Assembly and forward trafficking of NMDA receptors (Review)

Abstract

N-Methyl-D-aspartate (NMDA) receptors are a subclass of the excitatory, ionotropic L-glutamate neurotransmitter receptors. They are important for normal brain function being both primary candidates for the molecular basis of learning and memory and in the establishment of synaptic connections during the development of the central nervous system. NMDA receptors are also implicated in neurological and psychiatric disorders. Their dysfunction which is primarily due to either hypo- or hyper-activity is pivotal to these pathological conditions. There is thus a fine balance between NMDA receptor-mediated mechanisms in normal brain and those in diseased states where receptor homeostasis is perturbed. Receptor activity is due in part to the number of surface expressed receptors. Understanding the assembly and trafficking of this complex, heteromeric, neurotransmitter receptor family may therefore, be pivotal to understanding diseases in which their altered activity is evident. This article will review the current understanding of the mechanisms of NMDA receptor assembly, how this assembly is regulated and how assembled receptors are trafficked to their appropriate sites in post-synaptic membranes where they are integral components of a macromolecular signalling complex.

• NMDA receptors
• ionotropic glutamate receptors
• trafficking
• scaffold proteins
• PSD-95

Abbreviations
AMPA	=	α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid
COPII	=	coatomer protein II
CREB	=	cAMP response element binding
ER	=	endoplasmic reticulum
ERK	=	extracellular regulated kinase
FRET	=	fluorescence resonance energy transfer
LAO-QBP	=	lysine-arginine-ornithine and glutamine binding protein
LIVBP	=	leucine-isoleucine-valine binding protein
LTD	=	long-term depression
LTP	=	long-term potentiation
MAGUK	=	membrane associated guanylate kinase
NMDA	=	N-methyl-D-aspartic acid
SAP	=	synapse associated protein.

Introduction

N-Methyl-D-aspartate (NMDA) receptors are one of the major mediators of fast excitatory neurotransmission in the central nervous system. They are a subclass of ionotropic glutamate neurotransmitter receptors, a gene family that also includes the α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA), kainate and orphan receptors. NMDA receptors play a central role in long-term potentiation (LTP) and long-term depression (LTD), thought to be molecular mechanisms of learning and memory; in pain perception and also in synaptogenesis during the development of the central nervous system. They are atypical receptors in that they are activated by the binding of two co-agonists, L-glutamate and glycine, and by the alleviation of a voltage-dependent blockade by magnesium ions. NMDA receptors are often referred to as coincidence detectors since their activation is indicative of both pre- and post-synaptic neuronal activity [1]. Receptor activation results within milliseconds in the opening of the integral ion channel that has a high permeability for calcium ions, neuronal depolarization and subsequent activation of NMDA receptor downstream signalling pathways. These include activation or inhibition of both the Ras-dependent extracellular regulated kinase (ERK) and cAMP response element binding (CREB) dependent gene expression pathways that mediate either the apoptotic or the LTP/LTD effects of receptor activation [2], [3]. NMDA receptors are implicated in a variety of neurological, neurodegenerative and psychiatric disorders [4]. For example, over-activation of NMDA receptors leads to an excessive influx of calcium ions and resultant excitotoxic and apoptotic neuronal cell death as in ischaemia; Memantine, a non-competitive NMDA receptor antagonist and thus an anti-excitotoxic agent, is licensed for use in the treatment of Alzheimer's disease. Conversely, NMDA receptor hypofunction is associated with schizophrenia. There is thus a fine balance between NMDA receptor-mediated mechanisms in normal brain and those in neuropathological conditions where receptor homeostasis is perturbed. Receptor activity is in part related to the number of surface expressed receptors. NMDA receptors are a family of complex, heteromeric, integral membrane proteins whose respective cell surface expression and subcellular localization is tightly regulated. In this article, we will review the assembly and trafficking of NMDA receptors. It is important to delineate these processes since in the long term, it may lead to improved understanding of the mechanisms of the diseases in which NMDA receptors are implicated and ultimately, in improved therapies.

