Role of Distinct NMDA Receptor Subtypes at Central Synapses

See allHide authors and affiliations

Science's STKE  19 Oct 2004:
Vol. 2004, Issue 255, pp. re16
DOI: 10.1126/stke.2552004re16


Most excitatory synapses in the brain use the neurotransmitter glutamate to carry impulses between neurons. During fast transmission, glutamate usually activates a mixture of N-methyl-d-aspartate (NMDA) and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors in the postsynaptic cell. Experimental scrutiny of NMDARs provides insight into their involvement in excitatory synaptic transmission and related processes such as as synaptic plasticity, neural development, and pain perception. There is increasing awareness that subtle variation in NMDAR properties is imparted by specific receptor subunits, and recent studies have started to provide perspective into some of the discrete tasks carried out by individual receptor subtypes.


Most excitatory synapses in the brain use the transmitter glutamate to carry impulses between neurons. During fast transmission, glutamate usually activates a mixture of N-methyl-d-aspartate (NMDA) and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors in the postsynaptic cell. The intense interest in NMDA receptors (NMDARs) arises from the fact that experimental scrutiny of their properties continues to provide remarkable insight into their involvement in excitatory synaptic transmission and related processes involving long-term changes in neuronal function. There is increasing awareness that subtle variation in NMDAR properties is imparted by specific subunits. Furthermore, recent studies have started to provide valuable perspective into some of the discrete tasks carried out by individual receptor subtypes in processes as diverse as synaptic plasticity, neural development, and the perception of pain.

NMDARs exhibit remarkable properties that distinguish them from other types of ligand-gated ionotropic receptors. First, their ion channel is subject to a voltage-dependent block by physiological levels of extracellular Mg2+. The relief of block at depolarized potentials confers on the receptor the capacity to act as a molecular coincidence detector—permitting ion flow only when pre- and postsynaptic cells are excited in unison. Second, and closely related to block by Mg2+, is the fact that NMDAR channels display a high permeability to Ca2+ ions. Ca2+ influx through the NMDAR channel contributes to a cascade of intracellular events that may trigger the long-term potentiation (LTP) or long-term depression (LTD) of synaptic currents. The intense NMDAR activation that occurs in ischemia (for example, during stroke) can lead to more extreme Ca2+ entry and cell death. Third, under physiological conditions, the receptor’s response to neurotransmitter is modified by an impressive array of molecules encountered in the extracellular milieu. The initial activation of the NMDAR requires the presence not only of glutamate, but also of the coagonist glycine. Furthermore, NMDAR activity is dramatically modified by extracellular H+, Zn2+, and polyamines. The excitatory postsynaptic current that results from NMDAR activation (NMDR-EPSC) exhibits an exceptionally slow rise (~10 ms) and decay time (>100 ms), reflecting the receptor’s unusual kinetics. ­

NMDAR subunit expression varies with respect to brain region, ontogenetic period, and activity. This gives rise to a family of NMDARs that differ in sensitivity to endogenous modulators, permeation or block by divalent cations, channel kinetics, and interaction with intracellular proteins. From the functional viewpoint, it is now a major challenge to understand the distinct roles played by the individual NMDAR subunits and the various NMDAR subtypes.

NMDAR Channel Properties Depend on Subunit Composition

Glutamate activates an impressive repertoire of signals in different central pathways, reflecting the enormous variety of glutamate receptor subunits (including a variety of NMDAR, AMPAR, and kainate receptor subunits). Information on the mechanisms underlying NMDAR signaling diversity has come from a combination of functional, pharmacological, and molecular techniques. To date, three main families of NMDAR subunits have been identified—namely NR1, a family of NR2 subunits (NR2A, -2B, -2C, and -2D), and a pair of NR3 subunits (NR3A and NR3B) (1). Most native NMDARs appear to function only as heteromeric assemblies composed of two NR1 subunits and two NR2 subunits. As shown in Fig. 1A, glutamate binds to the S1 and S2 regions of NR2 while glycine binds to the S1 and S2 regions of NR1; binding occurs in the crevice between S1 and S2 from the same subunit, rather than between adjacent subunits (Fig. 1A). Individual subunits contain an extracellular N terminus (part of which forms S1), an intracellular C terminus, and an extracellular loop between M3 and M4 that constitutes S2 (Fig. 1A). The channel-lining domain is formed by a re-entrant pore loop (M2) that dips into the channel from the cytoplasmic side and is thought to form a narrow constriction located roughly half way across the transmembrane electric field; the channel gate appears to be near this constriction (Fig. 1A) (2). Selectivity of the NMDAR channel for Mg2+ block and Ca2+ permeability is dependent on a critical asparagine residue located within the re-entrant pore loop. This site is directly homologous to the Q/R (glutamine-arginine) site in non-NMDAR subunits, which controls the Ca2+ permeability of synaptic AMPARs and kainate receptors.

