Molecular Characterization of the Family of the N-Methyl-D-Aspartate Receptor Subunits*

cDNA clones for four different N-methyl-D-aspar- tate (NMDA) receptor subunits Control hybridization experiments were carried out in adjacent sections by using the same probes in the presence of an excess amount of unlabeled probes.

cDNA clones for four different N-methyl-D-aspartate (NMDA) receptor subunits (NMDARPA-NMDARZD) were isolated through polymerase chain reactions followed by molecular screening of a rat brain cDNA library. These subunits are only about 15% identical with the key subunit of the NMDA receptor (NMDARl) but are highly homologous (-50% homology) with one another. They also commonly possess large hydrophilic domains at both amino-and carboxyl-terminal sides of the four putative transmembrane segments. NMDARZA and NMDARZC expressed individually in Xenopus oocytes showed no electrophysiological response to agonists. However, these subunits in combined expression with NMDARl markedly potentiated the NMDARl activity and produced functional variability in the affinity of agonists, the effectiveness of antagonists, and the sensitivity to Mg2* blockade. Thus, NMDARl is essential for the function of the NMDA receptor, and multiple NMDARZ subunits potentiate and differentiate the function of the NMDA receptor by forming different heteromeric configurations with NMDARl. Northern blotting and in situ hybridization analyses revealed that the expressions of individual mRNAs for the NMDARP subunits overlap in some brain regions but are also specialized in many other regions. This investigation demonstrates the anatomical and functional differences of the NMDARZ subunits, which provide the molecular basis for the functional diversity of the NMDA receptor.
The diverse functions of glutamate neurotransmission in the mammalian central nervous system are mediated by a variety of glutamate receptors that are classified into two major groups termed ionotropic and metabotropic glutamate receptors (1). The ionotropic receptors can be subdivided into two distinct types of receptors, the receptors for N-methyl-D-* This work was supported in part by research grants from the Ministry of Education, Science, and Culture of Japan, the Ministry of Health and Welfare of Japan, the Senri Life Science Foundation, and Yamanouchi Foundation for Research on Metabolic Disorders. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The NMDA receptor plays a key role in many functions of glutamate transmission in the central nervous system. This receptor is essential for inducing long term potentiation, a long lasting change in neuronal responsiveness that is thought to underlie learning and memory (1,6). It also plays a critical role in pathophysiological processes such as epilepsy and acute and delayed ischemic neuronal cell death, as well as some neurodegenerative diseases (7). The NMDA receptor is distinguished from other ionotropic receptors by the actions of a number of selective agonists and antagonists and also by its peculiar properties, including high Ca2+ permeability, modulation by glycine, voltage-dependent Mg2+ blockade, and inhibition by Zn2+ and several selective open channel blockers (1). The intracellular Ca2+ increase through the activation of the NMDA receptor is thought to be the key event in evoking both glutamate-mediated neuronal plasticity and neurotoxicity (6,7). Voltage-dependent Mg2+ blockade is also postulated to be crucial for changing synaptic efficacy as occurs in long term potentiation (6).
Using a Xenopus oocyte expression system combined with electrophysiology, our group has cloned and characterized the key subunit of the NMDA receptor (NMDARl), a 938-amino acid protein with four putative transmembrane (TM) segments following a large extracellular amino-terminal domain (3). The cloned receptor expressed in oocytes faithfully reproduces the pharmacological and electrophysiological properties characteristic of the NMDA receptor by forming a homooligomeric assembly (3). However, the existence of different subunits of the NMDA receptor was also suggested on the basis of the findings of not only the low efficacy in NMDA responses elicited by homomeric expression of NMDARl but also some disparities between the radioligand-binding sites and the NMDARl mRNA expression sites (3). Recently, the isolation of cDNA clones encoding three different subunits of the NMDA receptor in the rat and mouse has been reported from two laboratories (8-10). These subunits, termed NMDARBA-NMDAR2C (also referred to as 61-63, respectively), have been shown to exhibit no NMDA receptor activity in a homomeric configuration but to greatly potentiate the response to NMDA when co-expressed with NMDARl (8-10). In this investigation, we extended our molecular screening of the NMDA receptor subunits and identified four cDNA The abbreviations used are: NMDA, N-methyl-o-aspartate; D-APV, ~-(-)-2-amino-5-phosphonovalerate; TM, transmembrane; PCR, polymerase chain reaction; bp, base pairs.

