptors experience greater conformational freedom. As illustrated in Fig. 5, the modeled loop conformations of D2.63¹⁷⁶A, K373A, and D2.63¹⁷⁶A-K373A position the EC-3 loop away from the center and are more directly above TMHs 6 and 7. It is noteworthy that these three mutants have very similar EC-3 loop conformations and that these conformations are fundamentally different from the WT EC-3 loop conformation.

Fig. 5.

Extracellular (EC) loop conformations of WT and D2.63¹⁷⁶A, K373A, or D2.63¹⁷⁶A-K373A mutant CB₁ receptors in the active (R*) state. CP55,940 is shown in green; D2.63¹⁷⁶ and K373 are shown in orange; WT EC loops are shown in blue; D2.63¹⁷⁶A, K373A, and D2.63¹⁷⁶A-K373A mutant EC loops are shown in red, yellow, and purple, respectively. In the WT model, the putative ionic interaction between D2.63¹⁷⁶ and K373 has formed; this promotes an EC-3 loop conformation that is pulled over the top of the receptor. In the alanine-substitution models, the putative interaction does not form, and the EC-3 loops are away from the bundle core. (A) Viewpoint is from EC with intracellular (IC) portions of TMHs, IC loops, and the N and C termini omitted to simplify view. (B) Viewpoint is from lipid looking between TMH1 and 7. Note: The IC portions of TMHs, IC loops, EC-1, EC-2, part of TMH1 and 7, and the N and C termini have been omitted here to simplify the view.

Unlike the D2.63¹⁷⁶A, K373A, or the D2.63¹⁷⁶A-K373A mutant, the D2.63¹⁷⁶K-K373D swap mutant can form the putative ionic interaction. In agreement with the experimental results, the model of this mutant includes an ionic interaction between D2.63¹⁷⁶K and K373D (see Fig. 6). This ionic interaction causes the EC-3 loop to pull across the top (EC side) of the receptor. As observed in Fig. 6, the WT and the D2.63¹⁷⁶K-K373D EC loops have a remarkable degree of conformational similarity in their EC loops. The formation of the putative ionic interaction (despite having switched the residues at 2.63¹⁷⁶ and 373) explains how the D2.63¹⁷⁶K-K373D swap mutant is capable of promoting an EC-3 loop conformation that is very similar to the WT EC-3 loop conformation.

Fig. 6.

Extracellular (EC) viewpoint of EC loop conformations of WT CB₁ R* and the D2.63¹⁷⁶K-K373D swap mutant. CP55,940 is shown in green; D2.63¹⁷⁶ and K373 are shown in orange; WT EC loops are shown in blue; the D2.63¹⁷⁶K-K373D swap mutant EC-3 loop is shown in red (see Fig. 5 legend for further details). (A) In the WT model, the putative ionic interaction between D2.63¹⁷⁶ and K373 has formed; this promotes an EC-3 loop conformation that is pulled over the top of the receptor. (B) In the D2.63¹⁷⁶K-K373D mutant model, the putative ionic interaction has been formed, promoting an EC-3 loop conformation that is very similar to WT.

In addition, the negatively charged K373D may be able to form ionic interactions with K370 and K7.32³⁷⁶ (see Fig. 1). These interactions may reduce the frequency of the D2.63¹⁷⁶K-K373D ionic interaction. Energetically favorable interactions with K370 and K7.32³⁷⁶ are highly unlikely for K373 (in WT), as its positive charge would be repelled by the positive charge on K370 and K7.32³⁷⁶.

CP55,940/Receptor Pairwise and Total Interaction Energies.

The results of the saturation and competitive binding assays demonstrate that all mutations do not significantly affect the binding affinity of the ligands studied. Therefore, to test whether our models agreed with these results, pairwise and total interaction energies were calculated for the WT and mutant models (the total interaction energies are listed in Table 4; the complete pairwise interaction energies are listed in Supplemental Tables 1–5). Only five residues contribute at least 5% of the total interaction energy between CP55,940 and each of the models (Supplemental Tables 1–5). Strikingly, these five important residues are the same in the WT and mutant models (Q1.32¹¹⁶, F2.57¹⁷⁰, K3.28¹⁹², S7.39³⁸³, and L7.43³⁸⁷). This consistency (in which residues contribute at least 5% of the total interaction energy) qualitatively suggests that CP55,940 binds WT and mutant receptors similarly. Quantitatively, Table 4 shows that none of the mutations resulted in significant change in the total interaction energy between CP55,940 and the receptor. These results indicate that our computational models are consistent with the results of the binding assays.

TABLE 4.

