I. Introduction

Top Next References

This article does not aim to review in detail the properties of -aminobutyric acid_A (GABA_A)^b receptors, because recent accounts of that topic are available. In this same journal, a review of the binding properties and pharmacology of these receptors has been published (Sieghart, 1995). Other reviews have dealt with their ion channel properties as well as their pharmacology (MacDonald and Olsen, 1994; Mohler et al., 1996a,b), whereas others have concentrated on their molecular biology and protein structure (Wisden and Seeburg, 1992; Smith and Olsen, 1995; Stephenson, 1995; McKernan and Whiting, 1996). Further, two recent books have provided many short review articles on the functional, behavioral, and psychopharmacological aspects of GABA receptors (Tanaka and Bowery, 1996; Enna and Bowery, 1997) and an account of these latter aspects will not be repeated here.

Building on that background, we will consider here how our knowledge of GABA_A receptor structure and function could lead to a classification system. Such a system is not immediately obvious from those previous accounts, as it probably would have been with a one-subunit receptor of the G-protein-coupled class. It is surely no accident that all the present series of NC-IUPHAR reports in Pharmacological Reviews on the nomenclature of individual receptor types have so far concerned G protein-coupled receptors.^c Certainly the G protein-coupled receptor class covers by far the largest numbers of receptor types; it includes most of the cases in which known clinical drug applications can be related so directly to those types that they have given a great impetus to the receptor analyses. However, beyond those considerations, a major reason for the success in classification of those types must surely be the distinction which can be made therein between the subtypes of a receptor, based on the fact that each will be created by a single polypeptide with a pharmacology which is encoded solely by its own sequence. This is not to say that the classification of any of the receptors previously surveyed in this series has been entirely obvious or without complexities. Nonetheless, the problems involved are generally concerned with borderline cases in which the sequence data or the discriminatory pharmacological tools were historically less satisfactory. How one longs for such a one-to-one correspondence in the case of the GABA_A receptors!

The discussion here, therefore, is the first in the classification series to tackle a receptor of their class, i.e., the multisubunit, heteromeric ion channels directly activated by the transmitter. The combinatorial principle of receptor construction (to be discussed below) for these ionotropic receptors, which also is used extensively in glutamate and nicotinic acetylcholine receptors, introduces a higher order of complexity. The functional unit is not the single polypeptide, and further, the functional properties contributed by a given subunit can vary with its interactions with the particular set of subunits in each receptor molecule. This complexity renders the recognition of the structures of receptor subtypes in their natural setting extremely difficult (in fact, at present, usually unattainable). Thus, it is not possible to construct a classification comparable with the comprehensive scheme for native receptor subtypes obtained in the previous articles in this series. Instead, a provisional version is presented which relies on the wealth of sequence and functional data available on the recombinant GABA_A receptors.

A. Earlier Classifications of -Aminobutyric Acid Receptors

1. -Aminobutyric acid_A and -aminobutyric acid_B receptors. GABA has been accepted as a neurotransmitter (in mammals and down to crustacea) for several decades. It is now evident that GABA mediates most inhibitory transmission events in the vertebrate brain. It was long clear that the fast, bicuculline-blocked response to GABA observed was caused by direct activation of an intrinsic anion channel in an entity subsequently termed the GABA_A receptor. GABA_B receptors were recognized later as bicuculline-insensitive, baclofen-stimulated metabotropic GABA receptors (Hill and Bowery, 1981) linked to G proteins. Confirmation by the deoxyribonucleic acid (DNA) cloning of a GABA_B receptor, as a 7-transmembrane domain protein, has been accomplished recently (Kaupmann et al., 1997). The complete structural and functional distinction between GABA_A and GABA_B receptors has a clear parallel to that between nicotinic and muscarinic acetylcholine receptors, between 5-HT₃ and metabotropic serotonin receptors, ionotropic and metabotropic glutamate receptors, or ionotropic P_2X and G protein-coupled P_2Y receptors for nucleotides.

