Sustained treatment with an α5 GABA A receptor negative allosteric modulator delays excitatory circuit development while maintaining GABAergic neurotransmission

Abstract

α5 subunit GABA type A receptor (GABA_AR) preferring negative allosteric modulators (NAMs) are cognitive enhancers with antidepressant-like effects. α5-NAM success in treating mouse models of neurodevelopmental disorders with excessive inhibition have led to Phase 2 clinical trials for Down syndrome. Despite in vivo efficacy, no study has examined the effects of continued α5-NAM treatment on inhibitory and excitatory synapse plasticity to identify mechanisms of action. Here we used L-655,708, an imidazobenzodiazepine that acts as a highly selective but weak α5-NAM, to investigate the impact of sustained treatment on hippocampal neuron synapse and dendrite development. We show that 2-day pharmacological reduction of α5-GABA_AR signaling from DIV12–14, when GABA_ARs contribute to depolarization, delays dendritic spine maturation and the NMDA receptor (NMDAR) GluN2B/GluN2A developmental shift. In contrast, α5-NAM treatment from DIV19–21, when hyperpolarizing GABA_AR signaling predominates, enhances surface synaptic GluN2A while decreasing GluN2B. Despite changes in NMDAR subtype surface levels and localization, total levels of key excitatory synapse proteins were largely unchanged, and mEPSCs were unaltered. Importantly, 2-day α5-NAM treatment does not alter the total surface levels or distribution of α5-GABA_ARs, reduce the gephyrin inhibitory synaptic scaffold, or impair phasic or tonic inhibition. Furthermore, α5-NAM inhibition of the GABA_AR tonic current in mature neurons is maintained after 2-day α5-NAM treatment, suggesting reduced tolerance liability, in contrast to other clinically relevant GABA_AR-targeting drugs such as benzodiazepines. Together, these results show that α5-GABA_ARs contribute to dendritic spine maturation and excitatory synapse development via a NMDAR dependent mechanism without perturbing overall neuronal excitability.

1. Introduction

GABA type A receptors (GABA_ARs) are responsible for generating fast inhibitory neuronal transmission and are critical in the regulation of nervous system excitability, mood, memory, and sleep. While GABA_AR positive allosteric modulators (PAMs) such as the benzodiazepine drug class (BZ) are widely used to promote sedation, reduce seizure activity, and lessen anxiety, chronic BZ use is hampered by tolerance and withdrawal symptoms. On the other hand, pro-cognitive and memory enhancing therapeutics based on non-selective GABA_AR antagonists and negative allosteric modulators (NAMs) are greatly limited by the serious side-effects of convulsions, neurotoxicity, and anxiety. However, receptor subtype specific or preferring drugs show greater potential therapeutic utility as there are 19 different GABA_AR subunit genes with varied subcellular, neuronal, brain regional and developmental expression patterns. GABA_AR heteropentameric Cl⁻ channels typically contain two α (α1–6), two β (β1–3), and one γ (γ1–3) or δ subunit (Olsen and Sieghart, 2008). The primary BZ binding site, where many GABA_AR--targeting PAMs and NAMs bind, is formed at the extracellular interface between γ2 and α(1, 2, 3, or 5) subunits. Genetic and pharmacological approaches have identified a specific role for α5 GABA_ARs in learning and memory and BZ-induced cognitive deficits (Collinson et al., 2002; Crestani et al., 2002; Martin et al., 2009). This led to the development of selective α5 GABA_ARs NAMs as cognitive enhancers (Atack, 2010, 2011; Atack et al., 2006a, 2006b, 2009).

The α5 NAMs later showed success in treating pre-clinical mouse models of neurodevelopmental disorders that present excessive inhibition (Ballard et al., 2009; Hausrat et al., 2015; Martinez-Cue et al., 2013; Schulz et al., 2019), leading to trials with Down syndrome patients by F. Hoffmann-La Roche (Liogier d’Ardhuy et al., 2015). In addition, α5 NAMs show antidepressant actions in multiple rodent behavioral assays that are predictive of human anti-depressant efficacy (Carreno et al., 2017; Fischell et al., 2015; Zanos et al., 2017). Like BZ, α5 NAMs bind at the high-affinity benzodiazepine-binding site on GABA_ARs (Maramai et al., 2020). Among the most well studied α5 NAM is the imidazobenzodiazepine L-655,708, a highly selective but low efficacy α5 NAM showing 50- to 100-fold greater affinity for α5 GABA_ARs over α1, α2, and α3 subunit-containing GABA_ARs (Quirk et al., 1996). At the neuronal level, L-655,708 enhances a cellular correlate of learning (long term potentiation, LTP) (Atack et al., 2006a) in a NMDAR dependent fashion (Xue et al., 2017) and can convert long term depression (LTD) to LTP (Martin et al., 2010). Importantly, relatively little is understood regarding the molecular mechanisms underlying the diverse therapeutic actions of α5 NAMs.

The α5 GABA_ARs only make up approximately 5% of the total neuronal GABA_AR population; however, these receptors are particularly abundant within the hippocampal CA1 and CA3 regions (Olsen and Sieghart, 2009). The most prevalent α5 GABA_ARs are α5β3γ2, although mixed α5/α1βγ or α5/α2βγ receptors have been identified (Araujo et al., 1999; del Rio et al., 2001; Ju et al., 2009; Sur et al., 1998). Interestingly, mixed α5/α1 or α5/α2 containing GABA_ARs preferentially assemble with an α5γ interface that confers α5 pharmacological properties at the benzodiazepine-binding site (Araujo et al., 1999; del Rio et al., 2001). In the hippocampus, α5 GABA_ARs are particularly well positioned to sculpt early circuit development due to their exceptionally high expression which peaks in the first two postnatal weeks both in vivo and in vitro (Bader et al., 2017; Liu et al., 1998; Ramos et al., 2004; Yu et al., 2014). Extrasynaptic α5 GABA_ARs play a large role in mediating tonic inhibition (Bonin et al., 2007; Brunig et al., 2002; Crestani et al., 2002; Glykys and Mody, 2006, 2007, 2007; Hausrat et al., 2015; Wu et al., 2021). However, growing evidence indicates the importance of synaptic α5 GABA_ARs (Collinson et al., 2002; Hausrat et al., 2015; Vargas-Caballero et al., 2010; Zarnowska et al., 2009). We previously reported that α5 GABA_ARs are expressed at a constant level at GABAergic synapses across development and associated with the synaptic scaffold protein gephyrin via a motif in the α5 subunit intracellular loop (Brady and Jacob, 2015). Interestingly and in support of a role for synaptic α5 GABA_ARs in early hippocampal circuit development, enhanced dendritic outgrowth was seen at the expense of spine maturation upon genetic redirection of α5 GABA_ARs to extrasynaptic sites (Brady and Jacob, 2015).

