Comparison of αβδ and αβγ GABA_A Receptors: Allosteric Modulation and Identification of Subunit Arrangement by Site-Selective General Anesthetics

Abstract

GABA_A receptors play a dominant role in mediating inhibition in the mature mammalian brain, and defects of GABAergic neurotransmission contribute to the pathogenesis of a variety of neurological and psychiatric disorders. Two types of GABAergic inhibition have been described: αβγ receptors mediate phasic inhibition in response to transient high-concentrations of synaptic GABA release, and αβδ receptors produce tonic inhibitory currents activated by low-concentration extrasynaptic GABA. Both αβδ and αβγ receptors are important targets for general anesthetics, which induce apparently different changes both in GABA-dependent receptor activation and in desensitization in currents mediated by αβγ vs. αβδ receptors. Many of these differences are explained by correcting for the high agonist efficacy of GABA at most αβγ receptors vs. much lower efficacy at αβδ receptors. The stoichiometry and subunit arrangement of recombinant αβγ receptors are well established as β-α-γ-β-α, while those of αβδ receptors remain controversial. Importantly, some potent general anesthetics selectively bind in transmembrane inter-subunit pockets of αβγ receptors: etomidate acts at β⁺/α⁻ interfaces, and the barbiturate R-5-allyl-1-methyl-5-(m-trifluoromethyl-diazirynylphenyl) barbituric acid (R-mTFD-MPAB) acts at α⁺/β⁻ and γ⁺/β⁻ interfaces. Thus, these drugs are useful as structural probes in αβδ receptors formed from free subunits or concatenated subunit assemblies designed to constrain subunit arrangement. Although a definite conclusion cannot be drawn, studies using etomidate and R-mTFD-MPAB support the idea that recombinant α1β3δ receptors may share stoichiometry and subunit arrangement with α1β3γ2 receptors.

Graphical abstract

1. Introduction

γ-Aminobutyric acid type A (GABA_A) receptors are major mediators of inhibition in the CNS [1, 2]. The GABA_A receptor is a chloride ion channel, a member of the pentameric Cys-loop superfamily of ligand-gated ion channels (pLGICs), which also include nicotinic acetylcholine, glycine, 5-HT₃ and zinc-activated receptors [3]. Multiple GABA_A receptor subunit subtypes and splice variants have been identified: α1-α6, β1-β3, γ1-γ3, δ, ε, π, θ and ρ1–3 [4, 5]. Each subunit is composed of a N-terminal extracellular hydrophilic domain, four transmembrane α-helices (M1-M4), three inter-helix loops, and a C-terminal extracellular domain [6, 7]. GABA_A receptors mostly consist of two α subunits, two β subunits and one additional subunit from either a γ or δ [4]. The composition, stoichiometry and subunit arrangement of receptors containing ε, π and θ are unknown. The predominant GABA_A receptor isoforms in the brain are αβγ and αβδ [8]. αβγ receptors, with α1β2γ2 as the most common isoform, are widely distributed in the brain [8, 9] and at the cellular level are mostly localized in synapses [10, 11], although α5βγ receptors [12–14] and a small fraction of other αβγ receptors are also extrasynaptic [10, 11]. In contrast, αβδ receptors only constitute a small proportion of GABA_A receptors in the brain [8, 15] and at the cellular level are located outside of synapses [10, 16]. The δ subunits co-assemble with α6 subunits in the cerebellum [17–19] and with α4 subunits in the hippocampus, striatum, thalamus and cortex [19–23]. Co-assembly of δ and α1 subunits is observed in the hippocampus [24]. Although α1 and δ subunit mRNAs are found in the thalamus [25], co-immunoprecipitation of α1 and δ subunit protein is undetectable in this brain structure [21]. The most common isoform of δ subunit-containing GABA_A receptors in the CNS is the α4βδ receptor [8].

Structural models of heteropentameric GABA_A receptors (Figure 1) are based on high-resolution structural data from snail Lymnaea stagnalis acetylcholine binding protein [26], bacterial and nematode homomeric pLGICs, including Gloeobacter violaceus ion channel (GLIC) [27, 28], Erwinia chrysanthemi ion channel (ELIC) [29, 30] and Caenorhabditis elegans glutamate-gated chloride channel α (GluCl) [31, 32], a human GABA_A homomeric β3 receptor [33] and a zebrafish α1 glycine receptor [34]. The interfacial surfaces of each subunit are denoted as minus (−) and plus (+) viewed from above and scanning anticlockwise. There are two GABA binding sites on typical synaptic αβγ GABA_A receptors, each located in extracellular domains between α and β subunits (β⁺/α⁻ interfaces) [35]. GABA binding induces a series of conformational changes that propagates to transmembrane domains, leading to opening of the channel gate [36].

Figure 1. Structural model of α1β2γ2 synaptic GABA_A receptors.

Panel A depicts a side view of α1β2γ2 receptor homology model, showing ribbon representation of extracellular domains (ECDs) and transmembrane domains (TMDs). Subunits are color-coded: α = gold, β = blue, γ = green. Intracellular domains (between M3 and M4) have been shortened -- their actual size is variable and their structures are not well established. Panel B shows the transmembrane domain of the homology model viewed from the extra-synaptic space (with the ECD removed). Panel C is a schematic representation of panel B, with transmembrane domains of each subunit and intra-subunit clefts (black) labeled. In this and the following figures, we observed the “Guidelines for preparing color figures for everyone including the colorblind” [179].

