Functional expression of a recombinant unitary glutamate receptor from Xenopus, which contains N-methyl-D-aspartate (NMDA) and non-NMDA receptor subunits.

A cDNA encoding a 100-kDa subunit (XenNR1) of the N-methyl-D-aspartate (NMDA) glutamate receptor type has been cloned from Xenopus central nervous system. When XenNR1 is coexpressed in a mammalian cell line with a recently cloned 51-kDa non-NMDA receptor subunit (XenU1), also from Xenopus, it forms a functional unitary receptor exhibiting the pharmacological properties characteristic of both NMDA and non-NMDA receptors. Firstly, XenU1 can replace NR2 subunits, in complementing XenNR1 to introduce the ligand binding properties of a complete NMDA receptor. Second, responses to both NMDA and non-NMDA receptor agonists and antagonists were obtained in patch-clamp recordings from the cotransfected cells, but no significant responses were recorded when the cells were singly transfected. Third, from solubilized cell membranes from the cotransfected cells, an antibody to the NR1 subunit coprecipitated the binding sites of the non-NMDA receptor subunit. The unitary glutamate receptor has a unique set of properties that denote intersubunit interaction, including a glycine requirement for the responses to non-NMDA as well as to NMDA receptor agonists and voltage-dependent block by Mg2+ of the non-NMDA agonist responses.

Few distinctions drawn in neuroscience are so universally accepted as that between the N-methyl-D-aspartate (NMDA) 1 and the non-NMDA classes of vertebrate ionotropic glutamate receptors (1)(2)(3). This differentiation is now fundamental in the analysis of brain excitatory pathways. It was given a molecular basis through the discovery that the protein sequences of NMDA receptor subunits are very different from those of non-NMDA receptors (4). However, this distinction was challenged recently when a protein purified from the central nervous sys-tem of the toad Xenopus laevis was found to exhibit both NMDA and non-NMDA receptor properties (5,6). The central nervous system of X. laevis is an exceptionally rich source (7) of both the AMPA and the kainate binding sites of non-NMDA receptors. These abundant sites and their unusual stability after detergent solubilization (7) permitted the purification on an affinity column containing immobilized domoate (a high affinity ligand for kainate receptors) of some of the ionotropic glutamate receptors. After reconstitution into bilayers two classes were found: (i) a non-NMDA (AMPA/kainate) receptor type, and (ii) a protein that exhibited both NMDA and non-NMDA binding and channel activation properties, i.e. it behaved as a unitary ionotropic glutamate receptor (5,6). Although the majority of the extracted NMDA receptors was not retained by that column and behaved as purely NMDA-type receptors, a significant minority of them was found to be of the unitary receptor type. These unexpected findings clearly required consolidation by the identification, through DNA cloning and expression, of the subunits constituting that novel type.
A cDNA (XenU1) encoding a non-NMDA receptor subunit has recently been cloned from Xenopus central nervous system using nine peptide sequences obtained from the purified (type (i) above) AMPA/kainate receptor (8). This 51-kDa protein corresponded in size and peptide sequence to the subunit isolated from the latter receptor. It has only 36 -40% sequence identity with rat non-NMDA receptors, despite the presence of four hydrophobic domains that have high sequence identity to four equivalent domains in the COOH-terminal half of mammalian non-NMDA receptor subunits (8,9). However, XenU1 lacks most of the large NH 2 -terminal domain that characterizes the latter proteins. Although similar in size to certain vertebrate kainate-binding proteins (10 -12), XenU1 differs from them in its binding of AMPA (K d ϭ 62 nM) as well as kainate (K d ϭ 9.1 nM) when expressed in mammalian cells; NMDA is not a competitor for these on XenU1 (8). Expression of those AMPA/ kainate binding sites was readily achieved by transfecting cells with XenU1 cDNA; however, extensive testing for functional expression of XenU1 in Xenopus oocytes gave no significant response, suggesting that a native partner subunit is needed (8). We identify here such a partner: this is a Xenopus NMDA receptor type of subunit, XenNR1, which we show reproduces, when (and only when) coexpressed with XenU1, the behavior of the above described native unitary glutamate receptor protein.