2. -Aminobutyric acid_C receptors. A third type of GABA receptor, insensitive to both bicuculline and baclofen, was designated GABA_C (Drew et al., 1984). The GABA_C responses are also of the fast type associated with the opening of an anion channel; they are, however, unaffected by typical modulators of GABA_A receptor channels such as benzodiazepines and barbiturates (Sivilotti and Nistri, 1991; Bormann and Feigenspan, 1995; Johnston, 1996). Native responses of the GABA_C type have occurred in retinal bipolar or horizontal cells across vertebrate species (Feigenspan et al., 1993; Quian and Dowling, 1993; Lukasiewicz, 1996) and can be expressed by rat retinal messenger ribonucleic acid (mRNA) injection in the oocyte system (Polenzani et al., 1991).

3. Benzodiazepine receptors. The interaction with benzodiazepines (BZ) (fig. 1) has been a major influence in studies on GABA receptors because of the long history of therapeutic application of BZs as anxiolytics, anticonvulsants, sedative-hypnotics, and muscle relaxants. Although the BZs were introduced first into clinical practice in the early 1960s, it was not until 1975 that these drugs were recognized to act by potentiating the inhibitory action of GABA in the brain (Costa et al., 1975; Haefely et al., 1975). The presence of high-affinity, specific binding sites for BZs in the mammalian brain was then demonstrated (Braestrup and Squires, 1977; Mohler and Okada, 1977). Converging lines of evidence established that these sites are in the same macromolecule as the GABA sites and the chloride channel and that all three elements are coupled allosterically (Chang et al., 1981; Olsen, 1981; Paul et al., 1981; Sigel and Barnard, 1984). The term "GABA/BZ receptor" came into use for this complex (and is still encountered). Progress in this field until recently was driven by the synthesis of a vast range of BZs and BZ-like drugs, all acting at brain GABA_A receptors and possessing clinical anxiolytic or sedative potencies correlated to their binding affinities there (Haefely et al., 1985).

View larger version (39K): [in this window] [in a new window]   Fig. 1.   A diagrammatic representation of the spectrum of ligands with different efficacies, positive or negative, at the BZ binding site, and their allosteric actions on the GABA site (Haefely, 1989). The ligand efficacy depends on subunit composition. The evidence used was based on either whole animal responses or wild-type receptors. A similar profile would be obtained in, e.g., 1n2 recombinants but not at some other subtypes.

View larger version (20K): [in this window] [in a new window]   View larger version (15K): [in this window] [in a new window]   View larger version (10K): [in this window] [in a new window]   View larger version (16K): [in this window] [in a new window]   Fig. 2.   Structures of ligands discussed in the text. A, acting at the BZ site; B, C, and D acting at other sites.

4. Excitatory -aminobutyric acid_A receptors. Yet another apparent distinction between sets of GABA receptors arises from observations that GABA can be an excitatory transmitter at certain loci in embryonic and early postnatal life in the mammal (reviewed by Cherubini et al., 1991; Ben-Ari et al., 1997). The excitatory response also may mediate the observed trophic role for GABA in nervous system development (Ben-Ari et al., 1997). Another form of excitatory GABA response is seen in tonically stimulated adult hippocampal pyramidal neurons (Staley et al., 1995; Perkins and Wong, 1996; Kaila et al., 1997). All the evidence on these excitatory receptors indicates that they are GABA-activated anion channels, in general similar to the inhibitory GABA_A receptors. GABA_A receptors therefore should be classified as one general type, whether their transduction is a depolarization or a hyperpolarization of the cell membrane. The subunit composition of these excitatory receptors has not been determined yet. It is possible that a different subunit composition increases the permeability of bicarbonate relative to chloride through the receptor channel, or that the subtypes involved are not necessarily different from those well known in the adult but that the chloride gradient across the cell membrane is inverted at the sites in question. Either of these situations could explain the observed excitatory GABA_A receptor activity. The relative bicarbonate permeability of the channel rarely has been measured for any identified GABA_A receptor subtype, but the possibility that it is increased in a particular case has been supported by Staley et al. (1995) and Perkins and Wong (1996). It need not be assumed that this would involve a receptor outside the range of GABA_A receptor subtypes. Indeed, Kaila et al. (1997) have shown that activity-induced changes in intracellular chloride and bicarbonate and extracellular potassium, along with normal GABA_A receptor function, can account for the GABA-excitatory phase in the tonically stimulated adult hippocampus. Further, the intracellular chloride activity of developing neurons (of the rat nucleus basalis) has been measured using gramicidin-perforated patch recording and shows a large decrease from the immediately postnatal to the mature brain, sufficient to account for the excitatory and inhibitory responses, respectively (Akaike et al., 1996). Likewise, the internal chloride concentration can be measured locally by confocal imaging based on a chloride-sensitive fluorescence, and this has shown that dendrites on some hippocampal or cortical neurons can exhibit a higher value than somatic locations (Inglefield and Schwartz-Bloom, 1996), confirming earlier suggestions of such a gradient. This also must be distinguished from subtype difference as a potential cause of the excitatory behavior of GABA_A receptors on some dendrites. As further evidence for this, in the mature mammal the pituitary melanotropic cells are known to possess a very high internal chloride level, and activation of GABA_A receptors (of normal pharmacology) there also is depolarizing (Tomiko et al., 1983). For all these reasons, it is unnecessary to provide a specific designation for receptors that mediate excitatory neuronal responses to GABA.