Cloning and
clones encoding the rat NMDAR2 subunits. TWO of them corresponded to NMDAR2A and NMDAR2B. The third one, although corresponding to NMDAR2C, differed from the reported NMDAR2C not only in the carboxyl-terminal sequence of the corresponding rat subunit but also at a certain limited region of the mouse counterpart. The fourth one represented a novel type of the NMDAR2 subunits. We here report the structures, properties, and expression patterns of the four rat NMDAR2 subunits.
Polymerase Chain Reaction (PCR)-The isolation of cDNA fragments encoding the NMDAR2 subunits was attempted by PCR techniques using two successive approaches. In the first attempt, a forebrain cDNA library (3) was divided into 24 fractions, each consisting of -3000 independent cDNA clones and was used as DNA templates for PCR amplification. The sequence of the 5' primer was the T7 promoter sequence (5'-AATACGACTCACTATAG-3') contained in the vector DNA of the forebrain cDNA library. The sequence of the 3' primer was the 5'-CGGAATTCCGCNGCNAA/GA/ GTTNGCNGCNGTA/GTA-3' sequence (NP-3) made according to the amino acid sequence YTANLAA of residues 647-653 of NMDARl, which is conserved in all ionotropic glutamate receptors (3,11,12). PCR amplification was performed by using AmpliTaq DNA polymerase according to the following schedule: five cycles at 94 "C for 1 min, 46 'C for 1 min, and 72 "C for 2 min, followed by 26 cycles at 94 "C for 1 min, 55 "C for 1 min, and 72 "C for 2 min. An aliquot of the PCR products was electrophoresed on an agarose gel and blotted to a nylon filter. The filter was hybridized to a mixture of 32P-labeled cDNA probes encoding nine ionotropic glutamate receptors (GluRl-6, KA-1, KA-2, and NMDARl (3,11,12)). Several PCR-amplified DNA fragments that were seen by ethidium bromidestaining but not by hybridization with the above cDNA probes were identified and excised for subsequent subcloning and sequence determination. The deduced amino acid sequence of one of the cloned cDNAs (psN2) differed from any of the ionotropic receptors so far reported but shared a similar hydrophobicity profile and amino acid sequence with the other ionotropic glutamate receptors. This amino acid sequence also contained an asparagine residue characteristic of NMDARl at the corresponding position of the putative TM I1 segment and the WNGMIGE sequence that is conserved at equivalent positions of NMDARl (residues 498-504 of NMDARl). A degenerative oligonucleotide, 5'-TGGAAC/TGGNATGATGGGNGA-3' , corresponding to the conserved WNGMM/IGE sequence was synthesized, and the NP-3 and NM-3 oligonucleotides were used as the 5' and 3' primers for the second PCR amplification to isolate cDNA fragments encoding different NMDAR2 subunits. PCR amplification and subsequent identification of cDNA fragments were carried out according to the same procedures as described above, except that a mixture of oligo(dT)-primed cDNAs was synthesized from rat forebrain poly(A)+ RNA by Superscript" RNase H-reverse transcriptase (3) and used as templates for PCR amplification. Through this procedure, we identified three additional cDNA clones (psN3, psN4, to that of psN2. and psN5) whose deduced amino acid sequences were closely related cDNA Cloning--1.5 X lo5 phage clones of rat forebrain cDNA libraries containing cDNA inserts of either 3-5 kilobase pairs or more than 4 kilobase pairs were used for the isolation of full sizes of cDNAs encoding individual NMDAR2 subunits. The hybridization probes used were prepared from cDNA inserts psN2-psN5. The 4,3,24, and 4 hybridization-positive clones identified by hybridization with the respective cDNA probes were rescued into plasmids according to the in uiuo excision procedure. The plasmids containing the largest cDNA inserts (pNRZA, pNRZB, pNRZC, and pNR2D-1) were selected from each group of the hybridization-positive clones and were subjected to sequence determination. An additional clone (pNR2D-2) of the pNR2D group, which differed from pNR2D-1 by restriction mapping at a certain limited region of the inserted cDNA, was also subjected to sequence determination. Both strands of the cDNA sequences were determined by the chain termination method (13).
