Sodium ions have been shown to modulate dopamine receptors, and allosteric modulation by sodium ions has been shown to drive the conformational equilibrium of heterotrimeric G protein-coupled receptors (GPCRs) toward an agonist low-affinity state (for review, see Schetz, 2005). In dopamine receptors, like in other heterotrimeric GPCRs, the highly conserved and negatively charged aspartic acid at position 2.50 (the generic numbering system is defined in Ballesteros and Weinstein, 1995) has been identified as a sodium interaction site. For example, charge-neutralizing mutations in the D₂ or the D₄ receptor [e.g., D2.50(80)N or D2.50(80)A] make them sodium-insensitive, whereas a charge-sparing mutation [e.g., D2.50(80)E] retains much of the sodium sensitivity (Neve et al., 1991; Schetz and Sibley, 2001). D₂ receptor mutations at other positions [e.g., S3.39(121)A and S7.46(391)A] also diminish sensitivity, presumably by reducing the H-bonding capacity at the sodium binding site (Neve et al., 2001). The latter studies, in the structural context of the high-resolution crystalline structure of bovine rhodopsin (Palczewski et al., 2000), led to a revised model of the sodium binding site (Neve et al., 2001), in which sodium is at the center of a square-pyramidal hydrogen-bonding network whose vertices are formed by D2.50(80), S3.39(121), N7.45(390), and S7.46(391); sodium binding is thought to neutralize the negative charge at D2.50(80). Allosteric modulation of dopamine receptors by sodium has been shown previously to reduce the affinity of endogenous agonists and zinc, increase the affinity or binding capacity (B_max) of substituted benzamide antagonists, and alter the rate of chemical modification (Neve, 1991; Schetz et al., 1999, 2001; Vivo et al., 2006). However, despite the prevalence of studies across multiple GPCR families indicating allosteric modulation by sodium, an exhaustive search of the literature failed to identify a mechanism for the sodium-induced effects.

In the process of determining the source of large discrepancies in binding affinity reported for 1,4-disubstituted aromatic piperidines/piperazines (1,4-DAPs) for the D₂-V2.61(91)F mutant (Simpson et al., 1999; Floresca et al., 2005), we discovered a change, elicited by sodium binding, in the dynamic properties of the receptor that correlate with a dramatic increase in sodium sensitivity for both agonists and antagonists belonging to a similar structural class. This finding offers an opportunity to understand the allosteric mechanism of the effect produced by sodium binding. To this end, we carried out a normal mode analysis (NMA) of the dynamic properties of various D₂ receptor constructs using three-dimensional molecular models of the receptor. Normal modes are calculated from the molecular structure model and provide information about the component harmonic motions or vibrations of the molecule that characterize its dynamic fluctuation as it occupies a stable conformational state (e.g., inactive state, etc.). The modes constitute a set of orthogonal vectors ranked by energy (or the corresponding frequency), which indicates the direction in which each particle (the component atoms, or residues, or C_α) is moving at that particular level of energy (frequency). Thus, the superposition of all the normal mode vectors describes the entire intrinsic motion of the molecule based on its shape and molecular connectivity, but often one or a few low-frequency (low energy) modes contribute most significantly to this thermal “breathing” motion of the molecule. It was demonstrated for many proteins that the directions of the lowest frequency modes also tend to indicate the path of molecular movements associated with functionally relevant conformational changes (Cui and Bahar, 2006). Here, the comparison of the normal modes between sodium-bound and sodium-free structures of the receptor models allowed us to identify a specific sodium-responsive normal mode motion that indicates distinct dynamic changes in the environment of position 2.61(91) and accounts for the hypersensitivity in the mutant D₂-V2.61(91)F. Thus, when backbone movements associated with this mode are explored as an additional degree of freedom in the ligand docking process, the “open” conformations produced within the trajectory of this normal mode are found to promote ligand binding poses that are consistent with experimentally verified interactions. These findings connect the dynamic properties characterized by the NMA of the wild-type and highly sensitive mutant receptor with experimentally observed sodium-dependent allosteric effects and suggest a mechanism by which the presence of sodium alters ligand affinity.

Materials and Methods

Reagents. Cell culture media were purchased from HyClone Laboratories (Logan, UT). For radioligand studies, NET-856 (70–80 Ci/mmol) was purchased from PerkinElmer Life and Analytical Sciences (Waltham, MA), and wash buffer reagents were purchased from United States Biological (Swampscott, MA). The source of (-)-quinpirole and forskolin was Sigma-Aldrich (St. Louis, MO). Other drugs were purchased from Tocris Bioscience (Ellisville, MO).

Site-Directed Mutagenesis. Mutagenesis was accomplished using a QuikChange kit (Stratagene, La Jolla, CA). The integrity of mutations and the lack of unwanted mutations were confirmed by full-length sequencing at the University of Maine DNA sequencing facility (Orono, ME). Mutant receptors are named employing the system created by Ballesteros and Weinstein (1995) and other nomenclature conventions. In brief, for each residue in a transmembrane segment, the first digit denotes the transmembrane segment (TMS), followed by a period and a relative position index within the transmembrane segment. The most conserved amino acid in a TMS is assigned the position index 50, and the other amino acids within this TMS are numbered relative to the conserved amino acid. The number in parentheses is the residue number in the sequence of the rat D₂ dopamine receptor short isoform. Our naming system for the mutants begins with a letter to designate the species (e.g., “r” for rat or “h” for human) followed by the receptor subtype abbreviation, e.g., rD₂ for rat D₂ dopamine receptor. Next, the single-letter abbreviation for the amino acid is listed followed by its position and then amino acid substitution. For example, rD₂-V2.61(91)F denotes a rat D₂ receptor with valine at position 2.61(91) being substituted for phenylalanine (Fig. 1).

