Abstract
The resolved X-ray crystal structures of the glutamate-binding domain (S1/S2 domains) of the GluR2 and NR1 glutamate receptor subunits were used to model the homologous regions of the N-methyl-d-aspartate (NMDA) receptor's NR2 subunits. To test the predictive value of these models, all four stereoisomers of the antagonist 1-(phenanthren-2-carbonyl) piperazine-2,3-dicarboxylic acid (PPDA) were docked into the NR2B glutamate-binding site model. This analysis suggested an affinity order for the PPDA isomers of d-cis > l-cis > l-trans = d-trans and predicted that the 2-position carboxylate group of the cis-PPDA isomers, but not of the trans-PPDA isomers, may be interacting with histidine 486 in NR2B. Consistent with these predictions, cis-PPDA displays a 35-fold higher affinity for NR2B-containing NMDA receptors than trans-PPDA. In addition, mutating NR2B's H486 to phenylalanine decreased cis-PPDA affinity 8-fold but had no effect on trans-PPDA affinity. In contrast, the NR2B H486F mutation increased the affinity of the typical antagonists CGS-19755 [(2R*,4S*)-4-phosphonomethyl-2-piperidine carboxylic acid] and 4-(3-phosphonopropyl) piperidine-2-carboxylic acid. In the NR1-based NR2 models, there were only four subunit-specific amino acid residues exposed to the ligand-binding pocket (and six in the GluR2-based models). These residues are located at the edge of the binding pocket, suggesting that large antagonists may be necessary for subtype specificity. Of these residues, mutational analysis and modeling suggest that A414, R712, and G713 (NR2B numbering) may be especially useful for developing NR2C- and NR2D-selective NMDA receptor antagonists and that residues A414 and T428 may determine subunit variations in agonist affinity.
(S)-Glutamate is the primary excitatory neurotransmitter in the vertebrate central nervous system. The fast, excitatory actions of (S)-glutamate are mediated by three types of glutamate-gated ion channel receptors, the N-methyl-d-aspartate (NMDA), kainate, and AMPA receptors (Watkins and Evans, 1981; Monaghan et al., 1989; Watkins et al., 1990). NMDA receptors have been extensively studied because of their roles in synaptic plasticity (Collingridge and Bliss, 1995), developmental plasticity (Bear et al., 1990), and neuropathology (Choi, 1992). Functional NMDA receptors are a heteromeric complex composed of NR1 subunits [NR1a–h, representing eight alternative splice forms from one gene (Sugihara et al., 1992) and NR2 subunits (NR2A–D, from four distinct genes; Ikeda et al., 1992; Monyer et al., 1992; Ishii et al., 1993; Monyer et al., 1994)]. Current evidence supports a tetrameric structure involving two NR1 subunits and two NR2 subunits (Laube et al., 1998). In addition, a nonessential NR3 subunit has been identified (Sucher et al., 1995).
The glutamate-binding sites of the NMDA receptor complex are found on the NR2 subunits (Laube et al., 1997; Anson et al., 1998). Although the different NR2 subunits display generally similar pharmacological profiles (Ikeda et al., 1992; Buller et al., 1994; Laurie and Seeburg, 1994), some glutamate site agonists and antagonists can distinguish between the different NR2-containing receptors (Buller et al., 1994; Buller and Monaghan, 1997; Auberson et al., 2002; Feng et al., 2004). Accordingly, glutamate site agents can distinguish four different native NMDA receptors in brain that correspond both anatomically and pharmacologically to the four different NR2 subunits (Monaghan et al., 1988; Beaton et al., 1992; Christie et al., 2000). Since the different NR2 subunits are heterogenously distributed (Monyer et al., 1992; Watanabe et al., 1993; Buller et al., 1994) and display significantly different physiological properties (Monyer et al., 1994), antagonists that selectively block different NR2-containing NMDA receptors are likely to have markedly different physiological and therapeutic/adverse effects. To date, however, only a few glutamate site antagonists have been found to display subtype selectivity, and these compounds are only weakly selective. To facilitate the development of antagonists with greater subtype selectivity, we generated molecular models of the glutamate-binding site on rat and human NR2 subunits.