GABA_A receptor-mediated inhibition is both phasic and tonic (Figure 2A) [11, 37]. Phasic inhibition is evoked by transient activation of synaptic GABA_A receptors in response to brief high concentrations of GABA. Voltage-clamp recordings of neuronal inhibitory postsynaptic currents (IPSCs) measure the amplitude and kinetic properties of phasic currents. In contrast, tonic inhibition is mediated by extrasynaptic GABA_A receptors, which are continuously activated by slowly varying low concentrations of ambient GABA [11, 37–40]. These steady-state tonic currents are measured in voltage-clamped neurons as the holding current that is blocked by GABA_A receptor antagonists such as bicuculline or picrotoxin. Tonic inhibition plays a critical role in modulating neuronal excitability in many brain areas. For example, in thalamus and hippocampus, tonic currents account for ~ 75–90% of total GABA_A receptor-mediated inhibitory currents [37, 41, 42].

Figure 2. Anesthetic effects on synaptic vs. extrasynaptic neuronal GABAergic currents.

Panel A depicts a GABAergic synapse with a presynaptic neuron on the left and a postsynaptic neuron on the right. In response to action potentials triggering calcium influx and increased intracellular calcium in presynaptic boutons, GABA release into the synapse transiently reaches millimolar concentrations that activate most of the nearby postsynaptic GABA_A receptors. Transporter-mediated reuptake into neurons rapidly reduces synaptic GABA, resulting in postsynaptic receptor deactivation. Together, these events generate inhibitory postsynaptic currents (IPSCs; panel B, top left). In the presence of anesthetic drugs, receptor deactivation and IPSC decay is slowed (panel B, top right), increasing GABA-dependent chloride flux, and reducing GABA EC₅₀ (panel B, bottom). GABA leakage from synapses activates extrasynaptic GABA_A receptors (panel A, lower right), generating tonic currents (panel C, top left). In the presence of anesthetics, tonic currents increase (panel C, top right). Because GABA is a weak agonist at extrasynaptic receptors containing δ-subunit, anesthetics induce smaller “leftward shifts” and larger “upward shifts” in concentration-response experiments of these receptors (panel C, bottom).

Evaluating the properties of individual GABA_A receptor isoforms is carried out by heterologous expression of specific subunits in cells such as human embryonic kidney (HEK) cells and Xenopus oocytes. To mimic fast IPSCs and study fast components of receptor mediated currents, small cells are selected or macropatches are pulled from cell membranes, and a rapid solution-switching device (exchange times range from hundreds of µsec to several msec) is used to apply agonists and/or drugs [43, 44]. This “artificial synapse” approach assesses rates of fast current phases such as desensitization (current reduction during agonist application) and deactivation (current return to baseline after terminating agonist application) [7]. Pseudo steady-state GABA_A receptor responses are investigated using relatively slow solution exchange devices (solution exchange time > 10 msec) in HEK cells and larger cells [45, 46]. In classical GABA concentration-response analysis, GABA EC₅₀ (the GABA concentration that elicits 50% of maximal response) and Hill slope, a coefficient that may give information on the number of GABA interacting sites [47], are measured.

GABA_A receptors are allosterically modulated by a myriad of compounds. Widely used positive GABA_A receptor modulators include benzodiazepines, general anesthetics, and some anticonvulsants [6]. Inhibitors such as bicuculline, picrotoxin, gabazine and other convulsants, are mostly used as research tools. This review focuses on potent general anesthetics.

A variety of approaches have been used to locate and study general anesthetic sites on heteromeric GABA_A receptors, including photolabeling, mutagenesis, subunit chimeras and concatenated subunit assemblies. Receptor labeling with photoreactive anesthetic analogs has identified several binding sites on α1β2/3γ2 GABA_A receptors [35]. Mutant studies have helped define the functional roles of amino acid residues on GABA_A receptors [48]. Chimeric subunits constructed with complementary portions of subunit isoforms associated with different pharmacological sensitivity to anesthetics can identify subunit domains of interest [49]. Concatenated subunit assemblies are formed by tethering two or three subunits together using amino acid linkers to constrain subunit arrangement [50].

2. Structure and function of synaptic αβγ GABA_A receptors

The stoichiometry and subunit arrangement of recombinant αβγ receptors have been well established as β-α-γ-β-α (counterclockwise viewed from the extracellular space) (Figure 3A) [51–53]. Mutation of the conserved M2-9´ leucine to serine (L9´S) on α1, β2 or γ2 subunits induces a large reduction in GABA EC₅₀. Co-expression of receptor messenger RNAs encoding mixtures of wild type and L9´S mutant α1, β2, or γ2 subunits produced mixed populations of receptors and multi-phasic GABA concentration-responses [51]. From the number of phases in such concentration-response curves, it was inferred that α1β2γ2 receptors contain two α1, two β2 and one γ2 subunits [51]. Baumann et al examined the subunit arrangement of α1β2γ2 receptors by co-expressing concatenated subunit assemblies that constrain different subunit arrangements [54]. Electrophysiological studies demonstrated that the subunit arrangement of β2-α1-γ2-β2-α1 exhibits pharmacological properties similar to α1β2γ2 receptors assembled from free subunits [54]. Biochemical and fluorescence energy transfer experiments [52, 53] support this stoichiometry and subunit arrangement of α1β2/3γ2 receptors.