Transfections. DNA constructs subcloned into the pcDNA3.1 vector were transfected into human embryonic kidney (HEK) 293 cells by CaPO₄ precipitation. In brief, 20 μg of DNA was mixed with 60 μl of 2 M CaCl₂, and the mixture was added to an appropriate volume of sterile water to make 500 μl of solution. This DNA-CaCl₂ solution was then added drop-wise to 500 μl of 2× HEPES-buffered saline while bubbling the HEPES-buffered saline with a 1-ml serological pipet. The resultant final transfection mixture was allowed to sit for 30 min before drop-wise addition to 150-cm² culture dishes seeded with an appropriate number of cells. For stable transfections, HEK293 cells were seeded at a density of 200,000 cells/150-cm² culture dish. For transient transfections, COS7 cells were seeded at a density of 1.5 million cells/150-cm² culture dish. Both cell lines were allowed to grow overnight in 20 ml of sterile growth media containing DMEM supplemented with 10% bovine calf serum, 100 μM sodium pyruvate, and 1% penicillin/streptomycin (5000 units). This resulted in approximately 30% confluent COS-7 cells and less than 5% confluent HEK293 cells. The transfected cells were incubated overnight with the final transfection mixture, after which the media were replaced. Two to 4 h before transfection, the media were replaced with 20 ml of new sterile growth media. Again, the next day, the media were replaced. For the HEK293 cells, the media were supplemented with 2 mg/ml G-418 to allow for clonal selection. Stable clones of HEK293 cells containing the mutant receptor were generated after several weeks of G-418-selective pressure and expanded for further use in radioligand binding and functional assays.

Membrane Preparation. Cell membranes were prepared by first detaching healthy cells with lifting buffer (Dulbecco's phosphate-buffered saline without Ca²⁺ and Mg²⁺; 5 mM EDTA) and then pelleting the detached cells by centrifugation in a sterile conical tube for 10 min at 700g. After centrifugation, the supernatant was decanted from the pellet. The pellet was then resuspended in 10 ml of lysis buffer (5 mM Tris, 5 mM MgCl₂, pH 7.4 at 4°C) and allowed to lyse on ice for 5 to 10 min before transfer to an ice-cold Dounce homogenizer. Eight full strokes of the Dounce homogenizer were used to disrupt the whole cells by glass-on-glass homogenization. The resulting homogenate was poured into a centrifugation tube, balanced by the addition of cold binding buffer (50 mM Tris, pH 7.4 at 4°C), and then centrifuged at 28,000g for 45 min. The supernatant was decanted from the resulting membrane pellet. This membrane pellet was then resuspended in cold binding buffer and recentrifuged. The final membrane pellet was resuspended in an appropriate amount of cold binding buffer for the experiment, rehomogenized by four strokes in an ice-cold Dounce homogenizer, and then stored on ice for same day use.

Radioligand Binding Studies. Both receptor saturation experiments and radiolabeled competition assays were used to characterize the receptors in this study. In brief, the binding and wash buffers consist of 50 mM Tris, pH 7.4, at 25°C, with 1 N KOH used for the fine pH adjustment. For sodium shift assays, the binding and wash buffers were supplemented with 140 mM NaCl. The membrane density of the receptors and their affinity for the radioligand [³H]methylspiperone ([³H]MSP) was assessed by saturation isotherm analysis. For this type of assay, the cell membranes were allowed to equilibrate with increasing nanomolar concentrations of [³H]MSP in the presence or absence of 5 μM(+)-butaclamol, a dopamine receptor antagonist used to define the nonspecific interactions of [³H]MSP. After 90 min of equilibration at room temperature, the samples were rapidly filtrated and washed with ice-cold binding buffer (50 mM Tris, pH 7.4, at 0°C) through GF/C filters pretreated for 10 min with 0.3% polyethylenimine. The filters were allowed to dry before cutting them into vials. Vials were then filled with 3.5 ml of scintillation fluid and mixed before quantifying the amount of radioactivity in a scintillation counter. Radiolabeled competition assays were performed in a similar fashion to saturation assays, except that a fixed concentration of 0.5 nM [³H]MSP was utilized in conjunction with increasing concentrations of nonradiolabeled competitive ligand. Membrane protein concentration was determined by bicinchoninic acid assay (Pierce Chemical, Rockford, IL) according to the manufacturer's instructions.

cAMP Functional Assays. Intracellular cAMP concentrations were determined using a PerkinElmer Fusion plate analyzer and a cAMP Alphascreen detection kit (PerkinElmer Life and Analytical Sciences). The assay was performed essentially according to the manufacturer's specifications, except for adaptations we devised to measure cAMP levels in attached cells. In brief, HEK293 cells stably expressing mutant receptor were resuspended in sterile growth media (DMEM with 10% bovine calf serum, 1% penicillin/streptomycin, 100 μM sodium pyruvate) and then plated at a density of 50,000 cells/well in 96-well microtiter plates coated with poly-l-lysine (Sigma-Aldrich). The next morning, stimulation buffer (DMEM, 20 mM HEPES, 100 μM sodium metabisulfite, 30 μM Ro 20-1724, pH 7.4 at 25°C), cell lysis buffer (0.3% Tween 20, 20 mM HEPES, 1 μg/μl bovine serum albumin, pH 7.4 at 25°C), and bead buffer (20 mM HEPES, 30 μM Ro 20-1724, 1 μg/μl bovine serum albumin, 1× Hanks' balanced salt solution, pH 7.4 at 25°C) were freshly prepared and pH adjusted to 7.4 with 1 N cell culture tested sodium hydroxide (Sigma-Aldrich). To examine the G_i protein-mediated inhibition of adenylyl cyclase, the levels of cAMP were first raised with 6 μM forskolin, a direct stimulator of adenylyl cyclase. Drug dilutions were prepared in stimulation buffer, and 200 μl of dilution was added per well in an empty 96-well microtiter plate and allowed to equilibrate in the 37°C incubator for 30 min. Culture medium was removed from the cells and the temperature- and carbon dioxide-equilibrated drug dilutions were rapidly added to the cells using a multichannel pipet. Cells were then incubated the presence of the drug dilutions at 37°C for 20 min and then centrifuged at 1500g for 5 min. Drug dilutions were carefully removed by pipetting, and 100 μl of cell lysis buffer was added to each well. The cells were lysed by shaking at 600 rpm on a microtiter plate shaker for 1.5 h. A portion of the resulting lysate (30 μl) was transferred to an opaque 96-well Costar plate (catalog no. 07-200-309; Corning Inc., Corning, NY) and challenged with 0.5 units of acceptor and donor beads (9.35 and 12.5 μg/ml, respectively) containing 5 units of biotinylated cAMP (3.76 nM). Before reading, this reaction was allowed to equilibrate for 1 h with shaking at 600 rpm, protected from light with aluminum foil.

