Trafficking and synaptic anchoring of ionotropic inhibitory neurotransmitter receptors

Authors

  • Matthias Kneussel,

    Corresponding author
    1. Zentrum für Molekulare Neurobiologie Hamburg, ZMNH, Universität Hamburg, Falkenried 94, D-20251, Germany
    Search for more papers by this author
  • Sven Loebrich

    1. Zentrum für Molekulare Neurobiologie Hamburg, ZMNH, Universität Hamburg, Falkenried 94, D-20251, Germany
    Search for more papers by this author
    • Department of Brain Cognitive Sciences, Picower Institute for Learning and Memory, Massachusetts Institute of Technology, Building 46, Cambridge, MA 02139, U.S.A.


(email matthias.kneussel@zmnh.uni-hamburg.de).

Abstract

Neurotransmitter receptors are subject to microtubule-based transport between intracellular organelles and the neuronal plasma membrane. Receptors that arrive at plasma membrane compartments diffuse laterally within the plane of the cellular surface. To achieve immobilization at their sites of action, cytoplasmic receptor residues bind to submembrane proteins, which are coupled to the underlying cytoskeleton by multiprotein scaffolds. GABAARs (γ-aminobutyric type A receptors) and GlyRs (glycine receptors) are the major inhibitory receptors in the central nervous system. At inhibitory postsynaptic sites, all GlyRs and the majority of GABAARs directly or indirectly couple to gephyrin, a multimeric PSD (postsynaptic density) component. In addition to cluster formations at axo-dendritic contacts, individual GABAAR subtypes also anchor and concentrate at extrasynaptic positions, either through association with gephyrin or direct interaction with the ERM (ezrin/radixin/moesin) family protein radixin. In addition to their role in diffusion trapping of surface receptors, scaffold components also undergo rapid exchange to/from and between postsynaptic specializations, leading to a dynamic equilibrium of receptor—scaffold complexes. Moreover, scaffold components serve as adaptor proteins that mediate specificity in intracellular transport complexes. In the present review, we discuss the dynamic delivery, stabilization and removal of inhibitory receptors at synaptic sites.

Abbreviations used:
ERM

ezrin/radixin/moesin

GABAAR

γ-aminobutyric type A receptor

GABACR

γ-aminobutyric type C receptor

GABARAP

GABAAR-associated protein

GFP

green fluorescent protein

GlyR

glycine receptor

GODZ

Golgi-specific DHHC-domain-containing zinc-finger protein

HAP1

Huntingtin-associated protein 1

MAP

microtubule-associated protein

MoCo

molybdenum cofactor

PIP2

phosphatidylinositol 4,5-bisphosphate

PSD

postsynaptic density

RAFT1

rapamycin and FKBP12 (FK506-binding protein 12) target 1

TGN

trans-Golgi network

VASP

vasodilator-stimulated phosphoprotein

Introduction

Efficient neurotransmission and synaptic plasticity require the precise interplay of various proteins at pre- and post-synaptic compartments, as well as in perisynaptic regions. At postsynaptic sites, a large number of submembrane proteins form a highly interconnected structure that is known as the PSD (postsynaptic density). PSDs were originally described as electron-dense thickenings at glutamatergic spine synapses. However, from a molecular point of view, they resemble protein networks at the inhibitory shaft synapse. This present review focuses on proteins within the inhibitory postsynaptic plasma membrane and inhibitory PSD, two functionally collaborating adjacent layers that are specialized for ion flux, signal transduction, structural functions and the turnover of molecules. In addition, we discuss the membrane specializations of GABAARs (γ-aminobutyric type A receptors) at extrasynaptic sites, which are known to be critically involved in tonic inhibition.

A principal observation in recent years is the fact that both membrane receptors and submembrane proteins are subject to fast and constant protein turnover, underlying synaptogenesis and synaptic plasticity. The interplay of receptor diffusion within the plane of the lipid bilayer and active transport of cargo molecules along cytoskeletal elements recruits molecules in a timeframe of seconds to minutes within and between individual synapses.

The principles of inhibitory postsynaptic function are discussed below, with special focus on subsynaptic receptor transport and receptor clustering mechanisms.