	II. Approaches to the Classification of the -Aminobutyric AcidA Receptors

Top Previous Next References

It previously has been accepted in this series of receptor classifications (see Hoyer et al., 1994) that the most fruitful comprehensive system is one in which evidence from three approaches, operational, structural, and transductional, is applied. How can these be applied here?

For an ionotropic receptor the transduction (intrinsic ion channel opening or closing) is by definition the same for all of its subtypes. The alternative intracellular pathways used in other receptor classes have no counterpart here. However, in principle subtle differences within this single transduction pathway still could occur. Thus, for two subtypes being activated by the same agonist it might be possible to measure different kinetic constants in the channel opening or closing steps, or different desensitization behaviors, or different distributions of the open and closed channel states. Such cases are known for subtypes of other transmitter-gated channels, e.g., glutamate receptors or P_2X nucleotide receptors (reviewed by North and Barnard, 1997). Differences in the kinetic properties among GABA_A receptor subtypes have been investigated rarely so far. In one case, Angelotti and Macdonald (1993) found that some difference in the single-channel properties could be discerned in two recombinant GABA_A receptors expressed in a nonneural cell line. Likewise, expressed recombinant receptors containing the ₆ subunit exhibit, at least in some cases, distinctive channel properties (Ducic et al., 1995). However, it may be difficult in practice to find such discriminatory differences in channel properties for many of the subtypes, as well as cumbersome to apply those in classification. Further, in the native setting it will be difficult, if not impossible, to determine whether any such difference instead is not caused either by some intracellular secondary reaction (e.g., a phosphorylation) or by the availability of a native modulator. Therefore, we will not consider transductional criteria for classifying these receptors.

Selective antagonists have been the most powerful operational tools for discriminating subtypes in other receptor classes (Kenakin et al., 1992). However, for the GABA_A receptors, antagonists at the GABA site generally produce convulsions in vivo. Hence, therapeutic potential is limited and systematic exploration of antagonists has not been developed. A few compounds unrelated to GABA, such as certain arylaminopyridazines and cognate compounds (Heaulme et al., 1987; Melikian et al., 1992), have been developed as potent antagonists at GABA_A receptors generally. Olsen et al. (1990) have shown that the binding of such compounds can discriminate between some subtypes in the brain; their functional study to identify selective actions on recombinant subtypes could be rewarding.

GABA_A receptors are endowed with a variety of modulatory sites for which ligands have been found that can allosterically control the activation by GABA and/or the opening of the anion channel. With the possible exception of the N-methyl-D-aspartate subclass of glutamate receptors (another family of heteromeric ligand-gated ion channels ubiquitous in the brain), the number of different regulatory sites is greater than for any other receptor type. Modulatory sites offer the potential for discriminating among receptor subtypes, namely by the discovery or the design of agents that can act at these sites but can recognize differences in a given site as it occurs in different subunit combinations. Thus far, this possibility has been realized to some extent with the site at which BZ and molecules with BZ-like activity bind, as we shall see. Other established modulatory sites, which can exist on these receptors and which might be used thus include those for barbiturates, neuro-steroids, propofol, certain other anesthetics, furosemide, zinc, picrotoxin, and some other channel blockers, loreclezole, substituted pyrazinones, and dihydro-imidazoquinoxalines. Those compounds and the evidence of their interaction with GABA_A receptors are reviewed by Sieghart (1995), Im et al. (1993a,b), Wafford et al. (1994), and Korpi et al. (1995). Only occasional clues to subunit selectivity have been obtained for any of the latter sites.