Electrophysiology-Preparations of oocytes, in vitro RNA synthesis, and electrophysiological measurements were carried out as described (3). In order to achieve efficient translation in Xenopus oocytes, the 5"noncoding region of each cDNA clone was largely removed by digestion with either appropriate restriction enzymes or exonuclease 111. Unless otherwise stated, electrophysiological responses of different members of the NMDAR2 subunits (-5 ng of mRNA each) were analyzed in combined expression with the NMDARl subunit (-0.5 ng of mRNA). Oocytes were perfused by a constant stream of a standard solution (95 mM NaC1, 2 mM KCl, 2 mM CaClZ, 5 mM HEPES, pH 7.5), and drugs tested were applied by switching the flow. Dose-response curves for L-glutamate and NMDA, and inhibition profiles of various antagonists were determined by measuring steady-state currents after serial application of various concentrations of these compounds in BaZ+-Ringer solution (2 mM CaCIZ in the standard solution was replaced with 2 mM BaC12). Current-voltage curves were constructed by ramping voltage slowly (56 mV/s) from -100 to 40 mV during application of 100 pM NMDA and 10 p~ glycine in the presence and the absence of M e (either 100 p~ or 1 mM) in BazC-Ringer solution. Leakage current in the absence of agonists was subtracted from currents measured in the presence of agonists. Currents recorded were analyzed using pCLAMP (Axon Instruments).
Northern Blotting-The brains of adult male Sprague-Dawley rats were dissected into four portions (the cerebral cortex, hippocampus, subcortical regions, and cerebellum). Total RNA (10 pg) isolated from each portion was fractionated in a formaldehyde-agarose gel and blotted on Biodyne nylon membranes (3). 32P-labeled probes were hybridized to the membrane filter as described (3); the probes used were the 1048-bp BgnI fragment of pNRBA, the -800-bp Sac1 fragment of pNRZB, the 751-bp XhoI-EcoRV fragment of pNRZC, and the 1703-bp ApaI fragment of pNR2D-1. Filter washing was carried out in 30 mM NaC1, 3 mM sodium citrate, and 0.1% sodium dodecyl sulfate at 65 "C.
In Situ Hybridization-In situ hybridization was performed as previously described (3). Briefly, 35S-labeled anti-sense riboprobes for the NMDARBA-2D subunits were transcribed in vitro and hybridized with cryostat sections (10 pm) of adult rat brain. After washing and RNase A treatment, the sections were exposed to 0film (Amersham C o p ) and developed after a 1-week exposure. Control hybridization experiments were carried out in adjacent sections by using the same probes in the presence of an excess amount of unlabeled probes.

RESULTS
cDNA Cloning of Four NMDAR2 Subunits-The two-step approach was used for the isolation of cDNA fragments encoding the NMDAR2 subunits by PCR amplification. The sequence comparison of NMDARl and other ionotropic glutamate receptors revealed an identical sequence, YTANLAA (residues 647-653 of NMDARl), at the vicinity of the putative TM I11 segment of these receptors (3,11,12). We chose this amino acid sequence for the synthesis of a degenerative oligonucleotide used for the 3' primer of PCR. However, there was no further noticeably conserved amino acid sequence to be used for the 5' primer of PCR. Because the T7 promoter sequence is located upstream of cDNA inserts in the vector DNA of a rat forebrain cDNA library, we chose the T7 promoter sequence as the 5' primer of PCR. The forebrain cDNA library was subdivided into 24 fractions, each containing -3000 cDNA clones.
Each fraction was used as DNA templates for PCR amplification by using the above 5' and 3' primers. PCR-amplified products were analyzed by blot hybridization with probes consisting of a mixture of nine species of ionotropic glutamate receptors (GluR1-6, KA-1, KA-2, and NMDAR1). Many of the DNA fragments hybridized to these probes, but some did not. The unhybridized fragments were excised, subcloned, and sequenced. One of the cloned DNA fragments (psN2) did not correspond to any of the known ionotropic glutamate receptors but encoded a polypeptide that shared a similarity with this receptor family

Cloning and
Characterization of NMDA Receptor Subunits in both the hydrophobicity profile and amino acid sequence. In addition, the predicted amino acid sequence contained an asparagine residue characteristic of the NMDARl subunit at an equivalent position of the putative TM I1 segment. The amino acid sequence deduced from clone psN2 showed a n additional sequence conservation (WNGMI/MGE) with NMDARl at the preceding region of the TM I segment. As a second step of cDNA cloning of the NMDAR2 subunits, a degenerative oligonucleotide primer corresponding to this conserved sequence was synthesized, and this primer in combination with the above degenerative oligonucleotide primer was used for the second PCR amplification experiments. The homologous sequences were enriched by PCR amplification from a mixture of oligo(dT)-primed cDNAs of the rat forebrain mRNA, and amplified DNA products were electrophoresed, excised, and subcloned for sequencing. Through this procedure, we identified three additional cDNA clones (psN3, psN4, and psN5), which closely resembled psN2 in their deduced amino acid sequences. To obtain full sizes of cDNA inserts for individual receptor subunits, we screened a rat forebrain cDNA library by hybridization with the cDNA probes derived from psN2-psN5. More than one clone was isolated for each receptor subunit, and the nucleotide sequences of representative clones containing the largest cDNA inserts (pNRSA, pNR2B, pNR2C, and pNR2D-1) were determined by the chain termination method. In addition, the nucleotide sequence of a different cDNA clone pNR2D-2 of the pNR2D group was determined at the region where restriction patterns were different between pNR2D-1 and pNR2D-2.
