The notion that drug treatments can improve memory performance has moved from the realm of science fiction to that of serious investigation. A popular working hypothesis is that cognition can be improved by altering the balance between excitatory and inhibitory neurotransmission. This review focuses on the unique physiological and pharmacological properties of GABAARs [GABA (γ-aminobutyric acid) subtype A receptors] that contain the α5 subunit (α5-GABAAR), as these receptors serve as candidate targets for memory-enhancing drugs.
The GABAARs [GABA (γ-aminobutyric acid) subtype A receptors] belong to the cysteine-loop family of ligand-gated receptors, which also includes the nicotinic acetylcholine receptor, glycine receptor, glutamate-gated ion channel, zinc-gated ion channel and ionotropic serotonin receptor . GABAARs are membrane-spanning proteins that surround a central pore to form an ion channel in the membrane. Each GABAAR is assembled as a pentamer from a pool of 19 different subunits (α1-6, β1-3, γ1-3, δ, ε, π, θ, ρ1-3) . The combination of subunits is specific, and the majority of native receptors in the mammalian brain contain α1, β2 and γ2 subunits in a 2:2:1 stoichiometry . Distinct isoforms of GABAARs have different developmental, physiological and pharmacological properties and are localized to specific brain regions and subcellular compartments . As we discuss below, GABAARs that contain the α5 subunit (α5-GABAARs) generate a tonic form of inhibition, are expressed mainly in extrasynaptic locations and play a role in modifying learning and memory behaviors.
Receptor distribution, function and pharmacology
The distribution of α5-GABAARs is relatively sparse and compartmentalized in the mammalian brain . In total, approx. 5% of GABAARs contain the α5 subunit , although in the hippocampus, 20–25% of GABAARs contain this subunit . The α5 subunit is predominantly localized to the stratum radiatum and stratum oriens of the CA1 and CA3 regions . The distribution of the α5 subunit is also exceptionally high in the olfactory bulb , where 35% of neurons in the internal granule cell layer express α5-GABAARs . The function of α5-GABAARs in this region remains unknown. Other regions that express the α5 subunit include the neocortex , subiculum  and substantia gelatinosa ; this subunit is also found in sympathetic preganglionic neurons .
Studies from our laboratory have shown that α5-GABAARs generate tonic inhibitory conductance in CA1 hippocampal pyramidal neurons . This tonic conductance is significantly reduced in null mutant mice that have a genetic deletion of the gene that encodes the α5 subunit (Gabra5−/− mice); however, spontaneous synaptic GABAergic inhibition remains unchanged . The tonic conductance is sensitive to midazolam , but not to zolpidem , which is consistent with the presence of α5 and γ subunits . The α5-GABAARs display relatively high sensitivity to GABA, and they display slower desensitization kinetics than conventional synaptic GABAARs [11,14]. Accordingly, low ambient concentrations of GABA in the extracellular space are thought to activate a proportion of these receptors and generate tonic inhibitory conductance . α5-GABAAR-generated tonic conductance has also been reported in cortical neurons , dopaminergic neurons of the striatum  and neurons in the intermediolateral cell column in spinal cord slices from rats . Recombinant α5β3γ2 receptors have pharmacological properties similar to those of receptors that generate tonic current in CA1 hippocampal neurons [11,18], suggesting that this is the predominant combination of subunits in the hippocampus. Additionally, mass spectrometry identified that the α5 subunit associates with multiple α, β and γ subunits, but most frequently the β3 subunit .
Immunocytochemistry and in situ hybridization studies have indicated that α5-GABAARs are localized mainly, but not exclusively, to extrasynaptic regions of neurons [20–22]. Immunogold staining of the hippocampus showed that α5-GABAARs were located on the dendrites of pyramidal neurons in the CA1 region of the rat hippocampus and cerebral cortex . The clustering of α5-GABAARs is regulated by the binding of radixin, an actin-binding protein that anchors receptors to the cytoskeleton, to the activated form of the receptor .
The results of several electrophysiological studies suggest that α5-GABAARs also generate transient inhibitory synaptic potentials [25–27]. In neocortical pyramidal cells, a proportion of synaptic GABAergic events are reduced by the α5-GABAAR-selective inverse agonist α5IA . Inhibitory postsynaptic potentials generated by bistratified interneurons were potentiated by diazepam but not enhanced by zolpidem, which is consistent with an α5-GABAAR subtype . Additionally, α5-GABAARs may contribute to synaptic GABAergic events with slow kinetics in cortical and hippocampal pyramidal neurons . In hippocampal slices prepared from Gabra5−/− mice, the amplitude and decay time course of the evoked inhibitory postsynaptic currents were reduced, which suggests that deletion of α5-GABAARs may reduce synaptic inhibition . In contrast, inhibiting α5-GABAARs with low concentrations of the benzodiazepine inverse agonist L-655,708 did not alter spontaneous IPSCs (inhibitory postsynaptic currents) in the CA1 and CA3 regions of the hippocampus, suggesting that spontaneous synaptic currents are not readily influenced by α5-GABAAR activity [11,30].
Finally, α5-GABAAR activity can reduce neuronal excitability by shunting mechanisms and/or changes in membrane potential . In the hippocampus, α5-GABAAR activity reduces the excitability of individual pyramidal neurons  and networks of neurons  as well as the power of network oscillations . The regulation of network excitability may contribute to α5-GABAAR regulation of hippocampus-dependent behavioural processes.