The glutamate-binding site on the glutamate-gated ion channel receptors lies in two domains, S1 and S2, which together have homology to the distantly related bilobed bacterial amino acid-binding proteins leucine/arginine/ornithine-binding protein (LAOBP) and glutamine-binding protein (QBP) (Nakanishi et al., 1990; O'Hara et al., 1993). Previously, Laube and colleagues have used the resolved structure of LAOBP (Oh et al., 1993) to generate a model of the NR2B subunit's glutamate-binding site (Laube et al., 1997). This model, together with the use of site-directed mutagenesis, enabled the identification of amino acid residues that are critical for the binding of glutamate and glutamate site antagonists. More recently, the AMPA receptor GluR2 subunit (Armstrong et al., 1998; Armstrong and Gouaux, 2000) and the NMDA NR1 (Furukawa and Gouaux, 2003) subunit S1/S2 domain structures have been resolved by crystallography. Since the NR2 subunit primary structure is more closely related to NR1 and GluR2 than to LAOBP, we used the resolved NR1 and GluR2 crystal structures to generate homology models of the NR2 S1/S2 structures. In support of this approach is the observation that the secondary structures of LAOBP/QBP and the distantly related GluR2 are readily superimposed despite low sequence identity (Armstrong et al., 1998). To test the predictive value of the models, we evaluated mutations that were predicted to have specific effects on antagonist binding. We also tested point mutations of amino acid residues that were predicted to be subunit-specific to determine whether these residues can account for some observations of pharmacological subunit specificity.
Materials and Methods
Homology Modeling. NR2B sequences D403 to N542 (S1) and K669 to N803 (S2) and the corresponding NR2C sequence were aligned to the GluR2 sequence. The COMPOSER and BIOPOLYMER modules of the SYBYL 6.6 (Tripos, St. Louis, MO) software package were used for comparative molecular modeling of the NR2B and NR2C subunits based on the crystal structure determined for GluR2 (Armstrong et al., 1998).
Consistent with the GluR2 data, a disulfide bridge was formed between NR2B C746 and C801. Within the NR2 sequences there is a segment of 35 amino acids that has no homology to GluR2 (amino acids 422–456 of NR2B) and 21 amino acids of this NR2B segment are an additional insert relative to the GluR2 sequence. To model this loop, the loop searching method implemented in the SYBYL 6.6 was used to find the most appropriate protein loop from the Protein Data Bank. From the several candidates, the best fitting loop was chosen based on the high homology with the modeled sequence, the proper positioning of the anchor atoms, and the tight packing with the rest of the molecule. The constructed structural models were tested with the PROTABLE module and refined by energy minimizations using the AMBER (Kollman all-atom) force field implemented in SYBYL. To better model antagonist interactions with the NR2 subunits, the NR2B model was refit to the antagonist-bound conformation of the GluR2 subunit (Armstrong and Gouaux, 2000). Using these same methods, the recently resolved NR1 S1/S2 crystal structure in the antagonist-bound conformation was used to model each of the four human NR2 S1/S2 domains. Antagonist isomers were constructed in the ionized forms and energy-minimized using MMFF94 force field in SYBYL. Docking was performed with the DOCK software module of SYBYL 6.6.
Construction of Rat NR2B Point Mutations. NR1a cDNA in pBluescript was generously provided by Dr. Shigetada Nakanishi (Kyoto University, Kyoto, Japan), and the NR2B cDNA in pRK5 was kindly provided by Drs. Dolan Pritchett and David Lynch (University of Pennsylvania, Philadelphia, PA). NR2B A414R, NR2B G427E, NR2B T428G, NR2B G712S, NR2B R712P/G713R, and NR2B H486F mutations were made using the QuikChange Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA). Mutations were confirmed by sequencing on both strands by the Molecular Biology Core Laboratory at the University of Nebraska Medical Center.
In Vitro Transcription and Translation inXenopusOocytes. Plasmids were linearized with NotI(NR1a) or SalI(NR2B) and transcribed with T7 (NR1a) or SP6 (NR2B) in vitro using the mMessage mMachine kit (Ambion, Austin, TX). Xenopus laevis female frogs were obtained from Xenopus I (Ann Arbor, MI), and oocytes were isolated and prepared as previously described (Monaghan and Larson, 1997). The handling of frogs was performed in accordance with the Guide for the Care and Use of Laboratory Animals as adopted and promulgated by the U.S. National Institutes of Health. Oocytes were injected with 2 to 30 ng of NR1/NR2 RNA (1:3) in 9 to 50 nl and then incubated in ND-96 (96 mM NaCl, 2 mM KCl, 1.8 mM CaCl2, 1 mM MgCl2, and 5 mM HEPES, pH 7.6) at 17°C for 1 to 4 days.
