NMDA receptors regulate GABA_A receptor lateral mobility and clustering at inhibitory synapses through serine 327 on the γ2 subunit

Abstract

Modification of the number of GABA_A receptors (GABA_ARs) clustered at inhibitory synapses can regulate inhibitory synapse strength with important implications for information processing and nervous system plasticity and pathology. Currently, however, the mechanisms that regulate the number of GABA_ARs at synapses remain poorly understood. By imaging superecliptic pHluorin tagged GABA_AR subunits we show that synaptic GABA_AR clusters are normally stable, but that increased neuronal activity upon glutamate receptor (GluR) activation results in their rapid and reversible dispersal. This dispersal correlates with increases in the mobility of single GABA_ARs within the clusters as determined using single-particle tracking of GABA_ARs labeled with quantum dots. GluR-dependent dispersal of GABA_AR clusters requires Ca²⁺ influx via NMDA receptors (NMDARs) and activation of the phosphatase calcineurin. Moreover, the dispersal of GABA_AR clusters and increased mobility of individual GABA_ARs are dependent on serine 327 within the intracellular loop of the GABA_AR γ2 subunit. Thus, NMDAR signaling, via calcineurin and a key GABA_AR phosphorylation site, controls the stability of synaptic GABA_ARs, with important implications for activity-dependent control of synaptic inhibition and neuronal plasticity.

Synaptic inhibition plays a critical role in regulating neuronal excitability and information processing in the brain. The number of GABA_A receptors (GABA_ARs) in the surface membrane and at synaptic sites is an important determinant of inhibitory synapse strength (1), but the mechanisms that rapidly control synaptic GABA_AR number and stability remain poorly understood. Activation of Ca²⁺-permeable ionotropic glutamate receptors (GluRs) during plasticity and in pathology can result in down-modulation of inhibitory synapse strength and GABA_AR function (2–5) but the molecular and cellular mechanisms underlying GluR-dependent changes in the strength of GABAergic inhibition remain unclear.

A major mechanism for modulating GABA_AR activity is the direct phosphorylation of residues within the intracellular loops of GABA_AR subunits, which can regulate synaptic inhibition, GABA_AR channel kinetics, and trafficking (6–9). The rapid movement of neurotransmitter receptors (including GABA_ARs) (10–12) into and out of synapses has also recently emerged as an important mechanism for regulating synaptic strength (13). However, whether GABA_AR phosphorylation can directly regulate the synaptic stability of GABA_ARs and their lateral diffusion and movement into and out of synapses is unknown.

Here, by live cell imaging of surface GABA_AR clusters with pH-sensitive superecliptic pHluorin (SEP) and single GABA_ARs with quantum dots (QDs), we investigate the mechanisms that regulate activity-dependent control of the lateral diffusion, clustering, and stability of GABA_ARs at inhibitory synapses. We find that Ca²⁺ entry through NMDA receptors (NMDARs) leads to a reversible dispersal of GABA_AR clusters and an increase in GABA_AR lateral mobility. This mechanism can also allow for a localized GABA_AR dispersal in dendrites when only a few excitatory synapses are activated. The dispersal of GABA_ARs from synapses requires activation of the phosphatase calcineurin, and is dependent on serine 327, a key phosphorylation site in the GABA_AR γ2 subunit (14). Thus, activity-dependent control of phospho-dependent signaling can rapidly regulate the number of synaptic GABA_ARs with important implications for inhibitory synaptic plasticity, pathology, and information processing in the brain.

Results

GABA_AR Clusters Are Stable Under Basal Conditions, Although Single GABA_ARs Can Be Highly Mobile.

To image surface GABA_AR clusters in the dendrites of live cultured hippocampal neurons, we expressed SEP-tagged GABA_AR α2 subunits (α2^SEP-GABA_ARs) (Fig. S1A) (15), the majority of which formed bright fluorescent clusters along the neuronal dendrites in addition to lower levels of diffuse staining (Fig. 1A). Heterologously expressed α2-SEP subunits could not access the cell surface without the presence of coexpressed GABA_AR β- and γ2 subunits (Fig. S1 B and C) and thus cannot form homomeric surface GABA_ARs. In agreement with this, immunoprecipitating native GABA_ARs from α2-SEP subunit transfected neurons followed by Western blotting with GFP antibodies confirmed that α2-SEP subunits assembled with endogenous GABA_AR subunits, and the lack of any cleavage products confirmed that the fluorescence from α2^SEP-GABA_ARs was reporting only full-length α2-SEP subunits (Fig. S1D). α2^SEP-GABA_AR fluorescence was rapidly and reversibly eclipsed by transient exposure to low extracellular pH buffer, which completely eclipsed the signal (Fig. S1 E–H), confirming surface expression of the clusters. A large fraction of α2^SEP-GABA_AR clusters (84.3 ± 1.8%) were localized to inhibitory synapses marked by GAD-6 staining (Fig. S1I) and 65.3 ± 5.2% were apposed to FM4-64 labeled presynaptic terminals (Fig. S1J, in agreement with previous reports for the proportion of native GABA_AR clusters found opposite active FM dye-labeled presynaptic terminals) (16, 17). However, only 9.1 ± 1.3% were found apposed to excitatory synapses labeled with homer1 (Fig. S1K). Thus, α2-SEP fluorescence reports surface heteromeric GABA_ARs that accumulate in clusters at inhibitory synapses.