Fig. 1.

(A) Transmembrane architecture of an NMDA receptor. NR1 and NR2 subunits are composed of several conserved regions. The extracellular region includes the N-terminal domain (NTD) that can contain modulatory sites that bind Zn2+ or ifenprodil (dependent on the NR2 subunit type—see text). The S1 and S2 domains form the binding site for glutamate in NR2 (subunit shown in turquoise) and for glycine in NR1 (subunit in red). The channel-lining region (M2) is a re-entrant "pore loop" that enters the membrane from the intracellular surface. The ion channel is permeable to Na+, K+, and Ca2+. Extracellular Mg2+ causes voltage-sensitive block by binding deep within the pore. Residues in other M regions can also influence channel permeability. The C-terminal tail of NR1 and NR2 binds to synaptic kinases and structural proteins (gray). On the NR1 subunit, proteins particularly assemble with the C0 (α-actinin, calmodulin), C1 [Yotaio, Neurofilament-Long (NF-L), calmodulin], and C2 (PDZ-containing proteins) domains. The NR2 subunit contains intracellular binding sites for PDZ-containing proteins (SXV motif), α-actinin, Ca2+/calmodulin-dependent protein kinase II (CamKII), AP-2 (YEKL motif), as well as phosphorylation sites (Y, S) (28, 98, 99). (B) The type of NR2 subunits involved is critical in determining the receptor-channel properties. NMDARs form as tetramers composed of two copies of NR1 and two copies of NR2. In diheteromeric receptors, the single-channel conductance provides a "signature" for the type of NR2 subunits involved: NR2A and NR2B subunits (turquoise and green) are thought to give rise to channels with high-(40 to 50 pS) conductance states, whereas NR2C-(2235 pS) and NR2D- (~16 to 35 pS) containing NMDARs (orange and blue) are characterized by low-conductance openings. (C) Examples of single-channel recordings from a cerebellar Golgi cell patch. In an NR2D knockout animal (top two traces), the cell expresses only NR2B receptors. The lower two traces show records from a wild-type cell expressing both NR2B and NR2D subunits. The patch displays high- and low-conductance openings (lower two traces). These could arise from three possible receptor combinations (see cartoon: NR2B, green; NR2D, blue). Some 50-pS events arise from diheteromeric NR2B receptors (NR1/NR2B), and some from triheteromeric NR1/NR2B/NR2D receptors. Diheteromeric NR2D receptors activate low-conductance events that are clearly seen in the bottom trace [data from (20)].

There are eight functional isoforms of NR1 arising from a single gene (1). Combinations of three independent splice variations give rise to the eight isoforms. These are generated by the insertion of exon 5 in the N terminus, the deletion of exon 21 in the C terminus, and use of an alternative splice acceptor site in the C terminus (exon 22). When the latter is used, it produces an alternative C-terminal cassette.The type of NR1 splice variant in the receptor assembly is critical in influencing certain key features, such as inhibition by protons, potentiation by polyamines, and inhibition by Zn2+. Proton sensitivity appears to be directly determined by the presence of exon 5 in the N terminus. NMDARs containing NR1 isoforms that lack exon 5 are partially inhibited (~50%) at physiological pH (although channel conductance appears unchanged), whereas exon 5–containing isoforms are insensitive to pH changes. Because exon 5 forms a surface loop with structural similarities to polyamine, it has been proposed that this loop cassette may screen the proton sensor on variants that contain exon 5 (3). Consistent with this idea, these isoforms are neither potentiated by polyamines nor inhibited by Zn2+ (which appears to act on NR1 and has a binding site that shares structural determinants with the proton site), unlike isoforms that lack the cassette (4). Although there is compelling evidence for an influence of NR1 on the kinetic properties of the NMDAR (5), this critical issue remains largely unexplored. Furthermore, precisely which of the NR1 isoforms are involved in generating the NMDAR-EPSCs in the central nervous system (CNS) remains unclear.