EXPERIMENTAL PROCEDURES
Methods and materials not specified were as in Refs. 6 and 8. L-Glutamate, glycine, and NMDA were from Sigma Chemical Co. (purest * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM  DNA Cloning for an Abundant Form of NR1 of Xenopus-Poly(A) ϩ RNA was isolated from Xenopus central nervous system, and cDNA was synthesized as described previously (8). A Xenopus brain ZAP-II cDNA library (2.5 million primary plaques) (8) was screened with an EcoRV fragment of the pN60 plasmid (4) containing the full coding region of rat NMDAR1A. The final washing of the filters was in 2 ϫ SSC, 0.5% SDS at 40°C. A positive clone was identified with a ϳ3-kilobase cDNA insert which was found to be homologous to rat NMDAR1A. The library was rescreened twice under high stringency conditions, using synthetic oligonucleotides derived from the sequences found (5Ј-TTCTGTATA-ATATAGCTGCCATTGAAGATTCCAACCT-3Ј and 5Ј-GTACAGTT-GGGCGACCTGGGATGGTCTCATTGGGACCATTACAAA-3Ј). The final washings were in 2 ϫ SSC, 0.5% SDS at 60°C. The full-length XenNR1 cDNA (4,060 base pairs) was constructed using the three longest overlapping clones taken from nine independent and fully consistent partial cDNA clones obtained from 0.5 million plaques. All clones and the junction sites were sequenced twice or more. To check for existence of the alternatively spliced forms of XenNR1 mRNA, two pairs of primers, synthesized such that potential splicing sites in XenNR1 would lie between them, were used to rescreen the entire cDNA library (2.5 million plaques) and also for reverse transcriptase PCR. The primers were 5Ј-ATGCCATCCAGATGGCTCTATCTGT-3Ј and 5Ј-CCTTG-GCCTCAAGGAGAAGGGCAGT-3Ј (for the NH 2 -terminal splicing site) and 5Ј-TCTTGATATTTATTGAAATCGCCTA-3Ј and 5Ј-GGCCAGT-CATTAGCCGTCAGTACAT-3Ј (for the COOH-terminal splicing sites). The cDNA library was screened with these as probes under high stringency conditions; the final washings were in 2 ϫ SSC, 0.5% SDS at 60°C. PCR amplification conditions were 1 cycle of 5 min at 94°C, 2 min at 50°C, 10 min at 72°C and 40 cycles of 1 min at 94°C, 1 min at 60°C, and 3 min at 72°C followed by 10 min at 72°C. The PCR products were analyzed by electrophoresis in 1% agarose and 5% polyacrylamide gels and subcloned into the SmaI site of the pBluescript II SK Ϫ phagemid (Invitrogen).
Immunoprecipitation-Protein G-Sepharose beads, Fast Flow grade (Sigma), were incubated with the mouse anti-NMDAR1 monoclonal antibody (Pharmingen) (30 pmol of antibody/25 l of Sepharose) for 2 h at room temperature (25°C) and washed thoroughly with Tris-buffered saline (pH 7.6) and then twice with the solubilizing buffer (defined below). Membranes were solubilized in 50 mM Tris citrate buffer containing 1 mM EDTA, 1% n-octyl-␤-D-glucopyranoside and protease inhibitors (6) for 1 h at 4°C followed by 1-h centrifugation at 100,000 ϫ g. The immobilized antibody was added to the solubilized samples followed by an overnight incubation at 4°C with gentle rotary agitation. The beads were collected by centrifugation and washed thoroughly with the solubilizing medium. The wash fractions were saved and combined with the rest of the supernatant to estimate the depletion of [ 3 H]kainate binding sites therein. The amount of XenU1 subunit precipitated onto the beads was measured by [ 3 H]kainate binding (performed as above) to the washed Sepharose pellet.
