See Table 1 for information on antibodies and materials used in this study.

All procedures were approved by the University of Pittsburgh Institutional Animal Care and Use Committee. Cortical or hippocampal neuronal cultures were prepared using procedures similar to those previously described (Jacob et al., 2005; Sahu et al., 2019). Briefly, cortical or hippocampal tissue was dissected from embryonic day 18 Sprague Dawley rats and dissociated with papain and trypsin inhibitor before resuspension in Neurobasal Media supplemented with B27 (Gibco). For FRAP experiments, neurons were nucleofected (Lonza) at plating with mScarlet-gephyrin.FingR (Gross et al., 2013; Bensussen et al., 2020) and γ2^pHFAP (Lorenz-Guertin et al., 2017). Neurons were then cultured undisturbed until DIV 13–15, at which point they were treated with vehicle (0.1% DMSO) or 1 μM DZP (Sigma D0899) for 7 days and collected for experiments at DIV 20–22.

Whole-cell patch-clamp recordings were performed on cortical neuron cultures at DIV 20–22 following 7-day treatment with vehicle or 1 μM DZP. Pyramidal neurons were visualized by IR-DIC video microscopy and identified by their apical dendrites and large triangular soma. Patch electrodes (5–10 MΩ open-tip resistance) were filled with an intracellular solution containing (in mM): 140 CsCl, 2 MgCl₂, 0.1 CaCl₂, 10 HEPES, 10 phosphocreatine, 4 ATP-Mg, 0.3 GTP, and 1.1 EGTA; 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; bubbled with a 95% O₂/5% CO₂ gas mixture; pH ~7.3. Current recordings were performed with a Multi-Clamp 700A amplifier (Axon Instruments, Union City, CA, USA). 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, USA). Access resistance was 10–20 MΩ and remained relatively stable during experiments (≤ 30% increase). Recordings were corrected for the liquid junction potential. All currents were recorded at a holding potential of −70 mV. Miniature inhibitory postsynaptic currents (mIPSCs) were recorded in the presence of NBQX (20 μM), D-APV (50 μM), and TTX (1 μM) to inhibit AMPAR, NMDAR, and voltage-gated sodium channels, respectively. Miniature events were analyzed using the MiniAnalysis Program (Synaptosoft, Decatur, GA, USA) as previously described (Povysheva and Johnson, 2016). The averaged data per cell was used for analysis to compare 7-day vehicle and DZP-treated neurons. To assess the ability of BZDs to potentiate mIPSCs, neurons which had been 7-day treated with vehicle or DZP were acutely applied 1 μM diazepam during recording. The corresponding increase to mIPSC amplitude and tau of decay (τ_decay) was then measured each for 7-day vehicle and DZP-treated neurons. The percent DZP potentiation of mIPSC amplitude and τ_decay was then determined by the percent change from baseline upon acute application of 1 μM diazepam.

Super-resolution imaging was carried out using DNA-PAINT. In this method, primary antibodies against the target protein of interest are recognized by secondary nanobodies coupled to short single strands of DNA (docking strand). The complementary DNA single strands are coupled to a fluorescent dye (imager strand), which is added to the sample during imaging. Transient binding of the docking and imager strands produces fluorescent blinking events (protein localizations), many of which are captured over an extended imaging period and compiled (Schnitzbauer et al., 2017). This method effectively achieves protein localization with high spatial resolution. Here, DIV 20–22 neurons were collected at the end of the 7-day drug treatment, rapidly washed with DPBS, and fixed for 10 min in PBS containing 4% paraformaldehyde (PFA) and 4% sucrose. Primary antibodies against the γ2-GABA_AR subunit and gephyrin were each separately pre-incubated for 20 min with custom-made single-domain secondary nanobodies coupled to oligonucleotides (Massive Photonics) such that the nanobodies were in 2.5 molar excess of the respective primary antibody (Sograte-Idrissi et al., 2020). After blocking in blocking solution [DPBS containing 10% horse serum and 0.5% bovine serum albumin (BSA)], neurons were incubated overnight with the γ2 subunit antibody/nanobody pre-mix to identify surface γ2-GABA_ARs. The next day, neurons were permeabilized for 10 min with blocking solution containing 0.2% Triton X-100 then incubated overnight with the gephyrin antibody/nanobody pre-mix and a primary antibody against the vesicular GABA transporter VGAT, which was used to confirm synaptic localizations. The corresponding secondary antibody for the VGAT primary antibody was added for 1 h at room temperature the next day followed by a 10-min post-fix. Dishes were stored in PBS at 4°C for up to 2 weeks prior to image collection.

