GABA_AR trafficking-mediated plasticity of inhibitory synapses

Changes in GABA_AR subunit gene expression

Regulated expression of GABA_AR subunit genes determines cell type-specific and developmental changes in the subunit composition and function of GABA_ARs. In addition, significant changes in subunit mRNA levels are observed in adulthood. For example, the subunit gene expression of α4βδ receptors in granule cells of the dentate gyrus is dynamically altered during epileptogenesis in a rat model of epilepsy (Brooks-Kayal et al., 1998; Peng et al., 2004) and during the estrus cycle of the mouse (Maguire et al., 2005). The levels of mRNAs encoding subunits of these receptors in CA1 pyramidal cells of rats is changed during puberty (Shen et al., 2007; Shen et al., 2010a), at the end of pregnancy (Sanna et al., 2009) and in a progesterone withdrawal model of premenstrual syndrome (Sundstrom-Poromaa et al., 2002). These studies in rodents indicate that alterations in subunit mRNA levels are generally paralleled by corresponding changes in the surface accumulation and function of GABA_ARs that contribute to changes in neural excitability. Significant changes in subunit gene expression have also been described in postmortem brain of depressed patients and suicide victims (Klempan et al., 2009; Sequeira et al., 2009). Such changes are predicted to disrupt the subunit composition of GABA_ARs and are consistent with the GABAergic deficit hypothesis of major depression (Luscher et al., 2011).

Dynamic changes in the Cl⁻ equilibrium potential

The neural response to GABA_AR activation depends on the Cl⁻ equilibrium (E_Cl) potential, which determines the electrochemical driving force for Cl⁻. E_Cl is determined chiefly by the relative expression of the Cl⁻-transporters KCC2 and NKCC1, which increase and decrease, respectively, during animal development and neural differentiation (for reviews see Ben-Ari, 2002; Fiumelli and Woodin, 2007; Andang and Lendahl, 2008). The ensuing hyperpolarizing shift in E_Cl leads to a gradual conversion of GABAergic depolarization in immature neurons to mainly hyperpolarizing function in mature neurons. This switch in the function of GABA_ARs is essential for structural and functional maturation of neurons (Tozuka et al., 2005; Ge et al., 2006; Cancedda et al., 2007) and for termination of interneuron migration in the developing neocortex (Bortone and Polleux, 2009). Recent evidence further suggests that the E_Cl of mature neurons may be subject to synaptic input-specific modulation by the voltage- and Cl⁻ sensitive Cl⁻ channel ClC-2 (Foldy et al., 2010). The proposed function of ClC-2 is to prevent excessive accumulation of intracellular Cl⁻ following strong GABAergic stimulation. While GABAergic inputs to mature neurons are mostly inhibitory, depolarizing GABAergic effects are also common (reviewed by Marty and Llano, 2005; Kahle et al., 2008). In particular, the aforementioned axo-axonic synapses at the axon initial segment of cortical pyramidal cells (Szabadics et al., 2006), at hippocampal mossy fiber terminals (Jang et al., 2006), and on parallel fibers of the cerebellum (Stell et al., 2007; Pugh and Jahr, 2011) are depolarizing and excitatory due to the local absence of KCC2 (Szabadics et al., 2006). Moreover, dynamic changes in the functional expression of KCC2 can lead to pathophysiological adaptations of neural excitability. For example, chronic stress-induced downregulation of KCC2 results in a depolarizing shift of the chloride reversal potential of neurons in the paraventricular nucleus of the hypothalamus, which renders GABA inputs ineffective (Hewitt et al., 2009). This posttranslational mechanism is thought to contribute to hypothalamus-pituitary-adrenal (HPA) axis hyperactivity and to the neuropathology of stress-associated neuropsychiatric disorders. Moreover, KCC2 mRNA and/or protein expression is down-regulated following focal ischemia (Jaenisch et al., 2010) and status epilepticus (Pathak et al., 2007). The sustained loss of GABAergic inhibition observed following status epilepticus has been proposed to underlie injury-induced long-term increases in seizure susceptibility. Mechanistically, status-epilepticus-induced downregulation of KCC2 involves BDNF/TrkB-mediated activation of cAMP response element-binding protein (CREB) (Rivera et al., 2004) and phosphorylation of tyrosine residues in the KCC2 C-terminal domain, which triggers its lysosomal degradation (Lee et al., 2010).

ER to Golgi translocation of GABA_ARs

The γ2 subunit of GABA_ARs is subject to palmitoylation at cytoplasmic cysteine residues, and this modification regulates the accumulation of GABA_ARs at inhibitory synapses (Keller et al., 2004; Rathenberg et al., 2004). The Golgi-specific DHHC zinc finger protein (GODZ, zDHHC3) interacts with and palmitoylates the γ2 subunit in vitro (Figure 1C) (Keller et al., 2004; Fang et al., 2006). In brain, GODZ is selectively expressed in neurons and highly restricted to Golgi membranes, including Golgi outposts in primary dendrites (Keller et al., 2004). The protein is a member of a family of at least 23 structurally related palmitoyltransferases characterized by the presence of a DHHC motif-containing cysteine-rich domain (DHHC-CRD). Among these, only GODZ and its paralog SERZ-β (zDHHC7) are able to palmitoylate the γ2 subunit in heterologous cells (Fang et al., 2006). Reducing the expression of GODZ by shRNA or dominant negative constructs leads to selective loss of GABA_ARs at synapses, along with reduced GABAergic innervation and corresponding reductions in amplitude and frequency of miniature inhibitory synaptic currents (mIPSCs), as well as whole cell currents (Fang et al., 2006). Palmitoylation is a reversible posttranslational modification and therefore may dynamically regulate the association of cytoplasmic substrates with membranous structures. In the case of integral membrane proteins, however, palmitoylation may extend the effective length of an adjacent transmembrane domain, as suggested by analysis of the palmitoylation-dependent trafficking of the Wnt coreceptor LRP6 (lipoprotein receptor-related protein 6) (Abrami et al., 2008). The restricted localization of GODZ to Golgi membranes, together with the notion that ER membranes are thinner than Golgi and plasma membranes (Bretscher and Munro, 1993; Mitra et al., 2004), suggests that GODZ serves to facilitate ER to Golgi translocation of γ2-containing GABA_ARs (Figure 2).