Immunoblotting-10% (w/v) SDS-polyacrylamide gel electrophoresis was performed as described previously (17). Samples were prepared by adding boiling SDS-loading buffer (17) directly to thoroughly washed Sepharose pellets or to chloroform/methanol-precipitated solubilized material and incubating at 100°C for 2 min. The separated samples were electroblotted (18) to BA-85 nitrocellulose membranes (Schleicher & Schuell) and stained with the anti-NMDAR1 antibody (1 g/ml) for 2 h at room temperature. Immunoreactive XenNR1 bands were visualized using horseradish peroxidase-conjugated goat anti-mouse antibody (Sigma; dilution 1:5,000) and the ECL Western blotting detection reagents (Amersham) according to the manufacturer's instructions.

RESULTS
A Lower Vertebrate NR1 Subunit-Poly(A) ϩ RNA was isolated from Xenopus central nervous system and used in constructing a cDNA library (8). This was screened at low stringency with the rat NR1 subunit (4) cDNA. From the resulting clones a full-length cDNA was isolated which encodes an NR1type subunit, XenNR1 (EMBL accession no. X94156). The translation start site is assigned to the first methionine residue of the largest open reading frame (Fig. 1). The sequence immediately following this is that of a signal peptide. The cleavage site was predicted (19) to be after the first 20 residues, equivalent to that for the rat NR1 subunit (4). The mature XenNR1 subunit has a predicted sequence ( Fig. 1) of 884 amino acids and a calculated molecular mass of 99,600 Da. It shares high amino acid sequence similarity with mammalian NR1 subunits, almost all of the differences being concentrated in the first 448 residues of the 559-residue presumptive NH 2 -terminal domain. Identified NR1 subunits of demonstrated receptor functionality have hitherto been available for comparison only from mammals; it is interesting to see that there is absolute conservation in the NR1 sequence, from amphibian to man, of the entire region containing the four hydrophobic domains and the long loop between the third and the fourth hydrophobic domains (apart from two conservative changes), as well as of an adjacent 32-residue stretch of the COOH-terminal tail. A 91residue stretch just prior to the assigned (4) first hydrophobic domain is also completely conserved; there is evidence that this segment contributes to the agonist binding site of glutamate receptors (20). The COOH terminus of XenNR1 also has the sequence motif (21) denoting interaction with the postsynaptic density protein PSD-95.
XenNR1 corresponds to the G isoform ( Fig. 1), which in rat brain is the least abundant of the eight alternatively-spliced NR1 isoforms (22)(23)(24)(25). That isoform is characterized by a 21amino acid insertion within the NH 2 -terminal sequence region and a double peptide deletion that produces a COOH-terminal sequence different from that in all but one (the E form) of the other NR1 isoforms. Additional extensive screening of the Xenopus cDNA library was performed with oligonucleotides (as detailed under "Experimental Procedures") which recognize the sequences around those potential splice junctional sites encoded in XenNR1 cDNA; this did not reveal any alternatively spliced forms of the XenNR1. The search was extended by analyzing the mRNA from Xenopus central nervous system using reverse transcriptase PCR. No product corresponding to alternative splicing in the COOH-terminal part of the XenNR1 was detected. However, amplification of the cDNA corresponding to a region around the potential alternative splicing site in the NH 2 -terminal part of XenNR1 generated two different products, which were separated, cloned, and sequenced. The nucleotide sequence of the major band exactly matched the previously found XenNR1 cDNA. The other, very minor band, was of a shorter product, with the same sequence but with the 63-nucleotide deletion, thus resembling the NH 2 -terminal region of the mammalian NR1E subunit (22,24). When these PCR products were subcloned (without a size separation), 2 out of 20 randomly chosen clones represented the E isoform and 16 the G isoform. The predominance of the G isoform is the oppo-site of what is found in the rat central nervous system (22,25).
