GABA_A Receptor Subtypes

GABA AND COGNITIVE DYSFUNCTIONS

Cognition is an umbrella term that covers interrelated mental activities including attention, learning and memory, understanding, reasoning, social interaction, and goal-directed operations such as planning and decision making. Cognitive dysfunctions are highly relevant across psychiatric disorders (Figure 1) (1). In cortical operations, a large diversity of cells endows the neuronal circuits with the capacity to perform complex processes. In hippocampus, for instance, excitatory pyramidal cells encode spatial and other episodic memories. In these circuits, the relatively uniform pyramidal cells are supported by a rich diversity of GABAergic interneurons that regulate pyramidal cell activity in a domain-specific fashion. The somatic domain, the axon initial segment, and the dendrites each receive distinct GABAergic innervation with independent timing. The interneurons imprint a spatiotemporal conductance matrix on the pyramidal cells that supports synaptic temporal dynamics, network oscillations, and the selection of cell assemblies. These spatiotemporal specializations provide a basis for cognitive behavior (2–4).

GABA_A Receptors and the Temporal Dynamics of Network Operations

The diversity of GABAergic interneurons is paralleled by a diversity of GABA_A receptors that display structural, functional, and positional specifications geared toward the requirement of the respective synapse operation. The GABA_A receptor subtypes are therefore pharmacological targets that provide opportunities for modulating the spatiotemporal pattern of network activity. For instance, one population of GABA interneurons, the fast-spiking parvalbumin (PV)-positive basket cells, release neurotransmitter at somatic pyramidal cell synapses containing GABA_A receptors that are geared toward high-frequency γ-range responses: α₁-subunit-containing GABA_A receptors, also known as α₁ GABA_A receptors or simply α₁ receptors. In contrast, the cholecystokinin (CCK)-containing basket cells---which, like the PV-positive interneurons, target the perisomatic compartment---are not fast spiking and operate with structurally different receptors: α₂- and α₃-subunit-containing GABA_A receptors, also known as α₂ and α₃ GABA_A receptors or simply α₂ and α₃ receptors. Interneurons that preferentially target dendritic sites of pyramidal cells, such as Martinotti cells, influence the synaptic input by affecting input integration, synaptic plasticity, and the generation of calcium spikes. The output of cortical and hippocampal pyramidal cells can be specifically regulated by axo-axonic GABA interneurons (Chandelier neurons) operating largely with α₂-subunit-containing receptors. Sir John Eccles famously wrote, “I always think that inhibition is a sculpturing process. The inhibition, as it were, chisels away at the ... mass of excitatory action and gives a more specific form to the neuronal performance at every stage of synaptic relay” (5, p. 92).

Enhancing γ Oscillations via Extrasynaptic GABA_A Receptors

Extrasynaptic receptors are activated by synaptic spillover and enable neurons to sense low ambient GABA in order to generate tonic inhibition, which is capable of altering oscillatory network behavior and cognitive processes (6, 7). Tonic inhibition of cortical neurons (layer 5) and hippocampal CA3 and CA1 pyramidal cells is mediated predominantly by α₅-subunit-containing GABA_A receptors (also known as α₅ GABA_A receptors or simply α₅ receptors) (Figure 2) (8--11). Through their location at the bases of the spines and the adjacent shafts of the dendrites (12) (Table 1), α₅ GABA_A receptors are in a privileged position to modulate the excitatory input arising at the spines via N-methyl-D-aspartate (NMDA) receptors (7, 13, 14). In neocortex, in addition to extrasynaptic α₅ receptors, some synaptic α₅ receptors respond to the input from bitufted interneurons on pyramidal cells (15). α₅ receptors affect the dynamic range of hippocampal γ oscillations (16, 17). Larger-amplitude γ oscillations have been observed in hippocampal slices from α₅^–/– mice compared with wild-type mice, with no change in frequency (19, 20). In α₅^–/– mice, no compensatory upregulation of other α subunits was apparent apart from a residual tonic inhibition that was attributed to a homeostatic upregulation of δ GABA_A receptors (13). γ oscillation frequency can also be modulated by extrasynaptic δ GABA_A receptors located on GABA interneurons in CA3 (6). Tonic inhibition is mediated by δ-subunit-containing GABA_A receptors (e.g., α₄βδ, α₆βδ, or α₁βδ) in other neuronal cell types such as granule cells in the dentate gyrus and cerebellum, thalamic relay neurons, neocortical pyramidal cells, and striatal medium spiny neurons (Table 1) (7). Thus, tonic inhibition, in particular via α₅ GABA_A receptors, plays a major role in shaping network synchronization (7).

