Interactions between ifenprodil and the NR2B subunit of the N-methyl-D-aspartate receptor.

Ifenprodil is an atypical noncompetitive modulator of the N-methyl-D-aspartate (NMDA) receptor (NR) which demonstrates a 140-fold preference for NR2B over NR2A subunits, although the molecular basis for this subunit specificity is unknown. We have made chimeric receptors by fusing the murine forms of NR2A (epsilon 1) and NR2B (epsilon 2) to localize the high affinity determinants of ifenprodil inhibition on the 2B subunit. Binding experiments with 125I-MK-801 implicated the region between amino acids 198 and 356 of NR2B for high affinity ifenprodil interaction. Site-directed mutants at Arg-337 showed that this residue is absolutely required for high affinity ifenprodil inhibition. Polyamines also modulate the NMDA receptor with a preference for NR2B subunits, and the pharmacology of these agents overlaps with ifenprodil. Although the determinants of the polyamine enhancement of iodo-MK-801 binding also localize to the NH2 terminus of NR2B, the point mutants at Arg-337 form receptors that are polyamine-stimulated at wild type levels. In addition, polyamine stimulation depends on the expression of NR1 splice variants, whereas high affinity ifenprodil inhibition is independent of NR1 isoform expression. These studies provide evidence that ifenprodil and polyamines interact at discrete sites on the NR2B subunit.

Ifenprodil is a structurally unique modulator of the NMDA receptor which exhibits subunit-specific affinity for NMDA receptors (24,26). This phenylethanolamine derivative noncompetitively antagonizes the NMDA receptor either by stabilizing the closed-channel conformation of the ion channel or by causing a modal shift in the gating of the ion pore (27,28). By a different mechanism, ifenprodil can block NMDA receptors in a voltage-independent manner (24). Ligand binding experiments have also suggested two sites for ifenprodil interaction. Inhibition of both [ 3 H]MK-801 and 125 I-MK-801 and [ 3 H]TCP binding to rat brain by ifenprodil is biphasic (24,27,29), suggesting the presence of both high and low affinity ifenprodil sites. The high affinity site for ifenprodil, measured electrophysiologically, has a K i in the submicromolar range (0.2-1 M), whereas the K i for the low affinity site is 140 -300-fold higher (60 -100 M) (24). Binding and electrophysiologic studies show that ifenprodil exhibits a 140-fold preference for NR1A/NR2B (⑀ 2 ) receptors over NR1A/NR2A (⑀ 1 ) combinations (24). Therefore, the expression of NR2 subunits may underlie the biphasic nature of ifenprodil interactions with NMDA receptors.
There is much controversy over whether ifenprodil and polyamines interact at the same site on the NMDA receptor (13, 27, 30 -32). Polyamines such as spermidine or spermine interact with NMDA receptors by at least three mechanisms. Polyamines stimulate NMDA receptors by enhancing the receptor affinity for glycine (glycine-dependent stimulation) and, in saturating glycine, increase the probability of channel opening (glycine-independent stimulation) (25). At higher concentrations, polyamines can also block NMDA receptors at the channel pore in a voltage-dependent manner (25). Like ifenprodil, glycine-independent stimulation by polyamines is dependent on NR2 subunit expression. Electrophysiologic and binding paradigms both demonstrate that in saturating concentrations of glycine, polyamines stimulate NR1A/NR2B receptors whereas NR1A/NR2A combinations are polyamine-insensitive (25,(33)(34)(35)(36). Radioligand binding studies suggested additional linkage between ifenprodil and polyamine sites. Spermidine displaces both 125 I-and [ 3 H]ifenprodil from rat brains, although direct competition is unclear (37)(38)(39)(40). The polyamine stimulation of 125 I-MK-801 and [ 3 H]CPP binding can also be reduced by increasing ifenprodil concentrations (13,39), further suggesting overlapping sites. Conversely, electrophysiologic studies of ifenprodil and polyamine effects on NMDA receptors expressed in Xenopus oocytes have ruled out simple competitive interactions of ifenprodil and polyamines (24, 26 -28). The molecular determination of the ifenprodil and polyamine sites on the NMDA receptor could aid in understanding the interactions between these modulators.
