Mechanisms of inhibition and activation of extrasynaptic αβ GABAA receptors

Abstract

Type A GABA (γ-aminobutyric acid) receptors represent a diverse population in the mammalian brain, forming pentamers from combinations of α-, β-, γ-, δ-, ε-, ρ-, θ- and π-subunits¹. αβ, α4βδ, α6βδ and α5βγ receptors favour extrasynaptic localization, and mediate an essential persistent (tonic) inhibitory conductance in many regions of the mammalian brain^1,2. Mutations of these receptors in humans are linked to epilepsy and insomnia^3,4. Altered extrasynaptic receptor function is implicated in insomnia, stroke and Angelman and Fragile X syndromes^1,5, and drugs targeting these receptors are used to treat postpartum depression⁶. Tonic GABAergic responses are moderated to avoid excessive suppression of neuronal communication, and can exhibit high sensitivity to Zn²⁺ blockade, in contrast to synapse-preferring α1βγ, α2βγ and α3βγ receptor responses^5,7-12. Here, to resolve these distinctive features, we determined structures of the predominantly extrasynaptic αβ GABA_A receptor class. An inhibited state bound by both the lethal paralysing agent α-cobratoxin¹³ and Zn²⁺ was used in comparisons with GABA-Zn²⁺ and GABA-bound structures. Zn²⁺ nullifies the GABA response by non-competitively plugging the extracellular end of the pore to block chloride conductance. In the absence of Zn²⁺, the GABA signalling response initially follows the canonical route until it reaches the pore. In contrast to synaptic GABA_A receptors, expansion of the midway pore activation gate is limited and it remains closed, reflecting the intrinsic low efficacy that characterizes the extrasynaptic receptor. Overall, this study explains distinct traits adopted by αβ receptors that adapt them to a role in tonic signalling.

Fig. 1. α-Cobratoxin binding site on α1β3 GABA_A receptors.

a_, α-CBTx–Zn²⁺-bound α1β3 receptor cryo-EM map, showing top (top) and side (bottom) views. α-CBTx is bound to the β–α neurotransmitter pocket interface. Glycans (orange) are not resolved inside the vestibule (top). Mb25 is shown in lime green; nanodisc and ‘hanging’ β3-subunit thermostabilized apocytochrome b562RIL (BRIL) densities are in grey. b, Atomic model of α-CBTx (green) bound to the GABA_A receptor with finger II positioned at the β3 (blue)–α1 (red) interface. c_, Close up of the binding mode in b showing residue positions and interactions (β3 loop C residues in blue are Val199, Phe200, Ala201, Thr202 and Tyr205). d, Overlays of the GABA-bound model (white) and α-CBTx-bound model (β3 loop C in blue, α1 with Arg67 in pink and red, and α-CBTx finger II in green), showing that finger II does not directly overlap with the GABA binding pose but displaces loop C and Arg67 away (black arrows) so that they no longer support GABA binding. Dashed lines represent putative hydrogen-bond interactions.

Fig. 2. The Zn²⁺ binding site.

a, Bar chart showing inhibition of maximal (1 mM) and EC₂₀ (1 μM) GABA whole-cell current responses (I_GABA) by 66 nM free Zn²⁺ (controlled using the chelator tricine) for wild-type αβ (WT) and the α1β3 cryo-EM construct (α1β3_CryoEM) expressed in HEK 293 cells. Data are mean ± s.e.m. n = 7 for wild-type and cryo-EM constructs, from biologically independent patch-clamp experiments with individual cells. One-way analysis of variance (ANOVA) and Tukey multiple comparisons post hoc test showed no significant differences across groups, F(3, 24) = 0.6449; P = 0.5937. b, Corresponding current recordings for α1β3_CryoEM. c, Top view of C_α backbones of M2 pore-lining helices showing three 17′ β3 His267 (blue) residues coordinating Zn²⁺ across the pore (α1 17′ Ser272 residues in red). Cryo-EM map shown as white transparent. d, Side-on view of β3-subunit chain B and E M2 helices flanking the pore permeation pathway (blue dots) with narrowings (orange dots) for three closed ‘gates’ at the 17′ Zn²⁺ site, 9′ hydrophobic (activation) gate and −2′ intracellular (desensitization) gate to create a triple-gated closed pore.