Enhancing Cognitive Behavior by Genetic Targeting of α₅ GABA_A Receptors

A partial knockdown of hippocampal α₅ receptors [in α₅(H105R) mice] improved hippocampus-dependent performance, as shown by trace fear conditioning, appetitive conditioning, and novel object recognition (19, 20, 20a). Similarly, mice with a full deficit of α₅ receptors (α₅^–/– mice) showed improved spatial performance (in the Morris water maze) and trace fear memory (14, 21). In contrast, performance in delay fear conditioning, contextual fear conditioning, and two-way avoidance was unaltered (14, 20, 20a, 21), as was hippocampal long-term potentiation (LTP) (20a, 21) except under specific stimulation conditions (14). These results put α₅ GABA_A receptors at center stage as targets for pharmacological modulation of learning and memory.

Restoring Memory Deficits with α₅ GABA_A Receptor Inverse Agonists

In keeping with the results from genetic targeting, partial inverse agonists (also known as negative allosteric modulators or NAMs) acting at the benzodiazepine site of a α₅ GABA_A receptor were expected to improve performance in learning and memory (22). L-655708, a partial inverse agonist with preferential affinity for α₅ GABA_A receptors, reduced tonic inhibition, facilitated LTP, induced γ oscillations in CA3, and enhanced performance in the Morris maze without being proconvulsant (13, 23, 24). It also reduced alcohol consumption in rhesus monkeys (25). However, the anxiogenic activity of L-655708, presumably due to its inverse agonistic activity at GABA_A receptors other than α₅ receptors (see 24 for a review), prevented its use in humans.

Another partial inverse agonist acting at the α₅ GABA_A receptor, α₅IA, displayed preferential efficacy at (but not affinity for) α₅ GABA_A receptors compared with α₁, α₂, and α₃ GABA_A receptors (Figure 2). It enhanced spatial memory (encoding and recall in the Morris water maze) without anxiogenic activity (24, 26--28). In healthy volunteers, administration of α₅IA reduced ethanol-induced amnesia (29). Unfortunately, preclinical renal toxicity prevented further clinical development. MRK-016, another inverse agonist acting at the α₅ GABA_A receptor (but with greater intrinsic efficacy than α₅IA), was poorly tolerated in the elderly and was not further developed (reviewed in 24).

The imidazotriazolobenzodiazepine RO4938581 was the first partial inverse agonist with a highly selective preference in both affinity for and efficacy at α₅ GABA_A receptors versus α₁, α₂, and α₃ GABA_A receptors (30). RO4882224 displayed a similar receptor profile with slightly less preferential affinity for α₅ GABA_A receptors (31). Both compounds enhanced hippocampal LTP in vitro (24, 31). Behaviorally, both compounds rescued a scopolamine-induced hippocampus-dependent working memory deficit (delayed matching-to-position task) as well as a diazepam-induced spatial learning deficit (Morris water maze) in rodents (30, 31). Cognitive deficits induced by phencyclidine (object recognition, odor discrimination) were attenuated by RO4938581 (32). Furthermore, in cynomolgus monkeys, the performance of an object retrieval task, which is thought to require executive functions, was enhanced by RO4938581 beyond the physiological performance level (30). Most important, RO4938581 showed no anxiogenic potential (elevated plus maze, social approach avoidance) in rats and no proconvulsive potential (audiogenic seizures), underlining the critical importance of subtype selectivity in binding affinity and functional efficacy. A compound related to RO4938581 was entered into clinical trials in subjects with Down syndrome (see below). Thus, such agents hold the promise of novel treatments for neurological and psychiatric disorders involving cognitive dysfunctions and possibly conditions of hypofrontality such as schizophrenia (33).