Single molecule localization imaging was performed on an Olympus inverted microscope using a 100 × TIRF oil-immersion objective (1.5 NA). The microscope was equipped with a super-resolution Abbelight 360 SAFe dual-camera (Hamamatsu Fusion) system. The incident angle was manually adjusted for Highly Inclined and Laminated Optical (HILO) illumination to achieve brightest blinking signals. Built-in TrueFocus Red Z drift was used to maintain stability in the z-focal plane throughout image collection. Imager strands diluted to their final concentration (~1–2 nM) in PBS supplemented with 500 mM NaCl were added to prepared neuron dishes. Prior to imaging, a snapshot was taken with the 488 nm laser to identify VGAT clusters. 30,000 frames were then collected at 100 ms exposure with excitation using the 561 and 640 nm lasers. On the same day, separate dishes coated with TetraSpeck beads (Invitrogen) were imaged for 100 frames at 100 ms exposure to facilitate channel alignment during analysis.

Single molecule processing and analysis was performed using procedures similar to those previously described (Schnitzbauer et al., 2017; Chen et al., 2020). Image files were converted from .tif to .raw format in FIJI using the plugin raw-yaml exporter (https://github.com/jungmannlab/imagej-raw-yaml-export) to allow further processing in Picasso (https://github.com/jungmannlab/picasso). Picasso: Localize and Picasso: Filter were used to identify and refine localizations for each channel. Drift correction was then performed in Picasso: Render, where localizations persisting for more than one frame were linked. For each neuron, exported localizations from the two channels were then combined in Excel to facilitate import into MATLAB. Synapses were manually selected based on colocalization with VGAT, significant overlap of γ2-GABA_AR and gephyrin, high local protein density, and a size of ~100–800 nm. Selected synapses were filtered using the MATLAB function DBSCAN according to the following parameters to remove background localizations outside of the synapse boundary: γ2-GABA_AR epsilon = 40 nm, minimum points = 5; gephyrin epsilon = 30 nm, minimum points = 5. Areas of high local protein density (subsynaptic domains, SSDs) were then analyzed in MATLAB as previously described (Chen et al., 2020; Anderson et al., 2023). Briefly, identification of SSDs was based on having a local protein density greater than a specified threshold determined by comparison to a randomized cluster with bounding areas created using the alphaShape function (alpha radius: 150 nm). Enrichment index was defined as the average local density of protein a within a 60 nm range from an SSD peak of protein b, as previously described (Chen et al., 2020; Dharmasri et al., 2024).

Following 7-day treatment with vehicle or 1 μM DZP, DIV 20–22 neurons were rapidly washed with DPBS then immediately fixed for 10 min in PBS containing 4% PFA and 4% sucrose. For surface staining of γ2-GABA_ARs, neurons were blocked for 30 min then incubated under non-permeabilized conditions with primary antibodies overnight at 4°C. Permeabilization was performed after washing the next day by 10-min incubation with blocking solution containing 0.2% Triton X-100. This was followed by overnight intracellular staining for GAD65 at 4°C. Neuron coverslips were washed the next day, then incubated with secondary antibodies for 1 h at room temperature before mounting.

PLA is a highly sensitive technique for detecting protein-protein interactions or protein modifications. Proximity ligation (PL) signals are produced when two oligonucleotide-coupled secondary antibodies (PL probes) are within close proximity (< 40 nm), resulting in oligonucleotide hybridization that is then amplified and visualized by confocal microscopy as discrete, quantifiable dots (Weibrecht et al., 2010). For in situ PLA experiments, 7-day vehicle- or DZP-treated neurons were collected at DIV 21 by rapid washing in DPBS followed by immediate fixation in PBS containing 4% PFA and 4% sucrose for 10 min. Neurons were then permeabilized for 10 min in 0.2% Triton X-100. PLA was performed according to the manufacturer's protocol using the NaveniFlex Cell MR Atto647N kit (Navinci Diagnostics, Sweden). In brief, coverslips were blocked in kit-supplied blocking solution for 1 h in a humidity chamber at 37°C then incubated with primary antibodies overnight at 4°C. Oligonucleotide-conjugated secondary antibodies (Navenibodies) were added the next day for 1 h in a humidity chamber at 37°C, followed by washing and incubation in a ligase solution to permit hybridization of proximal Navenibodies. Subsequent addition of a polymerase solution containing fluorescently-labeled oligonucleotides promoted rolling circle amplification. Next, overnight counterstaining was performed with primary antibodies against microtubule-associated protein 2 (MAP2), to facilitate visualization of neuronal dendrites, and the inhibitory presynaptic marker VGAT, to identify synaptic signals. This was followed by secondary antibody incubation and DAPI nuclear staining.