Figure 3. Cartoons to depict the subunit arrangement of recombinant αβγ and αβδ receptors.

A, The stoichiometry and subunit arrangement of recombinant αβγ receptors are well established as β-α-γ-β-α [51–53], which is also illustrated in Figure 1B, C. B, Our data in oocytes using MWC model analysis support the idea that the subunit arrangement of α1β3δ receptors is β3-α1-δ-β3-α1 [46], which was also proposed by Kaur et al [76]. C, D, The other two subunit arrangements proposed by Kaur et al [76] are β3-α1-δ-α1-β3 and α1-β3-α1-β3-δ. E, F, Photolabeling studies indicate that the possible subunit arrangement of α4β3δ receptors stably expressed in HEK293 cells is either β3-β3-δ-β3-α4 or β3-β3-α4-δ-α4. (all cartoons counterclockwise viewed from the extracellular space).

Most recombinant αβγ receptors exhibit very infrequent spontaneous gating [55–57]. It is estimated, for instance, that only about 1 in 20,000 channels is spontaneously open in α1β2γ2 receptors [57]. Maximally GABA-activated (at 1–10 mM) peak currents in oocyte-expressed α1β2/3γ2 receptors are enhanced about 10–20% with addition of positive allosteric modulators [38, 58–60], indicating that GABA alone activates about 85% of the receptors. Both receptor desensitization and deactivation apparently contribute to shaping IPSC time course [61]. Receptor desensitization in α1β2/3γ2, α4β3γ2, α5β3γ2 or α6β3γ2 receptors displays multiple exponential components during prolonged GABA exposure [49, 62–67]. Receptor desensitization reduces the speed of deactivation (desensitization-deactivation coupling) [43] in α1β2/3γ2, α4β3γ2, α5β3γ2 or α6β3γ2 receptors [65, 67–69].

Co-expression of α1, β2 and γ2 subunits in oocytes or HEK cells could form a mixture of α1β2 and α1β2γ2 receptors. In oocytes, increasing the ratio of γ2 subunit mRNA/cDNA relative to α1 and β2 subunits (up to 10:1:1) facilitated the formation of more homogenous α1β2γ2 receptors [70, 71]. In contrast, in HEK cells, reduced ratio of γ2 subunit cDNA to α1 and β2 subunits (0.3:1:1) led to homogenous α1β2γ2 receptors, while higher γ2:α1/β2 ratios may result in incorporation of more than one γ2 subunit per receptor [72]. It is currently unknown how receptor assembly differs in various cell types. However, co-expression of concatenated subunit assemblies appears to constrain subunit stoichiometry and arrangement in αβγ receptors [54, 72–74].

3. Structure and function of extrasynaptic αβδ GABA_A receptors

The stoichiometry and subunit arrangement of αβδ GABA_A receptors is uncertain. Several approaches that were used to investigate αβγ subunit stoichiometry and arrangement were unsuccessful in αβδ receptors. For instance, δL9´T mutations produce small EC₅₀ shifts (< 4-fold; our unpublished observations) and GABA concentration-responses using mixtures of mutant and wild type subunits do not support strong inferences about stoichiometry of expressed αβδ receptors [75]. Comparison of concatenated subunit assemblies to free α1/4β2/3δ receptors gives no “best match” result based on GABA EC₅₀ and neurosteroid enhancement of maximal GABA-induced currents [76, 77]. Both dimer (β-α) and trimer (β-α-γ or β-α-δ) can form functional receptors alone. In αβγ receptors, the current amplitude resulting from β-α-γ and β-α co-expression is much bigger than that of receptors formed from either dimer or trimer alone. Thus, it is evident that pentameric β-α-γ/β-α receptor currents dominate when both dimer and trimer are co-expressed [74]. However, in αβδ receptors, the current amplitude resulting from co-expression of β-α-δ and β-α is similar to that from receptors formed from either dimer or trimer alone. Thus, it is uncertain if pentameric β-α-δ/β-α receptors are the predominant form when both dimer and trimer are co-expressed. Moreover, α4β2δ receptors display very different neurosteroid activation efficacy relative to GABA in oocytes and HEK cells, although results in the two cell types agree better when β2-δ concatemers are co-expressed with free α4 subunits. This implies that free subunits assemble differently in the two cell types [45].

Spontaneous activity is not observed in recombinant α1β2δ receptors, but has been reported in α4β3δ and α6β2δ receptors [78, 79]. Maximally activating GABA concentrations evoke peak currents from recombinant αβδ receptors that are enhanced over 20-fold by positive allosteric modulators [59, 64, 80, 81]. These observations led us to estimate that saturating GABA concentrations activate only 3–5% of α1β3δ receptors [46]. α4/6β2/3δ receptors are very sensitive to GABA, with GABA EC₅₀ less than 1 µM [64, 80, 82]. Neuronal tonic currents appear to be non-desensitizing even with addition of exogenous GABA [83], and α1β2/3δ receptors activated by maximal GABA exhibit very little or no desensitization [58, 59, 84]. Although α4β2/3δ and α6β3δ receptors desensitize in response to > 1 mM GABA [64, 67, 80, 85], the rate is slower than in α4β3γ2 and α6β3γ2 receptors [65, 85]. Deactivation of αβδ receptors is faster than that of αβγ receptors, consistent with low GABA efficacy.

