Abstract
N-Methyl-d-aspartate (NMDA) receptors play key roles in both physiological processes, particularly synaptic plasticity, and in neuropathological states such as epilepsy and acute neurodegeneration. R-(R*,S*)-α-(4-Hydroxyphenyl)-β-methyl-4-(phenyl-methyl)-1-piperidine propanol (RO 25-6981), is a high-affinity and selective blocker of NMDA receptors containing the NR2B subunit. Using site-directed mutagenesis, [3H]RO 25-6981 binding, Xenopus oocyte voltage-clamp recordings, and molecular modeling, we have identified several critical residues involved in the RO 25-6981 binding site within the N-terminal LIVBP-like domain of the human NR2B subunit. Two mutations, NR2B(D101A) and NR2B(F176A), resulted in a complete loss of [3H]RO 25-6981 binding and also abolished the high-affinity RO 25-6981-mediated inhibition of NMDA-induced currents. The mutation NR2B(T233A) led to a marked reduction in binding affinity by 13-fold. Mutations F182A, D104A, or K234A had a more moderate influence on the binding affinity (KD values increased by 8-, 7-, and 6-fold, respectively). In a three-dimensional model of the NR2B LIVBP-like domain based on the X-ray crystal structure of the amino-terminal domain of the mGlu1 receptor, the critical residues are located in the central cleft where interaction with RO 25-6981 may stabilize the closed structure of the domain. Our results suggest that the three amino acids Asp-101, Phe-176, and Thr-233 are important molecular determinants for the high-affinity binding of RO 25-6981 to the LIVBP-like domain of human NR2B. A possible binding mode for RO 25-6981 is proposed.
N-Methyl-d-aspartate (NMDA) receptors are a major class of ionotropic glutamate receptors mediating fast excitatory neurotransmission in the central nervous system (Dingledine et al., 1999). NMDA receptors are of particular interest due to their important physiological roles in neuronal development, synaptic plasticity, and learning and memory. Furthermore, overactivation of these receptors is thought to play a major pathological role in excitotoxic neuronal cell death after acute ischemic and traumatic brain injury (Choi, 1988). NMDA receptors are unique among ligand-gated ion channels in their requirement for an obligatory coagonist, glycine, in addition to the synaptically released neurotransmitter, glutamate, for receptor activation (Johnson and Ascher, 1987; Kleckner and Dingledine, 1988). The glutamate and glycine agonist binding sites, together with the many regulatory sites, have attracted attention as possible sites for pharmacological modulation of receptor function (Kemp and Kew, 1998).
NMDA receptors are heteromeric structures composed of an obligatory NR1 subunit and one or more of four NR2 subunits (NR2A–D) (McBain and Mayer, 1994). NR1 and NR2 subunits share a common membrane topology with other members of the ionotropic glutamate receptor (iGluR) family (Dingledine et al., 1999). Each subunit has a large extracellular amino-terminal domain, three transmembrane segments (TM1, TM3 and TM4), a pore lining P-loop region (initially identified as TM2) that loops into the membrane with both ends facing into the cytoplasm, an extracellular loop (L3) located between TM3 and TM4, and an intracellular carboxy terminal domain.
O'Hara et al. (1993) observed a structural similarity between the extracellular amino terminal domain of the metabotropic glutamate receptor mGlu1 and a family of bacterial periplasmic amino acid-binding proteins, in particular, the leucine, isoleucine, valine periplasmic amino acid-binding protein (LIVBP). Based on the crystal structure of LIVBP, they proposed a “Venus flytrap” model for the binding of glutamate to mGlu receptor. Subsequently, Stern-Bach et al. (1994) identified two discontinuous segments (S1, located on N-terminal adjacent to TM1, and S2, on L3 region) of the iGluR family, which are structurally related to the bacterial glutamine binding protein (GlnBP) that seem to form the glutamate binding pocket. The X-ray crystal structure of iGluR2 receptor S1-S2 segments in complex with the agonist kainate has since confirmed this GlnBP-like model of the glutamate binding site (Armstrong et al., 1998). Likewise, the glycine binding site on the NR1 subunit (Kuryatov et al., 1994; Wafford et al., 1995; Kew et al., 2000) and glutamate binding site on NR2A and -2B (Laube et al., 1997; Anson et al., 1998) were shown to exhibit structural homology to the two lobes of the bacterial lysine, arginine, ornithine binding protein (LAOBP), a close structural neighbor of GlnBP. Hence, a Venus flytrap mechanism for the binding of these coagonists to their respective NMDA receptor subunits was proposed and has recently been confirmed by the description of the cocrystal structure of the NR1 S1-S2 segment with glycine (Furukawa and Gouaux, 2003).
