GABA (Sigma, Saint-Quentin-Fallavier, France) was prepared in the standard oocyte solution (SOS, see composition hereafter) and FPN (Sigma, Saint-Quentin-Fallavier, France; Figure 1) was diluted in DMSO and then its concentration range was prepared in SOS medium. Picrotoxin (PTX—Sigma, Saint-Quentin-Fallavier, France) was firstly diluted in DMSO to a final concentration of 0.1%. 100 μM PTX was then prepared in SOS medium. Control experiments using DMSO (0.5%) were performed. Etifoxine (EFX hydrochloride, Biocodex, Gentilly, France) was dissolved in DMSO at a final concentration of 0.1%.

All animal procedures were carried out in accordance with the European Community council directive 2010/63/EU for the care and use of laboratory animals and were approved by our local ethical committee (N°A49007002 for rats) in addition to the French Ministry of Agriculture (authorization APAFIS#19433-2019022511329240 and B49071 for Xenopus). The NC3R’s ARRIVE guidelines were followed in the conduct and reporting of all experiments using animals. Four rats were killed to prepare cerebellum membranes (in four independent experiments) to attenuate the effects of individual polymorphism in GABA_AR microtransplantation experiments. For TEVC, the data were collected from oocytes collected from 12 distinct Xenopus females. In our animal facility, oocytes were collected twice a week, leading to the use of two different females. Each animal was reused after 9 weeks to allow a full recovery (healing and for animal welfare).

Rats and Xenopus laevis females were used for cerebellum and oocytes preparations, respectively. Wistar rats of 200–250 g were obtained from the Animal Facility Centre of the Hospital/University of Angers. Rats were maintained with ad libitum access to standard diet and tap water and accommodated in individual cages under controlled conditions of room temperature and illumination (12 h light/dark cycle).

Xenopus oocytes were prepared as previously described (Mattei et al., 2019). Briefly, adult female Xenopus laevis were purchased from Centre de Ressources Biologiques Xénopes (Rennes, France) and were bred in the laboratory according to the recommendations of the Guide for the Care and Use of Laboratory Animals of the European Community. Oocytes were harvested from Xenopus laevis frogs under 0.15% tricaine anesthesia. All animals recovered within 2–3 h. Each female is operated every 3 months, not less and no more than 5 times.

The cDNAs encoding the α6 and δ subunits used in this work were cloned in mouse as previously described (Mattei et al., 2019). pGW1 (= pRK5) plasmids containing cDNAs encoding mouse β3 and γ2S subunits were provided by Steven J. Moss (Department of Neuroscience, Tufts University, Boston, United States).

We adapted the method of membrane transplantation described previously (Miledi et al., 2004). Adult Wistar rats (male and female) were euthanized with CO₂ (5%) for 6 min. The brain and cerebellum were removed and stored at −80°C. Tissues were ground on ice with 200 mM glycine buffer (sucrose 300 mM—glycine 200 mM—NaCl 150 mM—EDTA 50 mM—EGTA 50 mM, supplemented with protease inhibitors). A portion of the homogenate was aliquoted, frozen in liquid nitrogen and stored at −80°C for use in the protein assay. After homogenization, the samples were centrifuged for 15 min at 9,500 g and 4°C (12132-H angular rotor, SIGMA2-16K centrifuge). The supernatant was recovered—a part was aliquoted for protein assay—and centrifuged for 2 h at 100,000 g and 4°C (MLA-130 angular rotor, Beckman Coulter OPTIMA MAX-XP ultracentrifuge). The new supernatant was aliquoted and stored at −80°C. The pellet, which contains membranes, was suspended with 10 μl glycine buffer 5 mM, and stored at −80°C. The membrane preparations can be used for protein assay or for microinjection. These living membranes, which carry the native α6-containing receptors, incorporate within the oocyte membrane, and the oocytes can be stimulated with GABA (Palma et al., 2005).

We used the same method as described previously (Crespin et al., 2016). Briefly, the protein content was determined with a BCA™ Protein Assay Kit (Pierce^®). Several dilutions were performed with samples to ensure detection in the range. Each dilution was injected twice in a 96-well plate. After incubation at 37°C, the plate was read at 570 nm using a plate reader (Multiscan Ascent Thermoscientific^®). The protein content was calculated from a standard range of BSA (0–80 μM). The preparation of cerebellum rat membranes yielded a membrane suspension of 24 ng/nl protein concentration.