NMDA receptor subunits

There are seven NMDA receptor genes. These encode the NR1, NR2A-NR2D and NR3A-NR3B NMDA receptor subunits. Each subunit has a similar domain organization and amino acid identity, ∼ 29% between NR1 and NR2 subunits; ∼ 30% between NR1 and NR3; ∼ 28% between NR2 and NR3; ∼38–52% between NR2A-NR2D and ∼ 55% between NR3A and NR3B. All subunits have an extracellular N-terminal region that can be subdivided into two discrete domains based on the amino acid sequence homology between the NMDA receptor subunits and amino acid binding proteins of bacteria. Thus, the first N-terminal ∼ 400 amino acids have homology with the bacterial periplasmic leucine-isoleucine-valine binding protein (LIVBP) whereas the region containing amino acids 420–550 shows similarity with the bacterial lysine-arginine-ornithine and glutamine binding protein (LAO-QBP). There are three transmembrane domains and a single re-entrant loop membrane domain that mutagenesis approaches have shown forms the inner lining of the Ca^2 + permeable cation channel. The conventional glycine and the glutamate binding sites have been localized to the S1-S2 domains of the NR1 and NR2 subunits respectively. S1 in the N-terminal domain corresponds to amino acids ∼ 420–550 (the LAO-QBP domain) that are proximal to the membrane and S2 is the extracellular loop found between the third and fourth membrane domains. Each subunit has an intracellular C-terminal tail that is involved in association with scaffolding proteins (NR2 subunits) or the regulation of receptor trafficking (NR1 and NR2 subunits). The NR1 subunit gene undergoes extensive splicing involving its exons 5, 21 and 22 to yield eight variants NR1-1a,1b to NR1-4a,4b. The NR1 ‘b’ forms all contain an additional exon, exon 5, in the N-terminal region that encodes for a 21 amino acid insert within the LIVBP domain which if present is inserted after amino acid 190. NR1 subunits which contain the C1 cassette (exon 21) are NR1-1a,b and NR1-3a,b. NR1 subunits which contain the C2 cassette (exon 22) are NR1-1a,b and NR1-2a,b. The absence of the C2 cassette results in the introduction of a STOP codon which yields a shortened C-terminal domain. This truncated C-terminal is referred to as the C2’ cassette and it is found in the NR1-3a,b and NR1-4a,b variants. The structures of the soluble NMDA receptor NR1 S1-S2 and NR2 S1-S2 loop constructs in the presence of glycine and glutamate respectively have been resolved [5], [6]. These S1 and S2 lobes form a Venus flytrap or clamshell structure that is opened following the binding of either glycine (NR1) or glutamate (NR2).

Functional NMDA receptors are formed from the co-assembly of the obligatory NR1 low affinity glycine binding subunit with the L-glutamate binding NR2 subunit and/or high affinity, glycine binding NR3 subunits. There has been much controversy in the literature as to whether NMDA receptor subunits assemble as (NR1)₃(NR2)₂ or (NR1)₂(NR2)₃ pentameric or (NR1)₂(NR2)₂ tetrameric complexes. At this time however, the weight of evidence supports a tetrameric assembly with the key findings supporting this model being (i) the electron microscopy studies of recombinant AMPA, GluR2 homomers [7] and native AMPA receptors [8] that both visualized tetrameric structures; (ii) the X-ray crystallographic studies that identified the [NR1-(S1-S2)]-[NR2-(S1-S2)] heterodimer as the functional unit of assumed tetrameric NMDA receptors [6], and (iii) the use of a novel, single molecule technique employing the photobleaching of green fluorescent tagged NR1 or NR2 subunits to demonstrate a tetrameric (NR1)₂(NR2B)₂ quaternary structure [9]. This tetrameric model is in accord with earlier electrophysiological studies which showed that the binding of two molecules of glycine (NR1 subunit) and two molecules of glutamate (NR2 subunit) are required for NMDA receptor channel activation [10]. Figure 1 summarizes the pertinent features of assembled NMDA receptors and NMDA receptor subunits.