Interestingly, the first 380 amino acids of the N-terminal domain preceding the agonist binding domain of NMDA receptor subunits has structural similarity to the bacterial LIVBP. Using molecular modeling and a mutagenesis scan, molecular determinants of high-affinity Zn2+ inhibition have been localized to the LIVBP-like domain of the NR2A subunit (Choi and Lipton, 1999; Fayyazuddin et al., 2000; Low et al., 2000; Paoletti et al., 2000).
Ifenprodil is an atypical noncompetitive antagonist of NMDA receptors that exhibits approximately 400-fold higher affinity for NR1/NR2B than for NR1/NR2A heteromeric receptors (Williams, 1993). Ifenprodil is neuroprotective both in in vitro (Graham et al., 1992) and in in vivo (Gotti et al., 1988) models of ischemia. It has been shown to antagonize the NMDA receptor via a novel state-dependent mechanism (Kew et al., 1996). Ifenprodil was originally thought to bind to NMDA receptor at the polyamine site (Carter et al., 1990); however, other evidence pointed to a distinct site (Reynolds and Miller, 1989; Gallagher et al., 1996), and it has since been demonstrated that ifenprodil and polyamines bind to separate sites that exhibit an allosteric interaction (Kew and Kemp, 1998). The ifenprodil binding pocket has recently been mapped to the LIVBP-like domain of the NR2B subunit (Perin-Dureau et al., 2002). However, in the absence of radioligand binding data, it was not possible to discriminate definitively between effects of mutations on ligand binding and downstream signal transduction.
RO 25-6981 is an ifenprodil analog with improved selectivity and affinity for NR2B-containing NMDA receptors (Fischer et al., 1997; Mutel et al., 1998). In the present study, we have used site-directed mutagenesis together with both binding and functional assay and molecular modeling to identify critical residues in the RO 25-6981 binding site within the N-terminal LIVBP-like domain of the human NR2B subunit.
Materials and Methods
Materials. [3H]R-(R*,S*)-α-(4-Hydroxyphenyl)-β-methyl-4-(phenyl-methyl)-1-piperidine propanol ([3H]RO 25-6981) was synthesized in the isotope laboratory of F. Hoffmann-La Roche (Basel, Switzerland) by Dr. H. Harder and was prepared as a fumarate salt with a specific activity of 27 Ci/mmol. RO 25-6981, dextromethorphan (RO 01-6794), and 1-(4-chlorophenyl)-2-methyl-6-methoxy-7-hydroxy-1,2,3,4-tetrahydroisoquinolin (RO 04-5595) were synthesized at F. Hoffmann-La Roche. dl-Amino-5-phosphonopentanoic acid was obtained from Sigma-Aldrich (St. Louis, MO). 1,3-di-o-Tolylguanidine (DTG) and NMDA were from Sigma/RBI (Rahn AG, Zürich, Switzerland).
Isolation of cDNAs and Construction of Point Mutants. hNR1c (also termed NMDAR1-2a or NR001) and hNR2B cDNAs were amplified by PCR from a human fetal brain cDNA library in pCMV.SPORT2 (Invitrogen, Carlsbad, CA). PCR reactions were performed using Pfu Turbo DNA polymerase (Stratagene, La Jolla, CA). For PCR, the primer set corresponding to nt 172 to 2988 of human NR1c sequence (accession no. D13515) and primer set corresponding to nt 210 to 4664 of human NR2B sequence (accession no. U90278) were used. In the design of both forward primers of hNR1c and hNR2B, a kosak consensus sequence (CCACC) was introduced upstream from ATG. The PCR fragments generated this way were cloned using pScript PCR cloning kit (Stratagene) according to the manufacturer's instruction. The nucleotide sequence of the inserts was determined by automated cycle sequencing (Applied Biosystems, Foster City, CA). The 2.8-kilobase pair hNR1c and 4.5-kilobase pair hNR2B fragments were then subcloned into the eukaryotic expression vector pCMV. The isolation of rat NR1c and NR2A cDNAs was described previously (Sigel et al., 1994). The cDNA clone for the rat NR2B was obtained from Prof. S. Nakanishi (University of Kyoto, Kyoto, Japan). All the point mutations in hNR2B subunit were constructed using QuikChange site-directed mutagenesis kit (Stratagene) as described previously (Malherbe et al., 2001). The entire coding regions of all the point mutants were sequenced from both strands using an automated cycle sequencer (Applied Biosystems).