We have investigated the molecular interactions of ifenprodil at NMDA receptors to define the determinants of NR2B which confer subunit-specific modulation, in an attempt to provide biochemical evidence that ifenprodil and polyamines bind to discrete sites on the NMDA receptor. We designed chimeric ⑀ 1 (NR2A)/⑀ 2 (NR2B) subunits, coexpressed them with NR1 subunits, and measured the dose-dependent inhibition of 125 I-MK-801 binding by ifenprodil for these receptors in order to localize the binding determinants of ifenprodil interaction on the NR2B subunit. The NH 2 -terminal 464 amino acids of NR2B contained determinants for both ifenprodil and polyamine interactions, and additional chimeras permitted localization of high affinity ifenprodil inhibition to amino acids 198 -356. Site-directed mutants of ⑀ 2 were then characterized to define the ifenprodil modulatory site at the molecular level. All substitutions at Arg-337 render the receptor low affinity for ifenprodil, whereas these mutant receptors maintain wild type polyamine stimulation, providing biochemical evidence for discrete modulatory sites. The distinction of polyamine and ifenprodil binding sites was corroborated by additional experiments with NR1 splice variants. Unlike polyamines (25,36), ifenprodil inhibition was shown to be unaffected by NR1 isoform expression, further suggesting distinct binding mechanisms. The molecular characterization of modulatory sites on the NMDA receptor could provide vital information for the design of novel therapeutic agents for the prevention of the neurodegeneration following cerebral ischemia.

Materials
(ϩ)-3-125 I-MK-801 was purchased from DuPont NEN. Spermidine, (Ϯ)-ketamine, ␤-mercaptoethanol, and polyethylenimine were obtained from Sigma. (ϩ)-MK-801 (hydrogen maleate) and ifenprodil (tartrate) were from Research Biochemicals International. Restriction enzymes and Taq DNA polymerase were purchased from either Life Technologies, Inc. or New England Biolabs. Sequencing was performed using the Sequenase II kit from U. S. Biochemical Corp. Monoclonal antibodies recognizing the NH 2 terminus of NMDAR1 were purchased from Pharmingen (San Diego). Goat anti-rabbit and anti-mouse peroxidaseconjugated secondary antibodies were purchased from Boehringer Mannheim. ECL chemiluminescence reagents were obtained from Amersham Corp. HEK293 cells were purchased from American Type Culture Collection (ATCC). All other reagents used were of the highest purity available and were obtained from standard commercial sources.

Chimeric NMDAR2 Subunit Construction
DNA encoding ⑀ 1 and ⑀ 2 (the murine forms of NMDAR2A and NMDAR2B, respectively) were subcloned into the SalI/EcoRI sites of the vector pRK7 downstream of the cytomegalovirus promoter, which yields high level expression of each of these subunits in mammalian cell culture systems (21,33,34,41). The first chimera (CH8) (see Fig. 1) was constructed by replacing the first 1393 base pairs of ⑀ 1 with the corresponding region of ⑀ 2 by digesting ⑀ 2 with SalI and AflII and ligating the 1.6-kilobase fragment into the same sites of ⑀ 1 . The complementary chimera (CH25) was constructed by ligating the SalI/AflII fragment of ⑀ 1 into ⑀ 2 . A chimera (CH5) containing the first 198 amino acids of ⑀ 1 and the rest being ⑀ 2 was constructed using PCR. The two primers, SP6 (5Ј primer) and REMXHO13, which introduces an XhoI site into ⑀ 1 , (5Ј-GTGATCACGTTCTCGAGATCCCAGCCC-3Ј), were used to amplify (95°C, 1 min; 55°C, 1 min; 72°C, 2 min; 30 cycles) the 5Ј region of ⑀ 1 . The 795-base pair PCR product was digested by SalI and XhoI and ligated into the corresponding sites of ⑀ 1 (cloned into the vector pRK7). CH6 was constructed by digesting CH5 with SalI/AflII and ligating the resulting 1.6-kilobase fragment into SalI/AflII-cut ⑀ 1 . An ⑀ 2 /⑀ 1 chimera containing the NH 2 -terminal 356 amino acids of ⑀ 2 , (CH48), was created using a three-step PCR method using overlapping primers. PCR reaction 1, (PCR1), used the primers SP6 (5Ј) and the overlapping primer OAPAL3E1 (5Ј-GCACGATCACC ACAAGCCTGGGGTGCACCTGG-TAGCCATCTTCACTGAAGG-3Ј), with ⑀ 2 as the template. The second PCR reaction (PCR2) used the overlapping primer OAPAL5E2 (5Ј-CCTTCAGTGAAGATGGCTACCAGGTGCACCCCAGGCTTGTGGTGA-TCGTGC-3Ј), and a 3Ј primer downstream of the AflII site, E1AFLII3 (5Ј-GTTCTGGACAGTT TCTTAAGGATGTC-3Ј), with ⑀ 1 as the template. The products of these PCR reactions anneal across a 44 base pair region containing the chimeric fusion point at the center. 1 L of each of the PCR reactions (PCR1 and PCR2) are diluted 1/10 in PCR reaction buffer and are combined, allowing homologous regions to anneal. After 4 cycles of PCR, a double stranded template is formed from the annealed products. The outside primers SP6 and E1AFLII3 are then added, and a 1.7-kilobase fragment is amplified. The final fragment is ligated into the SalI/AflII sites of ⑀ 1 . A chimera that substitutes the amino acid region 199 -356 of ⑀ 2 into ⑀ 1 (CH58) was derived from CH48 by replacing the SalI/XhoI fragment of CH48 with the SalI/XhoI fragment of CH5. A chimera encoding ⑀ 2 only at amino acids 356 -464 was created using the same overlapping primer method with the following primers and templates.
