|[Frontiers in Bioscience 1, d214-233, September 1, 1996]|
MOLECULAR BIOLOGY OF THE GABAA RECEPTOR: FUNCTIONAL DOMAINS
IMPLICATED BY MUTATIONAL ANALYSIS
Martin Davies1, Alan N. Bateson1,2 and Susan M. J. Dunn1,2
1 Department of Pharmacology
2Division of Neuroscience,
Faculty of Medicine, University of Alberta, Edmonton, Alberta,
Canada T6G 2H7
2Division of Neuroscience, Faculty of Medicine, University of Alberta, Edmonton, Alberta, Canada T6G 2H7
Received 07/16/96; Accepted 07/22/96; On-line 09/01/96
This section will briefly overview some of the models that have been proposed to explain the ligand binding and channel opening properties of the GABAAR and other LGICs. As described above, a combination of photoaffinity labelling studies and mutational analysis has led to the assignment of TM2 as a channel lining structure. Based on the results of nAChR studies, a theoretical model of the ion channel was proposed (118). This model postulated that the channel is made up by a series of amino acid rings, each composed of specific types of residues projecting from the a-helical TM2 region of each subunit into a central lumen. It was further suggested that specific rings of amino acids underlie the ion and charge selectivity of the channel (119). However, further mutational analysis of the TM2 regions of LGICs has provided contradictory data, particularly with respect to the location of the charge selectivity filter and gate (see Section 5). Furthermore, there is also evidence that the TM2 region may not be completely a-helical in nature, but may be made up of both alpha-helical and ß-sheet structures (120).
An early model of neurotransmitter binding sites by Cockroft et al. (65) was based on the molecular modelling of the conserved cystine loop in the extracellular N-terminal domain. This region was examined because the loop structure is conserved throughout the known LGIC family and because several residues within this loop are either invariant or homologous. Thus, it was reasoned that the loop functions as a prototypical ligand binding domain for this superfamily of receptors, with ligand specificity arising from the presence of unique residues. The loop was modelled as a rigid ß-hairpin. Residues that were proposed to interact with GABA were a conserved aspartate residue (D11) proposed to interact with the positively charged amino group and a conserved aromatic group at position 8 (phenylalanine in alpha1 and gamma2, tyrosine in ß2) proposed to interact with the negatively charged carboxyl moiety. The residue at position 6 of the loop was proposed to be the residue conferring the specificity of the ligand interaction. However, as described above (Section 4.2.2), mutational analysis of this region in both the GABAAR and nAChR showed that the presence of this loop is more important for the correct assembly of the receptor in the cell membrane than for ligand recognition. It is, therefore, unlikely that the residues proposed in the molecular modelling study actually interact with the charged groups of the GABA molecule.
A model of the ligand binding domains of GABA and benzodiazepines has been proposed by Smith and Olsen (78) based on both biochemical data and mutational analysis. They propose the existence of two homologous binding domains for GABA and benzodiazepines within a single alpha subunit. Each structural domain is formed by two ß-sheets separated by an alpha-helical region. Each ß-sheet region contains a TXDXFF domain, and includes residues that have been photoaffinity labelled by either [3H]muscimol or [3H]flunitrazepam, or have been implicated in ligand binding by mutational analysis. This model suggests that the binding of ligand induces conformational twists in the alpha-helical segment which are then transduced to the ion channel. The binding of benzodiazepines to the second site was suggested to modulate the degree of twisting induced by the binding of ligand at the GABA site, thus allosterically modulating GABA-induced chloride flux. This model takes into account the probable involvement of residues of adjoining subunits in forming the binding site and proposes that, as has been suggested for the nAChR, ligands bind at subunit-subunit interfaces. The residues proposed by Smith and Olsen (78) to form the agonist site include loop A and residues N-terminal to this domain, in addition to putative loop D of the nAChR model. The ß-alpha-ß model does not, therefore, include the residues in loops B and C that Amin and Weiss (50) identified as being important for GABA recognition. In addition, there is evidence to suggest that several determinants for benzodiazepine efficacy lie further C-terminal to the binding domains proposed in the model (see Section 4.3).