Expression was studied using subunit cDNA transfections into the HEK-293 cell line. In that system, expressed mammalian NR1 subunits do not form functional NMDA receptors unless combined with an NR2 subunit (15, 26 -29). Likewise, XenNR1 expressed alone formed functional receptors with a very low probability (see below) and yielded very low binding of [ 3 H]MK-801 ( Fig. 2A, left). [ 3 H]DCKA, a competitive antagonist at the glycine site of NMDA receptors (27), also gave little binding ( Fig. 2A, right). That ligand, however, binds well (27,30) to the rat NR1A subunit expressed alone in HEK-293 cells; the difference here may be because the rat G isoform has not been tested thus or because of its other sequence variations FIG. 1. Amino acid sequence of the X. laevis XenNR1 subunit. The deduced amino acid sequence of the X. laevis XenNR1 subunit is aligned with those of the nearest known mammalian homologs: rat NR1G (22), human hNR1-E (14), and also mouse NMDA 1 (or NR1A) (9). Only the differences from the XenNR1 subunit are shown. The two-residue gap allows the maximum homology. Dots show the alternatively spliced regions where no sequence exists in the forms described. Four proposed hydrophobic domains (1)(2)(3)(4) are double underlined. Potential glycosylation sites (ࡗ) and potential phosphorylation sites (q) are shown. The point of deduced signal cleavage is marked by a vertical arrow.
from XenNR1 (Fig. 1). However, when XenNR1 was coexpressed with a mouse NR2A subunit, binding of both MK-801 and DCKA was increased greatly ( Fig. 2A). In electrophysiological experiments on HEK-293 cells and Xenopus oocytes coexpressing XenNR1/NR2A (mouse) or XenNR1/NR2B (mouse), NMDA plus glycine (but not kainate or AMPA, with or without glycine) now activated abundant receptors. 2 Hence XenNR1 is a subunit of a functional NMDA receptor.
Complementation of XenNR1 by a Non-NMDA Receptor Subunit-When XenNR1 was coexpressed in HEK-293 cells with XenU1, the binding of ligands selective for NMDA receptors rose dramatically ( Fig. 2A). In fact, the formation of hybrid receptors by the XenNR1/XenU1 combination was consistently more efficient than that by the XenNR1/NR2-A (mouse) combination. The binding of [ 3 H]DCKA was saturable; its affinities at receptors containing Xenopus or mammalian subunits are compared in Table I. When XenNR1 was coexpressed with XenU1, the binding of [ 3 H]glutamate was sensitive to NMDA, with EC 50 ϭ 550 nM (Fig. 2B, inset). This contrasts with either XenNR1 or XenU1 expressed alone, where NMDA-sensitive [ 3 H]glutamate binding is absent (data not shown), as has been reported for rat NR1A alone (30). 38% of the binding of [ 3 H]glutamate (100 nM) was now due to NMDA-reactive receptors, and the rest was due to the kainate-reactive sites (Fig. 2B). A fraction of the kainate sites, however, may be attributable to the excess of XenU1 employed: the ratio of XenNR1 cDNA to XenU1 cDNA for optimal hybrid expression (assessed by [ 3 H]DCKA or [ 3 H]MK-801 bindings, data not shown) was found to be ϳ1:4, and this was routinely used. In further such experiments the ratio was reduced to 1:1.5 (XenNR1/XenU1 cDNAs), when total expression is suboptimal but the expression of hybrid receptors is presumably dominant. In those conditions the binding of [ 3 H]kainate to the hybrid receptor was still abundant and of affinity as high as that measured when XenU1 is expressed alone (8). The affinity for [ 3 H]kainate of the hybrid receptor was unchanged in the presence of 20 M glycine (Table I).
