|
Fig. 1. The wild-type α1β2γ2 pentameric GABAAR concatemer.a Schematic top view of the concatemer. The two β2/α1 ECD interfaces (SU1/SU2 and SU3/SU4) harbor the two GABA-binding sites, while the α1/γ2 ECD interface (SU4/SU5) contains the benzodiazepine-binding site. b Representative TEVC recording of a Xenopus laevis oocyte expressing the wild-type concatemer, CWT. c Depiction of the experimental values used to quantify desensitization: τfast and τslow are the time constants of fast and slow desensitization components, respectively; %Afast is the relative amplitude of the fast component; %Ires is the relative residual current after 1 min of 10 mM GABA application. Of note, the weighted desensitization time constant can be defined as τw = %Afast * τfast + (1−%Afast) * τslow. d Cryo-EM structure of the α1β3γ2 GABAAR (pdb 6I5317), as seen from the extracellular space. The β2 and β3 GABAA subunits are highly homologous, and both display an asparagine residue at the M3-5′ position. Note the central pore, lined by the M2 helices of the five subunits, forming the transmembrane channel. e Sequence alignment of the M3 segment of various pLGIC subunits. All sequences are the mouse orthologs, except GLIC (Gloeobacter violaceus), as well as the α4 and β2 nAChR subunits (human). The M3-5′ residues, mutated in the present study, are highlighted (gray box; bold characters for GABAA subunits). f Enlarged view of the α1β3γ2 GABAAR structure highlighting the location of the M3-5′ residue at the M2/M3 transmembrane interface as seen from the side of the channel, facing the M1–M2 linker of the adjacent subunit.
|
|
Fig. 2. Desensitization kinetics of α1β2γ2 concatemers harboring combinations of M3-5′ valine mutations.a Schematic top views of the C3 (left), C45 (middle) and C12345 (right) concatemers. b–d Representative TEVC recording of Xenopus laevis oocytes expressing the indicated concatemers. Note the change in timescale for recordings in panel (d). e Plot indicating the mean values for fast (red squares), slow (blue squares), and weighted (dark gray diamonds) desensitization time constants for the indicated concatemers. Error bars are standard deviations. See Supplementary Fig. 1 for individual data points and Supplementary Table 1 for numerical values, the number of cells, and number of independent series of experiments.
|
|
Fig. 3. General scheme for the simulation of desensitization: an asymmetric non-concerted model.The first part in the kinetic scheme is the binding of the agonist A to the resting state R, which favors the opening of the channel (AO state) with a gating efficacy E = β/α. Of note, unliganded openings do exist but are not taken into account for our kinetic modeling as they barely contribute to the electrophysiological response (see main text). We also only include one binding event, even though α1β2γ2 GABAARs contain two binding sites whose occupation is required for substantial activation. Upon channel opening, the receptor can then transit from a fully activated AO state to states where only one subunit enters its desensitized conformation (AD3, AD4, and AD5). From these states, a second subunit can also desensitize, before the final step leading to the state in which all subunits are desensitized.
|
|
Fig. 4. Model I: only the fully open state is conducting, and subunits move independently during desensitization.a We assume in this model that a single desensitized subunit is enough to shut the pore of the channel, leading to functional desensitization. Moreover, subunits SU3, SU4, and SU5 can undergo a desensitization rearrangement independent of the other subunits. Thus, desensitization rates (δ3+ for SU3, δ+ for SU4 and SU5) and recovery rates (δ3− for SU3, δ− for SU4 and SU5) do not depend on the conformation of the neighboring subunits. b Effect of M3-5′ valine mutations in Model I. Mutations are hypothesized to specifically increase the desensitization rates of the mutated subunits, without altering any other parameter. c–e Representative currents for CWT (panel c), C4 (panel d) and C45 (panel e), in black, are compared to the outcome of two distinct simulations. In simulation a (red), the mutation-induced increase in the desensitization rates of SU4 and SU5 is adjusted so that the simulation of single mutants C4 and C5 broadly fits the experimental data, as seen in panel (d). In simulation b (blue), the mutation-induced increase in the desensitization rates of SU4 and SU5 is adjusted so that the simulation of the double mutant C45 accounts for the experimental data, as seen in panel (e). f Bar graph summarizing the experimental data vs the predicted effects of SU4 and/or SU5 mutations on the kinetics of the fast desensitization component in simulations a and b. Experimental data are shown as means (bar graphs) and standard deviations (error bars), with individual data points indicated as circles. For panels c–f note that parameters from simulation a fail at describing the data for the double mutant C45, while parameters from simulation b largely overestimate the effect of single mutants. See Supplementary Table 1 for numerical experimental values, the number of cells and number of independent series of experiments; and Supplementary Table 3 for the numerical values of parameters.