Inhibitory neurotransmitter receptor dynamics

Newly synthesized neurotransmitter receptor subunits are thought to assemble into functional receptors in the endoplasmic reticulum (Hawkins et al., 2004) from where they are directed to the Golgi apparatus. Upon arrival in the TGN (trans-Golgi network), receptors are sorted into specific vesicular transport units, which encode cargo identity and contribute to the directed delivery to certain subcellular destinations (Tang, 2001) (Figure 1). Cargo recruitment of transmembrane proteins between the TGN and the cellular surface includes active transport processes along cytoskeletal elements, which serve as tracks for motor-protein-mediated delivery and/or removal. Motor-protein complexes involved in transport of neurotransmitter receptors and/or associated proteins include kinesins (Miki et al., 2005) and cytoplasmic dynein (Vallee et al., 2004), both of which migrate along microtubules, as well as unconventional myosins (Bridgman, 2004) that use microfilaments as tracks. Since microtubules are distributed throughout neurons, they are thought to mainly serve as rails for long-distance transport. In contrast, due to the enrichment of F-actin microfilaments at the cellular cortex, the latter, in addition, represent good candidates for local transport within distinct microdomains. For instance, dendritic spines are rich in microfilaments, but contain few, if any, microtubules (Kaech et al., 2001), whereas dendrite shafts contain both microtubules and microfilaments.

Figure 1.

Neurotransmitter receptor and scaffold dynamics at postsynaptic membrane specializations

Intracellular receptors that cycle between organelles and the plasma membrane (blue ovals) can be subject to microtubule-dependent transport through motor-protein—cargo transport complexes. Postsynaptic scaffold proteins (red lines) are often used as motor-protein—cargo adaptors at intracellular locations. Endosomal receptors (blue) that have been internalized either recycle back to the cell surface or are degraded within the lysosome. Receptors within the plasma membrane (yellow and green ovals) are either subject to membrane diffusion (yellow ovals) or, upon trapping through scaffold proteins, anchor and cluster at postsynaptic sites (green ovals). The number of synaptic receptors available at a given time contributes to strengthen (+) or weaken (−) synaptic transmission. Scaffold material (red lines) is constantly added and removed from postsynaptic scaffolds (red arrows), and is also subject to intersynaptic exchange (red double arrow) on a timescale of minutes. This has been particularly shown for gephyrin at inhibitory synapses (Maas et al., 2006). Since the size of postsynaptic scaffolds is proportional to the number of binding sites available for receptor trapping, scaffold dynamics is also thought to contribute to the regulation of synaptic strength.

Synaptic delivery of receptors is encoded by the combined use of motors, adaptor proteins and vesicular cargo (Kneussel, 2005). Since a limited amount of motor proteins in neurons transport large amounts of cargo throughout the cell, adaptor proteins are thought to be involved in specificity of transport. For instance, KIF5 (a kinesin) acts in both axons and dendrites; however, the use of different cargo adaptors encodes the directionality of transport to one, but not the other, direction (Verhey et al., 2001; Setou et al., 2002; Hirokawa and Takemura, 2005). Motor-protein complexes involved in the transport of inhibitory neurotransmitter receptors include vesicular GlyRs (glycine receptors) that are coupled, via the scaffold adaptor gephyrin, to the dynein motor complex and are subject to retrograde transport from synaptic sites to intracellular destinations (Maas et al., 2006). For GABAAR transport, candidate interactions for anterograde recruitment of β2 subunits have been identified. This putative complex consists of the GABAAR β2 subunit, GRIF1 (GABAAR-interacting factor 1) and the kinesin motor KIF5C; however, functional evidence for cargo movement is currently lacking (Smith et al., 2006). Notably, from several neurotransmitter receptor transport complexes it has been revealed that the same proteins, which mediate adaptor functions in cargo recruitment, also represent postsynaptic scaffold components at PSDs (Setou et al., 2000, 2002; Maas et al., 2006). This observation led to the hypothesis that receptor and scaffold adaptors might enter the surface membrane together and, subsequently, be subject to membrane diffusion to finally co-integrate into postsynaptic scaffolds (Kneussel, 2005).

Plasma membrane insertion of receptors is followed by lateral receptor recruitment within the lipid bilayer (Borgdorff and Choquet, 2002; Dahan et al., 2003). Upon diffusion trapping through interactions with submembrane proteins at axo-dendritic contacts, receptors are concentrated into clusters (Kneussel and Betz, 2000; Moss and Smart, 2001), underlying the formation of functional synapses with high amounts of receptor polypeptides in apposition to presynaptic transmitter release sites. The switching between binding and releasing involved in protein interactions between membrane receptors and submembrane scaffold proteins also represents a potential system to rapidly regulate receptor number at synapses at a given time. Upon release from scaffolds, receptors are able to diffuse to extrasynaptic positions, but might remain at the neuronal surface and be available for further synapse incorporation (Figure 1). Other mechanisms of synaptic plasticity at the molecular level include the internalization of receptors via endocytosis (Kittler et al., 2000; Ashby et al., 2004; van Rijnsoever et al., 2005), followed by the subsequent decision to either degrade the protein in the lysosome or to recycle it back to the synapse (Ehlers, 2000; Bedford et al., 2001; Steiner et al., 2005) (Figure 1).

