AMPA Receptor Trafficking: A Road Map for Synaptic Plasticity

  1. José A. Esteban
  1. Department of Pharmacology, University of Michigan Medical School, 1150 W Medical Center Dr., Ann Arbor, MI 48109

Abstract

Most excitatory transmission in the brain is mediated by α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)-type glutamate receptors (AMPA receptors). Therefore, the presence of these receptors at synapses has to be carefully regulated in order to ensure correct neuronal communication. Interestingly, AMPA receptors are not static components of synapses. On the contrary, they are continuously being delivered and removed in and out of synapses in response to neuronal activity. This dynamic behavior of AMPA receptors is an important mechanism to modify synaptic strength during brain development and also during experience-dependent plasticity. AMPA receptor trafficking involves an intricate network of protein-protein interactions that start with the biosynthesis of the receptors, continues with their transport along dendrites, and ends with their local insertion and removal from synapses. The molecular and cellular mechanisms that regulate each of these processes, and their importance for synaptic plasticity, are now starting to be unraveled.

Introduction

Most excitatory activity in the brain is mediated by two types of glutamate receptors: α-amino-3-hydroxy-5-methylisoxazole-4-proprionic acid (AMPA) and N-methyl-d-aspartate (NMDA) receptors. these two types of receptors have very different roles in synaptic function (1, 2). AMPA receptors (AMPARs) mediate most excitatory (depolarizing) currents in conditions of basal neuronal activity, and hence, they have a major influence in the strength of the synaptic response. NMDA receptors (NMDARs), on the other hand, remain silent at resting membrane potential (3), but they are crucial for the induction of specific forms of synaptic plasticity, such as long-term potentiation (LTP) (4) and long-term depression (LTD) (5).

Although AMPARs and NMDARs reside in the same synapses in most brain regions, they reach their synaptic targets through quite different programs. in the brain, soon after birth, most excitatory synapses in the hippocampus (68) and other brain regions (912) contain only NMDARs, whereas the prevalence of AMPARs increases gradually over the course of postnatal development. in fact, the delivery of AMPARs into synapses is a regulated process that depends on NMDAR activation and underlies some forms of synaptic plasticity in early postnatal development (13) and in mature neurons (1418).

Synaptic plasticity is thought to underlie higher cognitive functions, such as learning and memory (1923), and is also critical for neural development (2426). Thus, it is not surprising that alterations in synaptic plasticity have been implicated in the pathology of several neurological disorders, including schizophrenia (27), Alzheimer Disease (28), and epilepsy (29). Consequently, elucidating the mechanisms underlying synaptic plasticity, such as the regulation of AMPAR trafficking, will help us to understand the pathophysiology of multiple brain illnesses.

Most AMPARs are likely to be synthesized in the neuronal cell body, far away from their synaptic targets. Therefore, newly synthesized receptors have to accomplish a series of trafficking steps before being delivered into synapses. This review will summarize our current knowledge of the trafficking pathways that guide AMPARs along their long journey from the endoplasmic reticulum (ER) to their destination at excitatory synapses, with special emphasis in the regulatory steps that contribute to synaptic plasticity. Most of the experimental observations that are the basis for this review have been obtained from hippocampal principal neurons, although it is expected that most of the principles described here will be applicable to the regulation of AMPAR trafficking in multiple brain regions.

