The daytime biological activity of humans is based on a delicate equilibrium between excitatory and inhibitory signalling in the brain. The neurotransmitter molecule that mediates the main inhibitory signal is γ-aminobutyric acid (GABA), which binds to and activates ion channels, including a family known as GABAA receptors1 found in the cell membranes of neurons. The activity of these receptors is modulated by many drugs2, including sedatives, anxiolytics and sleeping pills. GABAA receptors can be assembled from several types of subunit, and the most abundant GABAA receptors in the adult human brain consist of two α1 subunits, two β2 subunits and one γ2 subunit. In a paper in Nature, Zhu et al.3 present two structures of this receptor isoform in complex with GABA and a drug molecule known as flumazenil, providing much-needed insight that will aid future drug-discovery efforts.
The GABAA receptors that mediate fast signalling are found in several locations, most importantly at the synaptic junctions between neurons. Once activated by GABA, the channels in the receptors open, and conduct chloride ions through their pores. Benzodiazepines are a widely used class of drug that acts at some subtypes of GABAA receptor by binding to a specific site4 distinct from that at which GABA binds. These compounds do not induce receptor activity on their own, but increase activity triggered by GABA, a process known as positive allosteric modulation.
Zhu et al. prepared the receptors for their study by expressing them in human embryonic kidney (HEK) cells in the presence of GABA and flumazenil, which is an antagonist for the benzodiazepine binding site (that is, it blocks the site). The authors used cryo-electron microscopy to obtain two high-resolution structures of the synaptic α1β2γ2 GABAA receptor (see Fig. 1 of the paper3), which they propose are in non-conducting, desensitized states — closed states of the channel that occur when the receptors have had prolonged exposure to GABA. Crucially, the structure confirms that the number of each subunit type and the arrangement of the subunits in the assembled pentameric receptor correspond to what had previously been postulated1.
The structures reveal differences between the GABA binding sites (which are found at the two β2–α1 subunit interfaces in the receptor), the benzodiazepine binding site (which is found at the α1–γ2 interface, at an equivalent position to the GABA binding site), and equivalent sites at the α1–β2 and γ2–β2 interfaces to which neither GABA nor benzodiazepines bind (Fig. 1). These differences help to explain why the different sites bind (or do not bind) GABA or benzodiazepines. Both the γ2–β2 and α1–β2 sites are potential targets for drug discovery. A compound that binds at the α1–β2 interface has already been reported5, but no compounds have yet been found for γ2–β2. The new receptor structures will provide a template for molecular-modelling studies aimed at discovering more compounds that bind at the different interfaces, or to any of the many other pockets present in the receptor.
The extracellular region of the receptor forms a wide chamber. Zhu et al. find that most of the space in this chamber is occupied by complex carbohydrate chains attached to some of the subunits, leaving relatively little space for ion permeation. If the same is true for receptors expressed in neurons, rather than in HEK cells, then this observation is mechanistically highly relevant. Ions might also enter the channel through openings found at the subunit interfaces. A structure of the α1β1γ2 GABAA receptor was recently published6 on a preprint server, and the authors of that paper suggest that the attachment of carbohydrates to the α1 subunit has a key role in receptor assembly.
Understanding the variations between benzodiazepine binding sites found in different GABAA-receptor isoforms might help researchers to find compounds that act selectively at just one isoform. This could allow researchers to realize the hope of developing drugs that produce desirable therapeutic effects (such as reducing anxiety) without the unwanted side effects (such as sedation) caused by existing drugs that target several isoforms of GABAA receptors. How diazepam — the archetypal benzodiazepine drug — binds to benzodiazepine sites is of particular interest. Computational studies7 suggest that there are three possible binding modes (BM I–III) of benzodiazepines, of which BM I was proposed to actually occur. By contrast, an alternative approach has been used to identify the amino-acid residues in α1β2γ2 GABAA receptors that come into direct contact with diazepam8, and this pointed instead to BM II. The information from that study was used to identify high-affinity ligands for the benzodiazepine site.
In Zhu and colleagues’ structures, the position of flumazenil in the benzodiazepine site is similar to BM I. The authors suggest that docking of diazepam to the observed site would be possible through a binding mode similar to BM I, but that diazepam docking similar to BM II would lead to clashes between the molecule and the receptor. It should be noted, however, that the receptor conformations stabilized by flumazenil must be different from that stabilized by diazepam, so the idea that diazepam would dock into the reported structure in a similar way to flumazenil is questionable. A recent study9 has shown that benzodiazepines whose structures are similar to that of flumazenil use BM I, whereas those similar to diazepam use BM II. Thus, the positioning of diazepam postulated by Zhu et al. will probably need to be revised.
In structural biology, there is always the possibility that the protein structure determined does not entirely correspond to a conformation adopted in nature. This is illustrated by one of Zhu and colleagues’ structures (described as conformation A), in which the transmembrane part of the γ2 subunit is squeezed into the receptor’s pore. This conformation is unlikely to occur in nature, and might have been caused by the type of detergent that Zhu et al. used to replace the natural membrane environment in their microscopy experiments. The water-soluble domain of the protein is not subject to this problem.
Proteins are dynamic structures that assume distinct conformational states. In each state, large parts of the proteins vibrate much like wobbly puddings. The binding of molecules to allosteric sites in proteins (such as the benzodiazepine site in GABAA receptors) can stabilize different conformational states. This explains how molecules can act as positive or negative allosteric modulators, or as antagonists. GABAA receptors exist in at least four conformational states: closed; ligand-bound but not open (known as the pre-activated state); open; and desensitized. Therefore, insight into allosteric modulation cannot be gained from structural studies alone, because protein structures are caught in static conformations. Structural-biology methods will need to be combined with other approaches to probe the dynamic properties of GABAA receptors.
Nevertheless, Zhu and colleagues’ structures could well be the key to addressing some of the crucial issues in the study of GABAA receptors. For example, the work might help to unravel the numerous conformations of the α1β2γ2 GABAA receptor, and the differences in the benzodiazepine binding site between different isoforms of GABAA receptors, as well as providing information on all the possible modulatory sites. It is to be hoped that this will, in turn, lead to the development of improved drugs that target these receptors.
Nature 559, 37-38 (2018)