Oocyte Electrophysiology. Electrophysiological responses of recombinant receptors expressed in Xenopus oocytes were measured by two-electrode voltage clamp (OC-725B Oocyte Clamp; Warner Instrument, Hamden, CT) at a holding potential of –60 mV. Unless indicated otherwise, NMDA receptor responses were evoked by bath application of 10 μM (S)-glutamate/10 μM glycine. Recordings were made in barium Ringer's solution to eliminate calcium-activated chloride currents. Only cells that generated stable plateau responses were used. Current responses to drug application were recorded on both a strip chart and by digital capture using an ITC-16 computer interface (InstruTECH Corporation, Port Washington, NY) and a MacIntosh computer with AxoData software (Axon Instruments Inc., Union City, CA). Dose-response curves for antagonist blockade of responses were fit (GraphPad Prism; GraphPad Software Inc., San Diego, CA) to the equation: I = Imax/[1 + (IC50/A)], where I is the current response, Imax is the current response in the absence of antagonist, and A is the concentration of antagonist. (S)-Glutamate affinities were obtained for wild-type and mutant receptors with the same set of solutions and used to correct antagonist IC50 values to the corresponding Ki values.
Materials. Antagonists were kindly provided by Dr. Paul Ornstein (Lilly Research Laboratories, Indianapolis, IN) [(2R*,4S*)-4-(3-phosphonopropyl)piperidine-2-carboxylic acid (PPPA)] and Dr. Richard Lovell (Novartis, Basel, Switzerland) (CGS-19755). Cis-PPDA, trans-PPDA, and (R)-CPPene were synthesized in our laboratory (Dr. Jane's). Other chemicals were obtained from Sigma-Aldrich (St. Louis, MO).
Results
The rat NR2B and NR2C structural models were built by homology modeling based upon the GluR2 crystal structure (Armstrong et al., 1998; Armstrong and Gouaux, 2000). These are shown in Fig. 1A superimposed upon the GluR2 crystal structure. The four human NR2A to D S1/S2 models were constructed using the resolved NR1 subunit structure (Furukawa and Gouaux, 2003); the human NR2B S1/S2 structural model is shown in Fig. 1F. The models underwent several rounds of energy minimizations with subsequent stereochemistry testing by the PROTABLE module of SYBYL. From 275 amino acid residues included in each of the models, four to nine residues, mostly from the less defined loop 422 to 456, had backbone conformations that fell into disallowed areas of the Ramachandran plot, whereas 12 to 19 residues were found within allowed but sterically constrained areas of the Ramachandran plot. Given the high degree of overall structural similarity between the resolved crystal structures for GluR2 and the distantly related LAOBP and QBP, the proposed structure for NR2s (which are much more closely related to GluR2 and NR1) may have predictive value. Recently, a model for the human NR2B subunit based upon the GluR2 subunit was generated (Tikhonova et al., 2002). It is not known, however, how this model compares with the NR2B models described here.