Calculations and Data Analysis. Data points for each experiment were sampled in triplicate, and each experiment repeated three times, except where noted. The geometric mean and S.D. are reported for each experiment; however, the errors in the graphs are S.E.M. values. The equilibrium dissociation constant (K_D) of [³H]MSP was determined from saturation isotherm analysis. The inhibition constants (K_i) for all radioligand competition assays were calculated with the Cheng-Prusoff equation: K_i = IC₅₀/(1 + [radioligand]/K_D), where K_D is the equilibrium dissociation constant of the radioligand. A K_0.5 value is reported in cases where the Hill slope is significantly different from unity. All data were analyzed using Prism version 4.0 (GraphPad Software Inc., San Diego, CA). For the inhibition assays, data from three or more assays were combined and then interpreted by extrapolating all dose-response curves to zero to generate IC₅₀ values. These were subsequently converted to K_i values before analysis by one-way ANOVA with a Dunnett's post hoc analysis. For saturation isotherm binding assays, specific binding curves were obtained by subtracting nonspecific binding [defined as binding in the presence of 5 μM (+)-butaclamol] from the total binding at each concentration of radioligand. Values for K_D and B_max were determined from the specific binding curve. Sodium shift binding assays were analyzed by comparing the receptor data with and without sodium in a paired two-tailed Student's t test. cAMP functional assays were assessed by first quantifying the amount of cAMP generated per milligram of protein in each sample and then normalizing this value as a percentage of the cAMP generated by unopposed 6 μM forskolin. Efficacy was determined by subtracting the best fit values for the bottom of the curve (lowest horizontal asymptote) from the top of the curve (highest horizontal asymptote). Functional assays are graphed as sigmoidal semilog dose-response curves. Statistical analyses of the curve fitting procedure included the run test, F test, and Pearson's correlation coefficient. Potency and efficacy values generated from three or more replicate curves were analyzed by one-way ANOVA with a Dunnett's post hoc analysis. Significance was established at the 95% confidence level (p ≤ 0.05).

Construction of Receptor Homology Models. A wild-type D₂ receptor model was constructed using Modeler 9v1 (Sali and Blundell, 1993) simultaneously using as templates the (1GZM) structure of bovine rhodopsin (Li et al., 2004) and the recently determined structure of the β₂ adrenergic receptor (Cherezov et al., 2007) (2RH1). Initially, 1000 D₂ receptor structures were generated and ranked by Modeler's objective function. The models were then structurally aligned and clustered using the GROMACS version 3.3 package (van der Spoel et al., 2005) gcluster utility (cutoff = 0.20 Å). The most representative (central) structure from the best scoring cluster was selected for further analysis and mutated to V2.61F(91) before energy minimization runs.

Building Sodium-Bound D₂ Receptor Models. Sodium-bound models were constructed with the sodium cation placed at the putative sodium binding pocket near D2.50(80) proposed by Neve et al. (2001). One negative control was constructed by positioning the sodium far from the TM region, at the intracellular carboxylate terminus (C415) where it would not affect the dynamic properties through direct interaction with the TM region. A second negative control used in this study was the sodium-free “null” system. All the structures were subjected to energy minimization runs in vacuo with the low-memory Broyden-Fletcher-Goldfarb-Shanno (quasi-Newtonian algorithm for energy minimization) method, in three stages, each carried out to convergence (F_max < 10 kJ/mol/nm) using GROMACS version 3.3 (van der Spoel et al., 2005), with the molecular systems parameterized according to the optimized potentials for liquid simulations all-atom force field (Jorgensen et al., 1996). Electrostatic interactions were treated by the particle-mesh Ewald method (Essman et al., 1995). In the first stage, all heavy protein atoms were restrained with half-harmonic force restraints (k = 1000 kJ/mol/nm), with only the sodium ion unrestrained. In the second minimization stage, sodium and all side chains of residues within 6 Å of the sodium were unrestrained. Finally, sodium and all atoms of residues within 6 Å of the sodium position were unrestrained during minimization, with residues outside of this region restrained.

Normal Mode Analysis. To examine the effect of sodium binding on the dynamic properties of the receptor molecule, we used NMA, which determines a spectrum of independent harmonic (vibrational) motions available to a particular stable molecular conformation within a harmonic approximation (for further description, see Introduction). Our analysis focused on the lower frequency modes that comprise more facile motions (thus, higher amplitude) along directions that coincide with the more shallow curvatures along the potential well (Tama and Sanejouand, 2001) and have been shown to indicate function-related dynamics of proteins (Cui and Bahar, 2006). Each minimized structure was submitted to the NOMAD-Ref Web server for NMA (Lindahl et al., 2006). Elastic network models (Tirion, 1996) were built from C_α positions, with the sodium ion represented as an additional C_α at its optimized binding position. The first 106 normal modes (M_1–106) were calculated for each elastic network model using default parameters, with the exception of the distance-weighting parameter in which a nondefault value of 3.0 Å was applied as recommended for C_α-only models.