Learning, memory and α5-GABAARs
The concept that a decrease in GABAAR activity modifies learning and memory is not new, because it is well-recognized that bicuculline, a non-selective competitive antagonist of GABAARs, enhances memory performance . Similarly, non-selective inverse agonists for the benzodiazepine site have been shown to enhance cognitive performance in animal models . However, these drugs have anxiogenic, convulsant and proconvulsant properties that limit their clinical utility . A key question is whether α5-GABAARs can be targeted by subtype-selective drugs to modulate memory without the adverse consequences of a global decrease in GABAergic inhibition.
Two mouse models, the Gabra5−/− mouse  and a point mutant (α5H105R) mouse with reduced expression of α5-GABAARs , have been used extensively to study the role of α5-GABAARs in cognition. Both types of mice have normal lifespans, breed normally and exhibit no overt compensatory change in other GABAAR subtypes. Initial studies showed that Gabra5−/− mice display enhanced acquisition in the matching-to-place version of the hippocampus-dependent water maze task , although this finding has not been replicated [36,37]. Furthermore, Gabra5−/− mice and α5H105R mutant mice show improved performance in the trace fear conditioning paradigm but perform similarly to wild-type mice in the non-hippocampus-dependent cued fear conditioning protocol [22,37].
A number of drugs have been developed that have a greater affinity for, or selective activity at, α5-GABAARs than for other GABAAR subtypes. These drugs include L-655,708, Ro15-4513, RY 080, RY 023 and RY 024 [38,39], which may improve memory with a relatively low occurrence of side effects . Inverse agonists with selective binding or preferred efficacy for α5-GABAARs inhibit receptor activity allosterically via the benzodiazepine-binding site. The α5-GABAAR-function inverse agonist α5IA improves water maze learning and synaptic plasticity . In vivo, this drug has no apparent convulsant, proconvulsant, or anxiogenic properties . Furthermore, an analogue of α5IA, α5IA-II, regulates encoding and recall but not consolidation of spatial information . The administration of α5IA-II either before training or immediately before memory-testing improved the performance of rats in the water maze, whereas α5IA-II injected following training had no effect. Administration of L-655,708 reduced the time required to find the platform in the Morris water maze and the amount of time spent in the correct quadrant during the probe trial .
Inhibitors of α5-GABAARs may be clinically important for the reversal of memory blockade induced by other drugs. For example, the general anesthetics etomidate  and isoflurane  robustly increase α5-GABAAR-mediated tonic conductance. This action probably contributes to the drugs' amnesic properties [36,45]. Notably, Gabra5−/− mice are resistant to the amnesic properties of etomidate , and L-655,708 prevents memory blockade by etomidate in wild-type mice . In human volunteers, the memory-blocking effects of ethanol on the recall of word lists were reversed by α5IA-II , although the effects of α5IA-II alone on memory performance were not demonstrated. Furthermore, memory impairment caused by the muscarinic antagonist scopolamine can be reversed with BiRY-080, a novel inverse agonist with 130-fold selectivity for α5-GABAARs . Inverse agonists including L-655,708 and α5IA are not currently available for clinical use; nevertheless, these drugs can serve as prototypes for drug development.
The molecular substrate for hippocampus-dependent learning and memory is thought to be the strengthening of synaptic connectivity, and brain slices have been used to study plasticity in hippocampal networks. LTP (long-term potentiation) of excitatory synaptic transmission in CA1 pyramidal neurons following stimulation of Schaffer collaterals is increased by non-selective inhibition of GABAARs . Interestingly, hippocampal slices obtained from Gabra5−/− mice and α5H105R mice showed no differences in synaptic plasticity after high-frequency stimulation, despite enhanced memory behaviours [22,29]. In contrast, application of L-655,708 and α5IA to brain slices prepared from rats at a concentration that induces preferential binding to α5-GABAARs increased the LTP induced by theta burst stimulation [41,43]. These conflicting results suggest that, although α5-GABAARs may play an important role in the pharmacological enhancement of LTP, their role in baseline plasticity LTP may be smaller.
On a final note, although α5-GABAARs appear to play an important physiological role in the learning and memory process, these receptors may also contribute to pathological conditions, including ethanol addiction , schizophrenia , autism  and epilepsy . Although there are major discrepancies that must still be addressed, α5-GABAARs will undoubtedly remain at the forefront of studies aimed at understanding the cellular basis of memory and the development of memory-modifying drugs.
Neuronal Glutamate and GABAA Receptor Function in Health and Disease: Biochemical Society Focused Meeting held at University of St Andrews, St Andrews, U.K., 21–24 July 2009. Organized and Edited by Chris Connolly and Jenni Harvey (Dundee, U.K.).
This research was supported by operating grants from the Canadian Institutes of Health Research, Ottawa, Ontario, Canada [grant numbers MOP 79428 and MOP 38028 to B.A.O.], a Canada Research Chair Award in Anesthesia, Ottawa, Ontario, Canada (to B.A.O), Canadian Institutes of Health Research Canadian Graduate Scholarship, Ottawa, Ontario, Canada (to L.J.M.) and a Natural Sciences and Engineering Graduate Scholarship, Ottawa, Ontario, Canada (to R.P.B).