AMPAR Synthesis and Regulated Exit from the Endoplasmic Reticulum

AMPARs are hetero-oligomeric molecules, possibly tetramers (30), composed of different combinations of GluR1, GluR2, GluR3, and GluR4 subunits (1). In the mature hippocampus, most AMPARs are composed of GluR1–GluR2 or GluR2–GluR3 combinations (31), whereas GluR4-containing AMPARs are expressed mainly in early postnatal development (13). These oligomeric combinations are formed in the ER through mechanisms that are not well understood but seem to depend on interactions between the luminal, N-terminal domains of the subunits (32, 33). GluR1–GluR2 hetero-oligomers exit the ER rapidly, and traffic to the Golgi compartment where they become fully glycosylated (34). In contrast, GluR2–GluR3 hetero-oligomers take much longer to exit (i.e., are retained longer in) the ER. In fact, a fraction of the GluR2 subunits seems to reside stably at the ER, as an intracellular, immature pool (Figure 1). Interestingly, this GluR2 pool seems to be retained within the ER in an active manner that depends on the presence of a charged arginine residue (R607) at the channel pore region of the GluR2 subunit, which is acquired through RNA editing of the original sequence coding for glutamine (35). GluR1, GluR3, and GluR4 mRNAs are not edited at this position; therefore, these subunits contain a non-charged glutamine residue at the channel pore, and do not undergo retention at the ER. The retention protein that prevents immature GluR2 from exiting the ER is not known; however, a fraction of AMPARs associates with the ER chaperones BIP and calnexin (36), and GluR2 colocalizes extensively with bip in the ER (34). Therefore, it is possible that chaperones residing at the ER are related with the retention mechanism.

  Figure 1.
View larger version:
    Figure 1.

    Early trafficking of AMPA receptors from the ER to the Golgi complex. GluR1–GluR2 oligomers exit the ER rapidly, in a process that may be assisted by the interaction between GluR1 C-terminus and the PDZ protein SAP97. GluR2–GluR3 oligomers are transferred to Golgi more slowly, and this process requires the interaction between the GluR2 C-terminus and the PDZ protein PICK1. In addition, a population of GluR2 subunits remains retained at the ER, probably complexed with luminal chaperones.

    Additionally, export of AMPARs from the ER may require the interaction of the cytoplasmic, C-terminal domain of the AMPAR subunits with other proteins. the GluR2 C-terminus has a PDZ consensus motif (SVKI) that interacts with several PDZ domain-containing proteins, including the protein interacting with C Kinase-1 (PICK1) (3739), which is thought necessary for GluR2’s exit from the ER (34).

    The GluR1 C-terminus also contains a PDZ motif (ATGL), which interacts with SAP97 (Synapse Associated Protein 97) (40). In this case it was found that SAP97 interacts with immature GluR1 early in the secretory pathway, probably while the receptor is still in the ER (41). This study did not address whether SAP97–GluR1 interaction was necessary for this AMPAR subunit to exit the ER, but it has been shown that the SAP97-interacting region of GluR1 is necessary for the receptor to reach its synaptic targets (14).

    AMPAR Trafficking Along Cytoskeletal Tracks in Dendrites and Spines

    The majority of the mRNA coding for AMPARS is found at the neuronal cell body, and therefore, it is reasonable to speculate that most AMPARS are synthesized far away from their synaptic targets. the long-range dendritic transport of AMPARS is likely to depend on the microtubular cytoskeleton that runs along dendritic shafts. the transport of membrane organelles on microtubule tracks is an active process powered mainly by motor proteins of the kinesin and dynein superfamilies (42). Therefore, membrane compartments bearing AMPARs are likely to be recognized and transported by some of these motor proteins. The molecular mechanisms underlying these processes are just being uncovered.

    The PDZ domain-containing protein Glutamate Receptor Interacting Protein 1/AMPAR Binding Protein (GRIP1/ABP) interacts directly with the heavy chain of conventional kinesin, as revealed by yeast two-hybrid screening (43). GRIP1/ABP binds to the C-terminal PDZ motif of GluR2 and GluR3 (44, 45), and hence, may serve as the link between AMPARs and microtubular motor proteins. in fact, the ternary complex formed by GluR2, GRIP1, and kinesin can be immunoprecipitated from brain lysates, and expressed dominant-negative versions of kinesin reduce the presence of AMPAR at synapses (43). Liprin-α also associates with GluR2–GRIP1/ABP1 (46), and a liprin-α mutant (unable to bind GRIP1) disrupts the targeting of AMPAR to synapses (46). Interestingly, liprin-α also interacts with a kinesin family member, KIF1 (Figure 2), and AMPARS can be coimmunoprecipitated with KIF1 from brain lysates (47). Another member of the liprin-α–AMPAR–GRIP complex is GIT1, which is also involved in AMPAR trafficking (48). Therefore, it seems that the GRIP1–AMPAR complex can be transported along dendrites by more than one type of kinesin motor. It seems likely that additional links between AMPARs and microtubular motor proteins will be discovered in the future, possibly mediated by adaptor scaffolding molecules.

