Auxiliary Subunits Assist AMPA-Type Glutamate Receptors

See allHide authors and affiliations

Science  03 Mar 2006:
Vol. 311, Issue 5765, pp. 1253-1256
DOI: 10.1126/science.1123339


Glutamate, the major excitatory neurotransmitter in the brain, acts primarily on two types of ionotropic receptors: α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors and N-methyl-d-aspartate (NMDA) receptors. Work over the past decade indicates that regulated changes in the number of synaptic AMPA receptors may serve as a mechanism for information storage. Recent studies demonstrate that a family of small transmembrane AMPA receptor regulatory proteins (TARPs) controls both AMPA receptor trafficking and channel gating. TARPs provide the first example of auxiliary subunits of ionotropic receptors. Here we review the pivotal role that TARPs play in the life cycle of AMPA receptors.

“Life is all memory, except for the one present moment that goes by you so quick you hardly catch it going.” This statement by Mrs. Goforth in The Milk Train Doesn't Stop Here Anymore, by Tennessee Williams, succinctly addresses what the brain does. We are little more than the compilation of our memories, and these memories make each of us unique. Thus, understanding how the brain acquires and stores information is one of the foremost challenges in neurobiology.

For more than a century, activity-dependent changes in synaptic strength have been postulated as critical for learning and memory. The discovery of long-term potentiation (LTP) of excitatory synapses in the hippocampus (1), a brain structure essential for certain forms of memory, provided the first decisive evidence; LTP remains intensively studied. Excitatory synapses release glutamate onto two types of ionotropic receptors, AMPA receptors (AMPARs) and NMDA receptors (NMDARs). Whereas at least two mechanistically distinct forms of LTP exist (2), the most widespread form requires the activation of NMDARs, augmentation of postsynaptic calcium, and the activation of the calcium/calmodulin dependent kinase II (CaMKII) (37).

The mechanisms underlying the changes that occur during LTP have been difficult to define (8). The discovery of silent synapses and the evidence that LTP unsilences these synapses (9, 10) have convinced most researchers that LTP involves the activity-dependent rapid recruitment of synaptic AMPARs. Direct support comes from physiologically tagged AMPAR protein subunits (1113) and experimentally uncaging glutamate onto single spines (14, 15).

AMPARs are heterotetramers comprising combinations of glutamate receptors 1 to 4 (GluR1–4) subunits. In hippocampal pyramidal cells, AMPARs primarily comprise either GluR1/2 or GluR2/3. Synaptic trafficking of AMPARs depends on their subunit composition; GluR2/3 receptors constitutively cycle into and out of the synapse, whereas the trafficking of GluR1/2 receptors require activity (12).

Stargazer Mice

To understand receptor trafficking, investigators have defined interactions of the cytoplasmic tails of GluR1–4 with cytosolic proteins such as GRIP/ABP (glutamate receptor–interacting protein/AMPA receptor–binding protein), PICK1 (protein interacting with C kinase), and NSF (N-ethylmaleimide–sensitive factor) (12, 1622). Another approach has been the screening of mutant mice with well-defined motor defects. The stargazer mouse is both ataxic and epileptic. This mouse selectively lacks functional AMPARs in cerebellar granule cells (Fig. 1, A and B) (23, 24). The mutated protein, stargazin (also known as γ-2), is a small tetraspanning membrane protein with some homology to a calcium channel subunit γ-1 (25). Transfecting stargazin into cultured cerebellar granule neurons from stargazer mice restores both synaptic and extrasynaptic AMPAR responses (Fig. 1, C to E) (26). Stargazin shares homology with a large family of proteins (27, 28), and a subset of four (γ-2, γ-3, γ-4, and γ-8) can also traffic AMPARs (29). These four transmembrane AMPAR regulatory proteins (TARPs) (29) are differentially expressed throughout the brain.

Fig. 1.