An important consideration for GABA_AR drug effects in the developing nervous system is that GABAergic signaling initially depolarizes due to high intracellular Cl⁻ concentration ([Cl⁻]_i) leading to Cl⁻ efflux through the receptors. Thus, early GABA_ARs help drive the establishment of excitatory glutamatergic signaling and play a critical role in dendritic maturation (Ben-Ari et al., 2007; Cellot and Cherubini, 2013; Deng et al., 2007; Rivera et al., 1999). This contrasts with GABA binding to receptors resulting in Cl⁻ influx and hyperpolarizing actions as the nervous system matures. Rising levels of the neuronal Cl⁻ extruder K⁺/Cl⁻ co-transporter 2 (KCC2) at the end of the second week of development in vitro (DIV 14) and in vivo postnatal day 14 promotes the transition to hyperpolarizing GABA_AR activity (Kaila et al., 2014). Along with GABA_ARs, the excitatory glutamatergic GluN2B-containing NMDA receptors (NMDARs) also regulate early circuit formation by promoting dendritogenesis (Bustos et al., 2014). As synapses develop, depolarization and Ca²⁺ influx results in an activity-dependent rise in GluN2A expression and a switch in NMDAR function from GluN2B-driven to predominantly GluN2A-driven (Barria and Malinow, 2002; Bellone and Nicoll, 2007; Quinlan et al., 1999; Williams et al., 1993), which is thought to be regulated by activity (Bellone and Nicoll, 2007; Quinlan et al., 1999) and to change the threshold for synaptic plasticity (Yashiro and Philpot, 2008). Relatively little is known about the mechanisms coordinating GABA_AR and NMDAR regulation as circuits emerge and the level of signaling integration between these pathways. Furthermore, the effects of α5 GABA_AR targeting drugs on developing circuits is poorly understood.

Here we applied a pharmacological approach to better understand the functional effects of reducing α5 GABA_AR signaling during hippocampal circuit formation in early development with depolarizing GABA and later during hyperpolarizing GABA. Using the α5-preferring NAM L-655,708, we investigated the impact of sustained treatment on dendrite, spine, and synapse development. The findings here further support a role for α5 GABA_AR signaling in dendritic spine maturation and indicate a NMDAR dependent mechanism. Importantly, although 2-day α5 NAM treatment altered surface synaptic NMDAR subtype expression, the maintenance of dendritic gephyrin, surface α5 GABA_AR distribution and levels, and phasic and tonic inhibition suggests α5 NAMs may have sustained efficacy with reduced tolerance liability, contrasting with non-selective GABA_AR targeting drugs such as BZ (Gouzer et al., 2014; Jacob et al., 2009, 2012; Ĺevi et al., 2015; Lorenz-Guertin et al., 2019; Nicholson et al., 2018).

2. Methods

2.1. DNA constructs and antibodies

The GluN1-GFP (Addgene plasmid # 17926; http://n2t.net/addgene:17926; RRID:Addgene_17926), GluN2B-GFP (Addgene plasmid # 17925; http://n2t.net/addgene:17925; RRID:Addgene_17925), and GluN2A-GFP (Addgene plasmid # 17924; http://n2t.net/addgene:17924; RRID:Addgene_17924) cDNA constructs were gifts from Stefano Vicini (Luo et al., 2002). The constructs are rat NMDAR subunits with an N-terminal GFP insertion site (after signal peptide) and were fully sequenced. The following primary antibodies were used for immunofluorescence experiments: α5 GABA A receptor subunit 1:500 224503 Synaptic Systems; Synapsin 1:500 106001 Synaptic Systems; Gephyrin (3B11) 1:300 147111 Synaptic Systems; GFP (dendrite morphology analysis) 1:1000 GFP-1020 Aves; GFP (NMDAR analysis) 1:2000 A6455 Invitrogen. Immunofluorescence secondary antibodies from Invitrogen were all used at 1:1000: anti-rabbit Alexa Fluor 568 A11036; anti-rabbit Alexa Fluor 488 A1103; anti-mouse Alexa Fluor 568 A11031; anti-mouse Alexa Fluor 647 1:1000 A21236, anti-chicken Alexa Fluor 488 1:1000 A11039. Antibodies used for western blotting: GluN1 1:1000 5704 (D65B7) Cell Signaling; GluN2A/NR2A 1:1000 1500 PhosphoSolutions; GluN2B/NR2B 1:1000 N59/36 Antibodies Incorporated; α5 GABA A receptor subunit 1:1000 224503 Synaptic Systems; PSD-95 1:2000 MA1-046 Thermo Fisher Scientific; GAPDH 1:1000 2118 Cell Signaling; β-actin 1:1000 A1978 Sigma. The following secondary HRP coupled antibodies were used: HRP anti-mouse IgG 1:1250 PI32430 Fisher Scientific and HRP anti-rabbit IgG 1:10000 934V GE Healthcare.

2.2. Neuron culture, drug treatment, and transfection

All procedures were approved by the University of Pittsburgh Institutional Animal Care and Use Committee. Hippocampal neuron cultures were prepared from embryonic day 18 rats. Neurons were treated for 48 h with either DMSO (vehicle) or 50 nM L-655,708 (Tocris) added to the culture media followed by fixation or biochemical experiments at either DIV 14–15 or DIV 21–22. For neuronal experiments with NMDA receptor subunits, GluN1-GFP or GluN2B-GFP were nucleofected (Lonza) (Jacob et al., 2005) and GluN2A-GFP was lipofected at DIV 11–12 according to the manufacturers protocol (Lipofectamine 2000, Invitrogen). To visualize dendritic morphology neurons were nucleofected with GFP (eGFP, Clontech).

2.3. Immunofluorescence, imaging, and analysis of neurons

Neurons were treated for 48 h with vehicle or L-655,708 then fixed immediately and stained at DIV 14–15 or DIV 21–22. Surface α5 GABA_AR or surface anti-GFP staining for GluN1, GluN2A or GluN2B NMDAR was done under non-permeabilized conditions, followed by permeabilization and antibody staining for gephyrin or synapsin as previously described (Jacob et al., 2005). Images were taken on a Nikon A1 Confocal microscope using a 60x oil immersion objective (NA 1.49) at a 2x zoom. Samples for confocal imaging were sequentially scanned with individual lasers (488, 561, 650 nm) and an appropriate emission band pass filter (500–550 nm, 575–625 nm), or long pass filter (650 LP nm) to avoid any spectral bleed-through between channels. For comparison between vehicle and drug treatments the researcher was blinded to the experimental conditions, and laser settings for acquisition and thresholding values for analysis were kept constant within a culture. For each independent culture, measurements from the drug-treated group were normalized to the vehicle-treated group average. Data were analyzed using NIS Elements software (Nikon, NY). For surface receptor area and intensity quantification, 20-μm length regions of interest (ROIs) from two proximal dendrites per neuron were used for analysis. Analysis of synaptic and extrasynaptic receptor populations was determined by colocalization or exclusion from synaptic regions as previously described for (Tretter et al., 2008). The area covered by receptors and the pixel intensity of receptors were quantified, as surface receptor changes can independently modify these parameters. For example, if more receptors are packed into a synaptic area, the pixel intensity can increase without the area; vice versa, receptors may cover a larger synaptic area without an increase in receptor density, resulting in no change in intensity. To analyze dendrite morphology, GFP expressing neurons were treated and fixed as described above, followed by permeabilization and anti-GFP staining. 3D reconstructions of confocal z series acquired with a 40x oil immersion objective, 1x zoom, were used for Sholl analysis, and analyzed using Fiji ImageJ software (Schindelin et al., 2012; Schneider et al., 2012). Spine morphology studies were performed as previously described (Jacob et al., 2009). Briefly, images were analyzed from 3D reconstructions of confocal z series acquired with a 60x oil immersion objective using a 4x zoom. Spine length was determined using NIS Elements by measuring the length from the spine head to where the spine intersected the dendritic shaft. Spines were classified as either mushroom or filopodia, with a mushroom spine having a spine head at least twice the size of the spine neck, and the remaining spines classified as filopodia.

