Ezrin Mediates Tethering of the γ-Aminobutyric Acid Transporter GAT1 to Actin Filaments Via a C-Terminal PDZ-Interacting Domain

Molecular biology

The green fluorescent protein (GFP) tagged mGAT1, mGAT1₀-GFP construct was previously described, and was termed mGAT1-GFP (8). In GAT1₀-GFP, added C-terminal residues (a linker followed by GFP) disrupt the endogenous PDZ-interaction domain of GAT1 (C-terminal sequence, −AYI-CO₂⁻).

To generate the new fluorescent mutant, GAT1-YFP8, an intact PDZ-interaction motif is added at the C-terminus, following the yellow fluorescent protein (YFP) moiety. The wild-type mGAT1 open reading frame (ORF) is subcloned without its endogenous stop codon into the HindIII and EcoRI sites of the pcDNA3.1(+) expression vector (Invitrogen, Carlsbad, CA) multiple cloning site (MCS). The YFP (Clontech, Mountain View, CA) ORF into which we introduce the “monomeric” A206K mutation (15) is then inserted in the frame with the mGAT1 ORF at the NotI and XbaI sites of the pcDNA3.1(+) MCS. This results in a 12-amino-acid spacer between the end of the mGAT1 sequence and the beginning of the fluorophore. The method of Geiser et al. for the integration of polymerase chain reaction (PCR) fragments without the use of restriction enzymes was modified, to amplify the final eight codons of the hGAT1 C-terminal sequence from a source plasmid, using PfuTurbo C_x Hotstart polymerase (Stratagene, La Jolla, CA) with 5′ and 3′ extensions that corresponded to the 20–22 nucleotide regions that flank the intended site of insertion (16). The PCR product from this first PCR is integrated in the frame immediately after the YFP sequence, when used as the primers in a subsequent QuikChange II XL mutagenesis PCR reaction (100-ng product from the first PCR reaction: 20 ng destination plasmid; Stratagene).

The cyan and yellow fluorescent protein (CFP and YFP respectively) tagged ezrins, ezrin-CFP and YFP-ezrin, gifts from Dr. Forte and Lixin Zhu, were described by Zhu et al. (17). The YFP-syntaxin, a gift of Dr. W. Almers, was described by An and Almers (18).

Neuroblastoma-2a culture

Mouse neuroblastoma-2a (N2a) cells (ATCC, Manassas, VA) are grown at 37°C in 95% air, 5% CO₂, in N2a culture medium containing the following: 44.5% DMEM, 44.5% OptiMEM, 10% fetal bovine serum, and 1% penicillin/streptomycin (10,000 IU penicillin, 10,000 mg/mL streptomycin) (Invitrogen). After cells are grown to confluence, they are plated onto glass-bottom dishes at a density of 3.5 × 10⁵ cells/dish. Before plating, 14-mm glass-bottom dishes (No. 0, Mattek, Ashland, MA) are coated with 0.05% polyethylimine (PEI) (Sigma, St. Louis, MO), pH 8.4, in borate buffer. The transfection of a total of 1 μg DNA per plate is performed 16–24 h later with Lipofectamine and Plus reagent (Invitrogen). The composition of the N2a imaging solution is (mM): 128 NaCl, 2.4 KCl, 25 HEPES, 1.2 MgCl₂, 3.2 CaCl₂, 1.2 KH₂PO₄, and 10 D-glucose (19).

Nocodazole treatment

Nocodazole (Sigma, St. Louis, MO) was dissolved in DMSO at a concentration of 5 mg/mL, and stored at −20°C. We added 10 μM nocodazole to the N2a culture medium, and cells were incubated for 16–24 h at 37°C. This concentration and incubation time followed established protocols for maximally depolymerizing microtubules (20–22). Before imaging, N2a culture medium was aspirated from cells, and cells were washed with imaging solution containing 10 μM nocodazole. Cells were allowed to equilibrate at room temperature for 30 min in the 10 μM nocodazole containing imaging solution.

Latrunculin B and cytochalasin D treatment

Latrunculin B (Sigma-Aldrich, St. Louis, MO) was dissolved in DMSO and stored at −20°C at a stock concentration of 2.5 mM. We added 5 μM latrunculin B to the imaging solution, which was warmed to 37°C before imaging. Cells were washed with latrunculin B solution, and incubated for 1 h in the solution before imaging. Cytochalasin D (Sigma, St. Louis, MO) was dissolved in DMSO and stored at −20°C at a stock concentration of 1 mg/mL. We added 1 μg/mL cytochalasin D to the imaging solution, and warmed it to 37°C. Cells were washed with the cytochalasin D solution, and were incubated in the solution for 1 h before imaging.