To examine divergence in dynamical behavior between sodium-bound and control (null) structures, dot products were computed for each sodium-bound normal mode vector against all computed modes from the null system. A window of the null spectrum was then chosen by centering it at the most analogous mode (highest dot product with ). The dot product squares were then summed over the selected null spectrum window to provide a P_ij value (). A range of vectors from the control structures must be included because what might appear as a unique sodium-bound mode can be recapitulated by a set of such vectors, when combined. This overcomes a potential pitfall in comparative NMA resulting from a comparison of only pairs of corresponding modes between structures (Ming and Wall, 2005). P_ij was then plotted as a function of using various window sizes to identify the difference in normal mode(s) between sodium-bound structures and control.

Ligand Docking into D₂ Receptor Conformers from the Trajectory. The trajectory of the sodium-sensitive was selected to explore effects of sodium-related motions on ligand docking because: 1) its P_ij value demonstrates significant divergence from the conformational space of the other modes; 2) it was the lowest frequency mode among those showing divergence and represents a softer and, thus, higher amplitude collective domain motion; and 3) visual inspection revealed that as part of its characteristic motion, an open phase appears to widen the binding pocket, increasing accessibility (from the extracellular milieu) and its volume in the V2.61(91)F mutant. Therefore, we constructed a series of D₂ receptor models representing points in the trajectory of this particular mode by rebuilding all-atom structures on the C_α frames from the M₁₉ trajectory (output from NOMAD-Ref). Backbone and side chain atoms were built onto the fixed C_α template with Modeler 9v1, followed by minimization, a short 15-ps MD run, and minimization with C_α atoms fixed for each procedure. The structure was then minimized to convergence with positional restraints on the C_α carbons in GROMACS, as described under Materials and Methods.

Ligands were constructed in Discovery Studio (Accelrys, San Diego, CA), and their geometries were optimized with ab initio quantum mechanical calculations using the HF6–31G^** basis set in Gaussian03 (Frisch et al., 2004). The partial charges were set according to the AutoDockTools 1.4.5 automated Gasteiger partial charge assignment. Ligands were then docked 50 times into each M₁₉-based D₂ receptor frame using the Lamarckian genetic algorithm search routine in AutoDock 4.0 (Morris et al., 1998), with default parameters and a maximal number of energy evaluations of 4.0 × 10⁶. Selective side chain flexibility was allowed within the docking routine; we explored the rotation of dihedrals in the side chains of F2.61(91), F3.28(110), V3.29(111), D3.32(114), W6.48(358), F6.51(361), and H6.55(375), chosen because of their apparently critical position in the ligand binding site. After docking into each receptor frame, poses were clustered using an RMSD tolerance of 3.0 Å, and clusters were ranked by mean energies.

Results

When transiently expressed in COS-7 cells, the D₂-V2.61(91)F mutant receptor had a level of expression and an affinity for the moderately D₂-selective 1,4-DAP [³H]MSP that was similar to the wild-type D₂ receptor (1.1- and 4.1-fold, respectively) (Table 1; Fig. 2). It also had an affinity comparable with that of the wild-type receptor for the agonist (-)-quinpirole, which lacks a 1,4-DAP structural motif (1.2-fold, Table 1). The 1,4-DAP structural motif consists of two aromatic rings linked to positions 1 and 4 of a central six-membered piperidine or piperazine ring. L-745,870, for example, has two distinct aryl substituents that extend from positions 1 and 4 of the central piperazine ring directly and via a methylene spacer (Table 1). One of these ring nitrogens is likely protonated at physiological pH and interacts with the acidic pocket residue D3.32(114). For all ligands tested within this structure class, the D₂-V2.61(91)F mutant had affinities very similar to those measured for the wild-type D₂ receptor, including L-745,870 and six other D₄-selective 1,4-DAPs (1.1–3.3-fold changes, Table 1). All of these 1,4-DAPs have been tested in previous binding studies designed to investigate molecular determinants of ligand selectivity for D₄ receptors versus D₂ receptor subtypes (Schetz et al., 2000; Kortagere et al., 2004), but only one of them (L-745,870 or CPPMA) had been tested on the D₂-V2.61(91)F mutant, and it was reported to have a large improvement in binding affinity (Simpson et al., 1999).

In an effort to determine the source of the discrepancy between the reported 97-fold increase in affinity for L-745,870 for an N- and C-terminally epitope-tagged human D₂-V2.61(91)F mutant (Simpson et al., 1999) and the lack of change for the identical mutation in the rat receptor for the same ligand (Table 1) and several other ligands belonging to the same structural class (Table 1; Floresca et al., 2005), we systematically eliminated differences between the two experimental systems. Initially, we expressed our rat D₂-V2.61(91)F mutant in the same HEK293 cell line used in the previous report. Changing the cell background had only a moderate effect on the relative affinity measured for L-745,870 at the rD₂-V2.61(91)F mutant versus the wild-type rD₂ receptor that was difficult to accurately quantify because of low affinity and limited drug solubility under the conditions tested (∼5.5-fold increase based on estimates from extrapolated values, Fig. 2, A and B; Table 2).