Deduced Amino Acid Sequences of Four NMDAR Subunits-The nucleotide sequences determined for the four cDNA clones indicated that they contained large open reading frames with an overall sequence similarity with one another. The reading frame of pNRSB, however, was smaller in its 5'terminal portion than those of the other cDNA clones. The cDNA inserts in the remaining two clones of this group were analyzed by restriction enzyme and were found to similarly lack the 5"terminal portion. Around the time we completed the nucleotide sequence determination and characterization of four different species of cDNA clones ( P N R~A -~N R~D ) , two other laboratories reported the molecular cloning of three subunits of the NMDA receptor family. There are some differences between the amino acid sequences revealed in this study and those reported from the two laboratories as discussed below. However, because three of the four amino acid sequences reported in this study principally corresponded to those of the rat NMDAR2 subunits reported by Monyer et al. (9), we adopted the terms of NMDARSA-NMDAR2C in their report (9) for the corresponding NMDA receptor subunits revealed in this study; these subunits were also referred to as tl-c3 for the mouse counterparts, respectively (10). Fig. 1 shows the amino acid sequences of the NMDAR2 subunits on the basis of the sequence determination of our cDNA clones, pNR2A-pNR2D. The amino acid sequence deduced from clone pNR2A was identical with the NMDAR2A sequence reported by Monyer et al. (9) throughout the protein sequence except for 2 amino acid substitutions at residues 246 (Phe/Leu) and 758 (Thr/Ser). As expected, the sequence predicted from clone pNR2B lacked 255 amino acid residues at the amino-terminal portion of the NMDAR2B sequence reported (9). The sequence deduced from pNR2C agreed with the NMDAR2C sequence (9) at the amino-terminal portion, but there were noticeable differences between our sequence and those reported by Monyer et al. (9) and Kutsuwada et al. (10). 1) When the translation initiation codon was taken as the first ATG in the large open reading frame of our cDNA sequence, there was a 13-amino acid extension beyond the ATG codon that was assigned as the initiation codon by the two groups. 2) Our NMDARBC sequence agreed with that reported by Monyer et al. (9) up to residue 967, except for 4 amino acid substitutions at residues 149 (Gly/Thr), 201 (Ala/Arg), 227 (Arg/Pro), and 710 (Arg/ Gly). The following sequences, however, completely diverged from each other. The sequence reported by Monyer et al. (9) contained 8 amino acids following the common sequence and terminated at residue 975, whereas our sequence extended up to residue 1250. This sequence difference seems to result from the presence or absence of 1 nucleotide at 5 consecutive G residues in the open reading frame of the cloned cDNAs, because an alternative reading frame in our NMDARBC cDNA sequence results in the appearance of the 8-amino acid sequence and its following termination codon reported by Monyer et al. (9). 3) The mouse NMDAR2C sequence reported by Kutsuwada et al. (10) had a similar large carboxyl-terminal extension and was highly homologous to our sequence up to the carboxyl end (-95% amino acid homology). However, a significant sequence divergence was noted at a limited portion of the two sequences (from residue 988 to residue 1024 of our NMDAR2C sequence), and this sequence difference can again be explained by the frame shift of the nucleotide sequence in this region of our cloned cDNA. We determined the nucleotide sequences of four independent cDNA clones of the pNR2C group and confirmed that the four clones all possess an identical sequence at the above regions. Thus, the observed sequence differences may reflect DNA polymorphism in rats of different laboratories, species difference, or cloning artifacts during cDNA synthesis, cDNA cloning, or sequence determination. The amino acid sequence deduced from pNR2D clones showed an overall homology with the NMDARBA-NMDAR2C but differed from any of these NMDAR2 subunits. We thus named this subunit NMDARZD. Two independent pNR2D cDNA clones (pNR2D-1 and pNRBD-2), however, had 82-nucleotide deletions/additions at the region corresponding to the carboxyl terminus of NMDAR2D and thus encoded two different carboxyl-terminal sequences, as indicated in Fig. 1. On the basis of the sequence determination of our cDNA clones, we concluded that NMDARZA, R2B, R2C, R2D-1, and R2D-2 consist of 1464, 1482, 1250, 1356, and 1323 amino acid residues, respectively, all of which are considerably larger than NMDARl (938 amino acid residues).