A number of proteins have been identified that bind to GABAAR subunits at intracellular sites. This includes GABARAP (GABAAR-associated protein) (Figure 2), which binds to GABAAR γ2 subunits (Wang et al., 1999), to gephyrin (Kneussel et al., 2000) and to GRIP1 (glutamate receptor-interacting protein) (Kittler et al., 2004a); however, GABARAP has not been identified at synaptic sites (Kneussel et al., 2000; Kittler et al., 2001) and is not critically important for the synaptic delivery of GABAARs to synapses (O'Sullivan et al., 2005). On the basis of the sequence similarity of GABARAP with proteins involved in autophagy processes, this interaction might be involved in degradation processes. Another intracellular GABAAR interacting partner is Plic1, a ubiquitin-like protein that binds to most, if not all, of the α and β subunits (Bedford et al., 2001), participates in receptor turnover processes and has been reported to stabilize intracellular receptors and to promote their accumulation in the plasma membrane. Also GODZ (Golgi-specific DHHC-domain-containing zinc-finger protein) (Figure 2), a palmitoyltransferase, represents a specific binding partner of intracellular GABAAR γ2 subunits (Keller et al., 2004) and might be important for receptor coupling to vesicular membranes and trafficking of receptors. In addition, the observation that GABAAR β subunits bind to HAP1 (Huntingtin-associated protein 1) (Figure 2) points to another intracellular system that participates in the regulation of GABAAR trafficking (Kittler et al., 2004b). HAP1 binds to kinesin motor proteins (McGuire et al., 2006), as well as to Huntingtin (Harjes and Wanker, 2003). Furthermore, both Huntingtin and HAP1 associate with cytoplasmic dynein motors (Engelender et al., 1997; Harjes and Wanker, 2003). Whether the GABAAR—HAP1 interactions directly contribute to motor-protein complexes or represent optional modulators of transport requires further investigation.

Figure 2.

PSD at inhibitory synapses

Glycinergic (left) and GABAergic (right) synapses share a number of factors as described. Gephyrin (red lines) represents a multimeric protein with self-interaction domains at both its N- and C-terminus. It directly interacts with the large intracellular loop of GlyR β subunits (green ovals); however, direct interaction between gephyrin and GABAARs (blue ovals) has never been reported. Gephyrin functionally interacts with GABAAR γ2 subunits, suggesting that intermediate, currently unknown proteins (white rectangles, ‘?’), link these two factors at postsynaptic sites. Interaction screens identified RAFT1 (orange diamond), which is a candidate as a regulator of local translation, and the GDP/GTP exchange factor collybistin (pink rectangles), which is thought to regulate actin dynamics via Rho-type GTPases, as gephyrin-binding partners. In contrast, the palmitoyltransferase GODZ (small purple oval) binds to GABAAR γ2 subunits; however, this interaction partner has not been described at GABAergic synapses and instead acts at intracellular sites. Cytoskeletal attachment of the submembrane gephyrin scaffold is achieved through interactions with profilin (brown circle) and Mena/VASP (turquoise rectangle), both of which bind to microfilaments (F-actin; black lines). At intracellular locations, gephyrin (red lines) represents a transport adaptor that links vesicular GlyRs (green ovals) with the retrograde microtubule-dependent dynein motor complex. Similarly, Grif1 (magenta diamond) links GABAAR β2 subunits to the anterograde motor KIF5C. Intracellular GABAARs (blue ovals) bind a number of factors thought to regulate receptor targeting and turnover. These include GABARAP (yellow circle), which in turn binds gephyrin, GRIP1 (grey rectangle), HAP1 (small green oval) and the ubiquitin-related protein Plic1 (black diamond). Note that due to the regulated binding of individual factors and the overall protein dynamics at postsynaptic specializations not all proteins may be present at all times.

Neurons use a combination of several parallel systems to alter the number of neurotransmitter receptors at postsynaptic sites underlying plasticity: (i) trapping and release of surface-membrane diffusing receptors (denoted as fast ‘horizontal’ recruitment), (ii) endocytosis-mediated internalization and recycling (denoted as fast ‘vertical’ recruitment) and (iii) delivery/removal of cargo between the cell surface and cellular compartments specialized for synthesis/degradation (denoted as slow ‘vertical’ recruitment). To elucidate the mechanisms that regulate the interplay between these individual processes will be a future challenge for the further understanding of synaptic plasticity.