Cell Culture, Large-Scale Transient Transfection, and Membrane Preparation. Human embryonic kidney (HEK)-293 cells were adapted to grow in suspension in spinner flask at 95 to 105 rpm. For transfection experiments, a modified medium (DHI special, Invitrogen), with respect to the amino acid content, was used to minimize the glutamate activation of the cells during post-transfection. For the gene delivery, the new transfection reagent RO-1539 (Roche Diagnostics, Roche Applied Science, Rotkreuz, Switzerland), consisting of substances A and B, was used. Cells were cultured to a density of 6 to 10 × 105 cells/ml, centrifuged for 3 min at 600 rpm, and resuspended in glutamate-free DHI special medium. The cell density was adjusted to 5 × 105 cells/ml, and the culture was incubated for at least 3 h before transfection. The transfection complexes were generated in 1/10 of the culture volume in DHI-special at room temperature. For 1 ml of culture, first 0.4 μg of DNA (1:3 ratio of NR1:NR2) was added to 0.1 ml of medium, mixed, after 2 min followed by 0.15 μg of RO-1539 A, mixed, and followed after further 2 min by 0.5 μg of RO-1539 B. The mixture was incubated for 15 min at room temperature to allow DNA complex formation before it was added to the cells. Six hours post-transfection, 5% dialyzed fetal calf serum and NMDA antagonists dl-amino-5-phosphonopentanoic acid (50 μM, final), dextromethorphan (10 μM, final) were added for protection against NMDA receptor-induced cell death.
Forty-eight hours post-transfection, the cells were harvested and washed three times with cold phosphate-buffered saline and frozen at –80°C. The pellet was suspended in cold 20 mM HEPES-NaOH buffer containing 10 mM EDTA (pH 7.4) and homogenized with a polytron (Kinematica AG, Basel, Switzerland) for 10 s at 10,000 rpm. After centrifugation at 48,000g for 30 min at 4°C, the pellet was resuspended in cold 20 mM HEPES-NaOH buffer containing 0.1 mM EDTA (pH 7.4), homogenized, and respun as described above. The pellet was resuspended in a smaller volume of a cold 20 mM HEPES-NaOH buffer containing 0.1 mM EDTA (pH 7.4). After homogenization for 10 s at 10,000 rpm, the protein content was measured using the bicinchoninic acid method (Pierce Chemical, Socochim, Lausanne, Switzerland) with bovine serum albumin as the standard. The membrane homogenate was frozen at –80°C before use.
[3H]RO 25-6981 Binding. After thawing, the membrane homogenates were centrifuged at 48,000g for 10 min at 4°C, the pellets were suspended anew in the cold 50 mM Tris-HCl binding buffer containing 100 μM DTG (pH 7.4) to a final assay concentration of 25 μg of protein/ml. Saturation isotherms were determined (in presence of 100 μM DTG) by addition of various [3H]RO 25-6981 concentrations to these membranes for2hat4°C. At the end of the incubation, membranes were filtered onto Filtermate (unitfilter PerkinElmer Life Sciences, Boston, MA: 96-well white microplate with bonded GF/B filter preincubated 1 h in 0.1% polyethyleneimine) and washed three times with cold 20 mM HEPES-NaOH buffer containing 0.1 mM EDTA (pH 7.4). Nonspecific binding was measured in the presence of 10 μM RO 04-5595. The radioactivity on the filter was counted on a Packard Top-Count microplate scintillation counter with quenching correction after addition of 40 μl of microscint 40 (Canberra Packard S.A., Zürich, Switzerland). Saturation experiments were analyzed by Prism 3.0 (GraphPad Software Inc., San Diego, CA) using the rectangular hyperbolic equation derived from the equation of a bimolecular reaction and the law of mass action, B = (Bmax × [F])/(KD + [F]), where B is the amount of ligand bound at equilibrium, Bmax is the maximum number of binding sites, [F] is the concentration of free ligand, and KD is the ligand dissociation constant. The experiments were performed three times in quadruplet, and the mean ± standard deviation of the individual KD values was calculated.