GABA uptake

Uptake assays were performed using 12-well plates 48 h after transfection as follows. Culture medium was aspirated, and cells were equilibrated for 10 min at room temperature in 1 mL Krebs-Ringer (KRH) buffer containing (in mM): 130 NaCl, 1.3 KCl, 2.2 CaCl₂, 1.2 MgSO₄, 1.2 KH₂PO₄, 10 D-glucose, and 10 HEPES, pH 7.4 (23). GABA uptake was initiated on aspiration of the KRH buffer and the addition of 500 μL 2.5 μM GABA and 25 nM [³H]GABA in KRH at room temperature. Uptake was terminated by four rapid washes with KRH + GABA, using the same total GABA concentration as during the uptake step. Cells were lysed in 1 mL 2% SDS in phosphate-buffered saline (PBS). The [³H]GABA accumulation was assayed by liquid scintillation spectrometry. Specific uptake was normalized to the total protein in each well, as determined using the BCA protein assay (Pierce; Rockford, IL). Actin disruption was performed by incubating cells with 5 μM latrunculin B for 1 h at 37°C before uptake. We also included 5 μM latrunculin B in both the KRH equilibration and the [³H]GABA buffers.

Microtubule and nuclear labeling with TubulinTracker Green and Hoechst 33342

Microtubules were labeled according to the instructions of the supplier, using TubulinTracker Green reagent for live-cell tubulin labeling (Molecular Probes, Eugene, OR). Briefly, N2a nucleus was stained by the addition of 10 μg Hoechst 33342; microtubules were stained with TubulinTracker in Hanks' Balanced Salt Solution by incubating for 30 min at 37°C in a 5% CO₂ incubator. Cells were rinsed three times with warm imaging solution before imaging.

Actin labeling with rhodamine phalloidin

A 14-μM rhodamine-conjugated phalloidin (Cytoskeleton, Denver, CO) stock solution was prepared by dissolving the solid in methanol, and storing at −20°C. A fresh 4% solution of paraformaldehyde (Fluka, St. Louis, MO) at pH 7.4 was prepared. The medium was aspirated from the dishes, and cells were rinsed twice with 1 × PBS. Fixation and staining were performed by combining 4% paraformaldehyde, 0.1% Triton, and 2 units (5 μL) rhodamine phalloidin per dish (10 min, room temperature). After fixation, cells were washed twice with 1 × PBS.

Confocal imaging

All FRAP experiments were performed on an inverted Zeiss LSM 510 Meta (Carl Zeiss, Thornwood, NY), using a 63× Plan-Apochromat 1.4-NA oil-immersion objective. Fluorescence was collected using an HFT KP 700/514 nm beam-splitter (Carl Zeiss) and a 535–590-nm bandpass filter. The detector gain was modified to allow for near-saturating conditions for a given region of interest (ROI). There was no change to gain settings for a given ROI. For footprint photobleach, a circular ROI was chosen with an area of 9.4 ± 0.9 μm². For perimeter photobleach, a rectangular ROI was chosen with an area of 12.5 ± 1.2 μm². The ROIs were photobleached at 25% laser transmission, using the 514-nm laser line of a 40-mW argon laser. Photobleaching was optimized to obtain ∼80% photodestruction within the ROI. Prebleach and recovery images were collected at 0.25% laser transmission at 5-s intervals. Twelve-bit images were collected at a pixel resolution of 512 × 512. The pinhole diameter was set at 144 μm.