      Figure 2.
    View larger version:
      Figure 2.

      Trafficking of AMPARs along cytoskeletal tracks. The movement of AMPARs through the dendritic shaft along the microtubular cytoskeleton requires the interaction between the GluR2 C-terminus and the PDZ protein GRIP/ABP, which is connected to the motor protein KIF1 through liprin-α. The connection of the AMPARs with the actin cytoskeleton at dendritic spines may be mediated by protein 4.1N, which interacts with the actin–spectrin filaments and with GluR1 and GluR4 subunits.

      Most excitatory synapses in the adult brain occur on small dendritic protuberances called spines (49). Dendritic spines lack microtubular cytoskeletons, but they are rich in highly motile actin filaments (50), therefore, at some point, AMPAR-containing organelles, trafficking along microtubular tracks, must be transferred to the actin-based cytoskeleton for their final delivery into synapses (Figure 2). In fact, pharmacological depolymerization of actin filaments leads to the removal of AMPARs from dendritic spines (51) and from synapses (52). The molecular mechanisms that may mediate the actin-based movement of AMPARs are largely unknown; however, a possible adaptor between AMPARs and the actin cytoskeleton has been recently described: protein 4.1. The different members of the protein 4.1 family are known to link the spectrin–actin cytoskeleton to different membrane-associated proteins (53). In particular, the neuronal isoform 4.1N interacts directly with GluR1 (54) or GluR4 (55) through the juxtamembrane region of their cytoplasmic C-terminal tails. Importantly, the surface expression of these AMPAR subunits depends on their interaction with protein 4.1 in a heterologous cell expression system. Undoubtedly, further studies will be required to unravel what is likely to be a network of interactions mediating the transport of AMPARs along the actin cytoskeleton into synapses.

      Delivery of AMPARs Into Synapses: Two Distinct Pathways

      The last step in the long journey of AMPARs is their delivery into the specialized dendritic membrane that constitutes the postsynaptic terminal. Multiple studies have shown that AMPAR localization is a very dynamic process; AMPARs cycle in and out of synapses under a variety of situations related to synaptic plasticity and development (5658). Several excellent reviews on this topic have been recently published (1518). I will concentrate on the latest findings that have clarified some of the main principles governing AMPAR delivery into synapses.

      AMPARs can reach synapses by two distinct pathways, depending on their subunit composition. GluR2–GluR3 oligomers are continuously delivered into synapses in a manner largely independent from synaptic activity (59, 60), whereas GluR1–GluR2 (14, 61) and GluR4-containing receptors (13) are added into synapses in a manner dependent upon NMDAR activation. These subunit-specific rules for trafficking have led to a model in which GluR2–GluR3 receptors continuously cycle in and out of synapses such that the total number of AMPARs at synapses is preserved (constitutive pathway), whereas GluR1–GluR2 (and GluR4) receptors are added into synapses in an activity-dependent manner during plasticity (regulated pathway) (62) (Figure 3). According to this scenario, the constitutive pathway would serve to maintain synaptic strength despite protein turnover, and it would act in a relatively fast manner (half-time of minutes). the regulated pathway would act transiently upon plasticity induction, leading to the long-lasting enhancement of synaptic strength also known as long-term potentiation (LTP). LTP is one of the best-characterized paradigms of synaptic plasticity in the mammalian nervous system and is thought to arise, at least in part, from an increase in the number of AMPARs present at synapses. this conclusion is supported by a variety of electrophysiological and molecular biological studies on the hippocampus (6, 14, 60, 6370), visual cortex (71), somatosensory cortex (9), spinal cord (72), and frog optic tectum (10). This scenario seems to apply virtually to all cases where NMDAR-dependent LTP has been documented. nevertheless, this view is not free from controversy (73).

