A versatile optical tool for studying synaptic GABA_A receptor trafficking

γ-Aminobutyric acid (GABA) type A receptors (GABA_ARs) are ligand-gated ionotropic Cl⁻ channels responsible for most fast inhibitory neurotransmission in the adult central nervous system (CNS). This inhibition requires binding of the neurotransmitter GABA, which induces ion channel opening, Cl⁻ influx and subsequent membrane hyperpolarization. The majority of GABA_ARs are composed of two α subunits, two β subunits and a γ2 subunit forming a heteropentamer, but considerable subunit heterogeneity can exist (α1–6, β1–3, γ1–3, δ, ɛ, θ, π, ρ1–3) (Luscher et al., 2011). The γ2 subunit (encoded by GABRG2) plays a critical role in GABA_AR function, as it is necessary for receptor synaptic targeting and cluster maintenance (Alldred et al., 2005; Essrich et al., 1998; Rovo et al., 2014; Schweizer, 2003; Sumegi et al., 2012), coassembles with nearly all α and β subunits (Möhler et al., 2001), and is the only subunit characterized to undergo ubiquitylation leading to lysosomal degradation (Arancibia-Carcamo et al., 2009). The degree of γ2-mediated GABA_AR synaptic clustering directly impacts the strength of GABAergic synaptic inhibition and is dynamically regulated by changes in receptor trafficking (Jacob et al., 2008; Luscher et al., 2011). A number of endogenous and pharmacological agents including GABA, neurosteroids, ethanol and benzodiazepines are known to influence receptor trafficking (Calkin and Barnes, 1994; Carver and Reddy, 2013; Hablitz et al., 1989; Kumar et al., 2010; Liang et al., 2007; Shen et al., 2011; Tehrani and Barnes, 1988, 1991; Wan et al., 1997). Despite this knowledge, precise GABA_AR trafficking mechanisms induced by these and other clinically relevant compounds remain underexplored at the molecular level.

Real-time receptor trafficking measurements typically rely on genetically encoded fluorophores to track single protein localization and movement within living cells. In contrast to immunofluorescence-based live-cell techniques, which require reliable antibodies and are restricted to measurements of surface proteins (Arancibia-Carcamo et al., 2006), fluorescent tags allow for identification of a protein from synthesis through degradation. Despite this advantage, traditional fluorophores still remain limited in their ability to spatially resolve surface from internal populations without the use of total internal reflection fluorescence (TIRF)-imaging approaches (Kirchhausen, 2009). This issue led to the generation of pH-sensitive fluorescent proteins, such as pHluorin (pHGFP), which exhibit fluorescence in extracellular environments with alkaline pH (pH ∼7.4), but not in more acidic areas like intracellular vesicles (Sankaranarayanan et al., 2000). However, pHGFP is not without limitations as some discernible signal can be observed in the endoplasmic reticulum (ER) where pH can be roughly 7.2 (Asokan and Cho, 2002). One possible way to resolve protein trafficking occurring at the cell surface from the intracellular space is to use compartment-specific high-affinity labeling techniques. We previously described a labeling method utilizing a GABA_AR subunit genetically tagged with an α-bungarotoxin-binding site, which selectively binds cell-excluded fluorescent bungarotoxin, allowing for selective monitoring of receptor insertion and internalization (Brady et al., 2014). Unfortunately, the inherent fluorescence of Alexa Fluor dye-coupled bungarotoxins necessitates extensive washing after labeling to reduce background signal. Furthermore, bungarotoxins were recently shown to function as antagonists of the GABA_AR, further complicating the use of this reagent (Hannan et al., 2015). To overcome these obstacles associated with GABA_AR imaging, we have employed an innovative paired optical reporter system where two individually non-fluorescent components become highly fluorescent upon binding: a genetically encoded fluorogen-activating peptide (FAP tag) and exogenously applied Malachite Green (MG) dyes. Multiple FAPs are described as antibody single-chain variable fragments (scFvs) that selectively bind MG synthetic dyes with high specificity and affinity (Szent-Gyorgyi et al., 2008). These synthetic dyes are non-fluorescent in solution until bound by their respective FAP and can be modified to have distinct characteristics including cell permeability, pH-sensitivity and various fluorescence properties (Fisher et al., 2014; Grover et al., 2012; Perkins et al., 2017 preprint; Saunders et al., 2012; Zhang et al., 2015). The FAP–dye system offers many advantages for live-imaging: (i) the dyes can be added directly to a culture dish and saturate the target FAP in seconds; (ii) a number of distinct dyes can be used for the same genetically encoded FAP; and (iii) the dyes are highly specific for their target FAP. Moreover, high-affinity MG–FAP binding means that a stable fluorescent module that allows for measurements of receptors undergoing internalization and recycling is formed (Pratt et al., 2015; Szent-Gyorgyi et al., 2013; Yan et al., 2015). Measuring drug-induced changes in FAP-tagged receptor trafficking has proven largely successful (Fisher et al., 2010, 2014; Grover et al., 2012; Pratt et al., 2015; Snyder et al., 2015; Wu et al., 2014) placing this technique at the forefront of pharmacological screening.