The stable expression of these receptors in the HEK293 cell line also allowed us to study D₂-like dopamine receptor cAMP functional responses. Although useful for the initial binding studies, transient expression in a COS-7 cell line lacked a suitable functional response in our assay system (data not shown). The HEK293 cell line lacks endogenous receptors for dopamine but can mediate a cAMP functional response for transfected dopamine receptors (data not shown). In preparation for functional assays, the cell surface density of dopamine receptors expressed in several stable HEK293 clones was determined with [³H]MSP saturation isotherm analysis, after which selected clones were matched by receptor density to avoid discrepancies because of spare receptors. HEK293 cell lines expressing 8.6 ± 3.4 pmol/mg protein of the wild-type D₂ receptor and 11.6 ± 0.23 pmol/mg protein of the D₂-V2.61(91)F receptor were selected for use in all subsequent experiments. The corresponding [³H]MSP affinity values (K_D) were 74 ± 9.7 pM for the wild-type D₂ receptor and 95 ± 18 pM for the D₂-V2.61(91)F receptor. The absolute affinities of [³H]MSP for the D₂-V2.61(91)F mutant receptor expressed in COS-7 and HEK293 cells were similar (1.4-fold different), although small differences were found for the wild-type receptor (4.6-fold different). For this same reason, the relative differences between wild-type and mutant D₂ receptors within the same cell line are more pronounced in COS-7 than in HEK293 cells (4.1- versus 1.3-fold, respectively).

Our next step was to assess whether differences in the binding buffer could account for the large difference observed between our initial COS-7 binding data (Table 1) and published HEK293 binding data from Simpson et al. (1999). The striking finding was that the addition of 140 mM sodium chloride to the binding (and wash) buffer, to mimic the buffer conditions in the previous report (Simpson et al., 1999), resulted in a large (35-fold) increase in L-745,870 affinity for the rD₂-V2.61(91)F mutant receptor (Fig. 2B; Table 2), although the wild-type D₂ receptor displayed only a very limited sodium sensitivity for this ligand (∼7-fold, Fig. 2A; Table 2). These affinity values in the presence of sodium are consistent with those for hD₂ wild-type and hD₂-V2.61(91)F published in an earlier report (920 ± 200 and 9.5 ± 4.0 nM, respectively, Simpson et al., 1999). A similarly strong pattern of sodium-sensitive binding is evident for L-750,667, the iodinated derivative of L-745,870 (Fig. 2, C and D; Table 2); the affinity of L-750,667 for the rD₂-V2.61(91)F mutant was increased 37-fold in the presence of 140 mM sodium, even though there is only a moderate sodium sensitivity for this ligand at the wild-type receptor (>∼7-fold). Another D₄-selective 1,4-DAP, NGD 94-1, chosen because its structure is less similar to L-745,870 and L-750,667 (Table 1), still displayed comparably enhanced affinity for the D₂-V2.61(91)F mutant in the presence of sodium (Fig. 2, E and F; Table 2) but only a very small sodium-dependent increase in affinity for the wild-type receptor (Fig. 2E; Table 1). Similar patterns of sodium sensitivity were also observed for several other D₄-selective 1,4-DAPs and the sodium-sensitive substituted benzamide antagonist, (-)-raclopride (Fig. 3), and these patterns were not mimicked by the same concentration of N-methyl-d-glucamine, a sodium replacement ion (Fig. 3). Note that differences in (in the absence of sodium) affinities appear in Tables 1 and 2, but these are not comparable. The affinities for expressed wild-type and mutant D2 receptors in Table 1 are from membranes isolated from COS7 cells, whereas those in Table 2 are from membranes isolated from HEK293 cells.

In whole-cell attached functional assays, the full agonists dopamine and (-)-quinpirole both strongly reversed forskolin-stimulated increases in cAMP for both wild-type and D₂-V2.61(91)F mutant receptors (Fig. 4A). Spiperone was able to fully reverse the (-)-quinpirole-stimulated inhibition of cAMP (Fig. 4B), but neither (-)-quinpirole nor L-750,667 were able to reduce forskolin-stimulated cAMP accumulation in untransfected HEK293 cells (data not shown).

It is notable that we found that L-750,667 has partial agonist properties at the wild-type D₂ receptor (Fig. 4A) but acts as an antagonist at the D₂-V2.61(91)F mutant receptor (Fig. 4B). However, the chlorinated derivative, L-745,870, has no agonist properties at either receptor, whereas NGD 94-1 has agonist properties at both receptors (Fig. 4A). Concentrations of L-745,870 five times higher than that shown in Fig. 4 did not change its functional profile (data not shown). Despite these drugs having enhanced affinity in the presence of a high concentration of sodium (Fig. 2), no attempt was made to measure such sodium effects in the functional assays.

Dynamic Properties of the Receptor Constructs. NMA performed on molecular model constructs of the D₂ receptor was used to examine the dynamic response to sodium ion binding. In an effort to characterize changes in the receptor's harmonic motions imparted by the perturbation of sodium ion, we compared the spectrum of normal modes calculated for: 1) the sodium-bound structure [near D2.50(80)], 2) a negative control with sodium ion bound outside the protein core (at the carboxylate terminus), and 3) a system without any sodium, hereafter referred to as the null system. To ascertain which intrinsic low-frequency (and relatively high-amplitude) motions are sensitive to the perturbation from sodium binding within the receptor, we calculated the sum of dot product squares, (see Fig. 5) for each sodium-bound normal mode () over ranges (windows) of modes in the null spectrum (for details, see Materials and Methods). A plot of P_ij as a function of sodium-bound normal mode vectors () indicates that sodium ion binding influences the characteristic movements of only a few low-frequency modes. P_ij values < 0.9 are indicative of motions different from those described by the normal modes of the null system. Such values were observed for modes 19, 26, 29, and 31 in the sodium-bound structure, hereafter referred to as , , and so forth. We explored different window sizes (window = 7, 15, 21, and 106) in the null spectrum to see whether divergent could be recapitulated (P_ij ∼ 1.0) by comparison against larger windows of the null spectrum. , however, consistently provided P_ij values significantly lower than the other modes. P_ij calculations for the negative control spectrum () against the null spectrum (not shown) do not yield uniquely low P_ij values comparable with , which also suggests that the divergence is site specific in regards to sodium perturbation. had no clearly equivalent mode in the entire spectrum calculated for the null system. The modes and were most analogous, with values of 0.25 and 0.36, respectively. Together, these findings suggest that the different motion described by is because of the effect of sodium binding on the dynamic properties of the receptor molecule.