Fixed images were acquired using a Nikon A1 Confocal microscope equipped with a 60 × oil-immersion objective (NA 1.49) at a zoom of 2 × with sequential laser scanning. Image acquisition and laser settings were kept consistent within each culture and between treatment groups with the researcher blinded to the experimental conditions before data collection and throughout data analysis. Data were analyzed using NIS Elements AR 5.30.05 Software (Nikon, NY) with binary thresholding. For IF experiments, synaptic and extrasynaptic receptor quantification was performed as previously described (Nuwer et al., 2023). Briefly, synaptic γ2-GABA_AR signal was determined by binary intersection of the surface γ2-GABA_AR and GAD65 thresholds, while extrasynaptic γ2-GABA_ARs were defined by subtraction of the synaptic γ2-GABA_AR threshold from the surface receptor threshold. Prior to subtraction, the synaptic receptor binary threshold was dilated once. For each neuron, three 10 μm dendritic regions of interest (ROIs) were collected to analyze each binary threshold, with measurements of the number of clusters, binary area, mean intensity, and sum intensity exported for further analysis. The values of the three ROIs per cell were averaged prior to compiling. For PLA analysis, bright circular proximity ligation (PL) signals having a typical diameter of 0.50 μm, in agreement with the manufacturer-defined size of typical PL signals, were identified using Bright Spot Detection. Manual exclusion was used sparingly to remove non-specific signals that were not localized to any visible cell processes. Synaptic PL signals were defined using the binary operation “Having,” which isolated PL spot signals containing any pixels overlapping with the VGAT threshold. Whole field and synaptic measurements were exported for further analysis. The number of extrasynaptic PL signals was computed manually in Excel by subtraction of the number of synaptic PL signals from the total (whole-field) number of PL signals. Fluorescence intensity values for IF experiments, or PL signal measurements for PLA experiments, were normalized to the vehicle average for each independent culture.

Surface biotinylation experiments were performed as previously described (Nuwer et al., 2021). Briefly, 7-day vehicle- or DZP-treated neurons were rapidly washed twice with DPBS supplemented with 1 mM CaCl₂ and 0.5 mM MgCl₂. Dishes were then incubated with 0.5 mg/mL of cell-impermeant EZ-Link Sulfo-NHS-SS-Biotin (Thermo Fisher) for 15 min at 4°C. Excess biotin was quenched by three washes with 100 mM glycine followed by one wash in DPBS. Neurons were then lysed in RIPA containing 50 mM Tris-HCl at pH 8.0, 150 mM NaCl, 1% Igepal, 0.5% sodium deoxycholate, 0.1% SDS, 1 mM EDTA, 2 mM sodium orthovanadate, 10 mM NaF, and protease inhibitor cocktail (Sigma P8340). Lysates were sonicated, solubilized for 15 min at 4°C, then centrifuged (13,000 rpm, 15 min, 4°C) to remove cell debris. After quantifying protein concentrations by BCA Protein Assay (Thermo Fisher), equal amounts of protein were incubated with NeutrAvidin UltraLink Resin (Thermo Fisher) for 90 min at 4°C with rotation. This was followed by three washes with RIPA supplemented with 500 mM NaCl and elution of isolated biotinylated surface proteins with SDS loading buffer and heating (55°C, 10 min). Surface and total protein fractions were resolved by SDS-PAGE, with biological replicates per culture loaded into separate lanes. Proteins were then transferred overnight to supported nitrocellulose membrane (Bio-Rad). Membranes were incubated with primary antibodies overnight at 4°C. After washing with TBS supplemented with 1% Tween 20 (TBST) the next day, HRP-coupled secondary antibodies were added for 1 h at room temperature followed by chemiluminescent visualization. Analysis was performed in Image Lab 6.0 (Bio-Rad) using the volume tool to quantify immunoreactivities with global background subtraction. Within each independent culture, biological replicates were normalized to the vehicle-treated average. The absence of GAPDH signal in the surface fraction was used to confirm surface-specific labeling.