Suppressing Memory via α₅ GABA_A Receptors

The modulation of memory performance via α₅ GABA_A receptors is bidirectional. Positive allosteric modulators or an upregulation of α₅ GABA_A receptors impaired cognitive performance (32, 34, 35). Low-dose etomidate increased tonic inhibition, reduced LTP, and impaired memory performance (spatial memory, contextual fear memory) in wild-type but not α₅^–/– mice, whereas its sedative/hypnotic properties were retained in α₅^–/– mice (34, 36). Etomidate acts at a site independent of the benzodiazepine site of a GABA_A receptor (37). Following etomidate anesthesia, α₅ GABA_A receptors may contribute to postoperative cognitive and memory dysfunction (36).

Increasing the surface expression of α₅ GABA_A receptors is similarly associated with memory loss, as seen in acute systemic inflammation (35). Treatment with the inflammatory cytokine IL-1β or the endotoxin lipopolysaccharide induced memory impairment (contextual fear memory) in wild-type but not in α₅^–/– mice, whereas hippocampus-independent cued fear conditioning was unaffected. Interleukin-1β (IL-1β) increased tonic inhibition (through a p38 MAP kinase pathway), reduced LTP in CA1, and induced upregulation of the α₅-subunit protein level in hippocampus, whereas the subunit levels of α₂ and α₁ were unaltered or slightly decreased, respectively (35). In addition to the increased expression of α₅ GABA_A receptors (35), an increase in the extracellular GABA level has also been reported in inflammatory states (reviewed in 35).

Attempts to develop agonists that act selectively at the benzodiazepine site of a α₅ GABA_A receptor have found limited success so far. The seven-acetylene congener of diazepam, QH-II-066, which displays moderately preferential affinity for (11-fold versus α₁, 6-fold versus α₂ and α₃) (38) and partial agonistic activity at α₅ GABA_A receptors (39), impaired cognitive performance (object recognition) in rats at nonsedative doses (32). However, anticonvulsant activity had been reported earlier in the same dose range in mice (39). The imidazophenylbenzodiazepine SH-053-2′FR-CH3 (also termed SH-053-R-CH3-2′F) is another α₅-preferring partial agonist (40) that attenuates burst activity in a cultured neocortical slice (41) but produces sedation in some species (40). Clearly, more selective tools would be welcome. For studies in humans, a positron emission tomography ligand is available for imaging α₅ receptor occupancy (42).

Potential Adjuvant to Cognitive Behavioral Therapy?

Enhancing α₅ GABA_A receptor functions in hippocampus may potentially contribute to strategies aimed at erasing previously formed memory in anxiety disorders such as phobias and posttraumatic stress disorder. Memories for fear-associated cues and context are reconsolidated in the amygdala and the hippocampus, respectively (43, 44). In animal models of fear extinction, D-cycloserine, a cognitive enhancer acting as an NMDA receptor coagonist, enhanced fear extinction in a context-, dose-, and time-dependent manner (reviewed in 45). This finding was reproduced in humans, whereby administration of D-cycloserine improved extinction of phobia symptoms in a virtual reality setting (46). It is tempting to speculate that the same behavioral shift may be achieved by a selective reduction of tonic hippocampal inhibition through partial inverse agonists of α₅ GABA_A receptors.

DOWN SYNDROME

THE QUEST FOR SELECTIVE ANXIOLYTICS

DEPRESSION

Depression and related mood disorders are among the greatest public health problems today. The US lifetime prevalence of major depressive disorder (MDD) is 17% (124). Symptoms of depression include depressed mood, a reduced ability to experience reward (anhedonia), feelings of worthlessness or guilt, poor concentration, indecisiveness, thoughts of death, suicidal ideation, appetite/weight changes, sleep disturbances, changes in psychomotor activity, and fatigue (125).