Fig. 1.

Surface GABA_AR clustering under basal conditions. (A) α2^SEP-GABA_ARs expressed in cultured hippocampal neurons. (B) Kymograph (a line scan vertically projected over time) showing stability of α2^SEP-GABA_AR clusters in boxed region in A over the movie (duration 10 min). (C) Region boxed in A shown at t = 0 (Top), as a maximum projection over time (Middle) and with cluster trajectories shown (Bottom) (3× zoom). (Scale bars, 2 μm.) (D) Trajectory of a QD-tagged α2^SEP-GABA_AR (black, origin marked by gray arrowhead) undergoing lateral diffusion into and out of an α2^SEP-GABA_AR cluster (green). (E) Kymograph for trajectory in D. Note the period of low horizontal displacement corresponding to confined motion within cluster (position marked by green line). (Scale bar, 2 μm.) (F) α2^SEP-GABA_AR cluster diffusion coefficients (orange) and instantaneous diffusion coefficients for QD-α2^SEP-GABA_ARs inside (green) and outside of clusters (black). Single GABA_ARs are less mobile when within GABA_AR clusters [P = 4 × 10⁻¹³, Kolmogorov–Smirnov (K-S) test, n_in= 598, n_out= 3665].

Under resting conditions, continuous imaging revealed that the intensities and locations of α2^SEP-GABA_AR clusters were stable for periods of 10 min (Fig. 1 A–C) although interestingly, some clusters exhibited small, asynchronous lateral movements (Fig. 1 B and C). Tracking the mass centers of individual clusters gave trajectories that were used to determine their diffusivities. The low mean diffusivity of α2^SEP-GABA_AR clusters (6 × 10⁻⁴ μm²·s⁻¹, Fig. 1F and Movie S1) is similar to that previously reported for clusters of the inhibitory postsynaptic scaffold gephyrin (18), suggesting that α2^SEP-GABA_AR cluster movement may represent movement of the entire inhibitory postsynaptic apparatus.

Although imaging of α2^SEP-GABA_ARs allowed the behavior of synaptic GABA_AR clusters to be followed, it could not resolve the behavior of individual GABA_ARs. To simultaneously visualize synaptic GABA_AR clusters and single GABA_ARs, we combined SEP tagging with QD tracking (13, 19, 20). Single α2^SEP-GABA_ARs were labeled using a low concentration of GFP antibody to recognize the extracellular SEP tag on the α2 subunit in conjunction with an anti-mouse secondary coupled to 605-nm QDs (Fig. S2A). In contrast to the relative immobility of α2^SEP-GABA_AR clusters, single QD-labeled α2^SEP-GABA_ARs could be observed to rapidly diffuse into and out of GABA_AR clusters (Fig. 1 D and E and Movie S2). QD-labeled α2^SEP-GABA_ARs were less mobile when inside GABA_AR clusters (median D_in = 0.0078 μm²·s⁻¹, median D_out = 0.0192 μm²·s⁻¹, Fig. 1F) and exhibited confined motion within clusters (Fig. S2 B and C), presumably due to interactions with the scaffold gephyrin. α2^SEP-GABA_AR mobility and residency time were similar inside clusters whether or not these were apposed to FM 4-64 positive puncta (Fig. S2 D–F). Native GABA_ARs imaged by QD tracking could also be found either confined at inhibitory postsynaptic domains labeled by expression of GFP-tagged gephyrin or seen to rapidly diffuse into and out of these structures (Fig. S2 G–J). Thus, while clusters of synaptic GABA_ARs remain stable over time, single receptors can rapidly exchange between synaptic and extrasynaptic locations, providing a potential mechanism for rapid regulation of synaptic GABA_AR number.

Ca²⁺ Influx Through NMDARs Causes Rapid Dispersal of Surface GABA_AR Clusters.

