Subunit assembly of N-methyl-d-aspartate receptors analyzed by fluorescence resonance energy transfer.

N-methyl-d-aspartate (NMDA) receptors play major roles in synaptic transmission and plasticity, as well as excitotoxicity. NMDA receptors are thought to be tetrameric complexes mainly composed of NMDA receptor (NR)1 and NR2 subunits. The NR1 subunits are required for the formation of functional NMDA receptor channels, whereas the NR2 subunits modify channel properties. Biochemical and functional studies indicate that subunits making up NMDA receptors are organized into a dimer of dimers, and the N termini of the subunits are major determinants for receptor assembling. Here we used a biophysical approach, fluorescence resonance energy transfer, to analyze the assembly of intact, functional NMDA receptors in living cells. The results showed that NR1, NR2A, and NR2B subunits could form homodimers when they were expressed alone in HEK293 cells. Subunit homodimers were also found existing in heteromeric NMDA receptors formed between NR1 and NR2 subunits. These findings are consistent with functional NMDA receptors being arranged as a dimer of dimers. In addition, our data indicated that the conformation of NR1 subunit homodimers was affected by the partner NR2 subunits during the formation of heteromeric receptor complexes, which might underlie the mechanism by which NR2 subunits modify NMDA receptor function.

(NR2A-D) encoded by different but closely related genes (reviewed in Ref. 2). All of the subunits have a similar membrane topological structure, with three transmembrane domains plus a loop region, an extracellular N terminus, and an intracellular C terminus. Recently, NMDA receptors were thought to be tetrameric channels, for which the relevant evidence comes from the electrophysiological studies of NMDA and AMPA receptors (4 -6) and the evolutionary link between glutamate receptors and tetrameric K ϩ channels (7,8). The NR1 subunit contains a glycine-binding site (9,10) and is essential for the formation of functional NMDA receptor channels, whereas the NR2 subunit provides a glutamate-binding site (11,12) and modifies channel properties, such as current kinetics and channel conductance (13)(14)(15). The NR3 subunit does not form a functional NMDA receptor alone, but it can co-assemble with NR1/NR2 complexes to modulate the activity of the NMDA receptor (16 -18).
Previous studies have shown that NMDA receptor complexes may contain different NR1 splice variants (19,20) or different NR2 subunits (21)(22)(23)(24). Several distinct NMDA receptor subtypes have been identified in the central nervous system, differing in channel conductance, kinetic properties, sensitivity to pharmacological agents, and synaptic localization (2,25). Despite the increase in our understanding of the functional diversity of NMDA receptors, the cellular processes, mechanism of subunit assembly, and the structure-function relationships of the receptors are still far from being well understood. Biochemical studies indicate that the NR1-1a subunit expressed alone can form homodimers, and the first 380 amino acids in the N terminus are important for the association of NR1-1a with NR2A (26). In addition, cysteine 79 may be involved in the formation of NR1-NR1 homodimers (27). Data from NMDA receptor tandems suggest that the subunits making up the channel are arranged as a dimer of dimers with an NR1-NR1-NR2-NR2 orientation (28). Studies on AMPA receptors provide a two-step assembly model in which family-specific subunit dimerizations are initially formed and then these dimers are dimerized again to form tetramers (29 -31).
In this study, we used a biophysical approach based on fluorescence resonance energy transfer (FRET) to analyze the assembly of intact NMDA receptors directly. In recent years, FRET has been used widely to examine the stoichiometry, assembly, and conformational arrangement of many kinds of receptors and channels (32)(33)(34)(35)(36). FRET is a sensitive reporter of the proximity of two fluorophores (37,38), providing a convenient way to report subunit assembly. More importantly, FRET provides a non-invasive means of monitoring the subunit assembly of functional proteins in living cells. Furthermore, dynamic subunit assembly information can also be derived from FRET data. Our results suggest that the similar subunits are inclined to assemble into homomers before they form functional NMDA receptors with the other types of subunits. These find-ings indicate that dimerization may be a general principle guiding the assembly of NMDA receptors, as well as AMPA receptors (29 -31), K ϩ channels (39), and cyclic nucleotidegated channels (40).