Fractionation experiments were performed as previously described (Goebel-Goody et al., 2009; Lorenz-Guertin et al., 2023). Neurons were treated with vehicle or DZP for 7 days (~DIV 14–21) then lysed in sucrose buffer containing (in mM): 320 sucrose, 10 Tris-HCl, 1 EDTA, 2 Na₃VO₄, 10 NaF, and protease inhibitor cocktail (Sigma P8340). An initial slow-speed centrifugation (1,000 × g, 10 min) was performed to remove nuclear debris, and a small amount of supernatant (S1) representing the total fraction was set aside for downstream analysis. Subsequent centrifugation of S1 (15,000 × g, 30 min) generated a cytosolic fraction (supernatant S2) and crude membrane fraction (pellet P1). P1 was resuspended in 496 μL of H₂O containing phosphatase and protease inhibitors and incubated on ice for 15 min followed by addition of 3.75 μL of 1 M HEPES solution and another 15 min incubation. Samples were then spun at high speed (25,000 rpm, 20 min; Beckman Coulter Optima Max-E Ultracentrifuge), and the supernatant was discarded. The pellet (P2) was resuspended in sucrose buffer containing Triton X-100 (final concentration, 0.5%) and spun for 60 min at 53,000 rpm. The resulting Triton-insoluble pellet (P3), defined as the synaptic fraction, was resuspended in sucrose buffer and sonicated. SDS was then added (final concentration, 1%) to facilitate protein solubilization. The Triton-soluble supernatant (S3), defined as the extrasynaptic fraction, was concentrated by overnight incubation with 4 × volumes of acetone at −20°C. The resulting precipitate was isolated by centrifugation (15,000 × g, 10 min), resuspended in sucrose buffer, sonicated, and SDS added to a final concentration of 1% to solubilize proteins. Fractions were frozen at −80°C until downstream analysis. All steps were performed on ice, and all centrifugations were at 4°C. Pellets were rinsed twice between centrifugation steps with sucrose buffer containing inhibitors to minimize potential contamination between fractions. Protein concentrations for each fraction were determined by BCA Protein Assay (Thermo Fisher). Equal amounts of protein were resolved by SDS-PAGE and transferred overnight to supported nitrocellulose membrane (Bio-Rad). Membranes were then processed and analyzed as in section 2.8.

Neurons expressing γ2^pHFAP and mScarlet-Gephyrin.FingR were treated with vehicle or 1 μM DZP for 7 days then subjected to live-cell FRAP studies. Hippocampal neurons were used due to their improved longevity over cortical neurons following transfection. mScarlet-gephyrin.FingR is a transcriptionally controlled fibronectin intrabody generated with mRNA display (FingR) that selectively binds endogenous gephyrin without impacting protein levels or synaptic architecture (Gross et al., 2013; Bensussen et al., 2020). The γ2^pHFAP subunit construct has an extracellular pH-sensitive pHluorin tag, allowing surface-specific fluorescence, and a fluorogen-activating peptide (FAP) tag that binds malachite green (MG) dyes with high specificity. γ2^pHFAP assembles with endogenous subunits into receptors that show normal GABA response, DZP potentiation, and trafficking (Lorenz-Guertin et al., 2017, 2019). For live imaging, neurons were rapidly washed with, then transferred to, Hepes-buffered saline (HBS) imaging solution containing (in mM): 135 NaCl, 4.7 KCl, 10 Hepes, 11 glucose, 1.2 MgCl₂, and 2.5 CaCl₂ (adjusted to pH 7.4 with 1 N NaOH). To confirm synaptic localization of mScarlet-gephyrin.FingR clusters, live neurons were first incubated with CypHer5E-labeled VGAT for 1–2 h to allow uptake into recycling vesicles. Experiments were performed using a Nikon A1 Confocal microscope with a 60 × oil-immersion objective (NA 1.49) at 2 × zoom. Stage and objective heaters were set to 37°C throughout the imaging period. Following an initial acquisition phase, 4–6 synaptic regions and 1 extrasynaptic region per neuron were subjected to photobleaching for 1 min using the 488 and 561 lasers at 25% power. 10 nM MG-βTau was added to the dish immediately after photobleaching to re-identify surface γ2^pHFAP clusters as previously described (Lorenz-Guertin et al., 2019). Images were then taken every 2 min for the next 30 min to monitor fluorescence recovery. γ2^pHFAP signal was considered synaptic by colocalization with bright clusters of mScarlet-Gephyrin.FingR. Time series alignment was performed before analysis to correct for drift during image collection. Fluorescence recovery was calculated as previously described (Jacob et al., 2005) according to the following equation: (F_t – F₀)/(F_i – F₀), where F₀ is the fluorescence intensity within each ROI immediately after photobleaching, F_i is the average fluorescence intensity prior to photobleaching, and F_t is the measured fluorescence at each time point following bleaching.