Structural Features of Four NMDARB Subunits-The hydrophobicity analysis revealed an overall structural similarity between the NMDAR2 subunits and other ligand-gated ion channels (14). The NMDARB subunits possess five hydrophobic segments consisting of more than 20 uncharged amino acid residues, one at the amino terminus probably serving as a signal peptide and the others forming four transmembrane segments at the middle portion of the protein sequences. The NMDAR2 subunits, like NMDAR1, are thought to comprise four membrane-spanning domains preceded by a large extracellular sequence. These subunits, however, show a peculiar structural feature, as compared with other ligand-gated ion channels in that they all contain the large carboxyl-terminal extension following the TM IV segment.
The amino acid sequence homology is unexpectedly low (-15%) between each member of the NMDARB subunits and the NMDARl subunit, but considerably higher (-50%) among four members of the NMDAR2 subunits. This sequence homology is differently distributed according to the structural domains, and the homology is extremely high -N, potential N-glycosylation site; circled residues, possible phosphorylation sites. The conserved amino acid sequences at the carboxyl termini of the NMDAR2 subunits are underlined. In the NMDARZC sequence, the region that diverges between the sequence deduced from our cDNA clone and that reported by Monyer et al. (9) is indicated with a broken line; see the text for discussion of this sequence divergence.

NwJAR2c
+. Gs?*. F. w. G. m. Y- around the four transmembrane segments (80-90% homology), but moderate at the amino-terminal portions (45-60% homology) and very low at the carboxyl-terminal regions (20-30% homology). Interestingly, not only the sizes but also the amino acid sequences more closely resemble each other between NMDAR2A and R2B and between NMDAR2C and R2D than between the other pairs of these subunits, implying that four species of the NMDAR2 subunits can be classified into two subgroups. Under the assumed membrane topology, there are six, six, five, and six canonical Asn-X-Ser/Thr sequences for potential N-glycosylation sites (15) at the extracellular amino-terminal regions of NMDARZA, R2B, R2C, and R2D, respectively. Similarly, a large number of possible N-glycosylation sites are present at the extracellular carboxyl-terminal regions of NMDAR2A (12 sites) and R2B (10 sites), but only one and no such sites are observed at the carboxyl-terminal regions of NMDAR2C and R2D, respectively. At the carboxyl termini of these polypeptides, all but NMDARBD-1 share a common sequence, Ser/Pro-Ser-Leu/Ile-Glu-Ser-Glu/Asp-Val, which

Cloning and
Characterization of NMDA Receptor Subunits may play a certain role in the functions of these subunits. Another interesting feature is that the NMDAR2 subunits, like NMDARl, all possess many possible phosphorylation sites for Ca'+/calmodulin-dependent protein kinase type I1 and protein kinase C (16). These protein kinases have been reported to play a crucial role in the induction and maintenance of long term potentiation (17). The TM I1 segments of the ligand-gated ion channels are thought to line the channel pore and determine ionic conductance and ionic selectivity (14). Several structural characteristics of the TM I1 segment of NMDARl have been pointed out in our previous report (3). The TM I1 segment of NMDARl is flanked by a glutamic acid at the extracellular side and a stretch of these amino acids at the cytoplasmic side. These negatively charged amino acids are not present at either side of TM I1 segments of the NMDAR2 subunits. Instead, all subunits contain a positively charged lysine residue within the TM I1 segments. This segment of NMDARl comprises a threonine residue (residue 602) at the position where the threonine residue controls ion permeation of the nicotinic acetylcholine receptor (18). The threonine residue is conserved at the equivalent position of all four NMDAR2 subunits. NMDARl possesses an asparagine residue (residue 616) at the corresponding position where the glutamine/ arginine substitution determines the Ca2+ permeability and channel properties of the non-NMDA receptor (19)(20)(21)(22). This asparagine residue is present at the corresponding position of all NMDAR2 subunits, suggesting that an asparagine ring that controls the Ca2+ permeability and channel conductance is formed in a channel pore constituted by a heteromeric assembly of the NMDARl and NMDAR2 subunits.