        Figure 3.
      View larger version:
        Figure 3.

        Constitutive and regulated trafficking of AMPARs at synapses. Left. AMPAR oligomers containing GluR1 or GluR4 subunits are added into synapses in an activity-dependent manner (regulated delivery) during long-term potentiation (LTP). GluR2–GluR3 oligomers are continuously cycling (constitutive pathway) in and out of synapses. Right. The activity-dependent (regulated) removal of AMPARs from synapses (LTD) is likely to affect all receptor populations.

        Consistent with AMPAR trafficking being dependent on subunit composition, different subunits interact through their cytoplasmic C-termini with different sets of anchoring and regulatory proteins. For instance, GluR2 and GluR3 C-termini bind GRIP/ABP (44, 45) and PICK1 (3739). The GluR2 C-terminus also binds to NSF (N-ethylmaleimide-sensitive fusion protein) (7476) and to the clathrin adaptor ap2 (77). The C-terminus of GluR1 binds SAP97 (Synapse-Associated Protein-97) (40), and the C-termini of GluR1 and GluR4 bind the cytoskeletal protein 4.1N (54, 55). Most of these interactions have been proposed to control AMPAR trafficking, or to lead to the assembly of signaling complexes (or a combination of both) (78, 79). The precise trafficking steps controlled by these interactions are not well characterized, but as mentioned before, some of these interaction might mediate early intracellular trafficking of the receptor [e.g., the GluR1–SAP97 interaction (41)], or dendritic transport and accumulation in synapses [e.g., the GluR2–GRIP/ABP interaction (46, 47, 80, 81)]. The internalization of GluR2-containing receptors is modulated by their interaction with PICK1 (39, 8284) and AP2 (77), whereas GluR2–NSF interaction seems to be important for the recycling of AMPARs back into synapses (77, 84).

        Despite this wealth of information, it is still unclear how the intracellular trafficking of AMPARs is accomplished, and specifically, how synaptic activity controls receptor insertion and/or removal.

        Removal of AMPARs From Synapses

        The controlled removal of AMPARs from synapses is as important as the delivery of new AMPARs for synaptic function and plasticity. Similar to the constitutive and regulated pathways for synaptic delivery, there are also constitutive and regulated pathways for the trafficking of AMPARs out of synapses. The continuous movement of receptors from the synaptic membrane into extrasynaptic compartments was originally hypothesized from the fast reduction of AMPAR-mediated transmission observed when interfering with the exocytic machinery (85), or more specifically, when the interaction between NSF and GluR2 is prevented (76, 85). These experiments have now been corroborated and extended with molecular studies establishing that the GluR2–GluR3 subpopulation of AMPARs does not reside stably at synapses, but it is continuously cycling in and out of synapses (59, 60), as mentioned above. Although the mechanisms for this cycling are unknown, NSF might play a critical role in this process by regulating the interactions between GluR2 and PICK1 (84).

        In addition to the continuous movement of AMPARs out of synapses, there is also ample evidence for the activity-dependent removal of receptors, which leads to long-lasting depression of synaptic strength (long-term depression, or LTD). the regulated removal of AMPARs from synapses during LTD has been described in hippocampus (77, 8588) and cerebellum (89, 90). In contrast with the subunit-specific pathways for receptor delivery, it is much less clear which AMPAR subpopulations are affected by this regulated pathway. Hippocampal neurons lacking both GluR2 and GluR3 subunits can undergo normal LTD (91), suggesting that at least GluR1-containing receptors are subject to regulated removal. On the other hand, a switch of GluR2 binding partners from GRIP1/ABP to PICK1 accompanies LTD (39, 82, 83), suggesting that the GluR2 subunit is critical for LTD. It is possible that, in contrast to LTP, the regulated removal of AMPARs during LTD affects all subpopulations of AMPARs (77) (Figure 3).

