Epilepsy, E/I Balance and GABAA Receptor Plasticity

GABAA receptors mediate most of the fast inhibitory transmission in the CNS. They form heteromeric complexes assembled from a large family of subunit genes. The existence of multiple GABAA receptor subtypes differing in subunit composition, localization and functional properties underlies their role for fine-tuning of neuronal circuits and genesis of network oscillations. The differential regulation of GABAA receptor subtypes represents a major facet of homeostatic synaptic plasticity and contributes to the excitation/inhibition (E/I) balance under physiological conditions and upon pathological challenges. The purpose of this review is to discuss recent findings highlighting the significance of GABAA receptor heterogeneity for the concept of E/I balance and its relevance for epilepsy. Specifically, we address the following issues: (1) role for tonic inhibition, mediated by extrasynaptic GABAA receptors, for controlling neuronal excitability; (2) significance of chloride ion transport for maintenance of the E/I balance in adult brain; and (3) molecular mechanisms underlying GABAA receptor regulation (trafficking, posttranslational modification, gene transcription) that are important for homoeostatic plasticity. Finally, the relevance of these findings is discussed in light of the involvement of GABAA receptors in epileptic disorders, based on recent experimental studies of temporal lobe epilepsy (TLE) and absence seizures and on the identification of mutations in GABAA receptor subunit genes underlying familial forms of epilepsy.


INTRODUCTION
The convulsant effects of GABA and glycine receptor antagonists, and conversely the clinically relevant antiepileptic action of classical benzodiazepines, such as diazepam, led to the concept that epileptic seizures refl ect an imbalance between excitatory and inhibitory transmission in the brain (Bradford, 1995;Gale, 1992;Olsen and Avoli, 1997). This view was further supported by the strong epileptogenic effects of glutamate receptor agonists, in particular kainic acid (Ben-Ari et al., 1980;Sperk, 1994). Unlike acute drug effects, which occur in an intact system, epileptogenesis and recurrent seizures in chronic epilepsy likely refl ect pathological disturbances of neuronal circuits that may have multiple origins. Furthermore, the simple view that GABAergic transmission acts like a break preventing overexcitation of neuronal circuits has been challenged by the highly sophisticated anatomical and functional organization of GABAergic interneurons in cerebral cortex (Blatow et al., 2005;Markram et al., 2004). Rather, GABAergic function is required for fi netuning of neuronal circuits and its infl uence on cell fi ring and network oscillations is constrained spatially and temporally (Mann and Paulsen, 2007;Tukker et al., 2007). Furthermore, GABAergic transmission, while typically qualifi ed as being inhibitory, can also be depolarizing, even under physiological conditions, in the adult brain (Gulledge and Stuart, 2003;Szabadics et al., 2006). Finally, in epileptic tissue neuronal network undergo extensive rewiring that considerably changes the function of interneurons and their control over pyramidal cells (Cossart et al., 2005;Ratte and Lacaille, 2006). Therefore, the classical dichotomy between inhibitory and excitatory GABAergic/glutamatergic transmission has to be revised and the role of GABAergic transmission in epilepsy is much more complex than suggested by simple pharmacological experiments.
The purpose of this review is to summarize recent advances on the concept of E/I balance and its relevance for epilepsy and to discuss the signifi cance of GABA A receptor heterogeneity for the pathophysiology of epileptic disorders.