Docking of PPDA to NR2B. In recent studies, we have identified the compound cis-PPDA as a high-affinity NMDA receptor antagonist with novel subtype selectivity (Feng et al., 2004). For the structure of PPDA and other antagonists used and discussed in this study, see Fig. 2. As a test of the predictive value of the models generated, we modeled the binding of PPDA to NMDA receptors because PPDA has a bulky phenanthrene group attached to dicarboxypiperazine via a carbonyl group. Thus, this large and relatively rigid compound is expected to have relatively few permitted modes of docking with the NR2 glutamate-binding site. Furthermore, PPDA docking was of interest because PPDA is unusual for an NMDA receptor antagonist in having two closely spaced acid groups (3.4 Å) that do not conform to the previously defined NMDA receptor antagonist pharmacophore (6–7 Å between acid groups; Ortwine et al., 1992; Whitten et al., 1992). The energy-minimized (MMFF94 force field) (2S,3S)-, (2S,3R)-, (2R,3R)-m and (2R,3S)-isomers of PPDA were constructed. For each of these isomers, four initial conformations were used that differed in piperazine ring conformation (chair or boat) and in the position of the phenanthrene ring relative to the dicarboxylic acids (cis or trans). After energy minimization, each isomer of PPDA was manually docked into the NR2 glutamate-binding site using the DOCK module of SYBYL. To dock PPDA, the 4-amino and 3-carboxy groups of PPDA (corresponding to the α-amino and carboxylate groups of glutamate, respectively) were superimposed upon corresponding positions determined for the kainate molecule within the GluR2 crystal structure (while GluR2 was superimposed onto NR2B). After docking, the PPDA-receptor complex was energy-minimized to improve the fit. The three bonds in PPDA that can rotate were allowed to rotate, as were the ligand-binding amino acid residue side chains (see Fig. 1D). In docking the different isomers of PPDA and performing several rounds of energy minimizations for each isomer, we found that the two cis-isomers [(2R,3S)- and (2S,3R)-isomers] displayed lower energy fits (–231 and –274 kcal, respectively) than any of the trans-isomers [–167 kcal for (2S,3S)-isomer and –159 kcal for the (2R,3R)-isomer]. In Fig. 1B, the (2R,3S)-isomer (l-cis) of PPDA is shown docked. Interestingly, for all isomers/conformations, the large and rigid phenanthrene ring group was consistently fit into, or near, a groove at the bottom of the S1/S2 cleft formed between helices F and H on the S2 interface (Armstrong et al., 1998). This places the phenanthrene ring pointing directly out of the binding cleft and perpendicular to the two β strands that link the S1/S2 lobes.
The proposed placement of the phenanthrene group is identical to the position of the bulky aliphatic group of kainate (and presumably domoate) when bound to GluR2 (Armstrong et al., 1998). This orientation for a large hydrophobic group is also consistent with that proposed for the biphenyl group of the NMDA receptor antagonist EAB515 (Bigge, 1993). In antagonist molecular modeling studies, it was proposed that the linear biphenyl group of EAB515 is nearly perpendicular to the plane formed by the α-amino and the two carboxy groups of EAB515 and occupies a hydrophobic pocket (see also Cheung et al., 1996). Thus, the proposed hydrophobic pocket is probably the interface between the S1 and S2 lobes, and our model predicts that it might include the groove formed between helices F and H. In this docking, the end of the phenanthrene ring is flush with the surface of the S1/S2 globular structure and is in the vicinity of the subunit-specific residues NR2B R712 and G713.
Subsequent to docking the PPDA isomers, we were able to synthesize the racemic trans-isomer of PPDA and compare it with the racemic cis-isomer at recombinant NMDA receptors. The cis- and trans-forms of PPDA were tested at various concentrations for their ability to inhibit NR1a/NR2B NMDA receptor responses evoked by 10 μM glutamate and 10 μM glycine. Agonists were applied until a steady response was obtained, and then increasing concentrations of antagonist were applied, followed by an agonist-only application (for example, see Fig. 3 of Feng et al., 2004). As predicted, the cis-isomer displayed a significantly higher affinity (Fig. 3B). NR1a/NR2B affinity for cis-PPDA was 0.26 ± 0.05 μM and for trans-PPDA, 9.1 ± 1 μM. The NR2B wild-type 100% response was 81 ± 13 nA (mean ± S.E.M.); the NR2B H486F 100% response was 68 ± 13 nA.
As mentioned above, PPDA is distinctive among NMDA receptor antagonists in having the two closely spaced acidic groups that are separated by only three carbon-carbon bonds. Most NMDA receptor antagonists have two acidic groups separated by five or seven carbon-carbon bond lengths. The PPDA docking shown in Fig. 1B suggests that the closely spaced acidic groups in cis-PPDA may place the 2-carboxy group of cis-PPDA in a position to interact with NR2B H486. In contrast, lowest energy fits for trans-PPDA point the 2-carboxy group away from NR2B H486. Experiments using site-directed mutagenesis suggest that this residue has no effect on (R)-2-amino-5-phosphonopentanoate or (R)-3-((2-carboxypiperazin-4-yl)propyl-1-phosphonic acid binding (Laube et al., 1997). Thus, our model of cis-PPDA docking predicts that H486 may have a specific interaction with cis-PPDA that is not shared with the typical competitive antagonists. Furthermore, modeling predicts that trans-PPDA and cis-PPDA would be differentially affected by mutating H486.