In order to design a tool to track synaptic GABA_AR internalization and trafficking, we engineered a γ2 subunit encoding a pHGFP and the fluorogen-activating peptide dl5 (γ2^pHFAP) (He et al., 2016; Saunders et al., 2013; Szent-Gyorgyi et al., 2008, 2013). We find that γ2^pHFAP-containing GABA_ARs are trafficked to the cell surface in both HEK293 cells and primary neurons, and are localized appropriately at synapses. We further demonstrate how this construct can be combined with MG dye derivatives to measure alterations in surface localization and intracellular trafficking through high-resolution confocal microscopy approaches. Finally, using these FAP-based methods, we found that an in vitro seizure model induced rapid loss of dye-labeled synaptic GABA_ARs concomitant with enhanced targeting of internalized receptors to lysosomal compartments, key results that were not detectable when using the pHGFP signal alone. We therefore demonstrate an innovative tool to monitor multistage synaptic GABA_AR trafficking.

We began by introducing the dl5 FAP into a previously characterized N-terminal pHGFP γ2-construct (γ2^pHGFP) that functions comparably to wild-type γ2 (Jacob et al., 2005; Kittler et al., 2000b; Muir et al., 2010; Renner et al., 2012) (Fig. 1A). The FAP will selectively bind MG dyes with distinct biophysical and fluorescence properties. For example, the MG derivative 2-({4-[(2,5-dioxopyrrolidin-1-yl)oxy]-4-oxobutanoyl}amino)ethanesulfonate (MG-BTau) is cell impermeant, binds to dl5 FAP with a high binding affinity (K_d value less than 0.5 nM), and is non-fluorescent until dl5 FAP binding occurs, which causes activation of fluorescence in the far-red spectral region (680 nm) (Larsen et al., 2016; Pratt et al., 2015; Szent-Gyorgyi et al., 2013; Yan et al., 2015) (Fig. 1B). We first examined whether the full-length γ2^pHFAP construct was appropriately expressed in HEK293 cells. GABA_AR β3 subunits (encoded by GABRB3) were co-expressed in these experiments as a β subunit is required for trafficking of receptors to the cell surface (Connolly et al., 1996). Western blot analysis of non-transfected, γ2^pHGFP control and γ2^pHFAP-expressing cells reveal γ2^pHFAP is ∼25 kDa larger than γ2^pHGFP, consistent with the molecular mass of the dl5 FAP (Szent-Gyorgyi et al., 2013) (Fig. 1C). Next, we evaluated surface expression and specificity of MG-BTau dye labeling of GABA_ARs in cells transiently transfected with β3+γ2^pHFAP or β3+γ2^pHGFP subunits. Addition of MG-BTau to a live-cell culture should swiftly and selectively label surface GABA_ARs containing γ2^pHFAP. To test this dye-based labeling approach, living cells plated on glass-bottom Mattek dishes were pulse-labeled with 250 nM MG-BTau in HEPES-buffered saline (HBS) for 1 min at room temperature, washed five times with saline, and then immediately imaged by confocal microscopy. We found that HEK293 cells expressing γ2^pHFAP demonstrate selective receptor surface labeling by MG-BTau, while γ2^pHGFP control cells show no cell membrane MG-BTau fluorescence (although occasional low-level background MG-BTau binding can occur with cellular debris; Yan et al., 2015) (Fig. 1D). The majority of cells express heteromeric receptors composed of β3+γ2 subunits, although a few cells express the γ2 subunit alone, leading to subunit retention in the ER and no MG-BTau signal in γ2^pHFAP cells (Fig. 1D, yellow arrowheads). These results highlight two key advantages of FAP–fluorogen labeling over pHGFP alone: (1) a higher signal-to-noise ratio, and (2) the removal of pHGFP intracellular background fluorescence generated by ER-localized receptors where pH can be roughly 7.2 (Asokan and Cho, 2002).

We further investigated the ability of γ2^pHFAP to assemble into functional heteromeric α2β3γ2 GABA_ARs with normal Cl⁻ channel activity in response to the endogenous agonist GABA, which binds between the α and β subunits. Patch clamp recordings were used to determine the GABA dose–response curve for HEK293 cells expressing receptors composed of α2β3γ2^pHGFP or α2β3γ2^pHFAP. Concentration–response analysis revealed these receptors exhibit similar GABA EC₅₀ values (α2β3γ2^pHGFP, 22.0±6.2 µM; α2β3γ2^pHFAP, 14.3±8.9 µM; P>0.05) and Hill coefficients (α2β3γ2^pHGFP, 1.14±0.26; α2β3γ2^pHFAP, 0.90±0.29; mean±s.e.m., P>0.05) (Fig. 2A). The EC₅₀ values are consistent with previously reported values for receptors containing this subunit composition (Mortensen et al., 2011). To directly test whether addition of the FAP tag altered drug binding and functional properties of the γ2 subunit, we compared the potentiation response induced by the benzodiazepine diazepam (DZ), a positive GABA_AR allosteric modulator that binds at the interface of γ2 and specific α subunits (α1, 2, 3 or 5) (Fig. 2B) (Möhler et al., 2001). We found no significant difference in the DZ potentiation of the EC₂₀ GABA response between α2β3γ2^pHGFP- and α2β3γ2^pHFAP-expressing cells (Fig. 2C, fold DZ potentiation α2β3γ2^pHGFP=3.1±0.5, α2β3γ2^pHFAP=2.8±0.6; P>0.05). Finally, we tested whether binding of a MG dye to the γ2^pHFAP subunit would interfere with receptor function. We compared the DZ potentiation response of γ2^pHFAP before and after co-application of MG-BTau. Binding of MG-BTau to the receptors did not alter the response to DZ (Fig. 2D, fold DZ potentiation γ2^pHFAP without dye=4.0±0.4, γ2^pHFAP with dye=3.9±1.0; mean±s.e.m., P>0.05). Taken together, these results suggest that GABA_ARs incorporating γ2^pHFAP maintain normal receptor function and responsiveness to GABA and DZ in the presence of MG dye.