Functional Properties of Heteromeric Assemblies of the NMDA Receptor-We examined current responses to application of 100 p~ NMDA and 10 PM glycine in Xenopus oocytes injected with in vitro synthesized mRNAs for individual NMDAR2 subunits under voltage clamp at -80 mV. However, none of these subunits exhibited any electrophysiological current. When NMDARl was co-expressed in Xenopus oocytes, NMDAR2A and R2C greatly potentiated the NMDA response (Fig. 2). The peak amplitudes of heteromeric assemblies of NMDARl/R2A and NMDARllR2C were 487 f 62 (mean f S.E., n = 28) and 517 f 90 nA ( n = 16), respectively, and were more than 10 times larger than that obtained with the homomeric assembly of NMDARl. Neither uninjected oocytes nor water-injected oocytes showed any responses to NMDA. No appreciable potentiation of NMDA response was observed by co-expression of NMDARl and NMDARPD, even though two independent cDNA clones for NMDAR2D, NMDARBD-1, and R2D-2 were examined in this analysis. The reason for this failure remains to be determined, but it is possible that an unusually GC-rich sequence around the translation initiation site of the NMDAR2D mRNA may hamper efficient translation of this mRNA in Xenopus oocytes. We thus focused on the heteromeric formation of the NMDARl/ R2A and NMDARl/RBC subunits and examined and compared the electrophysiological and pharmacological properties of these two heteromeric receptors. The amplitude of current response was found to vary, depending on the ratio of the NMDARl and NMDAR2 mRNAs injected. In all subsequent experiments, we injected the NMDARl and NMDAR2 mRNAs in a fixed ratio of l:lO, respectively.
When 100 p~ NMDA, together with 10 pM glycine, was applied in a M$+-free solution, both NMDARl/R2A and NMDARl/R2C heteromeric receptors evoked a response with a rapid initial spike followed by a steady-state current (Fig.  2, a and d). The initial spike evoked by NMDA in oocytes results from the activation of an oocyte Ca2+-dependent C1channel through the NMDA receptor-mediated Ca'+ entry (3). The spike current, as well as the steady-state current, was, in fact, enhanced in a Na+,K+-free solution supplemented with 10 mM CaZ+ (Ca2+-Ringer) (data not shown). Furthermore, the spike current almost disappeared in Ba2+-Ringer, where 2 mM Ca2+ in the standard solution was replaced with 2 mM Ba2+ (data not shown). These findings indicated that both heteromeric receptors are efficiently permeable to Ca2+. In both cases, the NMDA response was greatly reduced by the removal of glycine (Fig. 2, a and d). The application of 10 pM glycine alone elicited an electrophysiological response (Fig. 2, a and d). (+)-MK-801 is a selective antagonist that acts as an open channel blocker of the NMDA receptor (23). In both heteromeric receptors, the NMDA response was rapidly inactivated by the addition of 1 PM (+)-MK-801, together with 100 p~ NMDA and 10 p~ glycine, and this response was hardly recovered by the following application of the agonists (Fig. 2, b and e). Thus, the NMDARlIR2A and NMDARl/ R2C receptors show a strong sensitivity to the (+)-MK-801 channel blocker. When Mg2+ inhibition was examined under voltage clamp a t -80 mV, the NMDA response in the NMDARl/RBA receptor was almost completely abolished by 100 FM Mg'+ (Fig. 2c). In contrast, the NMDARl/R2C receptor was very resistant to Mg'+ inhibition (Fig. 2f). The differ-

Cloning and Characterization
of NMDA Receptor Subunits 2841 ence in the sensitivity to Mg2+ blockade was also examined by ramping voltage slowly from -100 to 40 mV in BaZ+-Ringer solution in the absence and presence of M$+ (100 pM or 1 mM). In both heteromeric formations, current-voltage curves were almost linear at negative potentials in the absence of M$+, but the currents were reduced by the addition of Mg2+ under hyperpolarized potentials (Fig. 3). Thus, these two heteromeric receptors are blocked by M P in a voltagedependent manner. However, the NMDARlIRBA receptor was more markedly inhibited than the NMDARlIR2C receptor by M e at the two different concentrations.