Inhibitory postsynaptic scaffold dynamics

Multimeric assembly of scaffold proteins, which often harbour self-interaction domains for submembraneous network formation, provides binding sites for trapping of diffusing surface receptors. Considering that trapping does not require multistep activation to present the respective binding sites, the amount of scaffold material available for receptor trapping might be proportional to the number of synaptic receptors at a given time. Constant turnover of scaffold components through directed delivery and removal of individual units, which leads to growth and shrinkage of postsynaptic scaffold size, could provide an additional system to modulate receptor number at PSDs. Recent studies have revealed that the postsynaptic scaffold component gephyrin at inhibitory synapses is, indeed, the subject of constant delivery and removal to and from synaptic sites, as well as between individual active synapses over time (Maas et al., 2006). Consistent with representing particles in transit, a certain percentage of gephyrin puncta in neuronal dendrites do not co-localize with presynaptic terminal markers, but is mobile in time-lapse video microscopy assays (Figure 3) (Maas et al., 2006). Additionally, once located at synaptic sites, gephyrin scaffolds still display some kind of local mobility (Hanus et al., 2006), which has been referred to as ‘dancing behaviour’, a phenomenon that might account for constant growth and shrinkage reactions at different positions of the scaffold multimer.

Figure 3.

Gephyrin puncta in neuronal dendrites represent either postsynaptic scaffolds or en route transport complexes

(A) Immunocytochemical detection of endogenous gephyrin detects two populations of dendritic puncta with a significant size difference. Co-staining for the inhibitory presynaptic terminal marker VIAAT (vesicular inhibitory amino acid transporter) reveals that large clusters localize at synapses and therefore represent postsynaptic scaffold formations. In contrast, small clusters have no presynaptic counterpart. As the small particle population has been found to be frequently mobile in time-lapse video microscopy experiments, and to co-localize with molecular motors, it is most likely to represent transport complexes. Scale bar, 2 μm. (B) Relative particle size of immobile and mobile GFP—gephyrin puncta. (C) Relative particle size of synaptic and non-synaptic endogenous gephyrin puncta. Error bars represent the size variations of mobile or non-synaptic clusters. Reproduced from The Journal of Cell Biology, 2006, 172, 441–451. © 2006 The Rockefeller University Press.

The fact that scaffold material, such as gephyrin, can provide transport adaptor functions in microtubule-based transport complexes, does not exclude the possibility that gephyrin might be in contrast transported alone. Proteomics screening at PSDs of excitatory synapses has revealed that the number of scaffold components, which assemble into a three-dimensional lattice underneath the synaptic membrane (Sheng and Kim, 2000; Sola et al., 2004), could be approximately an order of magnitude higher than the number of membrane receptors that bind to these molecules (Peng et al., 2004).

Whether and to what extent scaffold components participate in synaptic plasticity remains to be functionally elucidated; however, initial experiments suggest that crosstalk occurs between synaptic transmission and the subcellular scaffold transport machinery. Upon blockade of GlyRs with their antagonist strychnine, the transport ratio of the postsynaptic GlyR-binding protein gephyrin is shifted towards retrograde transport away from synaptic sites and, furthermore, shows an increase in velocity, as compared with control conditions (Maas et al., 2006). Therefore the electrical state of individual neurons might activate signalling cascades that alter the transport of synaptic scaffold proteins, thereby generating a feedback situation for the regulation of synaptic receptor number.

Postsynaptic specialization at glycinergic synapses

This section focuses on protein interactions at the postsynaptic membrane and submembrane compartment of glycinergic synapses, which represent the main type of inhibitory synapses in spinal cord that are involved in reflex circuit inhibition. GlyRs are heteropentameric chloride channels, assembled out of α and β subunits (Betz, 1991), which display diffusion across the plane of the plasma membrane (Dahan et al., 2003) and constantly switch between a confined and diffuse state in the range of seconds (Meier et al., 2001). It has been demonstrated that the submembraneous scaffold protein gephyrin, which is widely expressed throughout the body (Prior et al., 1992), is essential for GlyR clustering at neuronal surface membranes in different neuronal tissues, including spinal cord (Kirsch et al., 1993; Feng et al., 1998), hippocampus (Levi et al., 2004) and retina (Fischer et al., 2000). Gephyrin co-expression with GlyR α and β subunits in HEK-293 cells leads to rearrangement of GlyR proteins into intracellular gephyrin aggregates (Kirsch et al., 1995). Vice versa, antisense depletion (Kirsch et al., 1993) and gene knockout in mice (Feng et al., 1998) dramatically interfere with GlyR clustering, leading to an almost complete loss of receptor accumulation at synaptic and extrasynaptic sites.