Figure 1.  A schematic diagram depicting pertinent features of NMDA neurotransmitter receptors. (A) summarizes the NMDA receptor subunits; (B) is a schematic of the NMDA receptor NR1, NR2 and NR3 subunits including all the NR1 splice variants; (C) shows a model of a tetrameric NMDA receptor comprising of two NR1 and two NR2 subunits; (D) shows the domain structure that is shared between both NR1 and NR2 subunits where the S1 and S2 domains form the glycine (NR1) or L-glutamate (NR2) binding sites, the Ca^2 + permeable pore is formed by the re-entrant membrane domain M2; (E) is a schematic depicting the Venus fly trap model for the opening of the gated ion channel and it is reproduced and adapted with permission from adapted from [5], [6], [57].

NMDA receptor subtypes

NMDA receptor diversity is generated by the association of the obligatory NR1 subunit with the different NR2 subunits. It is generally agreed that there are four major subtypes: NR1/NR2A; NR1/NR2B; NR1/NR2C and NR1/NR2D NMDA receptors that differ in their pharmacological, functional and biophysical properties, their developmental profiles and their neuronal cell type and subcellular distributions and most recently, in their lateral mobilities in post-synaptic membranes. The NR1/NR2A and NR1/NR2B are the major subtypes expressed in adult brain where the current dogma is that NR1/NR2A receptors are synaptic and NR1/NR2B, extra-synaptic receptors. NR1/NR2C and NR1/NR2D have a more restricted localization and NR1/NR2B receptors are thought to be of primary importance during development. However, both pharmacological studies in conjunction with functional methods and biochemical approaches have yielded evidence for receptors that contain two types of NR2 subunit. For example, the existence of NR1/NR2A/NR2B receptors was identified by sequential immunoprecipitations from detergent extracts of adult mammalian cerebral cortex [11] and recently, Thomas et al. [12] published evidence that these, NR1/NR2A/NR2B were synaptic. The relative abundances of NR1/NR2A, NR1/NR2B versus NR1/NR2A/NR2B receptors is disputed in the literature with support for trimeric, i.e., two types of NR2 assembled with NR1 subunits, being either a minor [e.g., 11], a major [13] or more recently, at least in the developing and adult hippocampus, almost a one third constituent [14]. The relative abundances of these subtypes is not an easy question to resolve since NMDA receptors are relatively resistant to detergent solubilization because of their association via scaffolding proteins with a post-synaptic macromolecular complex (for review see [15]). Discrepancies in the literature may thus be explained by differences in detergent solubilization conditions. One can envisage that these may result in the enrichment of certain NMDA receptor subpopulations which may be more readily extractable. Detergent extraction conditions that solubilize all NMDA receptors have not been established thus, this important issue has still to be determined. There are other examples in the literature of different combinations of NR2 subunits co-existing in assembled receptor complexes, i.e., NR1/NR2A/NR2C; NR1/NR2B/NR2D and NR1/NR2A/NR2D and this is reviewed in [15].

The role of the NMDA receptor NR3 subunit is unclear. In 2002, Chatterton et al. [16] reported that NR1/NR3 subunits formed a novel, glycine-gated, excitatory receptor. This was an exciting but controversial finding since, at the time, other groups had difficulty replicating this discovery. Recently however, two groups, Madry et al. [17] and Smothers and Woodward [18], have now succeeded in reproducing glycine activated excitatory currents following expression of either NR1/NR3A or NR1/NR3B in mammalian cells and in Xenopus oocytes. Further, recombinant NR1 and NR3 subunits were shown to co-assemble as (NR1)₂(NR3)₂ tetramers in oocytes [9]. The existence of such an in vivo receptor however, has yet to be proven. The NR3A subunit was shown to co-immunopreciptate with both NR1 and NR2B subunits from cerebrocortical extracts of P6, postnatal day 6, mouse brains [16]. Further, when exogenously expressed in hippocampal neurons, NR3B subunits associated with NR1 and NR2A but like recombinant NR1/NR2A/NR3 trimers, these complexes were retained in the endoplasmic reticulum (ER) suggesting that NR3 subunits per se may regulate receptor trafficking [19], [20]. A more recent report however was unable to demonstrate co-immunoprecipitaton of recombinant NR2 and NR3 subunits [21].