Statistical analysis and graphical representation of data were performed using GraphPad Prism 10.3.1. Data were assessed for normality using D'Agostino & Pearson, Anderson-Darling, Shapiro-Wilk, and Kolmogorov-Smirnov tests. For data that passed the normality tests, two-tailed unpaired t-tests were performed to compare vehicle- vs. DZP-treated groups; otherwise, two-tailed Mann-Whitney tests were conducted. Outliers were identified using Grubbs' (α = 0.05) or ROUT (Q = 1%) and removed as appropriate. All data are presented as mean ± standard error of the mean (SEM) unless otherwise stated. Additional information on specific statistical analyses can be found in the respective figure legends or Supplementary Table S1.

We first established a cultured neuron model of tolerance to evaluate the impact of chronic BZD treatment on basal inhibition and BZD potentiation. BZD binding in the presence of GABA stabilizes the pre-activation receptor conformation and increases the frequency of channel opening (Gielen et al., 2012; Mozrzymas et al., 2007), thus enhancing current amplitude and prolonging inhibitory currents (higher tau of decay, τ_decay). Following a 7-day treatment with either vehicle (0.1% DMSO) or 1 μM DZP, whole-cell recordings were performed in primary cortical neurons to measure miniature inhibitory postsynaptic currents (mIPSCs). In agreement with our prior report (Lorenz-Guertin et al., 2023), mIPSC parameters were unaltered after long-term BZD treatment (Figures 1Ai, B), indicating preservation of inhibitory synapse function. To next assess BZD potentiation, we acutely applied 1 μM diazepam to 7-day vehicle- and DZP-treated neurons (Figures 1Aii, C) and quantified the corresponding potentiation of mIPSC amplitude (Figures 1D, E) and τ_decay (Figures 1F, G). As expected, acute diazepam application to 7-day vehicle-treated neurons produced a 50% increase in mIPSC amplitude from baseline (Figures 1C–E) and a 75% percent increase in τ_decay (Figures 1C, F, G). In contrast, acute diazepam potentiation of mIPSCs was substantially diminished in 7-day DZP-treated neurons (Figures 1C–G). DZP potentiation of mIPSC amplitude was nearly completely lost, reduced to only ~7% (Figures 1D, E), and potentiation of τ_decay was reduced to ~25% (Figures 1F, G). Thus, these results validate our model of primary neuron culture 7-day treated with DZP as a suitable system for the investigation of neuronal plasticity associated with BZD tolerance.

Modern super-resolution microscopy has revealed synaptic proteins to be heterogeneously distributed into high-density protein clusters called subsynaptic domains (SSDs; MacGillavry et al., 2013; Nair et al., 2013; Specht et al., 2013; Crosby et al., 2019). SSDs facilitate efficient synaptic transmission and are subject to activity-dependent plasticity in response to altered neuronal activity or excitation/inhibition dysfunction (Dani et al., 2010; Specht et al., 2013; Tang et al., 2016; Pennacchietti et al., 2017; Werner et al., 2021; Yang and Annaert, 2021; Garcia et al., 2021). We hypothesized that chronic BZD treatment would disrupt the inhibitory synaptic nanoscale architecture and alter gephyrin and GABA_AR subsynaptic organization. To this end, we employed DNA Points Accumulation for Imaging in Nanoscale Topography (DNA-PAINT), a localization-based super-resolution microscopy technique that allows visualization of endogenous proteins with high spatial resolution (Figure 2A; Schnitzbauer et al., 2017). Using antibodies against gephyrin and an extracellular epitope of the γ2-GABA_AR subunit, we observed that gephyrin localizations were organized into highly concentrated clusters that aligned with vesicular GABA transporter (VGAT) puncta (identifying inhibitory presynaptic terminals) and largely overlapped with γ2-GABA_AR localizations, while smaller clusters of both gephyrin and γ2-GABA_AR were observed extrasynaptically (Figure 2B). Assuming roughly circular SSDs and synapses, the average diameters for γ2-GABA_AR and gephyrin synapses and SSDs (γ2-GABA_AR, SSD: 36–40 nm, synapse: ~210–250 nm; gephyrin, SSD: 73–82 nm, synapse: ~360–410 nm) were within previously reported ranges (Yang and Specht, 2019; Anderson et al., 2023), confirming the validity of our technique.