Western Blot Analysis. After thawing, the membrane homogenates were centrifuged at 18,000g for 10 min at 4°C, the pellets were solubilized in a buffer containing 20 mM HEPES, 150 mM NaCl, 1.5 mM MgCl2, 1 mM EGTA, 10% glycerol, 1% Triton X-100, 0.1% SDS, 10 μg/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride, and 10 μg/ml aprotinin, pH 7.5. The lysates were sonicated and incubated for 1 h at 4°C. After centrifugation at 26,000g for 20 min at 4°C, the protein concentration was determined as described above. For Western blot, the individual lysates were diluted in a Laemmli buffer containing 100 mM dithiothreitol to the same final protein concentration, heated for 10 min at 95°C, and sonicated. Five micrograms of membrane protein was separated by 7.5% acrylamide gel (Bio-Rad 161-1172; Bio-Rad, Hercules, CA) and electroblotted onto a nitrocellulose filter (0.45 μm) (Bio-Rad 162-0115). The filters were blocked in Tris-buffered saline-Tween with 3% skimmed dry milk and incubated with a solution containing 0.1 mg/ml anti-NR1 (goat polyclonal antibody, sc-1467; Santa Cruz Biotechnology, Inc., Santa Cruz, CA) or 0.2 mg/ml anti-NR2B (goat polyclonal antibody, sc-1469; Santa Cruz Biotechnology, Inc.). After washing, the filter was incubated in horseradish peroxidase-conjugated donkey anti-goat antibody (sc-2033, Santa Cruz Biotechnology, Inc.) (0.02 μg/ml), washed, and developed to reveal bound antibody using the Lumi-Light Western blotting kit (Roche Applied Science).
XenopusOocyte Voltage-Clamp Recordings.Xenopus laevis oocytes were prepared, nuclear injected, and voltage-clamped as described previously (Paoletti et al., 1995). Oocytes were injected with 30 nl of a mixture of NR1 and NR2 cDNAs (ratio 1:2, each cDNA at a concentration of 10 ng/μl) and evoked currents were recorded over the next 1 to 2 days. The external superfusing solution contained 100 mM NaCl, 5 mM HEPES, 0.3 mM BaCl2, and 10 mM tricine (used to chelate contaminating zinc) (Paoletti et al., 1997). The pH was adjusted to 7.3 with KOH. l-Glutamate and glycine were applied at saturating concentrations (100 μM). The holding potential was –60 mV. Mean data points were calculated from two to six individual experiments and used to fit inhibition curves according to the Hill equation with baseline, y = 1–(a/(1 + (X/IC50)nH)), where nH is slope factor and a is the maximum inhibition level.
Alignments and Molecular Modeling. The sequence alignments were essentially adapted from the alignments presented by Paoletti et al. (2000) and Perin-Dureau et al. (2002). Escherichia coli LIVBP and rat mGlu1 receptor (rmGlu1) structural data were accessed from LIVBP, GenBank accession number 230609; Protein Data Bank, PDB coordinates, 2LIV (Sack et al., 1989); rmGlu1, GenBank accession number P23385; and PBD coordinates, 1EWT (Kunishima et al., 2000). The 3D structure of the NR2B LIVBP-like domain was modeled by homology to the structure of the liganded form of mGlu1 (PDB coordinates, 1EWK) on the basis of the sequence alignment. All modeling calculations were made on a Silicon Graphics Octane with a single R12000 processor using our in-house modeling package Moloc (Gerber and Muller, 1995) (http://www.Moloc.ch).