In some recombinant expression systems, δ subunits may not efficiently incorporate into ternary αβδ receptors [46, 86–88]. This likely explains why one study found identical general anesthetic modulation in both α4β3δ and α4β3 receptors [89]. The subunit stoichiometry of surface αβδ receptors apparently depends on the ratios of α:β:δ mRNAs or cDNAs used for heterologous expression [72, 88, 90, 91]. By varying the subunit ratios and the concentrations of mRNAs injected into oocytes, we observed that increasing the δ subunit ratio (α:β:δ ratio = 1:1:3) and reducing total injected mRNA (10-fold from the 5 ng/oocyte typically used for α1βγ2 receptors) maximized current enhancement by allosteric modulators, suggesting optimal incorporation of the δ subunit into α1β3δ receptors [46]. It is unknown if this approach can be used to monitor the incorporation of the δ subunit into α4βδ and α6βδ receptors. In expression studies of α1β2δ in HEK cells, 10-fold less δ subunit cDNA is required to eliminate the functional signature of α1β2 receptors [72]. Thus, consistent incorporation of δ subunit into heterologously expressed receptors requires close attention to experimental conditions.

4. Functional effects of anesthetics on synaptic αβγ GABA_A receptors

Many potent general anesthetics allosterically enhance the activation of GABA_A receptors [7, 92–95]. In studies of phasic inhibition associated with αβγ receptors, etomidate [96, 97], propofol [98, 99], the barbiturate pentobarbital [100, 101], and the neurosteroids tetrahydrodeoxycorticosterone (THDOC) [102] and alphaxalone [103] all prolong IPSC decay. Potent general anesthetics also enhance GABA-evoked activation of oocyte-expressed α1β2/3γ2 receptors (Figure 2B), producing large “leftward shifts” (decreased EC₅₀) and a small “upward shift” (increased maximal GABA current response) in concentration-responses [46, 60, 69, 104, 105]. This pattern of anesthetic-induced changes indicates high GABA intrinsic efficacy in αβγ receptors, consistent with previous single channel studies [59, 106, 107]. At high concentrations, general anesthetics also directly activate αβγ receptors (GABA-independent activation or allosteric agonism) [7]. Both positive modulation of orthosteric agonist responses and direct receptor activation by the anesthetic etomidate display similar pharmacological characteristics [57, 108, 109]. In α1β3γ2 receptors, the direct effect of pentobarbital appears bi-phasic, with maximal currents due to direct agonism at ~1 mM and inhibition at higher concentrations via a distinct inhibitory site [59].

In rapid kinetic studies of α1β3γ2 receptors, pentobarbital, propofol and THDOC do not alter or slightly reduce peak currents evoked by maximal GABA [58, 59, 64]. Pentobarbital and propofol decrease the extent of desensitization of α1β3γ2 and/or α6β3γ2 receptors [49, 59, 64], while THDOC and etomidate do not alter desensitization of α1β2/3γ2 receptors [58, 66]. General anesthetics prolong the deactivation of α1β2/3γ2 and/or α6β3γ2 receptors [58, 59, 64, 66]. Thus, desensitization may contribute to the differences in apparent maximal intrinsic efficacy of GABA in αβγ receptors studied in Xenopus oocytes vs. HEK293 cells.

5. Functional effects of anesthetics on extrasynaptic αβδ GABA_A receptors

Extrasynaptic αβδ receptors are also important targets for general anesthetics [7], which enhance tonic currents in multiple brain regions [41, 42, 83, 98, 110–117]. General anesthetics alter GABA-dependent responses in oocyte-expressed αβδ receptors (Figure 2C) by inducing a small “leftward shift” and a large “upward shift”, contrasting with the pattern in αβγ receptors [46, 81]. The large increase in maximal GABA currents indicates low GABA intrinsic efficacy in αβδ receptors. Consistent with this, single channel recordings in the presence of high GABA show smaller mean channel open times in αβδ than in αβγ receptors [58, 59, 106, 118, 119]. Application of general anesthetics with maximal GABA increases the mean channel open time by inducing long-lived open states in α1β3δ receptors in single channel recordings [58, 59, 68, 119]. General anesthetics at high concentration also directly activate αβδ receptors. Pentobarbital alone directly activates α1β3δ receptors, with maximal responses at ~3 mM [59].

Artificial synapse experiments also show that general anesthetics substantially enhance peak currents of α1β3δ or α4β3δ receptors expressed in HEK cells [44, 58, 59, 64, 85, 89, 120]. Propofol and THDOC appear to evoke a greater augmentation of maximal GABA currents in α1β3δ than in α4/6β3δ receptors [58, 64, 76, 85, 89]. Different anesthetics evoke variable changes in αβδ receptor desensitization. Etomidate reduces the extent of desensitization for α1β3δ receptors [44]. This effect contrasts with studies showing apparent acceleration of α1β3δ receptor desensitization by both pentobarbital and THDOC [49, 58, 59]. Propofol does not alter desensitization of either α1β3δ or α6β3δ receptors [64]. Etomidate, pentobarbital and THDOC prolong deactivation of α1β3δ receptor currents [44, 49, 58, 59]. Propofol prolongs the deactivation of α6β3δ receptors but not α1β3δ receptors [64].