PCR3-The primers were as follows. 5Ј: E1XHOI53; 3Ј: E1AFLII3. The templates were products of PCR1ϩPCR2 elongated by four cycles of PCR.
Cleavage of the 1.7-kilobase fragment from PCR3 with XhoI and AflII and subcloning into CH5 cut with the same enzymes yielded CH84. The DNA sequences for all chimeras were verified by double-stranded dideoxy sequencing.

Site-directed Mutagenesis
The point mutations at Arg-337 were constructed using a similar three-step PCR technique. Mutagenic primers were constructed with a degeneracy at the codon for Arg-337 to allow simultaneous generation of a panel of mutations.
The products from PCR1 and PCR2 were then digested with BclI and ligated together. A third PCR product was generated using the product of the ligation as the template and the primers: 5Ј, SP6; and 3Ј, AF-LIIE23. This product was digested with SalI and AflII and cloned back into ⑀ 2 . Sequencing the final product confirmed the mutants R337A, R337P, R337K. The mutant R337Q was made using the same multistep PCR technique with the more restrictive primer, E2R337EQ (5Ј-GT-GACGTTGATCAGATA(GC)TCATTCAGCAT GTTA GACTG-3Ј, substituted for the primer E2ARGX35 above.

Cell Culture and Transfection of HEK 293 Cells
Human embryonic kidney cells (HEK293) were propagated in minimal Eagle's medium containing 5% horse serum and 5% fetal bovine serum, 2 mM glutamine, and 1 ϫ penicillin/streptomycin in a humid 95% air, 5% CO 2 atmosphere. Cells were transfected by the calcium phosphate precipitation method of Chen and Okayama with the modification that 10 M MK-801 is present during transfection to protect cells from NMDA-induced excitotoxicity (41,42).

Polyclonal Antibody Production
Polyclonal antibodies were raised against the COOH terminus of the rat NMDAR2A protein. The sequence encoding amino acids 1115-1465 was subcloned into the pRSET vector (Invitrogen) at the unique BglII site. The expression construct was transformed into the Escherichia coli strain BL21(DE3), and high levels of fusion protein were induced with isopropyl-1-␤-D-thiogalctopyranoside, (Boehringer Mannheim). Fusion proteins were purified by nickel affinity chromatography, with elution in 6 M guanidine. Following further purification by SDS-polyacrylamide gel electrophoresis, isolated bands were crushed and eluted in 100 mM NaCl, 50 mM Tris buffer, pH 7.5, and subsequently sent to Rockland Inc. for polyclonal antibody production. Initial boosts of between 250 and 500 g of fusion protein in complete Freund's adjuvant were followed at 2-week intervals by 200-g injections in incomplete Freund's adjuvant. Antibodies in the sera were capable of detecting 20 -50 ng of fusion protein by enzyme-linked immunosorbent assay screening. Further purification of the polyclonal antibodies was accomplished by affinity chromatography. Affinity columns were made by coupling 200 g of fusion protein to 6 ϫ Reacti-Gel (Pierce) in 50 mM sodium borate buffer, pH 8.5. Antiserum was applied to 3-ml columns in 20 mM Hepes, pH 7.5, and the purified antibodies were eluted with 6 M guanidine, pH 3.0, directly into 1 M Tris to elevate the pH rapidly. The yield of purified antibody from 100 ml of antiserum was approximately 100 g.

Western Blotting
Membranes from 293 cells transfected with combinations of NR1 and NR2 subunits were prepared by 2-3-s sonication in isotonic medium (0.32 M sucrose, 4 mM Hepes, pH 7.5, 1 M phenylmethylsulfonyl fluoride, 1 M pepstatin, 1 M leupeptin), followed by centrifugation at 14,000 rpm for 20 min to remove the nuclear fraction. Samples were boiled in SDS sample buffer containing ␤-mercaptoethanol and loaded on 6% polyacrylamide gels. Protein was transferred to nitrocellulose using a submarine transfer unit (Hoeffer Scientific), and blots were incubated in 2.5% dry milk in 100 mM NaCl, 50 mM Tris (1 ϫ TBS), pH 7.5, for 30 min before incubation with primary antibody. Blots were incubated in either polyclonal NMDAR2A or monoclonal NMDAR1 antibodies (3 g/ml in TBS, 2.5% dry milk) for 4 -6 h with constant agitation. After excess primary antibody was removed (once, 5 min, 0.01% Triton X-100/TBS; twice, 5 min, TBS), peroxidase-conjugated goat anti-rabbit (or mouse) secondary antibody was added at a 1/30,000 dilution in 1% dry milk, 1 ϫ TBS. Multiple washes in 0.01% Triton, 0.1% Tween 20, and TBS were followed by detection with the ECL chemiluminescence detection system.