        Signaling Cascades Controlling AMPAR Synaptic Trafficking

        It is now well established that the opening of NMDARs with the concomitant entry of Ca2+ ions in the postsynaptic terminal triggers both the regulated addition and removal of synaptic AMPARs. Multiple signaling cascades are thought to be activated by this rise in postsynaptic calcium, and it is likely that complicated interactions between different signaling pathways determine whether the final output results in a net increase or decrease of synaptic AMPARs. I will mention here some of the signaling cascades that seem to be central to the regulated trafficking of AMPARs.

        The CaMKII and cAMP-dependent Protein Kinase (PKA) Pathways for the Synaptic Delivery of Ampars

        There is abundant molecular and electrophysiological evidence supporting a critical role for the small guanosine triphosphatase (GTPase) Ras, the Ca2+-calmodulin–dependent protein kinase II (CaMKII) and the mitogen-activated protein kinase (MAPK) in synaptic plasticity, and in particular, in LTP (92, 93). Recent findings have offered a clearer picture of how the activation of this signaling pathway leads to the addition of new AMPARs into synapses (94) (Figure 4). LTP is mediated by the activation of CaMKII, which in turn activates Ras by inhibiting a synapse-localized ras GTPase-activating protein (SynGAP) (95, 96). Active Ras would then lead to synaptic delivery of AMPARs via activation of its downstream effectors p42–44 MAPK (97) and/or phosphatidylinositol 3-kinase (PI3K) (98). Still, the mechanistic links between MAPK and PI3K activation and the insertion of AMPARs into synapses remain unclear.

          Figure 4.
        View larger version:
          Figure 4.

          Signaling cascades mediating the regulated delivery and removal of AMPARs at synapses. Left. Ca2+ entry upon NMDAR opening leads to CaMKII activation, which in turn inhibits a RasGAP by phosphorylation. The decrease in RasGAP activity results in the accumulation of active Ras (Ras-GTP), which triggers the p42 MAPK and phosphatidylinositol-3 kinase (PI3K) leading to AMPAR synaptic delivery (LTP). Right. NMDAR-dependent Ca2+ entry activates a Rap-GEF, leading to an increase in the levels of active Rap (Rap-GTP). Subsequently, Rap activates the p38 MAPK cascade leading to the removal of AMPARs from synapses (LTD).

          The PKA signaling pathway is also involved in the regulation of synaptic plasticity. In particular, phosphorylation of GluR1 by PKA is required for AMPAR synaptic delivery (61, 99), and also controls the recycling of receptors between the plasma membrane and endosomal compartments (58). Interestingly, the signaling cascades that control the delivery of AMPARS to synapses change during postnatal development (61). Thus, early in postnatal development of the hippocampus, the regulated delivery of AMPARs involves GluR4-containing receptors (13), and cAMP-dependent protein kinase– (PKA)-mediated phosphorylation of GluR4 is necessary and sufficient for triggering delivery (61). Later in postnatal development, the regulated delivery of AMPARs requires PKA phosphorylation of GluR1, but this event is no longer sufficient, and the activation of the above mentioned cascade (CaMKII–Ras–MAPK) is also required for delivery (61). Therefore, the number of signaling pathways that need to be activated for AMPAR delivery increases during development, in agreement with the empirical observation that synaptic plasticity is more difficult to trigger later in life (100).

