Abstract
The strength of synaptic inhibition can be controlled by the stability and endocytosis of surface and synaptic GABAA receptors (GABAARs), but the surface receptor dynamics that underpin GABAAR recruitment to dendritic endocytic zones (EZs) have not been investigated. Stabilization of GABAARs at EZs is likely to be regulated by receptor interactions with the clathrin-adaptor AP2, but the molecular determinants of these associations remain poorly understood. Moreover, although surface GABAAR downmodulation plays a key role in pathological disinhibition in conditions such as ischemia and epilepsy, whether this occurs in an AP2-dependent manner also remains unclear. Here we report the characterization of a novel motif containing three arginine residues (405RRR407) within the GABAAR β3-subunit intracellular domain (ICD), responsible for the interaction with AP2 and GABAAR internalization. When this motif is disrupted, binding to AP2 is abolished in vitro and in rat brain. Using single-particle tracking, we reveal that surface β3-subunit-containing GABAARs exhibit highly confined behavior at EZs, which is dependent on AP2 interactions via this motif. Reduced stabilization of mutant GABAARs at EZs correlates with their reduced endocytosis and increased steady-state levels at synapses. By imaging wild-type or mutant super-ecliptic pHluorin-tagged GABAARs in neurons, we also show that, under conditions of oxygen–glucose deprivation to mimic cerebral ischemia, GABAARs are depleted from synapses in dendrites, depending on the 405RRR407 motif. Thus, AP2 binding to an RRR motif in the GABAAR β3-subunit ICD regulates GABAAR residency time at EZs, steady-state synaptic receptor levels, and pathological loss of GABAARs from synapses during simulated ischemia.
Introduction
GABAA receptors (GABAARs) are the main mediators of inhibitory neurotransmission in the CNS and play an essential role in maintaining the excitatory/inhibitory balance required for correct brain function (Smith and Kittler, 2010). Modulating the abundance of GABAARs at synaptic sites is a key mechanism for determining the strength of synaptic inhibition (Arancibia-Cárcamo and Kittler, 2009). However, the exact molecular processes that underlie changes in GABAAR numbers at synapses and hence synaptic efficacy under normal or pathological conditions remain poorly understood.
GABAARs are dynamic entities, in terms of both their lateral mobility at the plasma membrane (Bannai et al., 2009; Muir et al., 2010) and their trafficking to and from the cell surface (Smith and Kittler, 2010). Clathrin-mediated endocytosis of GABAARs is dependent on the clathrin-adaptor AP2 and rapidly modulates synaptic GABAAR number, inhibitory synaptic strength, neuronal excitability, and, ultimately, animal behavior (Kittler et al., 2000, 2005; Chen et al., 2006; Kurotani et al., 2008; Tretter et al., 2009). Although a number of recent studies have investigated the diffusion of GABAARs into and out of synapses, far less is known about GABAAR membrane dynamics at endocytic zones (EZs) and how GABAARs are recruited to and stabilized at these regions.
Disrupted GABAAR trafficking contributes to altered information processing in pathological conditions such as epilepsy and ischemia, in which acute receptor surface downmodulation and loss of synaptic GABAARs leads to compromised neuronal inhibition and altered excitability states, but the underlying mechanisms remain unclear (Mielke and Wang, 2005; Naylor et al., 2005; Tan et al., 2007; Arancibia-Cárcamo et al., 2009). Indeed, although both clathrin-dependent and -independent (e.g., caveolar-dependent) pathways for GABAAR endocytosis have been described (Kittler et al., 2000; Bradley et al., 2008), the specific contribution of AP2-dependent pathways for synaptic receptor loss in epilepsy or ischemia remains unclear. Deciphering the molecular mechanisms and endocytic pathways responsible for synaptic GABAAR downmodulation is therefore central to understanding how synaptic inhibition is regulated in the healthy brain and could lead to identification of potential target pathways of pathological receptor downmodulation for therapeutic intervention during disinhibition in disease.
Here, we characterize the molecular interactions that underlie internalization of GABAARs containing β3 subunits, which are present in a large proportion of receptor subtypes in the hippocampus and cortex, regions that are particularly vulnerable to excitotoxicity (Lo et al., 2003). Using a combination of molecular, biochemical, and imaging approaches, we identify a three arginine (RRR) motif within the GABAAR β3-subunit intracellular domain (ICD) that mediates the clathrin-dependent endocytosis of GABAARs via direct binding to AP2. Furthermore, by combining single-particle tracking with mutagenesis of the RRR motif, we reveal that GABAARs are reversibly stabilized at EZs, depending on interactions with AP2. Additionally, we demonstrate that loss of synaptic GABAARs in an in vitro model of ischemia is dependent on AP2-binding via this motif and that blocking this mechanism can reduce cell death. Thus, direct binding of GABAAR ICDs to AP2 regulates GABAAR recruitment into EZs, steady-state GABAAR levels at synapses, and provides a key mechanism for receptor downmodulation in pathology.
Materials and Methods
cDNA cloning.
The β3 GABAAR-subunit mutants were made by site-directed mutagenesis on a pRK5mycβ3 backbone. The ICDs of these mutants were amplified by PCR, and the resulting fragment was inserted into the BamH1/Not1 sites of pGEX–4T3. SEPγ2 (Superecliptic pHluorin) has been described previously (Muir et al., 2010). Untagged α1 was in pRK5 and was described previously (Twelvetrees et al., 2010). mRFP–rab5 [Addgene plasmid 14437 (Vonderheit and Helenius, 2005)] and clathrin-light chain polypeptide–EYFP [Addgene plasmid 21741 (Chen and Zhuang, 2008)] were from Addgene. EGFP was from Clontech.
Antibodies.
The following primary antibodies were used: mouse monoclonal anti-Myc was obtained from 9E10 hybridoma cells [immunofluorescence (IF) and ELISA, supernatant 1:100], anti-AP50 [Western blot (WB), 1:250; BD Biosciences], anti-GFP (IF, 1:500; clone N86/38; Neuromab), rabbit polyclonal anti-myc [IF, 1:200; quantum dot (QD) imaging, 1:100; Santa Cruz Biotechnology], rabbit anti-vesicular inhibitory amino acid transporter (VIAAT) [a kind gift from B. Gasnier, Universite Paris, Descartes, France (Dumoulin et al., 1999); IF, 1:1000], anti-α1 GABAAR subunit (WB, 1:5; clone N95/35; Neuromab), anti-β-adaptin (WB, 1:500; Sigma), vesicular GABA transporter (VGAT)–Oyster 488 (live imaging, 1:1000; Synaptic Systems), and anti-myc (QD imaging, 1:100; Millipore Bioscience Research Reagents).
Cell culture.
Cultures of cortical and hippocampal neurons were prepared from E18 Sprague Dawley rat embryos of either sex as described previously (Kittler et al., 2004). Neurons were transfected by nucleofection before plating as described previously (Kittler et al., 2004; Macaskill et al., 2009; Twelvetrees et al., 2010). COS-7 cells were maintained in DMEM (Invitrogen), supplemented with 10% heat-inactivated fetal bovine serum and penicillin–streptomycin, and transfected by electroporation (Kittler et al., 2004) for biochemistry and by nucleofection (Amaxa) for immunocytochemistry.
GST fusion protein pull-down assays from rat brain homogenate.
Pull-downs from brain were performed as described previously (Smith et al., 2008, 2010). Briefly, adult rat brain was homogenized in pull-down buffer (50 mm HEPES, pH 7.5, 0.5% Triton X-100, 150 mm NaCl, 1 mm EDTA, and 1 mm PMSF with antipain, pepstatin, and leupeptin at 10 μg/ml) and solubilized for 2 h. Solubilized material was ultracentrifuged at 66,000 × g for 40 min at 4°C, and the supernatant (solubilized protein) was exposed to 10–20 μg of GST fusion protein attached to glutathione–agarose beads for 1 h at 4°C. Beads were then washed extensively and analyzed by SDS-PAGE and WB.
In vitro translation of [35S]methionine-labeled protein and GST pull-down assays.