Translocation of GABA_ARs from Golgi to the plasma membrane

The brefeldin A-inhibited GDP/GTP exchange factor 2 (BIG2) interacts with a sequence motif in the intracellular loop of GABA_AR β subunits that overlaps with the PLIC binding site (Figure 1C) (Charych et al., 2004b). BIG2 is a Sec7 domain-containing guanine exchange factor (GEF) that catalyzes GDP/GTP exchange of class I ADP-ribosylation factors (ARF) 1 and 3 (Morinaga et al., 1997; Togawa et al., 1999). GEF activation of these G-proteins is required for membrane budding of vesicles from the Golgi apparatus, thereby enabling proteins to proceed through the trans-Golgi network (TGN) towards the plasma membrane (Shin et al., 2004) (Figure 2). Coexpression of BIG2 with the β3 subunit in heterologous cells promotes the translocation of this subunit to the cell surface. Consistent with a role in exocytosis of GABA_ARs, BIG2 immunoreactivity is concentrated in the TGN and has been detected in somatic and dendritic vesicle-like structures, as well as in the postsynaptic density of both inhibitory and excitatory synapses (Charych et al., 2004b). Interestingly, independent studies have identified BIG2 as a component of recycling endosomes and provided evidence that BIG2-mediated activation of ARFs contributes to the structural integrity of this trafficking compartment (Shin et al., 2004; Shin et al., 2005; Boal and Stephens, 2010). Thus, BIG2 is implicated in facilitating the exit of GABA_ARs from the Golgi towards the plasma membrane as well as in endocytic recycling of GABA_ARs.

The GABA_AR associated protein (GABARAP) represents the first GABA_AR interacting protein isolated and accordingly has received considerable attention (Wang et al., 1999; reviewed in Chen and Olsen, 2007). It belongs to a family of ubiquitin-like proteins that in mammals includes the paralogs GEC-1 (guinea-pig endometrial cells-1, also known as GABARAP-like 1, GABARAPL1), GATE-16 (Golgi-associated ATPase enhancer of 16 kDa, also known as ganglioside expression factor-2 or GABARAPL2), GABARAPL3, GABARAPL4 and the more distantly related MAP-LC3 (microtubule-associated protein light chain 3). GABARAP interacts with all γ subunits and with microtubules in vitro and in vivo (Figure 1C) (Wang et al., 1999; Nymann-Andersen et al., 2002b). The protein is enriched in Golgi and other somatodendritic membrane compartments but absent at synapses (Kneussel et al., 2000; Kittler et al., 2001). Upon overexpression in hippocampal neurons, GABARAP facilitates the translocation of GABA_ARs to the cell surface (Leil et al., 2004). Interestingly, GABARAP-mediated trafficking of GABA_ARs involves an evolutionarily conserved lipid conjugation and delipidation cycle first described in yeast (Tanida et al., 2004). The attachment of phosphatidyl ethanolamine (PE) to the C-terminus of GABARAP family proteins involves activating, conjugating, and deconjugating enzymes analogous to the ubiquitin conjugation system (Hemelaar et al., 2003; Tanida et al., 2003; Kabeya et al., 2004). Experiments in transfected cultured neurons indicate that conjugation to PE is required for dendritic accumulation of GABARAP and for GABARAP-induced cell surface expression of GABA_ARs (Chen and Olsen, 2007).

Activity-induced translocation of GABA_ARs to the plasma membrane

Acute knockdown of GABARAP by siRNA in cultured neurons has revealed a role of GABARAP in rapid NMDA-induced functional plasticity of inhibitory synapses (Marsden et al., 2007). NMDA receptor-mediated Ca²⁺ influx following moderate stimulation of neurons with NMDA leads to a rapid increase in the number of postsynaptic GABA_AR clusters and mIPSC amplitudes (see also further below and Figure 5C). In addition to GABARAP this mechanism involves Ca²⁺ calmodulin-dependent kinase II (CaMKII), the vesicular trafficking factor N-ethylmaleimide-sensitive factor (NSF), and glutamate receptor interacting protein (GRIP). The rate of GABA_AR endocytosis following treatment with NMDA was unaltered, suggesting that GABARAP-dependent potentiation of inhibitory synapses involves increased exocytosis rather than reduced endocytosis of GABA_ARs (Marsden et al., 2007). These findings represent the so far only loss-of-function experiments showing an essential role for endogenous GABARAP in GABA_AR trafficking. The relevant protein-protein interactions and CaMKII phosphorylation targets have so far not been identified. However, experiments in heterologous cells allow speculation that this mechanism might involve CaMKII-induced phosphorylation of the β3 subunit at S383 (Houston et al., 2007).

GABARAP - a multitasker?

The data summarized thus far suggest that GABARAP promotes the regulated, activity-dependent and CaMKII-mediated translocation of GABA_ARs from intracellular compartments to the somatodendritic plasma membrane. However, a more general role of GABARAP in exocytosis of GABA_ARs is difficult to reconcile with other findings. First, GABARAP has been proposed to contribute to rebound potentiation, a neural activity-induced postsynaptic form of long-term potentiation (LTP) of inhibitory synapses on Purkinje cell neurons (Kawaguchi and Hirano, 2007). Using electrical stimulation of cultured Purkinje cells to mimic rebound potentiation, the authors found evidence that this form of plasticity is critically dependent on a CaMKII-dependent conformational alteration of GABARAP. However, LTP of GABAergic synapses occurred without measurable changes in the cellular distribution and cell surface expression of GABA_ARs. Given that GABARAP is absent at synapses (Kneussel et al., 2000; Kittler et al., 2001) the exact role of GABARAP in this form of plasticity requires further clarification. Second, the aforementioned PE conjugation of GABARAP is critically involved in autophagy, an evolutionarily conserved form of bulk transport of membranes and cytoplasm to lysosomes for protein degradation (Tanida et al., 2004). Consistent with a role of GABARAP in autophagy, there is evidence that GABA_ARs are subject to autophagy in worms. Body wall muscle cells of C. elegans lacking proper GABAergic and cholinergic innervation show selective accumulation of GABA_ARs, but not nicotinic acetylcholine receptors in autophagosomes. This indicates that in worms autophagy represents a pathway for endocytic lysosomal degradation of GABA_ARs (Rowland et al., 2006). Third, GABARAP knock-out mice show normal expression and punctate distribution of γ2-containing GABA_ARs (O’Sullivan et al., 2005), possibly due to functional redundancy of GABARAP with GEC1 (Mansuy-Schlick et al., 2006) and other GABARAP family members. Third, the function of GABARAP is complicated by its interactions with a very large number of other proteins. Among these the ER luminal Ca²⁺-dependent chaperone calreticulin stands out in that it binds GABARAP with exceptionally high affinity (Mohrluder et al., 2007). Compared to calreticulin the interaction of GABARAP with γ2-derived peptides shows low affinity, suggesting that GABARAP might promote protein trafficking unspecifically along the secretory pathway (Knight et al., 2002).