2.4. Lysate preparation, Western blot

Hippocampal neuronal cultures were treated with vehicle or the α5 GABA_AR NAM L-655,708 at 50 nM for 48 h at DIV 12–13 or 19–20. Neurons were biotinylated as described previously (Saliba et al., 2007) using NHS-SS-Biotin (Thermo Fisher Scientific) at 0.5 mg/ml for 15 min at 4 °C. Neurons were lysed in RIPA containing 50 mM Tris-HCl (pH 7.6), 150 mM NaCl, 1 mM EDTA, 1% Igepal, 0.5% Na-Deoxycholate, 0.1% SDS, 10 mM NaF, 2 mM sodium orthovanadate, and protease inhibitor cocktail (Sigma, St. Louis, MO), sonicated, solubilized for 15 min at 4 °C, and spun at 13000 rpm for 15 min at 4 °C. Isolated biotinylated surface proteins and total proteins were resolved by SDS-PAGE, followed by western blotting. Total protein samples were 10% of the biotinylated protein input and β-actin or GAPDH were used as loading controls. The absence of the cytoplasmic protein β-actin was used to validate surface specific labeling in the biotinylated samples.

2.5. Electrophysiology

Following 48-h treatment with either DMSO (vehicle) or 50 nM L-655,708, whole cell patch clamp recordings were performed on hippocampal neuron cultures. Pyramidal neurons were visualized by IR-DIC video microscopy and identified by their apical dendrites and large triangular somata. Cells were perfused with a 95% O₂/5% CO₂ gas mixture. Patch electrodes (5–10 MΩ open-tip resistance) were filled with an intracellular solution containing (in mM): 105 Cs-gluconate, 2 MgCl₂, 10 NaCl, 10 HEPES, 10 phosphocreatine, 4 ATP-Mg, 0.3 GTP, and 10 BAPTA; pH 7.25. Extracellular Ringer solution of the following composition was used (in mM): 126 NaCl, 24 NaHCO₃, 2.5 KCl, 1.25 NaH₂PO₄, 1 MgSO₄, 2 CaCl₂, 10–20 glucose; pH ~7.3. Voltage and current recordings were performed with a Multi-Clamp 700A amplifier (Axon Instruments, Union City, CA). Signals were filtered at 2 kHz and acquired at a sampling rate of 10 kHz using Clampex 10.2 software (Molecular Devices Corporation, Sunnyvale, CA). Access resistance was 10–20 MΩ and remained relatively stable during experiments (≤30% increase). Liquid junction potential (−13 mV) was corrected.

Miniature (action potential independent) postsynaptic currents were recorded in the presence of tetrodotoxin (TTX; 0.5 μM) to inhibit voltage-gated Na⁺ channels. Miniature excitatory postsynaptic currents (mEPSCs) were isolated by holding neurons at −70 mV (close to the reversal potential of inhibitory responses, as previously described (Povysheva and Johnson, 2016). As expected, at this holding potential, all visible events were eliminated by application of the AMPAR and kainate receptor blocker 2,3-dihydroxy-6-nitro-7-sulfamoylbenzo(F) quinoxaline (NBQX; 20 μM) and the NMDAR receptor blocker D-2-amino-5-phosphopentanoic acid (AP-5; 50 μM). Miniature inhibitory postsynaptic currents (mIPSCs) were isolated by holding neurons at +12 mV (close to the reversal potential of excitatory responses. At + 12 mV, all visible events were eliminated by application of the GABA_AR antagonist gabazine (10 μM). Gabazine, NBQX, and AP-5 were purchased from Ascent Scientific LTD (Bristol, UK); TTX from Sigma (St. Louis, MO).

Miniature events were analyzed using the MiniAnalysis Program (Synaptosoft, Decatur, GA) as previously described (Povysheva and Johnson, 2016). Peak events were first detected automatically using an amplitude threshold of two times the average RMS noise. Membrane properties were analyzed in Clampfit (Molecular Devices Corporation, Sunnyvale, CA). To characterize the passive membrane properties of neurons, hyperpolarizing current steps were applied for 500 ms in 10 pA increments at 0.5 Hz. Input resistance was measured from the slope of a linear regression fit to the voltage-current relation in a voltage range hyperpolarized of resting potential. The membrane time constant was determined by single-exponential fitting to the average voltage responses activated by hyperpolarizing current steps (5–15 pA).

In the tonic current experiments, the baseline was recorded in the presence of NBQX (20 μM), AP-5 (50 μM), and TTX (0.5 μM). The α5-dependent tonic GABA_AR current was measured as the shift in baseline produced by the application of 50 nM α5 NAM L-655,708 (L6; Tocris). The baseline current was measured in the Statistics package of Clampfit during five consecutive 0.2-min intervals before the application of L6 and 3 min after its application (the time point after L6 application when baselines reach plateau in all cells). Total tonic GABA_AR current was measured as the difference in the baseline before L6 application and 2 min after GABA_AR blocker Picrotoxin (100 μM) application. The tonic current data was analyzed with a mixed-effects ANOVA analysis to compensate for missing values in the dataset, as repeated measures ANOVA cannot handle this situation. GraphPad Prism uses a compound symmetry covariance matrix and fits the data with a Restricted Maximum Likelihood (REML). This approach produces the same p-values as a repeated measures ANOVA in the absence of missing data points as well as comparable p-values when a given dataset contains missing values.

2.6. Statistical analysis

Statistical analyses were performed using GraphPad Prism 9.1 (San Diego, CA). Immunofluorescence and morphological studies were performed on 3–5 independent neuron cultures, 12–18 neurons per treatment group per culture, and analyzed using unpaired two-tailed t-tests with Welch’s Correction or two-way ANOVA with Sidak’s post hoc comparisons. For analysis of immunofluorescence intensity, data were normalized to the vehicle-treated group and represented as percent change. Biochemical studies were performed on 3–7 independent cultures, 2–3 replicate plates per treatment group per culture, and analyzed using unpaired two-tailed t-tests with Welch’s Correction. Electrophysiology studies were performed on 2–4 independent cultures, 3–4 neurons per treatment group per culture, and analyzed using unpaired two-tailed t-tests with Welch’s Correction (mIPSC & mEPSC) or a mixed-effects ANOVA with Geisser-Greenhouse’s correction and Sidak’s post hoc comparisons (tonic current). All values are given as mean ± SEM. Detailed statistical information can be found in the figure legends.