To directly examine the influence of neuronal activity on GABA_AR clustering in live neurons we investigated the effect of altering excitatory synaptic activity by activating GluRs. Perfusion of 30 μM glutamate and 1 μM glycine for 4 min caused a rapid dispersal of surface α2^SEP-GABA_AR clusters (Fig. 2 A–D and Movie S3). At t = 9 min, α2^SEP-GABA_AR cluster intensities were significantly decreased [fluorescence in clusters at t = 9 min normalized to t = 0 (cluster F/F₀); control, 0.95 ± 0.02; glut/gly, 0.65 ± 0.05, P = 0.0001]. Loss of clustered α2^SEP-GABA_AR fluorescence on GluR activation was significantly reduced in zero extracellular Ca²⁺ (cluster F/F₀: 0.87 ± 0.03, P = 0.005) or in the presence of the NMDAR antagonist APV (cluster F/F₀: 0.86 ± 0.02, P = 0.0003), confirming that synaptic α2^SEP-GABA_AR cluster dispersal driven by GluR activation was dependent on extracellular Ca²⁺ entry via NMDARs (Fig. 2 A–C). Interestingly, GABA_AR activation was not required for GluR-dependent α2^SEP-GABA_AR cluster dispersal, which was also seen in the presence of the GABA_AR antagonist bicuculline (100 μM) (cluster F/F₀: 0.74 ± 0.04, P = 0.001, Fig. 2C).

Fig. 2.

GABA_AR clusters disperse on GluR activation. (A) α2^SEP-GABA_AR clusters disperse on GluR activation (blue bar), which requires Ca²⁺ influx through NMDA receptors but not GABA_AR activity. (B) Time course of α2^SEP-GABA_AR cluster F/F₀: control, black, n = 6 cells; glut/gly, dark blue, n = 6; in presence of APV, gray, n = 8; in 0 [Ca²⁺]_e, light blue, n = 5; in presence of bicuculline, purple, n = 4. (C) Summary of cluster F/F₀ at t = 9 min after initial GluR activation. Loss of fluorescence in α2^SEP-GABA_AR clusters on GluR activation is significant compared with control (P = 0.0001) and is significantly blocked in 0 [Ca²⁺]_e (P = 0.005) and by APV (P = 0.0003), but not by bicuculline (P = 0.001 compared with control). Overall fluorescence in processes is not altered by GluR activation (P > 0.05). (D) Variance normalized to t = 0 (Var/Var₀) in processes at t = 9 min after initial GluR activation. Variance in pixel intensity is decreased significantly by GluR activation (P = 0.001). (E) α2^SEP-GABA_AR clusters disperse on GluR activation when endocytosis is blocked. (F) Time course of α2^SEP-GABA_AR cluster F/F₀: vehicle, black, n = 7 cells; glut/gly after pretreatment with dynasore, pink, n = 7. (G) Cluster F/F₀ at t = 9 min. Loss of fluorescence in α2^SEP-GABA_AR clusters is significant compared with vehicle control (P = 3 × 10⁻⁵). (H) Close-up of α2^SEP-GABA_AR dispersal on GluR activation (blue bar). Intensity is represented by a custom look-up table. (I) F/F₀ in perisynaptic region surrounding cluster (gray) increases as F/F₀ in cluster (black) decreases, indicating receptor dispersal. (J) Blot showing surface biotinylated GABA_AR γ2 subunit and total γ2 subunit for control and glut/gly treatment. (K) Surface GABA_AR-γ2 fraction is unaffected by glut/gly treatment (91.2 ± 6.2%, P > 0.05, paired t test, n = 4).

In contrast to the decrease in fluorescence of α2^SEP-GABA_AR clusters, the total fluorescence in dendrites remained unaltered (Fig. 2C, F/F₀, control, 0.96 ± 0.01; glut/gly, 0.92 ± 0.01, P > 0.05), suggesting a redistribution of GABA_ARs at the plasma membrane (i.e., out of clusters and into extrasynaptic regions), rather than their removal from the cell surface. Dynamin and AP2-dependent endocytosis can rapidly modify the number of synaptic GABA_ARs (21–23) and occurs predominantly at extrasynaptic sites (24). Pretreatment of neurons with the dynamin inhibitor dynasore (10 min, 80 μM) (25, 26) did not prevent the loss of clustered α2^SEP-GABA_AR fluorescence (Fig. 2 E–G, cluster F/F₀: vehicle, 0.93 ± 0.01; dynasore + glut/gly, 0.69 ± 0.04, P = 3 × 10⁻⁵), confirming that the observed loss of GABA_AR clusters is due to cell surface dispersal of α2^SEP-GABA_ARs and not endocytosis. In agreement with this, α2^SEP-GABA_AR fluorescence in perisynaptic regions surrounding clusters could be seen to transiently increase on GluR activation (Fig. 2 H and I), and the variance in the distribution of α2^SEP-GABA_AR fluorescence intensities in neuronal processes at t = 9 min was significantly decreased by GluR activation (normalized to variance at t = 0: control, 0.97 ± 0.03; glut/gly, 0.73 ± 0.03, P = 0.001, Fig. 2D), suggesting a more diffuse GABA_AR distribution (cluster dispersal). The above results were confirmed for native receptors as GluR activation did not change the total intensity of native surface GABA_ARs in dendrites but significantly decreased the variance of pixel intensities in dendrites, suggesting that surface clusters of native GABA_ARs are also dispersed upon GluR activation (Fig. S3). Moreover, surface biotinylation to directly quantify the effect of GluR activation on native GABA_AR levels revealed no change in the surface fraction of the γ2 subunit (glut/gly: 92 ± 6.2% of control, P > 0.05, paired t test, Fig. 2 J and K). These data suggest that GABA_AR cluster dispersal upon GluR activation is due to lateral movement of GABA_ARs out of synaptic clusters and into the extrasynaptic plasma membrane.