Residues 156–433 of the μ2 subunit of AP2 were in vitro translated using TNT SP6 Quick Coupled Transcription/Translation system (Promega) and labeled with [35S]methionine (GE Healthcare) following the protocol of the manufacturer (Smith et al., 2008, 2010). Five microliters of labeled protein were incubated with 10–20 μg of GST fusion protein in pull-down buffer (as above) for 2 h at 4°C. The beads were washed and resolved by SDS-PAGE followed by staining with Coomassie blue. Radioactivity was detected with a phosphor-storage screen and PhosphorImager. Bacterially expressed wild-type (WT) or mutant (mut) His-tagged μ2 were used for nonradioactive in vitro pull-downs with GST fusion proteins and analyzed by SDS-PAGE, followed by staining with Coomassie blue (Smith et al., 2008).
Surface plasmon resonance.
Surface plasmon resonance (SPR) experiments were performed as described previously (Kittler et al., 2008) on a BIAcore 2000 instrument (BIAcore).
Immunofluorescence.
For receptor internalization assays in COS-7 cells, receptors were live labeled with 9E10 anti-myc antibody in HEPES buffer (in mm: 25 HEPES, 140 NaCl, 5.4 KCl, 1.8 CaCl2, and 15 glucose, pH 7.4) at 4°C. Labeled receptors were allowed to internalize for 30 min at 37°C, and surface antibody was stripped away using 0.2 m acetic acid and 0.5 m NaCl before fixation as described previously (Arancibia-Cárcamo et al., 2009). Internalized receptors were identified using Alexa Fluor-594-conjugated or Alexa Fluor-488-conjugated anti-mouse secondary antibody (1:1000; Invitrogen) and confocal microscopy. Quantification of receptor internalization was performed by determining the area of the internalized receptors and normalization to the cell area (as given by GFP fluorescence; described by Hanley and Henley, 2005) or by measuring the colocalization of the internalized receptors with the early endosomal marker rab5.
Receptor internalization assays in neurons were performed similarly to what has been described previously (Scott et al., 2004), with modifications. Briefly, transfected neurons were incubated at room temperature with anti-9E10 antibody and placed at 37°C for 30 min to enable internalization. After washing, cells were fixed and surface receptors were stained with secondary Alexa Fluor-488 secondary antibody. Cells were permeabilized and stained with Alexa Fluor-594 secondary antibody to label the total receptor population and then visualized by confocal microscopy. For quantification of receptor internalization, red fluorescence that was not colocalized with green fluorescence (surface receptors) was calculated as the internalized receptor population and normalized to total receptor population.
Neurons for surface staining were fixed with 4% paraformaldehyde (PFA)/4% sucrose/PBS, pH 7, for 4 min and blocked with block solution (PBS, 10% horse serum, and 0.5% BSA) for 10 min. Neurons were incubated for 1 h with anti-9E10, followed by permeabilization with block solution containing 0.2% Triton X-100. Neurons were then incubated with anti-VIAAT antibody and subsequently with Alexa Fluor-488 and Alexa Fluor-594 secondary antibodies (1:1000; Invitrogen). After extensive washing, coverslips were mounted on microscope slides using ProLong Gold antifade reagent (Invitrogen) and sealed with nail varnish. Images were acquired using a Carl Zeiss Pascal confocal microscope and digitally captured using LSM software.
Surface receptor ELISA assay.
Quantification of surface myc-tagged GABAARs was performed by colorimetric cell ELISA as described previously (Mielke and Wang, 2005). Briefly, 8–9 DIV cultured cortical neurons were fixed with 4% PFA/sucrose for 5 min and then blocked under nonpermeabilizing conditions (PBS and 3% BSA). Cells were incubated for 1 h with anti-9E10 (1:100) and then for 1 h with HRP-conjugated anti-mouse secondary antibody (1:2000; Rockland). After extensive washing, cells were incubated with 1 vol of o-phenylenediamine substrate (Sigma) for 3 min. Reactions were stopped using 0.2 vol of 3N HCl, and 1 ml of supernatant was used to determine the absorbance at 492 nm. For N-ethylmaleimide-sensitive factor (NSF) inhibition experiments, COS-7 cells expressing myc-tagged GABAARs were incubated with a TAT-fused NSF inhibitory peptide (NSFi) or scrambled control peptide for 75 min (Anachem) (Morrell et al., 2005; Chou et al., 2010), followed by the ELISA procedure described above.
Live-cell imaging and oxygen–glucose deprivation.
Hippocampal neurons were nucleofected at time of plating with mycβ3 (WT, RRR/AAA or 7A), SEPγ2, and untagged α1 in a ratio of 2:1:1. To assess the cotransfection of mycβ3WT, mycβ3RRR/AAA, and mycβ37A with the SEPγ2 in hippocampal neurons, neurons were live labeled for 15 min with a mouse anti-GFP antibody diluted in HEPES buffer (in mm: 25 HEPES, 140 NaCl, 5.4 KCl, 1.8 CaCl2, and 15 glucose, pH 7.4) to label SEPγ2 clusters. Cells were then fixed and blocked as described above and labeled with rabbit anti-myc antibody to label mycβ3-subunit-containing receptors. Coverslips were washed and mounted as described above. Quantification of cotransfection was performed by assessing whether the SEPγ2-transfected neurons in a 40× field of view were also expressing mycβ3 subunits. This was then expressed as a percentage of the SEPγ2-transfected neurons.
Cells were imaged at 12–15 DIV as described previously (Arancibia-Cárcamo et al., 2009). Carbonate-buffered solution contained 126 mm NaCl, 2.5 mm KCl, 1 mm MgCl2, 2 mm CaCl2, 1 mm NaH2PO4, 24 mm NaHCO3, and 10 mm glucose and was gassed with 95% O2, 5% CO2. For oxygen–glucose deprivation (OGD), 7 mm sucrose was used in place of glucose, and the solution was gassed with 95% N2, 5% CO2. Both solutions were bubbled for 30 min before addition of CaCl2 to prevent precipitate formation. Solutions were perfused at a constant rate of ∼2 ml/min and heated to 35–37°C. For OGD experiments, the solution was switched from control to OGD buffer at t = 0 min. Super-ecliptic pHluorin fluorescence was captured using an Olympus microscope (BX51WI) with a 60× Olympus objective coupled to an EM-CCD camera (Ixon; Andor). Excitation was provided by a mercury-spiked xenon arc lamp (Cairn). Bandpass filters for excitation and emission were used (470/40 and 525/50 nm, respectively). Images (averages of eight consecutive exposures of 150 ms) were acquired every 10 min. The StackReg macro (Thévenaz et al., 1998) was used to correct for minor x–y drift that may have occurred between time points. Using NIH ImageJ, regions of interest were drawn around each non-somatic fluorescent cluster. After background subtraction, the average intensity across SEP-GABAAR clusters in each time point was used as readout. This value was normalized to that at t = 0 min.
For cell death assays, hippocampal cultures were transduced with peptides (β3 peptide, KTHLRRRSSQLK; control peptide, INRVDAHENAAAAPM) for 1 h using Chariot reagent (active motif) following the instructions of the manufacturer (Kittler et al., 2008; Twelvetrees et al., 2010). After peptide transduction, cultures were exposed to 40 min OGD using a Billups-Rothenberg modular incubator chamber as described previously (Dixon et al., 2009) using the control and OGD buffers described above. To monitor cell death, 24 h after OGD treatment, neurons were stained with propidium iodide (PI) (4 μg/ml) for 1 h, followed by fixation with 4% PFA/sucrose and staining with DAPI. For each experiment, six to eight fields of view were imaged per condition, and the number of DAPI-positive nuclei that were PI stained was determined.
Single-particle tracking.