Radioligand Binding
Cells were prepared, as described previously (33,34), for 125 I-MK-801 binding by washing in 100 M glycine, 100 M glutamate, 300 M Mg 2ϩ , 20 mM Hepes, pH 7.5 (three times, 30 min, 32°C) to remove the unlabeled MK-801 used to reduce cell death during transfection. Binding to membranes was performed in the presence of 125 I-MK-801 (150 pM), glycine (100 M), Mg 2ϩ (300 M), glutamate (100 M), and spermidine (100 M) to facilitate channel opening and ligand binding. Increasing concentrations of either ifenprodil or spermidine were added as appropriate. Incubations were for 3 h at 37°C. Each concentration was measured in duplicate with a corresponding blank containing 10 M unlabeled MK-801. Specific binding was measured on a Beckman gamma counter (model 5500B) following rapid filtration (Brandel).

RESULTS
Since the NMDA receptor subunits are large polypeptides (about 1,500 amino acids) the characterization of the molecular determinants of ifenprodil binding to NR2B was best accomplished using chimeric NR2A-NR2B subunits. Because the protein sequences of NR2A and NR2B are 50% identical (18 -20), their sequences can be effectively fused and could be expected to retain the functional properties of wild type receptors.
The murine forms of NR2A (⑀ 1 ) an NR2B (⑀ 2 ) were chosen for chimera construction due to the presence of unique restriction sites and because bacterial synthesis of the NR2B gene consistently yields low quantities of DNA. 2 Even though there is a 99% identity at the amino acid level between the mouse and rat subunits (18 -20), we needed to eliminate the possibility of dissimilar pharmacologies. Cotransfection of rat NR1A subunits with either ⑀ 1 or ⑀ 2 creates receptors with similar pharmacology as pure rat channels, consistent with previous work with NR1A/⑀ 1 receptors (21). The K d values of the receptors NR1A/⑀ 1 and NR1A/⑀ 2 for 125 I-MK-801 were within statistical error with pure rat receptors (34,35), both exhibiting values of 150 pM (data not shown). In addition, these same receptors retain the subunit specific effects of ifenprodil inhibition and polyamine stimulation. NR1A/⑀ 1 receptors have an IC 50 for ifenprodil inhibition of 59 M; NR1A/⑀ 2 receptors have an IC 50 of 0.16 M, comparable to previous results for pure rat combinations (34) (Fig. 1). 1A/⑀ 1 receptors were also not polyaminestimulated, unlike 1A/⑀ 2 combinations. Since the murine NR2 subunits retain the properties of their rat homologs, this validates the strategy of designing chimeric ⑀ 1 /⑀ 2 subunits to map the subunit-specific determinants of ifenprodil and polyamine interaction. In the present study we have designed seven chimeric ⑀ 1 /⑀ 2 receptors (Fig. 1), which were used to map the determinants of ifenprodil inhibition on the NR2B subunit.
Expression of NMDA Receptor Subunits by Western Blot-Western blot analysis demonstrated that NR1, NR2, ⑀ 1 , ⑀ 2 , and the ⑀ 1 /⑀ 2 chimeras all exhibit high levels of expression in our transfection system (Fig. 2, A-C). The splice variants of NR1A-G are all recognized by the monoclonal antibody against NR1A, and membranes analyzed by Western blot demonstrate comparable levels of expression of all forms ( Fig. 2A). The monoclonal against NR1 also detects the fainter breakdown products of the NMDAR1 protein, exhibiting different patterns depending on the splice variant transfected. A polyclonal antibody raised against the COOH terminus of the NR2A was used to verify NR2 subunit expression. This antibody recognizes NR2A but not NR2B or NR2C or any of the splice variants of NR1 (Fig. 2B). Fig. 2B shows a Western blot of two of the chimeras: CH8, which contains the COOH terminus of ⑀ 1 , and CH25, which contains the COOH terminus of ⑀ 2 . Bands running at the predicted molecular mass (165 kDa) demonstrated high level expression of both chimeras. All of the other five chimeric NR2 subunits showed similar levels of expression by Western blot (data not shown). The total 125 I-MK-801 binding for all chimeras, which ranged from 32 to 55 fmol/mg protein (with 150 pM 125 I-MK-801), was comparable to wild type receptors (1A/⑀ 1 , 58 fmol/mg protein; 1A/⑀ 2 , 87 fmol/mg protein), demonstrating high level expression for chimeric NR2 subunits (Fig. 2C). Based on total radioligand binding, all eight splice variants have expression levels comparable to combinations containing NR1A (Table I).