To test these predictions, we constructed the NR2B H486F mutation and evaluated the potency of cis- and trans-PPDA and the antagonists CGS-19755 and PPPA. The latter two antagonists are partially constrained in an extended form by a piperidine ring and display five and seven carbon-carbon bonds between their acid groups, respectively. Changing the histidine to phenylalanine resulted in functional NMDA receptors with a 6-fold reduced affinity for (S)-glutamate [wild type (WT), 2.3 ± 0.2 μM; H486F, 13.5 ± 0.7 μM]. This compares well with the results of Laube et al. (1997) and Anson et al. (1998), which reported an 8- to 10-fold reduction in glutamate affinity with this mutation. Although the antagonists CGS-19755 and PPPA displayed a 2- to 3-fold increase in receptor affinity for NR1/NR2B-H486F (Figs. 3 and 4), cis-PPDA displayed more than an 8-fold reduction in affinity (WT, 0.26 ± 0.05 μM; H486F, 2.2 ± 0.3 μM). Also, as predicted by the modeling, trans-PPDA affinity was not affected by the H486F mutation (WT, 9.1 ± 1.0 μM; H486F, 8.4 ± 0.6 μM).
Subunit-Specific Amino Acids in the NR2 Glutamate-Binding Pocket. Evaluation of the NR2 models based on GluR2 indicated that the ligand-binding pocket is highly conserved between different subunits. As shown in Fig. 1C, there are a large number of amino acids that vary among at least one of the NR2 subunits (magenta-shaded residues). However, only eight residues of the 39 residues lining the ligand-binding pocket are not identical among all four NR2 subunits. All of these residues are at the edge of the ligand-binding pocket and are not likely to directly interact with (S)-glutamate or similarly sized antagonists. Of these residues, the two flanking NR2B I490 have side chains that are not directed toward the pocket. The NR1-based homology models predicted the same subunit-specific residues in the ligand-binding pocket as the models based on GluR2, except for NR2B residues G427 and T428. These two residues are on the loop that is not present in GluR2 but is partially present in NR1. The NR1-based modeling procedure placed this loop outside of the ligand-binding cleft region (Fig. 1F). The remaining subunit-specific amino acid residues in the ligand-binding pocket were (NR2B numbering): A414, which corresponds to an arginine on NR2C; R712, corresponding to a proline on NR2D; and G713, which corresponds to a serine on NR2C and an arginine on NR2D (Fig. 1, D and E; Fig. 4). The amino acids lining the binding pocket are also well conserved between human and rat; R711 of rat NR2A is the only ligand-binding pocket residue to differ in the human sequence (lysine).
Pharmacological Effects of Select Point Mutations. Our results suggest that most antagonists and agonists generated to date are unlikely to interact with the proposed subunit-specific residues. Nevertheless, we sought to determine whether any of these residues could explain some of the subunit specificity observed to date.
A414R. The homology models generated suggest that NR2B's A414 is near the glutamate-binding pocket (Fig. 1, D and E) and is replaced by an arginine in NR2C and NR2D (Fig. 5). Both NMDA and (S)-glutamate display higher affinity at NR2C and NR2D than at NR2B. Converting NR2B's alanine to arginine, as found in NR2C and NR2D, significantly increased the affinity of both of these agonists (Fig. 6A), (S)-glutamate (WT, 2.3 ± 0.2 μM; A414R, 0.95 ± 0.10 μM) and NMDA (WT, 29.9 ± 0.2 μM; A414R, 9.7 ± 0.4 μM). However, the potency of the antagonist LY233536, which has lower affinity at NR2C and NR2D than at NR2B (Buller and Monaghan, 1997), was not affected by the A414R mutation (WT, 0.56 ± 0.01 μM; A414R, 0.50 ± 0.01 μM). The average 100% response of the NR1a/NR2B A414R receptor was 51 ± 4 nA (mean ± S.E.M.). The average 100% response of the NR1a/NR2B receptor observed with this and the below mutation studies was 73 ± 11 nA.
G427E. The residue adjacent to NR2B's 428 is also subunit-specific with the NR2A subunit displaying a glutamate residue, whereas the other NR2 subunits each have glycine at this position. The antagonist LY233536 is distinctive in having a several fold higher affinity for receptors containing NR2B subunits than for NR2A subunits (Buller and Monaghan, 1997). G427 does not appear to contribute significantly to LY233536 selectivity since the mutating G427 to glutamate caused only a small change in LY233536 affinity (WT, 0.56 ± 0.01 μM; G427E, 0.78 ± 0.02 μM), nor did it greatly affect agonist affinity [NMDA, WT, 29.9 ± 0.2 μM; G427E, 37.8 ± 2.3 μM; (S)-glutamate, WT, 2.3 ± 0.2 μM; G427E, 2.9 ± 0.2 μM] (Fig. 6B). The average 100% response of the NR1a/NR2B G427E receptor was 63 ± 7 nA.