Specificity for the Subunit Types in the Hybrid-Weak expression of a hybrid receptor resulted from cotransfection of XenU1 cDNA with each of two (rat and human) NR1A cDNAs. This conclusion is based upon the presence of a small amount of [ 3 H]MK-801 binding ( Fig. 2A, left). Expression was improved significantly when XenU1 cDNA was cotransfected with cDNA for the human NR1E subunit (14), but it was still less efficient than with XenNR1 ( Fig. 2A, left). In contrast, NR1 subunit expression in 293 cells is not affected by other known functional non-NMDA receptor subunits. Thus, when either XenNR1 or human (14)  Testing for Functional Expression of XenNR1 or XenU1 Alone-Receptor function was studied in whole cell current recordings obtained from transfected HEK-293 cells. Transfection with XenNR1 alone rarely led to the expression of functional homo-oligomeric receptors, i.e. only 1 of 27 cells tested responded to application of 100 M NMDA (plus 10 M glycine) and this with a consistently small (ϳ20 pA) inward current at V H ϭ Ϫ50 mV. Rat NR1 subunits, likewise, give little or no functional expression alone in HEK-293 cells (26,27). Cells transfected with XenU1 alone also showed little functional expression; i.e. in response to 100 M kainate, AMPA, or glutamate only 2 out of a total of 89 cells tested gave small currents, these being 20 pA or less at V H ϭ Ϫ50 mV. Addition of glycine (10 M) did not increase the amplitude of these currents. Treatment with cyclothiazide, wheat germ agglutinin, or concanavalin A, which can suppress the desensitization  of mammalian AMPA and kainate receptors (31,32), neither increased the amplitude of these responses nor increased the proportion of responding cells. Functional Expression of XenNR1/XenU1 in Heteromeric Combination-Consistent responses to NMDA and non-NMDA agonists were obtained when XenNR1 and XenU1 were cotransfected into HEK-293 cells. Inward currents of up to ϳ1,000 pA at V H ϭ Ϫ50 mV were elicited by 100 M NMDA. Inward currents of up to 800 pA at V H ϭ Ϫ50 mV were also obtained from these cells in response to kainate or AMPA (see Fig. 3, A and B, for typical responses). The response was at a maximum at 100 M agonist, but it was elicited down to 0.3 M AMPA or less (Fig. 3C). The cells also responded to 100 M glutamate, with inward currents at V H ϭ Ϫ50 mV which sometimes exceeded 1,000 pA (see Fig. 3A for a typical response). The responses were seen in ϳ30% of cells, which corresponds to the transfection efficiency, as estimated in parallel transfections made with a reporter plasmid expressing ␤-galactosidase. Whenever tested, a cell responding to NMDA also responded to any of the non-NMDA agonists. The responses to kainate, but not those to AMPA, declined progressively in amplitude with repeated application of agonist at 30-s intervals (Fig. 3, D and  E). This change could not be reversed by extensive washing of the preparation with agonist-free saline. This kainate-specific "run-down" was unaffected by coapplication of cyclothiazide or pretreatment of cells with lectins (31,32). This run-down blocked a subsequent response to AMPA. AMPA and not kainase was used in further studies.
A Glycine Requirement for Activation by non-NMDA Agonists-Surprisingly, the responses of the heteromeric combination to NMDA and to non-NMDA agonists (Figs. 3B and 4A) were glycine-dependent, although glycine alone, even as high as 1 mM, never evoked a response. Dose-response relationships for this potentiation of the responses to glutamate, NMDA, and AMPA showed a high glycine sensitivity (Fig. 4A), different for the NMDA and the AMPA agonist sites. DCKA, an antagonist of the mammalian NMDA receptor glycine site, completely antagonized (at 10 M) the potentiation by glycine of the responses to non-NMDA agonists (Fig. 4B).
Further Functional Properties of the Heteromeric Receptor-The dose-response curve for NMDA (plus 10 M glycine) had an EC 50 of 74 Ϯ 16 M (Fig. 4C). If the non-NMDA agonists were acting at the same receptor where NMDA acts, the NMDA response should be less than additive with a response to coapplied AMPA. Indeed, even at the maximum response to NMDA the addition of AMPA produced no extra current but in fact decreased the NMDA response (Fig. 4D). The decrease observed could in theory be due to one of several causes, but to offer a detailed interpretation of it would not be justified here until we have data on the single channel conductances evoked by NMDA and by AMPA for the unusual case of the channel in this hybrid assembly.
Responses to NMDA (plus glycine) showed only slight desensitization and that was not Ca 2ϩ -sensitive, although the shape of the current-voltage relationship for this agonist was Ca 2ϩsensitive. When the Ca 2ϩ concentration was raised from nominally 0 mM to 1 mM and then to 10 mM, the amplitude of the NMDA-induced current was reduced progressively at highly negative V H (Fig. 5A). This change was accompanied by a shift of the current reversal potential to more positive values. These features indicate that the Ca 2ϩ permeability of the XenNR1/ XenU1 heteromeric channel is similar to that of the mamma- The NMDA-induced currents were (as expected) antagonized noncompetitively and voltage-dependently by Mg 2ϩ (100 M to 1 mM) (Fig. 5B) and by MK-801 (Fig. 5C). As in mammalian NMDA receptors, the NMDA site ligand D,L-2-amino-5-phosphonopentanoic acid competitively antagonized the NMDA responses (data not shown). Interestingly, spermine, even at 1 mM, did not potentiate the responses to NMDA (plus glycine) (data not shown). In this respect the hybrid receptor behaves as the mammalian NMDA receptor when it has, as here, the NR1 subunit with the alternatively spliced NH 2 -terminal insert (23).
