1. BZD Modulation of GABA Binding
Benzodiazepine (BZD) binding to the classical extracellular domain (ECD) site does not, by itself, result in any appreciable opening of closed channels but rather modulates the ability of agonists such as GABA to evoke pore opening. The enhanced GABAergic inhibition conferred by BZD PAMs largely stems from their potentiation and prolongation of GABA
AR postsynaptic currents
[1][2][3]. Strikingly, potentiation of current responses by BZDs occurs only for activation by subsaturating but not saturating concentrations of GABA, resulting in a leftward shift of the GABA concentration–response relation. This suggests that the mechanism underlying this potentiation involves enhancement of GABA binding
[4][5][6][7][8]. At lower subsaturating GABA concentrations, where not all sites are occupied, enhanced GABA binding effectively results in larger current responses equivalent to those elicited by a higher concentration of GABA. In contrast, at saturating GABA concentrations, where all sites are already occupied, enhanced binding has no further effect. Slower GABA unbinding due to the stabilization of the bound complex can also explain the prolongation of the current decay by BZDs. Consistent with this idea, DZ has been shown to primarily affect the frequency of channel opening, which depends on the rate of GABA binding, whereas DZ has little effect on the duration of each opening event, which is governed by the energetics of pore closure downstream from the initial binding events (but see
[6] as well as
[4][9]).
For an effect on GABA binding, crosstalk between GABA and BZD sites is required. The enhanced binding of radiolabeled DZ in the presence of GABA in brain membranes supports the idea that agonists and BZD PAMs are costabilizing
[10][11]. Observations of the rate of modification by methanethiosulfonate (MTS) reagents of introduced cysteines in GABA or BZD ECD sites indicate that binding at either the GABA or BZD sites induces distinct structural changes at the other site
[12][13]. The dependence of these changes on the nature of the bound ligand (agonist or PAM, antagonist, or inverse agonist or NAM) suggests that these motions are relevant to the action of the ligand and indicate that GABA increases the aqueous accessibility to the BZD binding pocket.
For all ligand-gated ion channels, ligands that open the channel (agonists) bind more tightly to open (conducting) versus closed (non-conducting) channel conformations. Thus, any perturbation (e.g., BZD binding) that increases the probability of a channel to adopt an open conformation will necessarily increase the receptor’s apparent affinity for GABA, even if it has no effect on the affinities for each distinct closed and open state. Indeed, this can severely complicate the interpretation of equilibrium binding measurements
[14][15]. Although BZDs clearly induce shifts in the apparent affinity for GABA, it is not yet fully understood to what degree this reflects direct changes in binding affinities vs. shifts in the closed–open equilibrium.
2. BZD Modulation of Pore Gating
Although some single-channel observations are consistent with the idea that BZDs primarily modulate GABA binding
[6], other studies observed changes in macroscopic desensitization or single-channel open durations that are inconsistent with effects solely on agonist binding
[9][16][17]. In addition, two lines of evidence strongly indicate that BZDs alter channel gating steps downstream of GABA binding. First, peak current responses to saturating concentrations of partial agonists are potentiated by BZDs
[18][19][20][21]. Partial agonists bind at the GABA sites but elicit smaller current responses than full agonists such as GABA. Nonetheless, at saturating concentrations the sites are fully occupied, and thus the increased maximal current response conferred by BZDs must reflect the modulation of gating conformational changes in the partial agonist-bound receptor, regardless of any effects on the binding step itself.
Second, BZDs directly gate mutant receptors in the absence of GABA
[18][19][20][21][22]. A mutation in the pore gate confers a channel that spontaneously exchanges between closed and open conformations
[23][24]. The closed–open equilibrium of such spontaneously active mutants is modulated by BZDs alone, an effect that necessarily does not involve the binding of agonists such as GABA. Although these observations require mutating the receptor in such a way as to significantly alter the resting equilibrium, it is likely that the mutant still gates by the same mechanism as wild-type channels. In support of this idea, a single-channel analysis of mutants and combinations thereof that confer such spontaneous unliganded gating in homologous nicotinic acetylcholine receptors (nAChRs) showed that these mutants alter the closed–open equilibrium independent of agonist-elicited gating
[25]. This suggests that the chemical energy from BZD binding does affect the pore gate, but the amount of energy is insufficient to elicit appreciable channel opening without additional energy from either agonist binding or another perturbation, such as a mutation.