Several GABARAP-interacting proteins contribute to GABA_AR trafficking independently of GABARAP. The aforementioned NMDAR-induced and GABARAP-dependent increase in GABA_AR clustering also depends on the synaptic PDZ domain-containing protein GRIP (Marsden et al., 2007), which interacts with GABARAP in vitro and in vivo (Kittler et al., 2004a). GRIP was first described as a trafficking factor of AMPARs (Dong et al., 1997). It is present at both glutamatergic and GABAergic synapses, consistent with functions at both types of synapses (Dong et al., 1999; Charych et al., 2004a; Kittler et al., 2004a; Li et al., 2005). GABARAP further interacts with the phospholipase C-related catalytically inactive proteins 1 and 2 [PRIP1/2, PRIP1 was previously named p130 (Kanematsu et al., 2002)], a pair of GABA_AR-associated adaptor proteins for phosphatases and kinases (Figure 3A) (Kanematsu et al., 2002; Uji et al., 2002). Likewise, GABARAP and its paralog GATE-16 (Sagiv et al., 2000) interact with NSF, an ATPase and chaperone of SNARE complexes that is critically important for regulated neurotransmitter release and also involved in trafficking of neurotransmitter receptors (Morgan and Burgoyne, 2004; Zhao et al., 2007).

Both PRIP1/2 and NSF interact with GABA_ARs indirectly through GABARAP and directly via GABA_AR β subunits (Figure 1C) (Kanematsu et al., 2002; Kittler et al., 2004a; Terunuma et al., 2004; Goto et al., 2005). PRIP1/2 double knock-out mice exhibit reduced expression and altered behavioral pharmacology of GABA_ARs, suggesting deficits in mainly γ2-containing GABA_ARs (Kanematsu et al., 2002; Kanematsu et al., 2006; Mizokami et al., 2007). Brain extracts of these mice further show reduced association of GABA_ARs with GABARAP, indicating that PRIP facilitates indirect association of GABARAP with GABA_ARs (Mizokami et al., 2007). Moreover, PRIP and the γ2 subunit compete for binding to the same binding site on GABARAP (Kanematsu et al., 2002; Uji et al., 2002). The minimal interaction site for GABARAP in the γ2 subunit further overlaps with a γ2 domain that interacts with itself and with γ1, γ3 and β subunits in vitro (Figure 1C)(Nymann-Andersen et al., 2002a). Circumstantial evidence suggests that GABA_ARs are endocytosed as dimers (Kittler et al., 2008). It is therefore conceivable that overexpressed GABARAP and γ2 subunit peptides designed to compete for interaction with GABARAP affect the surface expression of GABA_ARs in part by competing for other protein interactions (receptor dimerization, interaction of receptors with PRIP and NSF) that facilitate the endocytic trafficking of GABA_ARs.

Clathrin-mediated endocytosis

Regulated endocytosis of neurotransmitter receptors is known to underlie physiological and pathological adaptations of neural excitability. Endocytosis of GABA_ARs occurs primarily via clathrin- and dynamin-dependent mechanisms that are facilitated by interactions of the GABA_AR β and γ subunits with the clathrin adaptor protein AP2 (Kittler et al., 2000; Kittler et al., 2005; Kittler et al., 2008)(Figure 4). Accordingly, blocking the function of dynamin results in increased accumulation of postsynaptic GABA_ARs along with increased mIPSC amplitudes (Kittler et al., 2000). When measured in about one-week-old cultures, approximately 25% of cell surface GABA_ARs are endocytosed within 30 minutes, and 70% of these receptors are recycled back to the cell surface within one hour. On a slower time scale (6 h), about 30% of neuronal GABA_ARs are targeted to late endosomes where they become subject to lysosomal degradation (Kittler et al., 2004b). However, constitutive endocytosis of GABA_ARs is significantly reduced in mature neurons as indicated by studies that analyzed the diffusion dynamics of GABA_ARs within the plasma membrane (discussed later in this review).

The search for sequence motifs important for AP2/clathrin/dynamin-mediated endocytosis of GABA_ARs first lead to the identification of a di-leucine motif in β subunits that is critical for receptor internalization in heterologous cells (Herring et al., 2003; Herring et al., 2005). However, whether this particular mechanism operates in neurons has not been established. A second motif that is important for AP2/clathrin/dynamin-mediated GABA_AR internalization in neurons has been mapped to a highly basic 10-amino acid sequence motif that includes a major phosphorylation site conserved in the cytoplasmic loop region of β1–3 subunits (S408, S409 in β3, Figure 1C) (Kittler et al., 2005; Kittler et al., 2008). Importantly, interaction of the AP2 μ2 subunit with this site is negatively regulated by phosphorylation of GABA_AR β subunits, indicating that AP2 binds GABA_ARs with high affinity and triggers their internalization preferentially when this site is dephosphorylated. Accordingly, perfusion of neurons with an unphosphorylated β3-derived peptide that competes with this interaction results in enhanced mIPSCs and whole cell currents (Kittler et al., 2005).

Regulation of endocytosis by phospho-sensitive interactions with GABA_AR β subunits

The AP2 interaction site on β1 and β3 subunits (Figure 1C) can be phosphorylated by both protein kinase A (PKA) (McDonald et al., 1998) and protein kinase C (PKC) (Brandon et al., 2000; Brandon et al., 2002), while the same site in the β2 subunit is phosphorylated by PKC only (McDonald et al., 1998; Brandon et al., 2003), allowing for receptor subtype-specific modulation of GABA_AR endocytosis. However, the same site can also be phosphorylated by CaMKII (McDonald and Moss, 1994) and Akt (also known as PKB) (Wang et al., 2003b; Xu et al., 2006). The latter is discussed further below in the context of insulin-induced exocytosis of GABA_ARs. PKC-mediated phosphorylation is facilitated by stable interaction of this kinase with β subunits, either directly as shown for the PKC-βII isozyme or indirectly through the receptor for activated C-Kinase (RACK-1), which recognizes a binding site in the β1 subunit adjacent to the PKC binding site (Brandon et al., 1999; Brandon et al., 2002). Reductions in the PKC-mediated phosphorylation of GABA_AR β subunits are implicated in the dramatic loss of GABAergic inhibition in animal models of status epilepticus, which is thought to underlie pharmaco-resistance to benzodiazepines following prolonged seizures in epileptic patients (Terunuma et al., 2008).

The β subunit phosphostate-dependent endocytosis of GABA_ARs is further regulated by interaction of β subunits with PRIP1/2 and their function as adaptors for the serine/threonine-specific phosphatases PP1α and PP2A (Yoshimura et al., 2001; Uji et al., 2002; Terunuma et al., 2004; Kanematsu et al., 2006; Kanematsu et al., 2007). Phosphorylation of PRIP at a threonine residue (T94 in PRIP1) leads to dissociation of the catalytically inactive PRIP/PP1α complex and activation of PP1α and hence dephosphorylation of the β3 subunit at the AP2 interaction site (Terunuma et al., 2004). Unlike PP1α, PP2A is constitutively active when bound to PRIP (Kanematsu et al., 2006). Consistent with a role of PRIP-associated phosphatases in endocytosis of GABA_ARs, the PRIP/PP1α/PP2A complex can be coimmunoprecipitated with AP2 and clathrin from brain extracts (Kanematsu et al., 2007). Moreover, PRIP facilitates GABA_AR endocytosis in transfected heterologous cells. The association of PRIP with PP2A (Kanematsu et al., 2006) is implicated in brain-derived neurotrophic factor (BDNF)-induced downregulation of GABA_ARs (Jovanovic et al., 2004), as discussed in further detail below. The end effect of PRIP on GABA_AR cell surface expression appears to depend on the cellular state of several other signal transduction pathways. The aforementioned phenotype of PRIP1/2 double knock-out mice, which includes functional deficits of GABA_ARs, suggests that PRIP primarily facilitates the exocytosis or cell surface stability of GABA_ARs (Kanematsu et al., 2002; Kanematsu et al., 2006; Mizokami et al., 2007). Recent evidence summarized further below indicates that PRIP1/2 also serve as an adaptor for the serine/threonine kinase Akt, which promotes the de novo insertion of GABA_ARs into the plasma membrane (Fujii et al., 2010).

