Astrocytes as gatekeepers of GABAB receptor function

Abstract

The long-lasting actions of the inhibitory neurotransmitter GABA result from the activation of metabotropic GABA(B) receptors. Enhanced GABA(B)-mediated IPSCs are critical for the generation of generalized thalamocortical seizures. Here, we demonstrate that GABA(B)-mediated IPSCs recorded in the thalamus are primarily defined by GABA diffusion and activation of distal extrasynaptic receptors potentially up to tens of micrometers from synapses. We also show that this diffusion is differentially regulated by two astrocytic GABA transporters, GAT1 and GAT3, which are localized near and far from synapses, respectively. A biologically constrained model of GABA diffusion and uptake shows how the two GATs differentially modulate amplitude and duration of GABA(B) IPSCs. Specifically, the perisynaptic expression of GAT1 enables it to regulate GABA levels near synapses and selectively modulate peak IPSC amplitude, which is primarily dependent on perisynaptic receptor occupancy. GAT3 expression, however, is broader and includes distal extrasynaptic regions. As such, GAT3 acts as a gatekeeper to prevent diffusion of GABA away from synapses toward extrasynaptic regions that contain a potentially enormous pool of GABA(B) receptors. Targeting this gatekeeper function may provide new pharmacotherapeutic opportunities to prevent the excessive GABA(B) receptor activation that appears necessary for thalamic seizure generation.

Figure 1.

GABA_B-mediated IPSCs during GABA transporter blockade. A, Schematic of thalamic slice preparation in which whole-cell recordings of TC neurons in the VB were obtained while directly stimulating neurons of the RT. GAT antagonists were applied via local perfusion in VB. B, IPSCs during complete GAT blockade. GABA_B IPSCs were evoked once/10 s during control (∼5 min), GAT1 blockade (∼15 min), combined GAT1 and GAT3 blockade (∼20 min), followed by a washout period. NO-711 was used by block GAT1, whereas SNAP-5114 was used to block GAT3. B1, IPSCs during GAT1 (NO-711), and then GAT1 plus GAT3 blockade (NO-711 plus SNAP-5114). B2, Time series plot showing evolution of IPSC decay during GAT1, and then combined GAT1/3 blockade. B3, Quantification of decay (τ_dw) during combined blockade compared with GAT1 block alone, which had no effect on decay. Raw (left) and normalized (right) data are shown. The combined blockers increased decay by 981%. C, IPSCs during either GAT1 or GAT3 blockade. C1, Example GABA_B IPSCs during either NO-711 (left) or SNAP-5114 (right) application. The insets shows IPSCs normalized to peak amplitude. C2, Time series measuring IPSC amplitude during either GAT1 or GAT3 blockade. SNAP-5114 (○; n = 13) had a greater effect on the IPSC amplitude than NO-711 (•; n = 13). D, Quantification of IPSC amplitude (D1) and decay (τ_dw) (D2) during GAT1 versus GAT3 blockade. Plots show raw (left) and relative (i.e., normalized-to-control; right) changes. The normalized amplitude and decay changes were larger during SNAP-5114 versus NO-711 (p < 0.0005, t test). The two points of each continuous line in raw plots constitute the control (left point) and GAT blocker (right point) for an individual VB cell recording. The three bars in normalized plots correspond to data from control (white), GAT blockade (black), and wash (gray) conditions. Example IPSCs are the average of 10 responses during the final ∼1.5 min of each condition. Error bars indicate SEM. See Tables 1–3 for raw data values.

Figure 2.

Distinct anatomical localization of GAT1 and GAT3 in the VB nucleus of thalamus. A, Thalamic slices were stained for gephryin (or VGAT), a scaffolding protein associated with GABAergic synapses, and then costained for either GAT1 (left) or GAT3 (right). Slices were analyzed for GAT proximity to GABAergic synapses. Confocal images of stained sections are shown. Scale bar, 3 μm. B, Schematic describing radial analysis. GAT staining density was measured within concentric shells surrounding each synaptic marker to determine the number of GAT-immunopositive particles per shell volume. C, Plotted is the average GAT1 (blue) and GAT3 (red) density profile derived from radial analyses of both gephyrin- and VGAT-stained slices in the VB nucleus. Data from both gephyrin and VGAT-stained slices were combined because length constants were not different (see D). Curves were normalized to peak value. D, Weighted length constants (λ_w) were calculated to describe GAT1/3 density decay as a function of distance from putative synapses. Shown are λ_w values for GAT1 and GAT3 density in individual VGAT- and gephyrin-stained slices (D1), and corresponding average λ_w values (D2). λ_w was shorter for GAT1 density than for GAT3 density. See supplemental Figure S2 (available at www.jneurosci.org as supplemental material) for more detailed description of data analysis. Error bars indicate SEM.

Figure 3.