Originally, gephyrin was isolated as a co-purifying protein together with the GlyR (Schmitt et al., 1987); however, it was later found to represent a high-affinity GlyR-binding protein. Binding to gephyrin occurs via an 18-amino-acid minimal gephyrin-binding motif within the intracellular loop (transmembrane domains III—IV) of GlyR β subunits (Meyer et al., 1995; Kneussel et al., 1999a). For efficient trapping of GlyRs diffusing in the plasma membrane, gephyrin can be in different oligomeric states (Sola et al., 2004) and, finally, forms a hexagonal structure, which provides binding sites for GlyRs and a number of subsynaptic proteins involved in cytoskeletal attachment and subsynaptic signalling. The actin-binding proteins Mena/VASP (vasodilator-stimulated phosphoprotein) (Giesemann et al., 2003) and the profilins 1 and 2 (Mammoto et al., 1998; Giesemann et al., 2003) have been shown to form a complex together with gephyrin (Figure 2). As high amounts of F-actin microfilaments localize at the cellular cortex, these interactions are likely to anchor GlyR—gephyrin scaffolds at synaptic membrane positions. Earlier studies identified gephyrin as a factor that co-purifies with tubulin (Kirsch et al., 1991); however, whether the distance of microtubules to the postsynaptic membrane is close enough to account for gephyrin scaffold attachment remains a matter of debate. In any case, analysis of an intracellular GlyR—gephyrin transport complex that is actively recruited along microtubules via the motor-protein cytoplasmic dynein (Fuhrmann et al., 2002; Maas et al., 2006) is consistent with indirect gephyrin—tubulin interactions (Figure 2).

Gephyrin-binding partners involved in subsynaptic signalling at glycinergic synapses include the protein RAFT1 [rapamycin and FKBP12 (FK506-binding protein 12) target 1] (Sabatini et al., 1999), which mediates effects of the immunosuppressant rapamycin and is an important regulator of mRNA translation. How the interaction of gephyrin and RAFT1 allows for rapamycin signalling, and if gephyrin influences translation of certain mRNAs either locally at the inhibitory synapse or in the cell body, is presently unknown. In addition, gephyrin binds the GDP/GTP exchange factor collybistin (Kins et al., 2000; Harvey et al., 2004), which activates the Rho family GTPase Cdc42 (Xiang et al., 2006) and is a candidate factor for actin cytoskeleton rearrangement at synaptic sites (Figure 2). Collybistin has been suggested to promote gephyrin and GlyR clustering (Kins et al., 2000), therefore a possible scenario is that synaptic microfilament rearrangement is required to locate protein scaffolds at subsynaptic membrane positions.

Apart from its role in synaptic GlyR clustering and intracellular GlyR transport function, gephyrin participates in the biosynthesis of MoCo (molybdenum cofactor) (Stallmeyer et al., 1999), which is important in xanthine dehydrogenase, and sulfite and aldehyde oxidase enzymatic functions (Mendel and Hansch, 2002; Schwarz and Mendel, 2006). Gephyrin knockout mice are not viable and display functional loss of MoCo-dependent enzymes (Feng et al., 1998); however, vice versa, knockout mice of MOCS1, an essential MoCo biosynthesis protein that acts prior to gephyrin in the MoCo pathway and also causes functional loss of MoCo-dependent enzymes, display normal GlyR and gephyrin clusters in neuronal tissues (Lee et al., 2002). In addition, transgenic expression of the Arabidopsis thaliana protein Cnx-1, a protein that represents the gephyrin homologue in plants, partially restores sulfite oxidase activity in gephyrin-deficient mice; however, it does not rescue GlyR clustering (Grosskreutz et al., 2003). Together, these data point to independent functions of gephyrin in GlyR clustering and MoCo biosynthesis; however, they do not the exclude the possibility that MoCo-dependent activities participate in distinct pathways at synapses other than receptor clustering. Interestingly, a recently identified mutation in the MOCS1 gene causes hyperekplexia (Macaya et al., 2005), a disease that is also caused by mutations in GlyR α and β genes (Breitinger and Becker, 2002). Whether, and to what extent, the GlyR and MoCo pathways functionally overlap therefore remains an important question and certainly requires further investigations.

Postsynaptic specialization at GABAergic synapses

The major sites of fast synaptic inhibition in the brain are GABAergic synapses (Figure 2). GABAARs assemble as pentameric formations from different subunits, which belong to five different classes with each class consisting of several proteins that are derived from separate genes: α1–α6, β1–β3, γ1–γ3, δ, ε, π and ρ. Individual subunits are expressed in a spatio-temporal manner, narrowing down the many possible combinations for a functional GABAAR. The most common subunit combinations in brain have been reviewed by McKernan and Whiting (1996).

Despite their role in phasic inhibition at synapses, certain GABAAR subtypes mediate tonic inhibition (Kullmann et al., 2005), thereby critically adjusting neuronal excitability. Accordingly, up to approx. 50% of all GABAARs are located at extrasynaptic positions (Scotti and Reuter, 2001). This section reviews GABAARs and associated proteins at postsynaptic sites, whereas the following section discusses extrasynaptic GABAARs.