Cell surface and synaptic AMPAR responses in stargazer (stg/stg) cerebellar granule cells require stargazin and stargazin C-terminal PDZ binding sites, respectively. (A) EPSCs elicited by mossy-fiber (MF) stimulation in granule cells from the wild-type (+/+) and stg mutant (stg/stg) mice at holding potentials of +40 mV (top panels), which records primarily the NMDA EPSC, and –70 mV (bottom panels), which records primarily the AMPA EPSC. Each trace is a single-sweep record, and several traces are superimposed for each record. (B) Current-voltage (I-V) relationships of MF-EPSCs from wild-type and stg mutant mice measured at the peak (open circle) and 50 ms after (closed circle) the stimulus. The EPSC amplitudes were normalized to the mean value at –40 mV in each experimental condition. Each data point and attached error bar represent mean and SEM. (A) and (B) are from (24). (C to F) Examples of glutamate-evoked (100 μM) whole-cell currents recorded at –80 mV, which evokes primarily AMPAR responses, and at +60 mV, which evokes primarily NMDA responses. The perfusion solution contained tetrodotoxin (1 μM) and glycine (20 μM). At –80 mV, glutamate evoked an inward current in a +/stg neuron (C), which is absent in stg/stg neurons (D). Stargazin rescues AMPAR responses and synaptic AMPAR responses in stg/stg granule cells (E). StargazinΔC rescues AMPAR responses but not synaptic AMPAR responses in stg/stg granule cells (F). Calibration bars are, from left to right, 25, 50, 25, and 25 pA (y axes), and 1 s (x axis) (top) and 25, 50, 50, and 25 pA, and 1 s (bottom). The calibration is the same for all synaptic responses.

Stargazin Binds to AMPARs

The immunoprecipitation of brain extracts shows that TARPs robustly and uniquely interact with all AMPAR subunits (2931). Prolonged incubation of these immunoprecipitates with high concentrations of glutamate, which may occur during excitotoxicity, causes stargazin and AMPARs to dissociate (30).

Separation of cerebellar extracts by native polyacrylamide gel electrophoresis revealed two populations of AMPAR complexes, and stargazin comigrated with the larger complex (32). The higher molecular weight AMPAR complex was absent in extracts from stargazer cerebellum. This suggests that the higher weight AMPAR complex consists of the tetrameric receptor bound to stargazin, whereas the lower weight form represents apo-AMPARs. Structural analyses of purified AMPARs at ∼40 Å resolution show that TARPs contribute to the density representing the transmembrane region of the forebrain AMPAR complex (33). These biochemical studies establish that TARPs bind selectively and stoichiometrically to AMPARs. However, the number of TARPs in a tetrametric AMPAR complex remains uncertain. A number of cytosolic proteins have been reported to bind to the C-terminal tails of AMPAR subunits (6, 7), and it will be important to find conditions that preserve these interactions in brain extracts.

AMPAR Maturation Requires Stargazin

The amount of GluR2 protein in the cerebellum, which is prominently expressed in granule cells, is reduced by about 20% in the stargazer mouse. Membrane proteins undergo a regulated biosynthetic progression through the endoplasmic reticulum (ER) and Golgi. Glutamate receptors receive high mannose glycosylation in the ER and are later modified with more complex sugars in the Golgi. The GluR2 protein that remains in the stargazer cerebellum has an immature ER-type glycosylation, suggesting that stargazin traffics AMPARs early in the biosynthetic pathway (29). Further evidence that TARPs stabilize AMPAR proteins is that the deletion of stargazin induces an ER unfolded-protein response in cerebellar granule cells (34).

Stargazin Traffics AMPARs by Two Mechanisms

Why do stargazer cerebellar granule cells lack synaptic AMPARs? First, AMPAR translocation from an intracellular site to the cell surface requires stargazin (26). In the Xenopus oocyte system (3537), stargazin enhances the surface expression of all AMPAR subunit combinations. By contrast, stargazin does not traffic closely related kainate receptors (35). Second, the last four amino acids of stargazin bind to the PDZ domains of a number of synaptic scaffolding proteins, including PSD-95 (26). This PDZ interaction mediates synaptic targeting of surface receptors. Thus, transfecting a stargazin construct lacking the last four amino acids (stargazinΔC) into stargazer granule cells rescues surface, but not synaptic, AMPARs (Fig. 1F). Furthermore, transfecting this construct into wild-type granule cells reduces synaptic AMPAR responses, presumably because of a dominant negative effect (26).

Although synaptic transmission in the stargazer hippocampus is normal, stargazin transfections have dramatic effects in hippocampal neurons (38). The overexpression of stargazin increases the number of extrasynaptic AMPARs but has no effect on AMPAR-mediated synaptic transmission. By contrast, transfecting stagazinΔC selectively reduces synaptic AMPAR transmission. Overexpression of PSD-95 enhances AMPAR-mediated synaptic responses (3841). Because PSD-95 does not bind to AMPAR subunits, it seems that stargazin, which binds both to AMPARs and to PSD-95, mediates AMPAR enhancement by PSD-95 (38). These results suggest that the interaction of PSD-95 or related PDZ proteins with TARPs participates in the synaptic targeting of AMPARs in many neuronal types (Fig. 2).