Chronic DZP treatment resulted in shrinkage of the inhibitory postsynaptic area, reducing the total synapse area of γ2-GABA_AR from ~50 × 10³ nm² to ~35 × 10³ nm² and gephyrin from ~130 × 10³ nm² to ~99 × 10³ nm² (Figures 2C, D). γ2-GABA_ARs reorganized at higher density within this smaller area without overall loss of receptors, as indicated by a significant increase in γ2-GABA_AR synaptic localization density (Figure 2E). Conversely, gephyrin localization density was unchanged, suggesting that chronic DZP treatment reduced total synaptic gephyrin levels (Figure 2E). Consistent with this, gephyrin SSD area was also reduced by DZP treatment (Figures 2F, H). However, 7-day DZP treatment did not alter the number of SSDs per synapse for either gephyrin or γ2-GABA_AR (Figures 2F, G), and γ2-GABA_AR SSD area was also unchanged (Figures 2F, H). These data indicate that chronic DZP treatment triggers a nanoscale redistribution of gephyrin and γ2-GABA_ARs without severely disrupting the inhibitory synaptic architecture. This is further supported by similar SSD localization density (Figure 2I) and SSD/total synapse area ratios (Figure 2J) between vehicle- and DZP-treated neurons for both gephyrin and γ2-GABA_AR. Finally, to determine whether the apparent loss of synaptic gephyrin altered its alignment with GABA_ARs, we calculated the enrichment index for γ2-GABA_AR and gephyrin, which is greater than one when the positioning of two proteins is closely correlated (Chen et al., 2020; Dharmasri et al., 2024). We found that chronic DZP treatment did not substantially alter the enrichment indices (Supplementary Figure S1), consistent with intact gephyrin-GABA_AR synaptic associations.

To investigate potential mechanisms by which chronic DZP treatment reduces synaptic gephyrin, we examined gephyrin phosphorylation at Ser270, which regulates gephyrin cluster size (Tyagarajan et al., 2011, 2013) and is increased after short-term (24 h) DZP treatment (Lorenz-Guertin et al., 2019). Gephyrin phosphorylation was assessed using in situ proximity ligation (PL) assay (PLA), a novel fluorescence-based technique which detects protein modifications with improved sensitivity and reduced non-specific signals over traditional phospho-antibody immunofluorescence (IF). Here, we performed PLA using an anti-gephyrin mAb7a antibody, specific for phospho-Ser270, paired with a total anti-gephyrin (3B11) antibody (Figure 3A). MAP2 and VGAT counterstaining were used to identify neuronal dendrites and inhibitory synapses, respectively (Figure 3B). As a control, we confirmed that minimal PL signal was observed under conditions of either primary antibody alone or with no primary antibodies (Supplementary Figure S2). In 7-day DZP-treated neurons, we observed trends consistent with an increase in the number of whole-field PL signals (Figure 3C; p = 0.0757), the number of synaptic PL signals (Figure 3D; p = 0.0751), and whole-field PL signal intensity (Figure 3E; p = 0.0877), while synaptic PL signal intensity was significantly increased by ~60% after chronic DZP treatment (Figure 3F). These data indicate that a higher proportion of the existing synaptic gephyrin is phosphorylated at Ser270 in neurons following 7-day DZP treatment.

Ser270 phosphorylation increases gephyrin susceptibility to calpain-mediated cleavage and proteolysis (Tyagarajan et al., 2011). Therefore, we next assessed chronic DZP-induced alterations to full-length and cleaved gephyrin expression using a biochemical fractionation technique followed by downstream western blotting (Figure 4A). Integrity of the isolated synaptic membrane, extrasynaptic membrane, and total protein fractions was validated by immunoblotting with several synaptic and extrasynaptic markers (Supplementary Figure S3). Chronic DZP treatment reduced full-length gephyrin expression in the total fraction to only ~80% that of vehicle-treated neurons (Figure 4B). Consistent with our DNA-PAINT analysis (Figure 2), this occurred with a near-significant decrease in synaptic full-length gephyrin (Figure 4B; Veh = 100.0 ± 7.01%, DZP = 75.16 ± 9.64%; p = 0.0559). In contrast, extrasynaptic full-length gephyrin was unchanged (Figure 4B). In line with enhanced Ser270 phosphorylation, we observed three-fold higher levels of cleaved gephyrin in 7-day DZP-treated neurons (Figure 4C). Surprisingly, however, this was restricted to the extrasynaptic membrane fraction (Figure 4C), despite an increase in the cleaved/full-length gephyrin ratio in both the synaptic and extrasynaptic membrane fractions (Figure 4D). Overall, these data are consistent with a reduction in full-length gephyrin mediated by increased proteolytic cleavage after chronic DZP treatment.