To confirm that γ2^pHFAP is fully expressed in cultured neurons, we compared non-transfected rat cortical neurons to those transfected with γ2^pHFAP or γ2^pHGFP control at plating (Fig. 3). Neurons were lysed 14 days after transfection and subsequently immunoblotted with anti-GFP antibody (Fig. 3A), revealing substantial expression of γ2^pHFAP and γ2^pHGFP, as seen in HEK293 cells (Fig. 1C). We then tested whether neurons with γ2^pHFAP GABA_ARs could selectively bind and activate MG-BTau dye fluorescence. Cortical neurons at 12 days in vitro (DIV) expressing γ2^pHGFP or γ2^pHFAP were pulse-labeled with 250 nM MG-BTau dye for 1 min at room temperature, then immediately washed and used for live-cell imaging. Only γ2^pHFAP-expressing neurons demonstrate dye activation, and this signal colocalizes with the surface synaptic clusters marked by pHGFP (Fig. 3B). Furthermore, it is evident that MG-BTau labeling in neurons also has a higher signal-to-noise ratio than pHGFP due to, first, the continually generated pHGFP signal of diffuse newly inserted extrasynaptic γ2^pHFAP receptors that are not yet clustered at synapses, which contrasts with the small MG-BTau pulse-labeled extrasynaptic population (if the MG dye was continually present rather than being washed away, its extrasynaptic signal would be more similar to pHGFP) and, second, the low but observable pHGFP background from ER-resident γ2^pHFAP subunits, whereas there is no ER signal from MG-BTau labeling. These data establish MG dyes as a selective label for synaptic γ2^pHFAP-containing GABA_AR populations in living primary cortical neurons.

Synapse formation and receptor clustering are critical for neuronal development and regulation of inhibitory neurotransmission. To determine whether γ2^pHFAP constructs cluster normally at GABAergic synapses in mature neurons, we transfected cortical neurons at DIV 0 with γ2^pHGFP control or γ2^pHFAP and fixed the cells for immunofluorescence studies at DIV 15. We assessed receptor synaptic localization by measuring colocalization of the pHGFP-tagged receptors with the vesicular GABA transporter (VGAT), a presynaptic marker of GABAergic synapses (Fig. 3C). Expression of γ2^pHFAP did not alter presynaptic GABAergic input as indicated by unchanged VGAT levels (intensity) and area compared to that in the control (Fig. 3D). Additionally, the size of postsynaptic GABAergic synapses was not significantly different between constructs, although the mean intensity of γ2^pHGFP synapses was 20.7% (95% confidence interval) greater than the intensity of γ2^pHFAP, likely due to higher overall expression of the control subunit. These results indicate that the full-length γ2^pHFAP construct is expressed, assembles with endogenous subunits into receptors, traffics to synapses and does not disturb neuronal development.

Having validated and established the application of FAP technology to generate a synaptic GABA_AR reporter system, we first explored the experimental flexibility afforded by the distinct characteristics of available MG dyes to study receptor trafficking. First, we examined whether MG-BTau-labeled surface receptors could be identified in endosomal pathways following internalization. DIV 12–13 γ2^pHFAP neurons were pulse-labeled with 100 nM MG-BTau, transferred to 10°C or 37°C HBS for 30 min, and were then fixed and immunostained for the early endosome marker EEA1 (Fig. 4A). Neurons maintained at 37°C demonstrated greater mean intensity and area of MG-BTau colocalized at EEA1 vesicles compared to those kept at 10°C to inhibit internalization, indicating MG-BTau labeling can also be used to track internalized GABA_AR pools (Fig. 4B).