	III. The Structures of the -Aminobutyric AcidA Receptors

Top Previous Next References

Cloning from cDNA libraries or genomically so far has generated 19 related GABA_A receptor subunits in mammals, which are each encoded by different genes. These now comprise 6, 4, 3, 1, 1, 1, and 3 mammalian types [for references see Burt and Kamatchi, 1991; plus () Davies et al., 1997 and Whiting et al., 1997; () Heblom and Kirkness, 1997; (_1-3), see Section I.A.2.; for database accession numbers see fig. 3]. These polypeptides are all ~50,000 daltons in size, and each carries four putative transmembrane hydrophobic segments (TM1-4). Figure 3 illustrates the seven different sequence families into which these fall structurally and their relationships. A mammalian counterpart of the avian ₄ subunit (Harvey et al., 1993) has not yet been isolated by cDNA cloning and so is not included here. However, the ₄ subunit gene, likewise discovered in the chicken (Bateson et al., 1991), has been shown more recently in humans (Levin et al., 1996).

View larger version (16K): [in this window] [in a new window]   Fig. 3.   Dendrogram depicting the relatedness of amino acid sequences between GABAA receptor subunits. Amino acid sequences were retrieved from the SWISS-PROT or NCBI databases. Most GABAA receptor sequences are available from the rat and so were used for the analysis, except for the 4 and  subunits. The chicken 4 subunit sequence was incorporated because the human homologue has been identified recently (Levin et al., 1996), although only a partial human sequence has yet been published. Similarly, the human  subunit has been identified (Davies et al., 1997), but the rat sequence has yet to be published. Thus, accession numbers of GABAA receptor subunit sequences used were: 1, P18504; 2, P23576; 3, P20236; 4, P28471; 5, P19969; 6, P30191; 1, P15431; 2, P15432; 3, P15433; , P18506; , U66661; 1, P23574; 2, P18508; 3, P28473; , U95368; 1, P50572; 2, P47742 and 3, P50573. Outgroup sequences (see below) were all from the rat: nicotinic acetylcholine receptor  (nAChR ), P25110; nicotinic acetylcholine receptor 2, P12389; and 5-hydroxytryptamine type 3 receptor, P35563. The predicted signal-peptide cleavage sites of all subunits were determined by the method of Nielsen et al. (1997). These did not always correspond to those previously indicated in the corresponding database entries and in such cases the newly determined cleavage sites were taken to be the more accurate. These were (signal peptide length in parentheses): 5 (25), 3 (25),  (21), and 3 (25). A multiple alignment of the predicted mature peptides and the consequent phylogenetic tree descriptive file were created using CLUSTAL W version 1.7 (Thompson et al., 1994) under the default parameters. NJPLOT (Perriere and Gouy, 1996) was used to generate the graphic output of the gene tree. The branch root was determined by including in the analysis sequences of two non-GABAA receptor subunits from the same superfamily (Barnard, 1996b). Shown in this figure is the tree generated from the rat nicotinic acetylcholine receptor  subunit and the 5-hydroxytryptamine type 3 (5HT3) receptor subunit as outgroup representatives. No significant differences were found when either of these were substituted with the rat nicotinic acetylcholine receptor 2 subunit sequence. The sum of the horizontal branch lengths connecting any two sequences represents the fractional divergence in their amino acid sequence, the scale bar corresponding to 10% sequence divergence. Vertical branches connecting groups are presented only for clarity, and their lengths do not infer differences between separate sequences, or groups of sequences, on the tree.