The identity of the NR2 subunits is critical in determining many biophysical and pharmacological properties of the receptor—including its high affinity for glutamate, modulation by glycine, sensitivity to Mg2+, fractional Ca2+ current, and channel kinetics (6). Work on recombinant (7, 8) and native NMDARs (9, 10) has shown that the decay of the response (deactivation time) to a brief glutamate pulse is characteristic of the type of NR2 subunit present. In general, the diheteromeric NMDARs (NR1/NR2) exhibit deactivation times that span a ~50-fold range, with the following order (from fastest to slowest): NR2A < 2C = 2B << 2D (Fig. 2A). Thus, NR1/NR2D-containing receptors produce responses with decay time constants of about 4 to 5 s, which is exceptionally slow for ligand-gated receptors. If present at a synapse, these receptors would produce NMDAR-EPSCs with similarly slow decay times. On the other hand, NR2C-containing receptors display more rapid deactivation kinetics—similar to that of NR1/NR2B receptors, but slow relative to the NR2A-containing receptors that predominate at many mature synapses. Patch-clamp studies of native and recombinant NMDARs have demonstrated that two other functionally important properties—single-channel conductance and block by Mg2+ ions—also depend on subunit composition (1116). As depicted in Fig. 1B, the frequently described 50-pS openings that display a high sensitivity to Mg2+ block are associated with NR2A- or NR2B-containing receptors, whereas low-conductance (~35 and 18 pS) openings, with a low sensitivity to Mg2+, arise from NR2C- and NR2D-containing receptors (7, 12, 15).

Fig. 2.

(A) Whole-cell responses from recombinant NMDARs illustrating the different deactivation kinetics of the various NR2 subunits (NR2A, turquoise; NR2B, green; NR2C: orange; NR2D: blue). The time constants of deactivation in response to a 1-ms pulse of 1 mM glutamate are roughly as follows: NR2A, 100 ms; NR2B, 250 ms; NR2C; 250 mS; NR2D, 4 s. Currents were activated by a 1-ms pulse of 1 mM glutamate [data from (8)]. (B) Different inputs within the same neuron can express different NMDAR subtypes. Evoked cortical-cortical and callosal-cortical EPSPs recorded from a cortical pyramidal cell arise from distinct NMDAR subunit combinations. Application of the NR2B antagonist, ifenprodil, partially blocked the cortical input, without affecting the callosal synapses, consistent with the presence of the NR2B subunit at cortical inputs [data from (50)]. (C) Spontaneous release of a single transmitter vesicle activates NMDARs located in the central part of the synapse. Release of multiple vesicles increases transmitter concentration in the cleft, activating perisynaptic NMDARs. As shown in the figure, experimental results suggest that the subtype of receptors located close to release sites may be different from those located more peripherally (55, 100).

Although NR3 is not an essential ingredient of most native NMDARs, there is evidence that NR3A can act as a regulatory subunit in some regions during early development. When combined with NR1/NR2, NR3 influences surface expression and reduces both Ca2+ permeability and current flow through the channel (17). Surprisingly, when NR3A or NR3B is coexpressed with NR1 alone, it can form a type of excitatory (cation permeable) glycine receptor that is impermeable to Ca2+, resistant to Mg2+ block, and unaffected by glutamate, NMDA, or NMDAR antagonists. On the other hand, it is selectively activated by glycine (18).

A number of drugs can distinguish between certain of the NMDAR subtypes. The best-characterized of these is probably ifenprodil (19) and its analogs. Ifenprodil selectively blocks NR2B diheteromeric receptors with a concentration giving 50% inhibition (IC50) that is roughly 400-fold lower than that for other NMDAR subtypes. This difference in sensitivity has been used to distinguish between NR2B- and NR2A-containing NMDARs in the synaptic and extrasynaptic membrane (10). Unfortunately, little is known about the action of subunit-selective drugs on NMDARs containing more than one type of NR2 subunit (triheteromeric receptors) (20). This is currently a severe limitation when attempting to use pharmacological approaches to investigate whether expression of NR2B receptors is reduced (for example, during development). Based on ifenprodil sensitivity alone, a reduction in NR2B receptors cannot be readily distinguished from a masking of NR2B by its increased coassembly with other subunits. A number of drugs act selectively on NR2A (10, 21, 22) or NR2C- or 2D-containing receptors (23), although so far these have not been widely examined. The recently described NR2A antagonist (NVP-AAM077) could prove of particular value (22) (see below).

How Many NMDAR Subtypes Are Present in the CNS?