MATERIALS AND METHODS
Plasmids-The expression vector for cyan fluorescence protein (CFP)-or yellow fluorescence protein (YFP)-tagged NR1-1a has been constructed by inserting a CFP or YFP cDNA fragment in-frame with the NR1-1a subunit between the third and fourth codons after the predicted sequence for signal peptide. CFP-or YFP-tagged NR2A and NR2B at the N termini were constructed by replacing the GFP sequence of pGFP-NR2A and pGFP-NR2B (41) with the CFP or YFP fragment of pECFP-N1 or pEYFP-N1 (Clontech). YFP-tagged GluR1 was generated by inserting the YFP cDNA fragment in-frame with the GluR1 subunit between the third and fourth codons. cDNA encoding CFP or YFP was amplified by polymerase chain reaction from plasmid pECFP-N1 or pEYFP-N1 (Clontech, Palo Alto, CA). GABA A ␣1 subunit with a YFP tag at the C terminus (␣1-YFP) was generated by replacing the GFP sequence of pGFP-GABA A ␣1 (provided by Dr. S. Vicini, Georgetown University School of Medicine, Washington, D. C.) with the YFP fragment of pEYFP-N1 (Clontech) between the HindIII and KpnI sites. The vector pECFP-YFP coding for the CFP-YFP fusion protein was generated by inserting the YFP cDNA obtained from pEYFP-N1 by PCR using the primers 5Ј-ggagCTCGAGATGGTGAGCAAGGGCGAGGA-3Ј and 5Ј-gCggGTCGACGCTCCCTTGTACAGCTCGTCCATGCG-3Ј into the SalI and XhoI sites of pECFP-N1 (Clontech), resulting in a GASTV-PRARDPPVAT spacer between the fluorescence proteins ( Fig. 1). All plasmids were sequenced to ensure correct reading frame, orientation, and sequence. The spectra of CFP and YFP fusion proteins were measured using a Fluoview FV1000 confocal microscope (Olympus, Tokyo, Japan). The spectrum between 480 -600 nm was recorded with a wavelength increment of 2 nm and an excitation wavelength of 405 nm.
HEK293 Cell Culture and Transfection-Human embryonic kidney (HEK)293 cells were grown in minimal essential medium, supplemented with 10% fetal bovine serum, antibiotics, and glutamine (all from Invitrogen) in a 5% CO 2 incubator. Exponentially growing cells were dispersed with trypsin (Sigma), seeded at 2 ϫ 10 5 cells/35 mm dish in 1.5 ml of culture medium, and plated on 12-mm glass coverslips (Fisher Scientific, Pittsburgh, PA). HEK293 cells were transfected with appropriate plasmids (ϳ3-4 g/35-mm dish) using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. The transfection mixture was replaced ϳ3-5 h later with fresh culture medium. Ketamine (0.5 mM, Sigma) and kynurenic acid (1 mM, Sigma) were added to the medium to protect the cells from NMDA receptor-mediated toxicity. The cells were observed within 2 days after transfection in extracellular medium composed of (in mM): 145 NaCl, 5 KCl, 2 CaCl 2 , 5 glucose, 0.01 glycine, and 5 HEPES at pH 7.4 with NaOH. For acceptor photobleaching measurement, HEK293 cells were fixed in 4% paraformaldehyde for 20 min before imaging. Surface immunostaining and electrophysiology were carried out as previously described (41) to determine surface expression and functional properties of recombinant NMDA receptors in transfected HEK293 cells.
Immunocytochemistry -Cultured HEK293 cells were fixed in methanol for 10 min at Ϫ20°C and permeabilized in phosphate-buffered saline (PBS) containing 0.4% Triton X-100 and 5% bovine serum albumin for 30 min at room temperature. The cells were then incubated in primary anti-ER monoclonal antibody (Sigma) in PBS containing 5% bovine serum albumin for 1 h. After washing with PBS three times, the cells were incubated with Alexa-546-conjugated secondary antibody (Molecular Probes, Eugene, OR) in PBS containing 5% bovine serum albumin for 1 h. After washing with PBS three times, the cells were observed using a Fluoview FV1000 confocal microscope (Olympus). The primary antibody was used at 1:200, whereas the secondary antibody was used at 1:2000.