This heterogeneity is increased by alternative exon splicing of the pre-mRNA, which generates two forms of the ₂ subunit from one gene (Whiting et al., 1990; Kofuji et al., 1991), which can be distributed differently in the brain (Glencorse et al., 1992). Two such forms are also known for the ₂ and ₄ subunits (Bateson et al., 1991; Harvey et al., 1994). In each case, the longer and shorter products were designated "L" and "S," and differ by some form or other of a short peptide in the long intracellular loop between TM3 and TM4. Splicing also occurs to express two alternative forms of exon-1 of the ₃ subunit (Kirkness and Fraser, 1993). Three potential forms of the ₅ subunit mRNA also exist (Kim et al., 1997) but with unchanged protein sequence. Another product of alternative splicing deletes a short sequence at the N-terminus of the ₆ subunit (Korpi et al., 1994), although this abolishes the functional receptor activity in all the combinations tested so far. Therefore, in assessing possible combinations of subunit types (other than ) to form a GABA_A receptor, we must consider in a given mammalian species, including the splice variants, at least 7  forms, 7  forms, 4  forms, 1 , 1 , and 1  form. The recently discovered  and  subunits in each case can combine with  and  subunits to form a functional, BZ-insensitive receptor (Davies et al., 1997; Heblom and Kirkness, 1997; Whiting et al., 1997). The  subunit has been detected clearly so far only in certain peripheral tissues (Heblom and Kirkness, 1997), and its range of combinations has not been defined yet. The GABA_A receptors in the central nervous system (CNS) are formed, on present knowledge, by combinations of both  and  subunits with one or more of the ,  or  subunit types (or possibly, exceptionally, of  and  types alone). In addition there are three known  subunits that occur in the retina: ₁ (Cutting et al., 1991); ₂ (Cutting et al., 1992; Kusama et al., 1993b); ₃ (Ogurusu and Shingai, 1996; Shingai et al., 1996). In co-expressions, evidence was not obtained to show that a  subunit can participate in combinations with the aforementioned , , or  types (Shimada et al., 1992; Kusama et al., 1993a), although more recently a ₁₂ heteromer forming in heterologous expression was suggested (Pan et al., 1997). In the rat retina, however, a recent study by immunofluorescence microscopy showed punctate localizations of non- GABA_A receptors and of -containing receptors, which occur at different synapses and do not overlap (Koulen et al., 1998). Hence, a pool of at least 20 subunit types may be used in forming combinatorially the CNS GABA_A receptors, plus at least 3  subunit types which assemble in a restricted manner.

To understand the construction of GABA_A receptor subtypes from this repertoire of subunits, it is necessary first to establish the total number of subunits in each receptor molecule, then to know whether this number is constant for all the native compositions, and finally to know the stoichiometry of the subunit types within that number. Regarding the number of subunits per receptor, the suggestion often has been made that this will be the same (five subunits) as for another transmitter-gated ion channel where the composition has been established unequivocally. Thus, the GABA_A receptor subunits share a low but definite (~25%) amino acid sequence homology with the subunits of the nicotinic acetylcholine receptors, both being in the same superfamily of the transmitter-gated ion channels (Schofield et al., 1987; Barnard, 1996b). The muscle type of that receptor occurs in Torpedo electric organ at such a high density in large postsynaptic membrane sheets that it is possible to prepare membranes containing a surface lattice of the receptors, from which a low-resolution three-dimensional structure of the molecule could be obtained by electron optical diffraction techniques (Toyoshima and Unwin, 1988; Unwin, 1993). Those studies clearly showed that the muscle type nicotinic receptor is pentameric, with the ion channel located in the center of a rosette formed by five homologous subunits (with the stoichiometry ₂ ).

For the GABA_A receptors, the situation is necessarily more complex, because the unique situation in the Torpedo postsynaptic membranes does not recur in the mammalian CNS and because there are many types of subunits involved, in varying combinations, in the receptor population. It is preferable, therefore, to use the natural GABA_A receptor population from the brain rather than a selected recombinant composition expressed in a nonneural cell, which may or may not be representative of the native population; further, when direct analyses are made on the latter, these will not be limited by an assumption of the subunit classes to be taken as co-assembling. Using purified GABA_A receptors from pig brain cortex and image analysis in the electron microscope, dispersed single receptor molecules can be visualized and analyzed (fig. 4). This method yields a power spectrum for each particle with a peak at its dominant symmetry. Figure 4 illustrates that this symmetry is five-fold, over the population of particles analyzed (Nayeem et al., 1994). Further, the negatively stained images obtained for all the receptor particles indicated a central pore in the pentameric rosette. These data correspond to the images observed with negatively stained Torpedo receptor particles, because of a central channel in the membrane enclosed within the pentameric receptor in the latter case (Toyoshima and Unwin, 1988). The particles isolated from brain will comprise a variety of GABA_A receptor subtypes. These data show that at least the majority of those receptors are pentameric; a deviating small minority with an atypical subunit number would not be distinguished from the experimental noise. Independent evidence to support the pentameric structure has been obtained in several ways. Hydrodynamic estimates of the size of GABA_A receptors, either native (Mamalaki et al., 1989) or ₁₃₂ recombinants (Tretter et al., 1997), in solution are consistent with the pentameric molecular weight. Further, the integral ratios of the subunits combined in several forms of functional recombinant receptors, as determined by diverse methods, fit best in each case with a subunit total of five (Im et al., 1995; Chang et al., 1996; Tretter et al., 1997). For parallel evidence, a method similar to that of Nayeem et al. (1994) has been used for the native 5HT₃ receptors by Boess et al. (1995) and there is supporting evidence by other methods for the neuronal nicotinic receptors and the glycine receptor (reviewed by Barnard, 1996b), all of these being in the same superfamily and all being deduced to be pentameric. In view of this concurrence with other receptors in the same superfamily, it is presumed that the pentameric structure that has been observed, within experimental error, for GABA_A receptors holds for at least the great majority of that receptor type.