One of the main reasons for the large number of potential NMDAR subtypes is that NR1, and most of the NR2 family, exist as various isoforms. This is tempered by the fact that different neurons express distinct complements of subunits. Although there is evidence for the coexistence of NR2B and NR2A (in cerebellar granule cells and hippocampal cells), NR2A and NR2C (in cerebellar granule cells), and NR2B and NR2D (in cerebellar Golgi and stellate cells), it remains to be seen whether other combinations, such as NR2B/NR2C, NR2A/NR2D, or NR2C/NR2D, coassemble in central neurons.

The structural arrangement of subunits within the receptor assembly places further constraints on the number of possible subunit assemblies. NMDAR subunits are thought to form tetrameric receptors (24), assembling as a pair of dimers formed from two glycine-binding NR1s, and two glutamate-binding NR2s (25, 26). It is generally agreed that the NR2 dimer can be composed of identical or different NR2 subunits, giving rise, respectively, to diheteromeric (e.g., NR1/NR2B) or triheteromeric (e.g., NR1/NR2B/NR2D) assemblies. Patch-clamp studies have clearly demonstrated that the presence of these various native di- and triheteromeric assemblies may have functional consequences, because they display distinct properties (Fig. 1C) (9, 20).

How many functional NMDAR subtypes are present in central neurons? Even in the simplest situation, the native NMDARs could arise from a possible eight NR1 isoforms. Any of these could, in principle, coassemble with one of four different homomeric NR2 dimers or one of six heteromeric NR2 dimers. This would seem to allow for the presence of at least 80 NMDAR subtypes in the CNS. These may not all be functionally distinct, or transported to cell regions where they are activated by glutamate. If we take into account the possibility that different splice variants of NR1 and NR2 subunits may, in some instances, influence channel properties and that NR3, when present, may further modify the receptor, the number of potential NMDAR subtypes is large. To date, more than a dozen functionally distinct NMDAR-channel subtypes have been described.

Trafficking and Surface Expression of the NMDARs

Whether particular NMDAR subunits contribute to the synaptic population and influence the NMDAR-EPSC depends on their interaction with intracellular molecules involved in assembly, transport, targeting, and anchoring of receptors (27, 28). NR1 and NR2 play distinct roles in the processes of intracellular trafficking and surface expression of receptors. The NR1/NR2 complex can assemble before leaving the endoplasmic reticulum (ER), although specific individual subunits can be exported from the ER on their own. The NR1 subunit, in particular, appears to play a vital role in the release of the NMDAR from the ER. Thus, a conditional NR1 knockout leads to retention of NR2 within the ER and a marked reduction of NR2A and NR2B in the dendrites (29). Furthermore, different splice variants of NR1 display selective targeting to different regions of the cell surface (soma versus distal dendrites); in addition, the overall activity of a neuron can alter the surface expression of NMDARs by modifying their egress from the ER (30). It is of note that although NR2A and NR2B subunits are apparently unable to reach the cell surface unless coassembled with NR1, most splice variants of NR1 can reach the cell surface in the absence of NR2 (3133).

In contrast, NR2 appears to play a dominant role in later stages of NMDAR delivery to the cell membrane (including the synaptic membrane). Specific interactions between various dendritic and postsynaptic proteins and NR2 are vital for dendritic transport, synaptic localization, and trafficking of receptors (28). These processes are controlled, primarily, by interaction between the C-terminal tail of NR2 and PDZ domains of intracellular scaffolding proteins and kinases. One family of putative scaffolding proteins, the membrane-associated guanylate kinases (MAGUKs, which include PSD-93, PSD-95, SAP-97, SAP-102, and chapsyn-110), contains PDZ domains responsible for binding and stabilizing NMDARs in the plasma membrane (27). Although it seems likely that the MAGUKs are involved in the trafficking and synaptic localization of NMDARs, the details of these roles have yet to be fully defined. In this respect, it is of interest that genetic manipulations of individual MAGUKs (such as PSD-95) often have minimal effects on the NMDAR EPSC (34). One possible explanation of this observation may be that other proteins, within the PSD, provide functional redundancy for the missing protein.

Recent research has identified specificity in the interaction between MAGUKs and NR2 subunits. Following the production of NR2B within the ER, delivery of the receptor to the cell surface involves an interaction between SAP-102 and the PDZ-binding domain of the exocyst protein Sec8 (35). The exocyst complex is an intracellular component thought to be involved in directing membrane vesicles to sites of fusion in the plasma membrane. The interaction of these exocyst proteins, including SAP-102 and Sec8, promotes the membrane delivery of NR2B-containing NMDARs (35). In contrast, PSD-95 binds to the C-terminal tail of NR2A, increasing its expression at the synapse while at the same time depressing synaptic expression of NR2B (36). During development, the overall expression of PSD-95 is increased, favoring an increase in NR2A and a decrease in NR2B at the synapse (37). In this respect, it is notable that the initiation of synaptic activity during development of cortical neurons leads to a rapid increase in the proportion of NR2A receptor subunits bound to PSD-95 (38). Surprisingly, transgenic mice in which PSD-95 has been truncated have enhanced LTP without any apparent change in synaptic NMDAR currents (34).