Dye-labeled N2a cells were imaged on an inverted Zeiss LSM 510 Meta at 12 bits and a 512 × 512 pixel resolution. Stack slices were at 100-nm intervals. Hoechst 33342 nuclear-labeled cells were imaged using a Plan-Apochromat 100× 1.4-NA oil-immersion objective, and were excited using a two-photon laser tuned to 700 nm at 2.5% laser transmission. Fluorescence was collected using the HFT KP 700/488 beam-splitter and FT 510 and BP 390-465 IR filters (Zeiss). Stack slices were at 100-nm intervals, and the pinhole diameter was maximal (693 μm). TubulinTracker Green-labeled N2a cells were imaged using a 100× 1.4-NA oil-immersion objective. Cells were excited with a 488-nm argon laser at 2.5% laser transmission. Fluorescence was collected with the HFT 488 beam-splitter and NFT 490 and BP 500-550 IR filters. Stack slices were at 100-nm intervals, and the pinhole diameter was 213 μm. Rhodamine phalloidin-labeled N2a cells were imaged with a Plan-Neuofluar 63× 1.25 oil-immersion objective. Cells were excited with a 543-nm helium neon laser at 7.75% laser transmission. Fluorescence was collected using a 488/543 beam-splitter and NFT 545 and bandpass 565–615 filters. The zoom was 2×, stack slices were at 550-nm intervals, and the pinhole was 121 μm.

The FRET experiments were performed on a Nikon C1si laser-scanning confocal microscope system (Nikon Instruments, Melville, NY), equipped with spectral imaging capabilities and a Prior (Rockland, ME) remote-focus stage, as described in Drenan et al. (24). Spectral FRET analysis was performed as described in Nashmi et al. (25). Data are reported as mean ± SE.

FRAP analysis

Raw data were processed using algorithms written in-house (MATLAB, The MathWorks, Natick, MA) to obtain the ratio fluorescence = $f\left(t\right)-b\left(t\right)/\left(f_{NB}\left(t\right)-b\left(t\right)\right)\times \left(f\left(t_o\right)-b\left(t_o\right)\right)$. This expression was obtained by subtracting the background fluorescence, b(t), correcting for the photobleaching that occurs during recovery scanning by obtaining the fluorescence in a region not bleached, f_NB (t), and normalizing to the prebleach fluorescence intensity f(t_o). Normalized traces were averaged. Error bars are reported as the standard error of the mean. Normalized data were fitted to a double-exponential, five-parameter regression in SigmaPlot (Systat Software, Inc., San Jose, CA) of the form $F\left(t\right)=F_{bo}-a{e}^{-bt}-c{e}^{-dt}$, where F_bo is the initial value of fluorescence after photobleaching, and a, b, c, and d are constants obtained through fitting. The mobile fraction was determined by applying regression data to the equation $M_f=\frac{F_{\mathrm{\infty }}-F_{bo}}{1-F_{bo}}$, where M_f is the mobile fraction and F_∞ is the asymptote of the recovery curve. Values for $t_{\frac{1}{2}}$ were determined by applying regression data to the equations $F_h=\frac{F_{\mathrm{\infty }}+F_{bo}}{2}$, $t_{\frac{1}{2}}=t_h-t_{bo}$, where F_h is the value for fluorescence when half the recovery has occurred, t_h is the corresponding time for F_h, t_bo is the time after photobleaching when the first recovery image is obtained, and $t_{\frac{1}{2}}$ is the halftime of equilibration (26).

Real-time quantitative reverse-transcription PCR

Total RNA was prepared using the RNeasy Plus Mini Kit (Qiagen, Valencia, CA). The RNA concentrations were quantified using the ND-1000 UV-Vis Spectrophotometer (Nanodrop Technologies, Wilmington, DE). The primer and probe sets for Pals1, ezrin, β-actin, and γ-actin were designed by the Roche Universal Probe Library (UPL) Assay Design Center (Roche Applied Science, Indianapolis, IN). Primer sets were obtained from Integrated DNA Technologies (Coralville, IA). Table 1 displays the primer and probe sets used for the real-time quantitative reverse-transcription PCR (qRT-PCR) analysis of each gene in the N2a cells. One-step qRT-PCR was performed using a hydrolysis probe mix, 96-well plates, and the LightCycler 480 RNA Master system (Roche Applied Science), according to the supplier's instructions. Briefly, samples along with serially diluted mRNA (1:50, 1:100, 1:1000, 1:5000, and 1:10,000) were added to adjacent wells of the plate, centrifuged for 2 min at 1500 × g, and loaded into the instrument. The one-step qRT-PCR was performed as follows: one cycle of reverse transcription at 61°C for 12 min; one cycle of denaturation at 95°C for 30 s; 45 cycles of amplification at 95°C for 10 s, 57°C for 30 s, and 72°C for 1 s; and one cycle of cooling at 50°C for 10 s. Serial dilutions of mRNA provided the standard curves needed to quantify relative mRNA concentrations.