Results
Generation of Point Mutations and Their Expression in HEK-293 Cells. The structures of [3H]RO 25-6981 and ifenprodil are shown in Fig. 1. To identify amino acid residues that control RO 25-6981 affinity, six point mutations (Fig. 2) were introduced into the human NR2B subunit by site-directed mutagenesis. These amino acids were chosen from residues previously determined to be critical for the ifenprodil inhibition of mouse NR1/NR2B receptor function (Fig. 2) (Perin-Dureau et al., 2002). To produce large quantities of functional NMDA receptor, a combination of NR1c with NR2B wild-type (WT) or mutant cDNAs were transiently coexpressed in HEK-293 cells as described under Materials and Methods. To assess the level of the NR1c/NR2B expression, cell membranes from mock-transfected cells (pcDNA3) as well as from cells cotransfected with various plasmid combinations were subjected to immunoblotting using goat anti-NR1 and anti-NR2B polyclonal antibodies (Fig. 3). The anti-NR1 antibody recognized a species with a molecular mass of ∼130 kDa in HEK-293 cells transfected with NR1c/NR2B WT (lane 9), NR1c/NR2B mutants (lanes 1–6) or NR1c alone (lane 7), but not in the mock-transfected cells (lane 10) or cells transfected with NR2B alone (lane 8) (Fig. 3, left). Similarly, an immunoreactive band of 185-kDa apparent molecular mass was detected using an anti-NR2B antibody (Fig. 3, right) in cells transfected with NR1c/NR2B WT (lane 9) or NR1c/NR2B mutants (lanes 1–6), indicating that the mutated receptors were expressed at a comparable level to WT. No band was detected in the mock-transfected cells (lane 10) or cells transfected with either NR1c alone (lane 7) or NR2B alone (lane 8). The lack of immunoreactive band in the cells transfected with NR2B alone may indicate that the NR2B subunit is highly unstable in the absence of NR1. The 130- and 185-kDa bands that are detected for recombinant NR1 and NR2B proteins, respectively, are in good agreement with those observed for the native rat cortex receptor (lane 11, left and right). We noted a slightly increased electrophoretic mobility of the hNR2B(F182A) subunit (Fig. 3, right, lane 4); importantly, as the integrity of entire coding region of all constructs was verified this did not result from any undesired mutational artifact.
Effect of Point Mutations on the [3H]RO 25-6981 Binding. To characterize [3H]RO 25-6981 binding to the recombinant NMDA receptors, a series of preliminary experiments were performed using membranes from mock-transfected and NR1c/NR2B-cotransfected HEK-293 cells. Surprisingly, specific [3H]RO 25-6981 binding with a KD value of 83 nM was detected in membranes from mock-transfected cells. This binding was fully inhibited by the σ site ligand DTG, with an IC50 value of 19 μM and Hill value of 1.6. Therefore, all subsequent saturation bindings were performed in the presence of 100 μM DTG to eliminate this apparent σ site binding. Saturation experiments were performed, as outlined under Materials and Methods, with membranes containing recombinant hNR1c/NR2B receptor (Fig. 4A). The saturation isotherm was monophasic ([3H]RO 25-6981 concentrations 0.5 to 100 nM) and best fitted to a one-site model. Curve fitting yielded a KD value of 7.2 ± 1nM and Bmax value of 3.1 ± 0.16 pmol/mg protein. The determined affinity is in good agreement with our data for native receptors in the rat and human brain (KD values of 5 and 6.4 nM, respectively). In an earlier binding study using rat whole brain membranes (Mutel et al., 1998), there was a very small but significant increase in Ro 25-6981 affinity with no change in Bmax in the presence of 100 μM glutamate (KD = 2.33 ± 0.3 and 1.6 ± 0.2 nM in the absence and presence of 100 μM glutamate, respectively), whereas glycine did not affect [3H]RO 25-6981 binding affinity at up to 100 μM. In view of this limited effect of glutamate on ligand binding affinity, we did not attempt to address its effect on [3H]RO 25-6981 binding to the mutated NR2B receptors in the present work.