Fig. 3. Response of the TMD to GABA binding.

a, Top-down views of α-CBTx–Zn²⁺ (grey) and GABA-bound (red and blue) atomic model overlays showing the M2 helices, M2–M3 linkers, α1 Pro278 and β3 Pro273. The GABA binding β3 B–E subunit linkers respond and switch to the ‘outward’ conformation (arrows). b, Cross-section at the 9′ Leu ring showing expanded C_α pentagonal perimeter for GABA (purple) compared with α-CBTx–Zn²⁺ (grey). c, Side-on view of the permeation pathway (blue and orange dots) between opposing β3-subunit chain B–E M2 helices, showing closed 9′ and −2′ hydrophobic gates. The asterisk indicates the kink in the permeation pathway around the 17′ residue, which varies depending on mobile His side chain positioning, so the 17′ radius of 2.1 Å is indicative only. d, Pore radius along the permeation pathway.

Fig. 4. Mode of activation by GABA.

Top-down views of ECDs (top row) and cross sections of TMDs (helices shown as black circles) at the level of the 9′ Leu gate (bottom row) for αβ and αβγ receptors. Pore leucines are represented by black ‘fronds’ projecting from the innermost α-helix, M2. In response to GABA, the two binding β-subunits are the principal responders, with their ECDs twisting similarly anticlockwise (black arrows) for both αβ receptors and αβγ receptors. The downstream reaction of the TMD is limited in the αβ receptor and the 9′ gate remains mostly closed (red and orange circles), whereas for αβγ receptors, the TMD response is greater and the 9′ gate opens (green).

Extended Data Fig. 1. Local resolution maps, overall plotted resolutions, and global map-model agreements.

For the three structures, α-CBTx/Zn²⁺, GABA/Zn²⁺, and GABA-bound, a map on the left is coloured by local resolution (see methods). Maps of Fourier shell correlation (FSC) (upper right panels) and map-model FSC (lower right panels) plots are also shown. Relevant statistics for these maps are presented in Extended Data Table 1.

Extended Data Fig. 2. GABA responses and histamine potentiation.

a, GABA concentration response curves for αβ_WT (white symbols) and αβ_CryoEM (black symbols) in the absence (circles) or presence (triangles) of 3 mM histamine (HSM). Data was obtained in whole cell patch clamp experiments performed on transiently transfected HEK 293 cells. b, Histamine concentration response curves for potentiating the 300 µM GABA response for αβ_WT and αβ_CryoEM. For a, and b, points represent mean ± s.e.m. Curves generated are n = 4 and n = 5 for WT and EM constructs respectively. For a, One-way ANOVA showed no statistical difference across the 4 pEC₅₀ values, and for b, Two-sided unpaired t-test showed values were not statistically different, P = 0.51. c, Representative whole-cell patch clamp current responses to GABA and GABA + 3 mM histamine applications. Note that histamine was pre-applied before co-applying with GABA (blue lines). d–e, Bar charts showing average maximum (1 mM) GABA current response levels and rates of desensitisation, respectively, for αβ_WT and αβ_CryoEM. Individual values are shown as circles, and bars are means ± s.e.m. EM I_max n = 14, WT I_max n = 15, EM desens n = 11, WT desens n = 10. Two-sided unpaired t-test showed values were not statistically different for either property, P = 0.64 and 0.08, respectively. Each n = 1 value of an pEC₅₀, I_max and desensitisation value were from biologically independent patch-clamp experiments from individual cells.

Extended Data Fig. 3. Nb25 potentiation at α1β3 receptors.

a, Atomic models show no obvious distinctions for the β-β interface and the Nb25 binding pose of α1β3 receptors in the inhibited α-CBTx/Zn²⁺-bound conformation (3.0 Å, darker shades) versus the GABA-bound conformation (3.0 Å, lighter shades), consistent with any functional impacts exerted by Nb25 being subtle. Upper insets are viewing aids to highlight the region of the protein complex being viewed. Nb in green, β-subunits in blue, α-subunits in red. CDR3 is complementarity determinant loop 3 of Nb25. b, Representative currents of whole cell patch clamp responses to GABA and GABA + 10 μM Nb25 applications for α1β3_CryoEM versus α1β3γ2 wild-type. Note that Nb25 was pre-applied before co-applying with GABA (red lines). c, Bar chart showing average potentiation of EC₁₅ GABA current responses for αβ_CryoEM versus α1β3γ2 wild-type by Nb25, revealing a weak selective potentiation of α1β3 receptors due to the β-β interface, which is absent from α1β3γ2 receptors. Bars are means ± s.e.m. n = 3, each value being from biologically independent patch-clamp experiments from individual cells.