T428G. In NR2C, NR2B's T428 is replaced by glycine. In the GluR2-based models, these residues are at the edge of the S1/S2 cleft (Fig. 1D). To test if T428 might contribute to agonist or antagonist binding, we constructed the NR2B T428G mutation. In addition to testing glutamate potency, we evaluated PPDA potency because the T428 residue could potentially account for the observation that NR2C has the highest affinity of the NR2s for PPDA (Feng et al., 2004) since NR2C is the only NR2 that does not have a threonine at this position. For comparison, (R)-CPPene, which has higher affinity for NR2B than for NR2C (Buller and Monaghan, 1997), was also tested. The 100% response of the NR1a/NR2B T428G receptor was 48 ± 5 nA.
As shown in Fig. 6C, the T428G NR2B mutation increased glutamate affinity 8-fold (glutamate, WT, 2.3 ± 0.2 μM; T428G, 0.26 ± 0.4 μM) but had no effect on cis-PPDA affinity (WT, 0.31 ± 0.02 μM; T428G, 0.30 ± 0.5 μM) and little effect on (R)-CPPene affinity (WT, 0.24 ± 0.04 μM; T428G, 0.14 ± 0.02 μM). Thus, threonine 428 may be contributing to the subtype selectivity of (S)-glutamate for NR2C subunits but cannot account for cis-PPDA selectivity for NR2C or for (R)-CPPene selectivity for NR2B.
G713S and R712P,G713R. At the edge of the binding cleft there are two additional subunit-specific residues predicted by both sets of models, G713 (NR2B numbering) and R712 (Fig. 1, D–F). NR2A and NR2B have both these residues, NR2C has a serine at the NR2B G713 position, and NR2D has a proline at the R712 position (NR2B numbering) and an arginine at the G713 position. This has the effect of moving the arginine closer to the glutamate-binding site in NR2D. The mutations NR2B G713S and NR2B R712P,G713R were made to mimic NR2C and NR2D, respectively. (S)-Glutamate and cis-PPDA both display higher affinity at wild-type NR2C and NR2D than at NR2B. Both mutation constructs appeared to increase both (S)-glutamate and cis-PPDA affinity. The increase in cis-PPDA affinity at the R712P, G713R mutation was statistically significant (WT, 0.31 ± 0.02 μM; R712P/G713R, 0.20 ± 0.01 μM), potentially accounting for about one-half of the PPDA's selectivity for NR2D. The 100% response of the NR1a/NR2B G713S receptor was 44 ± 4 nA and for the NR1a/NR2B R712P,G713R receptor, 91 ± 17 nA.
Discussion
The purpose of the present study was to identify structural differences in the glutamate-binding pocket that could potentially be utilized in the design of new subtype-selective antagonists. By taking advantage of the homology between GluR2/NR1 subunits and NR2 subunits and the recently resolved crystal structure for the GluR2 and NR1 extracellular amino acid-binding domains, we generated homology models of the NR2 glutamate-binding sites. The models described here predict that antagonists that are small and only probe the immediate glutamate-binding region are not likely to display significant NR2 subtype selectivity. Among the four NR2 subunits, 99 of the 275 amino acids (36%) in the S1/S2 domains differ between at least one of the NR2 subunits. However, our modeling suggests that only six (NR1-based model) or eight (GluR2-based model) of these variable residues are within the S1/S2 cleft (Fig. 5). Furthermore, in both sets of models, five of these residues are located near the protein surface distant to bound glutamate. Thus, antagonists with large side groups that can extend within the S1/S2 cleft toward the protein surface may potentially interact with one of the few variant amino acid residues and thus distinguish between NR2 subunits.