We next investigated the adaptability of γ2^pHFAP with other MG dye applications in live-imaging intracellular trafficking assays. The Cy3pH(S/SA)-MG dye is a dichromophore consisting of a pH-sensitive Förster resonance energy transfer (FRET) donor Cy3pH molecule and an acceptor MG (Fig. 5A; Fig. S1) (Perkins et al., 2017 preprint). Excitation of Cy3pH (561 nm excitation) results in highly efficient FRET to MG and emission is observed at 680 nm, while direct stimulation of MG (640 nm excitation) also results in 680 nm emission, allowing calculation of an emission ratio of MG₅₆₁:MG₆₄₀ (Fig. 5A). Increasingly acidic environments such as endosomes and lysosomes lead to protonation of Cy3pH and increased fluorescence intensity, enhancing FRET, and the ratio of MG₅₆₁:MG₆₄₀ (Grover et al., 2012; Perkins et al., 2017 preprint; Szent-Gyorgyi et al., 2010; Yushchenko et al., 2012). The pH sensor dye is cell impermeant like MG-BTau and thus allows for selective labeling of surface γ2-containing GABA_ARs. To characterize the pH sensitivity of Cy3pH(S/SA)-MG in our γ2^pHFAP neuronal system, we first performed live-cell imaging perfusion experiments using different pH solutions to simulate receptor progression through an increasingly acidic endolysosomal pathway, as similarly performed in other cell culture methods with Cy3pH(S/SA)-MG (Perkins et al., 2017 preprint). Neurons were pulse-labeled with pH sensor dye and then first imaged with pH 7.4 physiological saline, followed by pH 6.8 and pH 4.8 solutions, demonstrating a clear shift in the ratio of MG₅₆₁:MG₆₄₀ across treatments (Fig. 5B). Quantification of mean±s.e.m. MG₅₆₁:MG₆₄₀ ratios at surface synaptic cluster sites revealed pH 7.4 (0.58±0.01), pH 6.8 (0.69±0.01) and pH 4.8 (0.99±0.02) conditions gave significantly different values from one another (Fig. 5C). These data demonstrate that Cy3pH(S/SA)-MG-labeled γ2^pHFAP receptors can be readily identified in environments of different acidities based on their MG₅₆₁:MG₆₄₀ ratio.

Following pH characterization of Cy3pH(S/SA)-MG, we then utilized this dye to identify the localization of internalized γ2^pHFAP receptors in the endosomal-lysosomal system. Acidity increases as intracellular receptors undergo transition from early endosomal/recycling pathways to late endosomal/lysosomal pools (Fig. 6A). We therefore anticipated differences in mean MG₅₆₁:MG₆₄₀ ratios between these vesicular populations. To examine this effect, we co-transfected γ2^pHFAP neurons with the early endosome marker EEA1–GFP for live-imaging experiments. At DIV 12–13, 10 nM Cy3pH(S/SA)-MG was added directly to the cell culture medium and neurons were incubated for 30 min at 37°C prior to imaging. Cells were first perfused with pH 7.4 HBS to acquire MG₅₆₁:MG₆₄₀ data, then with pH 6.4 solution to eliminate surface γ2^pHFAP (pHGFP) signal (Fig. 6B, orange stars) to isolate intracellular EEA1–GFP for image analysis. Cy3pH(S/SA)-MG vesicles colocalized with EEA1–GFP (Fig. 6B, yellow arrows) demonstrated a mean MG₅₆₁:MG₆₄₀ ratio of 0.824±0.012, while all other identified vesicles not-associated with EEA1–GFP (Fig. 6B, white triangles) displayed a significantly higher ratio of 0.918±0.016 (Fig. 6C, other). The Cy3pH(S/SA)-MG found in non-EEA1 intracellular sites typically was present at locations with the lowest pH and was in comparably larger vesicular bodies, suggesting lysosomal compartments. These results indicate Cy3pH(S/SA)-MG-labeled γ2^pHFAP receptors can be used to decipher the localization of internalized GABA_AR pools along the endosome-lysosome axis.

Next, we wanted to elucidate whether Cy3pH(S/SA)-MG could be used to monitor the constitutive endolysosomal trafficking of GABA_ARs by using conditions that limit internalization. We compared γ2^pHFAP neurons pulse-labeled with 100 nM Cy3pH(S/SA)-MG dye either immediately after washing (t0) or 30 min later (t30) with or without the dynamin inhibitor dynasore (80 µM) at 37°C. Importantly, GABA_ARs have been previously characterized to undergo constitutive dynamin-dependent clathrin-mediated endocytosis (Herring et al., 2003; Kittler et al., 2000a). Images of neurons allowed 30 min of trafficking show enhanced dye-labeled GABA_AR accumulation within the cell body compared to t0 and to t30 with dynasore (Fig. 6D). The number of GABA_AR-positive vesicles inside the neuronal cell body was measured by using the spot detection feature in NIS Elements. Quantification confirmed that the total number of vesicles identified at t30 is 78.7±20% (mean±s.e.m.) greater than at t0 and that vesicle internalization is blocked by dynasore co-treatment (Fig. 6E). To better understand the receptor population distribution along the endosomal-lysosomal trafficking axis, we examined the respective MG₅₆₁:MG₆₄₀ ratio for each vesicle and plotted this data as a histogram (Fig. 6F). This analysis revealed that the increase in vesicles found at t30 were primarily at lower MG₅₆₁:MG₆₄₀ ratios, suggesting early endosomal trafficking. Moreover, the vesicle median MG₅₆₁:MG₆₄₀ ratios were significantly different between all conditions (t0=1.31; t30=0.767; and t30 with dynasore=0.826). Interestingly, we found that although dynasore reduced the number of internalized receptors detected back to t0 levels, dynasore-treated neurons demonstrated a unique vesicular MG₅₆₁:MG₆₄₀ ratio profile (Fig. 6E). Dynasore inhibition of clathrin-coated pit formation and endocytosis generates half formed ‘U’ and ‘O’ shaped pits/vesicles associated with the plasma membrane (Macia et al., 2006). The few vesicles present in t30 with dynasore neurons (Fig. 6D) are likely to be these structures undergoing partial acidification during the 30 min experimental period. In summary, γ2^pHFAP and the Cy3pH(S/SA)-MG dye reveal significant constitutive clathrin-dependent endocytosis of GABA_ARs that favors early endosomal pathways.