It is interesting that visual inspection of the trajectory shows that it deforms the binding pocket and as a result facilitates the potential contact of V2.61(91)F with the ligand (Fig. 6, A and B). The trajectory describes a concerted TM2 kinking motion at the proline kink (P2.59) along with the lateral TM3/TM4 motion away from the cleft in the plane of the bilayer, which could potentially disrupt the TM2-TM3 interhelical packing to expose V2.61(91)F for interaction with the ligand. Therefore, sodium induction of this particular mode must be considered relevant to the observed enhancement of L-745,870 affinity in the D₂-V2.61(91)F mutant receptor.

The dynamic response of the receptor to sodium binding, as measured by the divergence between normal mode spectra from sodium-bound and null structures, agrees with dynamic perturbation studies in other systems showing that significant changes in conformational equilibrium can be triggered by what would appear to be only small perturbations at “dynamical control points” (Ming and Wall, 2005, 2006). In the D₂ receptor, the allosteric site is likely coupled to instabilities near the proline kink in TM2. Despite sequence divergence for the templates rhodopsin and β₂-adrenergic receptor in this region, the TM2s are virtually superimposable. In the β₂-adrenergic receptor, an H-bond between W3.28 and the backbone carbonyl oxygen of V2.57 stabilizes the Pro kink. In our homology model of the D₂ receptor, the corresponding interactions are missing because of differences in sequence, but the dynamics of the TM2 region could still be coupled to the sodium binding site near D2.50(80) via intrahelical H-bonding and/or local conformational arrangements involving interhelical (TM2–3) side chain packing, interhelical H-bonding, or structural waters near the sodium binding pocket. To examine how dynamic flexibility expressed in the various receptor conformations visited by the trajectory may influence L-745,870 binding in the D₂-V2.61(91)F mutant receptor, we developed a docking protocol to explore a series of conformations determined by this trajectory.

Ligand Docking Guided by Conformations from . Based on preliminary results of L-745,870 docking into our initial model of the D₂ receptor (frame 0), the docking poses with the highest ranking lacked the expected interactions: π-stacking between the p-chlorophenyl moiety of L-745,870 and V2.61(91)F (Fig. 6, A and B) and the H-bond reinforcement of the ionic interaction between the piperazine ammonium proton and D3.32(114). This is because in the “closed” conformation of the D₂ receptor model, V2.61(91)F is adjacent to D₂-F3.28(110), with which it can maintain stable interactions (π-stacking or T-type interactions) that reduce the ability of the ligand to interact with it. In addition, the cleft in which the protonatable amine of the piperazine ring must fit to achieve this salt bridge is effectively occluded by residues F3.28(110), V3.29(111), and F6.51(361). These observations agree with our experimental data showing low affinity for L-745,870 in wild-type D₂ and D₂-V2.61(91)F in the absence of high concentrations of sodium.

To explore the changes in ligand binding attributable to the dynamic effects produced by sodium, we used the trajectory to construct as described under Materials and Methods, 15 receptor conformers (frames -7 to +7) representing the structure of the receptor along the motion described by (Fig. 7A). Each of these conformations was used to dock L-745,870. The protocol involves 50 separate dockings into each of these conformations, and the resulting ligand poses generated within each frame were binned into clusters based on similarity of binding position and orientation (Fig. 7B). Receptor frames -1 and -2 in Fig. 7A represent the closed interval (-7 being maximally closed) (see Fig. 6, A and B), and +1 through +7 represent the “opened” frames (+7 being maximally opened) (Fig. 7B). Because the amplitude of the trajectory was set arbitrarily in the NMA calculation, we were most interested in the smallest backbone movements away from the initial structure (±1, 2, 3) rather than the extremes (±7). Docking of L-745,870 into the various frames of the receptor oscillating along the harmonic motion shows that even a minimal excursion into the open phase of M₁₉ better accommodates the expected ligand binding poses (see Kortagere et al., 2004) and, hence, facilitates a direct interaction between either of the ligand's aryl moieties and the phenyl ring in the V2.61(91)F mutant. Within these receptor conformations visited by the trajectory, L-745,870 can achieve reasonable binding geometries that are not available in the static D₂ receptor model. It is notable that proper ligand accommodation in the open frames is because of the increase in the TM2 proline kink bend angle and TM3 lateral translation and increased accessible depth of the binding cleft that is noticeable in Fig. 7A (note the position of the gray regions that depict regions favorable to ligand occupancy). As the receptor opens (frames +1 and +2), this ligand occupancy region moves intracellularly into the vicinity of D3.32(114) and W6.48(358). Thus, the movement provides increased space for the ligand to access the entire binding cleft spanning from TM2,3,7 to TM3,5,6. Therefore, while maintaining the previously identified contacts in the D₂ receptor, L-745,870 can associate with a series of TM cleft-lining residues including F2.61(91), L2.64(94), C3.25(107), F3.28(110), V3.29(111), D3.32(114), V3.33(115), C3.36(118), F5.38(189), V5.39(190), S5.42(193), F5.47(198), W6.48(358), F6.51(361), T7.39(386), G7.42(389), and Y7.43(390) and potentially with e2 loop residues L143(171), E153(181), C154(182), I155(183), and I156(184) (when the loop is replaced after docking). This enhancement of the binding pocket by the effect of sodium binding on the dynamic properties of the receptor makes some previously inaccessible binding sites available to the ligand and results in the higher affinity observed in the presence of sodium.