As loss of synaptic gephyrin can impair GABA_AR synaptic clustering (Jacob et al., 2005; van Zundert et al., 2005; Yu et al., 2007; Carricaburu et al., 2024), we next used IF to examine γ2-GABA_AR surface expression and subcellular localization after chronic DZP treatment. Following 7-day vehicle or DZP treatment, neurons were fixed and surface stained for γ2-GABA_ARs, then permeabilized and stained for the presynaptic GABA-producing enzyme, GAD65 (Figure 5A). γ2-GABA_ARs were considered synaptic when colocalized with GAD65; otherwise, the signal was considered extrasynaptic. Chronic DZP treatment reduced the dendritic clustering density of synaptic γ2-GABA_ARs (Figure 5B) without loss of GAD65 clusters (Figure 5E), indicating a reduced proportion of inhibitory synapses expressing BZD-sensitive GABA_ARs. DZP treatment also decreased the γ2-GABA_AR area per synapse without altering signal intensity (Figure 5B). This is consistent with similar γ2-GABA_AR numbers contained within a smaller postsynaptic area per synapse, in agreement with our DNA-PAINT results (Figure 2). Concurrent with the loss of synaptic clusters, γ2-GABA_ARs were enriched extrasynaptically to 163% that of vehicle after chronic DZP treatment (Figure 5C). This occurred without change to surface γ2-GABA_AR expression (Figure 5D), which was confirmed by complementary surface biotinylation analysis (Figures 5F, G). These findings therefore suggest that DZP treatment induces a redistribution of synaptic γ2-GABA_ARs to extrasynaptic sites without altering surface expression. In contrast to short-term BZD treatments which promote γ2-GABA_AR internalization and degradation (Nicholson et al., 2018; Lorenz-Guertin et al., 2019), biochemical analysis here additionally revealed similar total protein levels of γ2-GABA_AR subunits in 7-day vehicle- and DZP-treated neurons (Figures 5F, G). Thus, these data are overall consistent with a model of individual synapse-specific losses of BZD-sensitive γ2-GABA_ARs induced by chronic DZP treatment. Given that BZD-insensitive α4-GABA_ARs are elevated in some neurodevelopmental disorders and are associated with BZD-resistant seizures (Talos et al., 2012; Sharma et al., 2021), we also assessed α4-GABA_AR synaptic and total subunit expression in neurons after chronic DZP treatment. However, α4-GABA_AR synaptic levels were similar between vehicle- and DZP-treated neurons, though total protein expression trended upward (Supplementary Figure S4; Veh = 100 ± 10%, DZP = 144 ± 18%, p = 0.0623).

To determine whether GABA_AR accumulation in the extrasynaptic membrane and reduced synaptic clustering is mediated by impaired gephyrin-GABA_AR interactions, we again employed PLA and paired a γ2-GABA_AR antibody with a total gephyrin (3B11) antibody (Figure 6A). PLA was performed under permeabilized conditions and thus included gephyrin-GABA_AR interactions both at the cell surface and intracellularly. As before, MAP2 and VGAT counterstaining was used to identify neuronal dendrites and inhibitory synapses, respectively (Figure 6B). Unexpectedly, chronic DZP treatment produced a near-significant increase in the total number of whole-field gephyrin-GABA_AR PL signals (Figure 6C; p = 0.0561), consistent with enhanced receptor-scaffold associations. Stratifying the PL signals into synaptic or extrasynaptic by colocalization with VGAT revealed similar numbers of synaptic PL signals between vehicle- and DZP-treated neurons (Figure 6D). Surprisingly, however, chronic DZP treatment produced higher numbers of extrasynaptic gephyrin-GABA_AR PL signals (Figure 6E). Therefore, these findings suggest that interactions of gephyrin with BZD-sensitive GABA_ARs are elevated specifically at extrasynaptic sites following chronic DZP treatment.

The preservation of synaptic gephyrin-GABA_AR associations (Figure 6; Supplementary Figure S1) may suggest that synaptic stability is intact even during chronic DZP treatment, which is conversely impaired after 12–24 h DZP exposure (Vlachos et al., 2013; Lorenz-Guertin et al., 2019). On the other hand, extrasynaptic gephyrin interactions with glycine receptors can slow their membrane diffusion (Ehrensperger et al., 2007). Thus, we hypothesized that the increase in extrasynaptic gephyrin-GABA_AR interactions (Figure 6) may similarly slow extrasynaptic γ2-GABA_ARs, potentially facilitating their extrasynaptic accumulation (Figure 5). To assess trafficking dynamics, we performed live-cell FRAP (fluorescence recovery after photobleaching) experiments in hippocampal neurons co-transfected with mScarlet-gephyrin.FingR and γ2^pHFAP constructs.