3. BZD Modulation of Closed-Channel Pre-Activation
Single-channel gating dynamics for full versus partial agonists at homologous nAChR and glycine receptors (GlyRs) can be rationalized by a mechanism whereby agonist binding is associated with the receptor adopting a “flipped” or “primed” pre-active conformation that precedes pore opening
[26][27][28][29]. This model discriminates between partial and full agonists based on their ability to bias the receptor towards this pre-active intermediate state, following which pore opening/closing is ligand-independent.
Strikingly, if one assumes that BZDs influence a similar “flipping” or “priming” pre-activation step in GABA
ARs
[30], apparently conflicting observations supporting BZD-modulation of either agonist affinity or pore gating can both be readily explained
[31][32][33]. By modulating the probability of receptor pre-activation, BZDs regulate the amount of time spent in a state from which the channel can open. Under conditions where pre-activation is not highly probable, enhanced pre-activation by BZDs will increase the frequency of channel opening. This explains the observed BZD-potentiation of currents evoked with either partial agonists or subsaturating GABA concentrations where only one of the two agonist sites are occupied. Conversely, in saturating GABA with both sites occupied, the channel is already so strongly biased towards its pre-active state that a further enhancement of pre-activation by BZDs has little to no additional effect on channel opening. This explains the lack of potentiation of currents evoked by saturating GABA. A similar mechanism can explain the modulation by non-BZDs such as the Z-drug zolpidem, which binds at the canonical BZD site
[34].
Given its explanatory power, it is highly plausible that such a mechanism underlies many of the observed effects of BZD binding in the ECD. Nonetheless, direct evidence for this mechanism in GABA
ARs is lacking. In GlyRs and nAChRs, the pre-active state preceding pore opening is very short-lived, and a similarly transient non-conducting intermediate in GABA
ARs challenges both functional and structural observations. Moreover, observations that single-channel open durations are altered by partial versus full agonists or BZDs are not explained by changes in pre-activation alone, which only affect the opening frequency
[9][35]. Some of these complications may reflect BZD binding to TMD sites, which based on their proximity, could directly affect the pore gate. Otherwise, the totality of the effects of BZDs at the canonical ECD site is at least a bit more complex than the satisfyingly simple approximation that they modulate channel pre-activation.
4. Structural Mechanism for BZD Modulation in the ECD
Advances in cryo-EM have begun to provide structural snapshots of heteromeric GABA
A(α1)
2(βX)
2(γ2)
1 receptors in complex with both agonists and BZDs
[36][37][38][39][40][41][42]. These studies validate the canonical high-affinity BZD site in the ECD, as identified by functional studies, and reveal details as well as subunit specificity of lower-affinity BZD sites in the TMD (
Figure 1A–C). Together with structures of homologous pentameric ligand-gated ion channels (pLGICs) such as nAChRs, GlyRs, 5-HT
3 receptors, and prokaryotic homologues
[43], these structural snapshots are generally in agreement with the overall gross subunit motions occurring during agonist gating. A comparison of agonist-bound structures with either unliganded or antagonist-bound structures indicate that agonist binding is associated with the ECDs adopting a more compact conformation with an increase in the buried surface area at the intersubunit interfaces and loop C moving inward over the agonist. In addition, the ECD of each subunit undergoes a rotation and tilt that results in an expansion at the ECD/TMD interface, where the M2-M3 linkers move radially outward and the top of the pore-lining M2 helices splay apart, resulting in the opening of the hydrophobic 9′ leucine ring near their center, which constitutes the activation gate.
Figure 1. Structures of synaptic GABAARs in complex with the BZD PAM diazepam (DZ) at both high-affinity ECD and lower-affinity TMD sites. (A) Top-down view with GABA bound at the β+/α− interfaces and DZ bound at the α+/γ− interface in the ECD. Cryo-EM map of GABAA(α1)2(β3)2(γ2)1 from PDB 6HUP. (B) Side view of the structure in A, additionally showing one of several binding sites for DZ in the TMD. Only three subunits are shown for clarity. (C) Same perspective as in (A) for a slice through the TMD. Cryo-EM map of GABAA(α1)2(β2)2(γ2)1 from PDB 6X3X. Three binding sites for DZ are highlighted at the β+/α− and γ+/β− intersubunit pockets between the TM helices and below the M2-M3 linker of one of the subunits. The central pore-gate-forming 9′ leucine residues are shown as sticks near the middle of the pore-lining M2 helices. (D) Biphasic modulation by DZ at the canonical high-affinity site in the ECD and lower-affinity sites in the TMD.