Saturation binding analysis was then performed on membranes isolated from HEK-293 cells transfected with NR1c and the various mutated NR2B subunits using 0.3 to 300 nM [3H]RO 25-6981. The dissociation constants (KD) derived from the saturation isotherms are given in Table 1. As expected, no specific [3H]RO 25-6981 binding was detectable in the cells transfected with hNR1c alone, hNR2B alone, or rNR1c/rNR2A. The binding affinity at rat NR1c/NR2B receptors (KD = 7.34 ± 2.4 nM) was similar to that of human NR1c/NR2B receptor (KD = 7.2 ± 1 nM). HEK-293 cells transfected with three subunits, rNR1c/NR2A/NR2B, also yielded a monophasic saturation isotherm with KD = 5.1 ± 1 nM (Fig. 4B). All six point mutations in hNR2B had a marked effect on the binding affinity of [3H]RO 25-6981. However, two mutations, hNR2B(D101A) and hNR2B(F176A), resulted in the complete loss of [3H]RO 25-6981 binding, up to the highest concentration tested (300 nM). Four other mutations, hNR2B(T233A), hNR2B(F182A), hNR2B(D104A), and hNR2B(K234A), led to 13-, 8-, 7-, and 6-fold decreases in [3H]RO 25-6981 binding affinity, respectively (Fig. 4; Table 1).
Effect of Point Mutations on RO 25-6981 Inhibition of NMDA Receptor Function. To assess the effect of point mutations in a functional assay, voltage-clamp recordings were performed from Xenopus oocytes coexpressing hNR2B wild-type or mutant subunits with the hNR1c subunit. In all cells expressing wild-type or mutant NMDA receptor subunits, coapplication of 100 μM glutamate and 100 μM glycine evoked inward currents. Exposure of cells expressing hNR1c/NR2B(WT) to RO 25-6981 resulted in a concentration-dependent inhibition of NMDA-evoked current with an IC50 = 9.6 nM (Fig. 5A), in very good agreement with both the determined KD value and our previous data with both recombinant and native rat receptors (9 and 15 nM, respectively) (Fischer et al., 1997). In addition, the measured on-rate of receptor blockade after application of 1 μM RO 25-6981 in the continual presence of glutamate and glycine (Fig. 5B; τ = 13.2 ± 1.4s, n = 4) was in good agreement with our previous observations with native rat receptors (Fischer et al., 1997). The RO 25-6981 concentration-dependent inhibition of NMDA-induced currents in cells expressing the mutated NMDA receptors is shown in Fig. 5A and their derived IC50 and slope values in Table 2. Inhibition curves for RO 25-6981 obtained from cells expressing the mutated NMDA receptors were all shifted to the right (Fig. 5A), reflecting a decrease in activity. As expected from the binding studies, in cells expressing the mutants hNR1c/NR2B(D101A) and hNR1c/NR2B(F176A) receptors, which did not bind [3H]RO 25-6981, the RO 25-6981 inhibition curves were markedly shifted resulting in large increases in IC50 values (11,000 and 1200 nM for D101A and F176A, respectively) (Fig. 5B). The mutation T233A, which had a marked effect on the binding of [3H]RO 25-6981 (KD value increased by 13-fold), had a similar influence on RO 25-6981 block of NMDA current response (IC50 value increased by 12.5-fold). However, a somewhat larger shift in IC50 value was observed for the F182A mutant (28-fold) relative to the observed effect on binding affinity (7.9-fold).
Molecular Modeling of the N-Terminal LIVBP-Like Domain of the NR2B Subunit. The sequence identity between the amino-terminal domain (ATD) of the mGlu1 receptor and NR2B is low (∼10%). The atomic coordinates of the mGlu1 ATD complexed with glutamate (PDB code 1EWK) (Kunishima et al., 2000) was used as a template to build the 3D model of the NR2B LIVBP-like domain. The alignment between mGlu1 and NR2B is shown in Fig. 2. None of the insertions/deletions seem to be near the binding site of the ligand. Whenever it was possible, the torsion angles of the amino acid side chains were kept similar to those in the crystal structure, otherwise they were optimized in the Moloc force field (Gerber and Muller, 1995). The model does not contain water molecules. A C-alpha representation of the homology model of NR2B LIVBP-like domain is shown in Fig. 6. Amino acids discussed in this work are shown in pink.