The NH 2 Terminus of ⑀ 2 Is the Site of High Affinity Ifenprodil Interaction-The region of NR2B which most likely mediates subunit specificity is in the more divergent NH 2 -terminal region, which is proposed to create the extracellular face of the receptor (28,36). The first chimeras (CH8 and CH25) were constructed by exchanging the NH 2 -terminal thirds of ⑀ 1 and ⑀ 2 (Fig. 1). 125 I-MK-801 binding to receptors formed by cotransfection of NR1A/CH8, which contains the NH 2 -terminal third of ⑀ 2 , is stimulated by spermidine, whereas 1A/CH25 (the complementary chimera) expression produced polyamine-insensitive receptors. Similarly, coassembly of CH8 with NR1A led to high affinity ifenprodil displacement (1.6 M) 125 I-MK-801 binding, whereas NR1A/CH25 receptors exhibited low affinity inhibition (65 M) (Fig. 3). These results implicated the NH 2 -terminal third of ⑀ 2 in both glycine-independent polyamine stimulation and high affinity ifenprodil inhibition.
High Affinity Ifenprodil Inhibition Is Mediated by Arg-337-An important determinant for high affinity ifenprodil inhibition was discovered by site directed mutagenesis of ⑀ 2 . Ifenprodil may act as an acid by losing a phenolic hydrogen in physiologic conditions; thus, efficacious ifenprodil binding could be mediated by a positively charged residue on the NR2B subunit such as lysine or arginine. The region of ⑀ 2 between amino acids 198 and 357 contains a single positively charged amino acid that was conserved in both ⑀ 2 and 2B, but not in ⑀ 1 or 2A, namely Arg-337. Site-directed mutants were then created at Arg-337 by the use of PCR mutagenesis, and all of these mutants showed expression levels comparable to wild type receptors (Table II). Mutation of Arg-337 to lysine (NR1A/ R337K) exhibited an IC 50 for inhibition of 125 I-MK-801 binding by ifenprodil of 120 M, even lower affinity than ⑀ 1 (Table II and Fig. 5). Arg-337 was also changed to alanine (R337A), proline (R337P), and glutamine (R337Q), which is the conserved amino acid in ⑀ 1 and NR2A. The IC 50 values for R337A, R337P, and R337Q were 83, 100, 65 M, respectively. As any residue other than arginine at position 337 attenuates high affinity ifenprodil inhibition, Arg-337 is a functional determinant of high affinity ifenprodil binding. Although the differences in IC 50 between ⑀ 1 and both ⑀ 2 and the Arg-337 mutants for ifenprodil are too great to be explained only by differences in their affinities for 125 I-MK-801, saturation analysis of 1A/⑀ 1 , 1A/⑀ 2 , and R337Q was performed. The K d values for 125 I-MK-801 of 1A/⑀ 1 , 1A ⑀ 2 , and R337Q were 124, 133, and 145 pM, respectively, demonstrating that the observed changes in ifenprodil potency do not result from changes in the affinity for 125 I-MK-801 and that receptor integrity is preserved in these mutants.
Splice Variants of NMDAR1 Distinguish between Ifenprodil and Polyamines-Although the ⑀ 2 -specific properties of polyamine stimulation and high affinity ifenprodil interaction map to similar regions, experiments with the splice variants of NR1 suggest that these agents bind to distinct sites. Since NR1 subunits containing the NH 2 -terminal insert, such as NR1B, are polyamine-insensitive (25), we studied the effects of NR1 isoform expression on ifenprodil inhibition to determine whether both ifenprodil and polyamines share splice variantspecific properties. We found that NR1B/⑀ 1 and NR1B/⑀ 2 receptors share the same subunit-specific effects as NR1A combinations, with IC 50 values for ifenprodil inhibition of 37 and 0.52 M, respectively (Table I and Fig. 6). Experiments with NR1C and NR1G yielded similar results (  1. Design of the ⑀ 1 /⑀ 2 chimeras. The schematic primary sequences depict how portions of ⑀ 1 (white regions) and ⑀ 2 (black regions) are fused to form chimeric NR2 subunits. The IC 50 values for ifenprodil inhibition were calculated using the best of either one-or two-sited fits using PROPHET. In all cases of two-sited fits the percent high affinity represented 65% of the total inhibition; low affinity was 35%. Data are representative of 4 -14 experiments. Polyamine stimulation is shown by either ϩ ϩ ϩ (150 -170% stimulation over baseline), ϩ (50% stimulation), Ϫ (Ͻ10% spermidine enhancement), or Ϫ Ϫ Ϫ (0% polyamine stimulation). even with chimeric receptors. Unlike polyamines, the inhibition of MK-801 binding by ifenprodil is not affected by NR1 isoform expression.