Comparisons of the structures of GABA
A(α1)
2(β2–3)
2(γ2)
1 receptors with and without classical BZDs such as DZ show, at most, only subtle conformational differences in the ECD
[38][39][40][44]. Based on this observation, Masiulis et al. suggested that BZD PAMs act at their high-affinity site to stabilize the α+/γ− subunit interface, which they proposed should aid in the global compaction of the ECD during gating and thereby promote channel opening
[39]. In contrast to PAMs, occupation of the α+/γ− ECD site by the BZD antagonist flumazenil confers a slight expansion in the ECD
[44].
Consistent with the idea that the stability of the α+/γ− subunit interface is important for BZD modulation, cysteine cross-linking between α and γ subunits in close proximity to the critical histidine residue at the back of the canonical BZD site in the α subunit β4–β5 linker mimics the action of BZDs and is reversible upon breaking the disulfide bond
[45] (
Figure 2ii). In contrast, the insertions of glycine residues to increase the flexibility of β4–β5 linkers in the α, β, or γ subunits all impair the BZD potentiation of GABA-evoked currents
[46] (
Figure 2iii). The importance of the linker in all subunits suggests that BZDs are associated with global changes that could couple BZD and agonist binding sites via interactions between the β4–β5 linker at the back of each binding site and the neighboring subunit beta barrel. Indeed, mutations affecting BZD modulation are not localized to a specific domain, supporting the notion of a more global effect on the channel conformation.
Figure 2. Canonical ECD binding α+/γ− interface, with diazepam (DZ) shown as white sticks bound behind loop C. Other loops implicated in binding or intersubunit interactions are indicated. Cryo-EM map for GABAA(α1)2(β3)2(γ2)1 from PDB 6HUP. Features discussed in the main text are indicated with spheres for residue Cα atoms and dashed lines for cross-linked residue pairs (numbered labels are referenced in the main text). The critical histidine residue in the β4−β5 loop within the BZD binding pocket (i) and the central valine residue in the α1 M2-M3 linker (viii) are shown as sticks.
Given the rigid body movement of the subunit ECD beta barrels, it is natural to expect much of the dynamics to occur at the interfaces between subunits, which include the GABA and BZD binding sites
[47]. Only one of the two GABA sites shares the same α subunit as the BZD site, raising the possibility that BZDs may induce local changes that preferentially modulate activation via a single GABA site. However, the serial disruption of each of the two GABA binding sites in concatenated receptors revealed that BZDs can modulate currents in response to GABA binding at either site
[48]. Apart from BZD- and agonist-binding α+/γ− and β+/α− interfaces, respectively, epilepsy-related mutations that disrupt electrostatic interactions at the canonical nonbinding γ+/β− interface reduce the stability of the BZD-receptor complex
[49]. Thus, the destabilization (or stabilization) of one interface results in a more global destabilization (or stabilization) at other interfaces, consistent with the idea that BZDs influence the stability of intersubunit contacts throughout the ECD.
Although it is likely that BZDs stabilize the α+/γ− interface, it is important to note that all structures of BZD-bound GABAA(α1)2(β2–3)2(γ2)1 receptors to date were obtained with both agonist sites occupied by GABA. Indeed, the inability of BZDs to potentiate currents evoked by saturating GABA is hypothesized to reflect a nearly complete biasing of the receptor to a pre-active conformation by the relatively larger energetic contribution from both bound agonists that effectively overshadows the much smaller contribution from BZD binding in the ECD. Thus, it is perhaps not surprising that little to no BZD-associated motions of the ECD were detected in this condition. Structures of BZD complexes with zero or one bound agonist or with partial agonists that may better inform on the motions associated with BZD binding are currently lacking.
Structural models of homologous GlyR and ACh binding protein in complex with partial agonists suggest that the initial pre-activation step involves loop C swinging towards the bound agonist to adopt a more compact conformation
[50][51]. Although the styrene maleic acid (SMA) copolymer used to solubilize GlyR in this research likely biased the conformations to some extent, this bias may have serendipitously aided the observation of a normally transient pre-active state. Taken together with the structures of GABA
A(α1)
2(βX)
2(γ2)
1 receptors, this suggests a hypothetical mechanism for BZD modulation via the canonical high-affinity site: BZD PAMs promote the adoption of a pre-activated more compact ECD conformation with loop C closed more tightly around the binding pocket by stabilizing or increasing global intersubunit interactions. It is straightforward to assume that NAMs might have an opposite effect by destabilizing intersubunit interfaces. Consistent with this idea, GABA
A(α1)
2(β2)
2(γ2)
1 receptor structures in complex with either a BZD antagonist or NAM exhibit a less compact ECD than complexes with PAMs
[40][42].