NMDAR interactions with intracellular scaffolding proteins have been further investigated by using mutant mice with truncated NR2 C-terminal tails. Although NR2B-truncated mice do not survive, truncation of the NR2A subunit is not lethal but causes a decrease in NR2A expression and reduces LTP at the CA3-CA1 synapse (39). Additional studies have demonstrated that truncation of both NR2A and NR2C causes a decrease in the open-channel probability of the NMDAR component of the EPSCs and a decrease in LTP at the cerebellar mossy fiber–granule cell synapse (40).

The subunit ratio at the synapse can be modified not only by receptor delivery, but also by receptor internalization. Activation of mGluRs (metabotropic glutamate receptors) can lead to the removal of both NMDARs and AMPARs from the cell surface (41). Furthermore, internalization of NR2B subunits can be initiated by the interaction of a YEKL motif in its C terminus with components of the endocytic machinery (42). Glycine binding to the NMDA receptor, without receptor activation, will "prime" the internalization of the receptor, decreasing NMDAR-mediated whole-cell responses (43). Thus, multiple distinct pathways exist for coordinating subunit trafficking at immature and mature central synapses.

Distribution of NMDAR Subtypes Within Individual Neurons

In the cerebellum, cortex, and hippocampus, the pattern of NR2 subunit distribution changes with age. Functionally, this is often manifested as a change in NMDAR-EPSC decay time, reflecting a switch in the composition of postsynaptic receptors (13, 44). Depending on the subunits involved, this can also result in a change in NMDAR sensitivity to Mg2+ (45). However, although a developmental speeding up of the NMDAR-EPSC occurs quite widely during maturation, this is not universal. For example, the initial acceleration of the NMDAR component of the EPSC in cerebellar granule cells (5, 13) gives way to a slowing of the current at the mature synapse when the initial increase in NR2A is overtaken by NR2C expression (45). However, in general, any early speeding up of the NMDAR-EPSC appears to result from an increase in NR2A, with accompanying reduction or loss of synaptic NR2B (5, 13, 4548).

In keeping with the view that NMDAR subunit composition is tailored to meet the needs of a particular pathway, differences have been noted in the NMDAR subtypes present at different inputs onto an individual cell (49, 50). In layer 5 pyramidal cells in the neocortex, the NMDAR-EPSCs arising from intracortical synaptic inputs are mediated primarily by NR2B-containing receptors, whereas those arising from callosal inputs to the same cells are carried by NR2A (Fig. 2B) (50). A similar dichotomy exists in CA3 pyramidal neurons of the adult hippocampus (49). It has been proposed that the biophysical properties of NR2B-containing receptors (slow decay) makes them more efficient at excitatory postsynaptic potential (EPSP) summation, bringing the neurons more rapidly to firing threshold (50). In contrast, the fast NR2A-containing NMDARs may be more effective as coincidence detectors. The distribution of NMDAR subunits within a neuron can also be asymmetric between the left and right hemispheres. Transection of the commissural axons can alter the expression pattern of NMDARs in the left and right hippocampi (51). In dentate gyrus granule cells, which display a continuum of synapses expressing either Ca2+-impermeable or Ca2+-permeable AMPARs, the NMDARs contributing to the EPSC differ in kinetics, subunit composition, and open probability. At synapses with Ca2+-permeable AMPARs, NMDARs have a higher NR2B composition and the EPSC displays a larger NMDAR component relative to those at synapses with Ca2+-impermeable AMPARs (52). Further, the different NMDAR compositions appear to correlate with distinct roles in synaptic transmission and plasticity, with substantial and robust temporal summation of synaptic events occurring at synapses with higher NR2B composition but not at those with lower levels of NR2B (52). One would expect more temporal summation at synapses with higher NR2B levels, simply because of the slower decay kinetics of this receptor. This may also be true for NR2C (and NR2D) subunits, although other cellular factors also have an impact on synaptic transmission in cells expressing NR2C (45).