View larger version (16K): [in this window] [in a new window]   Fig. 4.   Evidence for the pentameric structure of native GABAA receptors. The electron microscopic images of a population of pure GABAA receptor molecules analyzed to yield dominant symmetry for each particle were used in plotting the histogram (for details see Nayeem et al., 1994). Particles with only one-fold apparent symmetry, which is trivial, were rejected. The form of the distribution seen around the peak at five-fold symmetry is consistent with 100% being pentameric, because the apparent spread to some lower symmetries can be caused by tilted particles. The distribution shown was confirmed on a large number of the particles (from Nayeem et al., 1994; E. A. Barnard, personal communication).

It can be concluded that the repertoire of (at least) 24 mammalian subunit isoforms described above is drawn on to a total of 5 for each receptor molecule. Evidence discussed below will show that in most (but not all) GABA_A receptors , , and  subunits co-exist in one molecule, and that yet other combinations exist accommodating in special ways the other subunit types, , , , and (separately) . It is assumed that the -containing receptors are also pentameric, but this question has not been studied as yet.

The majority of GABA_A receptors contain, as noted, , , and  subunits, whereas the total number of subunits per receptor is five (fig. 4). Hence the receptors in this set can have at least one of three general compositions: 2.2.; 2..2; .2.2. Here, a notation is introduced in which the numeral represents the number of molecules of a given subunit class (, , etc.) present in one receptor molecule and not the isoform identity within that class; separating points are then used, and are absent when the stoichiometry is not being indicated. Such additional cases as 3.. and ..3 are theoretically possible, but measurements of an electrophysiological property determined quantitatively by the number of tagged recombinant subunits of each type forming the channel (Backus et al., 1993; Chang et al., 1996), in the cases of co-expression of the ₃₂₂ or ₁₂₂ subunits, have excluded (at least in those cases) the presence of three of any of those types in one receptor molecule. The next logical step in enumerating the potential combinations of subunits, therefore, is to ask whether two isoforms of  or of  or of  can occur in one receptor molecule, e.g., to produce compositions of the type (₁₂).2..

In the  subunits, there is a variety of evidence for such a co-occurrence of isoforms in a minority of GABA_A receptors. This evidence has come first from co-precipitation of a second  isoform when a brain-derived population of GABA_A receptors is treated with an antibody specific for a first  isoform. Receptors containing at least the pairs ₁₂, ₁₃, ₁₅, ₂₃, and ₃₅ have been detected thus (each in a minority, with the majority of receptors in the population containing a single  isoform) (Duggan et al., 1991; Luddens et al., 1991; Zezula and Sieghart, 1991; Endo and Olsen, 1993; Mertens et al., 1993; Pollard et al., 1993; Khan et al., 1996; McKernan and Whiting, 1996). Further, for the ₆ subunit that occurs (in the mature brain) only in the cerebellar granule cells (Laurie et al., 1992; Thompson et al., 1992) and in the similar granule cells of the cochlear nucleus (Varecka et al., 1994), antibody reactivities show ₁ and ₆ co-occurring in one cerebellar receptor (Pollard et al., 1995; Khan et al., 1996), although not for all the ₆ subunits there. For the  subunits, the similar use of isoform-specific antibodies has, on brain extracts or purified receptor preparations, shown evidence for the co-occurrence of ₂ with ₃ and also of _2L with _2S (Khan et al., 1994a,b; Quirk et al., 1994a).