Evidence that GABAARs, similar to other neurotransmitter receptors, display cell surface dynamics comes from two studies on α1 and γ2 subunits (Jacob et al., 2005; Thomas et al., 2005). In the first study, an electrophysiological approach revealed fast synaptic GABAAR receptor recovery after irreversible use-dependent inhibition at synaptic sites (Thomas et al., 2005). In addition, GABAAR subunits fused to a pH-sensitive version of GFP (green fluorescent protein) (pHlourin) revealed that RNAi (RNA interference)-mediated down-regulation of the postsynaptic anchoring protein gephyrin increased the lateral mobility of GABAAR subunits (Jacob et al., 2005).

Despite different results that show that gephyrin is a major component at synaptic sites (Triller et al., 1987), biochemical evidence for direct interactions between any GABAAR subunit and gephyrin is lacking. These observations gave rise to the speculation that bridging factors might exist that link both proteins at synapses (Figure 2), since heterologous co-expression studies and genetic approaches in mice clearly revealed that GABAAR clustering functionally depends on gephyrin (Kirsch et al., 1995; Essrich et al., 1998; Kneussel et al., 1999b, 2001). In gephyrin-deficient mice (Feng et al., 1998), synaptic clustering of α2 and γ2 subunits were drastically reduced in hippocampus (Kneussel et al., 1999b), spinal cord (Kneussel et al., 2001) and retina (Fischer et al., 2000); however, other subunits were either only moderately affected or completely unaffected upon the loss of gephyrin. The extent of GABAAR cluster reduction differed between low-density cultured neurons (Kneussel et al., 1999b) and neuronal tissue sections (Kneussel et al., 2001), suggesting that electrical parameters in the neuronal network might contribute to the effects observed. In agreement with this theory, the investigation of high-density neuronal cultures revealed less severe effects on α2 and γ2 subunits derived from gephyrin null mice, which, however, displayed significantly reduced mIPSC (miniature inhibitory postsynaptic current) amplitudes (Levi et al., 2004).

Further evidence for a functional interplay between GABAARs and gephyrin comes from GABAAR γ2 subunit knockout mice (Essrich et al., 1998). In this study, gephyrin clusters were drastically reduced upon the loss of the synaptic GABAAR γ2 subunit, confirming that both proteins functionally associate and that the loss of receptors might lead to reduced gephyrin scaffold size at submembrane positions. Interestingly, application of GABAAR α2 and γ2 chimaeric constructs identified the fourth transmembrane segment of the GABAAR γ2 subunit as critical for receptor clustering (Alldred et al., 2005). This observation could indicate an intramembrane binding partner of GABAAR γ2 subunits, which might provide the link to gephyrin; however, which protein could be involved in such a reaction is currently unknown. Overexpression of GABAAR γ3 subunits in mice (Baer et al., 1999) partially restored the loss of GABAAR γ2 subunits; however, due to the restricted expression pattern of γ3 subunits in brain (Laurie et al., 1992a, 1992b; Wisden et al., 1992), indirect gephyrin interactions with the major γ2 subunits are likely to represent the critical parameters for synaptic GABAAR clustering in vivo.

The only protein known so far to be capable of binding to both the GABAAR γ2 subunit and gephyrin is GABARAP (Wang et al., 1999) (Figure 2). GABARAP has a tubulin-binding motif and closely resembles the human factor GATE-16, a Golgi-associated protein, and the yeast autophagy protein Agp8p/Aut7 (Kneussel et al., 2000). However, although GABARAP was initially considered as a candidate for the GABAAR—gephyrin linkage at synapses, GABARAP is exclusively found at intracellular vesicle structures and is therefore considered as a factor that participates in receptor transport processes (Kneussel et al., 2000; Kittler et al., 2001). This view is supported by observations by Maas et al. (2006) which provide evidence that gephyrin also participates in intracellular transport. The recent analysis of GABARAP-deficient mice did not display alterations in GABAAR γ2 subunit cluster number at synapses (O'Sullivan et al., 2005), an observation which is in accordance with GABARAP representing a non-synaptic GABAAR-binding partner. However, due to the identification of close homologues to GABARAP (Paz et al., 2000; Xin et al., 2001), single knockout studies cannot exclude functional compensation through a closely related GABARAP-like protein in this respect.