Further pharmacological properties of the NMDARlIRBA and NMDARl/RZC heteromeric receptors were investigated by determining dose-response curves of various agonists and antagonists. The serial application of agonists was found to cause desensitization of the oocyte Ca2+-dependent C1-channel. We thus used Ba2+-Ringer to minimize the effect of secondarily activated C1-conductance and determined amplitudes of steady-state currents elicited by serial application of agonists or antagonists under voltage clamp at -80 mV. Doseresponse analysis of L-glutamate and NMDA indicated that the effective concentrations for half-maximal response (EC50) of L-glutamate were 3.7 and 1.0 p~ for the NMDARl/RXA and NMDARlIRBC receptors, respectively, whereas those of NMDA were 57 and 32 p~ for the respective receptors (Fig.  4a). Thus, the effectiveness of these agonists is slightly different between these two heteromeric receptors. D-APV is a competitive antagonist for the NMDA receptor by acting at the NMDA-binding site (24), whereas 7-chlorokynurenate is a noncompetitive antagonist that inhibits glycine binding (25). The inhibitory effects of these two antagonists were analyzed by increasing concentrations of the respective compound together with 100 p~ NMDA and 10 ~L M glycine. Effective concentrations for half-maximal inhibition ( ICEo) of D-APV were determined to be 2.5 PM for the NMDARl/R2A and 13 PM for the NMDARl/RSC (Fig. 46). The IC60 values of 7-chlorokynurenate were 0.79 p~ for the NMDARlIRBA and 3.2 /IM for the NMDARl/RBC (Fig. 4c). Thus, both antagonists more efficiently act on the former receptor than on the latter.
The electrophysiological characterization described above indicated that the two heteromeric receptors are considerably different in their sensitivities to M$+ blockade. This difference was confirmed by dose-response analysis of MgZ+ inhibition for the two heteromeric receptors (Fig. 5a). This analysis indicated that the ICso values for Mg2+ were 10 ~L M for NMDARl/RBA and 130 p~ for NMDARl/R2C. Because the inhibition of (+)-MK-801 was recovered very slowly, it was difficult to determine the value of (+)-MK-801 by serial application of this compound. In contrast, the recovery of inhibition by a similar channel blocker, desipramine, occurred more rapidly, probably due to fast dissociation from its binding to the channel pore (26). We thus determined doseresponse curves of desipramine for the two heteromeric receptors. The IC50 values of desipramine were very similar between the two receptors ( Fig. 5b), and this was in contrast to the different sensitivity to M e blockade between the two receptors.
Expression Patterns of the NMDAR2 mRNAs-Northern blotting and in situ hybridization analyses were performed to examine the expressions of four NMDAR2 mRNAs. On Northern analysis, the sizes of the mRNAs for NMDARBA, R2B, R2C, and R2D were estimated to be about 12,15,6, and 7 kilonucleotides, respectively (Fig.  6). This analysis also showed that both mRNAs for NMDAR2A and R2B are abundant in the cerebral cortex and hippocampus. The NMDAR2C mRNA is prominently expressed in the cerebellum, whereas the NMDAR2D mRNA is expressed highly in the subcortical regions.
I n situ hybridization revealed overlapping but distinct distribution of each NMDAR2 mRNA in the central nervous system. The NMDARBA mRNA is widely expressed in many brain regions, and this expression is prominent in the cerebral cortex, hippocampus, internal granular layer of the olfactory bulb, anterior olfactory nuclei, olfactory tubercle, some of thalamic nuclei, inferior colliculus, pontine nuclei, inferior olivary nuclei, and cerebellar cortex (Fig. 7a). The NMDARBB Hi, hippocampus; Hy, hypothalamus; IC, inferior colliculus; IO, inferior olivary nuclei; OR, olfactory bulb; Pn, pontine nuclei; SC, superior colliculus; SN, substantia nigra; St, striatum; Th, thalamus; Tu, olfactory tubercle; Ve, vestibular nuclei; V P , ventral pallidum. mRNA showed a wide but more restricted distribution in its expression pattern. Its expression is prominent in most of the telencephalic and thalamic regions but low in the hypothalamus, cerebellum, and lower brain stem regions (Fig. 7b). The distribution of the NMDAR2C mRNA is more localized, and its expression is extremely high in the granular layer of the cerebellum. Moderate expression of this mRNA is observed in the glomerular and mitral cell layers of the main olfactory bulb, some of thalamic nuclei, pontine nuclei, and vestibular nuclei (Fig. 7c). The NMDARZD mRNA is mainly expressed in the diencephalic and lower brain stem regions (Fig. 7d). Strong signals were observed in the glomerular layer of the main olfactory bulb, ventral pallidum, most of the thalamic nuclei, hypothalamus, superior colliculus, substantia nigra, vestibular nuclei, pontine nuclei, and deep cerebellar nuclei. Weaker signals were found in the cerebral cortical regions and the granular layer of the cerebellum.