3. Results

3.1. Spine maturation is dependent on α5 signaling

We previously showed that the α5 GABA_AR subunit plays a role in both dendritic outgrowth and spine maturation (Brady and Jacob, 2015). To better understand how impaired α5 GABA_AR signaling affects dendritic maturity across development, we turned to a pharmacological approach using the α5-selective NAM, L-655,708. We first assessed the dendritic morphology of cultured hippocampal neurons after a 2-day treatment with either vehicle (DMSO) or 50 nM L-655,708 both during depolarizing GABA (DIV 14) and later during hyperpolarizing GABA (DIV 21) (Fig. 1). At DIV 14 (Fig. 1A), dendritic spines were comparable in total number and length, between α5 NAM-treated and vehicle-treated neurons. L-655,708 treatment decreased the number of mushroom spines and increased the number of filopodial spines, resulting in a decrease in the ratio of mushroom to filopodia spines (Fig. 1A). At DIV 21 (Fig. 1B), again the total spine number and length was comparable between treatments. While mature mushroom spine number was unchanged by α5 NAM treatment, the number of filopodial spines was increased, representing an immature dendritic phenotype. Overall, this led to a decrease in the ratio of mushroom to filopodia spines (Fig. 1B). Next, we used Sholl analysis to assess the dendritic complexity of α5 NAM-treated and vehicle-treated neurons (Fig. 1C and D). Dendritic complexity was determined by measuring the number of dendritic intersections, the number of primary, secondary, and tertiary dendritic branch points, and the total dendritic length. At DIV 14 (Fig. 1C) and at DIV 21 (Fig. 1D), no change was seen in any of these parameters between α5 NAM-treated and vehicle-treated neurons. Together, these data show that treatment of hippocampal neurons in vitro with an α5 GABA_AR NAM causes a reduction in dendritic spine maturation both before and after reversal of GABA polarity (respectively, DIV 14 and DIV 21) without altering dendritic outgrowth or complexity.

Fig. 1. Sustained α5 GABA_AR negative allosteric modulation reduces dendritic spine maturation across development without affecting dendritic branch complexity.

GFP-transfected hippocampal neurons were treated with vehicle (Veh) or 50 nM α5 NAM L-655,708 (L6) for 48 h then fixed immediately at either DIV 14–15 (A,C) or DIV 21–22 (B,D) and stained for GFP under permeabilized conditions. A,B) High magnification confocal z-series through dendritic regions were obtained, and 3D reconstructions were used to analyze the morphology, density, and length of dendritic spines. Filopodia spines are represented by X; mushroom spines are represented by #. A) α5 NAM-treated DIV 14 neurons exhibited a greater number of filopodia spines (Veh = 2.6 ± 0.11, L6 α5 NAM = 3.5 ± 0.17, t(118.5) = 4.415, p < 0.0001) and fewer mushroom spines (Veh = 3.8 ± 0.18, L6 α5 NAM = 2.9 ± 0.19, t(135.0) = 3.124, p = 0.0022) compared to vehicle as well as a decrease in the ratio of mushroom/filopodial spines (Veh = 1.6 ± 0.11, L6 α5 NAM = 0.96 ± 0.070, t(113.2) = 5.495, p < 0.0001) [n = 66–71 neurons from 4 independent cultures]. B) α5 NAM treated DIV 21 neurons showed an increase in the number of filopodia spines (Veh = 3.3 ± 0.16, L6 α5 NAM = 4.0 ± 0.17, t (170.9) = 3.103, p = 0.0022) and an overall decrease in mushroom/filopodial spine ratio compared to vehicle (Veh = 1.3 ± 0.12, L6 α5 NAM = 0.94 ± 0.073, t (134.8) = 2.700, p = 0.0078) [n = 77–90 neurons from 4 independent cultures]. There was no difference in average spine length or the number of spines per 10 μm in treated neurons vs vehicle at DIV 14 (A) or at DIV 21 (B). C,D) Maximum projection of confocal z-series of vehicle vs α5 NAM treated neurons. Sholl analysis were used to analyze dendritic branching and outgrowth. There was no difference in Sholl Analysis, number of branches, or total dendritic length at either DIV 14 [C; n = 43–45 neurons from 3 independent cultures] or DIV 21 [D; n = 64–69 neurons from 4 independent cultures]. **p < 0.01 and ****p < 0.0001 using two-tailed t-tests with Welch’s Correction or 2-way ANOVAs with Sidak’s post hoc comparisons. Scale bars are and 5 μM for dendrite zoom images (A,B) and 20 μM for neuron images (C,D).

3.2. Prolonged α5 NAM treatment has minimal impact on inhibitory synapses and surface α5 GABA_AR localization or levels

Sustained exposure to non-selective benzodiazepine (BZ) binding-site full agonists results in multiple mechanisms that are implicated in tolerance: a reduction in surface and total levels of BZ-sensitive GABA_AR subtypes, reduction in gephyrin postsynaptic scaffold, and destabilization of BZ-sensitive GABA_AR synaptic tethering (Jacob et al., 2009, 2012; Lorenz-Guertin et al., 2019). Furthermore, treatment with GABA_AR antagonists that are not receptor subtype specific leads to an increase in surface levels of GABA_ARs (Gouzer et al., 2014; Jacob et al., 2009, 2012; Ĺevi et al., 2015; Lorenz-Guertin et al., 2019). To evaluate potential α5 NAM induced changes in inhibitory synapses and α5 GABA_ARs we analyzed α5 subunit surface expression, synaptic and extrasynaptic localization, and intracellular gephyrin, the primary synaptic scaffold for GABAergic synapses, using immunofluorescence and confocal microscopy. Following 2-day α5 NAM treatment, both DIV 14 and 21 neurons did not show a change in the area or intensity of synaptic surface α5 GABA_ARs (Fig. 2A and B). Similarly, the area and intensity of total surface α5 GABA_ARs was unchanged. The dendritic area covered by extrasynaptic receptors was not reduced in α5 NAM-treated neurons at DIV 14 or DIV 21, although there was a slight decrease in extrasynaptic intensity of α5 GABA_ARs at DIV 14. Both the area and intensity of the gephyrin scaffold was unaltered by α5 NAM treatment at either developmental stage (Fig. 2A and B). Together, these results suggest that while α5 NAM treatment leads to a more immature dendritic phenotype, the availability of α5 GABA_ARs and the architecture of inhibitory synapses is largely unaffected.

Fig. 2. Inhibitory synapse size, synaptic α5 GABA_AR and total dendritic α5 GABA_AR surface levels are not modified by α5 NAM treatment.

Non-transfected hippocampal neuronal cultures were treated with vehicle (Veh) or 50 nM α5 NAM L-655,708 (L6) for 48 h then fixed immediately at DIV 14–15 (A) or DIV 21–22(B). Fixed cells were first stained under nonpermeabilized conditions for surface α5 GABA_ARs, then permeabilized and stained for inhibitory postsynaptic marker gephyrin. Colocalization of surface α5 GABA_ARs with gephyrin was used to define synaptic receptors. A) At 14 DIV α5 NAM treated neurons did not show changes in total surface or synaptic α5 GABA_AR area or intensity. Only the intensity of dendritic extrasynaptic α5 decreased slightly (Veh = 100 ± 5.85, L6 α5 NAM = 80.88 ± 7.49, t(68.42) = 2.012, p = 0.0482). Gephyrin area and intensity were also unchanged by α5 NAM treatment (n = 36–41 cells from 3 independent cultures). B) At DIV 21 α5 NAM treated neurons did not show changes in surface total, synaptic or extrasynaptic α5 GABA_AR area or intensity. Gephyrin area and intensity were also unchanged by α5 NAM treatment [n = 38–42 cells from 3 independent cultures]. *p < 0.05 using two-tailed t-tests with Welch’s Correction. Scale bars are 20 μM for neuron images and 2 μM for dendrite zoom images.