Labeling of myc-tagged GABAARs (mycβ3WT and mycβ3RRR/AAA) with QDs was performed using a rabbit anti-myc antibody (Santa Cruz Biotechnology) or mouse anti-myc antibody (Millipore Bioscience Research Reagents) and anti-rabbit or anti-mouse 605 nm QDs (Invitrogen). Labeling was performed at room temperature with primary antibody (0.1 μg/ml, 1:100) for 8 min in block solution (imaging media containing 10% horse serum) and then QD incubation (0.5 nm) for 2 min in block solution as above. QD movies were recorded for 200 frames at 8.5 Hz (23.5 s duration). The imaging media for these experiments contained the following (in mm): 125 NaCl, 5 KCl, 1 MgCl2, 2 CaCl2, 10 d-glucose, and 10 HEPES, adjusted to pH 7.4. Labeling of inhibitory presynaptic terminals with VGAT–Oyster 488 (1:1000; Synaptic Systems) was performed by incubation of coverslips in 200 μl of imaging media for 30 min at 37°C, 5% CO2, and we found that spontaneous activity in the neuronal culture was sufficient to load the dye into inhibitory presynaptic terminals. Similar staining could be achieved with a depolarizing high KCl stimulus (60 mm, 5 min; data not shown).
For analysis of single GABAAR dynamics, trajectories were created automatically from movies of QD-tagged receptors. For each frame, an image segmentation algorithm was run that classifies image pixels based on their local symmetry at a specified scale. Pixels with maximal rotational symmetry (corresponding to local maxima in intensity) were classified as being attributable to QDs, and their locations were given to subpixel accuracy by taking the weighted average of connected target pixels. The pointing accuracy of the system is ∼50 nm. Tracks were generated by linking particle positions between frames, using homemade software written in Mathematica (Wolfram Research). Briefly, the Euclidian distance between frames was minimized using a function that allowed for particle blinking such that tracks could be connected across blinking events. Instantaneous diffusivities (Dinst) were calculated from interframe displacements along consecutive track segments according to the 2D diffusion law (x2 = 4Dinstt), and fitting to the first five points of the displacement curve was used to estimate the instantaneous diffusion coefficient. QD track segments were classified as being inside clathrin-coated pits (labeled by clathrin–YFP) if their midpoint was within 0.4 μm of the center of a pit. Cluster positions were found automatically from images taken immediately before the corresponding QD movie using the image segmentation algorithm described above. For imaging at inhibitory synapses, QD track segments were classified as synaptic if their midpoint was within 0.75 μm of the center of a VGAT punctum.
Statistical analysis.
All experiments were performed at least three times from different preparations. Unless otherwise stated, p values given are from two-tailed t tests, and values are given as mean ± SEM. Error bars represent SEM.
Results
AP2 binds to a patch of basic amino acid residues within the β3-subunit ICD
Little is currently known regarding the dynamics of GABAARs at EZs in neuronal dendrites for their subsequent endocytosis. Recruitment of GABAARs to these sites is likely to depend on direct associations between the μ2 subunit of the clathrin adaptor AP2 (μ2–AP2) and GABAAR-subunit ICDs (Kittler et al., 2005). To allow us to study the role of GABAAR β3 subunits in regulating the dynamics of GABAAR endocytosis, we set out to identify β3-subunit mutants no longer able to directly bind μ2–AP2. To this end, we made multiple GST fusion proteins of the β3-subunit ICD with sequential regions of the ICD deleted (Fig. 1A) and tested their ability to bind in vitro translated 35S-labeled μ2–AP2. Quantification of the radioactive protein bound to the various GST–β3 ICD deletions revealed that the binding of μ2–AP2 to deletion 6 (Δ395–414) was significantly reduced compared with binding to the other GST–β3 ICD deletions (n = 5, *p < 0.05; Fig. 1B,C). To ascertain whether a similar loss of binding was observed for μ2–AP2 from brain, we used a similar pull-down approach to assess the binding of GST–β3 deletion proteins to μ2–AP2 from rat brain homogenate. In agreement with our findings using radiolabeled μ2–AP2, the binding of brain μ2–AP2 to deletion 6 was significantly reduced compared with the other GST–β3 deletion proteins (n = 4, *p < 0.05; Fig. 1D,E), further confirming that the region deleted in GST–β3 deletion 6 is important for the binding of μ2–AP2 to the β3-subunit ICD.
Identification of the interaction domain within the GABAAR β3-subunit ICD that mediates association with AP2. A, Schematic showing deletions within the GABAAR β3-subunit ICD. B, In vitro translated 35S-labeled μ2 was subjected to pull-down experiments with deletion β3 ICD GST fusion proteins. IN, Input. C, Quantification of μ2 binding shows a significant decrease in binding to deletion 6 (Δ395–414) (n = 5, *p < 0.05). D, Pull-down experiments with rat brain lysate with the deletion β3 ICD GST fusion proteins. Protein complexes were analyzed by SDS-PAGE and Western blotting with an antibody to AP2. E, Densitometric quantification of endogenous AP2 binding to deletion β3 ICDs reveals a significant decrease in binding of AP2 to deletion 6 (Δ395–414) (n = 4, *p < 0.05). F, Amino acid sequence of the GABAAR β3-subunit ICD. Deletion 6 (Δ395–414) is shown by the black line. Basic residues that may be a putative AP2 binding site are highlighted in dark gray and are mutated to alanine in the β37A mutant construct. A putative dileucine motif is highlighted in light gray. G, In vitro translated 35S-labeled μ2 was subjected to pull-down experiments with β3-subunit mutant GST fusion proteins. H, Quantification of μ2 binding showing that the mutation of seven basic residues (β37A) within the region 395–414 significantly reduces binding of μ2 to the β3 ICD (n = 3, **p < 0.01). I, Pull-down experiments in rat brain lysate with the deletion β3-subunit mutant GST fusion proteins. Protein complexes were analyzed by SDS-PAGE and Western blotting with an antibody to AP2. J, Quantification of endogenous AP2 binding to β3-subunit mutant ICDs showing the mutation of seven basic residues significantly reduces the binding of endogenous AP2 to the β3 ICD (n = 6, *p < 0.05), but the mutation of the dileucine motif has no significant effect on AP2 binding.
The 20 aa region identified in Figure 1A–E (deletion 6) contains a stretch of amino acids enriched with basic lysine and arginine residues, similar to motifs identified in other AP2-binding membrane proteins, including synaptotagmin-1 (Syt-1) (Haucke et al., 2000), the α1β-adrenergic receptor (Diviani et al., 2003) and AMPA receptors (Kastning et al., 2007). To determine whether these residues within the β3 subunit (400KKTHLRRR407) were responsible for AP2 binding, we mutated the basic amino acids within this region to alanines (GST–β37A; Fig. 1F). A dileucine motif within the β2 GABAAR-subunit ICD has also been suggested to be a determinant of clathrin-dependent internalization of the receptor (Herring et al., 2003). To verify whether a conserved dileucine motif in the β3-subunit ICD was also important for AP2 binding, we also mutated this motif to alanine (Fig. 1F), to give GST–β3ILL/AAA, and additionally constructed a double mutant, GST–β37A+ILL/AAA with both sites mutated. These fusion proteins were used in pull-down assays with 35S-labeled μ2–AP2 to analyze the effect of these specific mutations on the AP2/β3 ICD interaction. Mutation of the seven basic residues greatly reduced binding of μ2–AP2 to the β3-subunit ICD by threefold (n = 3, **p < 0.01; Fig. 1G,H), to a similar level to the binding of μ2–AP2 to β3 deletion 6 (β3Δ395–414). However, the mutation of the dileucine motif to alanines had no significant effect on the binding of μ2–AP2 to the β3 ICD, and mutation of both motifs (β37A+ILL/AAA) had a similar effect to the 7A mutant alone, suggesting that the dileucine motif does not significantly contribute to the interaction with μ2–AP2. These experiments also corroborate the results from Figure 1B–E, which show no decrease in binding to AP2 with deletion 3, which contains the putative dileucine motif. Similar results were obtained from pull-down experiments with rat brain homogenate with binding of μ2–AP2 to the β3 ICD mutants mirroring the in vitro pull-down assays (n = 6, *p < 0.05; Fig. 1I,J). These results support a key role for a 7 aa motif (400KKTHLRRR407) within the GABAAR β3-subunit ICD in mediating the interaction with AP2.