(-)-Raclopride was also docked into the wild-type D₂ receptor conformations visited by the trajectory (Fig. 8). We find that receptor conformations generated along the opening path of this motion also increase the likelihood of proper binding interactions with (-)-raclopride; an H-bond-reinforced ionic interaction is achieved with D3.32, and the ligand's phenyl substituent penetrates more deeply into the primary binding pocket situated among TM 3, 5, and 6. These findings support our observation that (-)-raclopride affinity for wild-type D₂ receptor is enhanced in the presence of sodium (Fig. 3).

Discussion

Our finding that the affinity of L-745,870 for the rD₂-V2.61(91)F mutant is drastically increased in the presence of high sodium concentration (140 mM) is significant for two main reasons. First, it explains the discrepancy (∼100-fold differences) in the reported affinities of L-745,870 and several other structurally similar D₄-selective 1,4-DAPs for the D₂-V2.61(91)F mutant and thus resolves an apparent contradiction in the literature (Simpson et al., 1999; Floresca et al., 2005). Second, it demonstrates how key molecular interactions between a ligand and a specific GPCR microdomain are influenced by occupancy of an allosteric site. In our case, the interactions that become accessible through the allosteric effect of sodium binding involve a π-stack or T-type interaction between the ligand's aryl moiety and F2.61(91) of the receptor as we proposed previously (Kortagere et al., 2004). However, we found that in the D₂-V2.61(91)F receptor, this favorable interaction can occur only if an interhelical π-stack between the F2.61(91) and the adjacent F3.28(110), which forms a “hydrophobic brace,” is disrupted. We show that this disruption, achieved by sodium occupancy near D2.50(80), induces new dynamics that appear to widen the junction between TM2 and TM3 in the extracellular region of the receptor and thus disrupt the F2.61/F3.28 interaction. It is important that each of the 1,4-DAP compounds tested here (L-745,870, L-750,667, NGD 94-1, RBI-257, PD168,077, FAUC213, and Ro61-6270) is predicted from molecular models to engage in similar interactions when phenylalanine occupies position 2.61(91); in agreement, we found experimentally that all compounds tested display drastic sodium-dependent increases in affinity for the D₂-V2.61(91)F mutant.

Although previous studies have revealed the affinity relationship between the discriminant structural features of 1,4-DAPs and the 1,4-DAP D₄/D₂ selectivity-conferring positions 2.61(91), 3.28(110), and 3.29(111), the effects of mutations in this microdomain on the activity of 1,4-DAP ligands had not been described previously (Simpson et al., 1999; Schetz et al., 2000; Kortagere et al., 2004; Floresca et al., 2005). Therefore, because the D₂-V2.61(91)F receptor exhibits sodium-sensitive affinity changes to 1,4-DAPs, we examined here the functional properties of L-745,870, L-750,667, and NGD 94-1 at the wild-type and D₂-V2.61(91)F receptor. In contrast to the reports showing reduction in agonist affinities as a result of allosteric modulation by sodium, we report here sodium-dependent enhancement of affinities for the 1,4-DAP agonists (and antagonists) similar to that observed for the substituted benzamide antagonists (Neve, 1991). Moreover, we report here that although the D₂-V2.61(91)F mutant can be activated by dopamine and (-)-quinpirole, it cannot be activated by L-750,667, which exhibits weak partial agonist properties on the wild-type D₂ receptor. Rather, L-750,667 acts as an antagonist of the D₂-V2.61(91)F mutant. The relatively small influence of the D₂-V2.61(91)F mutation on the binding affinity of (-)-quinpirole and methylspiperone demonstrated here suggests that neither ligand is likely to directly contact 2.61(91). Furthermore, this mutation had little effect on the activation of the receptor by (-)-quinpirole and its reversal by methylspiperone. This suggests that the 2.61(91)-3.28(110) hydrophobic brace can prevent the receptor from being activated by some agonists if their binding brings them near position 2.61(91). Because the affinities of other 1,4-DAPs, which retained their agonist properties at the D₂-V2.61(91)F mutant receptor, were also enhanced in the presence of sodium, we reviewed the literature concerning the connection between agonist high- and low-affinity states, G protein coupling, and sensitivity to sodium and GTP.

In broken membrane preparations, high concentrations of either sodium ions (120 mM) or GTP and its related analogs (100 μM) decrease the affinity of agonists for catecholaminergic GPCRs (Paris et al., 1989; Neve, 1991; Neve et al., 2001; Schetz and Sibley, 2001). However, the similarities between the agonist affinity shift for sodium and GTP are coincidental because sodium binds an allosteric site on dopamine receptors accessible from the intracellular side (Neve, 1991; Neve et al., 2001; Schetz and Sibley, 2001), whereas GTP binds G proteins that then complex with the receptor at a different intracellular allosteric site. Under conditions (e.g., no GTP) in which a ternary complex is formed (receptor + G protein), kinetic studies of detergent-solubilized or broken membrane α adrenergic receptors demonstrate that both sodium ions and Gpp(NH)p accelerate the rate of agonist dissociation, and a synergistic increase in the rate is observed in the presence of both of these modulators (Limbird et al., 1982). In broken membrane equilibrium studies of α adrenergic receptors, agonist affinity decreases in the presence of either sodium or Gpp(NH)p, and a further decrease in affinity is observed in the presence of both of these modulators. These results were cited as evidence that sodium ions do not compete with GTP for its binding to G proteins (Limbird et al., 1982). Later studies by the same group demonstrated that substitution of the negative charge of the conserved aspartic acid in TM2 [D2.50(79)] to a neutral asparagine in the α2 adrenergic receptor resulted in a total loss of epinephrine's sensitivity to sodium (Horstman et al., 1990). When tested for its ability to stimulate GTPase activity, this same mutant receptor had a 7.5-fold decrease in potency for the agonist UK14304 with no change in efficacy (Ceresa and Limbird, 1994) and no change in affinity relative to the wild-type receptor. Increasing concentrations of Gpp(NH)p decrease the agonist p-[¹²⁵I]iodoclonidine's binding to the wild-type but not the mutant receptor (Ceresa and Limbird, 1994). These data suggest that the sodium-insensitive D2.50(79)N mutant α adrenergic receptor can still couple to G proteins but that the state of coupling (coupled or uncoupled) no longer influences agonist high- and low-affinity states of the receptor. Thus, it would appear that, at least in the case of α2 adrenergic receptors, sodium ions control the agonist high- and low-affinity states of the receptor but do not uncouple G proteins from their receptor.