Differential expression of NMDAR subunits has also been seen between synaptic and extrasynaptic regions of the dendrite (5) and can influence the functional properties of the postsynaptic response. This is particularly apparent in cells that lack NMDARs in the subsynaptic membrane, but express them extrasynaptically. In cerebellar stellate cells (53) and retinal ganglion cells (54), synaptic currents are normally mediated only by AMPARs. An NMDA-EPSC becomes apparent only when neurotransmitter release from these inputs is facilitated. It has been proposed that this spatial separation of AMPARs and NMDARs could provide a mechanism for generating a rapid change in EPSC properties that depends on the level of synaptic activity (53). Furthermore, spontaneous EPSCs in dentate gyrus granule cells are mediated by NR2A-containing receptors, whereas both NR2A and NR2B underlie the evoked synaptic responses, which may arise from a mix of synaptic and extrasynaptic receptors (see Fig. 2C) (55). As well as contributing to the synaptic signal, extrasynaptic NMDARs provide a receptor pool that can potentially be rapidly transported to and from the postsynaptic membrane (56). It is also of note that the Ca2+ influx associated with activation of extrasynaptic NMDARs during repetitive synaptic stimulation can trigger Ca2+-activated K+ channels, resulting in inhibition (57).

The location of NMDARs, and hence the route of Ca2+ entry into the cell, appears to play a role in the differential activation of various kinases necessary for normal cellular function. Ca2+ influx through NMDARs plays an important part in neuronal development, synaptic plasticity (LTP and LTD), and cell survival. Furthermore, Ca2+ influx during excessive NMDAR activation, associated with traumatic or ischemic events, can cause excitotoxic cell death (58). Whereas complete blockade of NMDARs can be neuroprotective against such insults associated with excess glutamate release, it appears to be detrimental to the overall health of the cell (59). Although the subunit composition of the NMDARs responsible for modulation of the intracellular Ca2+-signaling pathways associated with excitotoxicity and cell survival is unclear, there is evidence that dendritic versus somatic NMDAR activation produces contrasting effects. Activation of somatic NMDARs appears to be injurious, whereas activation of synaptic receptors may be neuroprotective (60). A clue to the basis of this difference could be that the synaptic NMDARs are composed predominantly of NR1/NR2A or NR1/NR2A/NR2B receptors (ifenprodil insensitive), whereas somatic NMDARs are composed predominantly of NR1/NR2B (5, 61). The C terminus of NR2B interacts with Ras–guanine nucleotide–releasing factor 1 (RasGRF1), activating the extracellular signal–regulated kinase (ERK) pathway (62). Ca2+ influx and subsequent ERK-pathway activation can lead to changes in cell proliferation and survival and the modulation of neuronal plasticity (63, 64).

Which NR2 Subunits Are Involved in Synaptic Plasticity?

NR2 subunit expression is also regulated by neuronal activity and sensory experience (44). For example, in sensory pathways, the developmental shift from NR2B to NR2A can be postponed by sensory deprivation (65). A number of studies have addressed whether alteration in the biophysical or molecular properties of NMDARs—resulting from insertion of NR2A and loss of the NR2B—sets a limit on the duration of the "critical period" for activity and experience-dependent fine-tuning of certain circuits (6567). The NR2B-containing NMDAR appears to be a dominant form whose activity can initiate anatomical and functional plasticity—including LTP (68), fear conditioning [in the amygdala (69)], C-fiber–evoked central sensitization or "wind up" in pain pathways (70), and development of somatotopic maps (71). However, important details of this relation are currently under scrutiny, and more details are emerging.

Several studies have addressed the relation between NMDAR-EPSC properties (or subunit composition) and the developmental reduction in plasticity seen in sensory cortex. In barrel cortex, the ability to generate LTP diminishes before NMDAR-EPSCs have been accelerated by expression of NR2A (72). This suggests that a switch in NMDAR-EPSC kinetics is not crucial for the reduction in plasticity. Despite this observation, there is evidence that loss of LTP induction ties in with the developmental reduction in ifenprodil sensitivity, apparently signifying a reduction in synaptic NR2B-containing receptors. This paradoxical observation suggests that termination of the "critical period" coincides with expression of an NMDAR subtype with "NR2B-type" slow channel kinetics, but lacking in ifenprodil sensitivity. One possible explanation is that a population of triheteromeric (NR2A/NR2B/NR1) receptors is transiently expressed at the end of the critical period.