Preparation of Hippocampal Neurons and Transfection-Hippocampal neuronal cultures were prepared by previously described techniques (41). Briefly, hippocampal tissue was harvested from one-day-old Sprague-Dawley rats, gently chopped, and digested in 0.28% trypsin for 15 min at 37°C. Dissociated cells were plated at a density of 2 ϫ 10 6 in a 35-mm dish with poly-L-lysine-coated coverslips in basal Eagle's medium (Invitrogen) containing 10% fetal bovine serum, 2 mM glutamine, 100 g/ml gentamycin, and 25 mM KCl and maintained at 37°C in 5% CO 2 . After 24 h in vitro, the medium was replaced with a half-and-half mixture of basal Eagle's medium and neurobasal medium containing 2% B27 supplement (Invitrogen), 1% antibiotic, and 0.25% glutamine. Subsequently, the culture medium was replaced every 3 days. At 4 days in vitro, cytosine arabinofuranoside was added at a final concentration of 10 M. Lipofectamine 2000 was used in neuronal transfection, but the method was slightly modified. Appropriate plasmids (ϳ3-4 g/35-mm dish) with 4 l of Lipofectamine 2000 was added, and after incubation for 3 h at 37°C, the cells were washed twice with serum-free medium and the original culture medium was returned. Expression of exogenous DNA was typically detected in 8 -10-day in vitro neurons.
Detection of FRET Using Three-cube FRET Measurement-The fluorescence imaging workstation consisted of a TE2000 inverted microscope (Nikon, Tokyo, Japan) equipped with a halogen lamp light source (100 W), Dual-View TM (Optical Insights, LLC, Santa Fe, NM) and a SNAP-HQ-cooled charge-coupled device camera (Roper Scientific, Trenton, NJ). MetaMorph version 5.0 software (Universal Imaging, West Chester, PA) was used to control the charge-coupled device camera and for analysis of the cell image data. Three-cube FRET filter cubes were listed as follows (excitation; dichroic; emission; company): YFP (S500/20 nm; Q515lp; S535/30 nm, Chroma); FRET (S430/25 nm; 455dclp; S535/30 nm, Chroma); and CFP (S430/25 nm; 455dclp; S470/30 nm, Chroma). Binning 2 ϫ 2 modes, 200 ms of integration time for HEK293 cells, and 500 ms of integration time for hippocampal neurons were used. Average background signal was determined as the mean fluorescence intensity from a blank area and was subtracted from the raw images before carrying out FRET calculations. FRET ratio (FR) was calculated according to Erickson et al. as Equation 1 (42), in which F AD represents the total YFP emission with 430/25-nm excitation, and F A represents the direct YFP emission with 500/20-nm excitation. In S CUBE (SPECIMEN), CUBE indicates the filter cube (CFP, YFP, or FRET), and SPECIMEN indicates whether the cell is express- are predetermined constants that require measurement of the bleedthrough of the emission of only CFP-or YFP-tagged molecules into the FRET channel and the emission of only CFP-tagged molecules into the YFP channel. In our system, exciting CFP in cells that only expressed CFP showed no fluorescence signal with the YFP filter set and vice versa (Fig. 2). Thus R D2 was zero, and FR could be calculated as Equation 2.
Detection of FRET Using Acceptor Photobleaching-Fluorescence signals were collected with a 510 LSM confocal microscope (Zeiss, Thornwood, NY) using laser excitation at 458 nm (CFP) and 514 nm (YFP) and emission windows of 470 -500 nm (CFP; BP470 -500; Chroma, Brattleboro, VT) and a 570-nm-long pass filter (YFP; LP530; Chroma). Preceding the emission filters in the light path was a NFT 515-nm dichroic mirror that split the CFP and YFP signals to distinct photomultiplier tubes. All fluorescence data were collected with a 40 ϫ 0.75 numerical aperture objective (Zeiss) and 2ϫ zoom. Photobleaching in selective regions was performed at 514 nm with a mean reduction of YFP emissions to Ͻ15% using the software-controlled 510 LSM confocal microscope. Images were acquired before and after photobleaching. Effective FRET efficiency (E EFF ) between CFP (donor) and YFP (acceptor) was quantified with acceptor photobleaching methods (43) using the following Equation 3, in which A 0 and A 1 are CFP emission with 458 nm excitation in the photobleached region before and after photobleaching with 514 nm excitation, respectively. Estimation of Physical Distance and Binding Affinity-This method was previously described by Erickson et al. (42,44). Briefly, the fraction of YFP-tagged molecules that are associated with CFP-tagged molecules, A b , is calculated as where K d is the dissociation constant and [D free ] is the concentration of free donor molecules. Note that FR ϭ 1 ϩ A b ⅐ (FR max Ϫ 1), in which FR max is the maximal FR value. When ⌬FR max ϭ (FR max Ϫ 1), we can use regression analysis to estimate A b in individual cells. From each cell, the FRET ratio (FR exp ) was experimentally determined. We then computed the predicted A b value by adjusting two parameters, ⌬FR max and K d,EFF (effective disso-ciation constant). A b was, in turn, used to give a predicted FR. By minimizing the sum of the squared errors (FR exp Ϫ FR predicted ) 2 using MATLAB software, the two free parameters, ⌬FR max and K d,EFF , can be determined; and FR max can be determined when A b is 1, which, in turn, is used to estimate the distance between donor and acceptor fluorophores.