Activity-Dependent GABA_AR Cluster Dispersal Is Reversible and Can Be Induced by Local, Synaptic Glutamate Release.

The activity-induced movement of GABA_ARs out of synapses and into the extrasynaptic space suggests a pool of redistributed surface receptors available for eventual replenishment of synaptic GABA_AR content. To further investigate this possibility we imaged α2^SEP-GABA_AR cluster fluorescence for longer periods after GluR activation to determine whether GABA_AR clusters could recover from dispersal (Fig. 3 A–C). GABA_AR clusters were dispersed at t = 10 min after initial drug application (cluster F/F₀, control, 1.00 ± 0.01; glut/gly, 0.70 ± 0.03, P = 0.001), but 40 min after GluR activation, α2^SEP-GABA_AR clustering had fully recovered (cluster F/F₀, control, 0.95 ± 0.03; glut/gly, 0.94 ± 0.06, P > 0.05, Fig. 3C), suggesting a return of α2^SEP-GABA_ARs back to synapses on longer timescales. Cluster recovery was not due to recycling of receptors that had been internalized upon GluR stimulation, as dynasore treatment did not block the recovery from dispersal (Fig. S4). To explore the possibility that synaptic activation could cause local GABA_AR dispersal, we used a patch pipette to locally stimulate axons crossing dendrites expressing α2^SEP-GABA_ARs (Fig. 3D). By maximally stimulating single axons, we found that fluorescence from clusters in stimulated regions (<15 μm from stimulated synapses) was rapidly decreased (at t = 240 s, cluster F/F₀ was 0.94 ± 0.02 in unstimulated regions and 0.79 ± 0.05 in stimulated regions, P = 0.014, Fig. 3 E and F, time course in Fig. S5), dependent on action potential firing and GluR activation (Fig. S5). Thus, GABA_AR dispersal is reversible and can be triggered by glutamate released from single or a few synapses with important implications for plasticity and control of local dendritic excitability.

Fig. 3.

GABA_AR cluster dispersal is reversible and can be induced by local glutamate release. (A) Schematic of recovery experiment. Images were taken at 0, 10, and 40 min. (B) α2^SEP-GABA_AR clustering can recover after dispersal induced by GluR activation. (C) Cluster F/F₀ at t = 0, 10, and 40 min. Loss of α2^SEP-GABA_AR cluster fluorescence is significant at t = 10 min (P = 0.001) but is not significantly different at t = 40 min (P > 0.05), n_control = 5 cells, n_glut/gly = 5 cells. (D) DIC image showing local stimulation of axon (colored green) crossing α2^SEP-GABA_AR containing dendrites (blue), using a glass microelectrode (lower left). (E) α2^SEP-GABA_AR fluorescence on dendrite in D before and after electrical stimulus (20 Hz, 1 min). Clusters are marked by gray arrows in stimulated regions (within 15 μm of apparent axonal contacts) and by black arrows in unstimulated regions. (F) F/F₀ at t = 240 s is significantly reduced in stimulated regions compared with unstimulated regions (P = 0.014, n = 8 cells).

GluR Activation Increases GABA_AR Lateral Mobility.