Intriguingly, the AP2 interaction site in the β1–3 subunits overlaps with the binding site for the vesicular ATPase and trafficking factor NSF (Figure 1C) (Goto et al., 2005). NSF interacts with phorbol ester-activated PKCε. Moreover, PKCε phosphorylates and activates the ATPase function of NSF. PKCε-mediated phosphorylation of NSF induces its translocation to the plasma membrane and to synapses and concurrently reduces the cell surface expression of GABA_ARs (Chou et al., 2010). PKCε knock-out mice are less anxious and produce lower levels of stress hormone than WT mice (Hodge et al., 2002), which is the opposite of the anxious-depressive-like phenotype of GABA_AR γ2 subunit heterozygous mice and therefore consistent with increased functional expression of GABA_ARs (Crestani et al., 1999; Luscher et al., 2011). Therefore, pharmacological inhibitors of PKCε activity may have therapeutic potential for the treatment of neuropathological conditions that involve deficits in GABAergic transmission. This NSF-dependent trafficking mechanism is reminiscent of aforementioned earlier experiments conducted in heterologous cells, showing phorbol ester-induced and PKC and clathrin-mediated endocytosis of GABA_ARs from the plasma membrane by a mechanism that is independent of GABA_AR phosphorylation (Chapell et al., 1998; Connolly et al., 1999). PKCε is one of seven PKC isozymes activated by phorbol esters. It therefore seems likely that PKCε contributes to phorbol ester-induced endocytosis of GABA_ARs. However, one might predict that PKCε and NSF-dependent endocytosis of GABA_ARs is counteracted by the aforementioned PKC-βII-mediated phosphorylation of β subunits, which limits endocytosis of GABA_ARs. Consistent with multiple PKC and PKA-regulated modes of GABA_AR trafficking, these kinases can have cell type specific and functionally opposite effects on mIPSC amplitudes in vivo (Poisbeau et al., 1999).

Regulation of endocytosis by phospho-sensitive interactions with the GABA_AR γ2 subunit

A third interaction of GABA_ARs with AP2 involves a bipartite motif in the intracellular loop region of the γ2 subunit (Figure 1C). It consists of a 12-amino-acid basic domain that is homologous to the AP2 binding site in β subunits and a more C-terminal γ2-specific YGYECL motif (Smith et al., 2008). These two domains interact cooperatively with separate domains in the μ2 subunit of AP2. The γ2-specific YGYECL motif is of particular interest as it exhibits high affinity for AP2 that is sensitive to phosphorylation at γ2 Tyr365/367 (Kittler et al., 2008). These residues are phosphorylated by Fyn and other Src kinase family members in vivo (Lu et al., 1999; Jurd et al., 2010). A non-phosphorylated YGYECL peptide effectively competes with the AP2-γ2 subunit interaction, thereby increasing the GABA_ARs surface expression and mIPSCs amplitude and showing that this site is constitutively phosphorylated in cultured neurons (Kittler et al., 2008). This mechanism is also important in vivo as evidenced by reduced expression and altered function of GABA_ARs in Fyn knock-out mice (Boehm et al., 2004), and by an embryonic lethal phenotype of knock-in mice in which the γ2 Tyr365/367 residues were mutated to phenylalanine, which interferes with AP2 binding (Tretter et al., 2009). Heterozygous γ2^Y365/7F mice, however, are viable. In the stratum pyramidale of the hippocampus they show a CA3-region-specific increase in the postsynaptic accumulation of GABA_ARs, suggesting different basal levels of γ2 Tyr365/367 phosphorylation in the CA3 versus CA1 region. The lethal phenotype of homozygous γ2^Y365/7F mutants indicates that excessive GABAergic transmission is detrimental during early development, probably due to excessive GABAergic excitation, which may interfere with normal neurogenesis and neural migration (Wang and Kriegstein, 2009).

Collectively, there is now conclusive evidence that GABA_ARs are subject to at least two major mechanisms of regulated endocytosis. These mechanisms involve different phospho-sensitive interactions of the clathrin adaptor AP2 with β and γ2 subunits, respectively. The phospho-states of the relevant β and γ2 subunit motifs are subject to regulation by multiple Ser/Thr and Tyr kinases, as well as phosphatases and their respective adaptor proteins. Dynamic changes in the phosphorylation state of NSF and PRIP and their interaction with the AP2 binding site of β subunits provide additional levels of regulation. Future experiments will need to address whether NSF and PRIP compete with AP2 for GABA_AR interaction and whether their interaction with GABA_ARs is regulated by phosphorylation of GABA_ARs.

Gephyrin

Arguably the most important protein for stabilization of GABA_ARs at synapses is gephyrin, the principal subsynaptic scaffold protein of both GABAergic and glycinergic synapses (Figures 3B, 5A) (Fritschy et al., 2008). Gephyrin was first identified as a 93-KDa polypeptide that copurified with affinity-purified glycine receptors (Pfeiffer et al., 1982), the principal inhibitory neurotransmitter receptors in the spinal cord. Molecular cloning and targeted deletion in mice revealed that gephyrin is a multifunctional protein that is broadly expressed and essential for postsynaptic clustering of glycine receptors and also for molybdenum cofactor (Moco) biosynthesis in nonneural tissues (Prior et al., 1992; Kirsch et al., 1993; Feng et al., 1998; Sola et al., 2004; Dumoulin et al., 2009). Gephyrin interacts with microtubules (Kirsch et al., 1995) as well as several regulators of microfilament dynamics including profilin I and II (Mammoto et al., 1998) and members of the mammalian enabled (Mena)/vasodilator stimulated phosphoprotein (VASP) family (Figure 3B, 5A) (Giesemann et al., 2003). The N-terminal gephyrin domain known as G-gephyrin assumes a trimeric structure (Schwarz et al., 2001; Sola et al., 2001), whereas the C-terminal E domain forms a dimer (Schwarz et al., 2001; Xiang et al., 2001; Sola et al., 2004). These domain interactions are essential for oligomerization and clustering of gephyrin at postsynaptic sites (Saiyed et al., 2007). The clustering function of gephyrin is regulated by select residues within the E-domain that are dispensable for E domain dimerization (Lardi-Studler et al., 2007). Moreover, the linker region between E and G domains of gephyrin is thought to interact with microtubules (Ramming et al., 2000). Thus, gephyrin has the structural prerequisites to form a microtubule and microfilament-associated hexagonal protein lattice that may organize the spatial distribution of receptors and other proteins in the postsynaptic membrane.