NR2 Chimeras and R337 Mutants Demonstrate Discrete Modulatory Sites-The chimeras CH5, CH6, CH48, CH58, and CH84 were used to determine if polyamine stimulation and ifenprodil inhibition localize to the same region of ⑀ 2 , although the results were not as straightforward as for ifenprodil inhibition. 125 I-MK-801 binding to 1A/CH5 and 1A/CH48 receptors is stimulated by spermidine to the same extent as 1A/CH8, whereas 1A/CH6 and 1A/CH58 are stimulated only 50% as much as wild type receptors (Fig. 4). This suggests that the determinants of polyamine stimulation, as for ifenprodil, are found in the region between amino acids 198 and 356, although a significant loss in polyamine stimulation occurs when the chimera fusion point is at amino acid 198. Surprisingly, there was a slight spermidine stimulation of 1A/CH84 receptors, which suggests that the binding determinants of polyamine stimulation may involve more regions of NR2B than for ifenprodil inhibition.
Although the determinants for glycine-independent polyamine stimulation map to the same general region of NR2B, results with the Arg-337 mutants gave the most conclusive evidence that ifenprodil and polyamine sites were indeed distinct. All four mutants at Arg-337 were found to be stimulated to the same extent as ⑀ 2 (Fig. 5 and Table II), thus Arg-337 is not a determinant of glycine-independent stimulation by polyamines.

DISCUSSION
Chimeric ⑀ 1 /⑀ 2 subunits, coupled with site-specific mutagenesis, permitted the localization of high affinity inhibition to the NH 2 terminus of NR2B and distinguished it from glycineindependent polyamine stimulation. To utilize the murine forms of NR2 subunits, we had to demonstrate that they exhibit the same pharmacology as their rat homologs. The murine forms show the same magnitude and subunit specificity for the ifenprodil inhibition and polyamine stimulation of 125 I-MK-801 binding as pure rat receptors (24,25,35). The ⑀ 1 and ⑀ 2 forms of NR2 subunits thus proved to be the ideal tool for chimera  construction and characterization of the effects of ifenprodil and polyamines. The expression of the murine forms of NR2 was comparable to that of NR2A and NR2B. Western blot analysis using a polyclonal antibody against NR2A demonstrated a high level of protein expression of both native ⑀ 1 and ⑀ 2 receptors and of all the chimeric NR2 subunits characterized. Our NR2A polyclonal was unable to discriminate between ⑀ 1 and ⑀ 2 , which may be due to 33% homology between ⑀ 2 and NR2A in the polyclonal recognition region (18 -20). The peptide sequences of NR2A and NR2B in this region are less than 20% similar and share no homology to NR2C or D, possibly explaining antibody specificity for the rat form of NR2A. By preabsorbing the antibodies in our polyclonal mixture on an ⑀ 1 affinity column it may be possible to isolate a population that demonstrates total specificity for NR2A and ⑀ 1 , obtaining a valuable tool for the biochemical characterization of NMDA receptors. Definitive evidence for high level expression was confirmed by the level of MK-801 binding, which was between 30 and 110 fmol/mg of protein for all forms of murine receptors when coexpressed with rat NR1A. Thus, expression of all of the chimeric subunits and the ⑀ 2 site-directed mutants was comparable to wild type receptors. Since we have shown previously that NR1A homomeric receptors bind insignificant levels of 125 I-MK-801 (34), which is generally consistent with [ 3 H]MK-801 binding experiments (43-45), our wild type binding levels confirm that our chimeric and mutant NR2 subunits efficiently coassemble to create intact MK-801 binding sites.