Aprison and colleagues have also recently described models of agonist binding sites (121, 122). Their computer modelling approach used information derived from sequence homologies and mutational analysis. In addition to suggesting which residues may be involved in agonist binding, the most recent model (122) proposes a mechanism by which agonist binding induces chloride ion flux. Residues predicted to form the agonist binding site were D146, E155, T202, Y205, R216 and R269. Residues D146, Y205, and T202 are proposed to form electrostatic interactions with the positively charged amine of GABA, while the two arginines form hydrogen bonds with the carboxyl group. E155 is proposed to form a charge-transfer complex with R216, resulting in stronger bonding with the carboxyl group. These interactions then result in channel opening, during which the guanidinium group of R269, proposed to attract Cl- ions, is pulled to the mouth of the channel. In the same study, a nearly identical model based on homologous residues is proposed for the glycine receptor. Although not explicitly stated within the text, the model presented by Aprison et al. (121) deals only with the ß-subunit. The residues identified as forming the GABA binding site reside in loops B and C, which is consistent with Amin and Weiss' mutational analysis of the GABA binding site (50). However, R269 which is predicted to line the mouth of the channel, has not previously been identified as an important determinant of GABA binding. This residue has a homologue in the GlyR alpha subunit at position 219, which mutational analysis has implicated in subunit processing rather than ligand binding (38). Furthermore, D146 that is proposed to interact with the positively charged amino group of GABA has already been shown to have little effect on GABA recognition (66). Only two of the residues (T202 and Y205) have been shown to be important in agonist interaction by mutational analysis (50).
An alternative approach to define the basis for ligand-induced ion channel gating, is to determine the three-dimensional structure of the proteins. So far only very low resolution structural information is available for the GABAA receptor. However, electron microscopy has shown that, like the nAChR and 5HT3 receptor, the GABAAR exists as a pentamer, with the subunits arranged to form a central ion channel (20). Using cryoelectron microscopy of crystallised Torpedo membranes, the structure of the nAChR as been resolved to a resolution of 9Å (19). Analysis of computer enhanced images, suggests that each subunit contains three alpha-helical rods running perpendicular to the membrane. In two subunits (presumed, though not proved, to be the alpha subunits), the rods contain cavities which have been suggested to be the ACh binding sites. In an attempt to identify structural differences between the closed and open states, the position of specific regions of the receptor in the presence and absence of agonist has been investigated (123). Unfortunately, in this study, there is no evidence to suggest that the channel had actually opened in the presence of agonist. However, the results have led to the proposal of a model by which agonist binding induces channel opening. Acetylcholine is proposed to bind to the two alpha subunits, one of which is adjacent to the gamma-subunit (designated alphagamma ) and the other borders the delta subunit (designated alphadelta ). In this model, the alpha subunits are separated by the ß subunit, although most current models favour the gamma subunit in this position (40). The binding of agonist is considered to induce a twist of the three helical rods in the alphadelta subunit and, to a lesser extent, in the alphagamma subunit. The small conformational change in the alphagamma subunit pulls the relatively rigid ß subunit away from the alphadelta subunit, allowing it to twist more freely. This is suggested to be the basis for the cooperativity in binding that is seen in channel gating. The region identified as TM2 in each subunit is proposed to line the channel and is alpha-helical except for a kink about halfway down the lumen. This kink, which may correspond to the non-helical portion of the TM2 domain identified by Akabas et al. (99), has been suggested to form the gate in the closed conformation of the receptor. When agonist binds, the resulting conformational change is translated through to the membrane-spanning region causing the helices to rotate. As a result, the kink moves away from facing the center of the lumen, and a region of constriction is formed at the cytoplasmic end of the channel. Unwin suggests that the kink is formed by the bulky side chain of L251 which projects into the channel in the closed state (123). However, this is contradicted by SCAM studies of the nAChR in which the gate is proposed to be at least as cytoplasmic as E241 (99; see Section 5). SCAM also suggests that L251 faces a hydrophilic environment only when in the open state (95), which is again inconsistent with the above proposal.