Statistical analysis

Values are expressed as mean ± SE, unless otherwise noted, and p < 0.05 is considered statistically significant.

Whole-footprint photobleach reveals lateral mobility

The question often arises whether FRAP actually detects lateral mobility, or whether it detects the fusion of vesicles that are actively trafficking protein onto the membrane. To address this question, we performed “whole footprint” FRAP. Fig. 1 A gives prebleach, postbleach, and recovery images of a whole-footprint FRAP experiment. The postbleach image shows that >90% of the footprint surface area is photobleached, and includes a line profile across the bleached region. The ends of the line overlap nonphotobleached regions. The kymograph in Fig. 1 B represents the line profile, as shown in Fig. 1 A, over time, and it shows that recovery occurs from regions not photobleached, with fluorophore entering from the edges. This qualitative analysis indicates that FRAP largely reveals the lateral mobility of GAT1. However, it is possible that the speckles that appear in the image after ∼60 s represent some contribution from vesicle exocytosis (9). In separate experiments, we visualized and tracked individual vesicles containing GAT1; these will be reported in a future study.

Fitting data to FRAP diffusion model

The diffusion model in Fig. 1 C is for a circular bleach spot, derived by Axelrod et al. (27), and modified by Sprague et al. (28) and Sprague and McNally (29). The model is as follows:

Based on the Axelrod-Sprague equation, γ is a constant equal to 0.88 for a circular beam with a Gaussian intensity profile, τ_D is the recovery time constant, ω is the radius of the bleach spot, D_eff is the effective diffusion coefficient, and A represents the recovery asymptote for GAT1-YFP8. Therefore, σ is a correction that allows the diffusion model to converge to the GAT1-YFP8 recovery asymptote. The Axelrod-Sprague equation does not accurately describe the behavior of GAT1-YFP8 recovery, suggesting that the mobility of the transporter is not governed by pure diffusion. In particular, the model exhibits sharp recovery at early time points. The recovery of GAT1-YFP8 is well-fitted as double-exponential decay (as described in Materials and Methods). Because the double-exponential fitting more accurately describes GAT1-YFP8 mobility on the time scale (∼150 s) of these experiments, t_1/2 values are used in this study, rather than diffusion coefficients.

Comparing and quantifying GAT1 mobility

To give a context to GAT1 mobility, we compare it with minimally mobile and maximally mobile proteins, represented respectively by soluble YFP and by YFP-syntaxin1A, a t-SNARE membrane protein (Fig. 2, A and B). Footprint photobleach is chosen, because YFP is not a membrane protein, and cannot be meaningfully analyzed with a perimeter photobleach. However, its mobility near the membrane can be obtained by focusing near the coverslip. Within the first seconds of detection, there is steep, unresolved recovery of YFP to ∼60% of its prebleach intensity; this limits the quantitative analysis of the YFP mobile fraction (Fig. 2 C). However, a qualitative observation of Fig. 2 C shows that the YFP recovery asymptote is greater than that of GAT1-YFP8, and by extrapolating the initial YFP photobleach to 20% (GAT1-YFP8 F_bo), we infer that YFP has a higher mobile fraction than GAT1-YFP8. As shown in Fig. 2 B, and as previously reported, YFP-syntaxin clusters to form exocytosis docking sites at the membrane, so it is not expected to have high membrane mobility (30). YFP-syntaxin has a mobile fraction of 65%, comparable to the GAT1-YFP8 mobile fraction of 60% (Fig. 2 C).

Photobleach regions

Variations in imaging parameters, such as laser intensity, bleach size/shape, and pixel dwell time, can affect the results of quantitative photobleach experiments (31). To provide additional confirmation of the FRAP results, GAT1 experiments were performed by photobleaching the cell footprint and perimeter, and comparing the trends seen between these regions of interest (Fig. 3, A and B). The approximate size for footprint photobleach is a circle with an area of 8 μm² (Fig. 3 A). Perimeter photobleach is performed by focusing the laser on a plane vertically midway through the cell, and a rectangular section, including the cell membrane (with an area of 13 μm²), is photobleached (Fig. 3 B). Perimeter photobleach recovery occurs in two dimensions, and the membrane fluorescence in the confocal slice is a line representation of this recovery. Footprint and perimeter photobleach revealed GAT1-YFP8 mobile fractions of ∼60% and 50%, respectively (Fig. 3 C). These results suggest that 40–50% of GAT1-YFP8 is immobilized on the time scale (∼150 s) of these experiments. Because a significant fraction of surface GAT1 is immobile, we explored the possibility of GAT1 tethering to the cytoskeleton.