Oocytes were harvested as previously described (Mattei et al., 2019). Briefly, oocytes were collected from female anesthetized X. laevis with tricaïn 0.15 M for 15 min and washed first in a solution of SOS (NaCl 100 mM, KCl 2 mM, MgCl₂ 1 mM, CaCl₂ 1.8 mM, HEPES 5 mM—pH 7,4), and then in a Ca²⁺-free SOS solution. They were incubated under gentle stirring with collagenase (2 mg/ml) and trypsin inhibitor (0.8 mg/ml) for 5–10 min and manually defolliculated. Then, oocytes were stored in SOS medium supplemented with antibiotics (gentamicin 0.04 mg/ml, penicillin/streptomycine/pyruvate 0.22 mg/ml) at 4°C.

Defolliculated stage V-VI oocytes were microinjected with rat cerebellum membranes or cDNA using a nano-automatic injector (Nanoject II, Drummond Scientific Company, Pennsylvania, United States). For optimal GABA-evoked current recordings (>10 nA), 55.2 nl of cerebellum membrane preparation corresponding to 1325 ng of proteins were injected by oocyte. The α6β3 (1:1), α6β3γ2S (1:1:5), α6β3δ (1:1:5) combinations were prepared at a concentration of 50 ng/μl for α and β, 250 ng/μl for γ2S, and δ to obtain. For each combination (Figure 2A), an amount of 450 pg cDNA was injected into the cell nucleus. Oocytes were incubated at 18°C and tested 24–48 h after injection. FPN was pre-applied for 45 s before any GABA application. To check ternary α6β3γ2S and α6β3δ GABA_ARs, and binary α6β3 GABA_ARs, controls were performed with GABA (5.10^–7 M) before and after addition of Zn²⁺ (10 μM), which does not affect ternary receptors, and inhibits GABA-induced current through binary receptors (Supplementary Figure 1). Control experiments were performed with the antagonist PTX (100 μM) (Supplementary Figure 1).

To monitor the activity of functional GABA_AR responses in Xenopus oocytes, we used a standard two microelectrode voltage-clamp technique as described previously (Mattei et al., 2019). Glass microelectrodes were made with a DMZ Zeitz puller and exhibited a resistance of 0.5–1.5 MΩ. They were filled with an intracellular solution containing 1 M KCl/2 M K acetate. Each oocyte was continuously bathed in a recording chamber with the SOS solution. The resting membrane potential of injected oocytes was about −30 mV and, in the voltage-clamp configuration, the holding current (Ih) was about 50 nA for a holding potential of −60 mV. We only chose oocytes with a stable resting potential. The resting membrane potential was measured at the end of each experiment. FPN was diluted in the perfusion solution and then directly applied in the oocyte-containing chamber. 48 h after cDNA injection, oocytes were tested using a two-electrode voltage-clamp amplifier (TEV-200A, Dagan Corporation, Minneapolis, United States), at a holding potential of −60 mV. Data were acquired with a pCLAMP system (Digidata 1440 and pCLAMP 10.0 software from Axon Instruments). Experiments were performed at room temperature. Control experiments using DMSO (0.5%) were performed. No change in holding currents were observed, when non-injected or injected oocytes were perfused with SOS solution with 0.5% DMSO.

The models for the three receptors were prepared depending on the closest available template. The β3 homopentamer was directly based on the PDB structure 4COF. For α6β3 and α6β3γ2 heteropentamers, an additional homology modeling step was required and was based on the structure of the α1β3γ2 heteropentamer (6HUG). The sequences of the human α6, β3, and γ2 GABA_AR subunits were aligned with those of the template using T-Coffee software (Notredame et al., 2000). The model was then prepared by homology modeling using Modeler version 9.19 software (Sali and Blundell, 1993) with default settings. One hundred models were prepared, and the best model, according to the Discrete Optimized Protein Energy function (DOPE), was selected. For the three models, side chains were improved with Scwrl4 (Krivov et al., 2009). The models were then evaluated with Molprobity and improvements on side chains were considered (Williams et al., 2018).

The docking has been performed with AutoDock Vina (Trott and Olson, 2010). The ligands and proteins were prepared with prepare_ligand4.py and prepare_receptor4.py scripts, respectively. The docking was restricted using a docking box of 15 A side, in the upper and lower binding sites identified previously (Charon et al., 2011). Figures were prepared with PyMOL (Schrödinger and DeLano, 2020).