Extended Data Fig. 4. α-Cobratoxin binding mode and αβ receptor conformation.

a, Atomic model fit in the cryo-EM map density of the α1β3 receptor bound by α-CBTx/Zn²⁺ (3.0 Å) for the receptor β3 subunit loop-C (blue) and toxin finger II (green); yellow segment is Cys26-Cys30 side chain Cys-bridge. b, Alternative view of the toxin to show side chain density for Phe29, and binding residues Arg33 and Arg36. c–e, Atomic models showing common binding poses for toxins against pLGICs at inter-subunit interfaces (AChBP PDB 1YI5; nAChR PDB 6UWZ). f, Overlays showing closely matching arrangements of α-CBTx atomic models for GABRα1β3, AChBP and apo-α-CBTx (PDB 1ZFM). g, h, Atomic models comparing α-CBTx finger II binding mode to GABA_A receptor β3 subunit loop-C versus AChBP. i, Receptor α1-subunit Arg67 side chain can move away from toxin finger II Ile32 to accommodate toxin binding. j, Mechanism of reduced toxin sensitivity for α2-GABA_AR subunit, caused by Lys68 in the equivalent position to α1 Ser69, which can be explained by Lys68 sterically and electrostatically hindering Arg67 movement away from toxin Ile32 to reduce accommodation of the toxin. k, Overlays to compare loop-C outward motion imposed by α-CBTx on αβ receptor (arrow; GABA-bound in pale blue, toxin-bound blue), and bicuculline on αβγ receptor (GABA-bound in white PDB:6HUO, bicuculline-bound in dark grey PDB:6HUK; bicuculline not shown). l, Overlays of cross-section of top of pentameric ECD for α1β3 α-CBTx/Zn²⁺ (blue/red) versus an inhibited state of α1β3γ2 bound by the specific antagonist bicuculline (grey). The only distinguishable difference is the exaggerated outward translation of the β3-subunit loop-C for α-CBTx/Zn²⁺ (red circles) caused by toxin binding (toxin not shown). Inter-subunit interfaces are indicated by orange dashed lines. m, Same as l except overlay of αβ versus GABA-activated state of α1β3γ2 to reveal a greater divergence in conformation (RMSD increased from 1.0 Å in l to 1.5 Å in m), in particular for the agonist-responding β-subunits (chains B/E), indicated by red arrows (RMSD increases to 1.8 Å). n, Overlays of the β1-β2 loops (inner) and β6-β7 loops (Cys-loops; outer) at the base of the ECD, which oppose the TMD, for α1β3 α-CBTx/Zn²⁺ (blue/red) versus the inhibited state of α1β3γ2 bound by bicuculline (pale shades, yellow for γ2 loops), showing occupation of the same positions for all subunits, including for the α1β3 Chain C β-subunit fitting the α1β3γ2 Chain C γ2-subunit position.

Extended Data Fig. 5. Zn²⁺ inhibition.

a, Free* Zn²⁺ inhibition curves from whole cell patch-clamp experiments performed in recording buffer supplemented with 10 mM tricine for αβ_WT (white symbols) and αβ_CryoEM (black symbols) expressed in HEK293 cells. Points represent mean ± s.e.m. Curves generated are n = 8 and n = 7 respectively of WT and EM constructs, of biologically independent patch-clamp experiments from individual cells. Two-sided unpaired t-test showed pIC50 values were not statistically different, P = 0.35. b, similar Zn²⁺ inhibition curve for αβ_CryoEM, but in the absence of Zn²⁺-chelating tricine, showing that contaminating Zn²⁺ in buffers was not impacting sensitivity in any way, n = 6. c–e, Atomic model fits in cryo-EM map density of β3 (chain E) subunit TMDs for α-CBTx/Zn²⁺ (3.0 Å), GABA/Zn²⁺ (2.79 Å) and GABA (3.04 Å) respectively. Zn²⁺ density at 17′ His is indicated. Note the 17′ density is absent when Zn²⁺ is not bound. f–h, Atomic model fits in cryo-EM map density for top-down slices of the pentamer at the 17′ pore position for α-CBTx/Zn²⁺, GABA/Zn²⁺ and GABA-bound structures, respectively. h, Density for 17′ His residues is absent when Zn²⁺ is not bound to coordinate them, indicating these side chains are highly mobile. Pore expands nominally to 4.1 Å diameter (variable depending on flexible His arrangement). *free Zn²⁺ concentration controlled and determined using the chelator, tricine (see “Methods”).