Although the neurobiology of depression has not been fully elucidated, increasing experimental and clinical evidence points to an association between MDD and GABAergic deficits (125a). Patients with MDD have reduced central nervous system (CNS) GABA concentrations (126, 127), which play a prominent role in the neural regulation of stress (125a). Preclinical evidence suggests that clinically used antidepressants exert their actions by ultimately counteracting GABAergic deficits (125a). Additionally, GABAergic neurotransmission plays an important role in hippocampal neurogenesis and neural maturation, both of which have been established as substrates of antidepressant therapies (128). Furthermore, mice with a heterozygous deletion of the γ₂ subunit (γ₂^+/− mice) exhibit trait anxiety (129) as well as heightened behavioral inhibition in response to behavioral despair, as evidenced by increased immobility in the forced swim test (FST) and in the tail suspension test (TST) (128). Desipramine, but not fluoxetine, normalized these deficits (130). γ₂^+/−₂ mice also have hypothalamic-pituitary-adrenal axis hyperactivity and antidepressant drug sensitivity reminiscent of melancholic forms of depression (130). Additionally, adult hippocampal neurogenesis was reduced in γ₂^+/− mice (128). In addition, the antidepressant effect of BDNF is mediated via GABA_A receptors (130a). These and other studies led to the “GABAergic deficit hypothesis of major depressive disorder” proposed by Lüscher and colleagues (125a).

With the γ₂ subunit contained in approximately 90% of all GABA_A receptors (131), the question arises as to which receptor subtype(s) as defined by the α subunit may be mediating the antidepressant-like actions. α₂ global knockout mice displayed increased immobility in the FST and TST, indicating a depressive-like phenotype (132). A major feature of depression is anhedonia, which can be modeled in rodents using the ICSS paradigm. Mice are implanted with an electrode into the medial forebrain bundle at the level of the lateral hypothalamus and trained to spin a wheel for rewarding stimulation, so that frequency-response curves can be constructed (133). Like cocaine, diazepam shifts these frequency-response curves to the left; i.e., it has a reward-enhancing action (110). In α₂(H101R) and α₃(H126R) mice with diazepam-insensitive α₂ and α₃ GABA_A receptors, respectively, the reward-enhancing effect of diazepam was absent, indicating that α₂ and α₃ GABA_A receptors are required for the reward-enhancing action of diazepam (111). Interestingly, in α₂(H101R) mice, diazepam becomes potentially aversive, as indicated by a shift of the frequency-response curve to the right (111). As α₂ GABA_A receptors may have an important role in antidepressant-like and reward-related behaviors, they may be a target for antidepressant drugs with a novel mechanism of action. The FST and TST models of behavioral despair use short-duration (on the order of minutes) stress in normal rodents, and antidepressants produce a rapid response in these models. In contrast, human depression is thought to develop in genetically susceptible individuals after chronic environmental stress exposure, and only chronic administration of antidepressants produces a response. It is therefore necessary to investigate the role of α₂-containing GABA_A receptors in rodent protocols involving chronic stress, such as the chronic social defeat paradigm, as well as in other depression-related tests (134).

In any case, there is already evidence that modulation of GABA_A receptors can have antidepressant effects. Alprazolam and adinazolam elicit antidepressant responses similar to widely prescribed antidepressants in MDD patients, but within the first two weeks of treatment (135, 136). Eszopiclone, a positive allosteric modulator (PAM) of GABA_A receptors, is used for the treatment of insomnia. In patients comorbid with MDD and insomnia, eszopiclone/fluoxetine cotherapy was associated not only with sleep improvement but also a faster onset of antidepressant response and a greater magnitude of the antidepressant effect (137). Interestingly, eszopiclone also facilitated the antidepressant efficacy of fluoxetine in the chronic social defeat mouse model of depression (138). Further studies---for example, studies using nonsedating α₂/α₃-selective PAMs in animal models of depression---are expected to further validate the GABA hypothesis of depression. Potentially, such α₂/α₃-selective PAMs or other related compounds could be fast-acting antidepressants.