          Protein Kinase C–Dependent Phosphorylation of GluR2 and AMPAR Removal from Synapses

          Historically, LTP was thought to be accompanied by phosphorylation of AMPAR, whereas LTD was thought to be associated with AMPAR dephosphorylation. this picture had emerged mainly from studies on GluR1 phosphorylation by CaMKII and PKA (99); however, the removal of AMPARs during ltd actually correlates with phosphorylation of GluR2 by protein kinase C (PKC) (83, 101, 102). The most accepted model for this regulated removal involves the preferential interaction of unphosphorylated GluR2 with the PDZ domain–containing protein GRIP1/ABP, which would favor the presence of the receptor at synapses. After phosphorylation by PKC, GluR2 would dissociate from GRIP1/ABP and bind PICK1, and this new interaction would retain GluR2 away from synapses [see (84) for a model that integrates NSF’s role in GluR2 recycling with its alternative binding to GRIP1/ABP and PICK1].

          The Rapp38 MAPK Pathway for the Removal of AMPARs

          The activation of p38 MAPK leads to long-term synaptic depression (103, 104). The elucidation of this signaling cascade moved one step upstream when Rap1, a small GTPase known to activate p38 MAPK (105), was shown to trigger the removal of GluR2-containing AMPARs from synapses (94). Therefore, according to this scenario, LTD is triggered by NMDAR opening and Ca2+ entry, leading to the activation of Rap1, possibly through a specific Ca2+ -dependent guanine exchange factor (106). Then, activated Rap1 stimulates p38 MAPK, resulting in the removal of synaptic AMPARs (Figure 4). However, the mechanisms that link p38 MAPK activation with AMPAR removal remain unknown.

          Conclusions and Future Directions

          The field of AMPAR trafficking has advanced at an incredible pace since its conception, a mere five years ago. From the initial observation that AMPARs are indeed mobile in hippocampal synapses, the number of neuronal systems that have been demonstrated to display plasticity because of AMPAR movement has increased steadily. in parallel, multiple AMPAR binding proteins have been characterized that control AMPAR transport and localization.

          We can envision two, somewhat divergent, directions where this rapidly moving field may take us. On one hand, the list of proteins potentially involved in AMPAR trafficking and synaptic plasticity has grown to a point where just proposing that another molecule affects these processes is not particularly meaningful (107). The field is now in a situation where defined mechanisms should be identified for each player proposed to control AMPAR movement. new studies will probably distinguish the initial steps in the dendritic transport of the receptor from the processes that orchestrate their dynamic behavior close to the synapse, where the regulated addition or removal of receptors is used to control synaptic strength. these investigations should also lead us to identify the cellular machinery that directly transports AMPARs, as well as the regulatory molecules that gate or modulate specific trafficking steps. this research direction will help us to integrate the problem of AMPAR targeting within the larger context of intracellular membrane transport and sorting. On the other hand, it needs to be established in a meaningful manner that the synaptic trafficking of AMPARs responds to specific brain functions. These studies will require more physiologically relevant approaches, where AMPAR trafficking can be correlated with a specific behavioral task or with a defined pattern of brain activity. In this sense, it is worth mentioning that there already reports showing that AMPAR delivery in the somatosensory cortex (108), as well as AMPAR phosphorylation and surface delivery in the visual cortex (109) can be controlled by direct sensorial experience in vivo. Although this kind of experiment is certainly challenging, it is likely that similar studies looking at other brain functions will follow.

          The two experimental directions mentioned above are not necessarily opposite, and they may actually prove to be complementary. For instance, the more we learn about the molecular details of receptor trafficking and synaptic plasticity, the better we will be able to pinpoint the molecular or cellular changes triggered by a particular neural function. In fact, given the pace at which this research field is moving, we may not be too far away from the long-sought goal of deciphering what memories are made of.

          References


          José A. Esteban, PhD, is an Assistant Professor at the University of Michigan Medical School. His research interests are focused on the molecular and cellular mechanisms that govern synaptic plasticity in the brain. E-mail: estebanj{at}umich.edu; fax (734) 763-4450.

          | Table of Contents