EPILEPSY
Epilepsy is a generic term encompassing multiple syndromes, with distinct symptoms, etiology, prognosis, and treatments. The role of GABA A receptors in the pathophysiology of epilepsy has been examined experimentally in most detail in two major diseases, namely absence epilepsy and temporal lobe epilepsy (TLE). In addition, the functional consequences of mutations associated with familial forms of epilepsies are now being analyzed in recombinant expression system and in vivo using transgenic mouse models carrying these mutations (Noebels, 2003).
Absence seizures can be genetically determined (GAERS, WAG/Rij rats) (van Luijtelaar and Sitnikova, 2006) or pharmacologically-induced, for example by treatment with a cholesterol biosynthesis inhibitor, AY-9944 (Snead, 1992). They are characterized by low frequency spike-and-wave discharges refl ecting impaired thalamo-cortical function. Typically, they are aggravated by benzodiazepine agonists. TLE is mimicked by induction of a prolonged status epilepticus (either upon repetitive electrical stimulation of sensitive regions of the temporal lobe or by injection of a convulsant, such as kainic acid or pilocarpine), which is followed in most cases by the occurrence of spontaneous recurrent seizures (reviewed in Coulter et al., 2002). Kindling, either electrical or chemical, is also used to model TLE, with the major difference that the animals do not present with recurrent seizures. In both TLE and absence epilepsy, alterations of GABA A receptor expression, pharmacology, and functional properties 2 have been studied in detail over many years. Recent results highlighting novel features will be discussed in more detail in the sections "Phasic/ tonic inhibition" and "GABA A receptor plasticity in epilepsy". Importantly, changes observed in experimental models need to be compared to alterations taking place in the brain of TLE patients. The availability of tissue resected at surgery from patients with intractable epilepsy represents an invaluable source for understanding the pathophysiology of the disorder (Loup et al., 2006;Magloczky and Freund, 2005).

GABA A RECEPTORS AND EXCITATORY/ INHIBITORY (E/I) BALANCE
The concept of E/I balance has gained much weight following the discovery of homeostatic synaptic plasticity, through which the level of activity of neuronal networks is maintained within a narrow window by locally adapting the strength and weight of synaptic transmission in response to external stimuli (Marder and Goaillard, 2006;Rich and Wenner, 2007;Turrigiano, 2007). Major factors contributing to homeostatic synaptic plasticity include intrinsic membrane properties of pre-and postsynaptic neurons, patterns of synaptic inputs, non-synaptic interactions with neighboring cells, including glial cells, ionic composition of the extracellular fl uid, and hormonal infl uences. Implicitly, an altered E/I balance, frequently postulated as mechanism underlying epileptogenesis and seizure generation, postulates a disturbance in homeostatic plasticity resulting from either insuffi cient or excessive compensatory mechanisms in response to a change in network activity.
In this review, we focus on three main factors underlying the contribution of GABA A receptors for homeostatic synaptic plasticity (Mody, 2005). The fi rst of these factors is tonic inhibition, mediated primarily by extra-or perisynaptic receptors. Although tonic inhibition typically is evidenced in patch clamp recordings by a reduction in holding current (typically 10-100 pA) upon application of a GABA A receptor antagonist, it represents a signifi cant fraction of GABA-mediated charge transfer and is therefore likely to have a strong impact on neuronal excitability. The second factor is the regulation of Cl − ion fl uxes upon opening of GABA A receptors, which are determined by specifi c potassium-chloride co-transporters such as NKCC1 and KCC2. In addition of being developmentally regulated (Rivera et al., 2005), these co-transporters undergo rapid changes in expression and function under pathological conditions, leading to chronic dysregulation of GABAergic inhibition (Price et al., 2005). The third factor is the activity-dependent regulation of GABAergic and glutamatergic synapse function, recently brought to light by systematic analyses of the effects of chronic epileptiform activity or axon potential blockade in vitro (Costantin et al., 2005;Marty et al., 2004;Rutherford et al., 1997).