Endogenous neurosteroids are mainly synthesized in glial cells [121, 122], and their concentrations in the brain (10–300 nM) are dynamically regulated [123, 124]. The endogenous neurosteroid THDOC at physiological concentrations selectively enhances tonic currents mediated by αβδ receptors [83]. Both expression and function of αβδ receptors change with fluctuating neurosteroid levels in rodent brain [125–130]. Sensitivity of tonic GABAergic currents to neurosteroid modulation varies among brain regions. In hippocampus, 10 nM THDOC reduces neuronal excitability by augmenting tonic αβδ receptor currents [83]. In thalamocortical neurons, although 100 nM THDOC enhances tonic currents, 10 nM THDOC does not [42].

Several factors likely contribute to region-specific modulation of tonic inhibition by neurosteroids: First, the distribution of GABA_A receptors in the brain is heterogeneous [8, 9]. The major αβδ receptor subtype expressed in thalamus is α4βδ, while α1βδ subtype is undetectable [20, 21]. In contrast, the hippocampus expresses both α4βδ and α1βδ receptors [22, 24]. α4βδ and α1βδ receptors exhibit differential sensitivity both to GABA [59, 80, 82] and to neurosteroids (see above). Second, the different effects of neurosteroids on tonic inhibition in the hippocampus and thalamus may be partly due to phosphorylation of αβδ receptors. The function and modulatory effect of α4β3δ receptors are regulated by both PKA and PKC [79, 131–133].

6. Shift analysis using an allosteric co-agonist model

A Monod-Wyman-Changeux (MWC) allosteric model was introduced by Chang & Weiss to describe the activation of wild type and gating mutant α1β2γ2 receptors [134]. In this model, estimated open probability (P_open) in a population of receptors is calculated by adding normalized spontaneous receptor activity to activated currents and correcting for the intrinsic efficacy of GABA. GABA efficacy is estimated by measuring how much maximal GABA currents are enhanced by strong positive modulators such as anesthetics or neurosteroids. MWC models allow and account for spontaneous channel activity, and were shown to account for effects on both GABA EC₅₀ and spontaneous activation associated with L9’S mutations [134].

The effects of general anesthetics on αβγ receptor function are similar to those of L9’S gating mutations [134]. The combined effects of GABA and anesthetics, including etomidate, propofol, and pentobarbital, can be quantitatively accounted for using MWC co-agonist models (Figure 4) [60, 104, 135, 136]. Because MWC models explicitly account for agonist intrinsic efficacy, MWC shift analysis also unifies leftward-shift, increased maximal response, and direct anesthetic activation to quantify anesthetic modulation in different GABA_A receptor subtypes [46, 136] or when a partial orthosteric agonist is combined with an anesthetic [135]. We modeled the allosteric modulation of etomidate on α1β3δ receptors using an MWC model that accounted for all of above effects when normalized responses were corrected for the low intrinsic efficacy of GABA in these receptors [46].

Figure 4. Monod-Wyman-Changeux allosteric co-agonism model in GABA_A receptors.

This equilibrium scheme depicts a receptor activation model with two canonical states (R = closed, O = open) that can be activated by the orthosteric agonist GABA (G) and allosteric agonists such as anesthetics (A). The number of GABA binding sites is 2, based on structural and functional studies of synaptic GABA_A receptors, and the number of anesthetic sites is a variable, n. GABA binding transitions and the GABA-activated state OG₂ are shown in black and anesthetic binding transitions and the anesthetic-activated state OA_n are shown in blue. The co-activated OG₂A_n is shown in orange. An equation for the equilibrium open probability (in a population of receptors) of this model is shown below the scheme.

The basal equilibrium between unliganded R and O states (R:O) is L₀ and basal open probability = (1+L₀)⁻¹; thus low L₀ values indicate a higher probability of spontaneous receptor activation. The dissociation constant for GABA interactions with closed receptors is K_G, and the dissociation constant for anesthetic interactions with closed receptors is K_A. Agonist efficacy in the model depends on the relative affinity of ligands for open vs. closed receptors. For GABA, the efficacy parameter is c and for anesthetics, it is d; c and d values less than 1.0 indicate agonist activity and lower values indicate greater relative affinity for open states and thus greater agonist efficacy. The influence of agonists on closed:open equilibria is shown in black. When GABA concentrations are high enough to occupy all receptors, the RG₂:OG₂ equilibrium is c²L₀ and P_open = (1+c²L₀)⁻¹. Similarly, anesthetic-bound receptors display an RA_n:OA_n equilibrium of dⁿL₀ and P_open = (1+dⁿL₀)⁻¹. When both GABA and anesthetics are bound, the RG₂A_n:OG₂A_n equilibrium is c²dⁿL₀ and P_open = (1+c²dⁿL₀)⁻¹. Modified from [104].