3.3. Impaired α5 GABA_AR signaling delays developmental synaptic NMDAR subunit switch

As α5 NAM treatment produced no definitive changes in inhibitory synapses or surface α5 GABA_AR, we next investigated potential effects on NMDARs, a critical regulator of dendritic maturation. Tetrameric NMDAR are formed by two obligate GluN1 and two variable GluN2 (A-D) subunits, with GluN2A and GluN2B being the vastly predominant subunits. GluN2B NMDARs promote dendritogenesis at early developmental timepoints, and as GluN2A expression rises, this process becomes inhibited (Bustos et al., 2014). Delaying the rise of GluN2A expression also leads to immature dendritic phenotypes (Akashi et al., 2009). We investigated the effect of α5 NAM treatment on NMDAR subunit expression in hippocampal cultures transfected with GluN1-GFP, GluN2B-GFP, or GluN2A-GFP. As in the above sections, neurons were treated with α5 NAM or vehicle for two days either before or after the GABA polarity reversal, followed by immunofluorescence staining to evaluate the resulting surface levels and distribution of NMDAR subunits. NMDAR subunit surface fluorescence intensity and area were measured, with the presynaptic marker synapsin (a regulator of the presynaptic vesicle pool) used for assessment of synaptic vs extrasynaptic localization.

At DIV 14, several distinct changes in surface NMDAR subtypes were seen in α5 NAM-treated neurons compared with vehicle while total synapsin expression was unchanged (Fig. 3). α5 NAM-treated neurons showed a slight decrease in total dendritic surface area for GluN1, with the decrease in area occurring in extrasynaptic regions, while no change was seen in the synaptic regions (Fig. 3A). Total surface levels of GluN2A containing NMDAR were also decreased by α5 NAM treatment, with loss measured both in area and intensity (Fig. 3B). Moreover, surface GluN2A decreased in both synaptic and extrasynaptic locations. Conversely, total surface GluN2B was increased with α5 NAM treatment, predominantly driven by enhancement of synaptic GluN2B area and intensity (Fig. 3C). This pattern of low GluN2A and high GluN2B expression parallels an immature development profile, suggesting that treatment with an α5 NAM while GABA has depolarizing effects leads to a delay in the developmental functional transition from GluN2B to GluN2A.

Fig. 3. Reducing α5 GABA_AR activity at DIV 14 delays the NMDAR synaptic maturation transition from GluN2B to GluN2A.

Hippocampal neuronal cultures were transfected with either GluN1-GFP (A), GluN2A-GFP (B), or GluN2B-GFP (C). Cultures were treated with vehicle (Veh) or 50 nM α5 NAM L-655,708 (L6) for 48 h then fixed immediately at DIV 14–15. The fixed cells were first stained under nonpermeabilized conditions to identify GluN1, GluN2A or GluN2B containing NMDARs (anti-GFP), then permeabilized and stained for the presynaptic marker synapsin. Colocalization of surface GluN1-GFP, GluN2A-GFP, or GluN2B-GFP with synapsin was used to define synaptic NMDAR subunit levels. A) L-655,708 decreased GluN1 total area (Veh = 1.89 ± 0.18, α5 NAM = 1.41 ± 0.15, t(112.5) = 2.043, p = 0.0434) and extrasynaptic dendritic area (Veh = 0.90 ± 0.14, α5 NAM = 0.55 ± 0.073, t(86.76) = 2.150, p = 0.034). No change in synapsin was found [n = 59–60 neurons from 4 independent cultures]. B) α5 NAM treatment decreased total surface GluN2A area (Veh = 4.49.29 ± 0.55, L6 α5 NAM = 2.65 ± 0.48, t(62.34) = 2.511, p = 0.0146) and intensity (Veh = 100.0 ± 10.8, L6 α5 NAM = 50.0 ± 7.1, t(55.54) = 3.861, p = 0.0003). GluN2A was decreased in synapses and extrasynaptic regions: synaptic area (Veh = 1.29 ± 0.12, L6 α5 NAM = 0.85 ± 0.11, t(65.25) = 2.660, p = 0.0098); synaptic intensity (Veh = 100.0 ± 10.0, L6 α5 NAM = 52.9 ± 6.19, t(53.24) = 3.992, p = 0.0002); extrasynaptic area (Veh = 3.23 ± 0.48, L6 α5 NAM = 1.85 ± 0.44, t(63.08) = 2.122, p = 0.0377) and extrasynaptic intensity (Veh = 100.0 ± 13.1, L6 α5 NAM = 43.32 ± 8.7, t(55.77) = 3.595, p = 0.0007). No change in synapsin was found [n = 32–35 neurons from 3 independent cultures]. C) L-655,708 increased surface GluN2B in synapses: area (Veh = 1.32 ± 0.11, L6 α5 NAM = 1.77 ± 0.15, t(114) = 2.440, p = 0.0162) and intensity increased (Veh = 100.0 ± 8.3, L6 α5 NAM 144.8 ± 14.4, t(94.60) = 2.703, p = 0.0081). The total surface GluN2B intensity was increased (intensity Veh = 99.6 ± 7.6, L6 α5 NAM = 126.3 ± 9.4, t(111.5) = 2.205, p = 0.0295). No change in synapsin levels was found [n = 58–63 neurons from 4 independent cultures]. *p < 0.05, **p < 0.01, and ***p < 0.001, using two-tailed t-tests with Welch’s Correction. Scale bars are 20 μM for neuron images and 2 μM for dendrite zoom images.

In contrast to DIV 14, α5 NAM 2-day treatment up to DIV 21 inverted the effects on GluN2A and GluN2B NMDAR, increasing synaptic levels of surface GluN2A (Fig. 4B) and decreasing GluN2B (Fig. 4C), although the changes were more subtle and constrained to synapses (Fig. 4). No α5 NAM induced changes in GluN1 were detected (Fig. 4A). Overall GluN2A surface levels showed a non-significant trend toward an increase in the α5 NAM treated neurons (Fig. 4B, intensity, p = 0.0763). At DIV 21, synapsin distribution and levels were comparable between treatment groups (Fig. 4A–C). In summary, α5 NAM in early circuit development leads to a delay in the NMDAR GluN2B to GluN2A functional switch, likely by reducing the depolarizing GABA_AR signaling that promotes synapse growth and maturation. However, at DIV 21, when GABA_AR functions to hyperpolarize and inhibit neuronal activity, α5 NAM treatment further decreases surface synaptic GluN2B levels while increasing synaptic GluN2A, likely by driving neuronal excitability and thus activity-dependent increases in GluN2A.

Fig. 4. Reducing α5 GABA_AR activity in an established circuit at DIV 21 increases surface synaptic GluN2A while decreasing synaptic GluN2B NMDARs.