Total interaction energy of CP55,940 at the WT and mutant CB1 R* models

Model	Total Interaction Energies (kcal/mol)
	Coulombic	VdW	Total
CB1 WT	−22.12	−38.84	−60.95
D2.63176A	−23.34	−36.94	−60.28
K373A	−23.34	−36.94	−60.28
D2.63176A-K373A	−23.61	−38.60	−62.22
D2.63176K-K373D	−24.72	−38.74	−63.46

Discussion

In our present study, we have used computational methods together with model-guided mutagenesis to evaluate the functional importance of a putative intramolecular ionic interaction within the CB₁ receptor. Our previous mutation studies demonstrated the importance of a negative charge at residue 2.63¹⁷⁶, possibly indicating its involvement in an essential ionic interaction (Kapur et al., 2008). Our previous modeling studies indicated that the EC-3 loop residue K373 might be the ionic partner to D2.63¹⁷⁶. To test this hypothesis, four substitution mutant CB₁ receptors were constructed—D2.63¹⁷⁶A, K373A, D2.63¹⁷⁶A-K373A, and D2.63¹⁷⁶K-K373D—to evaluate the effect of removing the putative ionic interaction. The charge-reversal mutant was designed to determine whether switching the positions of the ionic partners could rescue WT levels of function. Finally, computational methods were also used to explore how the putative ionic interaction influences receptor structure.

Ligand binding affinity was not significantly affected by any of the mutations performed here. These results are consistent with our prior characterization of D2.63¹⁷⁶, which was shown to be crucial for signal transduction but did not participate in high affinity agonist binding (Kapur et al., 2008). In addition, the binding affinity data reported here are consistent with the predictions made from our WT and mutant models. Specifically, CP55,940 was found to have a similar total interaction energy in the WT and mutant receptor models; this is not surprising, as neither D2.63¹⁷⁶ nor the EC-3 loop are part of the predicted CP55,940 binding pocket. Indeed, reports of EC loop mutations that only affect agonist efficacy (and not ligand binding) are well documented in the GPCR literature. Residues in EC-2 loop of the M₃ muscarinic receptor could be mutated without affecting ligand binding; however, a significant reduction in agonist efficacy was observed (Scarselli et al., 2007). Similar results have been observed in the EC-1 loop of the adenosine A_2B receptor (Peeters et al., 2011a). Analogously, our results suggest that the ionic interaction between D2.63¹⁷⁶ and K373 is not important for SR141716A (Table 1) or CP55,940 binding at the CB₁ receptor. Additionally, none of the mutations significantly affected the B_max for [³H]SR141716A. These results suggest that the mutations reported here did not cause the receptor to fold incorrectly or fail to express at the cell surface.

In contrast to the binding affinity results, the D2.63¹⁷⁶A, K373A, D2.63¹⁷⁶A-K373A or D2.63¹⁷⁶K-K373D mutations caused a significant change in CP55,940s EC₅₀ compared with WT. However, the EC₅₀s of the D2.63¹⁷⁶A, K373A, or the D2.63¹⁷⁶K-K373D mutants when compared with the D2.63¹⁷⁶A-K373A mutant were not significantly different. This is consistent with our hypothesis that it is the ionic interaction between the charged residues D2.63¹⁷⁶ and K373 (and not the residues independently) that is important to agonist efficacy. If D2.63¹⁷⁶ and K373 were independently important to function, one would expect that the EC₅₀ of the double-alanine mutant would be higher than either of the single-alanine mutants.

We previously reported that presence of a negatively charged residue at position 2.63¹⁷⁶ is crucial for receptor function (Kapur et al., 2008). In this study, we demonstrate that an ionic interaction between D2.63¹⁷⁶ and K373 (not simply the negative charge on D2.63¹⁷⁶ per se) is required for CB₁ WT function. The E_max values for either CP55,940 or WIN55,212-2 at the double-alanine mutant showed a significant decrease in function. This result suggests that the interaction between D2.63¹⁷⁶ and K373 is important for signaling at CB₁. This result is also reinforced by results for the charge-reversal mutant D2.63¹⁷⁶K-K373D, as both CP55,940 and WIN55,212-2 showed a restoration of function compared with the double-alanine mutation. Therefore, the ability to switch the residues at 2.63¹⁷⁶ and 373 (dramatically flipping the polarity on both residues) and preserve near WT levels of efficacy strongly supports the existence of a functionally required ionic interaction between D2.63¹⁷⁶ and K373.