Our MWC models are based on data collected from oocyte recordings and do not include desensitized states. As the solution exchange times are slow in oocyte recordings, fast components of desensitization are obscured. However, MWC models that include desensitized states are consistent with the kinetics of patch-clamp recordings obtained with rapid solution exchange [137].

7. Structural identification of selective anesthetic sites on synaptic αβγ GABA_A receptors

Multiple approaches have been used to locate general anesthetic sites on GABA_A receptors, revealing evidence of remarkable site selectivity for some drugs. Both photolabeling and chimera strategies are unbiased approaches that can locate potential sites, while approaches based on point mutations can provide hypothesis-based complementary evidence. Each of these approaches has multiple limitations and weaknesses. Here, we briefly review these approaches and how they have contributed to identifying GABA_A receptor sites for four potent anesthetics.

Chimeric constructs and point mutations have long been used to evaluate structure-function relationships of general anesthetics and other GABA_A receptor modulators [138–141]. However, these studies have limited value in identifying drug contacts. Chimeric approaches have proven useful in cases where two homologous subunits display widely different drug sensitivities. For example, differential etomidate sensitivity in GABA_A receptors containing β1 vs. β2 led to identification of the M2-15´ residue (β1S265 vs. β2N265) as key residues. However, care must be taken in evaluating efficacy and potency metrics--a set of β3/ρ1 chimeras that were constructed to locate pentobarbital sites showed varying effects on pentobarbital agonism vs. maximal GABA effects, but variance and single channel analyses indicated that GABA efficacy was altered without affecting pentobarbital efficacy [142]. In the case of point mutations, we tested whether the etomidate-insensitive phenotype of α1M236W and β2M286W mutations was generalizable to other anesthetic-photolabeled residues, and found that Trp mutations do not reliably identify drug contact vs. non-contact residues [143].

Photolabeling has provided unbiased regional identification of anesthetic binding sites on GABA_A receptors [3, 94, 144]. However, many of these studies used GABA_A receptors with only one or two subunit isotypes (e.g. β3 or α1β3) instead of the most common α1β2/3γ2 construct. These models limit data analysis and comparisons with functional studies [35, 145]. More generally, photoreactive groups, if appended to parent drug molecular structures in key binding regions, may disrupt pharmacological activity. Thus, photolabeling with many anesthetic derivatives likely fails to identify important contact residues.

Substituted cysteine accessibility modification-protection (SCAMP) was first used by Bali and Akabas to test propofol interactions in GABA_A receptors [146] and applied more extensively by us to identify contact sites for multiple anesthetics [35]. SCAMP involves mutation of a putative anesthetic contact residue to cysteine, introducing a free sulfhydryl. Formation of covalent bond formation after exposure to a sulfhydryl-reactive chemical agent is detected with electrophysiology. If the cysteine-substitution is in an anesthetic binding pocket, drugs occupying the site should reduce the rate of covalent bond formation through steric competition [35]. Our recent analysis shows that SCAMP results agree with photolabeling results, indicating that SCAMP is a reliable technique to identify new contact residues [143]. In SCAMP experiments, it is also important to demonstrate drug-specific effects on cysteine modification. If all anesthetics similarly affect cysteine modification rate, this may indicate an allosteric rather than a steric interaction [35].

Etomidate is a potent stereoselective imidazole ester anesthetic. R-(+)-etomidate positively modulates and directly activates α1β2γ2 receptors about 20-fold more potently than S-(−)-etomidate, similar to the relative anesthetic potency of etomidate enantiomers in animals [147]. Heterologous αβγ receptors with β2 or β3, but not β1 are modulated by etomidate [109]. This observation led to chimera and point mutant studies that identified an asparagine residue (N265) in the M2 helices of β2 or β3 subunits as a major determinant of etomidate sensitivity [108, 148]. Quantitative MWC co-agonist modeling of etomidate effects in α1β2γ2 receptors indicated the presence of two equivalent sites [135]. MWC analysis of etomidate effects in α1β2N265Sγ2 and α1β2N265Mγ2 receptors suggested effects on drug binding to open-channel states more than closed states [69]. Photoreactive etomidate derivatives incorporate at α1M236 in α-M1 and βM286 in β-M3, which are located in structural models at transmembrane β⁺/α⁻ interfaces [149, 150]. Tryptophan mutations at either α1M236 or β2M286 induce spontaneously active GABA-sensitive receptors that are weakly modulated by etomidate, suggesting that the Trp side chains both mimic etomidate binding and occlude the sites [151]. Concatemeric subunit β2-α1 and β2-α1-γ2 assemblies with α1M236W mutations produce equivalent functional changes in expressed receptors, supporting equivalence of the two etomidate sites [74]. Photolabeling also indicated etomidate binding at β3+/β3⁻ interfaces in α1β3 receptors [152], accounting for etomidate activation of β3 homomeric receptors [153]. SCAMP has mapped additional etomidate contact residues in α1β2/3γ2 receptors [143, 154], all of which map to transmembrane β⁺/α⁻ interfaces.

In summary, there is abundant evidence that etomidate binds selectively in the two β⁺/α⁻ interfaces of α1β2/3γ GABA_A receptors.