A second method for investigation of possible co-occurrence of particular isoforms is the application of isoform-specific antibodies in situ, i.e., in light or electron microscopic studies (table 1). Thus, by confocal laser microscopy with double or triple immunofluorescent staining, Fritschy et al. (1992) and Mohler et al. (1996a) have found that certain  pairs were co-localized on the membranes of various neurons (table 2). Co-localization of ₁ and ₆ subunits in single synapses of rat cerebellar granule cells also has been demonstrated by double-antibody labeling in postembedding electron microscopy (Nusser et al., 1996), although when used in a freeze-fracture method on such cells in culture a co-localization of ₁ and ₆ was not seen (Caruncho and Costa, 1994). Third, some electrophysiological properties of a recombinant assembly containing two isoforms of  can be distinct from those with either isoform separately, demonstrated with ₁₃ or ₁₅ pairings (Ebert et al., 1994; Verdoorn, 1994).

View this table: [in this window] [in a new window]   TABLE 1 Methods for recognition of -aminobutyric acidA receptor subtypes in situ

View this table: [in this window] [in a new window]   TABLE 2 Some of the -aminobutyric acidA receptor subtypes in specified rat neuronsa

Further, the  subunit often has been found to replace  subunits: Quirk et al. (1995) found that  and  are completely separable by antibodies [although Mertens et al. (1993) found some co-existence].  subunits were present in only 11% of all the receptors in rat brain but in 27% of those in rat cerebellum, from which both _6n and _6n2 combinations can be isolated (Quirk et al., 1995) and where Caruncho and Costa (1994) found in situ that the receptors contain either a  or a  subunit, but not both, by a label-fracture method. The subunit  (which has some similarities to ) also may replace  in some cells of the hypothalamus and hippocampus (Whiting et al., 1997).

Receptor gene knock-out can provide additional evidence. In favorable cases it can show the co-occurrence of certain pairs of subunits. Thus, the homozygous mice lacking the ₆ gene also lack the  subunit protein in the cerebellar granule cells and the results obtained support other evidence that ₆ and  are paired in receptors there and not ₁ and  without ₆ (Jones et al., 1997). The specific pharmacology of ₆-containing receptors was confirmed in vivo in this system (Mäkelä et al., 1997).

On the basis of the extensive evidence reviewed above, that two isoforms of the  subunit can sometimes occur in one receptor, the receptors are considered as having two  places in the pentamer. Pollard et al. (1995) supported this by quantitation in the ₁₆-containing cerebellar receptor. Likewise, Khan et al. (1994a,b) and Quirk et al. (1994a,b) found that two different  isoforms can co-occur, although Mossier et al. (1994) and Im et al. (1995) did not find this; if the former statement holds two  places can also be in the pentamer. However, it is not known whether any conclusion of this type would apply to the entire native population of GABA_A receptors. Because  has been observed at some sites (see above) to occur with  and  subunits only, a plausible model is that either  or  (and perhaps ) subunits can occupy the  places (in different receptors). We will give an illustration here, only of the basis on which the theoretical maximum number of receptor compositions may be assessed.

Some of the native GABA_A receptors may have the stoichiometry 2..2 (with the  subunits in some cases being replaceable by  or by ). Several lines of evidence support this. Thus, for the ₂-containing receptors, there is evidence from immunoprecipitation analyses (as noted above) that the ₂₃ pairing within one receptor molecule can occur in some cases (Khan et al., 1994b; Quirk et al., 1994a) and also that a subset of receptors in the cerebellum has the composition ₁₆.._{2S 2L} (Khan et al., 1994b, 1996). Further, Backus et al. (1993) have deduced, by incorporating mutant subunits with altered electrophysiological effects in the recombinant ₃₂₂ receptor [expressed in human embryonic kidney (HEK) 293 cells], that the 2..2 composition best fitted the properties found.