Gephyrin has long been established as a phosphoprotein (Langosch et al., 1992), although to date few studies have addressed the relevance of this modification. Zita et al. (2007) showed preliminary evidence that gephyrin is phosphorylated by proline-directed kinase(s) and that this is essential for interaction of gephyrin with the peptidyl-prolyl cis/trans isomerase Pin1 (Figure 5A). Pin1-induced conformational changes of gephyrin were found to be essential for maximal clustering of glycine receptors, suggesting a similar function for Pin1 in regulating gephyrin destined for GABAergic synapses. Recently, an unbiased proteomic screen using mass spectrometry mapped the first specific phosphorylation sites to S188, S194 and S200 of gephyrin (Huttlin et al., 2010). Treatment of cultured neurons with inhibitors of the phosphatases PP1α and PP2A caused a significant loss of gephyrin from inhibitory synapses (Bausen et al., 2010). However, mutation of S188/194 to alanine or glutamate resulted in only a modest change in gephyrin or GABA_AR cluster size, indicating that the effect of phosphatases was due to dephosphorylation of other PP1α/PP2A substrates.

An elegant study has identified glycogen synthase kinase 3β (GSK3β) as a proline-directed kinase that controls phosphorylation- and proteolytic cleavage-induced turnover of gephyrin (Figure 5A) (Tyagarajan et al., 2011). Using tandem mass spectrometry of gephyrin, the authors identified S270 as a residue that is basally phosphorylated in brain tissue. Transfection of cultured neurons with phosphorylation-deficient gephyrin^S270A increased the density of gephyrin clusters and the amplitude and frequency of GABAergic mIPSCs, indicating that gephyrin clustering is limited by phosphorylation at S270. However, mutations of S270 had no effect on cluster size. Using kinase-specific inhibitors in in vitro phosphorylation assays the authors identified GSK3β as an important kinase for S270. To address the mechanism by which phosphorylation might increase gephyrin turnover they focused on calpain-1. This Ca²⁺-dependent cysteine protease was previously shown to cleave gephyrin and to produce a stable C-terminal gephyrin fragment of 48–50 kDa (Kawasaki et al., 1997). Transfection of neurons with the natural calpain-1 inhibitor calpastatin increased the gephyrin cluster density (Tyagarajan and Fritschy, 2010). Moreover, this effect was enhanced in the presence of the phosphomimetic mutant gephyrin^S270E as a substrate, indicating that calpain-1-mediated degradation of gephyrin is triggered by phosphorylation of S270. Lastly, the authors showed that S270 phosphostate-dependent clustering of gephyrin is enhanced by chronic treatment of cultured neurons or mice with Li⁺, a potent inhibitor of GSK3β used as mood-stabilizing agent for the treatment of bipolar disorder. The findings strongly suggest that Li⁺-induced enhancement of GABAergic synaptic transmission contributes to the mood stabilizing effects of Li⁺ in patients (Tyagarajan and Fritschy, 2010). GSK3β is inhibited as a downstream target of both the canonical Wnt signaling pathway (Inestrosa and Arenas, 2010) and the insulin receptor signaling pathway. Both pathways promote the postsynaptic clustering of GABA_ARs by additional, gephyrin-independent mechanisms, as detailed further below. Gephyrin forms a stable complex with affinity-purified glycine receptors (Pfeiffer et al., 1982). By contrast, GABA_ARs in detergent-solubilized membrane extracts do not stably associate with gephyrin (Meyer et al., 1995). Moreover, a major subset of GABA_ARs comprised of α1βγ2 receptors can accumulate and cluster at synapses independently of gephyrin (Kneussel et al., 2001; Levi et al., 2004). Nevertheless, in brain gephyrin serves as a reliable postsynaptic marker for all GABAergic synapses (Sassoe-Pognetto et al., 1995; Essrich et al., 1998; Sassoè-Pognetto and Fritschy, 2000). Moreover, reducing the expression of gephyrin in cultured neurons or mice results in the selective loss of synaptic localization of GABA_ARs composed of α2βγ2 or α3βγ2 subunits (Essrich et al., 1998; Kneussel et al., 1999). These data indicate that the exact role of gephyrin at synapses is receptor subtype-specific. Conversely, however, GABA_ARs are essential for postsynaptic clustering of gephyrin at all synapses regardless of the GABA_AR subtype normally present (Essrich et al., 1998; Schweizer et al., 2003; Kralic et al., 2006; Studer et al., 2006; Patrizi et al., 2008).

Receptor subtype-specific functions of gephyrin may be explained at least in part by different modes of interaction of gephyrin with GABA_ARs. Tretter et al (2008) described a detergent-sensitive interaction of gephyrin with a hydrophobic motif in the cytoplasmic loop region of the receptor α2 subunit (Figure 1C). Yeast two-hybrid assays further suggest a similar interaction between gephyrin and the α3 subunit (Saiepour et al., 2010). Curiously, however, the gephyrin binding motif of the α2 subunit but not the homologous sequence of the α1 subunit is sufficient to target a heterologous membrane protein to synapses (Tretter et al., 2008). A lower affinity interaction between GABA_ARs and gephyrin than between glycine receptors and gephyrin is consistent with weaker synaptic confinement of GABA_A than glycine receptors (Levi et al., 2008).

The neuroligin-neurexin complex

The structural and functional maturation of synapses is critically dependent on synaptic adhesion complexes. One such complex involves a transsynaptic interaction of presynaptic neurexins and postsynaptic neuroligins (Figures 3D, 4, 5A) (Ushkaryov et al., 1992; Ushkaryov et al., 1994; Ichtchenko et al., 1995; Ullrich et al., 1995; Ichtchenko et al., 1996; Jamain et al., 2008). Overexpression of different neuroligins in neurons or heterologous cells cocultured with neurons can induce presynaptic development of glutamatergic and GABAergic synapses (Scheiffele et al., 2000; Chih et al., 2005; Chubykin et al., 2007; Dong et al., 2007; Fu and Vicini, 2009). Conversely, β-neurexins presented on beads or overexpressed in heterologous cells can induce the formation of separate postsynaptic GABAergic or glutamatergic hemisynapses in co-cultured neurons (Graf et al., 2004). Of special interest is NL2 as it is localized selectively at inhibitory synapses (Graf et al., 2004; Varoqueaux et al., 2004) and required for structural and functional maturation of subsets of GABAergic but not glutamatergic or glycinergic synapses in vivo (Varoqueaux et al., 2006; Gibson et al., 2009; Hoon et al., 2009; Poulopoulos et al., 2009). By contrast, NL3 is found at both glutamatergic and GABAergic synapses (Budreck and Scheiffele, 2007), while NL1 and NL4 are found primarily at glutamatergic (Song et al., 1999) and glycinergic (Hoon et al., 2011) synapses, respectively. A recent report has identified gephyrin as a direct interaction partner of NLs (Poulopoulos et al., 2009). Although this interaction lacks selectivity for NL2 (Poulopoulos et al., 2009), this finding provides important clues, detailed further below, for the mechanisms underlying the selective deposition of gephyrin at inhibitory synapses.