The hypothesis that larger antagonists are necessary to generate varied subunit selectivity is consistent with structure-activity studies. Typically, glutamate site NMDA receptor antagonists display slightly different affinities for NMDA receptor subunits in the order of decreasing affinity: NR2A > NR2B > NR2C > NR2D (Ikeda et al., 1992; Buller et al., 1994; Laurie and Seeburg, 1994). After surveying over 75 compounds we found only three classes of antagonists that display an atypical pattern of NR2 selectivity (Andaloro et al., 1996). These are the biphenyl piperazine compound (2R*,3S*)-1-(biphenyl-4-carbonyl)piperazine-2,3-dicarboxylic acid (PBPD), the biphenyl analog EAB515, and the bicyclic decahydroisoquinoline, LY233536. These compounds display an enhanced relative affinity for the NR2C and NR2D subunits, and some also show higher affinities for NR2B than NR2A subunits (Buller et al., 1994; Andaloro et al., 1996; Christie et al., 2000). Thus, all antagonists that displayed an atypical selectivity pattern were large multiring antagonists.
Recent studies with derivatives of PBPD support the hypothesis that minor variations in antagonist structure in the distal hydrophobic region of the antagonist may vary subunit selectivity (Feng et al., 2004). Modest changes in the biphenyl group of PBPD (such as halogenation) causes small, but significant, changes in subtype selectivity (Feng et al., 2004), and altering the ring system (e.g., biphenyl to phenanthrene) also causes significant changes in the subtype selectivity pattern. Likewise, adding a chlorine to the second phenyl ring of EAB515 causes a preferential increase in NR1/NR2B selectivity (Urwyler et al., 1996).
Derivatives of a new series of quinoxaline antagonists also show varied subtype selectivity after modifications to distal moieties. The antagonist PEAQX has 100-fold selectivity for human NR2A over human NR2B (Auberson et al., 2002) and about 10-fold selectivity at rat NR2A versus NR2B subunits (Feng et al., 2004). This compound has a quinoxaline group in the core binding region and a distal bromobenzene whose addition greatly increases subunit selectivity. Thus, the bromobenzene group may be extending to the edge of the S1/S2 cleft and interacting with subunit-specific residues. Furthermore, deleting the bromo group on the end of the benzene ring significantly reduces subunit selectivity (Auberson et al., 2002).
When any of the isomers of the atypical NMDA receptor antagonist PPDA was docked to the NR2B model (GluR2-based model), we observed that the bulky and rigid phenanthrene group could fit into the S1/S2 cleft along a groove formed by two helices (F and H; Armstrong et al., 1998). To accommodate this placement, it appears that histidine 486 may be interacting with the 2-carboxy group of cis-PPDA (corresponding to the β-carboxyl of aspartate or NMDA). This carboxy group of PPDA does not fit the pharmacophore models that have been generated for NMDA receptor antagonists, it is too close to the α-carboxy and amino groups (Ortwine et al., 1992; Whitten et al., 1992). Thus, H486 might be contributing to the distal acidic group binding site in the PPDA pharmacophore but not in that of conventional antagonists. This is consistent with the observation that the NR2B H486F mutation significantly reduces cis-PPDA affinity but causes a small (2-fold) increase in conventional antagonist-binding affinity (CGS-19755 and PPPA; Figs. 3 and 4). These results are also consistent with the previous report (Laube et al., 1997) that (R)-2-amino-5-phosphonopentanoate and (R)-3-((2-carboxypiperazin-4-yl)propyl-1-phosphonic acid affinities are not significantly affected by the H486F mutation. Thus, H486 appears to interact specifically with cis-PPDA but not with conventional NMDA receptor antagonists that have a larger spacing between acidic groups or with trans-PPDA. The distal acidic group of CGS-19755, PPPA, and other long-chain antagonists might be interacting with K485/K488 (Tikhonova et al., 2002) and/or with T691 (Anson et al., 1998).
(S)-Glutamate affinity was also significantly reduced by the H486F mutation. These results are consistent with the modeling study of Ortwine et al. (1992) in which they concluded that agonists bind in a folded form, whereas conventional antagonists have a component of their binding site that is distal to that found in the agonist-binding site. Thus, the second carboxy group of cis-PPDA could be binding to a site in the agonist pharmacophore (which includes H486) that is not usually occupied by antagonists.
Of the subunit-specific residues examined, NR2B A414 and T428 both may be contributing to the higher affinity displayed by (S)-glutamate for NR2C than for NR2B. These residues, as well as NR2B G427, did not affect antagonist binding significantly. According to the NR1-based models, NR2B G427 and T428 are well outside of the ligand-binding pocket and, thus, unable to interact with ligands. The effect of T428 on agonist affinity might be allosteric.