Statistics-All data were presented as the mean Ϯ S.E. Differences were tested using the Student's t test and considered significant at the p Ͻ 0.05 level.

RESULTS
FRET Reports Co-assembly of Different Subunits within the NMDA Receptor-CFP or YFP was inserted into the N terminus of the full-length NMDA receptor (NR) subunit to form CFP/YFP-NR1-1a, CFP/YFP-NR2A, and CFP/YFP-NR2B (see "Materials and Methods") ( Fig. 1). Spectral profiles of CFP-and YFP-tagged NR subunits were mostly identical to the reported spectra for the two fluorescent proteins. In addition, electrophysiological studies indicated that CFP-or YFP-tagged NR subunits were able to form functional channels when co-expressing with their partner subunits (data not shown). To measure steady-state FRET, cells transfected with different fusion proteins were imaged using an epifluorescence microscope through CFP, YFP, and FRET filter channels (Fig. 2). The FRET efficiency was quantified as the FR (see "Materials and Methods") (42,44). To ensure that our recording system could reliably detect FRET, we first carried out some control experiments. As shown in Fig. 3A, cells co-expressing CFP and YFP, which served as a negative control, showed no FRET with FR ϭ 1.08 Ϯ 0.05 (n ϭ 52). On the other hand, cells co-expressing the CFP-YFP concatemer, a positive control for FRET, showed significant increases in FR (FR ϭ 5.05 Ϯ 0.03, n ϭ 71) (Fig. 3A).
We next tested the interaction between NR1 and NR2 subunits by FRET. Immunostaining of living cells with anti-GFP antibody showed that fluorophore-fused NR1 and NR2 subunits formed receptors and targeted to the cell membrane, which further implies that they can be assembled as functional receptors. We found that co-expression of CFP-NR1-1a with YFP-NR2B or CFP-NR1-1a with YFP-NR2A produced significant FRET signals, with FR ϭ 3.48 Ϯ 0.20 (n ϭ 30) or 2.17 Ϯ 0.17 (n ϭ 21), respectively (Fig. 3A). Similarly, when YFP-NR1-1a was co-expressed with either CFP-NR2A or CFP-NR2B, significant FRET signals were also observed (data not shown). The result suggested that fluorescent protein-tagged NR subunits are capable of co-assembly just as the untagged subunits.
Could the apparent FRET signal come from unordered aggregates in compartments or random encounters between freely diffusing overexpressed channel subunits? To rule out these possibilities, we first co-expressed YFP-GluR1 with either CFP-NR1-1a or CFP-NR2A or CFP-NR2B. NR subunits expressed alone in HEK293 cells do not reach the cell membrane, and for AMPA receptor subunit GluR1, the majority of the homomeric assemblies are preferentially found intracellularly (45). Also, it is generally accepted that NMDA receptors do not assemble with AMPA receptors (2). As illustrated in Fig.  4, the transfected HEK293 cells with GluR1 or NR1 were fixed and immunostained by a specific antibody against ER, and it was clearly shown by confocal microscopy that GluR1 and NR1 were co-localized well with the ER marker, respectively. Despite clear co-localization, co-expression of various CFP-tagged NR subunits and YFP-GluR1 did not yield any FRET. The FR values were all close to 1. As a further negative control, we co-expressed NR subunits with GABA A ␣1 subunits, both of which are incapable of leaving the ER when expressed alone (46). As expected, co-expressing of CFP-tagged NR subunits with YFP-tagged ␣1 subunits yielded no FRET signal (FR ϭ 1.01 Ϯ 0.02, n ϭ 32). We conclude from these results that the FRET signal observed between NR1 and NR2 subunits arose from their close proximity within properly formed heteromeric NMDA receptors, rather than from occasional proximity by random encounters between singular channel subunits or from bulk protein aggregates.