In addition to its sensitivity to glycine, the responses of cotransfected cells to AMPA revealed other characteristics normally associated with NMDA receptors. They were antagonized noncompetitively by Mg 2ϩ , but only at negative potentials (Fig. 5D). Further, the form of the current-voltage plot in that condition is just as found for the NMDA response (comparing the 1 mM Mg 2ϩ curves of Fig. 5, B and D). This form reproduces exactly that which is characteristic for the rat NR1/ NR2 receptors (with NMDA) in HEK-293 cells (26). In contrast, the responses to 100 M AMPA (plus glycine), but not to NMDA (plus glycine), were completely blocked by 6,7-dinitroquinoxaline-2,3-dione (50 M), a competitive antagonist (33) of mammalian AMPA receptors (data not shown). This confirms that the exceptional properties of the response to AMPA are due to its action at the AMPA site itself, which is present (8) on XenU1.
Immunoprecipitation-An ability of an antibody specific to one subunit of a receptor to precipitate also a second subunit or its binding sites has recently become a powerful tool for identifying pairs of subunit types coexisting in ionotropic receptor molecules. In NMDA receptors it has been applied, for example, to show that in rat cortex extracts NR1 can be variously associated with NR2A or with NR2B subunits or with both (35). To precipitate non-NMDA-specific binding sites we have used a commercially available monoclonal antibody against a rat amino acid sequence that is completely conserved in Xenopus, located in the loop between the third and the fourth hydrophobic domains of NR1. This antibody recognizes in Western blots a band (ϳ110 kDa) derived from membranes from the Xenopus central nervous system or from cells transfected with XenNR1 cDNA, but not from cells transfected with XenU1 cDNA (Fig. 6, lanes A-C). Glutamate receptors from cotransfected 293 cells or from Xenopus central nervous system were extracted in n-octyl-␤-D-glucopyranoside medium (6) and equilibrated with the anti-NR1 antibody immobilized on Sepharose beads (Table II). This treatment precipitated a majority of the solubilized XenNR1 subunits (Fig. 6, lanes D and E). It also precipitated a significant number of the kainate binding sites (Table II). An extract of 293 cells transfected with XenU1 cDNA alone showed no depletion by the anti-NR1 antibody, and none of the kainate binding sites there became attached to the beads. The crossimmunoprecipitation occurred both from the extract containing heteromeric recombinant receptors and from the native tissue extract (Table II). DISCUSSION Properties of XenNR1-Among the NR1 subunits whose amino acid sequences are known in the mammals, there is extremely high conservation, e.g. 99% between rat and man (14). In view of the much greater phylogenetic distance from these to the amphibian Xenopus, it is interesting to see that the entire COOH-terminal half of the subunit is almost completely identical to the mammalian NR1G sequence, whereas in the first 450 amino acids of the NH 2 -terminal extracellular domain 13% are changed (Fig. 1). Although XenNR1 produces functional receptors when it is expressed in combination with rodent NR2 subunits, a fuller electrophysiological study is required to determine whether the sequence differences between the Xenopus and rodent NMDA receptor subunits are physiologically and/or pharmacologically significant.