To further explore GluR-driven modulation of GABA_AR surface movement we examined the dynamics of single QD-tagged α2^SEP-GABA_ARs, which allowed us to directly monitor the behavior of single GABA_ARs with respect to the positions of the synaptically localized GABA_AR clusters. GluR activation resulted in a significant increase in the surface mobility of GABA_ARs both inside and outside of α2^SEP-GABA_AR clusters [median D_in increased 1.87-fold from 0.0078 to 0.0146 μm²·s⁻¹, P = 8 × 10⁻⁸; median D_out increased 1.16-fold from 0.0192 to 0.0219 μm²·s⁻¹, P = 4 × 10⁻¹⁰, Kolmogorov–Smirnov (K-S) test, Fig. 4 A–C and Movie S4]. Although the mobility of both receptor pools increased, the difference in mobilities of single α2^SEP-GABA_ARs was more striking for receptors within clusters, suggesting that the effect of GluR activation on GABA_AR mobility was strongest for receptors at GABA_AR cluster locations (inhibitory synapses). Furthermore, GluR activation strongly decreased the average time spent by single QD-labeled GABA_ARs at GABA_AR clusters (control, 2.35 ± 0.41 s; glut/gly, 1.30 ± 0.17 s, P = 0.035, Fig. 4D). Using QD tracking to follow the behavior of single native GABA_ARs labeled with antibodies to either α2 or β2/β3 GABA_AR subunits and 605 nm QDs we found that GABA_ARs were notably more dynamic after GluR activation (Fig. S6). Median diffusivity increased 1.20-fold for the α2 subunit and 1.27-fold for the β3 subunit (both P < 1 × 10⁻¹⁰, K-S test). These results describe an effect of GluR activation on individual GABA_AR mobilities consistent with that of GABA_AR dispersal from synaptic clusters.

Fig. 4.

GluR activation increases GABA_AR lateral mobility and decreases synaptic residency time. (A) QD-tagged α2^SEP-GABA_AR trajectories (orange) before and after GluR activation (t = 9 min) overlaid on α2^SEP-GABA_AR image. (B and C) Instantaneous diffusion coefficients for QD-tagged α2^SEP-GABA_ARs before (solid line) and after (dashed line) GluR activation inside (B) and outside of clusters (C). Mobilities of both pools increased significantly: median D_in increased 1.87-fold (n_before = 598, n_after = 436, P = 8 × 10⁻⁸, K-S test); median D_out increased 1.16-fold (n_before = 3,665, n_after = 2,719, P = 4 × 10⁻¹⁰, K-S test). (D) Mean residency time of QD-tagged α2^SEP-GABA_ARs at clusters is significantly lower after GluR activation. Control, n = 10 cells; glut/gly, n = 6 cells, P = 0.035.

GABA_AR Cluster Dispersal and Increased Lateral Diffusion Depend on Activation of Calcineurin and Serine 327 in the GABA_AR γ2 Subunit.

To investigate the signaling mechanisms involved in NMDAR-mediated GABA_AR cluster dispersal, we asked whether Ca²⁺ influx through NMDARs could activate the phosphatase calcineurin, which has previously been implicated in activity-dependent down-modulation of synaptic inhibition (3, 27). We found that treating cells with a calcineurin autoinhibitory peptide (50 μM, 30 min) or cyclosporin A (20 μM, 10 min, also added in perfusate) did not affect α2^SEP-GABA_AR cluster intensity (Fig. S7 A–D), suggesting that calcineurin has no effect on GABA_AR clustering under basal conditions on this timescale. However, blocking calcineurin activity with the autoinhibitory peptide inhibited α2^SEP-GABA_AR cluster dispersal on GluR activation [cluster F/F₀, control, 0.92 ± 0.01; glut/gly, 0.71 ± 0.03 (P = 5 × 10⁻⁵); treatment, 0.86 ± 0.04, P = 0.018 compared with glut/gly, Fig. 5 A–C] as did cyclosporin A treatment (Fig. S7 E–G). This result suggests that calcineurin activation upon Ca²⁺influx through NMDARs is directly involved in GABA_AR dispersal.

Fig. 5.