The models generated also predict that the subunit-specific residues NR2B R712 (proline in NR2D) and G713 (serine in NR2C and arginine in NR2D) may also be near enough to the S1/S2 cleft to interact with appropriate antagonists. Although NR2D does not have an arginine at the position equivalent to NR2B's R712, the next residue in NR2D is an arginine that may be functionally equivalent to the NR2B R712 position. Since this residue is placed differently in the NR2D subunit, it may provide the basis for an NR2D-selective antagonist. That the 712 and 713 residues are sufficiently near the binding pocket to interact with ligands is also supported by the observation that the position corresponding to NR2B 714 in the GluR2 subunit is a residue known to interact with some AMPA receptor ligands but not with other ligands (Armstrong et al., 1998). The NR2B G713S and NR2B R712P,G713R mutations suggest that a portion of cis-PPDA's selectivity for NR2C and NR2D could possibly be due to these subunit-specific residues.
Given the high degree of structural homology observed between the bacterial amino acid-binding proteins and the S1/S2 domains of the GluR2 and NR1 subunits, it is expected that NR2 subunits are closely related in structure to the other ionotropic glutamate receptors. Consequently, it is likely that the GluR1-7 and KA1/2 subunits are also conserved in the general shape of S1/S2 glutamate-binding cleft. Thus, within these families, development of larger antagonists that probe the S1/S2 cleft may yield high-affinity antagonists. For the glycine-binding site of NR1, the bulky compound L-701,324 (Kulagowski et al., 1994) may represent such an antagonist. At the same time, it might be expected that large antagonists with side groups that reach the subunit-specific regions of the S1/S2 cleft could have modified subtype selectivity. One potential example of such a subtype-selective ligand is LY339434. This compound is a kainate receptor agonist derivative of (2S,4R)-4-methylglutamate (Small et al., 1998). By replacing the methyl group with a bulky naphthalene ring on a propenyl linker at the 4-position of glutamate, GluR5 affinity is reduced 5-fold, whereas GluR6 affinity is reduced over 1000-fold. Thus, addition of a distal hydrophobic ring system markedly enhances KA receptor subunit selectivity.
Overall, these studies suggest that there are relatively few amino acid residues in the NR2 subunits that may be capable of conferring subunit-specific antagonist binding. Importantly, however, there does appear to be at least one significantly different residue that is unique to each of the four NR2 subunits that may be near enough to interact with an appropriately designed antagonist molecule.
Acknowledgments
We thank Dr. Mike Sturgess for providing a copy of the NR2 model based upon the LAOBP structure; Drs. Shigetada Nakanishi, David Lynch, Dolan Pritchett, and Peter Seeburg for providing cDNA constructs; and Dr. Paul Ornstein and Dr. Richard Lovell for providing antagonist compounds.
Footnotes
-
This work was supported by National Institutes of Health Grant MH60252.
-
doi:10.1124/jpet.104.082990.
-
ABBREVIATIONS: NMDA, N-methyl-d-aspartate; AMPA, α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid; LAOBP, leucine/arginine/ornithine-binding protein; QBP, glutamine-binding protein; PPPA, (2R*,4S*)-4-(3-phosphonopropyl)piperidine-2-carboxylic acid; CGS-19755, (2R*,4S*)-4-phosphonomethyl-2-piperidine carboxylic acid; PPDA, (2R*,3S*)-1-(phenanthren-2-carbonyl)piperazine-2,3-dicarboxylic acid; (R)-CPPene, ((R,E4-(3-phosphonoprop-2-enyl)piperazine-2-carboxylic acid; EAB515, α-amino-5-(phosphonomethyl)[1,1′-biphenyl]-3-propanoic acid; WT, wild type; LY233536, (±)-6-(1H-tetrazol-5-ylmethyl)decahydroixoquinoline-3-carboxylic acid; PEAQX, (R)-[(S)-1-(4-bromo-phenyl)-ethyl-amino]-(2,3-dioxo-1,2,3,4-tetrahydroquinoxalin-5-yl)-methyl]-phosphonic acid; L-701,324, 7-chloro-4-hydroxy-3-(3-phenoxy)phenyl-2H-quinolinone; LY339434, (2S,4R,6E)-2-amino-4-carboxy-7-(2-naphthyl)hept-6-enoate.
-
↵1 Both authors contributed equally to this work.
- Received December 29, 2004.
- Accepted February 25, 2005.
- The American Society for Pharmacology and Experimental Therapeutics