The single-alanine mutants showed increased E_max values relative to the double-alanine mutant (E_max = 59% at D2.63A and E_max = 29% for CP55,940; E_max = 89% at D2.63A and E_max = 59% for WIN55,212-2) that are larger than what might be expected for disruption of the same ionic interaction as seen in the double-alanine mutant. Our models suggest that residues near the putative ionic interaction may help rescue function in these single-alanine mutants. There are two additional lysines (K370 and K7.32³⁷⁶, see Fig. 1) that are in close proximity to K373. These lysines may be able to form an ionic interaction with D2.63¹⁷⁶, thus partially rescuing function at the K373A mutant. Likewise, there is a negatively charged aspartate (D184) and two hydrophilic residues (H181 and D185) on the EC-1 loop that are in close proximity to D2.63¹⁷⁶. These residues may be able to form an ionic interaction (or a simple hydrogen bond in the case of H181 and S185) that enables the partial rescue of function at the D2.63¹⁷⁶A mutant. In the double mutant D2.63¹⁷⁶A-K373A, no such rescue would be possible because the polar residues at each site (D2.63 or K373) have been replaced with a nonpolar residue (Ala).

Involvement of EC Loops in GPCR Activation.

Results reported here suggest that the formation of an ionic interaction/salt bridge between the EC-3 loop and the EC end of TMH2 is important for CB₁ signaling. The hallmark of class A GPCR activation by an agonist is the “tripping” of the toggle switch within the binding pocket that allows TMH6 to flex in the highly conserved Cys-Trp-any amino acid-Pro hinge region and straighten. This straightening breaks the “ionic lock” between R3.50 and E/D6.30 at the IC end of the receptor. The result is the formation of an IC opening of the receptor, exposing residues that can interact with the C terminus of the Gα-subunit of the G protein (Hamm et al., 1988). There is increasing evidence that movements in the EC loops also occur subsequent to agonist binding and are integral to transmission of the activation “message” (or covalent ligand isomerization in the case of rhodopsin). Nuclear magnetic resonance studies of rhodopsin activation by light have indicated that activation triggers a simultaneous displacement of the EC-2 loop and TMH5. Motion of EC-2 may allow the EC end of the TMH6-EC-3-TMH7 segment to pivot toward the center of the protein and conversely allow the IC end of TMH6 to rotate outward (Ahuja et al., 2009). In some class A GPCRs, such as chemokine receptor 4, a specific interaction between the EC-3 loop and N terminus (disulfide bridge) acts as a “microswitch” that is crucial to the chemokine receptor 4 signaling (Rana and Baranski, 2010).

The computational results reported here illustrate how the ionic interaction between D2.63¹⁷⁶ and K373 causes the EC-3 loop to pull across the top (EC side) of the receptor. Notably, this EC-3 loop conformation is preserved in the charge-reversal mutant D2.63¹⁷⁶K-K373D. As described in Results, a strikingly different EC-3 loop conformation is observed in the three alanine-substitution mutants. These results suggest that the putative ionic interaction strongly influences the conformation of the EC-3 loop. This promoted EC-3 loop conformation could serve two important structural roles. First, this EC-3 loop conformation may contribute to forming a protected, closed EC surface, as has been reported in the crystal structures of rhodopsin (Li et al., 2004) and the sphingosine 1-phosphate receptor (Hanson et al., 2012). Second, this ionic interaction creates a noncovalent “tether” between the EC ends of TMHs 2, 6, and 7, allowing conformational changes that occur on one side of the receptor to be transmitted to the other side of the receptor. Thus, the alanine-substitution mutants are less capable of transmitting conformational changes throughout the receptor, and efficacy is consequently impaired. In conclusion, we have identified the EC-3 loop conformation that is mechanistically important in the signaling cascade in CB₁.

Abbreviations

B_max

maximal binding

CB₁

cannabinoid CB₁ receptor

CL

confidence limit

CP55,940

(−)-cis-3-[2-hydroxyl-4-(1,1-dimethylheptyl)phenyl]-trans-4-[3-hydroxyl-propyl] cyclohexan-1-ol

EC

extracellular loop

E_max

maximal effective response

GPCRs

G protein-coupled receptors

GTPγS

guanosine 5′-3-O-(thio)triphosphate

HEK

human embryonic kidney

IC loop

intracellular loop

K_d

equilibrium dissociation constant

K_i

inhibitory constant

SR141716A

N-(piperidiny-1-yl)-5-(4-chlorophenyl)-1-(2,4-dichlorophenyl)-4-methyl-1H-pyrazole-3-carboxamide hydrochloride

TMH

transmembrane helices

WIN55

212-2,(R)-(+)-[2,3-dihydro-5-methyl-3-[(4-morpholinyl)methyl]-pyrrolo[1,2,3-de]-1,4-benzoxazin-6-yl](1-naphthalenyl)methanone mesylate

WT

wild type

Authorship Contributions

Participated in research design: Marcu, Abood, Shore, Makriyannis, Reggio, Kapur.

Conducted experiments: Marcu, Trznadel, Kapur, Shore.

Performed data analysis: Marcu, Kapur, Shore.

Wrote or contributed to the writing of the manuscript: Abood, Reggio, Shore, Marcu.