Molecular determinants of the β3-subunit/AP2 interaction
To further identify the key residues within the 400KKTHLRRR407 motif that mediate AP2 binding, we sequentially mutated the seven basic amino acids (shown in Fig. 2A) within GST–β3 ICD to alanine, generating GST–β3KK/AA, GST–β3TH/AA, and GST–β3RRR/AAA mutants, which were then used in pull-down experiments with rat brain homogenate (Fig. 2B). Quantification of AP2 binding by densitometry showed a large decrease in the binding of AP2 to the GST–β3RRR/AAA mutant compared with the binding to the GST–β3WT, GST–β3KK/AA, and GST–β3TH/AA fusion proteins (n = 3, *p < 0.05; Fig. 2C), suggesting that these three arginine residues are essential for the binding of the β3 ICD to AP2. AP2 comprises a heteromer of α, β2, σ2, and μ2 adaptin subunits (Collins et al., 2002). We further confirmed that the RRR motif mediates interaction with heteromeric AP2 endogenous complexes as revealed by WB with a β2 adaptin antibody (Fig. 2B). Direct binding of μ2–AP2 to the RRR motif was further confirmed by in vitro pull-down assays using GST–β3WT and GST–β3RRR/AAA with bacterially expressed His–μ2 (Fig. 2D), which also showed a large decrease in the direct binding of His–μ2–AP2 to GST–β3RRR/AAA compared with GST–β3WT.
Molecular determinants of the AP2/GABAAR β3-subunit interaction. A, Sequence of the basic amino acid stretch of the β3-subunit ICD showing the basic amino acids (red) and their sequential mutation to alanine residues (blue). B–D, The arginine residues in the β3 subunit mediate its interaction with AP2. B, Pull-down experiments in rat brain lysate with WT, KK, TH, and RRR β3 ICD mutant GST fusion proteins. Protein complexes were analyzed by SDS-PAGE and Western blotting with antibodies to the β2 and μ2 subunits of the AP2 complex. C, Quantification of AP2 binding by densitometry showed a large 3.6-fold decrease in binding to GST–β3RRR/AAA (n = 3, *p < 0.05), whereas there was no significant decrease in binding to the KK or TH mutants. D, In vitro pull-down experiments with the β3-subunit mutant GST fusion proteins with and without His–μ2. Protein complexes were analyzed by SDS-PAGE and stained with Coomassie blue. E, In vitro pull-down experiments with the β3-subunit GST fusion protein and WT His–μ2 and lysine mutants of His–μ2. Protein complexes were analyzed by SDS-PAGE and stained with Coomassie blue. F, G, The β3-subunit ICD and EGFR bind distinct regions of μ2–AP2. F, In vitro pull-down experiments with Syt-1–GST, β3–GST, EGFR–GST, and GST as a control, and WT His–μ2, K354E–His–μ2, and Syt-1mut–His-μ2 (which is incapable of binding to Syt-1). Protein complexes were analyzed by SDS-PAGE and stained with Coomassie blue. G, SPR sensograms showing binding of WT and mutant His–μ2 to β3-basic patch and FYRALM peptides, demonstrating that the K354E mutation in μ2 reduces its affinity for the β3 peptide but not for the EGFR peptide. KD values are given in the figure. N/D, No interaction was detectable. RU, Response units.
The amino acid residues within the μ2 subunit of AP2 responsible for the interaction with the β3 ICD are not known. Atypical basic-patch motif containing membrane cargo (such as Syt-1) are known to interact with subdomain B of μ2–AP2 (residues 283–394). In the case of Syt-1, key residues within this region of μ2–AP2 are important for the binding of Syt-1 (Haucke et al., 2000). We therefore tested the binding of His–μ2, containing a series of lysine (K) residue mutations, to the GST–β3 fusion protein by performing pull-down assays with variants of μ2 containing key K residues in subdomain B mutated (Fig. 2E). These experiments revealed that mutation of K354 to glutamate (E) within His–μ2 is sufficient to abolish its binding to the β3 ICD (Fig. 2E), suggesting that this residue is critical for β3-subunit—AP2 complex formation. To support this finding, additional pull-down experiments were performed with wild-type His–μ2, mutant K354E–His–μ2 (its binding to β3 abolished), and, as an additional control, His–μ2–Syt-1mut (with key additional subdomain B residues mutated preventing its binding to Syt-1: K343A, Y344A, K345E, E348A, K354E, and K356E) and GST–β3 (Fig. 2F). This revealed that the K354E–His–μ2 and His–μ2–Syt-1mut mutants exhibited significantly reduced binding to GST–β3 (and with GST–Syt-1, as a control) but could still interact with GST fusion protein of the C-terminal tail of the EGF receptor (EGFR) (residues 995–1045), which associates with a different region of μ2–AP2 via tyrosine-based endocytic motifs (Owen and Evans, 1998) (Fig. 2F). The key role of His–μ2 for β3 binding was further supported by SPR experiments that compared the binding affinities of His–μ2 and His–μ2–K354E to a 401KTHLRRRSS409 peptide (the β3 ICD basic patch motif) or FYRALM peptide (an EGFR-derived tyrosine-based endocytic motif). Both endocytic sorting motif peptides bind to wild-type His–μ2 with a similar KD value of ∼1 μm. In agreement with the GST pull-downs (Fig. 2F), the SPR experiments showed that the μ2 K354E mutation selectively prevents binding to the basic atypical motif within the β3 ICD but not to the tyrosine motif within the EGFR peptide (Fig. 2G). These observations lead to the conclusion that the GABAAR β3-subunit binds μ2–AP2 via a minimal 405RRR407 motif, to a domain within μ2–AP2, which is distinct from the C-terminal pocket region of μ2–AP2 that binds tyrosine-based motifs such as those within EGFR.
AP2–β3-subunit interactions stabilize GABAARs at EZs but not synapses
Having identified a key AP2-binding site in GABAAR β3 subunits, we proceeded to investigate the role of β3-subunit-mediated AP2 binding for regulating the dynamics of GABAARs at EZs in hippocampal neuronal dendrites. As a prelude to investigating the role of AP2 adaptor interactions in stabilizing GABAARs to EZs, we initially characterized the membrane dynamics of β3-subunit-containing receptors in dendrites with respect to EZs. To test this, we visualized single GABAAR-containing myc-epitope-tagged wild-type GABAAR (mycβ3WT) subunits, using QD tracking (Muir et al., 2010), a well-established method for following the behavior of single receptors in the surface membrane (Triller and Choquet, 2008). To examine GABAAR mobility using this method in relation to regions of endocytosis in the neuronal plasma membrane, we simultaneously labeled EZs by cotransfecting neurons with YFP-tagged clathrin light chain, which has been used previously to label these subdomains in dendrites (Blanpied et al., 2002; Petrini et al., 2009). Single myc-tagged GABAARs were labeled using a low concentration of rabbit anti-myc antibody in conjunction with an anti-rabbit secondary coupled to 605 nm QDs. Single QD-labeled mycβ3WT-containing GABAARs could be observed to rapidly diffuse into and out of EZs (Fig. 3A,B) and were less mobile when at EZs compared with outside these subdomains (median Din = 0.008 μm2s−1, median Dout = 0.020 μm2s−1).