Phasic/tonic inhibition
Phasic and tonic neurotransmission are used to discriminate between the short, spatially restricted action of transmitters activating postsynaptic receptors, and the continuous activation of receptors localized peri-or extrasynaptically by transmitter spillover into the extracellular space. Given the prolonged duration of tonic transmission compared to the short openings of ion channels, most of the total charge transported by ligandgated ion channels occurs via tonic transmission, suggesting a major role in modulating neuronal activity (Farrant and Nusser, 2005;Mody and Pearce, 2004;Semyanov et al., 2004).
The subunit composition of GABA A receptors appears to be a major determinant of phasic and tonic GABAergic transmission. The γ2 subunit, which is present in the vast majority of GABA A receptor subtypes, is required for postsynaptic clustering of GABA A receptors and gephyrin (Essrich et al., 1998;Luscher and Fritschy, 2001;Schweizer et al., 2003), a cytoskeletal protein selectively concentrated in GABAergic and glycinergic synapses in the CNS (Sassoè-Pognetto and Fritschy, 2000;Triller et al., 1985). Multiple GABA A receptor subtypes are clustered at postsynaptic types in defi ned neuronal populations (α1-, α2-, α3-, and, in part, α5-GABA A receptors). A major additional property of the γ2 subunit is to confer diazepam sensitivity to receptor complexes containing these α subunit variants. It is important to note, however, that GABA A receptors containing the γ2 subunit are not only confi ned to postsynaptic sites. For instance, most α5-GABA A receptors are extrasynaptic (Crestani et al., 2002), contribute to tonic inhibition modulated by diazepam (Caraiscos et al., 2004;Glykys and Mody, 2006;Prenosil et al., 2006) and regulate the excitability of pyramidal cells (Bonin et al., 2007). Consequently, postsynaptic and extrasynaptic receptors formed with the γ2 subunit are diazepam-sensitive and contribute to the pharmacological profi le of classical benzodiazepine-site agonists.
Receptors containing the δ subunit, in contrast to the γ2 subunit, appear to be excluded from postsynaptic sites, as demonstrated by immunoelectron microscopy (Nusser et al., 1998;Wei et al., 2003). The δ subunit is associated mainly with the α4 subunit, e.g., in thalamus and dentate gyrus (Peng et al., 2002;Sun et al., 2004), or the α6 subunit in cerebellar granule cells (Jones et al., 1997). These receptors are diazepam-insensitive (Kapur and Macdonald, 1996;Makela et al., 1997) but are selectively modulated by GABA agonists such as gaboxadol and muscimol (Drasbek and Jensen, 2005;Storustovu and Ebert, 2006) as well as neurosteroids (Belelli and Herd, 2003;Belelli et al., 2005;Stell et al., 2003), pointing to possible novel target for drug therapy (Krogsgaard-Larsen et al., 2004). Importantly, GABA A receptors containing the α4 and/or δ subunit exhibit unique functional properties that may contribute to epileptogenesis and recurrent seizures upon altered expression, as occurs in TLE (Lagrange et al., 2007). δ subunit-null mice exhibit enhanced sensitivity to pentylenetetrazol-induced seizures, but it is not established whether this refl ects a reduced tonic inhibition in thalamocortical circuits or a reduced availability of binding sites for endogenous neurosteroids with anticonvulsant activity (Spigelman et al., 2002).
Ectopic expression of the α6 subunit under the control of the Thy-1.2 promoter has been used to assess the functional and pharmacological signifi cance of enhanced tonic inhibition. These transgenic mice overexpress α1/α6/β/γ2-GABA A receptors and exhibit a fi ve-fold increase in tonic inhibition in CA1 pyramidal cells (Wisden et al., 2002). Behaviorally, these mice are essentially normal, but are more sensitive than wild-type to the convulsant effects of GABA A receptor antagonists (Sinkkonen et al., 2004), suggesting an imbalance between phasic and tonic inhibition, with an overall decrease in GABAergic synaptic strength.