Discussion
The LIVBP-like domain of the NR2B subunit, which is formed from the first 390 aa preceding the GlnBP-like domain, represents an important regulatory region for allosteric modulation of receptor function by, e.g., protons, polyamines, and the noncompetitive antagonists ifenprodil and CP 101,606 (Gallagher et al., 1996, 1997; Brimecombe et al., 1998). Interestingly, the LIVBP-like domain exhibits substantial sequence divergence within the NR2 family. The NR2B LIVBP-like domain displays 55, 37, and 34% aa identity to that of NR2A, NR2C and NR2D, respectively, whereas the agonist binding domain (GlnBP-like domain) exhibits around 70% aa identity among the NR2 subunits. Recently, using a mutagenesis approach, Perin-Dureau et al. (2002) mapped the high-affinity binding pocket of ifenprodil to the LIVBP-like domain of the mouse NR2B subunit. However, in the absence of radioligand binding data, the study was not able to discriminate definitively between effects of mutations on ligand binding and downstream signal transduction. RO 25-6981, an ifenprodil derivative, belongs to the phenylethanolamine class and is a high-affinity antagonist, selective for NMDA receptors containing the NR2B subunit (>5000-fold higher affinity for heteromeric NR1/NR2B receptors than NR1/NR2A receptors). Similar to ifenprodil (Kew et al., 1996), RO 25-6981 exhibits an activity-dependent mode of antagonism and is neuroprotective against glutamate excitotoxicity in cell culture and in animal models of focal cerebral ischemia (Fischer et al., 1997). In the present work, we have used [3H]RO 25-6981 binding and RO 25-6981 inhibition of NMDA-evoked currents to elucidate critical residues mediating the high-affinity inhibitory activity at the human NR2B subunit.
Of the six mutations in the NR2B subunit that we characterized, we observed that the D101A and F176A mutations resulted in the complete loss of [3H]RO 25-6981 binding affinity and also abolished the high-affinity RO 25-6981 inhibition of NMDA-evoked currents. Our mutational analysis also identified Thr-233 as an important residue for the binding of [3H]RO 25-6981 to hNR1c/NR2B receptor. Mutagenesis of Thr-233 to alanine markedly decreased the affinity of RO 25-6981 for the receptor by 13-fold. The mutations F182A, D104A, and K234A all had a moderate influence on the binding affinity (KD values increased by 8-, 7-, and 6-fold, respectively). In general, our functional and binding data were in good agreement with the exception of the mutation F182A with which we observed a somewhat larger shift in the functional relative to the binding assay. In oocyte experiments using the F182A mutant, RO 25-6981 inhibition was partial up to 10 μM, the highest concentration assayed, and the derived slope from the fitted curve was also markedly less that that observed for the other mutant and WT receptors. Thus, it is possible that this mutation differentially affects RO 25-6981 efficacy versus affinity. The F182A mutation also exhibited a marked effect on the ifenprodil inhibition of the glutamate-induced NMDA receptor current response (IC50 value increased by >60-fold) (Perin-Dureau et al., 2002).
It is interesting to note that among the six mutated residues, Thr-233 is the only residue that differs between the NR2B and NR2A subunit, which has a serine at this position. In addition to this conservative substitution, differences in the primary sequence and tertiary structure of the LIVBP-like domains of NR2A and NR2B (which share only 55% sequence identity) could be critical determinants of the selectivity of RO 25-6981 for NR1/NR2B versus NR1/NR2A receptors. For example, the loop between the β1 and α1 helices (Fig. 2) contains residues important for Zn2+ binding at NR2A (H44 and to a lesser extent H42) (Paoletti et al., 2000) and for ifenprodil in NR2B (V42) (Perin-Dureau et al., 2002), yet, accepting the potential limitations of the alignments, the sequences are very different with an apparent insertion (containing the two histidines) in NR2A. Such a loop in NR2A might sterically hinder RO 25-6981 (or ifenprodil) binding.
To visualize the mutation data, we constructed a 3D model of the NR2B LIVBP-like domain using the alignment in Fig. 2 and atomic coordinates of the mGlu1 ATD complexed with glutamate (PDB coordinates, 1EWK) (Kunishima et al., 2000). Figure 6 shows the amino acids in the NR2B LIVBP-like domain found to affect the binding affinity of [3H]RO 25-6981. All the residues that we mutated in this work are located in the central cleft. Residues Asp-101 and Asp-104 are located in lobe I, whereas the residues Phe-176, Phe-182, Thr-233, and Lys-234 are positioned in lobe II, interestingly with respect to the differential effects of the mutation in functional and binding studies, with Phe-182 positioned toward the hinge region.