Native NMDA receptor diversity in addition to NR2 subunits, can also be generated by heterogeneity of NR1 subunits. In early studies, Sheng et al. [22] suggested that the NR1 splice forms were segregated, with the NR2A and NR2B subunits being preferentially associated with the NR1-1b and NR1-3b variants. However, a later study found no preferential assembly between NR2 subunits and particular NR1 splice variants [23] and both Blahos and Wenthold [23] and Chazot and Stephenson [24] showed that different NR1 splice variants can co-exist within the same NMDA receptor.

NMDA receptor assembly

For any multimeric, integral membrane protein, cell surface expression is dependent on transcription of the appropriate genes, their translation and co-assembly of the appropriately folded subunits in the ER, their post-translational modification and subsequent export to the cell surface along the secretory pathway. As for many other types of receptor, the cell surface trafficking of NMDA receptors has been extensively studied using as a model, expression of recombinant receptors of defined subunit composition in mammalian cell lines.

The NR1-1a subunit is the most abundant of the NR1 splice variants; most studies have been carried out with this NR1 isoform. NR1-1a when expressed alone is not trafficked to the cell surface. This is due to an ER retention signal in the NR1 C1 cassette; these same trafficking properties are also evident in the NR1-1b, -3a and -3b splice variants which also contain the C1 cassette [25–27]. Co-expression of these variants with NR2 subunits overcomes their ER retention and results in expression of cell surface receptors. The NR1-2a, 2b, 4a and 4b splice forms do not have the ER retention motif and indeed are efficiently trafficked to the cell surface independently of NR2 subunits in both heterologous expression systems and also, when over-expressed in neurons. There is no evidence to show that these purported NR1-2 and NR1-4 homomers are functional thus the significance of these observations is unclear. NR2 subunits when expressed alone are also not trafficked to the cell surface [28]. The possible existence of ER retention motifs in the NR2B subunit was investigated using Tac-NR2B C-terminal tail chimeric reporter constructs [29]. Deletions of the NR2B C-terminal tail resulted in increased cell surface expression and although three putative ER retention sites were identified, RRR (1110–1112), KRRK (1079–1082) and KKR (1090–1092), mutation of each did not result in an increase in surface expression. It was thus concluded that either multiple sites are involved in retention of the NR2B C-terminus or that a more complex structure is responsible for the NR2B ER retention. [29]. A more recent report studying co-expressed NR1 and NR2 subunits rather than Tac reporter constructs found that when NR1 and NR2B subunits were truncated immediately following the fourth membrane domain, when expressed individually, they did not reach the cell surface but when expressed together, robust cell surface expression was evident [30]. These findings suggested that ER retention motifs were present in regions preceding the C-terminal tail and that co-expression of NR1 and NR2 together masked these motifs. Further truncations showed that these ER retention motifs resided in the third membrane domain [30].

In addition to trafficking motif differences between NR1 and NR2 subunits, it appears that they also differ in metabolic stability [21]. In single subunit expression experiments, ∼ 65% of expressed NR1-1a subunits are stable in contrast to NR2A and NR3A where their respective half lives were determined as 5.7 h and 2.8 h, respectively [21]. Co-expression of NR2 with NR1 results in an increased NR2 stability [21].