Acceptor photobleaching is another simple and practical alternative to the three-cube FRET measurement. YFP was selectively bleached using a confocal microscope. Enhanced CFP emission due to disruption of FRET was measured. Using this method, we also found significant FRET between fluorophoretagged NR1 and NR2 subunits but not CFP-tagged NR subunits and YFP-tagged GluR1 subunits (Fig. 3B). The finding that E EFF (calculated from acceptor photobleaching) correlated well with FR (R 2 ϭ 0.94) for each combination provides additional evidence that NR subunits can form heteromers with one another (Fig. 3C).
FRET Measurement Reveals Homodimer Formation by NR1, NR2A, and NR2B Subunits-We co-expressed CFP-NR1-1a with YFP-NR1-1a in HEK293 cells. Surface immunostaining showed that the expression of NR1 subunits alone did not result in significant cell surface expression (data not shown), but the cells showed significant FRET signals with FR values of 2.40 Ϯ 0.10 (n ϭ 66) (Fig. 5). Similarly, when co-expressing CFP-NR2A with YFP-NR2A or CFP-NR2B with YFP-NR2B, we also observed FRET signals. The FR values for NR2A and NR2B subunits were 1.58 Ϯ 0.04 (n ϭ 36), and 1.92 Ϯ 0.08 (n ϭ 34), respectively (Fig. 5). The existence of FRET strongly suggests that NR1, NR2A, and NR2B subunits can form dimers or higher order multimers. Our results thus are consistent with NR subunits forming homodimers in the ER before forming functional tetrameric receptors.
Using the blue native PAGE system, McIlhinney and coworkers (26) show that NR1 subunits are able to form ho-modimers when expressed alone (26). Our FRET results support such a conclusion. However, our results are not consistent with the findings of McIlhinney and co-workers (26) in that NR2A may not form homo-oligomers similar to the NR1-1a subunit when expressed alone (26). When the cells were cotransfected with CFP-NR2A, YFP-NR2A, and the wild-type NR1 subunit, with CFP-NR2B, YFP-NR2B, and the wild-type NR1 or with CFP-NR1-1a, YFP-NR1-1a, and the wildtype NR2B, respectively, FRET signals were still detected (Fig.  5). This indicates that, in functional NMDA receptors, these subunits remain as homodimers.
NR2A and NR2B Subunits Can Form Heteromers-Co-expression of CFP-NR2A and YFP-NR2B yielded a much smaller but significant FRET signal in living cells (FR ϭ 1.87 Ϯ 0.10, n ϭ 27). When these two constructs were co-transfected with the wild-type NR1, a FRET signal was still observed (Fig. 5).
These results indicate that NR2 subunits can form heteromers without the help of NR1, and the addition of NR1 may not interrupt the interaction between different NR2 subunits. This result is consistent with previous findings using co-immunoprecipitation, which indicate that different NR2 subunits can occur within an assembly (21)(22)(23)(24).
FRET Measurement in Hippocampal Neurons-We further examined the subunit composition of NMDA receptors in cultured hippocampal neurons at 8 -10 days in vitro using threecube FRET. Because of the low expression levels of the fluorophore-tagged NR subunits, a 500-ms exposure time was used for imaging neurons, and the data were mainly obtained from the cell body and primary dendrites (Fig. 6A). It is known that hippocampal neurons express endogenous NMDA receptors, which makes the neuronal system more complex to analyze. Nonetheless, we still obtained valuable information as shown in Fig. 6B. In the cell bodies and primary dendrites of these mature hippocampal neurons, NMDA receptors did not coassemble with AMPA receptors. In addition, dimerization of NR subunits still occurred, whether retained in the ER or expressed in the cell membrane.
Dynamic Conformational Changes in the N Termini of NR Subunits during Subunit Assembly-FRET measurement is not limited to imaging protein-protein interaction. When each YFP-tagged molecule is associated with a CFP-tagged partner, tk;4FRET can also be used to estimate the physical distance between donor and acceptor fluorophores. According to the method suggested by Erickson et al. (42,44) (see "Materials and Methods"), we calculated the A b value from every cell expressing CFP-YFP concatemer and found that the FR data congregated at A b ϳ 1 (Fig. 7A), as expected for a molecule containing both CFP and YFP at a fixed 1:1 stoichiometry. In contrast, the A b value for co-expression of CFP-NR1-1a and YFP-GluR1 or co-expression of CFP-NR1-1a and YFP-␣1 was ϳ0 (Fig. 7, B and C). It further indicated that NR subunits did not form complexes with AMPA or GABA receptor subunits (Fig. 3). The A b values from cells co-expressing CFP-NR1-1a and YFP-NR1-1a or CFP-NR2A and YFP-NR2A are distrib-uted from 0 to 1, and the FR max value was estimated to be ϳ2.8 and 2.1, respectively (Fig. 7, D and E). In addition, the relative dissociation constant (K d,EFF ) calculated for CFP-NR1 and YFP-NR1 or CFP-NR2A and YFP-NR2A co-expression was very small (data not shown) and in accordance with the relatively high stability of the dimerization. We further extended this analysis to all FRET pairs, and the results are shown in Fig. 7F.