|
|
Fig. 5. Model II: Two desensitized subunits are required to occlude the pore.a Model II builds upon Model I by adding one key hypothesis: receptors with only one subunit in its desensitized conformation are still conducting, and desensitization occurs when at least two subunits are desensitized. Thus, states AD3, AD4, and AD5 are open states from a functional point of view. b In Model II, mutation of a subunit can affect both its desensitization and recovery, as shown here with an example in which both SU4 and SU5 are mutated (construct C45): c4+ and c5+ reflect the increase in desensitization rates, c4− and c5− reflecting the increase in recovery rates. c Simulated currents for CWT, C4 and C45. d Representative currents for CWT and C3 in black, are compared to their simulation counterparts in red. e–g Bar graphs summarizing the experimental data (in black) vs the simulations (in red) for the indicated concatemers on the kinetics (panel e) and the amplitude (panel f) of the fast desensitization component as well as the residual current after a 1 min long application of 10 mM GABA (panel g). Experimental data are shown as means (bar graphs) and standard deviations (error bars), with individual data points indicated as circles. Note that the results for the C5 construct are not displayed, since the experimental data are almost identical to that of C4 (see Fig. 2) and since the simulations for C4 and C5 are identical (see Supplementary Table 3). See Supplementary Fig. 5 for all simulation results; Supplementary Table 1 for numerical experimental values, the number of cells and number of independent series of experiments; and Supplementary Table 3 for the numerical values of parameters.
|
|
Fig. 6. Model III introduces inter-subunit coupling during desensitization.a For the wild-type receptors, Model III builds upon Model II by adding some coupling between adjacent subunits during desensitization. On the one hand, desensitization of SU3 accelerates the desensitization of SU4 by a factor γ, and reciprocally. On the other hand, desensitization of SU4 slows the recovery of SU5 by a factor ε, and reciprocally. b For mutated concatemers, Model III incorporates the additional hypothesis that the mutation of SU3 also affects the desensitization of SU4 by increasing both its desensitization and recovery rates, by ratios c34+ and c34−, respectively.
|
|
Fig. 7. Model III simulations are broadly consistent with experimental data.a–h Representative currents for the indicated constructs, in black, are overlaid with their simulation counterparts in red. Note the changes in timescales. i–l Bar graphs summarizing the experimental data (in black) vs the simulations (in red) for the indicated concatemers on the kinetics of the fast (panel i) and slow (panel j) desensitization components, the relative amplitude of the fast component (panel k) and the residual current after a 1 min long application of 10 mM GABA (panel l). Experimental data are shown as means (bar graphs) and standard deviations (error bars), with individual data points indicated as circles. See Supplementary Table 1 for numerical experimental values, the number of cells and number of independent series of experiments; and Supplementary Table 3 for the numerical values of parameters.
|
|
Fig. 8. States occupancies predictions and structural depiction of Model III.a The overall population of wild-type receptors in an active conformation is compared to the relative occupancies of the various desensitized states. As depicted by the red box, the early phase of desensitization is carried by the AD45 state. On longer timescales (blue box), slow desensitization is largely embodied by the entry in the AD345 state. The analysis of states occupancies was performed with QuB simulations. b In this simplified depiction of model III, we extracted the kinetically favored pathway for the desensitization of wild-type α1β2γ2 GABAARs. Upon agonist binding, the receptor is transiently stabilized in a fully open pseudo-symmetrical conformation. The two first subunits to rearrange during desensitization are the α1 and the γ2 subunits involved in the binding of benzodiazepines, namely SU4 and SU5 in our concatemers. While one desensitized subunit is not enough to occlude the pore, fast desensitization corresponds to the rearrangement of both SU4 and SU5 subunits, which are coupled. Slow desensitization is then driven by the slower rearrangement of the SU3 subunit, i.e. the β2 subunit opposite to the γ2 subunit.
|