DISCUSSION
We report here the isolation of cDNA clones encoding four different NMDA receptor subunits through PCR-mediated DNA amplification followed by molecular screening of a rat brain cDNA library. These subunits, termed NMDAR2A-NMDAR2D, share a structural similarity with other ligandgated ion channels and possess four putative transmembrane segments (TM I-IV). They are only 15% identical with NMDARl but highly homologous (-50%) within this receptor subunit family. The peculiar structural feature of this receptor family is the presence of large extensions of the putative extracellular domains a t both amino-and carboxyl-terminal regions that may play a certain role in the diverse function and regulation of the NMDA receptor. NMDARl in a homomeric configuration exhibits all electrophysiological and pharmacological properties characteristic of the NMDA receptor (3). In contrast, the NMDARP subunits alone show no ability to respond to glutamate or NMDA. However, when co-expressed with NMDAR1, NMDAR2A and NMDARBC, as well as NMDARBB (9, lo), markedly potentiate the NMDARl activity and confer the functional variability in the electrophysiological and pharmacological properties. In this investigation, we have failed to indicate the ability of NMDARBD to potentiate the NMDARl activity. However, this receptor is highly homologous to NMDARPC throughout the protein sequence. It is thus conceivable that NMDAR2D could also have an activity that is similar to the other members of this subunit family under appropriate expression conditions. Thus, NMDARl serves as a key subunit necessary for the NMDA receptor-channel complex, and the individual NMDAR2 subunits produce the functional diversity by forming a heteromeric configuration with NMDAR1.
The detailed characterization of the heteromeric receptors of NMDARl/RZA and NMDARl/R2C has shown that these heteromeric receptors bear all of the basic properties characteristic of the NMDA receptor: Ca2+ permeability, glycine modulation, voltage-dependent M e blockade, and selective inhibition by competitive and noncompetitive antagonists and open channel blockers. However, different combinations of the NMDARP subunits in a heteromeric configuration with NMDARl confer functional variability in the affinity for agonists, the effectiveness of antagonists, and the sensitivity to blockade of M e . Some of these results are consistent with those reported by Monyer et dl. (9) and Kutsuwada et al. (10). It is feasible that the structural heterogeneity in the extracellular domains of the NMDARP subunits is responsible for governing different affinities of agonists and antagonists that act at a glutamate-binding site and a glycine-modulatory site. Among various properties of the heteromeric receptors examined, the sensitivity to M e blockade is notably different between the NMDARlIR2A and NMDARl/RBC receptors. Our recent mutational analysis of the NMDARl has indicated that an asparagine at the putative channel-forming TM I1 segment plays a critical role in determining the Ca2+ permeability and sensitivity to M e blockade (27). Furthermore, mutation of this asparagine by replacing it with glutamine or arginine reduced both M e and desipramine blockades to similar extents, indicating that a putative asparagine ring formed by the assembly of the TM I1 segments is critical in determining binding of both M e and an open channel blocker. However, all NMDAR2 subunits contain an asparagine residue a t equivalent positions and also show a high sequence conservation at the putative channel-forming T M I1 segments. Thus, it seems unlikely that the putative asparagine ring in the middle of a channel is responsible for differentiating the sensitivity of Mg2+ blockade. Instead, as pointed out by Ascher and Nowak (28), there may exist a negatively charged surface potential that controls the accumulation of divalent cations at an extracellular vestibule of the NMDA receptor-channel complex. It is thus tempting to speculate that there is a heterogeneity in negative surface charges associated with the channel vestibule between different NMDAR2 subunits and that once these cations reach the channel vestibule, permeation and blockade of these ions are determined by the asparagine ring at the channel pore formed by the heteromeric NMDA receptor subunits.
The NMDARl mRNA is expressed ubiquitously in almost all neuronal cells throughout the brain regions (3). In contrast, the mRNAs for different NMDAR2 subunits show overlapping, but differential expression patterns in the rat brain. For example, the NMDARBA mRNA is prominently expressed in the cerebral cortex and hippocampus, whereas the NMDARPB mRNA is distributed in the forebrain. The NMDARPC and NMDARBD mRNAs predominate in the cerebellum and in the diencephalic/lower brain stem regions, respectively. Thus, the diversity of the physiological and pharmacological properties of the NMDA receptors at different brain regions is generated by anatomical and functional differences of multiple NMDAR2 subunits. Such functional diversity of the NMDA receptors in different brain regions has been reported by ligand-binding experiments, as well as biochemical and electrophysiological studies (29-33). Further molecular dissection of the NMDA receptors will provide much insight into the complex mechanisms of glutamate neurotransmission, synaptic plasticity, and neurotoxicity, as well as the molecular mechanisms of ion permeation and functional modulation characteristic of the NMDA receptor.