The divergent effects of the splice variants of NR1 on both ifenprodil inhibition and polyamine stimulation confirmed the distinct structural determinants between these modulators. Polyamine stimulation in electrophysiologic experiments depends on which splice variant of NR1 is expressed (16,17,24,46,47). Neither NR1B homomeric receptors nor NR1-NR2 combinations with the NH 2 -terminal insert (such as NR1B) display glycine-independent stimulation by polyamines. The effects of ifenprodil on receptors expressed with different splice variants had not been well characterized. We have shown that the high affinity inhibition by ifenprodil is not dependent on the NR1 subunit but is regulated by the NR2 subunit. All of the NR1 splice variants (A-H), when coexpressed with NR2 subunits, formed receptors whose modulation was governed only by NR2  expression. Conversely, polyamine stimulation was not observed for either NR1B/⑀ 1 or NR1B/⑀ 2 , whereas the NR1B/⑀ 2 receptor exhibited the same 140-fold greater affinity for ifenprodil as seen for combinations with NR1A. Clearly, the mechanisms by which ifenprodil and polyamines interact with the NMDA receptor differ. The use of chimeric receptors facilitated the discrete mapping of the site for high affinity ifenprodil inhibition. A major determinant of high affinity ifenprodil inhibition localizes to Arg-337 on the ⑀ 2 subunit. There are at least three possibilities for the mechanism of action for Arg-337. First, ifenprodil may directly bind to Arg-337 of the ⑀ 2 subunit. The localization of high affinity determinants to NR2B is consistent with the dramatic differences in affinity between 1A/⑀ 1 and 1A/⑀ 2 receptors and by the lack of NR1 splice variant-specific modulation. The fact that both in situ hybridization studies of NR2B mRNA and radiolabeled ifenprodil experiments show a strong correlation between high affinity ifenprodil binding and the developmental and regional profiles of NR2B expression also strengthens the argument that residues of NR2B interact directly with ifenprodil (26,39,48). Electrophysiologic evidence using outside-out patches has demonstrated that the high affinity ifenprodil site is located on the extracellular portion of the NMDA receptor (28). Theoretical models of the transmembrane architecture of the ⑀ 2 subunit are consistent with Arg-337 being present on this extracellular surface (36).
Mutants at Arg-337 could potentially alter the affinity of 1A/⑀ 2 for either the agonist (glutamate) or coagonist (glycine), affecting the association of MK-801 to open channels. This is ruled out by the fact that our assay system utilizes high excess concentrations of both glutamate and glycine and that the K d for 125 I-MK-801 of all our chimeras and point mutants were identical. The binding of MK-801 acts as a good internal control for receptor integrity. Functional high affinity MK-801 binding requires the presence of both functional glycine and glutamate sites and a structurally intact channel pore (33,34); thus the changes in ifenprodil inhibition mediated by Arg-337 must be distinct from effects on either agonist or coagonist sites. Significant reduction of either glycine or glutamate affinity in mutant receptors results in receptors that do not bind 125 I-MK-801 (34,35,49). The glycine coagonist site has recently been localized to the aromatic residues 390, 392, 466 and the charged residues 481 and 483 of the NR1 subunit. These residues are not only distal to Arg-337, but also are present exclusively on NR1 (50,51). Finally, Arg-337 may interact directly with the NR1 subunit where the true binding site for ifenprodil resides. NR1A mRNA injected into Xenopus oocytes yields functional homomeric channels with a high affinity for ifenprodil (0.28 M); thus the high affinity binding site for ifenprodil was thought to reside on the NR1 subunit or the association of multiple NR1A subunits (26). Homomeric channels, although functional, lack many of FIG. 5. Arg-337 of NR2B mediates high affinity ifenprodil inhibition. All of the site-directed mutants of Arg-337 cause at least a 300-fold shift in the IC 50 of Ifenprodil inhibition from wild type receptors. The ifenprodil inhibition curves for 1A/⑀ 1 (E), 1A/⑀ 2 (q), 1A/R337A (É), 1A/R337K (᭛), and 1A/R337Q (ࡗ) are shown. The lower panel shows that the mutants 1A/R337A and R337K are both polyaminestimulated to the same extent as 1A/⑀ 2 , providing direct biochemical evidence that ifenprodil and polyamines interact at discrete sites on the NR2B subunit.
FIG. 6. High affinity ifenprodil inhibition is not dependent on NR1 splice variant expression. The inhibition of 125 I-MK-801 binding by ifenprodil for 1B/⑀ 1 and 1B/⑀ 2 receptors shows the same 100-fold difference in apparent ifenprodil affinity as 1A/⑀ 1 and 1A/⑀ 2 . Binding data and theoretical curves for 1A/⑀ 1 (E) and 1B/⑀ 1 (᭛) both demonstrated low affinity ifenprodil inhibition, whereas 1A/⑀ 2 (q) and 1B/⑀ 2 (ࡗ) were high affinity. Conversely, neither of the 1B-containing receptors showed polyamine enhancement of iodo-MK-801 binding (lower panel), demonstrating that polyamine stimulation is dependent on NR1 isoform expression, whereas high affinity ifenprodil inhibition is not. the characteristics of native NMDA receptors and have not been conclusively shown to exist in vivo (21,26,33,34). Multiple ifenprodil binding sites, present on both NR1 and NR2 subunits, could also exist. There is a considerable sequence homology between NR1 and ⑀ 2 (NR2B) near Arg-337 (11, 14, 18 -20). An arginine residue exists in NR1 (Arg-344) at the comparable position of Arg-337 in ⑀ 2 and may be the site of high affinity ifenprodil inhibition found in homomeric receptors. New chimeras and site-directed mutants of both NR1 and NR2 subunits will help gain future insight into the mechanism by which high affinity ifenprodil binding occurs.