We analyzed our results to get more information about the distance and binding affinity between the two N termini of the NR subunits. Before forming functional acceptors, FR max and K d,EFF values for N termini of the dimers of the same NR2 subunit (CFP-NR2A and YFP-NR2A, CFP-NR2B and YFP-NR2B) were similar, with FR max being ϳ2. The distance between the N termini of the NR1 subunit was much shorter than that of the NR2 subunit, with FR max being ϳ2.8. After forming functional NMDA receptors, the FR max value for CFP-NR2A and YFP-NR2A decreased to ϳ1.6, whereas that for CFP-NR2B and YFP-NR2B increased greatly to ϳ2.5. As to YFP-NR1-1a and CFP-NR1-1a, when co-expressed with the wild-type NR2A subunit, FR max decreased to ϳ2.4, whereas when co- expressed with the wild-type NR2B subunit, FR max increased significantly to ϳ3.5 (Fig. 7F).

DISCUSSION
In this study, we took advantage of FRET between CFP and YFP that were genetically inserted into NR subunits as a reporter for the presence of dimerization between two fluorophore-tagged subunits. The results provided further evidence for the dimer-of-dimers assembly of NMDA receptors between the NR1 and NR2 subunits. Our data showed that NR subunits (not only NR1-1a but also NR2A and NR2B) formed dimers in living cells before they formed functional NMDA receptors. Evidence from the blue native PAGE system and sucrose-gradient centrifugation analysis suggests that the NR1 subunits expressed alone in heterogeneous cells can self-associate to form homodimers, whereas the NR2 subunits may not form homo-oligomers similar to the NR1-1a subunits (26), although the current model of NMDA receptor arrangement is a dimer of dimers based on the tandem and truncated experiment (28).
Here we addressed this issue by the FRET technique. CFP and YFP have previously been shown to function as a FRET pair with R 0 being 49 Å (47). This distance makes them optimal for reporting the protein-protein interaction in vivo in a non-invasive manner. In our experiments, no FRET signal was detected between fluorophore-tagged NR subunits and GluR1 or GABA A ␣1 subunits in HEK293 cells, even though the ␣1 subunit has been shown to accumulate in the ER (46), identical to that of NR subunits expressed alone (Fig. 4). In addition, we did not observe any FRET signals between CFP-tagged NR subunits and YFP-tagged GluR1 subunits in the hippocampal neurons, in which both subunits should share the same secretion pathway. However, when CFP-NR1-1a and YFP-NR1-1a were coexpressed in HEK293 cells, significant FRET signals were observed, indicating the formation of specifically assembled NMDA receptor complexes. The result was consistent with the previous conclusion that NR1-1a can form homodimers when expressed alone (26). In addition, when CFP-NR2A and YFP-NR2A or CFP-NR2B and YFP-NR2B were expressed in HEK293 cells, we also observed FRET signals. The data provided strong evidence that NR2A and NR2B subunits can form homomers when expressed alone in living cells. Because glutamate receptors, including NMDA receptors, are thought to be tetramers (4 -8) and the NR1 subunits can form homodimers (26), it is reasonable to infer that NR2A and NR2B subunits expressed alone can also form homodimers. The dimerization of the subunits occurring in the ER may be the general theme guiding the assembly of NMDA receptors, as that is guiding the assembly of AMPA receptors (29 -31), K ϩ channels (39), and cyclic nucleotide-gated channels (40). However, although extremely unlikely, our results cannot completely rule out the possibility that the matured tetrameric NMDA receptors are assembled in a symmetrical NR1-NR2-NR1-NR2 orientation. Indeed, work on AMPA receptors has indicated that the two GluR1 and two GluR2 subunits in one receptor may be in a symmetric arrangement with the orientation of 1-2-1-2 (48).