We last set out to investigate whether the γ2^pHFAP–dye system could be used to measure pharmacologically induced changes in GABA_AR trafficking in living neurons. Prolonged exposure of the GABA_AR antagonist bicuculline is proconvulsant due to sustained dampening of network inhibition, leading to seizure-type activity in vitro and in vivo (Chauvette et al., 2016; Colombi et al., 2013; Hongo et al., 2015; Khalilov et al., 1997). Multiple hyperexcitable neuronal states have previously been reported to enhance γ2-containing GABA_AR internalization and reduce total surface levels at 1 h post induction (Goodkin et al., 2008, 2005; Naylor et al., 2005; Terunuma et al., 2008). We investigated whether the γ2^pHFAP construct could be used to simultaneously examine multiple stages of receptor trafficking including receptor surface, synaptic and lysosomal levels following a bicuculline-induced seizure paradigm. At DIV 12–14 γ2^pHFAP neurons were pulse-labeled with MG-BTau dye and then returned to conditioned medium with or without 50 µM bicuculline at 37°C for 1 h. 50 nM LysoTracker was added 30 min prior to the end of treatment to identify association of receptors with lysosomes. Representative images indicate that MG-BTau labels synaptic GABA_AR clusters on the surface of dendrites as seen by colocalization of MG-BTau (blue) and pHGFP (green) (Fig. 7A,B). MG-BTau also reveals internalized receptors within the cell body in lysosomes (Fig. 7C, Lysotracker in red). These data demonstrate that the binding of MG-BTau to γ2^pHFAP GABA_ARs and its resulting fluorescence is stable even in very low pH environments, such as lysosomes, consistent with previous findings using different FAP-tagged receptors colocalized with LysoTracker in cell culture (Grover et al., 2012). Moreover, we detect sustained MG-BTau signal and labeling of surface γ2^pHFAP during extracellular pH 7.4 to 4.8 solution exchange experiments (Fig. S2). Image analysis uncovered no significant difference in total surface expression of γ2^pHFAP between DMSO control and bicuculline-treated cells when measuring the pHGFP signal (Fig. 7D). There was a trend towards a decrease in synaptic levels of γ2^pHFAP in bicuculline-treated neurons (75±13% of control; mean±s.e.m.), as determined by pHGFP cluster fluorescence, but this was not significant. In contrast, Fig. 7E shows that bicuculline treatment reduced total and synaptic MG-BTau signal by 41±10% and 67±8%, respectively, indicating the population of pulse-labeled γ2^pHFAP receptors had decreased. In support of enhanced receptor turnover, we found that bicuculline treatment increased association of labeled receptors with lysosomes by 107±41% compared to control. These findings suggest that bicuculline-induced seizure activity leads to augmented GABA_AR synaptic turnover, lysosomal targeting, and a compensatory increase in new (non-recycled) GABA_AR insertion to mitigate this response. These results robustly demonstrate the versatility of the γ2^pHFAP–dye system and its ability to measure numerous trafficking events to address complex biological questions.

Live-cell receptor tracking approaches offer critical information by revealing real-time alterations in protein trafficking. Here, we describe a flexible paired GABA_AR γ2 subunit optical reporter system that can be used to monitor multi-stage receptor trafficking. Compared to previously available methods that require reliable antibodies, conventional fluorophore tagging or fluorescent α-bungarotoxin labeling, the γ2^pHFAP–dye system can allow for simultaneous monitoring of surface, synaptic and intracellularly trafficking GABA_ARs in real-time. The strength of this method lies in the ability to simply switch FAP-compatible MG dyes to address the specific experimental question of interest.