GABA_AR dispersal is dependent on calcineurin activity and serine 327 in the γ2 subunit. (A) α2^SEP-GABA_AR fluorescence on GluR activation after pretreatment with calcineurin autoinhibitory peptide (CaN peptide). (B) Time course of cluster F/F₀; control, black, n = 6 cells; vehicle + glut/gly, blue, n = 5; glut/gly + CaN block, green, n = 6. (C) F/F₀ in α2^SEP-GABA_AR clusters at t = 9 min. GABA_AR dispersal seen under vehicle + glut/gly (P = 5 × 10⁻⁵) was blocked by calcineurin inhibitory peptide (P > 0.05 vs. control, P = 0.018 vs. vehicle + glut/gly). (D) Example blot with pS327 and γ2 antibodies from control and glut/gly conditions. Experiments were performed on hippocampal neurons. (E) Normalized to control, the ratio of pS327/γ2 decreased to 0.55 ± 0.09 on GluR activation; P = 0.035, n = 3. (F) Schematic of hypothesized mechanism of GluR-mediated changes in GABA_AR clustering and lateral mobility: dephosphorylation of S327 by calcineurin. (G) ^WTγ2-SEP and ^S327Aγ2-SEP fluorescence on GluR activation (blue bar). (H) Time course of cluster F/F₀; WT control, black, n = 6 cells; WT glut/gly, blue, n = 8; S327A control, gray, n = 7; S327A glut/gly, light blue, n = 7. (I) Cluster F/F₀ at t = 9 min. ^WTγ2-SEP clusters disperse on GluR activation (P = 6 × 10⁻⁵). However, ^S327Aγ2-SEP clusters were stable on GluR activation (P > 0.05). (J) Instantaneous diffusion coefficients for QD-tagged GABA_ARs containing ^WTγ2-SEP before (solid line) and after (dashed line) GluR activation. Median D increased 1.43-fold (P = 2 × 10⁻¹⁶, n_before = 8,085, n_after = 6,316, K-S test). (K) As in J, for QD-tagged ^S327Aγ2^SEP-GABA_ARs. Median D increased 1.06-fold (P = 2 × 10⁻¹⁶, n_before = 8,868, n_after = 7,327, K-S test). (L) Mean residency time of ^WTγ2^SEP-GABA_ARs at clusters is significantly lower after GluR activation (n = 7 cells, P = 0.04), but not significantly different for ^S327Aγ2^SEP-GABA_ARs (n = 7 cells, P > 0.05).

A major target site for calcineurin-mediated regulation of GABA_ARs is serine 327 (S327) on the γ2 subunit, dephosphorylation of which has been previously shown to lead to a reduction in inhibitory postsynaptic current amplitudes (27). We therefore tested whether S327 is important for NMDAR-mediated GABA_AR cluster dispersal. By Western blotting with a phospho-specific pS327 antibody (27) we found that GluR activation of hippocampal neurons caused a significant decrease in phosphorylation at S327 (normalized to control, pS327/γ2 ratio = 0.55 ± 0.09, P = 0.035, paired t test, Fig. 5 D and E). Using α2-SEP to report GABA_ARs as above, we then cotransfected a myc-tagged γ2 GABA_AR subunit, either with its S327 phosphorylation site intact (γ2-myc) or with S327 mutated to alanine (γ2^S327A-myc). Cotransfection efficiency was found to be ∼90% (Fig. S8 A–C). Neurons cotransfected with γ2-myc exhibited robust α2^SEP-GABA_AR dispersal on GluR activation, but coexpression of γ2^S327A-myc subunits partially blocked cluster dispersal (Fig. S8 D–H). Incomplete blockade of cluster dispersal was likely due to α2-SEP subunits being able to assemble with the pool of endogenous γ2 subunits (with their S327 sites intact); therefore, to directly test the influence of S327 on GABA_AR dispersal we imaged GABA_AR clusters with wild-type or mutant SEP-tagged γ2 subunits. S327A mutation did not affect the efficient targeting of ^S327Aγ2^SEP-GABA_ARs to synapses, cluster intensity, or cluster mobility compared with wild type (Fig. S9). However, whereas GABA_AR clusters containing ^WTγ2-SEP subunits were dispersed by GluR activation (cluster F/F₀, control, 0.96 ± 0.03; glut/gly, 0.74 ± 0.03, P = 6 × 10⁻⁵), GABA_AR clusters containing ^S327Aγ2-SEP subunits remained stable upon GluR activation (cluster F/F₀, control, 0.93 ± 0.02; glut/gly, 0.88 ± 0.02, P > 0.05) (Fig. 5 G–I).

We then examined the role of S327 in the γ2 subunit on GABA_AR diffusion dynamics in response to increased neuronal activity. By labeling γ2-SEP subunits with quantum dots via a GFP antibody, we were able to track single GABA_ARs containing either ^WTγ2-SEP or ^S327Aγ2-SEP subunits. Under control conditions, GABA_ARs containing ^S327Aγ2-SEP subunits were marginally more mobile than ^WTγ2-SEP containing receptors (median D_WT = 0.0161 μm²·s⁻¹, median D_S327A = 0.0187 μm²·s⁻¹). On GluR activation, mobilities of ^WTγ2^SEP-GABA_ARs were much more increased compared with their ^S327Aγ2^SEP-GABA_AR mutant counterparts (Fig. 5 J and K; median D_S327A increased only 1.06-fold from 0.0187 to 0.0199 μm²·s⁻¹, compared with median D_WT that increased 1.43-fold from 0.0161 to 0.0231 μm²·s⁻¹, P = 2 × 10⁻¹⁶, K-S test), suggesting that S327A mutation blocked an activity-induced increase in GABA_AR lateral diffusion. Similarly, the residency time of single ^WTγ2^SEP-GABA_ARs at clusters was significantly decreased by GluR activation (control, 2.29 ± 0.41 s; glut/gly, 1.55 ± 0.20 s, P = 0.04), whereas no significant change was seen for ^S327Aγ2^SEP-GABA_ARs (control, 2.16 ± 0.51 s; glut/gly, 2.19 ± 0.36 s, P > 0.05) (Fig. 5L). Together with the results above, this finding suggests that calcineurin-dependent dephosphorylation of S327 on the γ2 subunit controls activity-dependent changes in the lateral diffusion and clustering of surface GABA_ARs.