Other studies suggest that the entry of synapses into a "nonpermissive" state requires neither a change in NMDAR subunit composition, nor in NMDAR subtype in the barrel and visual cortex (73, 74). Furthermore, in knockout mice lacking NR2A, the end of the critical period for LTP induction still coincides with that found in wild-type mice (73), supporting a view that a change in molecular and biophysical properties of synaptic NMDARs is unlikely to be the sole trigger involved in terminating the critical period. However, because LTP can be blocked by ifenprodil in juvenile sensory cortex (73, 74), the NR2B subunit is apparently required for LTP induction in this brain region. It thus seems likely that the ending of the critical period may involve changes both in NR2 subunits and in other key molecular components.

In contrast to LTP at immature synapses in the sensory cortex, LTP induced at mature synapses in the adult perirhinal cortex depends on activation of NR2A-containing NMDARs, because it is suppressed by the selective NR2A antagonist NVP-AAM077 (75). In addition, LTD induced from the basal state requires NR2B-NMDARs and can be activated only in the presence of glutamate uptake blockers. This type of LTD appeared distinct from depotentiation LTD, which required the activation of NR2A-containing receptors (75). These findings suggest that transmitter spillover (enhanced in the presence of uptake blockers) onto extrasynaptic NR2B receptors may be required for the induction of LTD from the basal state.

Whether NR2A- and NR2B-containing NMDARs play special roles in the triggering of LTP and LTD at hippocampal CA1 synapses is a matter of considerable interest (76). Indeed, there appear to be fundamental differences in the subunit dependence of plasticity at hippocampal and sensory cortex synapses. The blockade of NR2A-containing receptors (with NVP-AAM077) prevents LTP induction at hippocampal synapses in young animals, without influencing LTD (22). The idea that activation of NR2A-containing NMDARs is an absolute requirement for LTP production at these synapses is consistent with experiments on genetically modified mice that lack NR2A or its C-terminal tail (77, 78). NR2B receptor blockers do not appear to affect LTP induction at hippocampal CA1 synapses, although such blockers can abolish LTD induced by low-frequency stimuli and pairing paradigms in young animals (22). It is, however, of note that ifenprodil has also been shown to enhance, rather than block, LTD at young hippocampal synapses (79). Furthermore, overexpression of NR2B produced a mouse with altered EPSC kinetics, enhanced LTP, and faster intitial learning in a number of behavioral tasks—but with an apparent lack of effect on hippocampal LTD (80). Although the evidence for a special role for NR2B subunits in triggering hippocampal LTD remains equivocal, a consensus is emerging for the importance of the NR2A subunit in LTP induction at mature hippocampal and cortical synapses (22, 75).

Cross Talk Between Synapses, and Activation of Presynaptic NMDARs

Their high affinity for glutamate makes NMDARs prone to activation through cross talk between synapses (81). Such cross talk can entail the activation of presynaptic receptors at excitatory or inhibitory synapses, to modify spontaneous and evoked transmitter release and alter the signaling process (82). It can also involve activation of extrasynaptic (perisynaptic or somatic) receptors (see above). Neuronal and glial glutamate transporters appear to be important in restricting the activation of extrasynaptic NMDARs, including those in the presynaptic terminal (8385). It has recently been estimated that in hippocampal CA1 pyramidal cells, up to 30% of NMDARs contributing to the EPSC arise from glutamate released from more than one synaptic release site (86). Surprisingly, such synaptic cross talk is abolished when NR2B-containing receptors are blocked. This has led to the proposal that NR2B-containing NMDARs may play a role in sensing global glutamate signals, whereas NR2A-containing NMDARs are normally involved only in direct synaptic transmission (86).

Presynaptic NMDARs appear to modulate transmitter release in the cerebellum, entorhinal cortex, neocortex, and spinal cord. In the entorhinal cortex and neocortex, these NMDARs are ifenprodil sensitive, indicating an involvement of NR2B (86, 88). In the cerebellum, ambient levels of glutamate appear sufficient to tonically activate presynaptic NMDARs on the terminals or axons of GABA-containing stellate and basket cells that provide inhibitory inputs to Purkinje cells (82). Retrograde activation of these receptors by glutamate produces a depolarization-induced potentiation of Purkinje cell inhibition (89).

Presynaptic NMDARs are also involved in LTD-type plasticity changes that affect excitatory transmitter release, indicating that presynaptic NMDA autoreceptors can detect and modify the release of glutamate. Presynaptic NMDARs appear to play a role in cerebellar LTD (90), whereas timing-dependent LTD in the neocortex requires simultaneous activation of presynaptic NMDARs and CB1 cannabinoid receptors (88). As described below, the activation of NMDARs on central terminals of spinal cord primary afferent fibers interferes with the propagation of the nerve-terminal impulse, increasing the latency of glutamate release (91). However, thus far, limited information is available about the NMDAR subtypes that are present presynatically.