Slowing GABA diffusion increases GABA_B IPSC amplitude. A1, Control IPSCs were evoked in TC neurons every 10 s for ∼5 min, followed by application of ACSF containing 5% dextran (∼10 min). Plotted here are responses for three cells that were maintained through the washout period. A2, Example GABA_B IPSCs from one experiment. A3, Quantification of data corresponding to complete recordings (full lines) and recordings that became unstable during the washout period (dotted lines) (*p < 0.05, paired t test) (left). Dextran increased IPSC amplitude by 21 ± 9% (right). B, A three-dimensional model was built to understand how GABA, GAT, and GABA_B receptor binding dynamics contribute to experimental GABA_B-mediated IPSCs (for model details, see Fig. 4 and supplemental Fig. S3, available at www.jneurosci.org as supplemental material). Here, the coefficient of diffusion for GABA (D_GABA) was varied in our model and the peak amplitude of simulated responses was measured. B1, Responses from a pair of simulations run with slow (left), moderate (middle), and fast (right) D_GABA values (in square micrometers per millisecond). B2, Plotted are peak amplitudes for simulated GABA_B responses over a range of D_GABA values. When D_GABA was >0.1 μm²/ms, slowing diffusion enhanced responses, whereas the opposite is true for values <0.1 μm²/ms. Each data point corresponds to the mean and SD of four simulations. Values for D_glutamate and D_acetycholine are shown for reference. B3, Example traces of control (D_GABA, 0.4 μm²/ms) and 40% slower (D_GABA, 0.24 μm²/ms) responses. B4, Slowing D_GABA by 40% from an initial value of 0.4 μm²/ms increased simulated responses by 29 ± 4%. The plot is subdivided to show raw values (left) and relative values compared with control (right). **p < 0.01 (unpaired t test). Error bars indicate SEM. C, Spatiotemporal dynamics of GABA concentration ([GABA]) and GABA_B receptor binding during simulations with different D_GABA values. [GABA] (C1–C3) and bound receptors (C4–C6) were radially binned according to distance from release sites at different time points (C1, C4, 10 ms after release; C2, C5, time of peak simulated GABA_B response; C3, C6, 100 ms after peak). Supplemental Movie S1 (available at www.jneurosci.org as supplemental material) shows continuous record. The line colors correspond to the colored data points in B2. The arrows point to sites of GABA release (i.e., synapses). Bound receptors appearing within synapses are an artifact of binning and reflect perisynaptic receptor binding—synapses were not populated with GABA_B receptors. The blue scale corresponds to the blue line (i.e., slowest D_GABA), whereas the black scale corresponds to both green and red lines. For detailed analysis methods, see supplemental Figure S4 (available at www.jneurosci.org as supplemental material).

Figure 4.

Modeling GABA diffusion, uptake, and GABA_B receptor binding at the RT–TC synapse. A1, Simplified three-dimensional model (left) based on anatomical characterization of inhibitory thalamic synapses between RT and TC neurons. A central cross-sectional view is also shown (right). One millimolar GABA was released directly underneath each release site (top, black) and diffused into/away from the synapse with a coefficient of diffusion of 0.4 μm²/ms. Cellular membrane (white) enveloped release sites and contained GAT1 (blue), GAT3 (red), and GABA_B receptors (green). The synaptic region did not contain receptors. A2, Left, The model was populated with GABA_B receptors at densities described by Kulik et al. (2002). Right, GAT density profiles. GAT density was highest (i.e., peak) in the region adjacent to synapses and then decayed according to the λ_w calculated in Figure 2. Shown is the GAT1/3 density profile when both peak GAT1 and GAT3 are 300 molecules/μm². See supplemental Figure S3 (available at www.jneurosci.org as supplemental material) for details regarding the methods used to populate the model with receptors/transporters. B, GABA_B receptor binding during simulated control (black), GAT1 blockade (blue), and GAT3 blockade (red) conditions. B1, The number of GABA_B receptors bound versus time for models containing different GAT1:GAT3 peak values (columns) and in various GIRK channel cooperativity schemes (rows). R¹, R², and R³ refer to no, 2, and 3 cooperativity schemes. B2, GAT1 peak density was maintained at 300 molecules/μm². From an initial control value of 0 molecules/μm², peak GAT3 density was gradually increased to 900 molecules/μm² in +100 molecules/μm² increments. Plotted are relative changes (i.e., percentage of control) in peak amplitude (left) and decay (right) observed during simulated GAT1 (blue) and GAT3 (red) blockade in various cooperativity schemes. Each data point represents the mean (±SD) values derived from four simulations. For comparison, the relative changes observed during experimental GAT1 and GAT3 blockade (Fig. 1) are also shown in each plot (horizontal lines). A peak GAT1/GAT3 density of 300:300 with a R³ cooperativity scheme (arrows) best described our experimental results [i.e., simulated amplitude/decay points intersect with experimental data (lines)].

Figure 5.