Interactions between postsynaptic NL2 and presynaptic neurexins are thought to contribute to proper alignment of pre- and postsynaptic molecules at inhibitory synapses. Nevertheless, NL2 is dispensable for clustering and synaptic localization of gephyrin in most brain areas (Varoqueaux et al., 2006; Hoon et al., 2009) (except dentate gyrus Jedlicka et al., 2010), suggesting that other so far unknown synaptogenic complexes might exist. A transsynaptic interaction between the postsynaptic dystrophin-associated glycoprotein (DG) complex and presynaptic neurexins might contribute to the structural integrity of a subset of inhibitory synapses (Sugita et al., 2001). The DG complex consists of the peripheral membrane protein α-dystroglycan, the integral membrane spanning protein β-dystroglycan, and the subsynaptic cytoskeletal component dystrophin. However, this complex appears late during synaptogenesis and is present at a subset of GABAergic synapses only (Knuesel et al., 1999). Moreover, the DG complex is dispensable for postsynaptic clustering of GABA_ARs and unable to promote the accumulation of GABA_ARs and gephyrin at synapses (Brunig et al., 2002b; Levi et al., 2002). Recently the synaptic scaffolding and PDZ domain-containing protein S-SCAM (also known as membrane-associated guanylate kinase inverted-2, MAGI-2) was isolated as a β-dystroglycan interacting protein that might physically link the DG complex to NL2 (Sumita et al., 2007). However, S-SCAM also interacts with NL1 and is found at both excitatory and a subset of inhibitory synapses, suggesting an unspecific role in maturation of synapses.

Collybistin

The gephyrin interacting protein collybistin (CB) is a member of the Dbl family of guanine nucleotide exchange factors (RhoGEFs) that selectively activates the small GTPase Cdc42 (Figure 3C, 4, 5A) (Reid et al., 1999; Kins et al., 2000; Grosskreutz et al., 2001). However, analyses of Cdc42 knock-out mice indicate that Cdc42 is dispensable for gephyrin and GABA_AR clustering (Reddy-Alla et al., 2010). In neurons, CB is colocalized with gephyrin at inhibitory synapses (Saiepour et al., 2010). When coexpressed with gephyrin in heterologous cells CB has the unique ability to transform cytoplasmic aggregates of gephyrin into submembrane micro-clusters that resemble postsynaptic gephyrin clusters of neurons (Kins et al., 2000). Moreover, CB is required for postsynaptic clustering of gephyrin and GABA_ARs, as shown by analyses of naturally occurring mutations of the CB gene (ARHGEF9) associated with hyperekplexia, epilepsy and mental retardation in patients (Harvey et al., 2004; Kalscheuer et al., 2009) as well as by CB gene knock-out in mice (Papadopoulos et al., 2007). Loss of gephyrin and GABA_AR clusters in CB knock-out mice is most pronounced in the hippocampus and amygdala. However, in brainstem, neocortex and several other brain areas the clustering of gephyrin and GABA_A or glycine receptors is unaffected, indicating that other GEFs can compensate for the loss of CB function in these brain areas.

Multiple CB splice variants exist that differ in their C-terminal structures and by the presence or absence of an N-terminal SH3 domain (Kins et al., 2000; Harvey et al., 2004). Intriguingly, the predominant CB isoforms detected in vivo contain an SH3 domain, which inhibits the aforementioned CB-dependent formation of submembrane gephyrin clusters, indicating that CB is negatively regulated by its SH3 domain (Kins et al., 2000; Harvey et al., 2004). However, cotransfection of CB^SH3+ and gephyrin with NL2 negates the inhibitory effect of the SH3 domain (Poulopoulos et al., 2009). CB splice variants invariably contain a pleckstrin homology (PH) domain that is required for its interaction with plasma membrane-restricted phosphoinositides and for clustering of gephyrin at inhibitory synapses (Harvey et al., 2004; Reddy-Alla et al., 2010). The data are consistent with a heterotrimeric membrane-associated complex that consists of NL2, CB^SH3+ and gephyrin and that enables the selective deposition of gephyrin at NL2-containing inhibitory synapses. Experiments in heterologous cells indicate that NL1 can potentially substitute for NL2 and similarly induce submembrane gephyrin clusters but only with constitutively active CB isoform that lack an SH3 domain. In addition, preliminary evidence suggests that the α2 subunit can substitute for NL2 and activate the gephyrin-clustering function of CB^SH3+ (Saiepour et al., 2010). This GABA_AR-dependent function of CB is specific for α2-containing receptors and abolished by a naturally occurring missense mutation (CB^G55A) that disrupts the clustering of α2-containing GABA_ARs and gephyrin in cultured neurons and associated with mental retardation, epilepsy and hyperekplexia in a patient (Harvey et al., 2004; Saiepour et al., 2010). NL1- and α2 subunit-mediated activation of CB might contribute to residual clustering of gephyrin seen in NL2 KO mice (Jedlicka et al., 2010).

Other gephyrin binding proteins that are concentrated at synapses include the Ser/Thr kinase mTor (mammalian target of rapamycin, also known as RAFT1 and FRAP1) (Sabatini et al., 1999) and the dynein light chains (DLC) 1 and 2 (Fuhrmann et al., 2002). mTor functions as an important regulator of mRNA translation, allowing for speculation that gephyrin might contribute to translational control of postsynaptic protein synthesis. This idea is supported by recent evidence that both gephyrin and collybistin are part of the eukaryotic translation initiation factor 3 complex (Sertie et al., 2010). However, whether gephyrin and collybistin play a role in translational control in dendrites remains to be elucidated. The interaction between gephyrin and the DLC is implicated in retrograde vesicular transport of gephyrin-glycine receptor complexes from glycinergic synapses (Maas et al., 2009). The significance of this interaction for GABA_ARs is unclear as intracellular GABA_AR-gephyrin-DLC complexes have not been described and the DLC-gephyrin interaction is dispensable for normal localization of gephyrin at GABAergic synapses (Fuhrmann et al., 2002).