Similar phenomena to those found in model expression systems with regard to the trafficking of NMDA receptors have been shown to occur in vivo. For example, in NR1 (-/-) mice where the NR1 gene knock-out was targeted to hippocampal CA1 pyramidal cells, NR2A and NR2B subunits were found to accumulate in the ER of these neurons [31]. A pool of unassembled NR1 C2 exon-containing NR1 subunits (i.e., NR1-1a,b and/or NR1-2a,b) has been demonstrated in wild-type rat brain [24]. Huh and Wenthold [32] also identified intracellular NR1 subunits that were not assembled with NR2 subunits and they showed that these subunits had a rapid degradation rate compared to assembled NR1 subunits. In cerebral cortical neurons where both NR1 and NR2 are both endogenously expressed, ∼ 75% NR2A subunit expression is stable over a 24 h period in contrast to the short half-life for single NR2A subunit heterologous expression. Thus it can be extrapolated from these observations that the first control point and rate limiting step in NMDA receptor biogenesis is transcription of the appropriate NR2 subunit genes. Following transcription, the translated NR2 subunits are able to assemble with the ready pool of NR1 subunits and quality control mechanisms operate to ensure that only NR2 subunits assembled with NR1 subunits exit the ER.

Three alternative pathways, each supported by experimental evidence, can be proposed as mechanisms by which NMDA receptor subunits co-assemble in the ER. These are shown in schematic form in Figure 2. The first of these is that NMDA receptors form NR1-NR1 and NR2-NR2 homodimers and that these pairs then co-assemble to form the hetero-tetrameric complex. The existence of NR1-NR1 dimers is supported by their detection by immunoblotting following size resolution by blue native-polyacrylamide gel electrophoresis (PAGE) [33] and also following cross-linking with the sulphydryl specific cross linking agent, bismaleimidohexane [21] or, more conventional sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE) under both reducing and non-reducing conditions [34]. NR1-NR1 homodimer formation is also inferred from the study of tandem NR1-NR2 constructs [35]. One report showed that NR1-NR1 subunits are covalently linked by disulphide bridging [34]. Support for NR2-NR2 dimer formation is from a fluorescence resonance energy transfer (FRET) approach using cyan fluorescent protein (CFP) and yellow fluorescent protein (YFP) tagged NR1 and NR2 subunits [36]. Also, NR2-NR2 dimer formation was inferred as above from the study of tandem NR1-NR2 constructs [35]. Atlason et al. [21] however using the chemical crosslinking approach were unable to detect significant NR2-NR2 dimer formation. This observation together with the relative metabolic instability of the NR2 subunits in the absence of NR1, led them to propose a model whereby an NR1-NR1 dimer forms a correctly folded stable complex to which two NR2 monomers are added sequentially to form the NMDA receptor tetramer. For both these mechanisms involving initial NR1-NR1 dimer formation, it would suggest that the unassembled ER pool of NR1 subunits should be present as dimers. However, gel filtration experiments are more consistent with this pool being monomers [24].

Figure 2.  Possible models for the assembly of NR1/NR2 NMDA receptors. (A) depicts three possible schemes as discussed in the text for the assembly of tetrameric, NR1/NR2 NMDA receptors. (B) depicts a possible scheme for the inter-subunit interactions for the assembly of NMDA receptors via the initial NR1-NR1 homo-dimer formation route. It should be pointed out however that in the publication by Schuler et al. [37] that appeared during the writing of this review, evidence was presented that the assembly pathway was via an initial NR1-NR2 hetero-dimer formation. This was not mediated by the NR1 LIVBP domain since the NR1^ΔNTD deletion construct (lacking the amino acid sequence 23–376, the LIVBP domain) was still able to form functional receptors with full-length wild-type NR2 subunits. This Figure is reproduced in colour in Molecular Membrane Biology online.