SCHIZOPHRENIA

Several lines of evidence suggest that inhibitory neurotransmission in the CNS plays important roles in modulating circuits in the brain that are involved in the manifestation of symptoms of schizophrenia. A decline in biosynthesis of cortical GABA leads to a downregulation of GABAergic cortical function in schizophrenia and a compensatory (but insufficient) upregulation of GABA_A receptors (139). Furthermore, a deficit in the glutamatergic activation of GABAergic interneurons in the prefrontal cortex, which synapse on pyramidal neurons at the axon initial segment, results in upregulation of the α₂ subunit of the GABA_A receptor in the axon initial segment (140). In rhesus monkeys, ketamine-induced working memory deficits were reduced by TPA023 (MK-0777) (141), which is a partial, α₂/α₃-selective PAM of GABA_A receptors [in vitro relative efficacy of 0.11 at α₂ and 0.21 at α₃ compared with chlordiazepoxide (1.0); see above and Reference 94]. Furthermore, this compound increased frontal gamma band power during a cognitive task in patients with schizophrenia (142). Whereas cognitive improvements were either relatively minor in the study by Lewis et al. (142) or undetectable in a larger study by Buchanan et al. (143), the relative efficacy of MK-0777 at α₂ GABA_A receptors is only 0.11 compared with that of chlordiazepoxide (1.0), a full agonist at the benzodiazepine site (94). A compound with higher efficacy might have greater cognitive effects.

The potential involvement of GABAergic inhibition in sensorimotor gating and cognitive function is further supported by the finding that α₃ global knockout mice and mice with reduced expression of α₅ in the hippocampus display deficits in prepulse inhibition of acoustic startle (144, 145). Moreover, the mice with reduced expression of α₅ in the hippocampus also displayed a latent inhibition deficit (146), indicating a role of α₅ GABA_A receptors in cognition. In line with these observations, an α₅-selective PAM, SH-053-2′F-R-CH3, applied systemically or directly into the hippocampus reverses dopaminergic hyperactivation in the methylazoxymethanol acetate (MAM) developmental model of schizophrenia (147).

The findings cited above would be compatible with a drug acting at α₂, α₃, and α₅ GABA_A receptors having both antipsychotic (via α₃ and α₅) and cognitive-enhancing (via α₂) effects, without sedative (via α₁) side effects. In retrospect, it is striking that the partial, nonselective PAM bretazenil, originally developed as a nonsedating anxiolytic, was efficacious as monotherapy in 44% of neuroleptic-free patients with acute episodes of schizophrenia (148).