R-5-allyl-1-methyl-5-(m-trifluoromethyl-diazirynylphenyl) barbituric acid (R-mTFD-MPAB) was developed as a barbiturate photolabel that was found to be both potent and stereoselective in α1β2/3γ2 receptors and in animals [155]. Photolabeling reveals that it binds specifically at transmembrane subunit interfaces between γ2-M3 and β3-M1 (γ⁺/β⁻) and between α1-M3 and β3-M1 (α⁺/β⁻) helices in α1β3γ2 receptors [156]. SCAMP studies have confirmed some of these contact residues in α1β3γ2 receptors and identified additional contacts [143]. The convulsant barbiturate photolabel S-1-methyl-5-propyl-5-(m-trifluoromethyl-diazirynylphenyl) barbituric acid (S-mTFD-MPPB) inhibits the function of α1β3γ2 receptors [157, 158], and also incorporates at γ⁺/β⁻ interfaces on α1β3γ2 receptors [159]. This suggests that the γ⁺/β⁻ interface can mediate allosteric channel gating shifts in opposing directions, perhaps depending on the specific orientation of hypnotic vs. convulsant barbiturates within the site.

In summary, m-substituted MPAB and MPPB derivatives selectively bind in α⁺/β⁻ and γ⁺/β⁻ interfaces of α1β2/3γ2 receptors. Studies are underway to establish if mutations in these sites display asymmetrical functional effects.

Propofol contacts were tentatively identified in mutant studies at β2M286 in β2-M3 [160]. MWC shift analysis indicates that propofol possesses more than two allosteric sites on α1β2γ2 receptors [104]. In line with this prediction, photolabeling studies demonstrated that propofol binds at β⁺/α⁻ and α⁺/β⁻ interfaces in α1β3 receptors, which are homologues of the inter-subunit sites for both etomidate and R-mTFD-MPAB, respectively [35, 161]. Mutant studies support this conclusion [141]. Thus, propofol shares binding sites with both etomidate and barbiturates. Photolabeling in β3 homomeric and α1β3 receptors also suggested propofol contact with β3H267 [162], and β3H267 may be close to the sites for R-mTFD-MPAB [163]. However, studies using mutational analysis and SCAMP techniques do not support the idea that β3H267 contacts propofol [163, 164].

Neurosteroids bind in sites that are distinct from those where etomidate, propofol, and barbiturates bind based on functional and photolabeling results. Neurosteroids enhance GABA_A receptor activation by etomidate and barbiturates, synergize with etomidate in anesthetizing animals [165], and enhance photolabeling by etomidate analogs [166]. Photolabeling in β3 homomeric receptors identified F301 in β3-M3 as a possible neurosteroid binding site [167]. Recent published data supports the hypothesis that neurosteroid enhancing sites are located in β⁺/α⁻ interfaces near the inner end of β-M3 and α-M1 helices. Crystal structures of homomeric chimeric channels that include α subunit transmembrane domains reveal THDOC interactions with αQ242 [168, 169], a site where mutations significantly alter neurosteroid sensitivity [170, 171]. Possible neurosteroid contact residues in α1β2γ2 receptors include S241 and Q242 in α1-M1 and N407 and Y410 in α1-M4 [170–173]. SCAMP experiments comparing contacts for propofol, etomidate and alphaxalone on β3-M3 in α1β3γ2 receptors show that alphaxalone protects cysteine substitutions at β3 residues V290, F293, L297 and F301, and some of these mutations also reduce neurosteroid sensitivity [174]. In addition, SCAMP studies confirm that neurosteroids do not interact with established etomidate and R-mTFD-MPAB contacts [105]. Neurosteroid sites are thus adjacent and intracellular to the sites where etomidate binds in β+/α⁻ interfaces, and the effects of mutations in these sites suggest that they account for the majority of neurosteroid effects on channel gating.

8. General anesthetics as structural probes in αβδ GABA_A receptors

Chimera studies have aimed at identifying the structures underlying the differential patterns of general anesthetic modulation in αβγ vs. αβδ receptors. For example, general anesthetics strongly enhance the peak currents of α1β3δ receptors but only slightly potentiate α1β3γ2 currents in the presence of saturating GABA [49, 58]. Feng and Macdonald [49] compared pentobarbital effects in receptors containing γ2/δ or δ/γ2 subunit chimeras vs. wild type α1β3γ2 and α1β3δ receptors. The relative efficacy of pentobarbital vs. GABA depended on δ sequences between the N-terminus and the middle of M1. However, GABA efficacy in the chimeric receptors was not assessed, so it remains uncertain whether the chimeras affected GABA efficacy, pentobarbital efficacy, or both. It is also possible that the chimeras affected subunit assembly.