Discussion

We have investigated the surface behavior of single GABA_ARs and GABA_AR clusters under resting conditions and during increased neuronal activity. Our work provides several important insights into the molecular mechanisms that underlie activity dependence of GABA_AR lateral diffusion and clustering. We show that synaptic GABA_AR clusters are stable under resting conditions whereas single GABA_ARs can be highly mobile. NMDAR activation can cause a rapid increase in surface GABA_AR mobility and a reversible and local dispersal of synaptic GABA_ARs into the extrasynaptic membrane. This mechanism for fast modulation of surface GABA_AR mobility depends on the activity of the phosphatase calcineurin and S327 within the intracellular loop of the γ2 subunit. Our results suggest a cellular mechanism through which GluR signaling can regulate synaptic GABA_AR accumulation on rapid timescales.

Imaging endogenous QD-labeled GABA_ARs does not allow simultaneous visualization of receptor clusters at synapses, and direct imaging of GABA_AR clusters and single receptors has not yet been reported. Simultaneously visualizing α2^SEP-GABA_AR clusters via SEP fluorescence and single α2^SEP-GABA_ARs labeled with QDs allowed us to image the behavior of individual receptors compared with synaptic GABA_AR clusters. Using this approach we found that single α2^SEP-GABA_ARs can be highly dynamic, both inside and outside clusters, and can also rapidly exchange into and out of synaptic GABA_AR clusters. Similar results were also obtained imaging GABA_ARs via a SEP-tagged γ2 subunit. The dynamic nature of single SEP-tagged GABA_ARs was similar to that of endogenous QD-labeled GABA_ARs. These observations are in agreement with other recent reports showing that GABA_ARs can be highly dynamic in the plasma membrane (12, 28) and are in line with previous observations of bulk exchange of receptors between synaptic and extrasynaptic locations determined using electrophysiological tagging or fluorescence recovery after photobleaching approaches (10, 11). In contrast to single receptors, GABA_AR clusters were stable for prolonged periods under basal conditions. We did, however, observe small asynchronous oscillatory movements of entire receptor clusters, but these movements were ≈50-fold slower than those of single receptors. The small oscillatory movements of α2^SEP-GABA_AR clusters are similar to those previously reported for the inhibitory scaffold gephyrin (18), suggesting that the entire inhibitory postsynaptic apparatus (receptor and scaffold) may move together. This observation fits well with the recent observation that GABA_AR α2 subunits interact directly with gephyrin (15) and suggests a tight coupling of the receptor aggregate and scaffold in live cells.

Live cell imaging of SEP-GABA_AR clusters allowed us to directly visualize in real time the role played by neuronal activity in regulating the stability of synaptic GABA_AR clusters. Ca²⁺ entry through NMDARs caused a rapid dispersal of GABA_AR clusters without affecting total surface receptor fluorescence, suggesting that dispersal was due to a redistribution of GABA_ARs out of clusters and into the extrasynaptic membrane. In agreement with this, blocking GABA_AR endocytosis, which is dynamin dependent (21), with the dynamin inhibitor dynasore did not block GABA_AR cluster dispersal. GluR-dependent dispersal of native synaptic GABA_ARs without surface down-modulation was also observed. Moreover, GluR-dependent cluster dispersal correlated with an increase in the lateral mobility of endogenous GABA_ARs and both α2^SEP-GABA_ARs and γ2^SEP-GABA_ARs as determined using QD tracking. Our results are in line with other recent observations demonstrating that increased neuronal activity generated with the potassium channel blocker 4-AP also leads to an increase in GABA_AR mobilities in dendrites (28).

Dependent on the levels of postsynaptic Ca²⁺influx, GluR activation can cause either an increase or a decrease in the number or activity of synaptic GABA_ARs, via activation of Ca²⁺-dependent kinases or phosphatases, respectively (2, 27–29). Calcineurin is known to dephosphorylate GABA_ARs on influx of Ca²⁺ through NMDARs (27). In support of a key role for calcineurin in regulating synaptic inhibition via GABA_AR trafficking, we found that this phosphatase underlies the NMDAR-mediated dispersal of GABA_ARs from inhibitory synaptic clusters. Interestingly, calcineurin activity was found to have little influence on the stability or intensity of GABA_AR clusters under resting conditions, in agreement with reports suggesting calcineurin is recruited to GABA_ARs only on increased neuronal activity (27).