NMDARs in Pain

NMDAR antagonists are known to be highly effective at suppressing the increase in excitability ("wind-up") of spinal cord neurons that follows stimulation of nociceptive fibers (70). There has been considerable interest in the possibility that NMDAR subtype–specific blockers could serve as useful analgesics if specific NMDAR subtypes are localized in particular afferent pathways. Indeed, NR2B-selective blockers are well known for their antinociceptive effects (92). There is evidence that the dorsal horn cells possess functional extrasynaptic NR2B- and NR2D-containing NMDARs (15, 91, 93). However, the NMDAR component of the dorsal horn EPSC does not appear to be mediated by either of these receptor subtypes. Rather, NR2A-containing (ifenprodil insensitive) NMDARs are activated during synaptic transmission, indicating that these receptors may participate in the conduction of acute noxious signals. In addition, immunocytochemical and electron microscopic evidence indicates that presynaptic NMDARs reside both on central and peripheral primary afferent terminals (94). These may also be involved in the sensitization process (95). It is suggested that presynaptic NMDARs in spinal cord C-fiber terminals trigger the release of substance P and BDNF (brain-derived neurotrophic factor) from a subset of nociceptive terminals. Recent evidence indicates that functional NMDARs are indeed present on (or near) the C-fiber central terminals, and that their activation inhibits glutamate release, affecting transmission of sensory input to the spinal cord (91). Based on their possible low sensitivity to extracellular Mg2+, it has been proposed that these presynaptic receptors may be NR2D or NR2A/B-containing assemblies (91).

How then do NR2B-selective blockers produce their well-known antinociceptive action (92)? It has been suggested that extracellular glutamate accumulation, following stimulation of nociceptive afferent fibers, may be sufficient to activate extrasynaptic NR2B receptors in spinal cord cells (93). As well as providing a role for extrasynaptic NR2B receptors in the conduction of chronic pain, this may account for the analgesic action of NR2B blockers. It is also of note that peripheral inflammation can result in a change in the Mg2+ sensitivity of NMDARs in dorsal horn neurons. Rather than representing a change in NMDAR subunit composition, however, this effect is thought to be mediated by PKC (protein kinase C) (96); it is suggested that Mg2+ blockade of NMDAR channels can be altered by elevation of PKC inside these cells (96). The growing number of reports that demonstrate the analgesic effects of NR2B-selective blockers supports the view that this subunit is important in mediating pain. However, the debate continues as to whether the spinal cord is the primary site of action of these blockers. It is therefore of particular interest that overexpression of forebrain NR2B gives rise to increased inflammatory pain, leading to the suggestion that forebrain-selective NMDAR blockers may be useful for treatment of persistent pain (96).


NMDARs represent a unique family of central synaptic receptors, with members that differ widely in their biophysical and pharmacological properties. Although much is now known about the expression, trafficking, and localization of NMDARs, the specific contribution made by particular subunits (or subunit combinations) remains enigmatic at many synapses. However, recent advances have provided a clearer understanding of how the expression of specific NMDAR subunits is tailored to match the needs of plasticity, neural development, and pain perception. Not only has this improved our insight into the role of NMDAR subunits in particular pathways and processes, it also offers possibilities for more selective targeting and treatment of certain central disorders.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
  44. 44.
  45. 45.
  46. 46.
  47. 47.
  48. 48.
  49. 49.
  50. 50.
  51. 51.
  52. 52.
  53. 53.
  54. 54.
  55. 55.
  56. 56.
  57. 57.
  58. 58.
  59. 59.
  60. 60.
  61. 61.
  62. 62.
  63. 63.
  64. 64.
  65. 65.
  66. 66.
  67. 67.
  68. 68.
  69. 69.
  70. 70.
  71. 71.
  72. 72.
  73. 73.
  74. 74.
  75. 75.
  76. 76.
  77. 77.
  78. 78.
  79. 79.
  80. 80.
  81. 81.
  82. 82.
  83. 83.
  84. 84.
  85. 85.
  86. 86.
  87. 87.
  88. 88.
  89. 89.
  90. 90.
  91. 91.
  92. 92.
  93. 93.
  94. 94.
  95. 95.
  96. 96.
  97. 97.
  98. 98.
  99. 99.
  100. 100.
  101. 101.
View Abstract

Navigate This Article