Tamiz et al. (1998) have reported an extensive structure-activity relationship (SAR) for a series of ifenprodil analogs. According to their study, the primary determinants of the ifenprodil analog potency at the NR1/NR2B receptors are 1) the phenolic OH group, which might serve as an H-bond donor; 2) the distance between two aromatic rings; and 3) an electrostatic interaction between the receptor and the basic nitrogen atom. Based on the information provided by this SAR study (Tamiz et al., 1998) and the proximity of three crucial residues, Asp-101, Phe-176, Thr-233 to glutamate in the 3D model of the NR2B LIVBP-like domain, we speculate on a possible mode of interaction for the high-affinity binding site of RO 25-6981 in which the protonated basic nitrogen atom of RO 25-6981 forms an electrostatic interaction with the carboxyl group of Asp-101 and the benzyl group of RO 25-6981 and the side chain of Phe-176 establishes an aromatic stacking (π/π) interaction (Fig. 7). The interaction of RO 25-6981 with the NR2B LIVBP-like domain might promote the closure of two lobes, resulting in the stabilization of the closed conformation of LIVBP-like domain similar to that shown for the binding of glutamate to mGlu1 (Kunishima et al., 2000). Due to the limited sequence homology between mGlu1 and NR2B ATD (∼10%), this model is likely to provide a simplistic view of the NR2B LIVBP-like domain and its interaction with RO 25-6981. Determination of the precise binding mode and the relative importance of the mutations studied here awaits the elucidation of the X-ray crystal structure of the NR2B LIVBP-like domain complexed with an antagonist.
In conclusion, in the present study we have used site-directed mutagenesis together with radioligand binding, functional assay, and molecular modeling to identify critical residues involved in the RO 25-6981 binding site of the human NR2B subunit. Our data suggest that three amino acids Asp-101, Phe-176, and Thr-233 are important molecular determinants for high-affinity RO 25-6981 binding. NMDA receptor dysfunction has been implicated in psychiatric and neurological disorders such as schizophrenia and Alzheimer's disease as well as in acute excitotoxic neuronal degeneration (Dingledine et al., 1999). The allosteric modulatory site on the NR2B LIVBP-like domain provides opportunity for the design of therapeutic compounds with negative, and perhaps positive, modulatory modes of action. Knowledge of the molecular determinants mediating this interaction could be instrumental in the design of therapeutic compounds with desirable profiles.
Acknowledgments
We are grateful to Klaus Christensen, Claudia Kratzeisen, Marie-Thérèse Zenner, Birgit Molitor, Agnès Nilly, and Sylvie Chaboz for excellent technical assistance.
Footnotes
-
ABBREVIATIONS: NMDA, N-methyl-d-aspartate; iGlu, ionotropic glutamate; TM, transmembrane; LIVBP, leucine/isoleucine/valine-binding protein; mGlu, metabotropic glutamate; GlnBP, glutamine binding protein; DTG, 1,3-di-o-tolylguanidine; PCR, polymerase chain reaction; nt, nucleotide; HEK, human embryonic kidney; RO 25-6981, R-(R*,S*)-α-(4-hydroxyphenyl)-β-methyl-4-(phenyl-methyl)-1-piperidine propanol; RO 04-5595, 1-(4-chlorophenyl)-2-methyl-6-methoxy-7-hydroxy-1,2,3,4-tetrahydroisoquinolin; 3D, three-dimensional; WT, wild-type; ATD, amino-terminal domain; CP 101,606, (1S,2S)-1-(4-hydroxyphenyl)-2-(4-hydroxy-4-phenylpiperidino)-1-propanol; SAR, structure-activity relationship.
-
DOI: 10.1124/jpet.103.056291.
-
↵1 Current address: Addex Pharmaceuticals SA, 12 Chemin Des Aulx, CH-1228 Plan Les Ouates, Switzerland.
-
↵2 Current address: Evotec Neurosciences GmbH, Schnackenburgallee 114, D-22525 Hamburg, Germany.
-
↵3 Current address: Psychiatry Centre of Excellence for Drug Discovery, GlaxoSmithKline, New Frontiers Science Park, Third Ave., Harlow, Essex CM19 5AW, UK.
- Received June 30, 2003.
- Accepted August 22, 2003.
- The American Society for Pharmacology and Experimental Therapeutics