Fig. 2.

Roles of stargazin-like TARPs in the trafficking of AMPARs. TARPs (blue) bind to AMPARs (red) early in the synthetic pathway (1) and are required for the trafficking of receptors to the surface (2). Interaction of the C-terminal PDZ binding site of TARPs with PSD-95 at the postsynaptic density (PSD) captures the surface AMPARs at the synapse (3).

Stargazin Determines Gating of Native AMPARs

Initial studies suggested that all stargazin effects involve receptor trafficking. However, recent studies combining biochemical and biophysical approaches revealed that the enhancement of AMPAR currents by stargazin exceeds the increase in number of surface receptors (36, 37, 42). Thus, stargazin also enhances the function of the AMPARs.

Stargazin increases the apparent affinity of AMPARs for glutamate (36, 37, 42). Stargazin causes a four- to sixfold increase of steady-state responses and slows the rate of desensitization and deactivation twofold (36, 37, 43). Furthermore, stargazin enhances the single-channel conductance of AMPAR (36). Analysis of neuronal AMPAR single channels reveals that they can open to four conductance levels. When AMPARs are heterologously expressed alone, most openings are to low-conductance levels, whereas coexpression with stargazin increases the prevalence of high-conductance openings. All of these effects on AMPAR properties would be expected to enhance AMPAR synaptic responses.

Stargazin also modulates AMPAR gating by pharmacological agents (36, 43). Kainate is a partial agonist when AMPARs are expressed alone, but kainate becomes a full agonist with stargazin. Structural studies of the isolated GluR ligand-binding domain indicate that agonists induce closure of the clamshell-shaped binding pocket and that this movement gates the channel. The extent of domain closure is greater for the full agonist glutamate than for the partial agonist kainate (44) or for other partial agonists (45). These findings predict that the degree of domain closure with glutamate, and especially with kainate, is enhanced by stargazin (Fig. 3).

Fig. 3.

A model for possible role of TARPs on AMPAR channel opening. AMPAR subunits consist of four domains: a large N-terminal domain (NTD), a ligand-binding pocket (S1-S2), transmembrane domains that form a channel pore (TMD), and a cytoplasmic domain. Upon glutamate binding, S1-S2 closes like a clamshell, which causes channel pore opening. TARPs bind to AMPAR and affect AMPAR channel opening either by inducing more closure of S1-S2–to–glutamate binding or more efficient coupling of domain closure to pore opening without any change in S1-S2 closure.

Are these findings applicable to native receptors? A systematic study of GluR subunit combinations that are likely to be present in cerebellar granule cells (46) failed to find a combination that showed the high-conductance openings to glutamate and kainate found in cerebellar granule cells (47, 48). By contrast, coexpression of GluR4, a prominent GluR subunit in cerebellar granule cells, and stargazin generated channels showing the neuronal-type high-conductance openings (36).

Are the size and kinetics of synaptic receptors governed by TARPs? This was addressed with chimeras of stargazin and γ-5, a related protein that does not affect AMPARs (36). The stargazin first extracellular loop modulates channel properties, whereas the C terminus mediates receptor trafficking. Thus, a stargazin chimera containing the first extracellular loop of γ-5 (Ex1) trafficked the receptors normally but had no effect on deactivation or desensitization, whereas a chimera lacking the stargazin C terminus did not traffic AMPAR but did modulate their channel properties. The expression of Ex1 in neurons should act as a dominant negative in terms of the biophysical properties of AMPARs. Indeed, in the presence of the Ex1 mutant, the amplitude of AMPAR excitatory postsynaptic currents (EPSCs) was reduced and the decay kinetics were increased. A rough estimate suggests that stargazin biophysical effects increase the charge transfer of synaptic AMPAR responses by 30%.