GAT1 interacts with cytoskeleton via actin but not via microtubules

Fig. 4 A (left) displays microtubule distribution in an N2a cell labeled with TubulinTracker 488. Fig. 4 A (right) includes nuclear labeling with Hoechst 33342. Applying nocodazole for 16–24 h maximally disrupts microtubules (20) (Fig. 4 B). The FRAP traces show that nocodazole treatment has no significant effect on either the recovery time constant or the size of the mobile population (Fig. 4, C and D).

Rhodamine phalloidin labeling of actin filaments in N2a cells confirms the actin disruption (Fig. 4, E and F). Modest differences were evident between the untreated and the latrunculin B footprint photobleach traces, which translate to a 20% increase in the mobile fraction (Fig. 4, G and I). More pronounced differences in the untreated and latrunculin B perimeter photobleach traces reveal a 60% increase in the time constant for recovery, and a 20% increase in the mobile fraction (Fig. 4, H–J). Cytochalasin D, another actin disruptor, confirms the latrunculin B results, with an increased t_1/2 and an increased mobile fraction compared with the untreated samples (Fig. 4, I and J).

Disrupting PDZ binding increases GAT1 lateral mobility

Typically, actin binding occurs through an adaptor protein, or by an interaction with a scaffolding protein such as a PDZ protein (13,32–34). To determine whether PDZ binding links GAT1 to the actin cytoskeleton, we performed quantitative photobleaching on GAT1₀-GFP, a fluorescent GAT1 that cannot interact with PDZ proteins. In GAT1₀-GFP, added C-terminal residues (a linker followed by GFP) interrupted the endogenous PDZ-interaction domain of GAT1 (C-terminal sequence, −AYI-CO₂⁻). As a result, homozygous GAT1₀-GFP knock-in mice showed a 70% reduction in surface localization and synaptosomal GABA uptake (8). Fig. 5 A illustrates the difference between the two fluorescent GABA transporter constructs in this study: GAT1₀-GFP lacks an intact PDZ-interacting domain, whereas we inserted a PDZ-interaction domain to GAT1-YFP8, downstream of the YFP moiety.

FRAP was performed on N2a cells transfected with GAT1₀-GFP. The GAT1₀-GFP was 30–40% more mobile on the membrane than GAT1-YFP8, indicating that the PDZ interaction domain contributes to GAT1 immobilization (Fig. 5 B). Latrunculin B treatment shows that depolymerizing actin does not significantly affect the mobile fraction (Fig. 5 B). Because the PDZ-binding domain of GAT1₀-GFP was disrupted and this led to no interaction with actin, we conclude that a PDZ-interacting domain in GAT1 is necessary, either directly or indirectly, for GAT1-actin interactions.

Actin disruption increases GABA uptake

Disruption of actin increases Na⁺-dependent GABA uptake in BGT1, a GABA transporter subtype that, like GAT1, is expressed in the central nervous system, but unlike GAT1, is also expressed in the kidney (35,36). To determine whether actin disruption can increase GABA uptake by GAT1, we performed GABA uptake on both actin-disrupted and nontreated cells. Actin disruption increased the GABA uptake of both GAT1-YFP8 and the poorly trafficking GAT1₀-GFP (Fig. 5 C). These statistically significant increases amounted to a 3.4-fold increase in GABA uptake by GAT1-YFP8, and a 3.1-fold increase in the much lower GABA uptake by GAT1₀-GFP. Both increases were relative to nonlatrunculin B-treated controls.

Ezrin and Pals1 are expressed in N2a cells

The FRAP results indicate that GAT1 interacts with the cytoskeleton through the PDZ-binding motif. Another membrane protein with this type of interaction is the Na⁺/H⁺ exchanger isoform 3 (NHE3), which is restrained to actin through interactions with the PDZ protein, the Na⁺-H⁺ exchanger regulatory factor (NHERF), and the adaptor protein ezrin (37). To determine whether ezrin could link GAT1 to the cytoskeleton, we performed qRT-PCR, testing for the expression of two genes: the PDZ protein known to interact with GAT1, Pals1 (38), and the linker molecule ezrin. The N2a cells expressed both ezrin and Pals1 mRNA at similar levels (Fig. 6 A). The mRNA expression was normalized to the β-actin control; γ-actin was also used as a control, with expression levels similar to those of β-actin (Fig. 6 A).