Extended Data Fig. 6. Impact of GABA binding on ECD conformation.

a, Upper panel: Overlay cross-sections of the top of pentameric ECDs of α-CBTx/Zn²⁺ (grey) versus GABA (α-red/β-blue)-bound atomic models. Greatest divergence is observed for the GABA binding β3-subunits (chains B/E), which have tilted/rotated in response to GABA binding, red arrows, and is reflected by RMSD being higher for these subunits, 1.8 Å, relative to the whole ECD, 1.3 Å. a, Lower panel: Overlay of the β1-β2 loops (inner) and β6-β7 loops (Cys-loops; outer) at the base of the ECD which oppose the TMD (not shown). This shows the resultant translation for the GABA binding β3-subunits (chains B/E; red arrows), caused by the motion in the upper ECD (upper panel). b, same as a, but for α-CBTx/Zn²⁺ (grey) versus GABA/Zn²⁺ (α-red/β-blue). Differences are the same because GABA induces the same ECD motions even with Zn²⁺ bound in the pore. c, same as a, but for GABA (α-red/β-blue) versus GABA/Zn²⁺ (grey). As these ECDs have undertaken the same motions in response to binding GABA, RMSDs are lower and β-subunit RMSDs do not increase relative to whole ECD. d, same as a, but for α1β3γ2 bicuculline-bound (grey; PDB 6HUK) versus α1β3γ2 GABA/Alprazolam-bound (α-red/β-blue/γ-gold; PDB 6HUO). The impact of GABA/Alprazolam binding versus the antagonist is the same as observed for the αβ receptor GABA binding versus antagonist (shown in a). NOTE: ligands are not shown.

Extended Data Fig. 7. TMD M2-3 loop conformations.

a, Atomic model fits in the cryo-EM map density for the M2-M3 loops of β3 chains B and E for α-CBTx/Zn²⁺ (3.0 Å, blue maps) and GABA-bound (3.04 Å, grey maps) respectively. Viewed looking down on to the M2-M3 loop reveals the switch to the ‘outward’ conformation in response to the ECD binding GABA, as highlighted by Pro276 repositioning to the other side of the dashed line. b, Top-down view of α1β3γ2 bicuculline-bound (dark grey; PDB 6HUK) versus α1β3γ2 GABA/Alprazolam-bound (α-pink/β-pale blue/γ-gold; PDB 6HUO) showing the M2-M3 loop positions. In response to GABA binding the β-subunit chain B/E M2-M3 loops switch to the ‘outward’ conformation, indicated by red arrows that highlight the motion of Pro273. The γ2-subunit is in the outward conformation in both states.

Extended Data Fig. 8. Pore arrangement.

a, Side-on views of subunit transmembrane helical bundle C_α-polypeptide for Chains A-E showing M2 helix tilt axis. α1β3 GABA model is coloured red (α-subunit) or blue (β-subunit), and the M2 helix axis is shown as a dim grey bar. α1β3 α-CBTx/Zn²⁺ model is white, M2 helix axis green. Pore axis is to right of each bundle, vertical black bar. M2 helix tilt and/or translation away from pore axis increases for each subunit when GABA is bound (angle values shown; translations not measured but visible by eye). b, Table showing M2 helix tilt angles for αβ and αβγ receptors in antagonist (black text) and agonist (GABA) bound (red text) conformations. αβγ M2 helix tilt angles shown as increase (+) or decrease (−) relative to the equivalent αβ M2 helix. The biggest increase is for the γ2 subunits (values bold, underlined). c, Electron density map slices of the pore conformation at the 9′ Leu gate (pore diameters given inside pore circles) for α1β3 α-CBTx/Zn²⁺, GABA/Zn²⁺ and GABA-bound structures. For comparison the cryo-EM map of EMD-0282 used to build 6HUO PDB of GABA+Alprazolam bound structure is shown. d, Cross-section at 9′ Leu hydrophobic gate showing C_α pentagonal perimeters for GABA-bound αβ receptor versus GABA+Alprazolam-bound α1β3γ2 (left panel) or GABA-bound α1β2γ2 (right panel). e, Side by side comparison of the αβ receptor β3 subunit Chain C M2 helix tilts versus equivalent αβγ receptor γ2 M2 helix tilts, which are more reclined (for PDB codes see table in, b).