Chloride ion homeostasis
A major facet of GABAergic transmission is the intimate link between GABA A receptor function and ion homeostasis. Therefore, multiple ATPdependent transport processes determine GABAergic signaling (Farrant and Kaila, 2007). GABA A receptors are primarily permeable to Cl − and HCO3 − anions. Cl − gradients are determined by two major pumps acting in opposite fashion, NKCC1 and KCC2, whereas bicarbonate is produced by carbonic anhydrases. The relative expression level of these molecules changes markedly during development, thereby rendering the reversal potential of Cl − more negative Kaila, 2007, Fiumelli andWoodin, 2007;Rivera et al., 2005). An opposite change in Cl − reversal potential affecting GABA A -mediated transmission has been suggested to occur in neurological disorders and following brain trauma (De Koninck, 2007, Huberfeld et al., 2007Payne et al., 2003;Toyoda et al., 2003), due to reduced expression or function of KCC2. It should be emphasized, however, that it remains largely unclear, whether GABA A -mediated depolarization after down-regulation of KCC2 has an excitatory effect on postsynaptic neurons, or whether shunting inhibition or deactivation of voltage-gated Na + channels predominate after GABA A receptor activation, resulting functionally in inhibition.
In any case, KCC2 expression and function are regulated on a short-time basis by activity-dependent mechanisms (Rivera et al., 2004), determined in particular changes in phosphorylation (Lee et al., 2007;Wake et al., 2007) and by the short half-life of this transporter. These specifi c properties of KCC2 provide a major tool for local and rapid adjustments of Cl − ion fl uxes, and therefore network activity, in adult brain.

Activity-dependent changes in synaptic structure and function
Pharmacological enhancement or blockade of neuronal activity in vitro represents a simplifi ed model of epileptiform activity or removal of afferents (as would happen after a lesion), respectively. The effects are evident on the molecular, functional, and structural level. To mention a few examples, enhanced activity has marked effects on postsynaptic receptor mobility, refl ecting diffusion within the plasma membrane. The effect is Ca ++ -dependent and likely mediated by interactions with the actin cytoskeleton (Hanus et al., 2006). The induction of epileptiform activity in vitro by application of GABA A receptor antagonists regulates synaptic function in hippocampal neurons by selectively favoring the loss of synapses on spines but not on dendritic shafts, resulting in increased GABAergic inhibition (Zha et al., 2005). In contrast, activity deprivation in hippocampal slices can induce epileptiform discharges (Trasande and Ramirez, 2007) and, during development, markedly affects the balance between glutamatergic and GABAergic synapses (Marty et al., 2000). Although multiple mechanisms are involved in these changes, neurotrophins and extracellular matrix proteins, including integrins or cell-adhesion molecules, for example, play an important role in synaptic plasticity and remodeling induced by chronic changes in network activity (Gall andLynch, 2004, Kuipers andBramham, 2006).

GABA A RECEPTOR PLASTICITY
Several mechanisms contribute to the dynamic regulation of GABA A receptor function, which is essential for fi ne tuning of neuronal networks and the generation of rhythmic activities under physiological conditions: 1. Regulation of GABA A receptor traffi cking, synaptic clustering, and cell-surface mobility (Kittler and Moss, 2001;Kneussel and Loebrich, 2007;Thomas et al., 2005). In particular, GABA A receptor internalization mediated by clathrin-coated vesicle endocytosis (Herring et al., 2003, van Rijnsoever et al., 2005 has emerged as a major mechanism of short-and long-term plasticity of GABAergic synapses. Unlike AMPA receptors (Man et al., 2000), GABA A receptor internalization is not triggered by agonist exposure but is regulated by phosphorylation (Kanematsu et al., 2006). In addition, several tyrosine kinase receptor ligands, such as TNF-α, insulin, or BDNF also modulate GABA A receptor cell surface expression by regulating its rate of internalization and/or membrane insertion (Brünig et al., 2001;Gilbert et al., 2006;Jovanovic et al., 2004;Wan et al., 1997). Next, synaptic clustering of GABA A receptors is largely inter-dependent on the scaffolding protein gephyrin. Thus, down-regulation of gephyrin expression by gene targeting or silencing leads to rapid disappearance of postsynaptic GABA A receptor clustering and loss of IPSCs (Essrich et al., 1998;Yu et al., 2007). Finally, cell surface mobility, refl ecting membrane diffusion, represents a major mechanism for the dynamic, short-term regulation of GABA A receptors available for synaptic transmission (Thomas et al., 2005). 2. Regulation of receptor functions by chemical modifi cation, with phosphorylation being one of the major covalent modifi ers. Increasing evidence indicates that chemical modifi cation affects receptor traffi cking and cell surface expression, as well as intrinsic functions of the ligand-gated ion channel (Kittler and Moss, 2003;Hinkle and Macdonald, 2003). GABA A receptor palmitoylation, selectively of the γ2 subunit, represents an additional mechanism for regulation of traffi cking, cell-surface expression and postsynaptic clustering (Keller et al., 2004;Rathenberg et al., 2004). 3. Regulation of subunit expression, at the transcriptional and translational level (Steiger and Russek, 2004); this mechanism determines the abundance and subunit composition of GABA A receptors in a given cell type or brain region and is of particular relevance for physiological alterations of network function, such as occurring upon hormonal fl uctuations during the ovarian cycle (Brussaard and Herbison, 2000;Maguire et al., 2005) and during puberty (Shen et al., 2007).