Interactions with AP2 stabilize GABAARs at EZs. A, Trajectories of QD-tagged mycβ3WT subunits (orange) shown overlaid onto the location of clathrin–YFP-labeled EZs (in white). Scale bar, 5 μm; imaging duration, 23.5 s. Arrows indicate examples in which QD-tagged GABAARs are within EZs. B, As in A but for QD-tagged mycβ3RRR/AAA subunits. C, Cumulative frequency plots of the instantaneous diffusion coefficients for QD-tagged mycβ3WT (solid line) and mycβ3RRR/AAA (dashed line) subunits inside EZs as marked by clathrin–YFP. GABAARs containing mycβ3RRR/AAA subunits are 3.3-fold more mobile than mycβ3WT within EZs (***p = 2.2 × 10−16, K–S test; mycβ3WT nin = 13,785; mycβ3RRR/AAA nin = 10,836). D, As in C, for mycβ3WT (solid line) and mycβ3RRR/AAA (dashed line) subunits outside EZs. GABAARs containing β3 subunits with RRR/AAA mutation are 1.63-fold more mobile than WT outside EZs (***p = 2.2 × 10−16, K–S test; mycβ3WT nout = 150,884; mycβ3RRR/AAA nout = 200,986). E, Section of example neuron showing GABAergic terminals loaded with VGAT-Oyster 488. F, Dendrite of neuron with GABAergic terminals loaded with VGAT-Oyster 488 and stained with GAD-6 antibody. G, QD-tagged GABAAR trajectories (orange) shown overlaid on VGAT loading (inhibitory synapses) for mycβ3WT receptors (top) and mycβ3RRR/AAA mutant receptors (bottom). Arrows point to trajectories in synaptic areas. H, Cumulative frequency plots of instantaneous diffusion coefficients for mycβ3WT and mycβ3RRR/AAA receptors at synapses (positions given by VGAT loading). Mobilities at inhibitory synapses are not significantly different (p = 0.77, K–S test; mycβ3WT nin = 10,341; mycβ3RRR/AAA nin = 8250). I, As in H but for extrasynaptic regions. mycβ3RRR/AAA receptors are 1.3-fold more mobile extrasynaptically (p = 2 × 10−16, K–S test; mycβ3WT nout = 127,452; mycβ3RRR/AAA nout = 91,211).
We then used QD tracking of mutant GABAARs to directly investigate the role of interactions with the AP2 complex on receptor lateral mobility at EZs. Strikingly, the lateral mobility of mutant mycβ3RRR/AAA-containing GABAARs inside EZs was 3.3-fold higher compared with wild-type receptors at EZs [mycβ3RRR/AAA median Din = 0.027, mycβ3WT median Din = 0.008 μm2s−1, ***p = 2.2 × 10−16, Kolmogorov-Smirnov (K–S) test (Fig. 3C); mycβ3RRR/AAA median Dout = 0.032, mycβ3WT median Dout = 0.020 μm2s−1, ***p = 2.2 × 10−16 (Fig. 3D)]. In agreement with mutant receptors being less stable at EZs, mutation of the three arginine residues caused a significant decrease in the mean time spent by GABAARs at EZs (mycβ3WT, 0.81 ± 0.18 s, n = 12 cells; mycβ3RRR/AAA, 0.42 ± 0.03 s, n = 11 cells; *p = 0.05), suggesting that mutation of the RRR motif disrupts the interaction required to stabilize GABAARs at dendritic EZs before their endocytosis.
We also investigated the mobility and residency time of individual mycβ3RRR/AAA- containing GABAARs at inhibitory synapses. Inhibitory presynaptic terminals were live labeled using an antibody to the luminal domain of the GABA transporter VGAT, coupled to the fluorophore Oyster-488 (Martens et al., 2008), which we confirmed were GAD-6 positive (Fig. 3E,F). QD–GABAAR trajectories were analyzed inside and outside of synaptic locations (as determined by VGAT puncta). Both mycβ3WT- and mycβ3RRR/AAA-containing GABAARs were slower at synapses compared with their extrasynaptic counterparts. Inside synapses, mycβ3WT and mycβ3RRR/AAA lateral mobility was very similar (mycβ3WT median D = 0.00882 μm2s−1; mycβ3RRR/AAA median D = 0.00887 μm2s−1; p = 0.77, K–S test), suggesting that the RRR motif is not important in stabilizing receptors at synapses. Outside synapses, mycβ3RRR/AAA receptors were 1.3-fold more mobile than mycβ3WT receptors (mycβ3WT median D = 0.0137 μm2s−1; mycβ3RRR/AAA median D = 0.0180 μm2s−1; ***p = 2 × 10−16, K–S test). This is likely attributable to the fact that EZs (at which mutant receptors are faster than wild-type receptors) are found extrasynaptically. In agreement with the finding that wild-type and mutant receptor mobilities are similar at synapses, we found no difference in the mean time spent by these receptors at synapses (mycβ3WT, 1.29 ± 0.22 s, n = 14 cells; mycβ3RRR/AAA, 1.36 ± 0.33 s, n = 11 cells; p = 0.70). Together, these experiments show the RRR AP2-binding motif in the β3 subunit is important in regulating GABAAR stabilization at EZs but does not affect the stabilization of individual receptors at inhibitory synapses.
Mutation of three arginine residues within the β-subunit ICD impairs GABAAR endocytosis
We predicted that, because mycβ3RRR/AAA mutant-containing GABAARs are less efficiently recruited to EZs and spend less time at these sites, the endocytosis of these mutant receptors would be disrupted. To test this, we performed antibody feeding experiments with COS-7 cells expressing α1mycβ3WT or α1mycβ3RRR/AAA GABAARs and GFP (Fig. 4A). Quantification of GABAAR internalization by normalization of the internalized receptor area to the total cell area (as given by GFP fluorescence, arbitrary units) revealed a significant decrease in α1mycβ3RRR/AAA receptor endocytosis (0.78 ± 0.07) compared with α1mycβ3WT receptors (2.92 ± 0.50, n = 3, *p < 0.05). Similar antibody feeding assays were performed on COS-7 cells cotransfected with other β3-subunit mutants analyzed for AP2 binding in Figure 1 (mycβ37A, mycβ3ILL/AAA, mycβ37A+ILL/AAA, and mycβ3Δ395–414). COS-7 cells expressing the α1mycβ37A (0.47 ± 0.22), α1mycβ37A+ILL (0.68 ± 0.11), and α1mycβ3Δ395–414 (0.56 ± 0.21) mutant GABAARs all displayed significantly reduced endocytosis levels compared with α1mycβ3WT and α1mycβ3ILL/AAA receptors (α1mycβ3WT, 1.5 ± 0.3; α1mycβ3ILL/AAA, 1.47 ± 0.3; n = 3, *p < 0.05; Fig. 4B).
Mutation of the RRR motif impairs GABAAR endocytosis and causes their surface accumulation. A, COS-7 cells coexpressing GABAARs (α1mycβ3 or α1mycβ3RRR/AAA) and GFP were live labeled with anti-myc antibody and placed at 37°C for 30 min to allow internalization of receptors. Remaining surface antibody was stripped away and cells were fixed, permeabilized, and labeled with secondary antibody. Quantification of GABAAR endocytosis by normalization of the internalized receptor area to the total cell area (as given by GFP fluorescence) showed decreased internalization of mycβ3RRR/AAA receptors compared with mycβ3WT receptors (mycβ3WT, 2.92 ± 0.50; mycβ3RRR/AAA, 0.78 ± 0.07; n = 3, *p < 0.05). B, Summary of data from similar antibody feeding experiments with mycβ3WT, mycβ37A, mycβ3ILL/AAA, mycβ37A+ILL/AAA, and mycβ3Δ395–414 containing GABAARs. C, Antibody feeding experiments in COS-7 cells coexpressing GABAARs (α1mycβ3WT or α1mycβ3RRR/AAA) and RFP–rab5. Receptor internalization was measured as the percentage colocalization of receptors with rab5 and showed a 50% decrease in internalization of α1mycβ3RRR/AAA mutant GABAARs compared with wild-type receptors (mycβ3WT, 24.0 ± 0.4%; mycβ3RRR/AAA, 12.5 ± 1.4%; **p < 0.01, n = 4). D, Summary of data from similar antibody feeding experiments with mycβ3WT, mycβ37A, mycβ3ILL/AAA, mycβ37A+ILL/AAA, and mycβ3Δ395–414 containing GABAARs. E, DIV 12–14 neurons expressing mycβ3WT or mycβ3RRR/AAA subunits were treated with an antibody feeding protocol. Fluorescence of internalized receptors was represented as a fraction of the total fluorescence of the cell and normalized to values for mycβ3WT-transfected cells. Red fluorescence indicates internalized receptors. The internalization of mycβ3RRR/AAA-containing receptors was reduced to 28.0 ± 5.0% compared with mycβ3WT receptors (n = 3, **p < 0.01). Arrows point to internalized receptor population (red). F, Quantification of similar antibody feeding experiments in neurons expressing mycβ37A, mycβ3ILL/AAA, mycβ37A+ILL/AAA, and mycβ3Δ395–414 receptors. Scale bars, 20 μm. G, Cluster counting analysis of 25 μm dendritic segments of neurons expressing mycβ3WT or mycβ3RRR/AAA revealed a 68.0 ± 23.2% (n = 3, **p < 0.01) increase in total surface receptor cluster area. Scale bar, 5 μm. H, ELISA experiments were performed with COS-7-expressing α1mycβ3WT or α1mycβ3RRR/AAA GABAAR subunits that were treated with either an NSFi or a scrambled control peptide (Scram). The NSFi peptide caused a 35.3 ± 7.2% increase in surface α1mycβ3WT GABAARs and a 47.7 ± 14.08% increase in surface α1mycβ3RRR/AAA GABAARs (n = 4, *p < 0.05).