We applied the pH sensor dye Cy3pH(S/SA)-MG to monitor internalization and constitutive endosomal and lysosomal trafficking of GABA_ARs in neurons. We found that Cy3pH(S/SA)-MG-labeled surface γ2^pHFAP receptors can be tracked to small EEA1-positive early endosomes (MG-BTau-labeled receptors were also identified at EEA1 endosomes; Fig. 4) and also to larger vesicular structures with low pH (likely lysosomes), generating high MG₅₆₁:MG₆₄₀ ratios (Fig. 6B,C). Synaptic GABA_AR endocytosis occurs primarily in a dynamin-dependent manner (Herring et al., 2003) and can be identified within 30 min when using an antibody-feeding approach (Kittler et al., 2000a). Our results in Fig. 6 demonstrate rapid internalization of receptors in endosomal/lysosomal pathways on this same time scale, a process reduced by the dynamin-inhibitor dynasore. As a result of the vesicle-level pH-sensitive measurement afforded by this tool, we were able to further generate a histogram plot identifying differences in the trafficking stages of γ2^pHFAP-positive vesicles. These analyses suggest that internalized receptors favor early endosomal pathways, a finding supported by a previous study that determined that over 70% of internalized GABA_ARs are recycled back to the cell surface within 1 h (Kittler et al., 2004). Furthermore, dynasore-treated neurons displayed vesicles with a distinct MG ratio profile, possibly resulting from acidification of endocytic pits/vesicles not released from the plasma membrane, non-dynamin-dependent endocytosis of receptors or off-target effects of dynasore. GABA_ARs undergo both clathrin-dependent and -independent endocytosis pathways in heterologous cells (Cinar and Barnes, 2001), with both mechanisms being recently identified in neurons for glutamateric α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors (Glebov et al., 2015; Zheng et al., 2015). Dynasore has also been shown to disrupt lysosomal fission by inhibiting dynamin-2 (Schulze et al., 2013) as well as having other non-specific dynamin-independent effects (Park et al., 2013; Preta et al., 2015) that could result in different vesicle MG₅₆₁:MG₆₄₀ ratio profiles. Further studies investigating additional time points are necessary to identify the rate, equilibrium and overall population dynamics of constitutive GABA_AR internalization.

To test the utility of γ2^pHFAP in detecting pharmacologically induced changes in GABA_AR trafficking, we exposed neurons to a bicuculline-induced seizure paradigm following a MG-BTau pulse-labeling protocol (Fig. 7). This unique dye labeling approach readily detected enhanced synaptic receptor turnover rates and lysosomal targeting that were not detectable by pHGFP fluorescence alone. Previous work has shown that more aggressive seizure protocols relying on depolarization via high external K⁺ and/or agonist application targeting highly Ca²⁺-permeable N-methyl-D-aspartic acid (NMDA) receptors reduces total surface GABA_ARs in cultured neurons (Goodkin et al., 2008). It is likely that the mild seizure paradigm used here does not affect total surface receptor levels, but instead enhances surface turnover that is offset by increased receptor insertion rates. In support of this argument, another recent study (Chaumont et al., 2013) found that a 1 h treatment with GABA_AR antagonists did not lead to a reduction in surface GABA_AR levels.

This paper lays the groundwork for future investigations of GABA_AR trafficking with the γ2^pHFAP sensor. Importantly, this technique can be further extended toward pharmacology-focused efforts in high-throughput screenings (Fisher et al., 2014; Snyder et al., 2015), assay development based on flow cytometry (Saunders et al., 2012; Wu et al., 2014) or high-resolution 96-well plate assays (Larsen et al., 2016) and for in vivo protein labeling (Liu et al., 2016; Zhang et al., 2015). In summary, our γ2^pHFAP is the first protein–FAP conjugate characterized in primary neurons, providing a unique tool to monitor multistage GABA_AR trafficking in living cells with relevance both for basic science research and applied pharmacology.

All procedures were approved by the University of Pittsburgh Institutional Animal Care and Use Committee. Cortical neurons were prepared from embryonic day 18 rats (Sprague Dawley rat, Charles River) and nucleofected (Lonza, Switzerland) at plating (Jacob et al., 2005). HEK293 cells were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS; Life Technologies, Carlsbad, CA) and were transfected by nucleofection.

The α2, β3 and pH-sensitive GFP-tagged γ2 subunit (γ2^pHGFP) plasmids have been previously described (Bedford et al., 2001; Jacob et al., 2005; Kittler et al., 2000b; Tretter et al., 2008). The flurogen-activating peptide dl5 (Szent-Gyorgyi et al., 2013) was inserted upstream of pHGFP (γ2^pHFAP) separated by a G-A-P-P-A amino acid linker. The EEA1–GFP construct was Addgene plasmid #42307 (deposited by Silvia Corvera) (Lawe et al., 2000). All constructs were sequenced to confirm the fidelity of final plasmids. The following primary antibodies were used: mouse anti-β-actin (1:2000, A1978, Sigma); rabbit anti-GFP (1:1000, A11122, Invitrogen); rabbit anti-VGAT (1:1000, 131002, Synaptic Systems); mouse anti-EEA1 (1:1000, 610457, BD Biosciences) antibodies. Immunofluorescence secondary antibodies were: goat anti-rabbit-IgG conjugated to Alexa Fluor 641 (1:1000, A21245, Invitrogen); goat anti-mouse-IgG conjugated to Alexa Fluor 405 (1:1000; A31553, Invitrogen).