The formalism of MWC allosteric models requires assessment of GABA efficacy and thus establishes a stronger framework for functional analysis of structural changes. Taking advantage of the unique binding sites of etomidate on αβγ GABA_A receptors, we inferred if putative αβδ receptors possess the same β⁺/α⁻ interfaces by quantifying the allosteric modulation of these two receptor isoforms by etomidate using MWC allosteric model. Our hypothesis was that if the stoichiometry and subunit arrangement of α1β3γ2 and α1β3δ receptors are different, the number of etomidate binding sites on both receptor isoforms would be different, which would result in divergent allosteric effects of etomidate [46]. We observed that etomidate at 3.2 µM produced a 10-fold leftward shift and 24% upward shift of GABA concentration-responses in α1β3γ2 receptors. However, the same concentration of etomidate evoked a 3-fold leftward shift and a 17-fold upward shift of GABA concentration-responses in α1β3δ receptors [46]. We quantitatively compared etomidate (3.2 µM) modulation of GABA-dependent activation in α1β3γ2 vs. α1β3δ receptors using MWC model analysis. Even though etomidate effects on GABA-dependent responses are very different in α1β3γ2 vs. α1β3δ receptors, after correcting data for the different GABA efficacies in these two receptors, our analysis indicated comparable allosteric shifts [46]. Comparable allosteric modulation of α4β2γ2 and α4β2δ receptors by alfaxalone was also observed using the similar approach [175]. Moreover, because etomidate selectively acts via two β⁺/α⁻ interfacial sites in α1β3γ2 receptors [35], this result may indicate that α1β3δ receptors also contain two β⁺/α⁻ interfaces. If this is the case, our data tend to support previous studies that δ simply replaces γ in the established synaptic subunit arrangement [72, 75, 88, 176]. These data also agree with a previous observation that pentobarbital evokes similar modulation in both α1β3δ and α1β3γ2 currents activated with low GABA concentrations [59]. Indeed, synaptic αβγ receptors may contribute significantly to modulation of tonic currents by general anesthetics [177].

Using site-selective anesthetics is also advantageous in studying GABA_A receptors formed from subunit concatemers. THDOC was used as a probe drug to seek information about receptor subunit arrangement of free α1/6β3δ receptors vs. various concatenated α1/6β3δ receptors [50, 76]. It was proposed that free α1β3δ receptors may display at least three subunit arrangements: β3-α1-δ-β3-α1, β3-α1-δ-α1-β3 and α1-β3-α1-β3-δ (Figure 3B, C, D) [76]. We further examined the allosteric modulation of concatenated β3-α1-δ/β3-α1 receptors (formed by injection of β3-α1-δ trimer and β3-α1 dimer mRNAs in oocytes) by etomidate at 3.2 µM and found that etomidate produced a 3-fold leftward shift and a 14-fold upward shift of GABA concentration-responses [46]. We quantitatively compared etomidate modulation of free α1β3δ receptors and concatenated β3-α1-δ/β3-α1 receptors, and found close agreement, further supporting the β3-α1-δ-β3-α1 subunit arrangement for oocyte-expressed α1β3δ receptors (Figure 3B) [46]. However, it should be noted that our data do not determine the definite stoichiometry and subunit arrangement of α1β3δ receptors. For example, it is possible that etomidate may bind to other intersubunit sites in α1β3δ receptors, such as δ⁺/α⁻ or β⁺/β– interface, which may exert similar allosteric effect to those β⁺/α⁻ sites. R-mTFD-MPAB acts in αβγ receptors via both α⁺/β⁻ and γ⁺/β⁻ sites, but its effects in αβδ receptors remain unexplored. We measured allosteric modulation by R-mTFD-MPAB in free and concatenated β3-α1/β3-α1-δ receptors and found them to be similar (unpublished observations). Further mutant studies to compare the contributions of α⁺/β⁻ sites vs. δ⁺/β⁻ sites and other possible interfacial sites are needed to strengthen our conclusion.

Photolabeling with site-selective anesthetics provides additional insight into αβδ receptor subunit arrangements. Chiara et al [178] identified azi-etomidate and R-mTFD-MPAB incorporation sites in human α4β3δ receptors that were expressed in HEK cells and purified in detergent. Neither photolabel incorporated into the δ subunit. Azi-etomidate incorporated in β3M286 (M3) and α4M269 (M1), consistent with β⁺/α⁻ sites. Both photolabels incorporated at β3M227 (M1), suggesting anesthetic sites in β3⁺/β3 interfaces. Assuming homogeneous assembly, possible subunit assemblies are β3-β3-δ-β3-α4 or β3-β3-α4-δ-α4 (Figure 3E, F) [178]. If α1β3δ receptors share similar stoichiometry and subunit arrangement with α1β3γ2 receptors, these data suggest that the stoichiometry and subunit arrangement of α4β3δ receptors may be different from those of α1β3δ receptors.

In summary, anesthetics, particularly etomidate, neurosteroids and R-mTFD-MPAB that display strong specificity for certain subunit interfacial sites, are informative structural probes for heteromeric GABA_A receptors. Some of these drugs apparently can also act via β⁺/β⁻ interfaces, requiring chimera studies to test inferences. Propofol acts via four sites per αβγ receptor, and is thus of no value as a probe of αβδ receptor subunit arrangements. Interpretation of anesthetic functional effects must include independent assessment of GABA efficacy, and formal MWC model analyses enforce this, while also enabling quantitative comparisons of drug effects in different receptor isoforms.

9. Conclusions

The structure of neuronal αβδ receptors remains uncertain, and the structure of expressed recombinant αβδ receptors may vary with cell type and other variables. Because certain general anesthetics act at well-defined αβγ subunit interfaces, they and subunit modifications that affect anesthetic sensitivity in αβγ receptors represent useful tools for probing the subunit arrangement of αβδ receptors. As we learn more about anesthetic binding sites, the value of these tools as structural probes will also grow.