Most of the NMDA receptors that can be extracted from Xenopus central nervous system do not contain the XenU1 subunit because about 70% of the total NMDA receptor binding sites present in an extract of this tissue could not be retained by the domoate affinity column (6). In fact the true proportion of such conventional NMDA receptors is likely to be higher than 70% since the initial n-octyl-␤-D-glucopyranoside extraction of all of the AMPA/kainate sites failed to solubilize all the NMDA receptor binding sites (7). Functional expression of both the conventional and the unitary NMDA receptor activities was seen previously in oocytes injected with poly(A) ϩ RNA from Xenopus brain, and this gave a similar proportion (5). We presume that this majority fraction of conventional NMDA receptors in Xenopus contains NR1/NR2 hetero-oligomers as in mammals; Western blotting tests on Xenopus brain membrane extracts with an antibody raised to a peptide sequence conserved in all NR2 subunits shows the presence of the latter (data not shown), but they have not yet been cloned. It seems reasonable to conclude that XenNR1 is a component of a major fraction of conventional NMDA receptors in Xenopus central nervous system and also a component of a smaller fraction there of unitary NMDA/non-NMDA receptors.
XenU1 as a Partner Subunit to XenNR1-In several respects, XenU1 is an exceptional non-NMDA receptor subunit. It is only about one-half the size of the mammalian non-NMDA receptors subunits (GluR1-7 and KA1-2), yet it has relatively low sequence similarity to the short kainate-binding proteins (10 -12) of non-mammalian vertebrates; unlike them, it also binds AMPA with high affinity (8). For the following reasons we conclude that XenU1 forms functional hetero-oligomeric receptors in Xenopus central nervous system by combining with XenNR1. (i) Functional receptor activity is introduced by the coexpression of these two subunits in HEK-293 cells and is strong and robust. (ii) The receptors formed by XenNR1 plus XenU1 acquire specific binding sites and agonist or antagonist selectivity properties that are characteristic of both NMDA and non-NMDA receptors. (iii) The partial complementation of human NR1E indicates that XenU1 can combine even with a mammalian NR1 subunit in stabilizing an NMDA receptor assembly. In contrast, NR1 subunits in HEK-293 cells did not interact with previously known non-NMDA receptor subunit types, when coexpressed. (iv) If XenNR1 and XenU1 expressed only independent homomeric receptors after their cotransfection, then additive currents would be anticipated during coapplication of NMDA and non-NMDA receptor agonists. In fact, for the current seen when NMDA is present, not only is there no increase with AMPA, but a decrease is produced. (v) Interaction between the two subunits comprising the XenNR1/ XenU1 receptor is evident. For example, a voltage-dependent Mg 2ϩ block occurs when the receptor is activated by non-NMDA as well as by NMDA ligands. (vi) A unique feature of the XenNR1/XenU1 receptor is the glycine requirement for responses to AMPA and kainate. This is clearly dependent on the presence of the XenNR1 subunit, since it is DCKA-sensitive and since glycine failed to enhance the responsiveness to non-NMDA agonists in the few cells expressing homo-oligomeric XenU1 receptors. An allosteric effect of glycine on the open time of the channel gated by non-NMDA agonists appears to be involved, since in the binding studies the affinity of kainate was not increased by the glycine. (vii) The XenNR1 and the XenU1 subunits can be coprecipitated from extracts of cotransfected cells and from Xenopus central nervous system by an anti-NR1 antibody. A considerable amount of receptor (Ͼ2 pmol/mg initial protein) was cross-precipitated in the latter case. This, as with features (i) to (vi), is only explicable in terms of a unitary receptor. It should be noted that the combined evidence here for the hybrid is equivalent to the similar evidence from which NR1/NR2 combinations have been accepted (2,3,23,35).
In addition to that cross-immunoprecipitation of the native receptor, further evidence for the in vivo occurrence of the unitary receptor is that, as noted in the Introduction, a native protein purified from Xenopus central nervous system and reconstituted to give glutamate-activated channels showed the same unitary behavior and interaction between NMDA and non-NMDA receptor sites (6). That protein contains the XenU1 subunit (8) and a subunit of XenNR1 size (6).
Conclusions-We conclude that XenNR1 and XenU1 form a unitary NMDA/non-NMDA receptor in vivo. Are unitary glutamate receptors peculiar to such amphibians? Evidence of direct interaction between native NMDA and non-NMDA subunits in other species, including mammals, has rarely been sought, but functional indications of this have in fact been detected in native preparations (31,36,37). The data presented herein suggest that a wider search is now merited.