MG dyes were kindly provided by Dr Alan S. Waggoner, Dr Ming Zhang, Dr Marcel P. Bruchez, and Dr Brigitte F. Schmidt at Carnegie Mellon University, Pittsburgh, PA. The MG dye MG-BTau was synthesized as described previously (Grover et al., 2012; Yan et al., 2015) and the MG dye Cy3pH(S/SA)-MG was prepared by the method of Perkins et al. (2017 preprint). The IUPAC name for Cy3pH(S/SA)-MG is 2-((E)-3-((Z)-3,3-dimethyl-5-sulfoindolin-2-ylidene)prop-1-en-1-yl)-1-(6-((3-(4-((4-(dimethylamino) phenyl)(4-(dimethyliminio)cyclohexa-2,5-dien-1-ylidene)methyl)phenoxy)propyl)amino)-6-oxohexyl)-3,3-dimethyl-5-sulfamoyl-3H-indol-1-ium. Structural, synthetic and analytical details for MG-BTau are as described previously (Larsen et al., 2016; Pratt et al., 2015; Yan et al., 2015).

Primary cortical neurons grown on glass coverslips were fixed at days in vitro (DIV) 13–14. Neurons were permeabilized and stained with anti-VGAT antibody. Images were taken on a Nikon Ti-E A1 confocal microscope equipped with a motorized Z-stage and perfect focus system (PFS) using a 60× oil immersion objective (NA 1.49) at 3× zoom. Data were analyzed by using NIS Elements software (Nikon, NY). Thresholds were set using binary masks to selectively identify brightly fluorescent objects above background (Jacob et al., 2005). Three dendritic 10 µm regions of interest (ROI) were drawn per neuron to measure synaptic colocalization of pHGFP signal with VGAT in γ2^pHFAP characterization studies. γ2^pHGFP control and γ2^pHFAP-expressing neuron pixel intensity and sum area of synapses in µm² was measured and values were normalized to control mean. Fixed EEA1 early endosome studies utilized a 2-min 100 nM MG-BTau dye pulse-labeling protocol in HBS at room temperature. After live-cell MG dye incubation steps, cells were washed five times to remove all unbound dye prior to returning the cells to 10°C or 37°C solution for 30 min followed by fixation and immunostaining. The total pixel intensity of MG-BTau-labeled receptors colocalized with EEA1 was measured within a cell body ROI. Laser settings were held constant across experiments. Image acquisition and data analysis was carried out by a researcher who was blind to the experimental details.

Transfected DIV 12–14 cortical neurons or HEK293 cells were plated on MatTek glass-bottom dishes (Ashland, MA). Imaging was performed in HEPES-buffered saline (HBS): 135 mM NaCl, 4.7 mM KCl, 1.2 mM MgCl₂, 10 mM HEPES, 2.5 mM CaCl₂, 11 mM glucose, pH 7.4 (Jacob et al., 2012). Low-pH HBS (pH 6.4–6.8) was similarly prepared, while pH 4.8 saline was a MES-buffered saline: (135 mM NaCl, 4.7 mM KCl, 1.2 mM MgCl₂, 10 mM MES, 2.5 mM CaCl₂, 11 mM glucose). All dye pulse-labeling steps were performed in HBS at room temperature. Following dye incubation steps, cells were washed five times to remove all unbound dye prior to treatment or imaging. All images were taken at room temperature using a 60× objective at a 3× zoom within 10 min of dye washout. Perfusion assays utilizing different pH saline buffers monitored a single cell per assay, while two or three cells were imaged in each experimental dish for all other experiments. Data analysis was carried out by a researcher who was blind to the experimental details where possible.

Cy3pH(S/SA)-MG dye was used for assays of constitutive trafficking in neurons. Neurons were pre-incubated in DMSO vehicle control (t0, t30) or 80 µM dynasore (2897, Tocris) 30 min prior to pulse-labeling. After dye exposure, neurons were either immediately imaged (t0) or returned to conditioned medium with or without the continued presence of dynasore for 30 min at 37°C prior to imaging. Cy3pH(S/SA)-MG dye acquisition measurements were taken sequentially to measure the Cy3pH-induced (561 nm excitation) MG emission (680 nm) followed by MG-induced (640 nm excitation) MG emission (680 nm). Laser settings were then switched to allow capture of pHGFP signals (488 nm excitation; 510 nm emission). Individual vesicles were identified using the NIS Elements Spot Detection tool and thresholding analysis. A ROI was drawn around the cell body and binary thresholds were set to selectively capture only pHGFP clusters and the MG_ex640 (MG signal excited by 640 nM) signal above background. Next, spot detection thresholds were set to selectively identify the MG_ex561 (signal excited by 561 nM) fluorescent objects with a minimum circular area of 0.55 µm² and signal above threshold. Colocalization of the MG_ex561 spot signal and MG_ex640 binary signal was considered as a γ2^pHFAP-positive vesicle. To remove contributions of surface synaptic GABA_ARs from these vesicle measurements, MG_ex561 and MG_ex640 signals colocalized with pHGFP signals were subtracted. The values reported reflect the total number of MG_ex561- and MG_ex640-positive objects (vesicles) identified and the MG₅₆₁:MG₆₄₀ ratio of each vesicle. Early endosome characterization assays measured colocalization of Cy3-MG objects with EEA1–GFP vesicles to determine the MG₅₆₁:MG₆₄₀ ratio in these compartments, while all non-associated Cy3-MG objects were placed in the ‘other’ category. A ROUT outlier test (Q=1.0%) was used in this experimental analysis. For surface Cy3pH(S/SA)-MG pH characterization perfusion assays, the ratio of MG₅₆₁:MG₆₄₀ was determined by selectively measuring synaptic clusters of γ2^pHFAP GABA_ARs.