Previous studies have shown that the NR2A subunits can be co-immunoprecipitated by the anti-NR2B antibody in detergent extracts of brain tissue under non-denaturation conditions and vice versa (21,22). Here, we found that, in living cells, a FRET signal was observed between NR2A and NR2B subunits, which further indicated that heteromers could be formed between different NR2 subunits. The NR1/NR2A and NR1/NR2B receptor subtypes have distinct functional properties (49). Evidence from recent studies further suggests that they are involved in different intracellular signaling machineries and different patterns of plasticity (50 -52). It is still unclear about the pharmacological properties and functions of NR1/NR2A/NR2B receptor subtypes in neurons. Our results suggest that more attention should be paid to this subtype of NMDA receptors in functional studies.
We further examined the interaction between NR subunits in living hippocampal neurons. Basically, the results were consistent with those from the HEK293 cells, indicating that the NR1, NR2A, and NR2B subunits can form homomers or heteromers with each other. It is important to mention that there are some particular advantages in the experiments using the primary cultured neurons. For example, it is very interesting to know whether different subtypes of NMDA receptors discrimi- nately distribute on individual synapses or whether the interaction between NMDA and AMPA receptors occurs in these particular sites.
When each YFP-tagged molecule is associated with a CFPtagged molecule, FRET can be used to estimate the distance between donor and acceptor fluorophores (53). Based on previous work and our data, the NMDA receptor is assembled as a dimer of dimers. Therefore, when co-expressing CFP-and YFPtagged NR1 or NR2 subunits with or without wild-type NR2 or NR1 subunits, each donor fluorophore was associated with one acceptor fluorophore, and it was possible to estimate the distance between donor and acceptor. Using the method suggested by Erickson et al. (42,44), we determined the FR max values to estimate the distances and the conformational changes of the N termini of NMDA receptors. Our results indicate that, before forming functional NMDA receptors, the distances between the N termini of the NR2A/NR2A and NR2B/NR2B dimers have no difference, but these distances are much larger than that of the NR1/NR1 dimers. After forming functional acceptors, the distance between the N termini of the NR1/NR1 dimer changed greatly. When forming NMDA receptors with the NR2A subunits, the distance decreased, whereas when forming NMDA receptors with the NR2B subunits, the distance increased. Also, when co-expressing with wild-type NR1-1a, the distance between CFP-NR2A and YFP-NR2A or between CFP-NR2B and YFP-NR2B was changed greatly. It is suggested from this observation that the N termini of NMDA receptors may undergo conformational change during heteromeric assembling. Furthermore, the distances between the N termini of the NR1 subunits are different when they are assembled with different NR2 subunits, which may also contribute to the difference in properties exhibited by NMDA receptors containing NR2A and those containing NR2B. We also noticed that a recent study using single-particle electron microscopy provided evidence for a variable separation of the two dimeric extracellular N-terminal domains of native AMPA receptors associated with glutamate binding (54). It would be interesting to observe distance change of the dimeric domains of NR1 and NR2 subunits by FRET in living cells during activation and deactivation of the NMDA receptors.
It would be very interesting to take the additional step and actually estimate the distance between the labeled proteins. However, there are limitations that must be attended to. First, the distance estimated here is the physical separation of GFP chromophores and not the proteins they label. However, the CFP or YFP is inserted in the N terminus of NMDA receptors, just following the signal peptide, and the distance between the fluorophore reflects that of the N termini. Second, changes in donor-acceptor distance can cause changes in FRET efficiency, but other aspects such as dipole-dipole orientation of the donor and acceptor fluorophores also affect FRET efficiency. Anisotropy experiments with fluorescence proteins tagged to various proteins suggest that these fluorophores retain significant mobility, which reduces or eliminates the orientation effect on FRET efficiency (55-58). Our fluorescence protein-fused receptor constructs were similar in design and were likely exhibiting similar mobility.
In summary, our data provide further evidence for the subunit assembly of NMDA receptors as a dimer of dimers and the existence of two different NR2 subunits in one NMDA receptor complex. In addition, our results suggest that the conformation of the NMDA receptor may be affected differentially by different subunit composition. Although the underlying mechanism for the assembly of the homodimers and that of the functional tetramers are still uncertain, we anticipate that future studies on this subtype-specific assembly will lead to refinement of the functional heterogeneity of NMDA receptors.