Broken membrane equilibrium studies demonstrate that dopamine's affinity for the D₂ receptor is decreased when pertussis toxin or GTP uncouple G proteins from the receptor and that the addition of sodium ions further decreases dopamine's affinity (Neve et al., 1989). This was cited as evidence that sodium ions interact directly with dopamine receptors. Later studies by the same group demonstrated that substitution of the conserved aspartate in TM2 [D2.50(80)] for alanine results in a mutant receptor whose affinities for agonists and substituted benzamide antagonists are sodium insensitive (Neve et al., 1991); however, this same mutant receptor is unable to efficiently couple to G proteins, making it impossible to carry out the same types of studies as described above for α2 adrenergic receptors. Finally, the rate of chemical modification of D₂ dopamine receptors by the thiolreactive agent N-ethylmaleimide is altered in the presence of sodium (Neve, 1991), and this is consistent with our finding [utilizing the D₂-V2.61(91)F mutant receptor as a molecular probe of the 1,4-DAP binding pocket] that the binding of sodium ions to the allosteric site [D₂-D2.50(80)] causes significant conformation changes in the receptor in the region of the orthosteric binding pocket.

The changes in dynamic properties of the receptor that can explain the observed pharmacological consequences of sodium binding are illustrated by our findings from NMA. Thus, the presence of sodium was found from this analysis to give rise to a unique, high-amplitude (low-frequency) normal mode that is not observed for the wild-type structure. Following the motions represented by this unique, sodium-related mode shows that it produces a wider cleft opening in the region responsible for ligand binding. These “opening dynamics” allow larger ligands to be accommodated in the pocket for ligand recognition and make room for better interaction of the ligand with the sites in the receptors at which it forms H-bonds, aromatic interactions, and hydrophobic matches. This leads to the measured improvement in the ligand's affinity for the receptor. Specifically, the effects of sodium binding near D2.50(80) on the dynamics in the microdomain surrounding the 2.61(91) position were shown by NMA to involve an increase in the scissors-like helical movements of the extracellular portions of TMs 2 and 3. This motion accounts for the enlarged opening and, consequently, for the large affinity changes observed at the D₂-V2.61(91)F mutant, where these movements disrupt the hydrophobic brace formed by the interaction between V2.61(91)F and F3.28(110). The result of these local rearrangements is a receptor conformation more suitable for accommodating a ligand like L-745,870, which can now make stabilizing interactions as found in the optimal docking poses. Similar sodium-enhanced motions are predicted to occur in the wild-type receptor, which lacks a hydrophobic brace, but cleft opening would lead to better ligand access to the conserved aspartic acid at D3.32(114) for a reinforced ionic bond interaction with the protonatable piperizinyl amine portion of the docked 1,4-DAP. This prediction is supported experimentally by our findings for the wild-type D₂ receptor whose affinities for L-745,870, L-750,667, and NGD 94-1 are somewhat increased (∼3–7-fold) in the presence of sodium. Moreover, the docking of the substituted benzamide (-)-raclopride, which belongs to a distinct structural class of D₂ antagonists, into conformations along the sodium-induced normal mode produced very similar results. Again, suitable binding poses were more readily achieved in wild-type D₂ receptor conformations visited along the open phase of the sodium-induced mode (Fig. 8), thus accounting for the enhanced affinity for (-)-raclopride (Fig. 3) and other substituted benzamides at the D₂ receptor when modulated by sodium (Neve, 1991).

How these dynamic changes affect ligand binding is clearly demonstrated by exploring the binding poses of L-745,870 in a series of receptor conformers built along the characteristic vector of . It is notable that involving NMA in a docking protocol to provide a basis for exploring collective backbone movements in ligand docking has also been used successfully to reproduce known conformations from ligand-bound crystal structures (Lindahl and Delarue, 2005). Here, we found that docking into conformations generated from the open interval of accommodates L-745,870 in the expected binding modes for 1,4-DAPs in D₄ dopamine receptors (Kortagere et al., 2004) because the opening motion disrupts contact between F2.61(91) and F3.28(110), which in turn creates access to these residues that appeared inaccessible to the p-chlorophenyl (“mode 1”) or pyrrolopyridinyl group (“mode 2”) of L-745,870 in the initial D₂ receptor homology model. The access is established by the dynamic changes and enables favorable aromatic-aromatic interactions between the ligand moieties and the receptor sites. This result suggests, therefore, that the specific perturbation propagated from the sodium binding site influences ligand affinity by dynamically changing the interactions in the orthosteric site. The backbone movement observed in our model might capture the phenomenon of sodium allostery by achieving the high-affinity receptor conformational state along the vector path and/or by disrupting the brace that stabilizes the low-affinity state without a significant, time-averaged, conformational change, i.e., through the sodium-induced amplitude increase along this vector (Cooper and Dryden, 1984; Popovych et al., 2006). In any case, the presence and effect of endogenous allosteric modulators (e.g., sodium, G protein, GPCR dimer partner, etc.) should be considered when selecting a receptor representation for in silico library screening. It is also important to point out that the use of target receptor conformations based on normal mode trajectories could inform the design of drugs working through allosteric mechanisms.