FRET suggests a GAT1-ezrin interaction

Figs. 6, B–H, display the results of FRET experiments that combined donor recovery after acceptor photobleach with spectral unmixing. After the photodestruction of GAT1-YFP8 with a 514-nm argon laser, there was a 23% increase in ezrin-CFP fluorescence (Fig. 6, B and E). This increase was seen most prominently along the cell membrane (Fig. 6 B). The FRET efficiency observed between GAT1-YFP8 and ezrin-CFP was 19% ± 2% (Fig. 6 F).

Inactive ezrin forms C-terminal to N-terminal dimers in the cytosol (17,39). Therefore, the previously studied FRET partners, ezrin-CFP/YFP-ezrin (17), were used as a positive control. As expected, ezrin-CFP and YFP-ezrin were localized in the cytosol, and exhibited a FRET efficiency of 20% ± 2% (Fig. 6, D and F).

Actin disruption diminishes GAT1-ezrin interaction

GAT1-YFP8 photodestruction resulted in little or no increase in ezrin-CFP fluorescence when actin was disrupted (Fig. 6, C and E). Latrunculin B exposure produced a FRET efficiency of 7% ± 5%, i.e., significantly lower than the FRET value for both untreated and ezrin-CFP/YFP-ezrin control samples (Fig. 6 F). These results indicate that actin depolymerization greatly destabilizes the interaction between GAT1 and ezrin.

There is no GAT1₀-ezrin interaction

If the interaction between GAT1 and ezrin requires a PDZ-binding motif, we expected to observe no such interaction between ezrin and a GAT1 molecule that lacks such a motif. GAT1₀-YFP provided an appropriate test. As expected, no detectable increase in ezrin-CFP fluorescence accompanied the photodestruction of GAT1₀-YFP (Fig. 6, G and H). This result supports the idea that the interaction between ezrin and GAT1 requires the binding motif.

Physiological significance of GAT1 movements in the membrane

One functional consequence of a disrupted actin cytoskeleton is increased GABA uptake, as evidenced by the several-fold increase in GABA uptake after actin depolymerization. Although Fig. 5 B also shows that depolymerizing actin does not significantly affect the mobile fraction of GAT1₀-GFP, Fig. 5 C shows that this treatment does increase the relatively small GABA uptake of GAT1₀-GFP, and by roughly the same ratio as for GAT1-YFP8. This increase in GABA uptake upon actin depolymerization suggests that actin plays an additional role in modulating transporter function. Further experiments beyond the scope of this study are required to determine whether the additional modulation occurs via a syntaxin isoform that interacts with both GAT1 and F-actin (53,54).

The GAT1 link to actin via ezrin may point to the role of this interaction in the pathophysiology and/or treatment of epilepsy. Although most studies of ezrin are focused on its role in epithelial structure, organization, and signaling, ezrin is also expressed in central nervous system neurons, and is increased, at both the mRNA and protein levels, in certain epilepsies. Ezrin is up-regulated in the hippocampus of patients with mesial temporal-lobe epilepsy (55,56). Likewise, in the lithium-pilocarpine-treated rat model of temporal-lobe epilepsy (57), ezrin is up-regulated within 48 h of seizure induction (58). Epilepsies associated with malformations of cortical development, such as focal cortical dysplasia (FCD_IIb and FCD_IIa) and gangliogliomas, are also marked by neuronal ezrin accumulation (59). Although the role of ezrin up-regulation in seizure progression has not been formally identified, recent evidence points to the role of ezrin as a downstream target in the phosphoinositide 3-kinase pathway (59,60).

The identification of interactions between GAT1, ezrin, and actin, and our quantitative study using FRAP, contribute to a further understanding of how transporters are clustered, regulated, and confined on the cell surface. Because GAT1 is an important drug target in the treatment of epilepsy, continued research is needed to determine whether ezrin up-regulation and GAT1 tethering play a neurotherapeutic role in epilepsy. Such protein interactions may lead to the discovery of additional therapeutics targeting GAT1 function (61–64).