Hippocampal neuronal cultures were transfected with either GluN1-GFP (A), GluN2A-GFP (B), or GluN2B-GFP (C). Cultures were treated with vehicle (Veh) or 50 nM α5 NAM L-655,708 (L6) for 48 h then fixed immediately at DIV 14–15. The fixed cells were first stained under nonpermeabilized conditions to identify GluN1, GluN2A or GluN2B containing NMDARs (anti-GFP), then permeabilized and stained for the presynaptic marker synapsin. Colocalization of surface GluN1-GFP, GluN2A-GFP, or GluN2B-GFP with synapsin was used to define synaptic NMDAR subunit levels. A) L-655,708 did not significantly affect surface GluN1. No change in synapsin was found [n = 48–56 cells from 3 independent cultures]. B) L-655,708 increased the intensity of surface synaptic GluN2A levels compared to vehicle treatment (Veh = 100.0 ± 10.3, L6 α5 NAM 176.3 ± 28.7, t(47.78) = 2.500, p = 0.0159). No change in synapsin was found [n = 39–45 cells from 3 independent cultures]. C) L-655,708 significantly decreased the intensity of surface synaptic GluN2B levels compared to vehicle treatment (Veh = 100.0 ± 9.1, L6 α5 NAM 70.6 ± 6.7, t(96.73) = 2.500, p = 0.0109). No change in synapsin levels was found [n = 49–55 cells from 4 independent cultures]. *p < 0.05, **p < 0.01 using two-tailed t-tests with Welch’s Correction. Scale bars are 20 μM for neuron images and 2 μM for dendrite zoom images.

3.4. Prolonged α5 NAM treatment primarily modulates surface signaling at excitatory synapses while α5 GABA_AR are unchanged

To complement these immunofluorescence and confocal microscopy studies, we used a biochemical approach to measure cell surface levels and total protein levels of key neurotransmitter receptors and scaffolding of both inhibitory and excitatory synapses. As before, hippocampal neuronal cultures were treated with 50 nM L-655,708 for 48 h followed by cell surface biotinylation and lysis at DIV 14 or DIV 21. Samples were analyzed by SDS-page and standard western blotting methods. 2-day α5 NAM treatment at both DIV 14 and 21 did not alter surface or total levels of α5 GABA_ARs (Fig. 5A and B). Total gephyrin levels were unchanged at DIV 14 and increased slightly at DIV 21. There was no statistical difference between total, surface, or surface/total protein levels following α5 NAM treatment at DIV 14 for NMDAR subunits GluN1, GluN2A, GluN2B, and excitatory synaptic scaffold PSD-95 (Fig. 5A). At DIV 21, there were no changes detected in GluN2A (surface, total or surface/total ratio), GluN1, or PSD-95. However, there was a significant decrease in surface and surface/total protein levels of GluN2B (Fig. 5B). These biochemical data, together with the imaging analysis (Figs. 2–4), show the primary α5 NAM-induced change is a reduction in GluN2B at DIV 21, with relatively minor perturbation of the overall Excitation/Inhibition balance as determined by excitatory and inhibitory synapse receptor distribution and levels.

Fig. 5. Prolonged α5 NAM treatment does not result in major perturbation of surface or total levels of key inhibitory or excitatory synapse protein levels.

Non-transfected hippocampal neuronal cultures were treated with vehicle (Veh) or 50 nM α5 NAM L-655,708 (L6) for 48 h then at either DIV 14 (A) or DIV 21 (B) surface proteins were biotinylated. Cells were then lysed and both surface fraction and total protein levels were probed for the α5 GABA_AR subunit, GluN2A, and GluN2B; total protein levels were additionally assessed for gephyrin, GluN1, and PSD-95 [n = 6–13 from 4 to 7 independent cultures]. A) At DIV 14, treatment with L-655,708 did not alter either surface, total, or surface-to-total protein levels. B) At DIV 21, treatment with L-655,708 decreased both GluN2B surface and surface-to-total protein levels (surface: Veh = 100.0 ± 3.3, L6 α5 NAM = 71.0 ± 7.4; t(7.0) = 3.567, p = 0.0091; surface/total: Veh = 101.5 ± 5.7, L6 α5 NAM = 63.9 ± 8.5; t (9.2) = 3.682, p = 0.0048) and increased gephyrin total protein levels (Veh = 100.0 ± 4.7, L6 α5 NAM = 118.6 ± 7.2; t(15.74) = 2.165, p = 0.0461). *p < 0.05, **p < 0.01 using two-tailed t-tests with Welch’s Correction.

3.5. Excitatory and inhibitory synaptic currents are unchanged by prolonged α5 NAM treatment, and α5 NAM-dependent inhibition of tonic current is maintained

To reveal the functional impact of 2-day α5 NAM treatment on basal neurotransmission at both inhibitory and excitatory synapses, we assessed mIPSC and mEPSC parameters in vehicle or α5 NAM-treated hippocampal neurons at DIV 14 (Fig. 6A and B) and DIV 21 (Fig. 6C and D). Although mEPSCs are mediated predominantly by AMPARs, changes in NMDAR subunit expression or function can promote changes in AMPAR activity and excitation (Ferreira et al., 2015; Kato et al., 2007). In particular, GluN2B negatively regulates AMPAR trafficking (Ferreira et al., 2015; Hamada et al., 2014) and formation of functional synapses (Gray et al., 2011). α5 NAM treatment did not affect the amplitude or frequency of mIPSCs and mEPSCs at either time point (Fig. 6B,D). Additionally, there was no difference between groups in input resistance (DIV 14–15: Veh = 378.6 ± 53.02 MΩ, L6 = 347.4 ± 45.63 MΩ, t(12.93) = 0.4449; DIV 21–22: Veh = 299.8 ± 40.4 MΩ, L6 251.3 ± 26.41 MΩ, t(12.30) = 1.003) or tau of decay (14–15 DIV: Veh = 30 ± 9 ms, L6 = 33 ± 12 ms, t(11.43) = 0.6522; DIV 21–22: Veh = 31 ± 12 ms, L6 = 27 ± 8 ms, t(11.33) = 0.7061). Lack of mEPSC amplitude changes suggests AMPARs are not altered by α5 NAM treatment. Furthermore, the absence of changes in PSD-95 levels for α5 NAM-treated neurons (Fig. 5) is also consistent with no excitatory synapse functional potentiation (Ehrlich et al., 2007). The absence of functional effects in inhibitory phasic currents is consistent with no change in surface synaptic GABA_ARs maintenance and no loss of gephyrin synaptic scaffold (Figs. 2 and 5).

Fig. 6. Prolonged α5 GABA_AR NAM treatment either during or after circuit development does not change miniature excitatory or inhibitory postsynaptic currents.

Hippocampal neuronal cultures were treated with vehicle (Veh) or 50 nM α5 NAM L-655,708 (L6) for 48 h then whole-cell recordings of spontaneous miniature excitatory and inhibitory postsynaptic currents were obtained in the presence of TTX at DIV 14–15 (A,B) or DIV 21–22 (C,D), respectively, and analyzed using two-tailed t-tests with Welch’s Correction. Treatment with L-655,708 did not affect the amplitude or frequency of mEPSCs or mIPSCs at either timepoint [n = 7–9 cells from 3 to 4 independent cultures].