An alternative model of initial NR1-NR2 dimer formation may be put forward as a result of the structural studies of co-crystallized NR1-(S1-S2) and NR2-(S1-S2) soluble constructs [6] and two recent reports from the group of Laube [37], [38]. In the structural approach, it was shown that neither NR1-(S1-S2) nor NR2-(S1-S2) could form homo-oligomers but NR1-(S1-S2) and NR2-(S1-S2) could, leading to the deduction that this dimer is the functional unit of the receptor and that NR1-NR2 may be an assembly intermediate (but see below for more discussion). Laube et al. [37] employed FRET analysis of differentially tagged subunits and native polyacrylamide gel electrophoresis to show that NR1 subunits form homo-oligomeric aggregates but NR2 (and NR3) subunits could only form either homo- or hetero-oligomers in the presence of NR1. Further, if NR1 homo-oligomerization was prevented by the use of an N-terminal truncation construct i.e. NR1^ΔNTD (lacking the amino acid sequence 23–376, the LIVBP domain), this did not result in a loss in the formation of functional receptors [37]. This suggested that the initial step in receptor formation is not via NR1-NR1 dimer formation but via NR1-NR2 hetero-dimer formation followed by (NR1/NR2)-(NR1/NR2) tetramerization. It is not clear at this point how to rationalize the different findings from the various groups.

If NR1-NR1 homo-dimer formation is the initial step in NMDA receptor biogenesis, this implies that this must be mediated by preferred protein-protein binding domains. By analogy with the studies of Stern-Bach and colleagues on AMPA receptor biosynthesis [39], it could be predicted that it is the N-terminal LIVBP domain that mediates this NR1-NR1 dimer formation. It is within this domain that the NR1–NR1 inter-subunit disulphide bond via C-79 has been suggested to form [34]. Further, an epitope-tagged NR1-2a trafficking mutant where the tag was introduced after amino acid 81 in the LIVBP domain, impaired NR1-NR1 disulphide bonding and dimer formation [34]. The LIVBP domain is requisite for NR1-NR2 association since co-expression of NR2A with NR1 where the N-terminal 1–380 amino acids were deleted (i.e., NR1^{Δ 1–380}) did not yield functional NMDA receptors [33]. This NR1-NR2 protein-protein interaction mediated via the LIVBP domain may seem at odds with the ability of the soluble constructs, NR1-(S1-S2) and NR2-(S1-S2) to form dimers [6]. However, the crystallographic study did not contain the distal N-terminal LIVBP domain. One can envisage a scheme whereby initial NR1-NR2 association is mediated via the LIVBP domain with subsequent stabilization via the NR1-NR2 S1-S2 domains and that multiple protein-protein contacts maintain the final assembled complex (Figure 2).

Finally, NR2B and NR2A subunit genes are actively transcribed in the same neuron so a question not addressed at all so far is what controls (if indeed it is a regulated process and not one of random association), the preferred formation of NR1/NR2A, NR1/NR2B or NR1/NR2A/NR2B receptors? Do protein-protein binding domains mediate the formation of preferred complexes or is translation in the ER compartmentalized to ensure appropriate subunit co-assembly? Any one of the three schemes outlined in Figure 2 could be amended to accommodate heterogeneity of NR2 subunits. As yet, no information is available.

Trafficking of NMDA receptors

Once assembled in the ER, NMDA receptors are forward trafficked to the Golgi network followed by targeted insertion into synaptic or extra-synaptic membranes via the secretory pathway. The forward trafficking from the ER appears to be regulated since a motif, HLFY, found in the NR2B subunit immediately following the fourth membrane domain was shown to be essential for the exit of NR1/NR2B assembled receptors from the ER [29]. Mutation of this domain resulted in the retention of assembled receptors in the ER. Later studies refined this domain to three amino acids, i.e., EHL (the NR2B amino acid sequence is EHLFY), and it was also shown that this motif operates for NR1/NR2A NMDA receptors [40]. The corresponding sequence in NR2C and NR2D subunits is EHLVY which from the study of Yang et al. [40] would suggest that it may also regulate the forward trafficking of these receptor subtypes but so far, this has not been investigated. The rate of exit from the ER is also regulated. Mu et al. [41] showed that the rate of exit of NMDA receptors is dependent on the NR1 splice variant. NR1 splice forms containing the C2’ cassette (NR1-3a,b and NR1-4a,b) are exported more rapidly than those containing the C2 cassette (NR1-1a,b and NR1-2a,b). A novel ER export sequence, TVV, was identified at the C-terminus of the NR1 C2’ cassette [25], [41]. The coatomer protein II (COPII), a protein involved in the trafficking from the ER to the Golgi of secretory proteins in general, was shown to bind to this TVV motif [41]. Interestingly, splicing of the NR1 RNA was shown to be activity-dependent with the C2’ variant mRNA predominating following neuronal stimulation [41].