Simulated GABA_B responses during repetitive release. In all portions of this figure, black lines/points correspond to control conditions, whereas blue and red lines/points correspond to GAT1- and GAT3-block conditions, respectively. A1, Five release events were delivered at 300 Hz to mimic RT neuron activity during action potential bursts. The 300:300 peak GAT1/GAT3 densities were incorporated into the model and D_GABA was set to 0.4 μm²/ms. Shown are simulated GABA_B responses during control and GAT1 and GAT3 blockade. The insets show normalized responses. Scale bar, 200 ms. A2, Relative changes in amplitude, decay, and time-to-peak during GAT1 and GAT3 blockade. Although the simulated responses were larger than those generated by single release events, the relative differences in amplitude, decay, and rise time observed during GAT1 and GAT3 blockade were maintained. B, Three-dimensional map of X–Y plane projecting through model center. Shown is [GABA] on a log₁₀ scale during control, GAT1-, and GAT3-blocked simulations at the time of the peak GABA_B response. C, Spatiotemporal dynamics of [GABA] and GABA_B receptor binding during control and GAT1 and GAT3 blockade conditions. Similar to plots in Figure 3C, [GABA] and bound GABA_B receptors were radially binned according to distance from release sites. In a practical sense, the [GABA] plots in C are binned, and numerical representations of the planes transecting the plots are shown in B (e.g., control). [GABA] (C1–C3) and bound receptors (C4–C6) are shown for three time points: 10 ms after release (C1, C4), time of peak GABA_B response (C2, C5), and 200 ms after peak GABA_B response (C3, C6). Supplemental Movie S2 (available at www.jneurosci.org as supplemental material) shows continuous record. The raw values (left) and relative changes (right) are shown in each plot. Largest relative differences were observed in distal model regions. The horizontal lines above raw bound receptor lines show spatial boundary circumscribed by 80% of bound receptors (i.e., 80% width). The inset bar graphs in raw bound receptor plots show percentage of total bound receptors that are located >3 μm from synapses. D, Cumulative plots were generated to show how response properties change as model volume (i.e., spatial bins) increases. D1, Example control, GAT1-, and GAT3-blocked responses generated by only summing bound receptors within 2 μm (left) and 10 μm (right) of model center. D2, Relative response amplitude (y-axis) as a function of cumulated 0.5 μm bins [x-axis (i.e., “X” in D1)]. Relative amplitude changes observed during simulated GAT1 and GAT3 blockade depended steeply on model regions within ∼2.5 μm of the model center. D3, Relative response decay as a function of cumulated 0.5 μm bins. Simulated GAT3 blockade required incorporation of distal model regions to achieve the ∼30% prolongation in response decay. Error bars indicate SEM.

Figure 6.

Partial GAT3 blockade increases GABA_B IPSC decay. A–C, Partial GAT3 blockade prolonged experimental GABA_B IPSCs. A, Example of evoked GABA_B IPSCs during control conditions and during 10 μm SNAP-5114. The inset shows normalized traces. B, Time series showing relative change in IPSC amplitude during application of 10 μm SNAP-5114 (n = 10). C, The 10 μm SNAP-5114 increased GABA_B IPSC amplitude by 42 ± 6% (C1), a relative change similar to that observed during full GAT1 blockade (p = 0.23, t test) (Fig. 1). Unlike GAT1 blockade, however, 10 μm SNAP-5114 also increased IPSC decay (τ_dw) by 26 ± 10% (C2) (control, 175 ± 25 ms; SNAP, 211 ± 27 ms; wash, 194 ± 31 ms; p < 0.005; n = 10). D, Simulated GABA_B responses are prolonged during partial (50%) GAT3 blockade. The mean amplitude of responses generated during partial GAT3 blockade (red) was 46% greater than control (n = 4; p < 0.0001, t test). This level of GAT3 blockade also prolonged responses by 17% (n = 4; p < 0.0001, paired t test). Error bars indicate SEM.

Figure 7.

GABA concentrations are highest during GAT3 blockade. A–D, The actions of the low-affinity GABA_B receptor blocker 20 μm CGP-35348 were assessed on control IPSCs (A), and IPSCs evoked during either GAT1 (•) or GAT3 blockade (○) (B). C, Examples of the actions of CGP-35348 on control GABA_B IPSCs (left), and GABA_B IPSCs in the presence of GAT1 (middle) and GAT3 (right) blockers. D, Raw (left) and normalized (right) amplitudes during CGP-35348 application. The relative change in amplitude induced by CGP-35348 was smallest during SNAP-5114 application. E–H, The high-affinity GABA_B receptor antagonist CGP-54626 equally reduced GABA_B IPSCs evoked during control and GAT1 and GAT3 blockade. E, The actions of the high-affinity GABA_B receptor blocker 20 nm CGP-54626 were assessed on control IPSCs (E) (n = 6), and IPSCs evoked during either GAT1 (•; n = 7) or GAT3 blockade (○; n = 7) (F). G, Examples of the actions of CGP-54626 on control GABA_B IPSCs (left), and GABA_B IPSCs in the presence of GAT1 (middle) and GAT3 (right) blockers. H, Raw (left) and normalized (right) amplitudes during CGP-54626 application. The relative change in amplitude during CGP-54626 was similar during all three conditions. Error bars indicate SEM.