In electrophysiology, the amplitude of each current response was expressed as a% of the response to GABA EC₅₀. The EC₅₀ and the Hill coefficient (nH) were determined by non-linear regression (Figure 2B) using the Langmuir equation with variable slope. We excluded data (i) in case of potential drift (>0.6 mV) after pulling out the electrodes from the oocytes and (ii) when current amplitudes were < 10 nA or > 2 μA. When required, current densities were calculated using membrane capacitance. GraphPad Prism 7.02 (GraphPad Software, San Diego, United States) was used for all graphs and statistical analyses. Normality of data distribution was validated using Shapiro-Wilk test to choose a parametric or a non-parametric test. Statistical significance tests between groups were performed using variance analysis (one-way ANOVA) followed by Tukey’s post hoc test for comparison of all groups or non-parametric Kruskal-Wallis procedure, followed by the post hoc Dunn’s test when appropriate. All data are presented as mean ± SEM of individual oocytes from at least two separate female Xenopus. Differences with p < 0.05 were considered significant (* for p < 0.05, ^** for p < 0.01, ^*** for p < 0.001, ^**** for p < 0.0001).

To assess the effect of FPN on native receptors embedded in their biological membranes, we used rat cerebella. In this context, GABA_ARs are fully functional as they can be activated by increasing concentrations of GABA (Figure 3A). We challenged microinjected oocytes with GABA (0.1 mM). FPN dose-dependently inhibited GABA-induced currents (Figure 3B). The inhibition level was 33.0 ± 3.9, 53.8 ± 7.5, and 58.8 ± 7.4% when GABA was applied with 1, 10 and 100 μM FPN, respectively (Figure 3C). Because the cerebellum mainly contains α6-, but also α1-, GABA_ARs (Wisden et al., 1992), we chose to express the ternary α6β3γ2S and α6β3δ receptors in Xenopus oocytes for investigating FPN putative antagonist activity and to decipher the role of the third subunit on its inhibitory effects.

To see if the FPN mode of action relies on the subunit composition, we first generated GABA concentration-response curves for the different GABA_AR isoforms expressed in Xenopus oocytes, i.e., synaptic α6β3γ2S, extrasynaptic α6β3δ, and α6β3 (Figure 2A). For the synaptic α6β3γ2S, extrasynaptic α6β3δ and α6β3 GABA_ARs, we determined GABA EC₅₀ values (Figure 2B) for subsequently evaluating FPN effects on each of these receptors. Our GABA EC₅₀ values (see caption) are in accordance with data obtained by Mortensen et al. (2012), who transiently expressed various synaptic and extrasynaptic GABA_AR in HEK-293 cells and reported a very similar overall ranking of GABA sensitivity. We obtained an EC₅₀ of 0.75 μM for α6β3δ (Figure 2B), and Mortensen et al. (2012) reported EC₅₀ of 0.17 μM for this same subunit combination. The EC₅₀ was 3.85 μM for α6β3γ2S and 0.43 μM for α6β3.

To compare the inhibitory effects of FPN on extrasynaptic and synaptic receptors, we expressed ternary α6β3δ and α6β3γ2S combinations. FPN alone (10 μM) did not induce any current. We analyzed the effects of FPN (10 μM) with increasing concentrations of GABA (Figures 4A,C). For this purpose, we first applied 10^–9 to 10^–4 M GABA in the absence of FPN, which served as the control. This was followed by another set of experiments where the same concentrations of GABA were applied in the presence of 10 μM FPN. We measured the current density of each oocyte recorded. The concentration-response relationships for GABA in the absence and presence of FPN are shown in Figure 4. Our data show that the FPN effect is similar between the GABA_AR combination: for both receptors, the addition of FPN did not modify EC₅₀ values, whereas the maximal effect was significantly decreased (Figures 4B,D and Table 1). These observations agree with the fact that FPN behaves as a non-competitive antagonist (Li and Akk, 2008), which means that it targets an allosteric site on these two mammalian α6-containing receptors.