Assembled NMDA receptors do not exit the ER as discrete entities but they have been shown to associate with other proteins into macromolecular complexes that appear to be then transported as a single unit. The pivotal proteins in the formation of these complexes are the post-synaptic density (PSD-95) family of membrane associated guanylate kinase (MAGUK) scaffold proteins that provide a link between NMDA receptors and membrane targeting proteins (Figure 3). There are four PSD-95 MAGUKs, PSD-93, PSD-95, synapse associated protein 97 (SAP97) and SAP102. Each can associate with NMDA receptors via an intracellular ES(E/D)V domain found at the distal C-termini of all NR2 subunits [reviewed in [42]]. The developmental expression profiles of the MAGUKs together with some supporting evidence suggested that PSD-95 associates in vivo with NR1/NR2A and SAP102 with NR1/NR2B receptors respectively [43], [44]. This dogma was however recently challenged by Al-Hallaq et al. [14] who found no discrimination between NR1/NR2A, NR1/NR2B and PSD-95, PSD-93 and SAP102 in adult rat hippocampal neurons. Association of NMDA receptors with these scaffold molecules is known to occur in the ER. It is not known if the scaffold associates with assembled NR1/NR2 subunits or alternatively (although unlikely), it may associate with NR2 subunits prior to their assembly with NR1. The ER retention of the NR1-3a and NR1-3b variants can be over-ridden by association with PSD-95 (NR1-3 subunits have a C-terminal four amino acid PDZ-interacting domain) [25], [41]. But, because this variant also contains the TVV export sequence that promotes forward trafficking due to association with COPII, this leads one to question which of the two different interactions, i.e., NR1 C2’-PDZ-domain containing protein or NR1-C2’-COPII, may be responsible for ER exit. By mutating the three C-terminal amino acids of a Tac NR1-3 chimera, Mu et al. [41] showed that disruption of the COPII binding site but not the PDZ binding site resulted in a decreased forward trafficking thus concluding that the TVV ER export signal recruits receptors to ER exit sites.

Figure 3.  A schematic diagram showing the organization of the post-synaptic NMDA receptor macromolecular signalling complexes. Reproduced and adapted with permission from [57]. This Figure is reproduced in colour in Molecular Membrane Biology online.

SAP102 and PSD-95 mediate the formation of the NMDA receptor/targeting protein complexes, NR1/NR2B/SAP102/Sec8/m-Pins or NR1/NR2B/PSD-95/Sec8/m-PIns [45], [46]. The Sec8 protein is part of the exocyst complex, a large complex of proteins that are essential in the secretory pathway being associated with the ER, the Golgi apparatus and trans-Golgi network. Disruption of the SAP102-Sec8 protein-protein interaction resulted in a decrease in NMDA receptor cell surface delivery in recombinant systems and in neurons, a decrease in the delivery of synaptic NMDA receptors. mPins (also named LGN since they contain a series of Leu-Gly-Asn repeats) belongs to a conserved family of proteins which mediate G-protein signalling. It is implicated in protein-protein interactions and in Drosophila, it associates with Discs large (Dlg) and is a key protein in the Frizzled signalling pathway that regulates the establishment of cell polarity and asymmetric cell division. It is not clear if this same complex traffics all NMDA receptor subtypes since so far only the NR1/NR2B receptor has been investigated.