In similar experiments with COS-7 cells cotransfected with RFP–rab5, an early endosomal marker (Fig. 4C), quantification of the colocalization of internalized receptors with rab5-labeled compartments revealed that RRR mutant GABAARs displayed significantly decreased colocalization with rab5 compared with WT receptors (mycβ3WT, 24.0 ± 0.4%; mycβ3RRR/AAA, 12.5 ± 1.4%; n = 4, **p < 0.01; Fig. 4C). This was also the case with the other mutant receptors, revealing decreased colocalization with rab5-labeled compartments for the α1mycβ37A, α1mycβ3ILL/AAA, and α1mycβ3Δ395–414 receptors [α1mycβ37A, 13.17 ± 0.55% (*p < 0.05); α1mycβ37A+ILL/AAA, 20.02 ± 3.21% (*p < 0.05); and α1mycβ3Δ395–414, 9.48 ± 1.50% (**p < 0.01)] compared with their wild-type counterparts (30.42 ± 1.53%; Fig. 4D). However, α1mycβ3ILL/AAA mutant receptors displayed almost identical levels of colocalization with rab5 as wild-type receptors (30.68 ± 2.98%, p = 0.7), suggesting that the dileucine motif does not play a significant role in the endocytosis of α1β3-containing receptors to the early endosome compartment in COS-7 cells.
To also test whether a similar mechanism of GABAAR endocytosis occurs in neurons, we transfected neurons with mycβ3WT or mycβ3RRR/AAA constructs, which assemble with endogenous subunits in neurons (data not shown), and performed a similar antibody feeding protocol. Surface receptors were labeled with myc antibody, left to internalize at 37°C for 30 min, and then fixed and blocked under nonpermeabilizing conditions. Surface receptors were labeled with Alexa Fluor 488-conjugated secondary antibody (green; surface), and, after permeabilization, all myc-tagged receptors were labeled with Alexa Fluor-594-conjugated secondary antibody (red; total), enabling the internalized receptors to be quantified (Fig. 4E). Quantification of the internalized pool of receptors (red fluorescence) showed a 72.0 ± 5.0% decrease in the endocytosis of the mutant mycβ3RRR/AAA-containing GABAARs compared with wild-type receptors (n = 3, **p < 0.01; Fig. 4E), suggesting that the RRR motif is essential for GABAAR internalization in neurons. Neurons expressing mycβ37A, mycβ37A+ILL/AAA, and mycβ3Δ395–414 mutants also exhibited a similar decrease in percentage endocytosis (mycβ37A, 78.14 ± 4.70%; mycβ37A+ILL/AAA, 62.14 ± 6.06%; mycβ3Δ395–414, 74.24 ± 5.00% decrease in the extent of receptor endocytosis compared with mycβ3WT control levels; n = 3, **p < 0.01; Fig. 4F, images not shown). In contrast, the mycβ3ILL/AAA mutant subunits were not significantly altered in their endocytosis compared with mycβ3WT-subunit-containing receptors (32.03 ± 26.23% decrease, p = 0.3). Therefore, in both COS-7 cells and neurons, the mutation of the RRR motif within the β3 subunit to alanine residues disrupts the endocytosis of GABAARs.
In agreement with these findings, whole-cell ELISAs (Bedford et al., 2001; Mielke and Wang, 2005) in cortical neurons expressing either mycβ3WT or mycβ3RRR/AAA constructs revealed that neurons transfected with the mycβ3RRR/AAA mutant displayed a significant 61.6 ± 8.0% increase in surface GABAARs compared with neurons expressing mycβ3WT receptors (n = 3, **p < 0.01). To support these data, we then used confocal microscopy to quantify the total area of GABAAR clusters in neurons expressing either the mycβ3WT or mycβ3RRR/AAA constructs. Similarly, the total area of GABAAR clusters in neurons expressing the mycβ3RRR/AAA mutant was 68.0 ± 28.2% greater than that of neurons expressing the mycβ3WT construct (n = 3, **p < 0.01; Fig. 4G). Therefore, mutation of residues (405RRR407) within the β3 subunit leads to steady-state accumulation of GABAARs at the cell surface and at synapses as a result of a decrease in clathrin-dependent endocytosis.
The observed alterations in surface levels and lateral mobility of mycβ3RRR/AAA mutant GABAARs could also possibly be attributable to the effects of NSF, an ATPase that binds to a potentially overlapping region of the β3 ICD as AP2 (Goto et al., 2005) and is implicated in GABAAR trafficking (Chou et al., 2010). To confirm that NSF function was not involved in the observed increased surface expression of mycβ3RRR/AAA mutant GABAARs, we treated COS-7 cells expressing α1mycβ3WT or α1mycβ3RRR/AAA GABAARs with an NSF inhibitory peptide (NSFi; fused to the TAT sequence), shown previously to increase GABAAR surface expression, or a scrambled control peptide (Morrell et al., 2005; Chou et al., 2010). ELISA experiments were then used to verify surface GABAAR levels. The NSFi peptide treatment caused a 35.3 ± 7.2% increase in surface α1mycβ3WT GABAARs compared with scrambled peptide treatment (Fig. 4H). A similar effect on NSF inhibition was also observed for surface α1mycβ3RRR/AAA GABAARs [47.7 ± 14.08% increase in surface α1mycβ3RRR/AAA GABAARs on NSFi treatment compared with scrambled peptide treatment (n = 4, *p < 0.05)]. These results confirm that the mycβ3RRR/AAA mutant still undergoes regulatory control by NSF-dependent trafficking mechanisms and that the observed increase in surface mycβ3RRR/AAA mutant expression is not attributable to altered regulation by NSF via this motif. NSF-dependent interactions were also unlikely to play a role in the observed alterations in surface GABAAR lateral mobility of mutant β3RRR/AAA receptors, because QD-imaging experiments revealed that blocking NSF activity had very little effect on the lateral mobility of GABAARs at the plasma membrane (data not shown). We therefore conclude that the effects of the RRR/AAA mutation on GABAARs at the neuronal cell surface are attributable to the specific disruption of the β3/AP2 binding site.
β3-subunit AP2 interactions are critical for GABAAR depletion from synapses during ischemic insult
An imbalance between excitatory and inhibitory synaptic transmission is thought to contribute to excitotoxicity and neuronal cell death during ischemic insult. Several studies have reported a decrease in surface and synaptic GABAAR expression during ischemia, suggesting an increase in receptor endocytosis (Schwartz-Bloom and Sah, 2001; Mielke and Wang, 2005; Zhan et al., 2006; Kelley et al., 2008; Arancibia-Cárcamo et al., 2009; Liu et al., 2010), but whether this is via an AP2-dependent pathway remains unclear.