Besides the GABAARs, GABAC receptors (GABACRs) also constitute GABA-gated chloride channels (Feigenspan and Bormann, 1998; Lukasiewicz et al., 2004). Unlike their counterparts, GABACRs are more restricted in their expression and assemble through a combination of only three subunits, ρ1–ρ3 (Lukasiewicz et al., 2004). Most prominently, GABACRs are expressed in the retina where they mediate feedback inhibition from amacrine cells to bipolar cells. It has been demonstrated that GABACRs are distributed diffusely during early postnatal development, but cluster from postnatal day 10 onwards (Koulen et al., 1998). Interestingly, these synaptic clusters do not co-localize with GABAARs, GlyRs or gephyrin clusters (Koulen et al., 1998). A yeast two-hybrid screen using the intracellular loop of the ρ1 subunit as bait identified MAP-1B (microtubule-associated protein 1B) as a specific GABACR interactor (Hanley et al., 1999). MAP-1B co-immunoprecipitates and co-localizes with GABACRs at synaptic sites in the inner plexiform layer of the retina and was suggested to exert a clustering function through binding to both receptors and microtubules. On the other hand, MAP-1B-deficient mice (Meixner et al., 2000) did not display loss of GABACR clustering, a finding which could either be due to compensatory effects through other MAPs or might point to a role of MAP-1B at GABACR-positive synapses other than clustering.

GABAAR clustering at extrasynaptic sites

The prominent GABAAR subunits that locate at extrasynaptic sites are cerebellum-specific α6 subunits (Nusser et al., 1995; Jones et al., 1996), as well as α5 subunits that display predominant expression in hippocampus and olfactory bulb (Laurie et al., 1992a, 1992b; Wisden et al., 1992). Notably, mutation or depletion of GABAAR α5 in mice facilitates trace-fear conditioning or improves spatial learning (Collinson et al., 2002; Crestani et al., 2002). Whether GABAAR α5 subunits exclusively mediate tonic inhibition in neurons is currently a matter of debate. Since tonic inhibition would not necessarily require local receptor concentration, the finding that α5-subunit-containing GABAARs are, indeed, organized into cluster formations at extrasynaptic sites could either point to neuron—glia contacts or to reserve pool functions for synaptic—extrasynaptic exchange (Figure 4). In this respect, studies that address membrane diffusion and anchoring of this receptor subtype will be required to understand the behavioural phenotype of GABAAR α5 mouse mutants at the molecular level.

Figure 4.

GABAAR clusters at synapses (left) and extrasynaptic sites (right)

Synaptic receptors (blue ovals) are linked to gephyrin and associated proteins, as described in Figure 2. Extrasynaptic GABAAR clusters, which contain the α5 subunit, are mainly expressed on hippocampal pyramidal cells and carry β3 and γ2 or γ3 subunits. Cluster formation of α5-subunit-containing GABAARs requires active radixin, an ERM family protein that links receptors to the actin cytoskeleton. Intramolecular interaction of the N- and C-terminal domains of radixin leads to an inactive molecule. Upon PIP2 binding and phosphorylation, radixin switches into its active conformation, which is required for both receptor and F-actin binding. The reason why α5-subunit-containing GABAARs, which are known to mediate tonic inhibiton, form clusters at extrasynaptic sites is currently unknown. Neuron—glia contact or a reserve pool function for synaptic receptor exchange could account for this extrasynaptic association of GABAAR α5 molecules; however, both of these ideas are currently speculative.

A recent study, indeed, identified a crucial player for GABAAR clustering in a protein—protein interaction screen (Loebrich et al., 2006). Here, the ERM protein radixin was found to be required for α5-subunit-containing receptor anchoring at the actin cytoskeleton (Figure 4). Both radixin and α5-subunit-containing GABAARs are mainly located at extrasynaptic sites (Brunig et al., 2002; Loebrich et al., 2006), but to a lesser extent are also found at GABAergic synapses (Christie and de Blas, 2002; Loebrich et al., 2006). Consistent with an essential function of radixin in receptor cluster formation, both radixin antisense depletion or gene knockout in mice causes loss of GABAAR α5 puncta in cultured neurons and hippocampal tissue sections (Loebrich et al., 2006) (Figure 5).

Figure 5.

Loss of GABAAR α5 receptor clustering in radixin-knockout mice

Immunohistochemical analysis of wild-type and radixin-knockout hippocampal slices. Left-hand panels: TOTO staining to reveal the overall tissue structure. Middle and right-hand panels: GABAAR α5 subunit staining reveals loss of GABAAR α5 receptor clusters, with remaining diffuse signals upon radixin deficiency. Scale bars: middle panel, 200 μm; right-hand panel, 10 μm. Reprinted by permission from Macmillan Publishers Ltd: The EMBO Journal, Loebrich et al. © 2006. (http:www.nature.comemboj).