Neuron surface and lysosomal-association assays utilized MG-BTau dye for surface receptor pulse-labeling. For bicuculline-induced seizure assays, neurons were either immediately imaged following 100 nM MG dye incubation (t0) or returned to conditioned medium in the presence of 50 µM (–)-bicuculline methiodide (2503, Tocris) or DMSO vehicle control. To label lysosomes, neurons were incubated at 37°C in 50 nM LysoTracker Blue DND-22 (Life Technologies) in conditioned medium 30 min prior to imaging. For image analysis, independent ROIs were drawn to capture the soma, three 10 µM sections of dendrite and the whole cell. Binary thresholds and colocalization measurements were performed as above to identify MG-BTau, pHGFP synaptic GABA_AR clusters and lysosomes. Total surface pHGFP expression was determined by taking the entire cell surface signal following background subtraction.

Transfected and non-transfected DIV 13-14 cortical neurons were lysed with RIPA buffer containing: 50 mM Tris-HCl (pH 8.0), 50 mM NaCl, 1 mM EDTA, 1% Igepal, 0.5% sodium deoxycholate, 0.1% SDS, 10 mM NaF, 2 mM sodium orthovanadate and protease inhibitor cocktail (Sigma, St. Louis, MO), solubilized for 15–30 min at 4°C, and spun at 13,500 g for 15 min to remove the nuclear pellet. A BCA protein assay was performed on the supernatant, and equivalent amounts of protein were loaded for SDS-PAGE analysis. After electrophoresis and transfer to nitrocellulose membrane, samples were probed with primary antibody overnight followed by the appropriate horseradish peroxide (HRP)-coupled secondary antibody. Blots were visualized by using a Biorad Chemicdoc XRS+ following ECL development (Thermo Scientific).

Whole-cell voltage-clamp recordings were performed on HEK293 cells 12–48 h after nucleofection with GABA_AR subunits. Cells were transfected with (1) EGFP:α2_:β3:γ2^pHGFP or (2) EGFP:α2_:β3:γ2^pHFAP in a 1:1:1:3 ratio to favor the production of γ2 subunit-containing receptors. Pipettes were pulled from borosilicate capillary tubing (Sutter Instruments) to a resistance of 2–5 MΩ on a Sutter Instruments-Flaming Brown P-97 electrode puller and fire polished. Unless otherwise indicated, the extracellular solution contained (in mM): 140 NaCl, 4.7 KCl, 1.2 MgCl₂, 2.5 CaCl₂, 10 HEPES and 11 D-glucose, and was pH adjusted to 7.4±0.05 with NaOH. The intracellular solution contained (in mM): 140 CsCl, 0.1 CaCl₂, 10 HEPES, 1.1 EGTA, 2 MgCl₂, 2.5 phosphocreatine, 2 ATP-Mg and 1 GTP-Na, and was pH adjusted to 7.2±0.05 with CsOH. Recordings were made from cells expressing EGFP as identified by epifluorescence illumination on an inverted Zeiss Axioscope microscope. Cells were held at a membrane potential (V_m) of −50 mV for all experiments. V_m was corrected for an empirically determined liquid junction potential between the extracellular and intracellular solution of −4 mV. Whole-cell currents were recorded using an Axopatch 200A patch-clamp amplifier (Molecular Devices), low-pass filtered at 5 kHz and sampled at 10 kHz in pClamp10.7 (Molecular Devices). Series resistance was compensated for with the prediction and correction circuitry to at least 75% in all experiments. Rapid solution exchange was achieved using an in-house-fabricated 10-barrel fast perfusion system connected to gravity-fed reservoirs similar to the system previously described (Blanpied et al., 1997). All experiments were performed at room temperature. Peak currents were measured relative to baseline current prior to agonist application using Clampfit 10.7.

Statistical analysis was performed using GraphPad Prism 6 (GraphPad Software, San Diego, CA) or Microsoft Excel. Unpaired Student's t-tests or one way ANOVA with subsequent post hoc Tukey's test were used to determine significance in indicated imaging studies. Significant differences in data distribution for Cy3pH(S/SA)-MG experiments were determined using the Mann–Whitney test. Paired or unpaired two-tailed Student's t-tests were used to determine significance of differences in electrophysiological studies. All data are reported as mean±s.e.m. unless otherwise indicated in the text.

We would like to thank Megan Brady for her technical assistance during the design and generation of the γ2^pHFAP construct.