On the other hand, Chang et al. (1996), using a similar principle (in oocytes and using ₁, not ₃ subunits), found that there the evidence apparently favors the 2.2. composition. The same stoichiometry also was derived for ₁₃₂ receptors, when expressed in HEK 293 cells, from the staining ratios of those subunits when separated in Western blots (Tretter et al., 1997). Moreover, the co-occurrence of ₁ with ₃, and of ₂ with ₃ (but not ₁ with ₂), isoforms has been indicated in some of the receptors from rat cortex by immunopurification (Li and De Blas, 1997) and likewise in rat cerebellum (Jechlinger et al., 1998). Benke et al. (1994) compared the fractions from rat whole brain containing ₁, ₂, or ₃ subunits by immunoprecipitation and also excluded the ₁₂ combination; however, in contrast to the findings just noted, they found that the ₁₃ or ₂₃ pairings also were absent. Overall, it is desirable to allow for possible 2.2. forms in the nomenclature. In view of this situation and of the evidence for 2..2 combinations, Li and De Blas (1997) suggested that the ratios of the  and  subunits in the molecule (within the total of 5) may vary with the isoforms selected.

In the expression of recombinant receptors in either cultured cells or oocytes, any ternary combination of the _{i j k} type tested so far can yield a functional receptor in the membrane (e.g., Kirsch et al., 1995). The limit to the number of ternary subtypes in vivo apparently is not set by barriers to the co-assembly in certain cases but by the program for gene expression of different isoforms in a given cell. However, in a case where this was tested (Angelotti and Macdonald, 1993), when such an ++ set is expressed the ternary combination assembles (as far as the subunits are available) and is maintained at the cell membrane to the exclusion of binary combinations. Within the ternary assemblies, there are no obligate combinations or exclusions of pairings known from the co-distribution data at the present resolution limits. However, some exclusions are known (see above) at the / position. Moreover, the  subunit has a more restricted expression in the brain than  subunits (Wisden and Seeburg, 1992) and has fewer co-occurrences with other subunits; the same is true for  (Whiting et al., 1997), whereas  is clearly detectable in certain peripheral tissues only (Heblom and Kirkness, 1997). Hence those subunits cannot be included on the same basis as the others in permutations of the possible compositions.

An enumeration is obtained on the basis that, for a given subunit set which will form one receptor, there will only be one arrangement and stoichiometry in the molecule. This is found to be so with all other heteromeric proteins containing tightly-bound subunits; for example, there is only one cyclic order of the subunits, .,.,.,.,, present in the population of Torpedo acetylcholine receptors (Karlin, 1991). Moreover, with those subunits one does not find that the same receptor type, in a variety of skeletal muscles, can contain another stoichiometry. That constancy and the circular order of subunits around the rosette are fixed by the interactions between the interfaces of different subunits. In the case of a GABA receptor (the recombinant _112S), supporting evidence for a single configuration in the population, from the homogeneity of the channel properties, has been reported (Angelotti and Macdonald, 1993).

Therefore we do not count all possible permutations (n = 36) of 2  isoforms present (out of the 6), but only those for a fixed order (n = 21); likewise, for the others. Splice variants could increase these numbers on the same basis, except that two alternatives from one subunit cannot be assumed to be able necessarily to co-occur. From what is known so far of excluded compositions and the restricted co-occurrence of subunits in certain cases, Barnard (1996a) has suggested that a maximum of the order of 800 combinations, of the types observed so far, would then be calculated. The true number is likely to be far smaller than this, but still much larger than for other known receptor subtypes.

The  subunits apparently assemble separately from the others (discussed below). They do not affect the enumeration above, but can add a few separate subtypes in the total noted above.

	IV. Principles of the Classification

Top Previous Next References

A. Application of Selectivities at the Binding Site for Benzodiazepines and Their Functional Analogs

1. The choice of a classification system. As noted above, this modulatory site presently offers by far the richest pharmacology for distinguishing subtypes of the GABA_A receptors. BZs have no intrinsic activity on mammalian GABA_A receptors, unlike some of the other modulators such as anesthetics (although such a direct effect of some BZs has been found at invertebrate GABA_A receptors; Zaman et al., 1992). Most BZs act to enhance the action of GABA by increasing the frequency of channel openings and their bursts (Rogers et al., 1994). This can be explained partly by the ability of BZs to increase the affinity of GABA at its binding site. However, in tonic activation of hippocampal neurons by low GABA concentrations they also can increase the channel conductance (Eghbali et al., 1997).