To function as a receptor clustering factor that anchors GABAAR α5 subunits through simultaneous binding to F-actin (Loebrich et al., 2006), radixin molecules, such as other ERM family members, require intramolecular activation. Both docking to membrane-associated PIP2 (phosphatidylinositol 4,5-bisphosphate) and subsequent phosphorylation at a conserved C-terminal threonine residue (Fievet et al., 2004) are essential to open and activate ERM protein conformations that are subsequently required for radixin-mediated receptor—cytoskeleton anchoring (Figure 4). Although efforts have been undertaken to identify the kinases that mediate phosphorylation of ERM proteins (Matsui et al., 1998; Cant and Pitcher, 2005; Koss et al., 2006), little is known about the regulatory pathways that activate radixin and subsequently regulate GABAAR α5 subunit clustering in vivo.

Since GABAAR α5 subunit mouse mutants display characteristic behavioural phenotypes, future studies on radixin knockouts that lack GABAAR α5 clusters will have to unravel whether, and to what extent, these effects require receptor cluster formation. Notably, clustering mechanisms of radixin and gephyrin seem to operate independent of each other. In accordance with this view, radixin and gephyrin do not bind and co-localize with each other (Loebrich et al., 2006), and radixin knockout does not lead to postnatal death (Kikuchi et al., 2002; Kitajiri et al., 2004), as compared with studies for gephyrin knockouts (Feng et al., 1998).

Currently, inspite of the data obtained for GABAAR α5 and F-actin, our knowledge about radixin-binding partners in neurons is very limited. The identification and functional characterization of additional radixin-binding partners might therefore be an option to shed light on: (i) the mechanisms that cause clustering of α5-subunit-containing GABAARs; (ii) the reason why such clusters form at extrasynaptic sites; and (iii) whether extrasynaptic cluster formations could be subject to recruitment into postsynaptic sites.

Conclusions

The regulation of neurotransmitter receptor numbers at postsynaptic sites at a given time requires the precise interplay of membrane and/or submembrane transport, diffusion and clustering. To alter the number of receptors available for ligand binding at axo-dendritic contacts, neurons use three major mechanisms: (i) motor-protein-mediated cargo transport between intracellular organelles and the plasma membrane; (ii) intramembrane receptor diffusion, followed by subsequent trapping and/or release through submembrane scaffolds; and (iii) internalization through endocytosis, followed by recycling and/or degradation of receptor molecules. In addition, the size of postsynaptic scaffolds that are available for receptor trapping might be proportional to the number of synaptic receptors. Consequently, turnover of scaffold molecules could indirectly contribute to synaptic plasticity. Similar to the case of radixin, which clusters extrasynaptic GABAARs, intramolecular activation of receptor-binding proteins, which can be regulated via various triggers (phosphorylation, ubiquitylation, palmitoylation, PIP2 binding etc.), also participates in regulatory processes at synapses.

At inhibitory postsynaptic sites, a number of proteins has been described and functionally characterized that either mediate structural roles, such as cytoskeletal coupling (gephyrin, radixin, Mena/VASP and F-actin), or are involved in signalling cascades modulating synaptic function (RAFT1, collybistin). Whereas certain proteins exclusively co-localize with receptors at intracellular positions (GABARAP, GODZ and Plic1), others are localized to both submembrane and intracellular compartments (GRIP1 and gephyrin), thereby representing suitable candidates for trafficking and communication between membrane specializations and intracellular structures.

Transport of proteins to, from and between postsynaptic sites might substantially differ between inhibitory synapses that are located at the shaft of dendrites and excitatory synapses, which are enclosed in a dendritic spine compartment, representing a micro-environment with limited exchange of material with its parent dendrite. For example, despite their important role in tonic inhibition, pre-formed clusters at extrasynaptic sites might represent receptor reserve pools for synaptic recruitment upon certain physiological triggers.

Notably, the GABAAR binding protein HAP1 (Kittler et al., 2004b) also binds to molecular motor proteins and to Huntingtin (Harjes and Wanker, 2003), a protein that harbours an abnormally elongated polyglutamine tract in Huntington's disease (Harjes and Wanker, 2003; McGuire et al., 2006). Whether Huntingtin, HAP1 and molecular motors are involved in the transport of GABAARs to or from synaptic sites, and whether the cognitive impairment in Huntington's disease involves abnormal transport of synaptic receptors, is currently unknown.

In any case, further identification of new proteins that are located at inhibitory PSDs, as well as functional understanding of postsynaptic protein turnover, will be an important aspect for our understanding of synaptic plasticity in learning, memory and disease.

Acknowledgements

This work was supported by grants from the Deutsche Forschungsgemeinschaft (SFB444/B7, DFG 556/1-3, DFG556/3-1) to M.K.

Ancillary