AUTISM

In a model that has become increasingly popular, Rubenstein & Merzenich (149) have postulated that genetic and environmental factors can lead to an increased excitation/inhibition (E/I) ratio, which may be of pathophysiological relevance for at least some forms of autism. Optogenetic activation of prefrontal cortical pyramidal neurons not only resulted in increased spiking but also caused a substantial reduction in information processing in these cells, a social interaction deficit, and impaired fear conditioning (150). These findings indicate impaired cognition, providing support for the hypothesis that an increased E/I imbalance is sufficient to elicit neuropsychiatric disease-related symptoms. A meta-analysis revealed that in several mouse models of autism spectrum disorders (ASD), the number of PV-positive GABAergic interneurons was reduced in the neocortex, suggesting that PV-circuit disruption may be relevant in the development of ASD (151). For example, in a Cntnap2^–/– mouse model of cortical dysplasia-focal epilepsy, a syndromic form of ASD, the number of GABAergic interneurons was reduced, and the network activity was abnormal (152). Evidence that dysfunction of GABAergic neurotransmission is associated with ASD is increasing. For example, loss of MeCP2 from a subset of GABAergic neurons in the forebrain results in a reduced inhibitory quantal size and in the recapitulation of many features of Rett syndrome, indicating that subtle dysfunction of GABAergic neurons potentially contributes to ASD (153). Furthermore, in Scn1a (Na_v1.1) heterozygous knockout mice (i.e., haploinsufficient mice)---which are a model of Dravet's syndrome, also a syndromic form of ASD---GABAergic neurotransmission is decreased, and mice exhibit hyperactivity, stereotypies, social interaction deficits, and impaired context-dependent spatial memory (154). This phenotype is essentially replicated by a heterozygous knockout of the Scn1a gene specifically in GABAergic forebrain neurons, suggesting that dysfunction of GABAergic neurons is the essential deficit in the global Scn1a^+/− mice and potentially also in patients. Interestingly, a very low dose (0.0625 mg/kg) of clonazepam---a PAM of GABA_A receptors containing the α₁, α₂, α₃, or α₅ subunit---completely rescued abnormal social behaviors and deficits in fear memory (154). At this low dose, clonazepam is not anxiolytic or sedative (154). Dysfunction of GABAergic neurons may result in upregulation and/or sensitization of postsynaptic GABA_A receptors, which would explain why such low doses of clonazepam are effective. Which GABA_A receptor subtypes mediate the phenotypic rescue observed in the Scn1a^+/− mice is unknown. Given the observation that PV-positive interneurons are reduced in multiple mouse models of ASD, it is tempting to speculate that α₁-containing GABA_A receptors might play a role, as synapses between PV-positive basket cells and principal neurons typically contain α₁ GABA_A receptors. Collectively, the findings cited above support the notion that an elevated E/I imbalance is an essential factor in the pathophysiology of ASD and that correction of this imbalance may be a powerful therapeutic strategy. Given the efficacy of the nonspecific GABA_A receptor modulator clonazepam in Scn1a^+/− mice, knowledge of the GABA_A receptor subtype(s) that correct the autistic-like phenotype would allow for the development of much more specific compounds with a pharmacological profile different from that of classical benzodiazepines.

Support for the hypothesis that a dysregulated GABA system may play an important role in the pathophysiology of autism comes from postmortem studies, which found reduced expression of GABA_A receptor subunits α₄, α₅, and β₁ as well as and GABA_B receptor subunit β₁ (155). Furthermore, a recent positron emission tomography (PET) pilot study points to lower levels of α₅ GABA_A receptors in the brains of patients with ASD (156), strengthening the emerging connection between deficits in the GABA_A receptor system and ASD. The important role of the E/I imbalance in the pathophysiology of ASD suggests that subtype-selective modulation of GABA_A receptors may be a promising novel therapeutic approach once the functions of individual receptor subtypes in animal models of ASD have been elucidated.

CONCLUSION

GABA_A receptor subtypes offer the promise of a new CNS pharmacology beyond classical benzodiazepines. The most striking impact comes from the recognition that α₅ GABA_A receptors modulate cognitive behavior. Partial inverse agonists (i.e., NAMs of the benzodiazepine site) acting at α₅ receptors overcome deficits in attention, learning, and memory induced either pharmacologically in rodents or nonhuman primates or genetically in a murine Down syndrome model. A clinical trial in individuals with Down syndrome has been initiated. Concerning anxiety circuits, α₂ (and potentially α₃) GABA_A receptors play a key role. Despite promising initial trials, an anxiolytic devoid of sedation is still being awaited. Ligands with selective affinity and efficacy for the α₂ GABA_A receptor subtype might be needed for a breakthrough. This is the more urgent matter, as ligands of this receptor subtype are expected also to ameliorate depressive symptoms and impaired cognitive behavior in schizophrenia. In ASD, attempts to rebalance the increased E/I ratio also points to a role for enhancing the GABA system. Neurosteroids act preferentially at benzodiazepine-insensitive GABA receptors. By stimulating the synthesis of neurosteroids, the drug XBD173 showed antipanic effects in human volunteers. Thus, with an increasingly subtype-specific pharmacology, the GABA system holds expanding therapeutic potential for CNS disorders.