Extended Data Fig. 9. Single channel current analysis for GABA_AR heteromers.

a, epochs of GABA single channel currents recorded from outside-out patches of HEK293 cells expressing α1β3_WT and α1β3γ2_WT receptors, activated by 30 and 100 μM GABA respectively (~EC₉₅ for each receptor isoform) at low (upper trace) and higher time resolution (lower traces). C – closed and O – open state; closed state marked by dashed line; downward deflections are transitions to open state. b, Examples of open and closed state dwell time distributions for single cells expressing α1β3_WT or α1β2γ2_WT receptors. Single exponential component fits (green lines) and summed fits from a mixture of exponentials (red lines) are shown. In the example shown for open times, a single exponential fit was sufficient to account for the α1β3_WT open state distribution, whilst for α1β3γ2_WT a mixture of two exponentials was required. Mean exponential τ values (with SEM, and percentage area) determined from analysing multiple patches are: τ1 = 0.65 ± 0.15 ms (A1 = 83 ± 6%, n = 6), τ2 = 4.3 ± 1.1 ms (A2 = 17 ± 6%, n = 4 – two cells did not show long open times); for α1β3γ2_WT: τ1 = 0.78 ± 0.07 ms (A1 = 39 ± 3%, n = 6), τ2 = 4.8 ± 0.7 ms (A2 = 61 ± 3%, n = 6). For closed state dwell time distributions both receptor isoforms required a mixture of four exponentials of similar magnitudes, however α1β3 favoured the longer duration closed states, whereas α1β3γ2 favoured the shortest closed states which normally appear within bursts of openings. Mean τ values (and SEM, including percentage areas) from multiple patches are: α1β3_WT; : τ1 = 0.17 ± 0.05 ms (A1 = 36 ± 1%, n = 3), τ2 = 2.8 ± 0.3 ms (A2 = 62 ± 10%, n = 6), τ3 = 21 ± 5 ms (A3 = 22 ± 6%, n = 4), τ4 = 52 ± 3 ms (A4 = 20 ± 2%, n = 3)); α1β3γ2_WT: τ1 = 0.50 ± 0.04 ms (A1 = 68 ± 3%, n = 6), τ2 = 3.0 ± 0.3 ms (A2 = 21 ± 2%, n = 6), τ3 = 24 ± 4 ms (A3 = 9 ± 2%, n = 6), τ4 = 165 ± 25 ms (A4 = 2 ± 0.5%, n = 3). c, Bar graph showing percentage distribution between short versus long open state dwell times for α1β3 and α1β3γ2. Points represent mean ± s.e.m. n = 6, except for α1β3 long openings n = 4 (no long openings observed for two of the cells). Two-sided unpaired t-test comparisons of open dwell times t(8) = 7.53, p < 0.0001, and shut dwell times t(10) = 7.01, p < 0.0001. d, Bar graph showing the open probability (P_o; the average fraction of time spent in the open state), measured as the total open time divided by the total length of the recording. Recordings were taken from patches showing limited or no channel stacking (see Methods). Individual values are shown as circles/squares with associated error bars (mean ± s.e.m.), n = 6, two-sided unpaired t-test comparison, t(10) = 4.43, P = 0.0013. ** signifies as statistically different (P < 0.01)for bars linked by black lines. Each n = 1 value of an open time, shut time or P_o were from biologically independent patch-clamp experiments from individual cells.

Extended Data Fig. 10. Probability of activation for α1β3 and α1β3γ2 receptors.

a, whole-cell patch clamp recordings from HEK293 cells expressing either α1β3 EM, α1β3 WT, or α1β3γ2 WT, showing responses to increasing concentrations of GABA and saturating GABA + 1 mM pentobarbitone in order to measure the probability of activation (P_A). Dashed lines indicate the baseline current (grey), the maximum activation by saturating GABA alone (black) and activation by saturating GABA + 1 mM pentobarbitone (red). Accompanying concentration response curve plots are provided (mean ± s.e.m, α1β3EM n = 8, α1β3WT n = 5, α1β3γ2 WT n = 6). b, Bar chart showing similar P_A values for α1β3 EM and α1β3 WT, which were lower than for α1β3γ2 WT. Individual values are shown as circles/squares and bars are means ± s.e.m. (α1β3EM n = 8, α1β3WT n = 5, α1β3γ2 WT n = 6). One-way ANOVA comparing abWT vs abCryoEM vs abgWT: F(2,16) = 18.78; P < 0.0001. Post-hoc Tukey. Each n = 1 value of a P_A were from biologically independent patch-clamp experiments from individual cells.