Changes in subunit composition
Alterations in GABA A receptor subunit expression and composition in epilepsy are well documented in human (Loup et al., 2000(Loup et al., , 2006 and in animal models (Gilby et al., 2005;Li et al., 2006;Nishimura et al., 2005;Peng et al., 2004;Roberts et al., 2005). The latter studies extend previous work by demonstrating a major contribution of extrasynaptic GABA A receptors to the changes in inhibitory function that might underlie epileptogenesis and occurrence of chronic recurrent seizures. For example, in the mouse pilocarpine model of TLE, a profound decrease in δ subunit immunoreactivity was observed, correlating with a redistribution of the γ2 subunit from synaptic to perisynaptic sites, where it assembled with the α4 subunit, which is normally associated with the δ subunit (Zhang et al., 2007). A down-regulation of the α5 subunit also occurs in CA1 pyramidal cells of pilocarpine-treated rats (Houser and Esclapez, 2003), resulting in a loss of diazepam-sensitive tonic inhibition seen upon blockade of GABA reuptake (Scimemi et al., 2005). Despite this change, tonic inhibition is enhanced in pyramidal cells, suggesting compensatory up-regulation of other extrasynaptic GABA A receptors, possibly containing the α4 subunit. Quite recently, region-specifi c changes in GABA A receptor function and expression have been reported in models of absence epilepsy (Bessaih et al., 2006;Li et al., 2006;Liu et al., 2007). In the pharmacological model of absence seizures induced by neonatal treatment with the cholesterol biosynthesis inhibitor AY-9944, a reduced expression of the α1 and γ2 subunit has been reported (Li et al., 2006), with distinct sex differences and temporal profi les, correlating with the higher incidence of absence seizures in female rats (Li et al., 2006). Electrophysiologically, in the GAERS strain, GABA A receptor-mediated currents are altered selectively in the thalamic reticular nucleus, but not in ventrobasal complex or somatosensory cortex, with mIPSCs exhibiting enhanced amplitude and reduced decay kinetics (Bessaih et al., 2006). Such changes might be accounted for by expression of the β1 subunit (Huntsman and Huguenard, 2006). Finally, in WAG/Rij rats, a loss of GABA A receptor α3 subunit-immunoreactivity has been shown to occur without alteration in mRNA expression in the reticular thalamic nucleus (Liu et al., 2007), suggesting a local and highly specifi c defi cit in GABA A receptor function as a possible cause of absence seizures in these mutant rats.