α5-containing GABA_ARs generate a large component of the tonic current in hippocampal CA1 and CA3 pyramidal cells, as determined from both knockout experiments and measurements using the α5 NAM L-655,708 (Glykys and Mody, 2006), with the remaining tonic current generated by δ subunit containing GABA_ARs (Glykys et al., 2008). To determine if prolonged L6 treatment had altered the tonic current, we performed whole-cell recordings on 48-h treated neurons. We perfused L-655,708 to reveal the α5 GABA_AR component of the tonic current, followed by perfusion of the GABA_AR channel blocker picrotoxin to measure the total tonic current (Fig. 7A). The α5 GABA_AR component of the tonic current and the total GABA_AR tonic current were not significantly different between neurons after 48 h L6 or vehicle treatment (Fig. 7B). Neurons treated for 48 h with L6 or vehicle were sensitive to α5 NAM-dependent inhibition of the tonic current (Fig. 7C), with the α5 NAM-sensitive component being about half of the total tonic current in both treatment groups (Vehicle = 39.6 ± 3.6% and L6 = 44.5 ± 3.4%, p = 0.3512, t(7.63) = 0.9929).

Fig. 7. The α5 GABA_AR-mediated tonic current and α5 NAM-dependent inhibition of the tonic current is maintained after prolonged α5 NAM treatment.

Hippocampal neuronal cultures were treated with vehicle (Veh) or 50 nM α5 NAM L-655,708 (L6) for 48 h then whole-cell recordings of GABA_AR tonic current were performed [n = 4–6 neurons from 2 independent cultures]. A) Representative traces from Veh or L6 treated neurons obtained during perfusion of L-655,708 (50 nM) and the GABA_AR channel blocker Picrotoxin (100 μM), as indicated by the horizontal lines above traces. The α5 GABA_AR component of the tonic current was measured as the difference in the baseline after L6 application. The total tonic GABA_AR current was measured as the difference in the baseline 2 min after Picrotoxin application. B) Tonic currents in neurons treated for 2 days with L6 or vehicle were sensitive to L-655,708 and picrotoxin. The α5 GABA_AR component of the tonic current (triangles) and the total GABA_AR tonic current (circles) were not significantly different between neurons after 48-h L6 or vehicle treatment (Main effect of treatment: F(1,5) = 0.4817, p = 0.5186; post hoc between groups: L-655,708: veh-treated = 8.88 ± 1.71 pA, L6-treated = 7.98 ± 0.275 pA, t(4.00) = 0.5686, p = 0.8400); picrotoxin: veh-treated = 18.8 ± 1.5 pA, L6-treated = 21.7 ± 3.3 pA, t(3.00) = 0.6161, p = 0.8248). C) The α5 NAM-sensitive tonic current was significantly lower than the total GABA_AR tonic current for both L6 and Veh-treated neurons (Main effect of drug perfusion: F(1,5) = 85.79, p = 0.0002; post hoc within groups: Veh-treated: L-655,708 = 8.88 ± 1.71 pA, picrotoxin = 18.8 ± 1.5 pA, t(5.00) = 7.053, p = 0.0018; L6-treated: L-655,708 = 7.98 ± 0.275 pA, picrotoxin = 21.7 ± 3.3 pA, t(3.00) = 7.298, p = 0.0106). *p < 0.01 and **p < 0.001 using mixed-effects ANOVA with Geisser-Greenhouse’s correction and Sidak’s post hoc comparisons.

Overall, this functional data suggests that although α5 NAM treatment delays spine maturation and shifts GluN2B and GluN2A NMDAR subtype synaptic content, it does not drastically perturb circuit establishment and excitation or inhibition. The subtle effect of α5 NAM-increased GluN2A synaptic levels at 21 DIV without changes in GABA_AR phasic and tonic currents are consistent with human and rodent studies indicating pro-cognitive activity with lack of anxiogenic or epileptogenic activity (Atack et al., 2006a, 2009). Maintained sensitivity to α5 NAM-dependent inhibition of the tonic current also suggests that tolerance to α5 NAMs over 2-day treatment is minimal.

4. Discussion

Drugs that target α5 GABA_ARs are being tested for their effectiveness in treating neurodevelopmental disorders, mild cognitive impairment, schizophrenia, and depression; however, little is known about the mechanism by which these drugs exert their desired effects and the results of sustained drug exposure. In this study, we investigated the effects of α5 GABA_AR negative allosteric modulation at two key developmental time points to better understand the functional effects of reducing α5 GABA_AR signaling during initial circuit establishment (up to DIV 14) or after synaptic circuitry is well established (DIV 21). α5 NAM treatment reduced spine maturation (Fig. 1), supporting our previous findings that α5 GABA_AR signaling regulates spinogenesis and dendritogenesis (Brady and Jacob, 2015). Two-day treatment with α5 NAM up to DIV 14 (during a period of predominantly depolarizing GABA) enhanced total surface, synaptic, and extrasynaptic GluN2B while decreasing GluN2A (Fig. 3), delaying the synaptic maturation switch between NMDAR GluN2B and GluN2A. In contrast, NAM-induced reduction of hyperpolarizing α5 GABA_AR activity at DIV 21 led to an enhancement of the ratio of synaptic GluN2A/GluN2B (Fig. 4). Somewhat surprisingly, our imaging studies show that α5 NAM treatment did not perturb total surface levels or synaptic distribution of α5 GABA_ARs at either age (Fig. 2). Similarly, the inhibitory postsynaptic gephyrin scaffold was also unchanged (Fig. 2). Biochemical experiments confirmed that α5 NAM treatment does not alter the surface or total protein levels of endogenous α5 GABA_ARs (Fig. 5). Interestingly, a slight increase in gephyrin overall protein levels of α5 NAM-treated neurons was measured in hippocampal lysates at DIV 21. Total protein, surface, and surface/total ratio of GluN1, GluN2A, and GluN2B were unchanged at DIV 14. We did not detect a decrease in surface GluN2A at DIV 14 from biotinylation; perhaps lower levels of GluN2A at this developmental timepoint impaired sensitivity of this method. Biochemical studies of 2-day α5 NAM-treated neurons at DIV 21 showed a decrease in surface GluN2B (Fig. 5), consistent with enhancing the GluN2A/GluN2B ratio. The protein levels of the glutamatergic synapse scaffold PSD95 was not modified by α5 NAM treatment at either age. Lastly, electrophysiological studies in hippocampal neurons showed no changes in mIPSCs or mEPSC parameters after 2-day treatment with L-655,708 at either developmental timepoint, indicating α5 NAM treatment does not overtly perturb either inhibitory or excitatory synapse basal function. Importantly, the α5 GABA_AR-mediated tonic current as well as the total GABA_AR tonic current were also unchanged by 2-day α5 NAM treatment (Fig. 7). Both vehicle and L6-treated neuron groups, showed equivalent sensitivity to α5 NAMs. Together, our findings provide initial mechanistic insight into α5 GABA_AR signaling control of dendritic spine and circuit maturation via modulating synaptic GluN2A/GluN2B distribution. Importantly, α5 NAMs may produce minimal tolerance as 2-day treatment did not decrease α5 GABA_AR levels, reduce inhibitory synapse size, or impair phasic or tonic inhibition. This is in contrast to non-selective PAMs and NAMs, as well as GABA_AR agonists and antagonists, that produce quantifiable changes in receptor and/or gephyrin levels and distribution (Crosby et al., 2019; Gouzer et al., 2014; Gutierrez et al., 2014; Jacob et al., 2009, 2012; Ĺevi et al., 2015; Lorenz-Guertin et al., 2019; Nicholson et al., 2018; Wilhelm and Wenner, 2008).