The 300-fold difference in ifenprodil affinity between NR2Aand NR2B-containing receptors can best be explained by an electrostatic interaction occurring at the high affinity ifenprodil binding site. The chemical structure of ifenprodil contains no obvious ionizable groups such as amines but does possess a phenyl ring with a hydroxyl group attached. Tyrosyllike groups may become phenolate ions following the loss of a proton from the phenyl hydroxyl group (52)(53)(54)(55)(56). The O Ϫ is stabilized through conjugation with the double-bond structure of the phenyl ring. An electrostatic interaction between ionized ifenprodil and one or more basic amino acid residues of the NMDA receptor could be proposed. The energy loss from the disruption of an electrostatic interaction is believed to be approximately 3-5 kcal (57). This change in apparent binding energy would account for a change in K d of approximately 150 -4,000-fold. The 300-fold difference in IC 50 is consistent therefore with the loss of a weak electrostatic interaction in ⑀ 1 -bearing receptors. Arg-337 is the only basic amino acid residue that is conserved in both ⑀ 2 and 2B between amino acids 198 and 356, whereas glutamine is found at this position in ⑀ 1 and 2A. Surprisingly, even substitution of the basic residue lysine at position 337 renders the receptor low affinity, suggesting that not only is a positively charged residue necessary at residue 337, but specifically arginine. Since the orientations of the positively charged moieties of lysyl and arginyl side chains differ, it is likely that the precise positive charge alignment of Arg-337 is required for efficacious high affinity ifenprodil binding. Glutamine substitution at this position exhibits the least detrimental effect on the IC 50 of ifenprodil inhibition, presumably because the glutamine side chain has a surface volume most similar to that of arginine. Additional point mutations will be necessary to define further the involvement of Arg-337 in NMDA receptor modulation.
Although some components of the NR2B-specific effects of ifenprodil and polyamines overlap, results of the NR1 splice variant experiments and the Arg-337 mutation experiments provide biochemical evidence for distinct polyamine and ifenprodil binding sites. The dissimilar structures of spermidine and ifenprodil make competitive binding arguments unlikely. Even though the long aliphatic chain and amine group of spermidine differ from the phenylethanolamine structure of ifenprodil, some of the pharmacologic properties of polyamines and ifenprodil overlap. Ifenprodil blocks the stimulatory effects of polyamines on both TCP and MK-801 binding and inhibits the increase in [ 3 H]CPP binding facilitated by spermidine (39). Polyamines antagonize the partial displacement of [ 3 H]glycine by ifenprodil (32). These overlapping effects may be explained by the determinants of polyamine stimulation on the ⑀ 2 subunit being between amino acids 198 and 293, which is potentially close to the site of high affinity ifenprodil binding. Curiously, the homologous region of NR1 (amino acids 190 -211) is the location for the 63-residue insertion that renders splice variants such as NR1B polyamine-insensitive (16,17,25). Sitedirected mutants of this region of NR1A eliminate polyamine stimulation (46). Although the determinants of polyamine stimulation on ⑀ 2 have not yet been characterized at the amino acid level, it seems likely that the region from amino acids 198 to 464 will include some component of the glycine-independent polyamine stimulation region of the ⑀ 2 subunit. The binding sites for ifenprodil and polyamines are biochemically distinct, although their determinants on the NR2B subunit are at least allosterically linked if not overlapping. Since the binding of polyamines and possibly ifenprodil involves both the NR1 and NR2 subunits, biochemical information about NR1-NR2 interactions could be studied by closer examination of the allosteric linkage between ifenprodil inhibition and polyamine stimulation.
There is much current interest in the ability of ifenprodil to act as a neuroprotective agent during focal cerebral ischemia and as an anticonvulsive agent (13,25,58). The interaction between ifenprodil and the NMDA receptor may underlie this neuroprotective ability. Unlike many other neuroprotective agents, ifenprodil and the derivative SL 82.0715, which has a better oral bioavailability, do not cause any behavioral effects and have already been used clinically for the treatment of hypertension and cerebral ischemia (13,58). The location and mechanism of ifenprodil action on the NMDA receptor are still not completely understood. By identifying more residues like Arg-337, which directly participate in modulating the function of NMDA receptors, and by characterizing these modulatory sites at the molecular level, it will be possible to design additional novel therapeutic agents to combat the neurodegeneration that follows events such as stroke. Acknowledgments-We give special thanks to Dr. Michael Robinson, Dr. Brian Basckai, and Elfrida Grant for helpful comments on this manuscript.