Radixin-mediated extrasynaptic clustering of α5βγ2 receptors

Among different γ2-containing GABA_ARs the α5βγ2 receptors are unique in that they are localized mostly extrasynaptically, as mentioned earlier. Interestingly, even extrasynaptic α5βγ2 receptors are clustered at the plasma membrane (Christie and de Blas, 2002) (Figure 5B). Loebrich et al (2006) have identified radixin as a α5 subunit-interacting protein that is essential for extrasynaptic clustering of α5βγ2 receptors. Radixin is a member of the ERM (ezrin, radixin, moesin) family of proteins, which are known to link transmembrane proteins to the actin cytoskeleton. Transfection of neurons with a dominant negative radixin construct abolishes the clustering of α5-containing receptors but does not affect GABA_AR surface expression nor GABAergic tonic and phasic currents (Loebrich et al., 2006). The data suggest that radixin-independent mechanism prevent α5-containing receptors from accumulation at synapses. The functional relevance of α5βγ2 receptor clustering in the extrasynaptic membrane is not known.

NMDAR- and calcineurin-mediated long-term depression of GABAergic synapses

Experiments that tracked the lateral movement of quantum dot-labeled, single GABA_AR molecules showed that the diffusion coefficient of postsynaptic receptors is about half of that of nonsynaptic receptors (Bannai et al., 2009). Increasing neural activity with a K⁺-channel blocker increased the diffusion coefficient of both synaptic and extrasynaptic GABA_ARs and decreased the postsynaptic cluster size of gephyrin and GABA_ARs, concomitant with a reduction in the amplitude of mIPSCs (Figure 5B). This effect of increased neural activity was dependent on Ca²⁺ influx and activation of calcineurin, did not involve receptor internalization, and was reversed when normal neural activity was restored. These results are consistent with EPSC-induced long-term depression (LTD) of unitary IPSCs observed in association with high frequency stimulation-induced LTP of the Schaffer collateral-CA1 pathway (Lu et al., 2000; Wang et al., 2003a). LTD of IPSCs required NMDA receptor-dependent recruitment of calcineurin to the GABA_AR complex and calcineurin-mediated dephosphorylation of S327 of the γ2 subunit. The findings by Wang et al. and Bannai et al. were confirmed by a recent study that combined live imaging of fluorescently tagged GABA_AR clusters with single molecule tracking of quantum dot-labeled single GABA_AR molecules (Muir et al., 2010). As expected, glutamate-induced dispersal of GABA_AR clusters and enhancement of GABA_AR mobility was critically dependent on NMDA receptor and calcineurin activation and independent of dynamin and therefore did not involve endocytosis of GABA_ARs. Moreover, Glu-induced and calcineurin-mediated dephosphorylation of γ2 S327 increased the lateral mobility and reduced the synaptic residency time of quantum dot-labeled single GABA_AR molecules (Muir et al., 2010). Future experiments will need to address how γ2 S327 regulates interaction of GABA_ARs with the synaptic protein scaffold.

The NMDAR- and calcineurin-mediated form of LTD of inhibitory synapses (Wang et al., 2003a; Bannai et al., 2009; Muir et al., 2010) at first seems in conflict with the aforementioned NMDAR-mediated potentiation of mIPSCs (Marsden et al., 2007). However, more recent evidence suggests that opposite functional effects observed in these two sets of experiments reflect different neuronal stimulation intensities (Marsden et al., 2010). NMDAR-dependent LTD of hippocampal pyramidal cells associated with calcineurin-dependent diffusional dispersal of GABA_ARs reflects robust stimulation of both NMDA and AMPA receptors achieved either by high frequency stimulation of glutamatergic afferents, or treatment of neurons with K⁺-channel blockers or glutamate (Figure 5B). These conditions result in activation of both CaMKII and calcineurin. However, calcineurin inhibits the targeting of CaMKII to inhibitory synapses (Marsden et al., 2010). By contrast, NMDAR-mediated membrane insertion of GABA_ARs and potentiation of mIPSCs is observed following more moderate chemical stimulation of NMDARs only (Figure 5C). Moderate stimulation of neurons with NMDA seemingly fails to activate calcineurin and thereby allows the activation and translocation of CaMKII to inhibitory synapses (Marsden et al., 2010), As mentioned earlier, NMDAR-induced de novo insertion of GABA_ARs into the plasma membrane is further dependent on GABARAP, NSF and GRIP (Marsden et al., 2007). Thus, the directionality of neural activity-induced trafficking of GABA_ARs is strictly stimulus intensity-dependent.

Insulin-regulated cell surface expression of GABA_ARs

Signaling by pancreatic insulin is pivotal for the regulation of peripheral glucose and lipid metabolism. However, insulin is also produced in brain (Havrankova et al., 1981; Stevenson, 1983) and released from neurons in an activity-dependent manner (Clarke et al., 1986). Signaling by insulin receptors contributes to structural maturation of neuronal dendrites, as well as functional synaptic plasticity (reviewed in Chiu and Cline, 2010). In addition, insulin signaling leads to a rapid increase in the cell surface accumulation and function of postsynaptic GABA_ARs (Wan et al., 1997; Wang et al., 2003b).

A first line of investigation indicates that insulin-induced translocation of GABA_ARs to the cell surface requires activation of the serine-threonine kinase Akt, a primary target of insulin signaling downstream of phosphoinositide 3 kinase (PI3K, Figure 6A) (Wang et al., 2003b). PI3K-mediated phosphorylation of membrane lipids is established as a mechanism that leads to recruitment of Akt to the plasma membrane where it is phosphorylated and activated by the serine-threonine kinase, phosphoinositide-dependent kinase 1 (PDK1) (Cantley, 2002). In vitro assays showed that activated Akt phosphorylates a conserved phosphorylation site present in all three β subunits of GABA_ARs (S409 in β1, S410 in β2, S408/409 in β3) (Wang et al., 2003b; Xu et al., 2006). Cotransfection of Akt with α2β2γ2 receptors increased the cell surface expression of these receptors in HEK293 cells. Lastly, phosphorylation of β2 S410 was shown to be essential for Akt-induced surface expression of corresponding receptors in transfected neurons (Wang et al., 2003b). Curiously, the Akt phosphorylation site of β1–3 subunits is identical with the aforementioned motif in β subunits that regulates clathrin-mediated endocytosis of GABA_ARs. One might therefore conclude that insulin-induced surface expression and function of GABA_ARs reflects reduced clathrin-mediated endocytosis of GABA_ARs. However, insulin-induced potentiation of GABA-evoked currents was completely abolished by pretreatment of neurons with brefeldin A (BFA), an inhibitor of anterograde transport from ER to Golgi (Fujii et al., 2010). In the presence of BFA, insulin induced a modest run-down of GABA-evoked currents, thereby facilitating rather than inhibiting GABA_AR endocytosis. These results are consistent with Akt-mediated insertion of newly synthesized GABA_ARs (and to a lesser extent increased endocytosis of GABA_ARs). Together with the aforementioned studies by Kittler et al. (2005), these findings indicate that phosphorylation of a single site in GABA_AR β subunits can have different effects on trafficking of GABA_ARs depending on the kinase involved, most likely reflecting different subcellular compartments where phosphorylation of GABA_ARs occurs. Interestingly, Fujii et al (2010) found that the effects of insulin on GABA-evoked currents are absent in neurons from PRIP1/2 double knock-out mice. PRIP1 interacts with Akt directly and is required for insulin-induced association of Akt with GABA_ARs. Thus, PRIP serves as a multifunctional adaptor for both kinases and phosphatases and thereby contributes to both regulated membrane translocation and endocytosis of GABA_ARs (Fujii et al., 2010). Interestingly, insulin-mediated activation of Akt further results in inhibitory serine-phosphorylation of GSK3β (Cross et al., 1995). GSK3β in turn promotes postsynaptic GABA_AR clustering and mIPSCs by reducing calpain-1-mediated cleavage of gephyrin (Tyagarajan et al., 2011) (Figures 5A, 6A).