In addition to mediating the formation of the exocyst complex, PSD-95 was shown to be the link in the formation of a motor protein complex, NR2B/PSD-95/m-Lin/KIF17 [47], [48]. This complex was shown to transport NR2B subunits to dendrites and then to deliver them to extra-synaptic localizations. It is speculated that these trafficking complexes would be formed following exit from the Golgi network. Indeed, Washbourne et al. [49] imaged mobile transport packets containing NR1 subunits moving along microtubules at a velocity of ∼ 4 µm/min before and during synaptogenesis. For the NR2B/PSD-95/m-Lin/KIF17 study, it should be noted that co-association of NR2B with NR1 subunits was not investigated. So, although the assembly of both NR1 and NR2 subunits is requisite for exit from the ER as discussed above, the possibility intriguingly remains, that NR2B subunits may be trafficked alone in the absence of NR1 subunits. Assembly with NR1 may then be envisaged as occurring locally and on demand. There are no reports to date of the association of other NMDA receptor subunits with the m-Lin/kinesin protein complex.

The existence of both synaptic and extra-synaptic NMDA receptors is well documented in the literature and the subcellular compartmentalization of defined subtypes is critical for the activation of downstream signalling pathways resulting in either LTP or LTD, neuroprotection (synaptic) or neurotoxic (extra-synaptic) pathways. Although there is now an understanding of how NMDA receptors are delivered to membranes, the mechanism or the determining factor in the subcellular destination remains elusive. The scaffold proteins are perhaps the prime candidates for a role in this fate since they not only mediate the formation of trafficking complexes but once receptors are in situ, they are pivotal for the anchoring and clustering of receptors in addition to coupling receptors to regulatory proteins such as kinases and phosphatases. The major splice variant of PSD-95, PSD-95α, is palmitoylated at residues C3 and C5. Palmitoylation at these cysteines was shown to be essential for membrane targeting [50]. It is not clear whether this is a synaptic targeting signal. It is of note that SAP102 is not N-terminally palmitoylated since it lacks C3 and C5 suggesting that other factors must also be involved in synaptic membrane localization.

In addition to being forward trafficked to the membrane, NMDA receptors have lateral mobility in post-synaptic membranes [51–53] and they undergo constant recycling via endocytosis [reviewed in [54–56]]. It was recently demonstrated that the two major subtypes, NR1/NR2A and NR1/NR2B have different lateral mobilities. Both NR2A and NR2B subunits were found within synapses but whereas NR2A subunit-containing receptors were relatively immobile, NR2B subunit-containing receptors were observed to move rapidly between extra-synaptic and synaptic regions. Since both NR2A and NR2B subunits contain the distal C-terminal ESDV PSD-95 binding motif, it is unclear why NR2B subunit-containing NMDA receptors are freely mobile and not anchored by PSD-95 or SAP102 within synaptic domains.

Concluding remarks

This review has summarized the current status of our knowledge of the assembly of NMDA receptors and their targeted trafficking to the post-synaptic membrane. It is clear that whilst much information has been accrued to demonstrate that these mechanisms are highly regulated, results from different laboratories using disparate methodologies need to be reconciled to produce a unified understanding. It is also evident that some fundamental questions still need to be addressed. Amongst others, these include the definitive determination of the subunit compositions and relative abundances of NMDA receptor subtypes. Is there a clear demarcation between the properties of synaptic and extra-synaptic receptors? Does this correlate with subunit composition and activation of NMDA receptor downstream signalling pathways? What determines synaptic targeting, is it local protein synthesis and delivery or is it more long range, kinesin-mediated transport mechanisms that determine the specialized subcellular destinations. The NMDA receptors are such important and pivotal proteins in the central nervous system that these questions need to be answered to enable meaningful interpretation of changes in NMDA receptors and their associated proteins observed in animal models of neurological disease and in human brain.

Acknowledgements

Research in the authors’ laboratory was funded by the BBSRC (UK). Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.