TARPs in the Hippocampus

To test whether TARPs may modulate AMPAR throughout the brain, the gene for TARP γ-8, which is highly expressed in the hippocampus, was deleted (49). The γ-8 knockout (KO) mice were born in a Mendelian ratio, and they did not differ from littermate controls in terms of weight and gross behavior. However, AMPAR proteins (GluR1 and GluR2) were reduced by 80 to 90% in the hippocampus of the γ-8 KO mice. Immunohistochemistry also showed a profound loss of GluR1 and GluR2 in the hippocampal dendrites, with some GluR1/2 protein remaining in neuronal soma. AMPAR-mediated synaptic transmission was reduced by 35%. The density of extrasynaptic AMPARs, measured electrophysiologically, was reduced by 90%. These observations resemble those found for the GluR1 KO mouse (50) and suggest that hippocampal pyramidal cells, when confronted with a limited number of AMPARs, sequester the remaining receptors to synapses. Finally, LTP was greatly reduced, but long-term depression (LTD) was normal in the γ-8 KO mouse.

A number of questions concerning the γ-8 KO mouse remain. For instance, what accounts for the persistent synaptic AMPARs? Are they “TARPless,” or are the other TARPs in hippocampal pyramidal cells playing a role? What is the basis for the defect in LTP? Does γ-8 play a direct mechanistic role in LTP, or is the defect due to the loss of extrasynaptic AMPARs that might supply synaptic AMPARs during LTP?

Stargazin and Synaptic Plasticity

Because protein kinases play a central role in LTP induction, TARP phosphorylation was explored. Phosphopeptide mapping revealed that the C-terminal tail of stargazin has nine phosphorylated serines (51). These serines are conserved in all TARPs. Both CaMKII and protein kinase C, which have been implicated in LTP, phosphorylate some of these serines. In addition, these serines are dephosphorylated by the phosphatases PP1 and PP2b, which have been implicated in LTD. To evaluate the role for these serines in LTP induction, we mutated them to alanine, thereby preventing stargazin phosphorylation. This stargazin mutant still traffics AMPARs; however, LTP could not be generated in neurons expressing this construct. As a converse approach, the serines were mutated to aspartic acid to mimic phosphorylation. This construct effectively delivered receptors to the membrane surface, but, in addition, it drove AMPARs to the synapse in an activity-independent manner. Furthermore, neurons expressing this construct did not generate LTD, presumably because stargazin was locked into a phospho-mimic state.


Remarkable progress has been made over the last decade in understanding the molecular machines responsible for trafficking AMPARs to excitatory synapses. This progress has been fueled, in large part, by the realization that synaptic plasticity is mainly mediated by changes in the number of synaptic AMPARs. Stargazin and related TARPs have emerged as primary AMPAR auxiliary subunits that control both AMPAR trafficking and channel gating.

A number of issues remain unresolved. Are TARPs required for all AMPAR trafficking in the central nervous system? In heterologous systems, some AMPAR trafficking occurs without TARPs. Does this also occur in neurons? Generating mutant mice lacking all four TARPs can address this. Might AMPARs have other essential subunits? Functional AMPARs in Caenorhabditis elegans require suppressor of lurcher 1 (SOL-1) protein, which is structurally unrelated to stargazin (52). Whether a SOL-1–like mechanism operates in mammalian neurons remains unclear.

The existence of four TARPs and four AMPAR subunits raises intriguing possibilities. Might TARP subtypes differ in their capacity to traffic AMPARs and to affect channel gating? This may indeed be the case (36, 43). In some neuronal types, AMPARs of different subunit compositions localize to different synapses in the same neuron (53). Might TARP subtypes expressed in these neurons differentially traffic the AMPARs? Do other iontropic receptors have auxiliary subunits? Because AMPAR trafficking plays a central role in plasticity, perhaps the TARPs evolved to assist in this specialized dynamic mechanism. It has been reported that the cytosolic scaffolding protein PSD-95 can alter the gating of NMDARs (54) and the desensitization of kainate receptors (55). Another critical question is whether TARP-dependent trafficking shows any GluR subunit specificity. A central challenge will be to determine what role TARPs might have, if any, in the subunit specific control of AMPAR trafficking.

Finally, this Review focused on stargazin and its homologs solely in control of AMPAR function. Do these proteins have other functions? Another possible role is the control of calcium channels, which was originally proposed for these proteins (25). It would be intriguing if TARPs could serve as auxiliary subunits for both voltage-gated channels and ionotropic channels. As is the case with any rapidly emerging field, there are many more questions than answers. We can undoubtedly expect many surprises over the next few years.

References and Notes

View Abstract

Navigate This Article