To highlight the role of the complementary subunit in FPN-induced inhibition of GABA currents, we decided to challenge ternary and binary receptors. Oocytes injected with α6β3δ extrasynaptic ternary receptors were first subjected to GABA EC₅₀ in the presence of increasing concentrations of FPN (Figure 5A). FPN (300 μM) produced a maximal inhibition of 46.0%, with an IC₅₀ of 20.7 μM, which denotes a partial antagonist effect on extrasynaptic receptors (Table 2). To make sure that the current observed was mostly due to the ternary receptors (α6β3δ) and not binary receptors (α6β3), Zn²⁺-containing SOS was applied to inhibit GABA-evoked currents elicited by binary GABA_ARs, without altering GABA-evoked currents from ternary GABARs (Draguhn et al., 1990). The addition of ZnCl₂ (10 μM) in a solution containing GABA 5.10^–7 M triggered currents with similar amplitude to those obtained with the application of GABA alone (Supplementary Figure 1). The currents generated by the binary receptors were thus negligible, suggesting that most receptors expressed in the oocyte membrane were ternary α6β3γ2S. Furthermore, co-application of PTX (100 μM) and GABA (5.10^–7 M) was performed to check if the observed currents were effectively GABA-driven. PTX induced an almost complete inhibition of current (Supplementary Figure 1).

We then challenged the effect of FPN on α6β3γ2S synaptic GABA_ARs using the calculated EC₅₀. FPN concentration-dependently antagonized α6β3γ2S (Figure 5A). Inhibition of the current was 94% with 300 μM FPN, and the calculated IC₅₀ was 20.2 μM. This result shows a discrepancy in the inhibitory effect of FPN, as a function of the complementary subunit: on the one hand, IC₅₀ values are similar between α6β3γ2S and α6β3δ; on the other hand, the inhibition is full for α6β3γ2S, and partial for α6β3δ. Again, to verify that the current recorded was due to the ternary α6β3γ2S receptors, we used Zn²⁺. The subsequent addition of ZnCl₂ in the medium did not modify the GABA-induced current, while PTX (100 μM) inhibited these currents (Supplementary Figure 1).

To highlight the role of the complementary γ/δ subunit in FPN mode of action, we performed the previous experiment again using the binary α6β3 receptor. Representative current traces are shown in Figure 5B. Unexpectedly, FPN did not fully antagonize the α6β3-driven current. The IC₅₀ is 2.4 μM—10-fold lower than ternary receptors—and the maximal inhibition is 46.8%, close to what is observed with α6β3δ receptors. Our data suggest that FPN differentially antagonizes GABA_ARs, as a function of their subunit composition, and the complementary subunit appears to play a crucial role in this inhibition. Altogether, the stoichiometry of these receptors drives their pharmacological properties toward the phenylpyrazole insecticide FPN, since the current inhibition was almost complete for synaptic receptors and partial for extrasynaptic and binary receptors.

The β3 subunit of mammalian GABA_ARs is of particular importance: although it has never been identified in vivo, it was the first GABA_AR structure solved at high-resolution (Miller and Aricescu, 2014). This subunit has been shown to influence the binding of FPN to GABA_ARs (Ratra et al., 2001). The β3 homopentamer stands in an open conformation and proved to be insensitive to GABA or muscimol but can be positively modulated by pentobarbital (Wooltorton et al., 1997). The β3 leak currents could be inhibited by EFX (Supplementary Figure 1C), as it has been shown previously (Hamon et al., 2003). Then, this homopentamer was challenged to increasing concentrations of FPN which elicits dose-dependent currents (Figure 6). We can notice that these currents display small amplitudes (∼15 nA), compared to GABA-induced currents elicited by binary or ternary GABA_ARs (∼200–1,000 nA). In addition, FPN did not induce any current in oocytes expressing binary or ternary GABA_ARs, indicating that β3 homopentamers did not influence the GABA-mediated currents of Figures 2, 4.

Hence, FPN behaved as a pseudo-agonist of β3 GABA_ARs, after we have shown its role as an antagonist on binary and ternary receptors. This again highlights the versatile pharmacological properties of FPN on GABA_ARs, depending on their subunit composition, from non-competitive antagonist to positive allosteric modulator.

We generated homology models of α6β3 and α6β3γ2S GABA_AR to predict how FPN binds to their receptor site (Figure 7). The docking was guided by the prior knowledge that FPN is an open channel blocker and has two putative binding sites in the ion channel, lined by the M2 transmembrane segments (Perret et al., 1999; Charon et al., 2011). Indeed, our docking finds two binding modes, located nearby Val257 and Ser272 (Figures 7C,D). We reasoned that these two putative binding sites may explain the different pharmacological properties observed at Figure 5, because FPN appears in contact with residues that differ depending on the subunit (Figure 7). Both sites could be accessible to FPN and occupied simultaneously. The model presented is speculative and based on a possible docking. Consequently, our hypothesis deserves to be verified with mutagenesis experiments.