A second proposed mechanism involves direct interaction of the insulin receptor target PI3K with GABA_ARs (Figure 6B) (Vetiska et al., 2007). The PI3K-GABA_AR complex was found to be present constitutively in brain tissue, presumably in an intracellular inactive state. When brain slices were treated with insulin the abundance of this complex rapidly increased as well as its association with phosphorylated membrane lipids. This suggests that the complex is translocated to the plasma membrane in concert with PI3K-mediated phosphorylation of lipids. In vitro analyses revealed that the PI3K-GABA_AR complex involves binding of the PI3K p85 subunit SH2 domain to phospho-tyrosines (Tyr 372/379) in the intracellular loop region of β subunits. These Tyr residues were essential for insulin-induced surface expression of β2-containing receptors in transfected neurons. However, several aspects of this mechanism remain to be resolved. First, it is unclear whether this mechanism applies to GABA_ARs independently of the type of β subunit. Second, it is not known whether Tyr phosphorylation of β subunits is itself modulated by insulin, which Tyr kinase is involved in β subunit phosphorylation, and whether interaction of PI3K with β subunit phospho-tyrosines contributes to activation of PI3K enzyme function. Lastly, it is not known whether and how insulin-induced interaction between PI3K and GABA_ARs corroborates with the aforementioned Akt- and PRIP-dependent downstream effects of PI3K.

BDNF-regulated trafficking of GABA_ARs

Signaling by BDNF and its cognate receptor (receptor tropomyosin-related kinase B, TrkB) is critically important for neurogenesis (Bergami et al., 2008; Li et al., 2008) and inhibitory synapse formation (Seil, 2003). BDNF is further involved in structural and functional neuronal plasticity in the adult nervous system (Xu et al., 2000). At GABAergic synapses, BDNF leads to a rapid and transient increase followed by a lasting reduction in the amplitude of mIPSCs (Brunig et al., 2001; Jovanovic et al., 2004). The lasting reduction in mIPSC amplitude is correlated with reduced surface expression of GABA_ARs (Brunig et al., 2001). Mechanistically, BDNF-induced up- and downregulation of mIPSCs involves a biphasic modulation of the Ser408/409 phosphorylation state of β3 subunits (Jovanovic et al., 2004). Initial rapid phosphorylation is correlated with a transient association of GABA_ARs with PKC and the receptor for activated C-kinase (RACK-1). Subsequent dephosphorylation of the β3 subunit is predominantly mediated by PP2A. As discussed earlier, dephosphorylation of β3 Ser408/409 by PP2A promotes the association of GABA_ARs with AP2, which in turn facilitates clathrin-mediated endocytosis of GABA_ARs (Kittler et al., 2005) and explains the lasting effects of BDNF on GABA_ARs surface expression and mIPSCs. Interestingly, the recruitment of PP2A to GABA_ARs is critically dependent on the phosphatase adaptor PRIP (Kanematsu et al., 2006). Treatment of hippocampal PRIP1/2 double knock-out neurons with BDNF resulted in a steady rise in β3 phosphorylation accompanied by increased GABAergic whole cell currents, indicating that PKC-mediated phosphorylation remained intact while the subsequent PRIP-dependent and PP2A-mediated dephosphorylation step was disrupted (Kanematsu et al., 2006). Thus, PRIP plays essential roles both in BDNF-induced downregulation and insulin-induced potentiation of GABAergic postsynaptic function.

Wnt-regulated cell surface expression of GABA_ARs

Wnt signaling is critically involved in diverse aspects of embryonic development, neural differentiation and adult synaptic plasticity (reviewed by Inestrosa and Arenas, 2010; Budnik and Salinas, 2011). Wnt proteins encoded by 19 different genes act through several different frizzled family receptors to induce multiple signal transduction pathways. The canonical Wnt pathway involves inhibition of GSK3β in the axin/GSK3β/APC complex, which leads to accumulation and nuclear translocation of β-catenin and activation of β-catenin-dependent gene expression. By contrast, two non-canonical Wnt pathways activate either c-Jun N-terminal kinase (Wnt/JNK pathway) or CaMKII (Wnt/Ca²⁺ pathway) as downstream targets. All three pathways are implicated in the regulation of synaptic plasticity, primarily of excitatory synapses and both pre- and post synaptically (Inestrosa & Arenas, 2010). In addition, Wnt-5a was recently shown to result in rapid (5 min) and significant (+40%) upregulation of GABA_AR clusters in cultured neurons (Cuitino et al., 2010). This effect was due to postsynaptic changes as it was paralleled by increased amplitudes but not frequency of mIPSCs recorded from cultured neurons. Consistent with this interpretation, the time course and paired-pulse relationship of evoked IPSCs recorded from hippocampal slices were unaffected by Wnt-5a. The effect of Wnt-5a on mIPSCs was further potentiated by blockade of endocytosis with a dynamin inhibitor peptide. In addition, Wnt-5a treatment reduced the pool of previously surface biotinylated and internalized GABA_ARs, suggesting that increased clustering of GABA_ARs reflected enhanced recycling of endocytosed receptors. In support of this mechanism, treatment of neurons with a Wnt-5a-mimicking peptide (Foxy-5) that specifically activates non-canonical Wnt pathways replicated the Wnt-5a effect on GABA_AR clustering. Moreover, co-treatment with Foxy-5 and pathway-specific pharmacological inhibitors allowed the conclusion that Wnt-5A-induced clustering of GABA_ARs involved the non-canonical Wnt/Ca²⁺ pathway and CaMKII. The CaMKII targets that are phosphorylated in response to Wnt-5a have so far not been determined. In addition to the Wnt/Ca²⁺ pathway the canonical Wnt pathway is strongly implicated in the regulation of GABAergic inhibition by the aforementioned effects of Li⁺ and GSK3β on the stability and postsynaptic clustering of gephyrin (Tyagarajan et al., 2011). However, in apparent conflict with this study, the canonical Wnt ligand Wnt-7A and Li⁺ had no significant effect on GABA_AR clustering in the study by Cuitino et al (2010).