To monitor the role of AP2 interactions in GABAAR trafficking, we used a super-ecliptic pHluorin (pH-sensitive GFP) N-terminally tagged γ2-subunit construct (SEPγ2), cotransfected with β3 subunits to visualize surface and synaptic GABAAR clusters under normal and ischemic conditions using live-cell imaging techniques. This cotransfection also enabled us to study effects of mutations within both γ2 and β3 subunits. Hippocampal neurons were cotransfected with SEPγ2, α1, and either the mycβ3WT or the AP2-binding-deficient mycβ3RRR/AAA or mycβ37A constructs. Quantification of the cotransfection efficiency of these DNAs revealed that 93.3 ± 3.3 and 96.6 ± 3.3% of SEPγ2-transfected cells were cotransfected with mycβ3WT or mycβ3RRR/AAA, respectively, indicating a high and similar level of cotransfection for these experiments (data not shown). Similar transfection efficiencies were also obtained for the mycβ37A construct (data not shown).
We imaged SEPγ2 and mycβ3WT cotransfected 12–15 DIV hippocampal neurons under control conditions or under conditions of OGD, an in vitro experimental model used to mimic cerebral ischemia (Kelley et al., 2008; Arancibia-Cárcamo et al., 2009). The fluorescence of individual SEPγ2 clusters in neuronal dendrites was imaged over a 30 min time period. Analysis of SEPγ2 cluster fluorescence in neurons cotransfected with mycβ3WT showed that the SEPγ2 cluster fluorescence remained stable under control conditions (Fig. 5A,B). However, during the OGD treatment, there was a significant 21.3 ± 4.0% reduction in surface cluster fluorescence over time compared with the control (n = 12, **p < 0.01; Fig. 5A,B). Using this approach, we then investigated whether this downmodulation was dependent on β3-subunit/AP2 interactions. In contrast to the mycβ3WT/SEPγ2-containing receptors, the surface SEPγ2 cluster fluorescence was not reduced on OGD treatment in neurons cotransfected with either the mycβ37A or mycβ3RRR/AAA mutant subunits with impaired endocytic ability (p = 0.71 and p = 0.36, respectively; Fig. 5C–F,I).
GABAARs that are unable to bind AP2 are not removed from synapses during OGD. A, C, E, Twelve to 15 DIV hippocampal neurons expressing SEPγ2 and mycβ3WT, mycβ3RRR/AAA, or mycβ37A were imaged over time under control or OGD conditions. Twenty micrometer sections of dendrite are shown. B, D, F, Fluorescent clusters in dendrites were monitored for loss of GFP fluorescence over time and normalized to t = 0. B, mycβ3WT control t = 30 min, 0.94 ± 0.04 (n = 12); mycβ3WT OGD t = 30, 0.74 ± 0.04 (n = 12, **p < 0.01). D, mycβ3RRR/AAA control t = 30 min, 0.87 ± 0.05 (n = 7); mycβ3RRR/AAA OGD t = 30, 0.89 ± 0.03 (n = 6, p = 0.71, nonsignificant). F, mycβ37A control t = 30 min, 0.95 ± 0.04 (n = 4); mycβ37A OGD t = 30, 1.02 ± 0.05 (n = 6, p = 0.36, nonsignificant). Arrows point to SEPγ2 clusters. G, H, Twelve to 15 DIV hippocampal neurons expressing SEPγ2AECL with either mycβ3WT or mycβ3RRR/AAA subunits were imaged over time under OGD conditions. G, mycβ3WT/SEPγ2AECL control, 0.90 ± 0.03, n = 5; OGD, 0.75 ± 0.05, n = 5; *p < 0.05. H, mycβ3RRR/AAA/SEPγ2AECL control, 0.89 ± 0.04, n = 5; OGD, 0.91 ± 0.05, n = 8; p = 0.77. Twenty micrometer sections of dendrite are shown. I, J, Bar graphs summarizing the fluorescent cluster analysis. Scale bars, 5 μm.
A GABAAR AP2-binding domain (YECL) in the γ2 subunit also exists (Kittler et al., 2008). In contrast, mutating the YECL AP2-binding domain in the γ2 subunit (to AECL) led to significant decrease in GABAAR cluster intensity under OGD conditions (*p < 0.05; Fig. 5G,J). However, in neurons cotransfected with mycβ3RRR/AAA and SEPγ2AECL, GABAARs were stable at the cell surface over the OGD time course (p = 0.77; Fig. 5H,J). Thus, the RRR motif within the β3 subunit is more important for control of AP2-mediated GABAAR internalization during OGD insult than the YECL motif on the γ2 subunit. Therefore, these experiments show that the interaction of the β3 subunit with AP2 plays a key role in the loss of synaptic GABAARs observed during ischemia. Furthermore, these data reveal that specific targeting of this GABAAR RRR AP2-binding motif could provide resistance for GABAARs against their accelerated endocytosis and removal from synaptic sites induced by ischemic insult.
Blockade of the β3/AP2 interaction reduces neuronal cell death during ischemic insult
We have revealed that preventing the interaction of GABAAR β3 subunit and AP2 can maintain surface GABAAR clusters during OGD. To investigate the functional relevance of this finding, we used a peptide overlapping the AP2 binding region in the β3 subunit to compete the β3/AP2 interaction in hippocampal neurons (β3 peptide). The β3 peptide and a control peptide (to a region of the β3 ICD distinct from the AP2 binding site) were transduced into 12–15 DIV hippocampal neurons for 1 h before OGD insult using a membrane-permeable carrier peptide (Twelvetrees et al., 2010). Delayed neuronal death was analyzed 24 h later using PI and DAPI staining, and the number of PI-positive nuclei were counted and the cell viability determined (Fig. 6A,B). OGD treatment of neurons transduced with the control peptide caused a 30.5 ± 1.6% decrease in cell viability (n = 3, ***p < 0.001) compared with neurons pretreated with the β3 peptide that exhibited an insignificant 6.4 ± 1.4% reduction in cell viability (n = 3, p = 0.09). These data show that specifically blocking the β3/AP2 interaction in neurons can prevent OGD-induced delayed cell death, strongly suggesting that GABAAR endocytosis via the RRR motif is important in mediating this process.
Blockade of GABAAR endocytosis reduces OGD-induced delayed cell death. Twelve to 15 DIV hippocampal neurons were transduced with control or β3 peptides for 1 h before a 40 min OGD treatment. Twenty-four hours later, neurons were stained with PI and DAPI, and the PI-positive nuclei were counted. A, Representative confocal images of control and β3 peptide transduced neurons under control and OGD conditions. Scale bar, 40 μm. B, Bar graph summarizing the pooled data (n = 3, ***p < 0.001).
Discussion
Regulated trafficking of postsynaptic neurotransmitter receptors is essential for determining the strength of synaptic transmission and for ensuring the correct balance of excitation and inhibition of neuronal circuits (Groc and Choquet, 2006; Arancibia-Cárcamo and Kittler, 2009). Tightly controlled clathrin-dependent endocytosis of surface GABAARs has been specifically shown to control synaptic efficacy in neurons, providing a fast mechanism to modulate synaptic strength and therefore synaptic inhibition (Kittler et al., 2000, 2008). Reduced inhibition attributable to alterations in GABAAR trafficking can also contribute to disrupted information processing and to the detrimental excitotoxicity thought to exacerbate neuronal cell death observed in conditions such as stroke and epilepsy (Smith and Kittler, 2010). In this study, we have investigated the internalization mechanisms of β3-subunit-containing GABAARs under normal and pathological conditions. We have identified a specific RRR motif (405RRR407) within the β3-subunit ICD that is essential for the endocytosis of GABAARs. We show that these residues mediate the direct interaction with μ2–AP2, although K354 within μ2–AP2 is also critical for this interaction. Mutation of the β3-subunit RRR motif reduces recruitment of GABAARs to EZs and impairs their endocytosis, leading to their steady-state accumulation at the neuronal cell surface and at synapses. We also show that downmodulation of GABAARs from dendritic clusters during OGD treatment (to simulate ischemia) is dependent on an AP2-dependent pathway for cell-surface removal of receptors. Moreover, blockade of this pathway can reduce the delayed neuronal death observed during ischemic insult.