We first confirmed that the majority of mScarlet-gephyrin.FingR clusters were synaptic by live labeling of inhibitory presynaptic terminals with a fluorescently tagged antibody to VGAT (VGAT CypHer5E; Supplementary Figure S5A), which was added to the neuron dish for 1–2 h to allow uptake into synaptic vesicles. In agreement with previously reported values of ~DIV 21 neurons (Danglot et al., 2003), ~90% of analyzed gephyrin clusters were colocalized with VGAT (Supplementary Figure S5B). Synaptic γ2^pHFAP signals were thus subsequently defined by colocalization with bright clusters of mScarlet-gephyrin.FingR.

Following an initial pre-bleach acquisition phase to establish baseline fluorescence, we photobleached synaptic (Figure 7A) and extrasynaptic (Figure 7C) regions of neurons expressing γ2^pHFAP and mScarlet-gephyrin.FingR. Fluorescence recovery within these regions was monitored every 2 min for the next 30 min. MG-βTau, a cell-impermeable MG dye that is non-fluorescent until FAP binding (Yan et al., 2015), was added immediately after photobleaching to confirm surface expression of the γ2^pHFAP GABA_AR clusters (Supplementary Figures S5C, D). Chronic DZP treatment resulted in higher fluorescence recovery of synaptic mScarlet-gephyrin.FingR (Figures 7A, B) compared to vehicle-treated neurons. This is consistent with a reduction in the population of stable, immobilized gephyrin at synapses and is potentially due to faster gephyrin diffusion and/or increased gephyrin forward trafficking. Despite this destabilization of gephyrin, the synaptic dynamics of γ2-GABA_AR were unchanged by chronic DZP treatment (Figure 7B). Additionally, trafficking of extrasynaptic mScarlet-gephyrin.FingR was unchanged (Figures 7C, D). However, in agreement with our hypothesis, the fluorescence recovery of γ2^pHFAP was significantly lower in extrasynaptic regions of DZP-treated neurons (Figures 7C, D). Because GABA_ARs are primarily exocytosed extrasynaptically (Bogdanov et al., 2006), yet surface γ2-GABA_AR expression was unchanged by DZP treatment (Figure 5), these data are consistent with reduced lateral receptor movements within the extrasynaptic membrane rather than decreased forward trafficking or altered diffusion from synaptic to extrasynaptic sites.

There were several limitations in this study. Firstly, the findings could be strengthened by the addition of experiments in which co-treatment with the BZD-site antagonist flumazenil is used to confirm BZD-specific effects. Secondly, while the literature supports a high level of correlation between gephyrin Ser270 phosphorylation and calpain-mediated cleavage, we do not demonstrate a causal link between these observations. Calpain 1 and 2 (family of 15 calpain protease genes) are highly expressed in the nervous system, share many substrates that regulate synaptic plasticity, but often exhibit opposite functions and have differing levels of calcium requirement (reviewed in Baudry et al., 2023). Calpains also do not recognize a specific sequence or posttranslational modification, and the governing rules of recognition are still unclear (reviewed in García-Trevijano et al., 2023). Thus, multiple areas of future research are needed to define the calpain-dependent gephyrin cleavage mechanism, investigate the role for posttranslational gephyrin modifications (via gephyrin mutants), and study these events in the context of BZD tolerance. Thirdly, DNA-PAINT experiments would be improved by three-dimensional imaging to simultaneously visualize presynaptic active-zone proteins such as RIM, as the current data do not provide insight into trans-synaptic alignment of SSDs after chronic BZD treatment. Finally, while neuronal culture was used here to examine molecular mechanisms with high resolution, future studies will be needed to determine the relevance of these findings in vivo and assess potential sex-specific differences. This is particularly true given that our prior studies of 7-day DZP treatment in mice showed increased γ2-GABA_AR synaptic expression and increased extrasynaptic gephyrin expression without loss of synaptic or total scaffold (Lorenz-Guertin et al., 2023). Differential timelines of neuroadaptations in vitro vs. in vivo are a potential contributing factor, which could be elucidated in future research through the analysis of additional time points of BZD treatment in both systems.