Whether changes in GABA_AR phosphorylation state could influence GABA_AR lateral mobility was unknown. Importantly, we found that activity-dependent GABA_AR cluster dispersal and increased lateral mobility were dependent on a well-established GABA_AR phosphorylation site and known target for calcineurin-mediated dephosphorylation (27). GluR activation reduced the phosphorylation of the native γ2 subunit at S327, and mutation of S327 to alanine (S327A) blocked GluR-dependent GABA_AR cluster dispersal. Moreover, by QD tracking GABA_ARs containing mutant γ2 subunits, we found that S327 directly controlled the GluR-dependent increase in GABA_AR lateral mobility and parallel decrease in GABA_AR residency time at synaptic clusters. These results suggest the possibility of an active release mechanism of GABA_ARs from synapses due to S327 dephosphorylation. An interesting topic for future study will be to determine if there is also an activity dependence to GABA_AR lateral mobility in receptors containing nonsynaptic subunit combinations that contribute to tonic conductances (e.g., α5- or δ-containing GABA_ARs).

NMDAR-dependent synaptic GABA_AR cluster dispersal has important implications for information processing and is likely to underlie the long-term depression of inhibitory synaptic strength reported to occur during long-term potentiation at excitatory synapses (3, 27). Importantly, imaging SEP-GABA_ARs for longer periods following the activation of GluRs revealed that NMDAR-mediated GABA_AR dispersal is reversible because recovery of cluster fluorescence could be observed on longer (40 min) timescales. Return of receptors on this timescale fits well with the idea that whereas transient GABA_AR dispersal leading to reduced inhibition would favor long-term potentiation at excitatory synapses and coupling of increased excitability to increased action potential firing (E-S coupling) (3), an eventual “homeostatic” return of GABA_ARs back to synapses would prevent a decrease in synaptic inhibition that may lead to runaway excitation and the generation of seizures. We also demonstrate that GABA_AR dispersal can act locally when only a few glutamatergic synapses are activated. Thus synaptic GABA_AR cluster dispersal could act to locally enhance dendritic branch excitability and facilitate interactions between neighboring excitatory synapses, with important implications for clustered plasticity and information storage in dendrites (30, 31).

Our findings are also likely to be relevant to neurological disease where prolonged pathological glutamate release, as occurs in epilepsy or stroke, could cause a more sustained shift of GABA_ARs out of synapses and sustained dispersal. This would lead to disinhibition and facilitation of runaway excitation, contributing to neuronal excitotoxicity and disrupted information processing (32, 33). Indeed, NMDAR-mediated dispersal of GABA_ARs out of synapses and into extrasynaptic locations may be the first step that occurs before the activity-dependent down-modulation of surface GABA_ARs that leads to the generation of self-sustaining seizures in status epilepticus (5, 34). Thus, targeting calcineurin and S327-mediated synaptic GABA_AR dispersal could also provide a unique therapeutic target for pathological disinhibition in diseases such as epilepsy.

Materials and Methods

Constructs.

The N-terminally tagged α2-SEP DNA was a kind gift from S. Moss (Tufts University, Cambridge, MA) and has been described previously (15). For details of construction of mutants, see SI Materials and Methods.

Biochemistry.

See SI Materials and Methods.

Cell Culture and Transfections.

Rat hippocampal neurons were prepared and cultured from embryonic day 18 rat brains. Cells were transfected by Amaxa nucleofection as previously described (35). See SI Materials and Methods.

Live Cell Imaging.

Imaging media used for all experiments contained 125 mM NaCl, 5 mM KCl, 1 mM MgCl₂, 2 mM CaCl₂, 10 mM d-glucose, and 10 mM Hepes and was adjusted to pH 7.4 with NaOH before use. Cells were imaged under perfusion (4 mL/min) and heating (35–37 °C). See SI Materials and Methods.

Fixed Cell Imaging.

See SI Materials and Methods.

Image Analysis.

Images were analyzed in ImageJ. QD tracking was performed using detection and tracking algorithms written in Mathematica (Wolfram Research). See SI Materials and Methods.

Local Stimulation.

Local electrical stimulation of axons was made using a stimulating electrode via a patch pipette (resistance ∼1 MΩ) filled with imaging media (36). See SI Materials and Methods.

Statistical Analysis.

All experiments were performed on neurons from at least three individual preparations. Unless otherwise stated, P values given are from two-tailed t tests (type 2) and values are given as mean ± SEM. Error bars represent SEM.