Mutations affecting GABA A receptor assembly and traffi cking
Several mutations in GABA A receptor subunits have been associated with familial idiopathic epilepsies, including childhood absence epilepsy (CAE), generalized epilepsy with febrile seizures plus (GEFS+) and juvenile myoclonic epilepsy (JME) (Heron et al., 2007;Noebels, 2003). Missense and frame shift mutations in the GABA A receptor α1 subunit gene (GABRA1; 5q34) are associated with JME (Cossette et al., 2002) and childhood absence epilepsy (CAE) (Maljevic et al., 2006). By contrast, missense, splice site mutations, or deletions in the γ2 subunit gene (GABRG2; 5q34) have been found in families with GEFS+ and CAE with febrile seizures (Audenaert et al., 2006;Baulac et al., 2001;Harkin et al., 2002;Kananura et al., 2002;Wallace et al., 2001). In recombinant expression systems, these missense mutations typically affect single channel gating and/or cell surface availability of GABA A receptors. The precise mechanism underlying seizure generation remains in most cases ill-defi ned. The GABA A receptor γ2 subunit R43Q mutation has been reported to impair assembly and cell surface expression of GABA A receptors (Baulac et al., 2001;Bowser et al., 2002). The mutation causes an increase in intracortical excitability in patients compared to unaffected relatives (Fedi et al., 2007). Intriguingly, the effect of the mutation was shown to be temperature-dependent, with cell surface expression being reduced in vitro at temperatures higher than 37ºC . However, since most GEFS+ patients do not carry this mutation, such a mechanism alone is not suffi cient for explaining the onset of seizures. In fact, other studies have shown that the γ2(R43Q) mutation affects GABA A receptor cell surface traffi cking and subunit composition independently of temperature (Frugier et al., 2007). The reduction of cell surface expression mainly affects extrasynaptic receptors containing the α5 subunit, without altering phasic inhibition mediated by synaptic GABA A receptors (Eugène et al., 2007). In the same study, these effects were contrasted to the γ2(K289M) mutation, which accelerates decay kinetics of miniature and evoked postsynaptic inhibitory currents, but does not affect GABA A receptor traffi cking and cell surface expression. Another mutation, α1(A322D), is characterized by reduced subunit expression due to enhanced proteosomal degradation, probably due to protein misfolding (Gallagher et al., 2007). Finally, two susceptibility variants (E177A and A220H) have been found in the δ subunit gene (GABRD; present primarily in extrasynaptic GABA A receptors, see section "Phasic/tonic inhibition"), affecting channel kinetics and cell surface expression in recombinant systems (Dibbens et al., 2004;Feng et al., 2006). However, no segregation of A220H with epilepsy could be found in a subsequent analysis of a large family (Lenzen et al., 2005), and the signifi cance of these mutations remains to be established.

CONCLUSIONS AND PERSPECTIVES
The heteregeneous molecular structure of GABA A receptors and their differential expression, traffi cking, localization, and function underscore their complex regulation. They contribute in multiple ways to the maintenance of E/I balance and the pathophysiology of epilepsy. Consequently, much work remains to be done to conceive therapeutic applications exploiting specifi c facets of GABA A receptor heterogeneity. So far, data sets obtained with different methods cannot be integrated into a single coherent picture, and multidisciplinary approaches will be required to grasp the signifi cance of GABA A receptors in homeostatic synaptic plasticity. The postulated "imbalance" between synaptic excitation and inhibition has been a motor for studying the functional properties of GABAergic and glutamatergic synapses in great detail. However, it is too simple a model for allowing conceptual advances about the pathophysiology of complex brain diseases, such as epilepsy disorders. While the present review focused solely on GABA A receptors, it is evident that other mechanisms contributing to synaptic homeostasis will have to be included in a global concept as a prerequisite for understanding and preventing epileptogenesis and ictogenesis. Yet, the central role played by GABAergic transmission in the regulation of neuronal networks justifi es the current interest given to GABA A receptors in studies of epilepsy.

CONFLICT OF INTEREST STATEMENT
The author declares that the research was conducted in the absence of any commercial or fi nancial relationships that could be construed as a potential confl ict of interest.

ACKNOWLEDGEMENT
The author's own research was supported by the Swiss National Science Foundation and the NCCR-Neuro.