We focused our study on the internalization of GABAARs containing the β3 subunit because of its high expression in the hippocampus and cortex and its presence in a large proportion of GABAAR subtypes found in these regions (Laurie et al., 1992; Wisden et al., 1992). To identify the molecular determinants of GABAAR internalization, we first identified the region of the β3-subunit ICD that mediated the interaction with the clathrin-adaptor protein AP2. This revealed a 20 aa stretch that harbored a seven residue atypical basic patch motif that, when mutated to alanine, prevented the binding of AP2 to the GABAAR β3-subunit ICD. Within this binding motif, additional mapping identified three arginine residues (405RRR407) that are essential for the interaction with μ2–AP2 and receptor internalization via an AP2- and clathrin-dependent mechanism. Because μ2–AP2 binds to GABAAR subunits β1–β3 and the RRR motif is conserved across all three β subunits, it suggests that these residues might be an internalization signal for all β-subunit-containing GABAARs. An arginine residue (R845) was also shown to be critical for AMPA receptor GluR2 subunit binding to AP2 (Lee et al., 2002; Kastning et al., 2007). Atypical basic patch motifs of both AMPA and GABAARs bind to the same region of μ2–AP2, subdomain B (Kastning et al., 2007; and this study), pointing to a similar clathrin-mediated internalization mechanism for both types of postsynaptic neurotransmitter receptors. In contrast, we did not find that a putative dileucine motif within the β3 subunit [as suggested for the β2 subunit (Herring et al., 2003)] was required for endocytosis or AP2 binding, suggesting that this motif is not a primary determinant of β3-subunit-mediated GABAAR internalization. Interestingly, the β3-subunit RRR motif is located adjacent to a phosphorylation site, Ser408/Ser409, which acts as a key site for GABAAR phosphomodulation by kinases (Brandon et al., 2002) and phosphatases (Jovanovic et al., 2004; Terunuma et al., 2008). Phosphorylation at this site has been shown to impair GABAAR endocytosis by preventing the interaction of the β3-subunit with AP2 (Kittler et al., 2005; Jacob et al., 2009). The addition of phosphate groups to Ser408/Ser409 would create a large negative charge that may disrupt the local charge interactions provided by the arginine residue cluster, thereby disrupting the AP2 binding site.
We have used QD tracking to measure the lateral mobility of GABAARs within EZs, specialized clathrin-rich regions along the dendritic plasma membrane, which mediate clathrin-dependent endocytosis (Blanpied et al., 2002). We found that QD-labeled GABAARs could be reversibly recruited to these clathrin-rich structures in which they displayed a reduced lateral mobility compared with receptors outside the EZ, suggesting that the receptors are being confined to these subdomains, before their endocytosis. This is similar to the confined behavior of GABAARs observed at the inhibitory postsynaptic site, with similar diffusion coefficient values for the mobility of the receptors previously recorded at the synapse to those observed at EZs (Bannai et al., 2009; Muir et al., 2010). This suggests that both inhibitory postsynaptic sites and EZs are specialized dendritic subdomains that share similar GABAAR confinement mechanisms via reversible interactions with submembranous synaptic or EZ proteins, such as gephyrin and AP2, respectively. In agreement with this, mutation of the RRR motif to alanine impaired the recruitment of GABAARs to EZs, as shown by the 3.3-fold increase in lateral mobility of non-AP2-binding mutant GABAARs at clathrin-coated pits compared with their wild-type counterparts. This increase in lateral mobility was accompanied by a significantly decreased residency time for the mutant receptors at EZs and a reduced likelihood of mutant receptors being endocytosed. Mutation of the β3-subunit AP2-binding motif did not alter the mobility or residency time of individual GABAARs at synaptic sites but did increase the steady-state levels of surface and synaptic receptors as a result of diminished receptor endocytosis. These results clearly demonstrate the key role of direct interactions of GABAAR ICDs with AP2 for recruiting and confining GABAARs at endocytic sites. This allows the trapping of mobile receptor populations at these membrane compartments within the plasma membrane to enable GABAARs to either mediate phasic inhibition at the synapse or undergo receptor endocytosis at the EZ.
During cerebral ischemia, overexcitation as a result of excessive glutamate release contributes to excitotoxicity and cell death in susceptible neuronal populations (Lipton, 1999; Lo et al., 2003). Research into therapeutics to treat ischemia has focused primarily on the glutamatergic system and the reduction of excess glutamatergic signaling during ischemic insult, but these have not proven to be clinically effective (Schwartz-Bloom and Sah, 2001). GABAergic transmission and GABAAR endocytosis are also significantly altered in ischemia, which may contribute to disinhibition and pathological excitability states. Because glutamatergic neurotransmission is counterbalanced by the GABAergic system, modulation of GABA metabolism and GABAAR function may be alternative targets of therapeutic intervention for ischemia. An in vitro model of cerebral ischemia, OGD, has been shown to modulate GABAAR trafficking, reducing surface expression of receptors (Mielke and Wang, 2005; Liu et al., 2010) and increasing ubiquitin-dependent lysosomal receptor degradation (Arancibia-Cárcamo et al., 2009). Previous studies using hypertonic sucrose treatment or dynamin inhibitors have suggested that pathological GABAAR endocytosis in OGD or status epilepticus is attributable to clathrin-dependent endocytosis (Goodkin et al., 2005; Mielke and Wang, 2005). However, hypertonicity has multiple cellular effects (Rosenmund and Stevens, 1996; Carroll et al., 1999), and dynamin inhibitors may block other (e.g., caveolar) endocytosis pathways (Nabi and Le, 2003; Doherty and McMahon, 2009). By imaging super-ecliptic pHluorin-tagged GABAARs, we show a decrease in synaptic GABAAR clusters during OGD treatment of hippocampal neurons for 30 min, which is consistent with other studies (Mielke and Wang, 2005; Arancibia-Cárcamo et al., 2009; Liu et al., 2010). Using the β3-subunit RRR motif mutant, we then directly tested the pathway of pathological GABAAR endocytosis during OGD and showed that acute loss of synaptic GABAARs during OGD is mediated by an AP2 and clathrin-dependent internalization pathway. Furthermore, blocking this internalization pathway with a competing β3 peptide can reduce delayed cell death after OGD insult, demonstrating the physiological relevance of this mechanism in neuronal function.
We also find that the tyrosine motif within the γ2 subunit has little effect on endocytosis of GABAARs during OGD, suggesting that the β3-subunit RRR motif is the primary determinant of internalization during simulated ischemic insult. This is probably attributable to the different binding affinities of the subunits for AP2 and their regulation by different kinase families (Kittler et al., 2005, 2008), allowing multiple signaling pathways to converge at these receptor ICDs, to enable the fine-tuning of endocytosis modulation in response to a variety of extracellular signals.
In summary, we have identified a triple arginine motif that mediates direct binding of the GABAAR β3 subunit to the clathrin adaptor protein AP2. Mutation of this motif impaired receptor recruitment to clathrin-coated pits, significantly reduced receptor endocytosis, and caused an accumulation of surface GABAARs in neurons. This AP2 pathway is likely to play a key role in regulating synaptic receptor number during inhibitory synaptic plasticity. In ischemia, synaptic GABAARs are also depleted by this AP2/clathrin-dependent pathway. This internalization pathway might therefore also be targeted therapeutically to maintain surface GABAAR populations and synaptic inhibition during pathological insults such as epilepsy and ischemic neuronal death.
Footnotes
This work was supported by the United Kingdom Medical Research Council (J.T.K.) and by the German funding agency Deutsche Forschungsgemeinschaft [Exc-257, HA2686/3-2 (FG806) (V.H.)]. J.M. and D.F.S. are in the CoMPLEX PhD Programme at University College London. We thank David Attwell for the use of his confocal microscope.
- Correspondence should be addressed to Dr. Josef Kittler, Department of Neuroscience, Physiology, and Pharmacology, University College London, Gower Street, London, WC1E 6BT, UK. j.kittler{at}ucl.ac.uk