These findings contribute to the growing body of evidence that supports a novel role for synaptic α5 GABA_AR in regulating synapse plasticity and maturation of neuronal morphology. We previously reported that a significant portion of GABAergic post-synaptic compartments contain α5 GABA_ARs both in vitro and in vivo, and that α5 GABA_AR subunit-gephyrin interaction can occur independent of other α subunits (Brady and Jacob, 2015). Furthermore, genetic experiments that specifically redistributed α5 GABA_ARs out of synapses reduced spine maturity. Here we found that treatment of hippocampal cultures with the α5 GABA_AR NAM L-655,708 promoted an immature filopodial spine phenotype, suggesting the α5 NAM primarily impairs synaptic α5 GABA_AR signaling. Supportive of α5 NAM action at synaptic GABA_ARs, α5 NAMs inhibited GABA evoked currents in neurons (Lecker et al., 2013) and inhibited IPSC amplitudes and accelerated IPSC decay rates in neuron-HEK co-culture expressing α5β1γ2, α5β2γ2, or α5β3γ2 GABA_ARs (Chen et al., 2017). Schultz et al. determined that SST interneurons preferentially target synaptic α5 GABA_AR on dendrites of CA1 hippocampal neurons to generate slow inhibitory postsynaptic currents (slowIPSCs) that regulate pyramidal neuron NMDAR activity and burst firing, while parvalbumin interneurons synapsing onto the perisomatic area showed negligible α5 GABA_AR contribution to IPSCs (Schulz et al., 2018). In these studies, L-655,708 reduced CA1 pyramidal neuron tonic current, while spontaneous currents showed only a reduction in the decay time constant without any change in amplitude or frequency. Detecting α5 NAM modulation of synaptic inhibition is challenging as spontaneous and miniature IPSCs are dominated by perisomatic synapses with minimal α5-subunit component (Schulz et al., 2018). Similarly, a recent study in hippocampal pyramidal neurons reported that all α5 NAMs (L-655,708, basmisanil, Ono-160, a5IA, and MRK-016) inhibited the tonic current generated primarily by extrasynaptic receptors with no effect on miniature inhibitory postsynaptic currents (mIPSCs) (Manzo et al., 2020). Further studies showed that the increased dendritic inhibition and decreased synaptic NMDAR availability in Down Syndrome model mice (Ts65Dn) could be rescued by an α5 NAM (Schulz et al., 2019). Thus, further work is needed to define α5 NAM effects on synaptic vs extrasynaptic receptors and GABAergic neurotransmission.

During development, glutamatergic synapse formation is driven by both GABA_AR and NMDAR signaling (Wang and Kriegstein, 2008). Prior to the establishment of glutamatergic synaptic activity, spontaneous GABA_AR transmission acts to depolarize and excite the neuron (Ben-Ari et al., 2007; Deng et al., 2007). This depolarizing GABA_A receptor transmission results in the influx of Ca²⁺ through NMDARs which activates CaMKII and the Ras-ERK pathway, both of which subsequently activate CREB signaling to promote dendritic outgrowth and synaptic maturation (Kim et al., 2005; Krapivinsky et al., 2003; Passafaro et al., 2001; Sepulveda et al., 2010; Shi et al., 2001; Tian et al., 2004). It is important to note that as the Cl⁻ gradient changes during development, GABA_AR activation can be both depolarizing and inhibitory, which also contributes to balanced circuit formation (Salmon et al., 2020). Our findings that GABA_AR α5 NAM treatment at DIV 14 delayed GluN2A upregulation and resulted in an immature spine phenotype reinforce a permissive role for α5 GABA_AR signaling in synaptic maturation. Interestingly, research into mechanisms underlying α5 NAM anti-depressant and anti-nociceptive actions in the mature nervous system also indicate plasticity at glutamatergic synapses. Investigation of α5 GABA_AR in pain thresholds show that intrathecal administration of α5 NAM decreased pain threshold, boosted basal synaptic transmission, facilitated NMDAR-dependent LTP, and increased NMDAR and AMPAR synaptic levels (Xue et al., 2017). Similarly, α5 NAM fast-acting antidepressant actions rely on enhancement of excitatory synaptic transmission and AMPAR activity (Bugay et al., 2020; Fischell et al., 2015; Zanos et al., 2017).

In summary, we present three key findings relevant to the therapeutic utility of α5 NAMs: 1) α5 NAMs mediate synaptic plasticity events that impact neuron morphological maturation primarily via modulation of NMDAR subtype synaptic surface localization; 2) prolonged α5 NAM exposure does not modify surface α5 GABA_AR expression or inhibitory synapse function; and 3) the α5 GABA_AR-mediated tonic current, total GABA_AR tonic current, and α5 NAM sensitivity are maintained with 2-day α5 NAM treatment. These results indicate that α5 GABA_ARs play a key role in excitatory hippocampal synapse development and α5 NAMs may provide continued efficacy in contrast with other non-subtype preferring GABA_AR drugs that act at the BZ binding site, such as BZs (Gouzer et al., 2014; Jacob et al., 2009, 2012; Ĺevi et al., 2015; Lorenz-Guertin et al., 2019). There is clearly great potential for widespread therapeutic benefits of α5 NAMs. However, further studies on α5 GABA_ARs in rodent neurons and human IPSC-derived neurons are needed to identify molecular mechanisms and post-translational modifications that regulate α5 GABA_AR signaling and the plasticity underlying α5 NAM therapeutic effects. While we observe no change in total or synaptic α5 GABA_ARs and phasic or tonic inhibition over 2-day treatment in hippocampal neurons, it is not known if long-term α5 NAM use has a similar outcome. In addition, it will be critical to assess in vivo effects of α5 NAM treatment on both developing and mature brain circuitry as aspects of plasticity may differ in the intact brain compared to cultured neuronal systems. This is clinically important, as long-term reductions in α5 GABA_AR levels or function in genetic mouse models and human patients produces pathologies including impaired cognition, increased anxiety, autism related behaviors, sleep disturbance and increased epilepsy susceptibility (Bakker and Oostra, 2003; Kazdoba et al., 2016; Mesbah-Oskui et al., 2017; Zurek et al., 2016). The recently published preclinical and early clinical profile data on the α5 NAM basmisanil (progressed to Phase II trials for neurodevelopmental and schizophrenia disorders), show drug selectivity, pro-cognitive effects in non-human primates, as well as safety, tolerability, and [¹¹C]-Ro 15–4513 PET analysis of maximum α5 GABA_AR occupancy in healthy human volunteers (Hipp et al., 2021). Further investigation of synaptic plasticity mechanisms and the effects of long-term treatment with α5 NAMs will be of great utility in determining therapeutic application of α5 GABA_AR-preferring drugs.